Radius of comparison and mean cohomological independence dimension (Q2161310)

The research done in this paper lies between the areas of topological dynamics and \(C^\ast\)-algebras theory. Given an action \(T: G\curvearrowright X\) of a discrete amenable group on a compact metrizable space by homeomorphisms, much attention goes to the structure of the crossed product \(C^\ast\)-algebra \(C(X)\rtimes_T G\) in terms of the internal structure of the dynamical input data. The focus of this work is the relationship between the so-called radius of comparison of \(C(X)\rtimes_T G\) (see [\textit{A. S. Toms}, J. Funct. Anal. 238, No. 2, 678--708 (2006; Zbl 1111.46041)]) and the mean dimension of \((X,T)\) (see [\textit{E. Lindenstrauss} and \textit{B. Weiss}, Isr. J. Math. 115, 1--24 (2000; Zbl 0978.54026)]). Although it is outside the scope of this review to introduce all the involved concepts rigorously, it is worth mentioning that the mean dimension of a dynamical system is a dynamical generalization of the covering dimension with values in \([0,\infty]\), which for instance obeys the formula \(\operatorname{mdim}\big( ([0,1]^d)^G,\text{shift} \big)=d\). The radius of comparison of a unital \(C^\ast\)-algebra is also a numerical value in \([0,\infty]\) measuring to what extent the Cuntz order relation on positive elements is determined by the rank functions induced by tracial states on the \(C^\ast\)-algebra. The relationship between these two a priori different concepts was specified in a conjecture by the second author of this article and Andrew Toms (in part inspired by the examples exhibited in [\textit{J. Giol} and \textit{D. Kerr}, J. Reine Angew. Math. 639, 107--119 (2010; Zbl 1201.46055)]), which asserts that for minimal actions, the radius of comparison for the crossed product is equal to half the mean dimension of the system. Some important work in this direction (see [\textit{G. A. Elliott} and \textit{Z. Niu}, Duke Math. J. 166, No. 18, 3569--3594 (2017; Zbl 1410.46046); \textit{Z. Niu}, J. Anal. Math. 146, No. 2, 595--672 (2022; Zbl 1502.46055)]) has focused on finding upper bounds for the radius of comparison in terms of the mean dimension, but the reverse relationship is not yet understood so well. This is the issue that the article aims to address, i.e., to find lower bounds for the radius of comparison in terms of the mean dimension. Since it would appear to be unclear at present how to achieve this goal directly, the authors introduce a few alternative versions of mean dimension that are better suited for this purpose. In drastically oversimplified terms, the authors propose to tweak the usual definition of mean dimension in such a way that the covering dimension of a space is replaced by the rational cohomological dimension. Recall that a compact metrizable space \(X\) has rational cohomological dimension \(d\) if \(d\) is the smallest natural number such that for any \(k>d\) and closed set \(Y\subseteq X\), the \(k\)-th relative Čech cohomology with rational coefficients vanishes, i.e., \(\check{H}^k(X,Y;\mathbb{Q})=0\). Using this basic idea, the authors can introduce the mean cohomological independence dimension \(\operatorname{mcid}(T;R)\) even with coefficients in any unital ring \(R\). The authors proceed to observe that one always has \(\operatorname{mcid}(T;R)\leq\operatorname{mdim}(T)\), although the two values do agree for some examples such as certain shift actions. More generally, the authors prove some results about subshifts where \(\operatorname{mcid}(T;R)\) can be estimated from below in terms of \(\operatorname{mdim}(T)\). At the same time, the authors exhibit some techniques for subshifts to estimate \(\operatorname{rc}(C(X)\rtimes_T G)\) from below in terms of the (rational) mean cohomological independence dimension of \(T\). The combination of these separate results leads to the inequality stated in Corollary 3.4 of the article, the precise assumptions and conclusion of which cannot be explained in this review in a self-contained manner because it relies on the specific construction in [\textit{J. Giol} and \textit{D. Kerr}, J. Reine Angew. Math. 639, 107--119 (2010; Zbl 1201.46055)]. Afterwards the authors exhibit yet another variant of their concept, called symmetric mean cohomological independence dimension, with the aim of obtaining even sharper lower bounds in this spirit. Using this concept, the authors can indeed prove the inequality (again slightly oversimplified for this review) \[ \operatorname{rc}(C(X)\rtimes_T G) \geq \frac12 \operatorname{smcid}(T;\mathbb{Q})-1 \] for any action \(T: G\curvearrowright X\) such that \(C(X)\rtimes_T G\) is simple. Lastly, the authors show that a subclass of examples constructed by \textit{J. Giol} and \textit{D. Kerr} [loc. cit.] have the property that the symmetrical mean cohomological independence dimension agrees with the mean dimension, which sheds new light on the Phillips-Toms conjecture for a reasonably large class of group actions. The paper ends with some interesting questions for further research.