Mean dimension, mean rank, and von Neumann-Lück rank (Q1637088): Difference between revisions

The mean dimension of topological dynamical systems was introduced by \textit{M. Gromov} [Math. Phys. Anal. Geom. 2, No. 4, 323--415 (1999; Zbl 1160.37322)] and later developed by \textit{E. Lindenstrauss} and \textit{B. Weiss} [Isr. J. Math. 115, 1--24 (2000; Zbl 0978.54026)] as a non-trivial invariant of systems with infinite topological entropy. Later work extended this notion to the setting of continuous actions of countable amenable groups. The von Neumann-Lück dimension was introduced for finitely-generated projective modules (and later for arbitrary modules) over the group von Neumann algebra of a discrete group by Lück as a generalization of \textit{M. F. Atiyah}'s \(L^2\) Betti numbers [in: Colloque ``Analyse et Topologie'' en l'honneur de Henri Cartan. Paris: Société Mathématique de France (SMF). 43--72 (1976; Zbl 0323.58015)] to arbitrary continuous actions. In this work, which builds on work of the first named author and \textit{A. Thom} [J. Am. Math. Soc. 27, No. 1, 239--292 (2014; Zbl 1283.37031)], where a similar link is found between entropy and \(L^2\) torsion, a connection is found between these two `dimension' invariants arising in different areas. The setting is an arbitrary \(\mathbb{Z}\Gamma\)-module \(M\) for a discrete amenable group \(\Gamma\). This structure gives rise to an induced \(\Gamma\)-action on the Pontryagin dual \(\widehat{M}\) by continuous automorphisms, giving an algebraic topological dynamical system for which the mean dimension is defined. On the other hand the module \(M\) itself has a von Neumann-Lück dimension. Here it is shown that these are both equal to a new invariant, the `mean rank' of \(M\), analogous to the rank of a discrete abelian group. Several applications are given, both to the mean dimension theory in dynamics and to the von Neumann-Lück rank. An addition formula is found for the mean dimension in the algebraic case (it is known it does not hold in general), showing in particular that in this setting the mean dimension cannot increase when passing to a factor. Further applications include extending work of Lindenstrauss giving a dynamical analogue of the Pontryagin-Schnirelmann theorem for actions of \(\mathbb{Z}\) to algebraic actions of countable amenable groups; using known cases of the strong Atiyah conjecture to constrain the possible range of values of the mean dimension of algebraic actions of certain groups (which parallels the result of \textit{D. Lind} et al. [Invent. Math. 101, No. 3, 593--629 (1990; Zbl 0774.22002)] relating the possible values of the topological entropy of algebraic \(\mathbb{Z}^d\)-actions to Lehmer's problem in algebraic number theory); finally, the converse of \textit{E. Lindenstrauss}' result [Publ. Math., Inst. Hautes Étud. Sci. 89, 227--262 (1999; Zbl 0978.54027)] showing that inverse limits of actions with finite topological entropy have zero mean dimension is shown for algebraic actions of elementary amenable groups under a condition on their finite subgroups. An appendix discusses the relation between the mean rank introduced here and the rank for finitely-generated \(\mathbb{Q}\Gamma\)-modules introduced by \textit{G. Elek} [Proc. Am. Math. Soc. 131, No. 11, 3477--3485 (2003; Zbl 1037.43001)].