Topological dimension and dynamical systems. Translated from the French by the author (Q2353889)

Topological entropy was introduced by \textit{R. L. Adler} et al. [Trans. Am. Math. Soc. 114, 309--319 (1965; Zbl 0127.13102)] as an invariant for topological conjugacy of continuous maps on compact topological spaces. Several ideas lay behind this work: the creation of a new invariant for the classification of topological dynamical systems; topological analogues of the measure-theoretic entropy notions of \textit{C. E. Shannon} in information theory [Bell Syst. Tech. J. 27, 379--423, 623--656 (1948; Zbl 1154.94303)] and of \textit{A. N. Kolmogorov} [Dokl. Akad. Nauk SSSR 124, 754--755 (1959; Zbl 0086.10101)], \textit{V. A. Rokhlin} [ibid. 124, 980--983 (1959; Zbl 0096.31405)] and \textit{Ya. G. Sinai} [ibid. 124, 768--771 (1959; Zbl 0086.10102)] in ergodic theory; finally the purely topological notion of open covers and (in the metric case) the Lebesgue covering lemma. For the case of a full shift (the map that shifts left in the coordinate) \(K^{\mathbb Z}\) for a compact metric space \(K\) the topological entropy is \(\log| K|\) if \(K\) is a finite space and is infinite if \(K\) is an infinite compact space, and so in those cases the topological entropy carries no information about \(K\) beyond the fact that its cardinality is infinite. On the other hand, it is clear that a great diversity of topological dynamical systems exist even of this specific type: if the full shift on \(K\) and the full shift on \(K'\) are topologically conjugate then their fixed point sets \(K\) and \(K'\) must be homeomorphic. In a different direction, a classical result of Beboutov which may be found in a survey of \textit{V. V. Nemycki} [Am. Math. Soc., Transl. 103, 85 p. (1954); translation from Usp. Mat. Nauk, n. Ser. 4, No. 6(34), 91--153 (1949; Zbl 0059.08002)] shows that a real flow (a continuous action of \(\mathbb{R}\) on a topological space) whose set of fixed points can be imbedded into \(\mathbb{R}\) can be imbedded as a whole into the space of continuous functions on \(\mathbb{R}\), respecting the natural action of \(\mathbb{R}\) on this space. From a different direction, a long-standing open problem in topological dynamics asked if any minimal topological dynamical system (which, in particular, has no constraints arising from the topological properties of the set of points of a given period) can be imbedded in the full shift on \([0,1]\). A different direction of enquiry sought for topological models of dynamical systems with prescribed properties, one strand of which culminated with the result of \textit{R. I. Jewett} that any weakly mixing system has a uniquely ergodic model [J. Math. Mech. 19, 717--729 (1970; Zbl 0192.40601)]. On the other hand, a sophisticated theory of dimension in topology developed, with several different notions of topological dimension and, for example, the general result of \textit{K. Menger} [Proc. Akad. Wet. Amsterdam 29, 1125--1128 (1926; JFM 52.0595.01)] and \textit{G. Nöbeling} [Math. Ann. 104, 71--80 (1930; JFM 56.0506.02)] showing that a compact space with topological dimension less than or equal to \(d\) can be embedded into \([0,1]^{2d+1}\) and the construction of exotic spaces like the example of Boltyanskiĭ of topological dimension \(2\) whose square has dimension \(3\) [\textit{V. G. Boltyanskij}, Dokl. Akad. Nauk SSSR, n. Ser. 67, 597--599 (1949; Zbl 0035.38703)]. Finally, \textit{M. Gromov} [Math. Phys. Anal. Geom. 2, No. 4, 323--415 (1999; Zbl 1160.37322)] suggested a possible invariant of topological dynamical systems that could distinguish between certain systems of infinite topological entropy. These diverse currents from dimension theory, embedding problems, and topological dynamics were brought together in the notion of mean dimension systematically developed by \textit{E. Lindenstrauss} [Publ. Math., Inst. Hautes Étud. Sci. 89, 227--262 (1999; Zbl 0978.54027)] and \textit{E. Lindenstrauss} and \textit{B. Weiss} [Isr. J. Math. 115, 1--24 (2000; Zbl 0978.54026)]. This is a topological invariant that has functorial properties with respect to embedding, giving a necessary condition for embedding into \([0,1]^{\mathbb{Z}}\) (and many other such questions), and showing that uniquely ergodic systems do indeed have distinct dynamical properties in that they must be of mean dimension zero. The original work developing the theory of mean dimension focused on the case of \(\mathbb{Z}\)-actions, but already indicated how the theory might be extended to the natural more general setting of discrete amenable group actions. A challenging aspect of this theory from a pedagogical point of view is that (for students of topological dynamics) the topological and dimension-theoretic machinery may seem daunting, and (for students well versed in topology and dimension theory) the dynamical motivation may seem opaque. This volume -- a revised and somewhat augmented translation of the earlier notes by the author [Dimension topologique et systèmes dynamiques. Cours Spécialisés (Paris) 14. Paris: Société Mathématique de France (2005; Zbl 1116.54019)] -- is therefore a welcome addition. The overall objective is to provide a self-contained introduction to the theory of mean dimension without assuming any prerequisites in dimension theory. The first part, comprising five chapters, is a rapid introduction to the essential features of dimension theory in topological spaces. The second part develops the theory of mean dimension, including some applications to embedding theorems and some important counterexamples constructed by Lindenstrauss and Weiss. The main change from the original French version is two additional chapters, Chapter 9 on the basic properties of amenable groups and Følner's characterization, and Chapter 10 developing the mean dimension theory for continuous actions of discrete countable amenable groups. The text contains a large number of exercises, with hints for some of the more challenging ones.