Nonhyperbolic step skew-products: ergodic approximation (Q1679725)

This work is an exhaustive study of transitive skew-product maps models over a complete shift of finite symbols whose fiber maps are circle \(C^1\) diffeomorphisms. These skew-products are typical models underlying nonhyperbolic (transitive) dynamical systems. The authors show that these systems exhibit, simultaneously, ergodic measures with positive, negative and zero exponents. The approach is axiomatic. Several axioms are given in order to capture the fundamental mechanism of nonhyperbolic robustly transitive dynamical systems with compact central fibers (circles in this case). This mechanism is based on the existence of a blending interval (also known as ``blender'') in the central fiber in addition to transitivity. It is shown that ergodic measures with zero fiber exponent can be approximated by hyperbolic ones, both in the weak-\(\ast\) topology and in entropy. This is a remarkable result. Moreover these ergodic measures with zero exponent belong to the intersection of the convex hulls of measures with positive and negative exponents. This also allows to perturb, in a direct way, measures with positive exponent measures with negative exponent. Some amount of entropy is lost in this process, but it is determined by the difference between the Lyapunov exponents of the measures. The paper is well written and mandatory to those researchers interested in nonhyperbolic dynamical systems. This work can be placed as an intermediate step in the comprehension of measure space, ergodic theory and other dynamical properties of nonhyperbolic dynamical systems. Examples of works in this vein are: [\textit{J. Bochi}, \textit{C. Bonatti} and \textit{L.J. Díaz}, Math. Z. 276, No. 1--2, 469--503 (2014; Zbl 1284.37044); \textit{L. J. Díaz} and \textit{A. Gorodetski}, Ergodic Theory Dyn. Syst. 29, No. 5, 1479--1513 (2009; Zbl 1184.37010); \textit{A. S. Gorodetski} et al., Funct. Anal. Appl. 39, No. 1, 21--30 (2005; Zbl 1134.37347); translation from Funkts. Anal. Prilozh. 39, No. 1, 27--38 (2005); \textit{V. A. Kleptsyn} and \textit{M. B. Nalsky}, Funct. Anal. Appl. 41, No. 4, 271--283 (2007; Zbl 1178.37026); translation from Funkts. Anal. Prilozh. 41, No. 4, 30--45 (2007)]. Of course, the hypothesis of a compact central fiber is a great simplification (for instance, partially hyperbolic diffeomorphisms do not need to support an integrable central distribution). Even with this simplification, the dynamics is rich and the analysis is not easy at all. In the first section the axioms defining a \textit{blending interval} are introduced. More precisely, a blending interval is an interval \(J\subset S^1\) (identify \(S^1\) with the central fiber) such that the following properties hold: (1) \textbf{CEC+}: For any interval \(H\subset J\), the forward iteration of the iterated function system (IFS) associated to the fiber maps, expands \(H\) so that, eventually, it covers \(J\). The number of iterations and the expansions must be uniform, and only depends in the size of \(H\); (2) \textbf{Acc+}: Forward iterations of \(J\) cover \(S^1\); (3) \textbf{CEC-, Acc-}: The same as the previous axioms for backward iterations. The authors also state the main theorems and define the basic concepts as Lyapunov exponent and (non)hyperbolic measure in this section. The second section is devoted to study the consequences of transitive skew products with a blending interval and a discussion about the necessity of the previous axioms. In the third section the authors introduce the tools to approximate ergodic measures (with positive entropy). These are consequences of the Brin-Katok, Birkhoff and Egorov ergodic theorems. In addition, some results of control of distortion in the neighbourhood of orbits with zero exponent are given. The fourth section introduces the skeletons. These are pieces of orbits that can be used to approximate both entropy and fiber exponents. These are used as blocks to construct hyperbolic sets with good ergodic and dynamical properties. The fifth section deals with the so-called ``multi-variable-time horseshoes''. These are based in the interval horseshoes defined in [\textit{M. Misiurewicz} and \textit{W. Szlenk}, Stud. Math. 67, 45--63 (1980; Zbl 0445.54007)]. One of the key points is to estimate the entropy of these horseshoes and to establish their relation with skeletons. The final section is devoted to combine the developed notions and tools to prove the main theorems. The exposition is clear and clever and the authors put a lot of effort in making the arguments accessible to the reader.