Local entropy averages and projections of fractal measures (Q431635): Difference between revisions

For a probability measure \(\mu\) on a metric space, the lower Hausdorff dimension \(\dim_{\ast} \mu\) of \(\mu\) is defined as \(\dim_{\ast} \mu = \inf \{\dim(A): \mu(A) > 0\}\). The present paper establishes the semi-continuity of the map \(\pi \mapsto \dim_{\ast} \pi\mu\) for \(\pi\in\Pi_{d,k}\), the space of orthogonal projections from \({\mathbb R}^d\) to its \(k\)-dimensional subspaces, when \(\mu\) is a measure on \({\mathbb R}^k\) displaying a certain ``local dynamics'', which roughly means that for \(x\in \text{supp}\,\mu\), one can form a sequence of cubes \(B_n\) descending to \(x\) and a sequence of restricted rescaling measures \(\mu^{x, n}\) which display stationary dynamics. The paper presents several significant applications. First of all, the conjecture of \textit{H. Furstenberg} is resolved in its full generality (the conjecture is not in print, but its related conjecture appears in [Probl. Analysis, Sympos. in Honor of Salomon Bochner, Princeton Univ. 1969, 41--59 (1970; Zbl 0208.32203)]). The following stronger form is proved: Let \(\mu\), \(\nu\) be Borel probability measures on \([0,1]\) which are invariant under \(T_m\), \(T_n\), where \(m\) and \(n\) are not powers of the same integer. Then, for every \(\pi\in\Pi_{2,1}\setminus \{\pi_x, \pi_y\}\), \(\dim_{\ast} \pi(\mu\times\nu) = \min \{1, \dim_{\ast}(\mu \times \nu)\}\). Here \(T_m\) is the \(m\)-fold map of \([0, 1]\): \(T_m(x) = mx\mod 1\), and \(\pi_x\) and \(\pi_y\) are the coordinate projections. This result leads to a short proof of Rudolph-Johnson theorem [\textit{D. J. Rudolph}, Ergodic Theory Dyn. Syst. 10, No. 2, 395--406 (1990; Zbl 0709.28013); \textit{A. S. A. Johnson}, Isr. J. Math. 77, No. 1--2, 211--240 (1992; Zbl 0790.28012)], which reads: If \(m\), \(n\) are not powers of the same integer, \(\mu\) is a probability measure on \([0, 1]\) invariant under \(T_m\) and \(T_n\), and all ergodic components have positive entropy for one of the maps, then \(\mu\) is the Lebesgue measure. The main result is applied to establish a generalization of the results by \textit{C. G. T. de A. Moreira} [Period. Math. Hung. 37, No. 1--3, 55--63 (1998; Zbl 0980.54025)], \textit{Y. Peres} and \textit{P. Shmerkin} [Ergodic Theory Dyn. Syst. 29, No. 1, 201--221 (2009; Zbl 1159.37005)] and \textit{F. Nazarov, Y. Peres} and \textit{P. Shmerkin} [Isr. J. Math. 187, 93--116 (2012)], concerning the Hausdorff dimension of the projections of products of self-similar measures on regular Cantor sets. They are also applied to problems concerning the Hausdorff dimension of the images of self-similar measures under \(C^1\) maps and the lower Hausdorff dimension of Bernoulli convolutions.