Local dimension of ergodic measures for two-dimensional Lorenz transformations (Q4495003)

The aim of this paper is to prove a formula which involves the local dimension, the entropy and characteristic exponents of ergodic invariant measures for two-dimensional Lorenz transformations. Without entering into all the details, a map \(F: [0,1]^2\to [0,1]^2\) is called a two-dimensional Lorenz transformation if \(F(x,y)= (T(x), g(x,y))\), where \(T\) is piecewise strictly monotonic, \(g\) is \({\mathcal C}^1\) (except on the set \(P\times [0,1]\), where \(P\) denotes the set of discontinuities of \(T\)), \(g(x,.)\) is a contraction (plus some extra hypotheses on the partial derivatives of \(g\)). Furthermore, the images under \(F\) of the product by \([0,1]\) of two distinct continuity intervals for \(T\) are supposed to be distinct. Let \(F\) be a two-dimensional Lorenz transformation and \(\mu\) an ergodic, \(F\)-invariant probability measure on \([0,1]^2\) with positive entropy \(h_\mu(F)\). Let \(l_\mu(X)\) denote the local dimension of \(\mu\) in the point \(x\in [0,1]^2\). Suppose that the partition by the intervals of continuity for \(T\) is a generator (for a precise definition, see the paper), \(\int -\log|\partial g/\partial y|d\mu< \infty\), and \(0<\int \log|T'|d\mu<\infty\). If the maps \(y\mapsto \varphi(x,y)\) are uniformly continuous for \(x\in P\) and \(1/|T'|\) has bounded \(p\)-variation for some \(p> 0\), then it is proved that \[ l_\mu(X)= h_\mu(F) \Biggl( \frac{1} {\int- \log|\partial g/\partial y|d\mu}+ \frac{1} {\int \log|T'|d\mu} \Biggr). \] For connected results on two-dimensional Lorenz maps, see the author [\textit{T. Steinberger}, Isr. J. Math. 116, 253-269 (2000; Zbl 0952.37004)].