Dimension theory for invariant measures of endomorphisms (Q985707): Difference between revisions

The paper is devoted to the dimension theory for invariant probability measures of a smooth non-invertible endomorphism of a compact Riemannian manifold. Its results extend several theorems by \textit{L. Barreira, Y. Pesin} and \textit{J. Schmeling} [Ann. Math. (2) 149, No. 3, 755--783 (1999; Zbl 0977.37011)] and \textit{F. Ledrappier} and \textit{L.-S. Young} [Commun. Math. Phys. 117, No. 4, 529--548 (1988; Zbl 0665.58020)] concerning diffeomorphisms to the case of endomorphisms. In particular, the author proves that if \(\mu\) is a hyperbolic and ergodic Borel probability measure for a smooth endomorphism \(f:M\to M\), then it is exact dimensional, i.\,e., \(\dim(\mu,x)\) exists for \(\mu\)-almost every \(x\in M\) and does not depend on \(x\), where \[ \dim(\mu,x) =\lim_{\rho\to 0}\frac{\ln\mu(B(x,\rho))}{\ln\rho}. \] Moreover, \(\dim(\mu,x) =\delta^s +\delta^u\) for \(\mu\)-almost every \(x\), where \(\delta^s\) and \(\delta^u\) are the dimensions of \(\mu\) along stable and unstable foliations. If \(\mu\) is not ergodic but only invariant then, analogously, \(\dim(\mu,x)\) exists almost everywhere and \(\dim(\mu,x) =\delta^s(x) +\delta^u(x)\). These results are obtained via consideration of the inverse limit space for the dynamical system \((M,f)\). In the case of surface endomorphism \(f\) they imply a certain relation between \(\dim(\mu,x)\), the measure entropy \(h_\mu(f)\), and the folding entropy \(F_\mu(f)\). For the case of general (not hyperbolic) ergodic measure it is proved that its dimension does not exceed its Lyapunov dimension. The author considers the dimension of invariant probability measures for random endomorphisms as well. Namely, for each ergodic measure \(\mu\) (with respect to a family of random endomorphisms) he defines a family of random sample measures and proves that under certain hypothesis the dimension of the sample measure is equal to the Lyapunov dimension of \(\mu\) with probability \(1\).