Loss of memory and moment bounds for nonstationary intermittent dynamical systems (Q2035935)

The authors deal with nonstationary dynamical systems with intermittency in the spirit of the study initiated by \textit{Y. Pomeau} and \textit{P. Manneville} [Commun. Math. Phys. 74, 189--197 (1980)]. They consider the maps introduced in [\textit{C. Liverani} et al., Ergodic Theory Dyn. Syst. 19, No. 3, 671--685 (1999; Zbl 0988.37035)], \(T : [0, 1]\rightarrow [0, 1]\) defined as \[ T(x)=\left\{ \begin{array}{cc} x(1+2^{\gamma }x^{\gamma }), & x\leq \frac{1}{2}, \\ 2x-1, & x>\frac{1}{2}, \end{array} \right. ,\] where \(\gamma\in (0, 1)\) is a parameter. Let \(\mu\) and \(\mu'\) be probability measures on \([0, 1]\) with Hölder densities and \(\nu\) be a probability measure on \([0,1]\) with density in \(\mathcal{C}_{*}\), which is a convex cone of functions. In this case, the authors prove that \[\left|(T_{1,n})_{*}(\mu -\nu )\right|=O(n^{-1/\gamma^*+1}),\] \[\left|(T_{1,n})_{*}(\mu -\mu^*)\right|=O(n^{-1\left/\gamma ^*\right.}).\] The optimal bounds for a general measure \(\mu\) with density in \(\mathcal{C}_{*}\) are given. By constructing a filtration \(\mathcal{B}_n\) approximating \(H\) with the Doob martingale, the authors estimate the sharp moment bounds for Birkhoff sums and, more generally, ``separately Hölder'' observables.