Mixing properties for toral extensions of slowly mixing dynamical systems with finite and infinite measure (Q2026743)

Let \((X,d)\) be a metric space and let \(f \;:\; X \longrightarrow X\) be a topological mixing map which preserves an ergodic Borel measure \(\mu\) . Let \(m\) be a positive integer and \(h \;:\; X \longrightarrow \mathbb{T}^m\), be a measurable map such that \[\big|h\big|_{\eta}=\sup_{x \neq y}\frac{\big|h(x)-h(y)\big|}{d(x,y)^\eta}<\infty,\; \textrm{with}\; \eta \in (0,1].\] We say that \(h\) is in \(C^{\eta}(X,T^m).\) Denote for \(\eta \in (0,1]\) by \(C^\eta (X \times \mathbb{T}^m )\) the space of functions \(v\) such that \(\big\|v\big\|=\big|v\big|_{\infty}+\big|v\big|_{\eta}<\infty,\) and for a natural integers \(m,p\) by \(C^{\eta,p}(X \times \mathbb{T}^m)\) the space of \(p\)-times differentiable with respect to the second coordinate with derivatives that lie in \(C^\eta (X \times \mathbb{T}^m ),\) the norm is given by \[\big\||v\big\||_{C^{\eta,p}}=\sum_{|j| \leq m} \Big\|\frac{\partial^j v}{\partial \psi^j}\Big\|_{C^\eta}.\] The toral extension of \(f\) is given by \begin{align*} f_{h} \colon X \times \mathbb{T}^m &\to X \times \mathbb{T}^m \\ (x,\psi) &\mapsto (f(x),\psi+h(x)). \end{align*} Let \(\ell \; : \; (0,+\infty) \to (0,+\infty) \) be a measurable function such that for each \(\lambda>0\), we have: \[\lim_{x \longrightarrow +\infty}\frac{\ell(\lambda x)}{\ell(x)}=1.\] For \(\beta \in (0,1]\), put \[\tilde{\ell}(n)=\begin{cases} \ell(n) &\textrm{if}\; \beta \in (0,1)\\ \sum_{j=1}^{n}\frac{\ell(j)}{j} &\textrm{if} \; \beta=1. \end{cases} \] Let \(Y \subset X\) be a subset with \(\mu(Y )\in (0, \infty)\) and \(v,w\) be an observables supported on \(Y \times \mathbb{T}^m,\) such that \(v_0=\int_{\mathbb{T}^m}v(y,\psi) d\psi\) is a suitable Banach space of observables \(\mathcal{B}(Y)\subset L^1 (Y)\), \(w \in L^{\infty}(Y \times \mathbb{T}^m)\) and \(v-v_0 \in C^{\eta,p}(Y \times \mathbb{T}^m).\) The correlation of order \(n\) of \(v\) and \(w\) is given by \[<v,w \circ f_h^n>=\int_{Y \times \mathbb{T}^m} v(x,\psi) w(f_h^n(x,\psi)) d\mu(x)d\psi,\] we put \[\tilde{v}=\int_{Y \times \mathbb{T}^m} v(x,\psi)d\mu(x) d\psi \; \textrm{and} \; \tilde{w}=\int_{Y \times \mathbb{T}^m} w(x,\psi)d\mu(x) d\psi.\] Under some suitable conditions including Dolgopyat's condition (the absence of approximate eigenfunctions), the authors prove that for the finite measure case, for all \(\epsilon>0\) there holds \[<v,w\circ f_h^n>-\tilde{v}\tilde{w} =\sum_{j>n}\mu(\tau>j)\tilde{v}\tilde{w}+O(n^{-q})\big\||v\big\|||w|_{\infty},\] where \(\tau\) is the first return time to \(Y\), \(q=\beta-\epsilon\) if \(\beta \geq 2\) and \(q=2\beta-2\) if \(\beta \in (1,2).\) In the infinite case, they obtain for \(\beta \in (\frac{1}{2},1]\), \[\lim_{n \longrightarrow +\infty} \tilde{\ell}(n) n^{1-\beta} <v,w \circ f_h^n>=\begin{cases} \frac{\sin(\beta \pi)}{\pi} \tilde{v}\tilde{w} & \textrm{if}\; \frac{1}{2}<\beta<1,\\ \tilde{v}\tilde{w} & \textrm{if}\; \beta=1. \end{cases} \] and for \(\beta \in (0,\frac{1}{2}]\), or \(\beta \in (0,1]\) and \(\tilde{v}=\tilde{w}=0\), and for any \(\epsilon>0\), \[<v,w\circ f_h^n>=O(n^{-(\beta-\epsilon)}\big\||v\big\|||w|_{\infty}).\]