First and higher order uniform dual ergodic theorems for dynamical systems with infinite measure (Q1955786)

Let \((X,\mu )\) be an infinite measure space and \(f:X\rightarrow X\) be a conservative measure-preserving transformation with transfer operator \(L:L^{1}(X)\rightarrow L^{1}(X)\). The transformation \(f\) is ``pointwise dual ergodic'' if there exists a positive sequence \(a_{n}\) such that \(a_{n}^{-1}\sum_{j=0}^{n}L^{j}v\rightarrow \int_{X}vd\mu\) a.e. as \(n\rightarrow \infty ,\) for all \(v\in L^{1}(X)\). Furthermore, if there exists \(Y\subset X\) with \(\mu (Y)\in (0,\infty )\) such that \(a_{n}^{-1}\sum_{j=0}^{n}L^{j}1_{Y}\rightarrow \mu (Y)\) uniformly on \(Y,\) then \(Y\) is referred to as a Darling-Kac set and one refers to \(f\) as ``uniformly dual ergodic''. The article is mainly devoted to intervals maps with indifferent fixed points, as Pomeau-Manneville-Liverani intermittency maps, studied by \textit{C. Liverani} [Ergodic Theory Dyn. Syst. 19, No. 3, 671--685 (1999; Zbl 0988.37035)] and \textit{M. Thaler} [Isr. J. Math. 46, 67--96 (1983; Zbl 0528.28011); ibid. 91, No. 1--3, 111--127 (1995; Zbl 0846.28007)], who have shown that the convergence \(a_{n}^{-1}\sum_{j=0}^{n}L^{j}v\;\)for the mapping \[ f(x)=\begin{cases} x(1+2^{\alpha }x^{\beta }), & 0<x<1/2 \\ 2x-1, & 0\leq x\leq 1 \end{cases} \tag{1} \] is uniform on compact subsets of \((0,1]\) for all observables of the form \( v=u/h,\) where \(u\) is Riemannian integrable and \(h\) is the density. In the paper under review, the authors generalize the proof of \textit{J. Karamata}'s theorem by the method of approximation by polynomials [M. Z. 32, 319--320 (1930; JFM 56.0210.01); J. Reine Angew. Math. 164, 27--39 (1931; Zbl 0001.27302)] to the operator case and present a simple proof of uniform dual ergodicity for a large class of dynamical systems with infinite measure and, jointly with author's results [Invent. Math. 189, No. 1, 61--110 (2012; Zbl 1263.37011)] strengthened the results in [M. Thaler, 1995, loc. cit.; \textit{R. Zweimüller}, Ergodic Theory Dyn. Syst. 20, No. 5, 1519--1549 (2000; Zbl 0986.37008)]. In particular, for the AFN maps the following theorem is proved. Theorem 1.1: Suppose that \(f:X\rightarrow X\) is an AFN map with regularly varying tail probabilities, \(\beta \in [ 0,1].\) Consider observables of the form \( v=\xi u,\) where \(\xi \) is \(\mu \)-integrable and bounded variation on \(X,\) and \(u\) is Riemann integrable. Then \[ \lim_{n\rightarrow \infty }a_{n}^{-1}\sum_{j=0}^{n}L^{j}v=\int_{X}vd\mu \tag{2} \] uniformly on compact subsets of \(X^{\prime }\), where \(a_{n}=D_{\beta}^{-1}n^{\beta }m(n)^{-1},D_{\beta }=\Gamma (1-\beta )\Gamma (1+\beta )\), \(m(n)=l(n)\) for \(\beta \in [ 0,1),m(n)=\tilde{l}(n):=\sum_{j=0}^{n}l(j)j^{-1}\) for \(\beta =1\), and \(l(n)\) is slowly varying. Theorem 1.3 states that in the case of the dynamical system \[ f(x)= \begin{cases} x(1+xe^{-1/x}), & 0<x<1/2 \\ 2x-1, & 0\leq x\leq 1 \end{cases} \tag{3} \] for a Hölder function \(v:[0,1]\rightarrow \mathbb{R}\) or a bounded variation one supported on compact subsets of \((0,1]\), the equality \[ c\sum_{j=0}^{n}L^{j}v=\log n\int_{0}^{1}vd\mu +O(1) \tag{4} \] holds uniformly on compact subsets of \((0,1],\) \(c:=\frac{1}{2}h(\frac{1}{2})\). In the case of the mapping (1) the authors obtain an error term of order \(O(n^{\beta }/\log n)\) (Theorem 3.2(a)). Theorem 1.5 refines this result as \[ \sum_{j=0}^{n}L^{j}v=(C_{0}n^{\beta }+C_{0}n^{\beta }+C_{0}n^{\beta }+\dots+C_{0}n^{\beta })\int_{0}^{1}vd\mu +O(\log ^{\tau }n) \tag{5} \] uniformly on compact subsets of \((0,1],\) where \(C_{j},j=1,2,\dots,k,\) are real nonzero constants, depending on \(f\) and \(k=\max \{j\geq 0:(j+1)\beta -j>0\}.\) Some interesting results are obtained by the authors during the process of proving Theorem 1.5. Namely, they obtain a generalization of a complex Tauberian theorem with remainder, Theorem 1.8, which eventually has not been previously known.