Cotlar's ergodic theorem along the prime numbers (Q895429)

Let \((X,\mathcal{B},\mu,S)\) be a dynamical system on a measure space \(X\), and assume that \(\mu\) is \(\sigma\)-finite and \(S: ~T \mapsto T\) is invertible and measure-preserving. By a classical result of \textit{M. Cotlar} [Zbl 0071.33402] the ergodic truncated Hilbert transform \[ \lim_{N \to \infty} \sum_{1 \leq |n| \leq N} \frac{f(S^n x)}{n} \] is almost everywhere convergent for all \(f \in L^r(\mu),~r \in [1,\infty)\). In the present paper the authors obtain a corresponding if the sum runs over the set of primes (with positive and negative signs) rather than over all non-zero integers, and the appropriate normalization factor is added. The proof uses Calderon's transference principle to move from the measure space \(X\) to the set of integers, where it has to be shown that a certain maximal operator is bounded in \(\ell^r(\mathbb{Z})\). The proof of this statement uses the Hardy-Littlewood circle method in a way which has been originally developed by Bourgain. Along the way the authors correct an error in a paper of \textit{M. Wierdl}, which dealed with closely related topics [Isr. J. Math. 64, No. 3, 315--336 (1989; Zbl 0695.28007)].