Compound Poisson statistics for dynamical systems via spectral perturbation (Q6931041)

The present paper investigates return-time statistics for dynamical systems. The authors use the technique of compound Poisson random variables. Given a complete metrizable space \(M\) and a probability space \((\Omega, \mathcal{F},m)\), a random variable \(Z:M\to\mathbb{N}\) is said to be {compound Poisson distributed} if there exists a Poisson random variable \(N\) and a sequence \(X_k:M\to\mathbb{N},\ k\ge 1,\) of i.i.d. random variables such that \(Z=\sum_{k=1}^N X_k\). The authors consider the dynamical system \(T^n_\omega=T_{\sigma^{n-1}\omega}\circ\cdots\circ T_{\sigma\omega}\circ T_{\omega},\ n\ge 1\), driven by a function \(\sigma:\Omega\to\Omega\), where \(T_\omega : M\to M\), \(\omega\in\Omega\), is a collection of maps. The desired compound Poisson approach in the deterministic setting via Keller-Liverani perturbation theory is sketched. To investigate return-time statistics in the general setting of random dynamical systems, a sequence of random target sets \(H_{\omega,n}\subset M\), such that \(H_{\omega,n+1}\subset H_{\omega,n}\), for any \(n\in\mathbb{N}\) and \(\omega \in\Omega\), is defined. The main result of this study states that, for an appropriate choice of the target sets, the random variable \[ \sum_{j=0}^{n-1}1_{H_{\sigma^j\omega,n}}(T^j_\omega)\] converges in distribution to a compound Poisson distribution, as \(n\to\infty\). Thus, it is proven that the characteristic function of the above random variable converges pointwise to the characteristic function of a random variable which is discrete and infinitely divisible, and therefore compound Poisson distributed. Suitable examples, in both the deterministic and random settings, are highlighted. The work is well anchored in the literature.