An analogue of Law of Iterated Logarithm for Heavy Tailed Random Variables

Abstract: We study heavy tailed random variables obtained by sampling an observable with a power singularity along an orbit of an expanding map on the circle. In the regime where the observables have infinite variance, we show that the set of limit points of the ergodic sum process, rescaled by $N^{1/s} (\ln N)^\alpha$, where $s$ is the index of limiting stable law, equals to the set of piecewise constant increasing functions with finitely many jumps. The proof relies on the multiple Borel-Cantelli Lemma proven in Dolgopyat, Fayad and Liu [J. Mod. Dyn. 18 (2022) 209--289].