Random countable iterated function systems with overlaps and applications

TL;DR

This paper investigates invariant measures for random countable conformal iterated function systems with overlaps, establishing their dimensional properties and applying results to random continued fractions and other classes of systems.

Contribution

It introduces new methods for analyzing infinite, random IFS with overlaps without separation conditions and provides dimension formulas and estimates for their invariant measures.

Findings

Invariant measures are dimensionally exact under finite entropy.

Dimension formulas are derived for random infinite conformal IFS with overlaps.

Results apply to random continued fractions and systems related to Kahane-Salem sets.

Abstract

We study invariant measures for random countable (finite or infinite) conformal iterated function systems (IFS) with arbitrary overlaps. We do not assume any type of separation condition. We prove, under a mild assumption of finite entropy, the dimensional exactness of the projections of invariant measures from the shift space, and we give a formula for their dimension, in the context of random infinite conformal iterated function systems with overlaps. There exist many differences between our case and the finite deterministic case studied in [7], and we introduce new methods specific to the infinite and random case. We apply our results towards a problem related to a conjecture of Lyons about random continued fractions ([10]), and show that for Lebesgue-almost all parameters \lambda > 0, the invariant measure \nu_\lambda is exact dimensional. The finite IFS determining these continued fractions is not hyperbolic, but we can associate to it a random infinite IFS of contractions which have overlaps. We study then also other large classes of random countable iterated function systems with overlaps, namely: a) several types of random iterated function systems related to Kahane-Salem sets; and b) randomized infinite IFS in the plane which have uniformly bounded number of disc overlaps. For all the above classes, we find lower and upper estimates for the pointwise (and Hausdorff, packing) dimensions of the invariant measures.