Multifractal analysis of measures arising from random substitutions
Andrew Mitchell, Alex Rutar
TL;DR
The paper develops a comprehensive multifractal framework for frequency measures arising from primitive random substitutions, introducing the inflation word L^q-spectrum and proving its equality with the true L^q-spectrum for all q≥0. Under recognisability, the authors obtain a closed-form L^q-spectrum and establish the multifractal formalism, including a variational principle for relative local dimensions and concave, analytic spectra. The results unify entropy, topological entropy, and L^q-spectra within a single symbolic-dynamical setting, and they delineate the roles of separation conditions and recognisability in ensuring sharp bounds and exact formulas. The work yields explicit formulas under disjoint or identical-set conditions, recovers known entropy results, and provides sharp negative-q results in the recognisable regime, supported by diverse examples and counterexamples.
Abstract
We study regularity properties of frequency measures arising from random substitutions, which are a generalisation of (deterministic) substitutions where the substituted image of each letter is chosen independently from a fixed finite set. In particular, for a natural class of such measures, we derive a closed-form analytic formula for the $L^q$-spectrum and prove that the multifractal formalism holds. This provides an interesting new class of measures satisfying the multifractal formalism. More generally, we establish results concerning the $L^q$-spectrum of a broad class of frequency measures. We introduce a new notion called the inflation word $L^q$-spectrum of a random substitution and show that this coincides with the $L^q$-spectrum of the corresponding frequency measure for all $q \geq 0$. As an application, we obtain closed-form formulas under separation conditions and recover known results for topological and measure theoretic entropy.
