Estimates for the lower and the upper dimension of a measure (Q1389183)

The function \(\tau\) which appears in multifractal analysis is known to be adapted to calculate the lower and the upper dimension of a measure, in particular when \(\tau'(1)\) does exist. The paper under review considers mainly the case where \(\tau\) has no derivative at 1. For that purpose, a notion of lower and upper dimension based on the Tricot dimension is introduced, which enables to interpret the left derivative of \(\tau\) at 1. Optimal estimates for the upper and the lower dimensions are thus obtained, for some good classes of measures, as well as upper and lower bounds in terms of Rényi dimensions. These results are applied to quasi-Bernoulli measures and yield the derivability of the function \(\tau\), precising results in [\textit{G. Brown}, \textit{G. Michon} and \textit{J. Peyrière}, J. Stat. Phys. 66, 775-790 (1992)]. This nice paper ends with well-chosen examples and an application to quasi-symmetric functions.