Local dimensions for the random \(\beta \)-transformation (Q374065): Difference between revisions

For a probability measure \(\mu\) on a metric space \((X,\rho)\), the local dimension at \(x \in X\) is defined by \(d(\mu,x) = \lim_{r\to0} \log \mu(B_\rho(x,r))/ \log r\), when the limit exists. Random \(\beta\)-transformations were defined in [\textit{K. Dajani} and \textit{C. Kraaikamp}, Ergodic Theory Dyn. Syst. 23, No. 2, 461--479 (2003; Zbl 1035.37006)]. By [\textit{K. Dajani} and \textit{M. de Vries}, J. Eur. Math. Soc. (JEMS) 7, No. 1, 51--68 (2005; Zbl 1074.28008)], they are isomorphic to the full shift and, when \(\beta\) is a generalised multinacci number, to a subshift of finite type. Here, ``generalised multinacci'' means that the \(\beta\)-expansion of \(1\) is finite and contains no zeros (except for the tail). The authors study local dimensions of random \(\beta\)-expansions for the invariant measure \(\nu_\beta\) given by the uniform Bernoulli measure on the full shift, for the Bernoulli convolution measure \(\mu_\beta\) (which is a projection of \(\nu_\beta\)), as well as for the invariant measure \(\nu_{\beta,p}\) given by an asymmetric Bernoulli measure \(m_p\). For a generalised multinacci number \(\beta\), the local dimension with respect to \(\nu_\beta\) (and a suitable metric) is given by the frequency of the switch regions; in particular, it is almost everywhere equal to \(\log_\beta\lceil\beta\rceil + (1-\mu_\beta(S)) \log_\beta 2\), where \(S\) is the union of switch regions, while it equals \(\log_\beta\lceil\beta\rceil + \log_\beta 2\) for points with unique \(\beta\)-expansion; the measure \(\nu_\beta\) is thus multifractal. When \(\beta\) is moreover a Pisot number, the local dimension at \(x\) with respect to \(\mu_\beta\) (and the projection of the latter metric) is given by the number of different \(\beta\)-representations of \(x\), thus almost everywhere equal to \(\log_\beta\lceil\beta\rceil - \log_\beta \gamma\) for some constant \(\gamma\). For the Euclidean metric, this result is due to \textit{D.-J. Feng} and \textit{N. Sidorov} [Monatsh. Math. 162, No. 1, 41--60 (2011; Zbl 1273.11018)]. When \(\beta\) is the golden ratio, similar results are obtained for asymmetric Bernoulli measures.