Multifractal formalism for infinite multinomial measures (Q1902809): Difference between revisions

This paper gives a rigorous mathematical frame for statements made in a joint paper of Mandelbrot, Evertsz and Hayakawa motivated by anomalies first observed with DLA. For an infinite family \((w_i)_{i\in\mathbb{N}}\) of contractions in \(\mathbb{R}^d\) consider the measure \(\mu\) defined by \(\mu=\sum^\infty_{i=1} p_i\mu(w_i^{-1}(.))\) for given probability weights together with certain subsets \(K_\alpha\) of its support \(K\). Let \(\lambda_i\) be the contraction ratio associated to \(w_i\) and let \(\tau\) be the function defined for the real argument \(q\) as the solution of \(\sum^\infty_{i=1}p^q_i \lambda_i^{\tau(q)}=1\) provided that one exists and to be \(\infty\) for \(q\) otherwise. The above sets \(K_\alpha\) are exactly those points in \(K\) which have Hölder exponent \(\alpha\), that is, the sequence of log ratios of local measures to diameter tends to \(\alpha\). Amongst other results there are given two conditions which ensure that the Hausdorff dimension of \(K_\alpha\) is equal to the Legendre transform \(f(\alpha)\) at \(\alpha\) of the function \(\tau\).