On the exceptional set for absolute continuity of Bernoulli convolutions (Q2510445): Difference between revisions

An iterated function system (IFS) is a family \(\mathcal F=(f_1,\ldots,f_m)\) of strict contractions on some complete metric space \(\mathbb R\). Then \(A=A(\mathcal F)\) is a corresponding self-similar set such that \(A=\bigcup_{i=1}^nf_i(A)\). Denote the open simplex in \(\mathbb R^m\) by \(\mathbb P_m=\{(p_1,\ldots,p_m)\colon p_i>0,\sum_{i=1}^mp_i=1\}\). Then there exists a unique measure \(\mu=\mu(\mathcal F,p)\) such that \(\mu=\sum_{i=1}^np_i\,\mu\circ f_i^{-1}\) with support \(A(\mathcal F)\). For \(p\in\mathbb P_m\) the similarity dimension is defined by \[ s(\mathcal F,p)={\sum_{i=1}^mp_i\log p_i\over\sum_{i=1}^mp_i\log r_i}, \] where \(r_i\) is the contraction ratio of \(f_i\). We denote the Hausdorff dimension by \(\dim_H\), and \[ \dim_\mu=\inf\{\dim_H(B)\colon \mu(B)>0\}. \] The Bernoulli convolution \(\nu_\lambda^p\) corresponding to a contraction ratio \(\lambda\in(0,1)\) and weight \(p\in(0,1)\) is the distribution of the random sum \(\sum_{n=0}^\infty\pm\lambda^n\), where \(P(+)=p\) and \(P(-)=1-p\). \bigskip Then the main results of this article are the following two theorems, and they are proved by the decay of the Fourier transform of self-similar measures and their convolutions. \medskip {Theorem 1.1.} There exists a set \(E\subset(1/2,1)\) of Hausdorff dimension 0, such that \(\nu_\lambda=\nu_\lambda^{1/2}\) is absolutely continuous for all \(p\), \(\lambda\) such that \(s(\lambda,p)>1\) and \(\lambda\in(1/2,1)\backslash E\). \bigskip {Theorem 1.2.} Let \(a_a,\ldots,a_m\) be distinct fixed numbers, and for \(\lambda\in(0,1)\), let \(\mathcal F_\lambda=(\lambda x+a_1,\ldots,\lambda x+a_m)\). There exists a set \(E\) of zero Hausdorff dimension, such that if \(\lambda\in(0,1)\backslash E\) and \(p\in\mathbb P_m\) is such that \(s(\mathcal F_\lambda,p)>1\), then \(\mu(\mathcal F_\lambda,p)\) is an absolutely continuous measure.