On inhomogeneous Bernoulli convolutions and random power series (Q416443)

The results of \textit{Y. Peres} and \textit{B. Solomyak} [Math. Res. Lett. 3, No. 2, 231--239 (1996; Zbl 0867.28001), Trans. Am. Math. Soc. 350, No. 10, 4065--4087 (1998; Zbl 0912.28005)] on absolute continuity and singularity of homogeneous Bernoulli convolutions are extended to inhomogeneous ones. The results are generalized to random power series given by inhomogeneous Markov chains. An Erdős-Salem type theorem for inhomogeneous Bernoulli convolutions is proved in the following form: fix a sequence \((p_n)_{n \in N_0}\) of probabilities \(p_n \in (c,1-c)\) for some \(c \in (0,\frac 12)\). The Fourier transform \(\hat{\mu}_{\beta}(\xi)\) of the convolution measure \(\mu_{\beta}\) does not tend to zero for \(\xi \to \infty\) if and only if \(\beta\) is the reciprocal of a Pisot number.