# Event № 598

*Abstract:*

Given $\lambda\in (0,1)$, consider the distribution of the random series $\sum_{n=0}^\infty \pm \lambda^n$, where the signs are chosen randomly and independently, with probabilities $(\half,\half)$.This is a probability measure on the real line, which can be expressed as an infinite Bernoulli convolution product. These measures have been intensively studied since the mid-1930's, because they arise, somewhat unexpectedly, in many different areas, including harmonic analysis, number theory, and number theory. The case of $\lambda < 1/2$ is simple: we get the classical Hausdorff-Lebesgue measure on a Cantor set of constant dissection ratio and zero length, hence the measure is singular. For $\lambda=1/2$ we get a uniform measure on $[-2,2]$, but the case of $\lambda>1/2$ is very challenging. The basic question is to decide whether the resulting measure is absolutely continuous or singular, which is still open. It was believed at first that since the support of the measure is an the entire interval $[-(1-\lambda)^{-1}, (1-\lambda)^{-1}]$, it should be absolutely continuous. This turned out to be false: P. Erdos showed in 1939 that the measure is singular for $\lambda$ reciprocal of a Pisot number, e.g. for $\lambda$ equal to the golden ratio $0.618...$

Since then, many mathematicians (including the speaker) worked on this problem, and much is known by now, but it is still an open question whether all numbers in $(1/2,1)$, other than reciprocals of Pisot numbers, give rise to absolutely continuous measures. In the last five years a dramatic progress has occurred, after a breakthrough by M. Hochman, followed by important results due to P. Shmerkin and P. Varju.In the first part of the talk I will outline this recent development.

Bernoulli convolution measures can be generalized in various directions, which leads to new interesting problems. In the second part of the talk I will report on the recent work, joint with M. Hochman, on the dimension of stationary (Furstenberg) measures for random matrix products, and time permitting, on a joint work with S. Saglietti and P. Shmerkin on absolute continuity of non-homogeneous self-similar measures with ``overlap''.