Arithmetic progressions in sets of fractional dimension (Q1034679)

Starting from an old question of Erdős, the authors consider the following modified problem: if there is a finite set such as \(\{0,1,2\}\) that might be universal. The authors prove that if \(\alpha\) is sufficiently close to 1, and \(E\) supports a probability measure obeying some dimensionality and Fourier decay conditions, then \(E\) contains non-trivial 3-term arithmetic progressions. The main result of the paper is the following theorem: Assume that \(E\subset [0,1]\) is a closed set which supports a probability measure \(\mu\) with the following properties: \[ (A)\;\;\mu([x,x+\varepsilon])\leq C_1\varepsilon^\alpha,\quad 0<\varepsilon\leq 1, \] \[ |{\widehat\mu}(k)|\leq C_2(1-\alpha)^{-B}|k|^{-\beta/2}, \quad k\neq 0, \] where \(0<\alpha<1\) and \(2/3<\beta\leq 1\). If \(\alpha>1-\varepsilon>0\), \(\varepsilon>0\), is a sufficiently small constant depending only on \(C_1, C_2, B, \beta\), then \(E\) contains a non-trivial 3-term arithmetic progression. The proof of the main theorem is difficult, it involves some originality, moreover the use of the technique of restriction estimate is successful, which is the key feature of the proof. The authors give a new construction of Salem-type sets satisfying the conditions of the above theorem, they discuss also Salem's construction and Brownian images of Kahane.