Long-range correlations of sequences modulo 1 (Q2071708)

This article deals with the distribution of the fractional parts \((\{ a(n) \})_{n\in\mathbb{N}}\) of real sequences \( (a(n))_{n\in\mathbb{N}}\), i.e., the sequence given by \(a(n)\ (\bmod{1})\) for \(n\in\mathbb{N}\). Classical examples of sequences \(a(n)\) studied in the literature are, e.g., \(a(n)=\alpha \sqrt{n}\), \(a(n)=\alpha n^2\), or, more generally, \(a(n)=\alpha n^\beta\) for some \(\alpha\in \mathbb{R}\setminus \{0\}\) and \(\beta>0\). In particular, the paper deals with a way of assessing the uniformity of distribution of such fractional parts of sequences by considering the distribution of gaps between neighboring points of the sequence. However, it is in general very challenging to directly study the gap distribution for a given sequence, which is why one frequently considers so called \(m\)-level correlations of the first \(N\) points of the sequence instead. The \(m\)-level correlation of a sequence is called Poissonian if it converges, for a function \(f\) in a certain class, to the integral of that function. The sequence is called Poissonian if all \(m\)-level correlations for \(m\ge 2\) are Poissonian, and it can be shown that the gap distribution of such a sequence is exponential. The current paper, however, does not directly study the \(m\)-level correlations, but a coarser measure, namely so-called long-range correlations, where the support of the functions \(f\) under consideration is increased. The author derives a general method to show that these long-range correlations are Poissonian, and gives a relation to Weyl sums. In particular, the method is applied to some concrete examples of sequences that are frequently studied in the literature.