Invariance principles in metric Diophantine approximation (Q1404807)

Let \(f(x)\) be a positive function tending monotonically to \(0\) as \(x\to\infty\). A classical theorem of Khintchine asserts, that for almost all \(\alpha\in(0, 1)\) (that is, for all \(\alpha\in (0, 1)\), except possibly a set of Lebesgue-measure zero), the inequality \[ \Biggl|\alpha- {p\over q}\Biggr|< {f(q)\over q^2}\tag{\(*\)} \] has at most finitely many solutions or infinitelly many solutions, as to the series \[ \sum^\infty_{q=1} {f(q)\over q}\tag{\(**\)} \] converges or diverges. The case of convergence is immediately settled by the Borel-Cantelli Lemma of probability theory. For the case of the divergence of the series \((**)\) Khintchine originally needed the theory of continued fractions, which, in a later proof he avoided, in order to obtain a generalization to more dimensions. If \((**)\) diverges, one can ask for inequalities about the number of solutions of \((*)\) below a limit. Choose at random, with a uniform distribution in \((0,1)\), an \(\alpha\) and define the random variables \(\xi_q\) \((q= 1,2,\dots)\) to be 1 if \((*)\) holds with an integer \(p\) and 0 otherwise. Then the dependence of the \(\xi_q\)'s is sufficiently weak to ensure the validity of the classical limit theorems of probability theory. As generalizations of results of several authors, the author proves the central limit theorem and the law of iterated logarithm under restrictions on the approximating fractions \(p/q\), like \((p, q)\leq d\) and \(q\equiv r\pmod r\).