On dispersion and Markov constants (Q5917759)

Let \(\{x_n\}\) be a sequence in \([0,1]\). Define \[ d_n= \sup_{0\leq x\leq 1} \min_{1 \leq n\leq N} |x-x_n |,\;D\bigl(\{x_k\} \bigr) = \limsup_{N\to\infty} Nd_N. \] In the paper the case \(x_n = \text{fr}(n\alpha)\) is considered, where \(\alpha\) is a real irrational number and \(\text{fr}(\cdot)\) denotes the fractional part. The main result is Theorem 2: For \((a,b,c)=1\), \(d= b^2 -4ac\), \(\alpha = (b+d^{1/2})/(2a) = [a_0; \overline{a_1, \dots, a_k}]\) (the last expression is the periodic regular continued fraction expansion of \(\alpha)\) we have \[ D(\alpha) = \sup_{1 \leq i\leq k} \bigl(2d+ (d+R_i) d^{1/2}/Q_i \bigr)/(4d), \] where the complete quotients of \(\alpha\) are \(\alpha_i = [a_i;a_{i+1}, \dots] = (P_i+d^{1/2})/Q_i\) and \(R_i\) can be expressed in terms of \(P_i\) and \(Q_i\). The proofs are carried out by establishing a relation between \(D (\alpha)\) and the Markov constant of \(\alpha\) defined by \(M(\alpha) = \liminf_{n\to\infty} n|n \alpha |\), \(|\dots |\) being the distance to the nearest integer.