On the distribution of the van der Corput sequence in arbitrary base (Q2407719): Difference between revisions

For an integer \(b\geq 2\), the base \(b\) van der Corput sequence \((x_n)_{n\geq 0}\) is defined by \(x_n=\sum_{j=1}^m a_jb^{-j}\), where \(a_j\in\{0,1,\dots,b-1\}\) are digits of the base \(b\) representation of the integer \(n\geq 0\) as \(n=\sum_{j=1}^m a_jb^{j-1}\). In the paper under review, the author studies the random aspects of the base \(b\) van der Corput sequence, by proving that the sum \(S(N)=\sum_{n=0}^{N-1}(\frac{1}{2}-x_n)\) satisfies the central limit theorem, asserting that for any integer \(M>b^2\) and any real number \(\lambda\) one has \[ \frac{1}{M}\#\left\{0\leq N<M:\frac{S(N)-c(b)\log_b{N}}{\sqrt{d(b)\log_b{N}}}<\lambda\right\}=\Phi(\lambda)+O\left(\sqrt[4]{\frac{\log\log_b{M}}{\log_b{M}}}\right), \] where \(\Phi(\lambda)\) denotes the distribution function of the standard normal distribution, \(c(b)=\frac{b^2-1}{12b}\) and \(d(b)=\frac{b^4+120b^3-480b^2+600b-241}{720b^2}\). The author proves similar result replacing \(S(N)\) by the \(L^p\) norm \[ \|\Delta_N\|_p=\left(\int_0^1|\Delta_N(x)|^p\,dx\right)^\frac{1}{p}, \] where \(\Delta_N(x)\) is the discrepancy function of \(x_n\), and also gives a method to generalize the above central limit theorem for sums of the form \(\sum_{n=0}^{N-1}f(x_n)\), where the function \(f:[0,1]\to\mathbb{R}\) satisfies some conditions.