Optimal bounds for exponential sums in terms of discrepancy
Let \(0\leq x_ 1,...,x_ N<1\) be real numbers and let \[ D^*_ N=D^*_ N(x_ 1,...,x_ N)=\sup_{0<t\leq 1}| (1/N)\sum^{N}_{x_ n<t,n=1}1-t| \] be their discrepancy. If \(c_ N\) denotes the least constant such that \(| \sum^{N}_{n=1}e^{2\pi ix_ n}| \leq c_ N D^*_ NN\) for all such sequences, then the inequalities \(4-10 N^{-2/3}\leq c_ N\leq 4- 2N^{-2/3}\) will be proved. The proof uses elementary methods like inequalities about the Sine-function as well as results of \textit{H. L. Montgomery} and \textit{H. Niederreiter} [Bull. Greek Math. Soc. 18, 291-301 (1977; Zbl 0409.10023)]. An alternative method which is based on Koksma's inequality is presented. This approach leads to the weaker estimation \(c_ N\leq 4-0.9N^{-2/3}.\)
- On discretizing integral norms of exponential sums
- scientific article; zbMATH DE number 5868119 (Why is no real title available?)
- On the exponent of discrepancies
- On Irregularities of Distribution II
- scientific article; zbMATH DE number 1774261 (Why is no real title available?)
- Weighted exponential sums and discrepancy
This page was built for publication: Optimal bounds for exponential sums in terms of discrepancy
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3813869)