Explicit constructions in the classical mean squares problem in irregularities of point distribution (Q2782018)

In this paper an explicit construction of distributions \(\mathcal{D}_N\) of \(N\) points in the \(k\)-dimensional unit cube \(U^k\) with the minimal order of the \(L_2\)-discrepancy \(\mathcal{L}_2[\mathcal{D}_N]<C_k(\log N)^{\frac{1}{2}(k-1)}\), where the constant \(C_k\) is independent of \(N\). As an essential tool ideas from coding theory are used. In particular, the authors consider codes over finite fields with large weights simultaneously in two different metrics -- the well-known Hamming metric and a new non-Hamming metric arising recently in coding theory.NEWLINENEWLINEFirst constructions of such distributions were given in dimensions \(k=2\) and \(k=3\) by Davenport (1956) and Roth (1979) and in arbitrary dimensions by Roth (1980). Until recently apart from Davenport's construction for dimension \(k=2\) all known constructions involve probabilistic arguments and are therefore not explicit. Very recently Larcher and Pillichshammer have studied the problem for the dimension \(k=3\) by an approach using point sets constructed by Faure (1982).NEWLINENEWLINEThe method of the present paper is a breakthrough in the theory of point distributions since it gives a complete and explicit solution of the problem. The proofs depend on a delicate analysis of Walsh series expansions.