Optimal \(L_{p}\)-discrepancy bounds for second order digital sequences (Q1678655)

Let \(d, N\in \mathbb{N}\) and let \(P_{N,d}\) be an \(N\)-element point set in \([0, 1)^d\). The discrepancy function of \(P_{N,d}\) is defined as \[ D_{P_{N, d}}(\boldsymbol{x})=\frac{1}{N}\sum_{\boldsymbol{z}\in P_{N, d}}\chi_{[\mathbf{0},\boldsymbol{x}]}(\boldsymbol{z})-x_1\cdots x_d, \] where \(\boldsymbol{x}=(x_1, \ldots , x_d)\in [0,1]^d\), \(\chi_{A}\) is the characteristic function of a set \(A\). The \(L_p\)-discrepancy of \(P_{N,d}\) is defined as \(L_{p,N}(P_{N,d})=\|D_{P_{N,d}}|L_{p}([0,1]^d)\|\) for \(p\in [1, \infty]\). For an infinite sequence \(S_d\) in \([0, 1]^d\), \(D^N_{S_d}(\boldsymbol{x})\) denotes the discrepancy function of the first \(N\) terms of \(S_d\). The \(L_p\)-discrepancy of \(S_d\) is defined as \(L_{p,N}(S_d)=\| D^N_{S_d}|L_p([0,1]^d)\|\) for \(p\in [1, \infty]\). \textit{P. D. Proinov} [C. R. Acad. Bulg. Sci. 39, No. 9, 31--34 (1986; Zbl 0616.10042)] proved that for \(p\in (1,\infty]\) and \(d\in \mathbb{N}\), \(d\geq 2\), there exists a positive constant \(c_{p,d}\) such that for every infinite sequence \(S_d\) in \([0, 1)^d\) \[ L_{p,N}(S_d)\geq c_{p,d}\frac{(\log N)^{\frac{d}{2}}}{N} \] holds for infinitely many \(N\in \mathbb{N}\). In this paper, the authors proved that for every order 2 digital \((t, d)\)-sequence \(S_d\) over \(\mathbb{F}_2\) with generating matrices \(E_i=(e_{i,k,\ell})_{k,\ell}\) for which \(e_{i,k,\ell}=0\) for all \(k>2\ell\) and for all \(i\in \{1, 2, \ldots, d\}\) we have \[ L_{p,N}(S_d)\ll _{p,d} 2^t \frac{(\log N)^{\frac{d}{2}}}{N} \] for all \(N\geq 2\) and all \(1\leq p<\infty\). Therefore it is shown that the lower bound by Proinov is best possible for all \(p\in (1,\infty)\).