On zero-sum subsequences of restricted size (Q678401)

\textit{P. Erdös}, \textit{A. Ginzburg} and \textit{A. Ziv} [Bull. Res. Council Israel 10F, 41-43 (1961; Zbl 0063.00009)] showed that if \(a_1,\dots, a_{2n-1}\) is a sequence of \(2n-1\) elements of integers, then there is a subsequence \(a_{k_1}\dots,a_{k_n}\) of \(n\) elements for which \(n\mid a_{k_1}+\cdots+ a_{k_n}\). This result has been generalized in several ways. Let \(C_n\) be the cyclic group of order \(n\) and let \(C_n^k= C_n\oplus C_n\oplus\cdots\oplus C_n\) (\(k\) times). Harborth proved if \(n=2^a3^b\) then any sequence of \(4n-3\) elements in \(C_n^2\) contains an \(n\)-element subsequence with sum zero. Kemnitz proved the same result for \(n=2^a3^b5^c7^d\) and conjectured that it is true for every positive \(n\). The author proves that if this conjecture is true for \(n\), and \(m\) is a positive integer for which \(n\geq ((3m-3)(m-1)m^2+3)/4m\), then the conjecture is true also for \(nm\). By the Kemnitz result the author gets: if \(n=2^a3^b5^c7^d m_1\dots m_r\) with \(m_1\geq\dots\geq m_r\) and \(2^a3^b5^c7^d\geq ((3m_1-4)(m_1- 1)m_1^2+3)/4m_1\), then the ``\(4n-3\) conjecture'' is true for \(n\).