A characterization of incomplete sequences in vector spaces (Q645958)

A sequence of elements in an abelian group is incomplete if there exists a group element which cannot be expressed as a sum of elements of \(A\). The paper characterizes incomplete sums in \(\mathbb{F }_p^d\). Theorem 1.2: For any positive integer \(d\) and positive constants \(\alpha, \beta\) there is a positive \(\varepsilon\) such that the following holds for every sequences \(A\) in \(\mathbb{F }_p^d\), where \(p\) is a sufficiently large prime. Either there is an \(m \leq \beta p\) such that \[ m^*A := \left\{ \sum_{x \in B}x: \;B \subset A, |B|=m \right\}= \mathbb{F }_p^d \] or \(A\) can be partitioned into disjoint subsequences \(A_0, \ldots , A_l\) such that 1) \(|A_0|\leq \alpha p\), 2) \(|A_i|= \lfloor \varepsilon p \rfloor, i \geq 1\), 3) there is a subspace \(H\) such that \(A_i\) \((1\leq i\leq l)\) is contained in a translate of \(H\), and \(m^* A_0\) contains a translate of \(H\) for some \(1 \leq m \leq |A_0|\). The proof makes use of ideas on sumset growth similar to those in \textit{N. Alon} and \textit{M. Dubiner} [Combinatorica 15, No. 3, 301--309 (1995; Zbl 0838.11020)] and the authors [J. Comb. Theory, Ser. A 116, No. 4, 936--959 (2009; Zbl 1196.11048)]. The result is applied to a new estimate on the Olson constant in \(\mathbb{F}_p^3\). For an abelian group \(G\), the Olson constant \(\text{OL}(G)\) is the smallest integer such that no subset of \(G\) of cardinality \(\text{OL}(G)\) is zero-sum free. It is known that \(\text{OL}(\mathbb{F }_p)\leq \lceil \sqrt{2 p}+5 \log p\rceil\) and \(\text{OL}(\mathbb{F }_p^2)=p +\text{OL}(\mathbb{F }_p)-1\), see \textit{W. D. Gao, I. Z. Ruzsa} and \textit{R. Thangadurai} [J. Comb. Theory, Ser. A 107, No. 1, 49--67 (2004; Zbl 1107.11014)]. In Theorem 1.3 the authors solve the inverse problem, i.e. they characterize zero-sum free sets of size \(p+\text{OL}(\mathbb{F }_p)-2\). For higher dimension \(d\), Gao, Ruzsa and Thangadurai conjectured that \(\text{OL}(\mathbb{F }_p^d)=p +\text{OL}(\mathbb{F }_p^{d-1})-1\). This implies that for every \(\gamma>0\) and sufficiently large \(p\) the following upper bound holds: \[ \text{OL}(\mathbb{F }_p^d)\leq (d-1+\gamma)p. \] The authors prove the latter in the case \(d=3\).