The structure of the minimum size supertail of a subspace partition (Q521856)

In this paper, the authors continue the study of the supertail of a subspace partition, building on earlier work [\textit{O. Heden} et al., Des. Codes Cryptography 69, No. 3, 305--316 (2013; Zbl 1278.51005)]. Let \(V(n,q)\) be a vector space of dimension \(n\) over the finite field of order \(q\). A \textit{subspace partition} \(\mathcal{P}\) of \(V\) is a collection of subspaces of \(V\) such that each non-zero vector is in exactly one subspace of \(\mathcal{P}\). Let \(d_1<\ldots<d_m\) be the distinct dimensions that occur in the subspace partition, and denote by \(n_i\) the number of spaces in \(\mathcal{P}\) of dimension \(d_i\), then \([d_1^{n_1},\dots,d_m^{n_m}]\) is called the \textit{type} of the subspace partition. Some necessary conditions on the parameters are known for the existence of a subspace partition of a certain type. For any integer \(s\) with \(1<s\leq m\), the \textit{\(d_s\)-supertail} of \(\mathcal{P}\) is the set of subspaces of \(\mathcal{P}\) of dimension less than \(d_s\). Denote by \(\sigma_q(n,t)\) the minimum size of a subspace partition \(\mathcal{P}\) of \(V\) such that \(t\) is the largest occurring dimension, then it is shown in [loc. cit.] that the number of subspaces at least \(\sigma_q(d_s,d_{s-1})\). In that same paper, it was shown that in case of equality, and if \(d_s\geq 2d_{s-1}\) the subspaces of the supertail form a \(d_s\)-subspace. In this paper, the authors investigate what happens in case of equality, in the case that \(d_s\) is not at least \(2d_{s-1}\). Their main theorem (Theorem 5) is the following: Theorem. Let \(\mathcal{P}\) be a subspace partition of \(V(n,q)\) of type \([d_1^{n_1},\ldots,d_m^{n_m}]\). Let \(2\leq s\leq m\) and suppose that the \(d_s\)-supertail of \(\mathcal{P}\) has \(\sigma_q(d_s,d_{s-1})\) elements, where \(d_s<d_{s-1}\). Furthermore, assume that one of the following conditions hold: {\parindent=0.7cm\begin{itemize}\item[(i)] \(s-1\leq 2\), that is, the supertail contains subspaces of at most \(2\) different dimensions. \item[(ii)] \(d_s=2d_{s-1}-1\). \item[(iii)] All the subspaces in \(\mathcal{P}\) that are not in the supertail have the same dimension. \end{itemize}} Then the union of subspaces in the supertail forms a subspace \(W\). Moreover, {\parindent=0.7cm\begin{itemize}\item[(a)] \(s-1=1\), \(n_1=q^{d_1}+1\), and \(\dim(W)=2d_1\), or \item[(b)] \(s-1=2\), \(n_1=q^{d_2}\), \(n_2=1\) and \(\dim(W)=d_1+d_2\). \end{itemize}} Furthermore, the authors conjecture that this theorem holds in all cases, and not only if (i), (ii), or (iii) holds. The proof of this theorem is given as a series of technical lemmas that consist of (sometimes complicated) combinatorial arguments.