A cross-intersection theorem for vector spaces based on semidefinite programming

Abstract

Let $\mathscr{F}$ and $\mathscr{G}$ be families of $k$- and $\ell$-dimensional subspaces, respectively, of a given $n$-dimensional vector space over a finite field $\mathbb{F}_q$. Suppose that $x \cap y \ne 0$ for all $x \in \mathscr{F}$ and $y \in \mathscr{G}$. By explicitly constructing optimal feasible solutions to a semidefinite programming problem which is akin to Lov\'{a}sz's theta function, we show that $|\mathscr{F}| |\mathscr{G}| \leq {n-1 \brack k-1} {n-1 \brack \ell-1}$, provided that $n \geq 2k$ and $n \geq 2\ell$. The characterization of the extremal families is also established.