Exact values and improved bounds on $k$-neighborly families of boxes

A finite family $\mathcal{F}$ of $d$-dimensional convex polytopes is called $k$-neighborly if $d-k\le \text{dim}(C\cap C^{\mathrm{\prime }})\le d-1$ for any two distinct members $C,C^{\mathrm{\prime }}\in \mathcal{F}$. In 1997, Alon initiated the study of the general function $n(k,d)$, which is defined to be the maximum size of $k$-neighborly families of standard boxes in ${\mathbb{R}}^d$. Based on a weighted count of vectors in $\{0,1{\}}^d$, we improve a recent upper bound on $n(k,d)$ by Alon, Grytczuk, Kisielewicz, and Przesławski for any positive integers $d$ and $k$ with $d\ge k+2$. In particular, when $d$ is sufficiently large and $k\ge 0.123d$, our upper bound on $n(k,d)$ improves the bound ${\sum }_{i=1}^k{2}^{i-1}\left(\genfrac{}{}{0px}{}{d}{i}\right)+1$ shown by Huang and Sudakov exponentially. Furthermore, we determine that $n(2,4)=9$, $n(3,5)=18$, $n(3,6)=27$, $n(4,6)=37$, $n(5,7)=74$, and $n(6,8)=150$. The stability result of Kleitman’s isodiametric inequality plays an important role in the proofs.