Supernormal vector configurations (Q596658): Difference between revisions

Let \(B=\{b_1,\dots,b_n\}\subseteq\mathbb{Z}^m\), and let cone\((B) \subseteq\mathbb{R}^m\) denote the cone spanned by \(B\). Then \(B\) is normal if cone\((B)\cap\mathbb{Z}^m\) consists of non-negative integer combinations of \(B\), and is supernormal if cone\((B')\cap\mathbb{Z}^m\) consists of non-negative integer combinations of \(B\cap\text{cone}(B')\) for every \(B' \subseteq B\). Finally, recall that \(B\) is unimodular if every \(B'\subseteq B\) is normal; supernormality is thus a condition lying between normality and unimodularity. Of particular interest here are the vector configurations of the form \[ B=\bigl\{(1,u,v) \in\mathbb{Z}^2\mid (u,v)\in P\cap \mathbb{Z}^3\bigr\}, \] where \(P\) is a lattice polygon in the plane. It is shown that these are all supernormal. The chamber complex of \(P\) is the common refinement of all lattice triangulations of \(P\). If \(\mu(P)\) denotes the maximum number of edges of any region in the chamber complex, an open problem posed here asks if there is a universal bound for \(\mu(P)\) (that is, independent of \(P)\). A note added in proof says that the answer has since been shown to be negative. Because the chamber complex of \(B\) is the Gröbner fan of the associated binomial (initial) ideal \(J_B\), there is thus no finite bound \(\varphi(3)\) on the number of facets of any of its cones; this has implications for the sensitivity of integer programming in dimension 3.