Equivalence classes of Latin squares and nets in \(\mathbb{CP}^{2}\) (Q2450836)

An \((n,k)\)-net in \(\mathbb{CP}^2\) consists of \(n\) disjoint \(k\)-subsets of lines \(\mathcal{A}_1\sqcup\cdots\sqcup\mathcal{A}_n\subset\mathbb{CP}^2\) together with a set of points \(\mathcal{X}\subset\mathbb{CP}^2\) satisfying the following two conditions: 1) for all \(1\leq i\neq j\leq n\) and all \(L\in\mathcal{A}_i\) and \(K\in\mathcal{A}_j\) the intersection point \(L\cap K\) belongs to \(\mathcal{X}\), and 2) for every point \(p\in\mathcal{X}\) and for every \(1\leq i\leq n\) there is a unique line \(L\in\mathcal{A}_i\) that contains \(p\). In the paper under review, the authors focus on the \(n=4\) case, motivated by the following question of \textit{S. Yuzvinsky} [Compos. Math. 140, No. 6, 1614--1624 (2004; Zbl 1066.52027), Problem 1]: Do there exist \((4,k)\)-nets in \(\mathbb{CP}^2\) that are not projectively isomorphic to the \((4,3)\)-net defined by the Hessian line arrangement? To quote the authors: ``In this note we take the first few steps towards showing that the only \(n=4\) net is the Hessian.'' To any \((4,k)\)-net in \(\mathbb{CP}^2\) there is an associated pair of orthogonal \(k\times k\) Latin squares. Conversely, every pair of orthogonal \(k\times k\) Latin squares is associated to the (possibly empty) realization space of a \((4,k)\)-net in \(\mathbb{CP}^2\). The authors define an equivalence relation on pairs of orthogonal Latin squares so that pairs in the same equivalence class correspond to the same (possibly non-existent) \(4\)-net, up to projective isomorphism. For small values of \(k\), the authors count these equivalence classes, and construct the corresponding realization spaces explicitly. Their main result states that there are no \((4,k)\)-nets in \(\mathbb{CP}^2\) for \(k=4,5,6\), and that the Hessian is the only \((4,3)\)-net in \(\mathbb{CP}^2\), up to projective isomorphism.