The Segre Determinant
Elizabeth Pratt
TL;DR
The paper introduces the Segre determinant $\mathrm{Seg}_{k,\ell}$, a bi-homogeneous invariant that encodes when $k\ell$ points in $\mathbb{P}(V)\times\mathbb{P}(W)$ lie on a hyperplane section of the Segre embedding. It develops its invariant-theoretic computation via bracket algebras, establishes its role as the Chow-Lam form of torus orbits in Grassmannians, and connects to Chow quotients and algebraic-vision applications. A central result shows that the Chow-Lam form of a torus orbit in $Gr(k,n)$ equals $\mathrm{Seg}_{k,\ell}(A,B)$ in primal coordinates, tying torus actions to Segre invariants. The paper further defines the Segre coefficient variety, proves a maximal-span property for the coefficients, and demonstrates both a concrete GIT correspondence for $k=2$ and non-uniqueness phenomena for larger $k$, highlighting the Segre determinant’s constructive and geometric utility.
Abstract
The Segre determinant is a polynomial which encodes the condition for points to lie on a bilinear hypersurface in the product of projective spaces. We study Segre determinants and compute them in various coordinate systems. We show that the Segre determinant represents the Chow-Lam form of a generic torus orbit in the Grassmannian. These Chow-Lam forms were introduced as a generalization of Chow forms for projective varieties, and enjoy many similar properties. We also present applications to algebraic vision and to Chow quotients of Grassmannians.
