The non-Lefschetz locus of conics

TL;DR

This paper studies the non-Lefschetz locus of conics in Artinian algebras, proving expected codimension results for general complete intersections and exploring connections to jumping conics in vector bundles.

Contribution

It establishes the codimension of the non-Lefschetz locus of conics for general complete intersections and links this to jumping conics in vector bundle cohomology.

Findings

Non-Lefschetz locus of conics has expected codimension in general complete intersections.

Complete intersections have the Strong Lefschetz Property at range 2.

Examples show non-Lefschetz locus varies in codimension for monomial cases.

Abstract

A graded Artinian algebra $A$ has the Weak Lefschetz Property if there exists a linear form $\ell$ such that the multiplication map by $\ell:[A]_i\to [A]_{i+1}$ has maximum rank in every degree. The linear forms satisfying this property form a Zariski-open set; its complement is called the non-Lefschetz locus of $A$. In this paper, we investigate analogous questions for degree-two forms rather than lines. We prove that any complete intersection $A=k[x_1,x_2,x_3]/(f_1,f_2,f_3)$, with $\text{char } k=0$, has the Strong Lefschetz Property at range $2$, i.e. there exists a linear form $\ell\in [R]_1$, such that the multiplication map $\times \ell^2: [M]_i\to [M]_{i+2}$ has maximum rank in each degree. Then we focus on the forms of degree 2 such that $ \times C: [A]_i\to [A]_{i+2}$ fails to have maximum rank in some degree $i$. The main result shows that the non-Lefschetz locus of conics for a general complete intersection $A=k[x_1,x_2,x_3]/(f_1,f_2,f_3)$ has the expected codimension as a subscheme of $\mathbb{P}^5$. The hypothesis of generality is necessary. We include examples of monomial complete intersections in which the non-Lefschetz locus of conics has different codimension. To extend a similar result to the first cohomology modules of rank $2$ vector bundles over $\mathbb{P}^2$, we explore the connection between non-Lefschetz conics and jumping conics. The non-Lefschetz locus of conics is a subset of the jumping conics. Unlike the case of the lines, this can be proper when $\mathcal{E}$ is semistable with first Chern class even.