On quasi-complete intersections (Q2644721)

Let \(C\subset {\mathbb{P}}_ k^ n=:X\) (k an infinite field) be a closed subscheme of pure codimension \(m<n\) with ideal sheaf \({\mathcal I}_ C\) (on X). C is called a quasi-complete intersection if there are \(m+1\) hypersurfaces \(F_ 0,...,F_ m\) in X such that \(C=F_ 0\cap...\cap F_ m\) (scheme theoretically) or, equivalently, if there is an epimorphism \({\mathcal O}_ X(-d_ 0)\oplus...\oplus {\mathcal O}_ X(-d_ m)\to {\mathcal I}_ C\) (with suitable integers \(d_ 0,...,d_ m)\). The aim of this paper is to extend numerical criteria for C to be a quasi- complete intersection to more general cases which are already known in special situations. It turns out that these results provide a complete characterization of quasi-complete intersections. Here is the main result: Assume that C is a generic complete intersection (i.e., \({\mathcal I}_{C,x}\) is a parameter ideal in \({\mathcal O}_{X,x}\) for all generic points x of C). Then C is a quasi-complete intersection iff the following conditions are fulfilled: (a) There are hypersurfaces \(F_ 0,...,F_ m\) in X of degrees \(d_ 0,...,d_ m\) containing C such that \(F_ 0\cap...\cap F_ m=C\cup D\) with \(co\dim_ XD\geq m+1;\) (b) \(H^{n-m-1}(C,{\mathcal O}_ C(p))=0\) for all \(p\ll 0;\) (c) \(d_ 0\cdot...\cdot d_ m=(d_ 0+...+d_ m-2n+m)\cdot h_ 0(C)+2\cdot h_ 1(C)\) where \(h_ 0(C)=\deg (C)\) and \(h_ 1(C)\) denote the highest coefficients of the Hilbert polynomial of C (with respect to the embedding \(C\subset X).\) This and related results are also extended to characterize so-called almost complete intersections.