Line bundles with partially vanishing cohomology (Q1949962)

The main result of this paper under review is the equivalence of three notions of \(q\)-ampleness for a line bundle \(L\) on a projective variety \(X\) (or more generally a projective scheme with mild assumptions) over a base field \(k\) of characteristic zero. The first notion, which the author called `naive' \(q\)-ampleness, is an analogy of Serre's characterization of ampleness. A line bundle \(L\) is `naively' \(q\)-ample if for every coherent sheaf \(E\) on \(X\), we have \(H^i(X,E\otimes L^{\otimes m})=0\) for every \(i>q\) and \(m\) at least equal to some \(m_0\) depending on \(E\). This notion involves vanishing of infinitely many cohomology groups. The second notion, called \(q\)-T-ampleness, only involves the vanishing of an explicit finite list of cohomology groups. The third notion, called uniform \(q\)-ampleness, was introduced by \textit{J.-P. Demailly, T. Peternell} and \textit{M. Schneider} [Isr. Math. Conf. Proc. 9, 165--198 (1996; Zbl 0859.14005)]. Demailly-Peternell-Schneider showed that uniform \(q\)-ampleness of \(L\) only depend on the numerical class of \(L\). There are several nice consequences of the equivalence. First of all, \(q\)-ampleness is Zariski open on families of varieties and line bundles in characteristic zero, which is not clear from the first and third definition. Secondly, \(q\)-ampleness defines an (not necessarily convex) open cone in the Néron-Severi vector space \(N^1(X)\) using the third definition. The author also pointed out that the equivalence of different notions of \(q\)-amplesness fails in positive characteristic.