On Reider's method and higher order embeddings (Q914770)

In this paper various higher order generalizations of very ampleness are considered. The primary generalization, k-spannedness, is defined as follows. A line bundle L on a smooth surfaces S is k-spanned for a non- negative integer k if given any length \((k+1)\), 0-dimensional subscheme (Z,\({\mathcal O}_ Z)\) of S which is curvilinear (i.e. which lies on a germ of a smooth curve), we have a surjection, \(\Gamma\) (L)\(\to \Gamma (L\otimes {\mathcal O}_ Z)\to 0\). A line bundle, L, is spanned at all points by global sections (respectively very ample) if and only if L is 0-spanned (respectively 1-very ample). A Reider type criterion is given for the adjoint bundle \(K_ S\otimes L\) to be k-spanned. This definition has the virtue that it is weaker than other reasonable definitions, and thus results proven using this definition will hold true for stronger k- th order notions of very ampleness. This has turned out to be important since a strengthened form of a conjecture stated at the end of the paper has been proved independently by different methods by the reviewers [``Zero cycles and k-th order embeddings of smooth projective surfaces'', in ``Projective Surfaces and their Classification'', 1988 Cortona Proc.: Symp. Math. INDAM] and also by Reider. This result, which subsumes the Reider type criterion of the paper under review, is stated as follows. Theorem. Let L be a nef line bundle on a smooth surface S. Let \(L\cdot L\geq 4k+1\). Then given any 0-dimensional subscheme (Z,\({\mathcal O}_ Z)\) of S of length \(\leq k\), either \(\Gamma (K_ S\otimes L)\to \Gamma (K_ S\otimes L\otimes {\mathcal O}_ Z)\to 0\) or there exists an effective divisor D such that L-2D is \({\mathbb{Q}}\)-effective, D contains some 0- dimensional subscheme \((Z',{\mathcal O}_{Z'})\) of (Z,\({\mathcal O}_ Z)\) of length \(k'\leq k\), and \(L\cdot D-k'\leq D\cdot D<L\cdot D/2<k'.\) Because of this result, it is better to work with k-very ampleness, a more natural generalization of very ampleness, where the curvilinear \(length\)-(k\(+1)\) 0-dimensional subschemes in the definition of k- spannedness are replaced by arbitrary \(length\)-(k\(+1)\) 0-dimensional subschemes of S. The results in the paper under review on the k- spannedness of powers of the canonical bundle, hold with k-very ample used in place of k-spanned. The only change in the proofs is to use the above criterion instead of the criterion in the paper. Some further papers on k-spannedness are (in chronological order) \textit{M. Beltrametti} and \textit{A. Sommese} [``On k-spannedness for projective surfaces'', Algebraic geometry, Proc. Int. Conf., L'Aquila/Italy 1988, Lect. Notes Math. 417, 24-51 (1990) and ``On the relative adjunction mapping'', Math. Scand. 65, No.2, 189-205 (1989)]; \textit{E. Ballico} [Proc. Am. Math. Soc. 105, No.3, 531-534 (1989; Zbl 0671.14019) and ``On k-spanned projective surfaces'', in Algebraic geometry, Proc. Int. Conf., L'Aquila/Italy 1998, Lect. Notes Math. 1417, 23 (1990)]; \textit{E. Ballico} and \textit{M. Beltrametti} [Manuscr. Math. 61, No.4, 447-458 (1988; Zbl 0661.14005)]; \textit{F. Catanese} and \textit{L. Göttsche} [``d-very-ample line bundles and embeddings of Hilbert schemes of 0-cycles'', Manuscr. Math. 68, No.3, 337-341 (1990)]; \textit{M. Andreatta} and \textit{M. Palleschi}, ``On the 2-very ampleness of the adjoint bundle'' (preprint).