On the 3 and 4-normality of codimension two subvarieties of \({\mathbb{P}}^ n\) (Q912176)

The Hartshorne conjecture about codimension two subvarieties \(X\subset {\mathbb{P}}^ n({\mathbb{C}})\) (with \(n\geq 6)\) is equivalent to the statement that X is k-normal for each \(k\geq 1\), i.e., the natural maps \(H^ 0({\mathbb{P}}^ n,{\mathcal O}(k))\to H^ 0(X,{\mathcal O}(k))\) are surjective for each \(k\geq 1\). \textit{F. L. Zak} has proved [Math. USSR, Sb. 44, 535-544 (1983); translation from Mat. Sb. 116(158), 593-602 (1981; Zbl 0484.14016)] that X is 1-normal for \(n\geq 5\), while \textit{L. Ein} [in Algebraic Geometry, Proc. Conf.,, Sundance/Utah 1986, Lect. Notes Math. 1311, 71-75 (1988; Zbl 0673.14025)] and \textit{T. Peternell}, \textit{J. Le Potier}, \textit{M. Schneider} [Invent. Math. 87, 573-586 (1987; Zbl 0618.14023)] have proved independently that X is 2-normal for \(n\geq 10.\) In this paper the authors prove that X is 3-normal for \(n\geq 16\) and 4- normal for \(n\geq 22\). The main tools used are the theorem of \textit{Mather} on generic projections and the Barth-Ogus theorem on formal completions. The same techniques give also some vanishing for \(H^ q(X,{\mathcal O}(3))\) and \(H^ q(X,{\mathcal O}(4))\). - The authors have improved these bounds by showing [J. Reine Angew. Math. 409, 35-40 (1990)] that X is k-normal for \(n\geq 6k-2, k\geq 2\) (it is possible to obtain even \(n\geq 4k+4)\).