Coniveau 2 complete intersections and effective cones (Q2269901)

The Hodge conjecture is one of the major problems in algebraic geometry. Let \(X\) be a smooth complex projective variety and \(L\subset H^k(X, {\mathbb{Q}})\) a Hodge substructure of coniveau \(c\). Griffiths conjectured that there exists a closed algebraic subset \(Z\subset X\) of codimension \(c\) such that \(L\) vanishes under the restriction map \(H^k(X, {\mathbb{Q}})\rightarrow H^k(U, {\mathbb{Q}})\), where \(U=X-Z\). If \(X\) is a complete intersection in \({\mathbb{P}}^n\) of \(r\) hypersurface of degree \(d_1\leq d_2\leq \dots \leq d_r\), then by using Griffiths residues and the comparison of pole order and Hodge filtration, the author proves: \(X\) has coniveau \(\geq c\) if and only if \(n\geq \sum_i d_i + (c-1)d_r\). We say that \(X\) has coniveau \(c\) if the Hodge structure on \(H^{n-r}(X, {\mathbb{Q}})_{prim}\) has coniveau \(c\). In this paper the author also proposes a strategy to prove the generalized Hodge conjecture for coniveau 2 complete intersections, which does not involve in the study of Chow groups and the construction of \(2-\)cycles. The method is based on the following result: when \(n\geq \sum_i d_i+d_r\), the subvariety \(F_G\subset F\) is very moving. In this result, \(X\subset {\mathbb{P}}^n\) is a generic complete intersection of multidegree \(d_1\leq d_2\leq \dots \leq d_r\). For a generic section \(G\in H^0(X, {\mathcal{O}}_X(n-\sum_i d_i -1))\) with zero set \(X_G\subset X\), \(F\) is the variety of lines contained in \(X_G\) and \(F_G\subset F\) is a subvariety. The definition of ``very moving'' is: A smooth \(k\)-dimensional subvariety \(V\) of a smooth projective variety \(Y\) is very moving if through a generic point \(y\in Y\), and given a generic vector subspace \(W\subset T_{Y, y}\) of rank \(k\), there is a deformation \(V'\subset Y\) of \(V\) in \(Y\) which is smooth and passes through \(y\) with tangent space equal to \(W\) at \(y\). An algebraic cohomology class is big if it belongs to the interior of the effective cone. The author conjectures that if \(V\) is a smooth and projective variety and \(W\subset V\) a very moving subvariety, then the class \([W]\) of \(W\) is big. Finally in the paper the author proves that if \(n\geq \sum_i d_i +d_r\) and the subvariety \(F_G\subset F\) has a big class, i.e., it satisfies this conjecture, then the complete intersections of multidegree \(d_1\leq \dots \leq d_r\) in \({\mathbb{P}}^n\) satisfy the generalized Hodge conjecture for coniveau 2. Thus the author gives a very different proof of the known result of \textit{A. Otwinowska} [C. R. Acad. Sci., Paris, Sér. I, Math. 329, No. 1, 51--56 (1999; Zbl 0981.14004)] concerning the generalized Hodge conjecture for coniveau 2.