Pseudo-effective classes and pushforwards (Q489947)

Let \(X\) be a complex projective variety, inside \(H_{2k}(X,\mathbb R)\) classes of irreducible subvarieties of \(X\) generate a closed convex cone, the pseudo-effective cone, whose elements are called { pseudo-effective classes}, and a real vector subspace \(H_{2k}(X,\mathbb R)_{\mathrm{alg}} \). Consider a morphism \(\phi:X\to Y\) and look at the family of those irreducible subvarieties \(Z\) of \(X\) which are {contracted by \(\phi\)}, i.e. \(\dim (\phi(Z))< \dim (Z)\). If \(k:=\dim(Z)\), this amounts to the requirement that the class \([Z]\in H_{2k}(X,\mathbb R)\) is in the kernel of \( \phi_*:H_{2k}(X,\mathbb R )\rightarrow H_{2k}(Y,\mathbb R ).\) The authors explain that their first main result is the proof of the statement that a pseudo-effective degree-2 homology class annihilated by \(\phi_*\) belongs to the vector space generated by classes of contracted curves. If \(X\) is smooth, it belongs to the closed convex cone generated by these classes and the same result holds for divisor classes. This provides a positive partial answer to the paper's \textit{Conjecture {A.}} Conjecture: Any pseudo-effective class \( \alpha\in H_{2k}(X,\mathbb R )\) such that \(\phi_*\alpha=0\) in \(H_{2k}(Y,\mathbb R )\) belongs to the real vector space (resp.\ the closed convex cone) spanned by classes of \(k\)-dimensional subvarieties of \(X\) which are contracted by \(\phi\). The theory of currents is one main tool used in the proof, the following variant is then proposed: \textit{ {Conjecture {C}}}. Let \(X\) and \(Y\) be smooth complex projective varieties and let \(\phi:X\rightarrow Y\) be a morphism. Then any weakly pseudo-effective class \( \alpha\in H^{2l}(X,\mathbb R )\) such that \(\phi_*\alpha=0\) belongs to the real vector space spanned by classes of \(l\)-codimensional subvarieties of \(X\) which are contracted by \(\phi\). Here an algebraic cycle is called \textit{weakly pseudo-effective} if its cohomology class is the class of a positive current, and consequently a real cohomology class is said to be weakly pseudo-effective if it is a limit of weakly pseudo-effective cycle classes. The second main result in the paper is the proof of the fact that {Conjecture C} implies a stronger form of {GHC}, the generalized Hodge conjecture of Grothendieck, for the case of smooth projective varieties \(X\) such that \(H^{i,0}(X)=0\) for \(i>0\). From {GHC} it follows that {{{when \(X\) is as above then there exist a proper closed algebraic subset \(Y\subset X\) such that the restriction map \(H^i(X,\mathbb Q)\rightarrow H^i(X\setminus Y,\mathbb Q)\) is zero for any \(i>0\).}}} The authors observe that if \(n=\dim (X)\), by the Künneth decomposition, the conclusion is equivalent to the fact that the cycle class \([\Delta_X- x\times X]\in H^{2n}(X\times X,\mathbb Q) \) vanishes in \(H^{2n}(X\times (X\setminus Y),\mathbb Q)\) or, equivalently, comes from a class in \(H_{2n}(X\times Y,\mathbb Q)\). The stronger version of GHC says that there exist a proper closed algebraic subset \(Y\subset X\) and an \(n\)-\textit{ {algebraic cycle}} \(W\in \tenit Z_n (X\times X)_{\mathbb Q}\) supported on \(X\times Y\) such that the class of \(W\) in \(H^{2n}(X\times X,\mathbb Q)\) is equal to \([\Delta_X- x\times X]\).