Unirational threefolds with no universal codimension \(2\) cycle (Q2516380)

This paper defines a new stable birational invariant and gives a very effective method in proving that a variety is not stably rational using this invariant. This method has already been applied by various authors to prove lots of new important results about non-stable rationality in a short time after the appearance of the paper. Given a smooth complex projective variety \(X\) of dimension \(n\), the author defines that \(X\) has an integral (cohomological) decomposition of diagonal if the following equality is true in the integral Chow(cohomology) group: \[ \Delta_X =x \times X+Z, \] where \(Z\) is a cycle of dimension \(n\) supported in \(X \times D\) for some divisor \(D\) in \(X\). It is easy to show that having an integral decomposition of diagonal is a stable birational invariant. There are many consequences of having such a decomposition. For example, the non-ramified cohomology groups with torsion coefficients of \(X\) (in particular, the Brauer group) are trivial. The author discovers that one can show the non-existence of an integral decomposition of diagonal using degenerations: given a family of varieties over a curve with central fiber \(X\) having ordinary double points as the only singularities, if \(X\) does not have an integral decomposition of diagonal, then a very general fiber does not have such a decomposition. It may happen that a smooth variety may degenerate to a mildly singular variety with only ordinary double points which does not have an integral decomposition of diagonal since it has a non-trivial Brauer group. This is the case with a double solid, a double cover of \(\mathbb{P}^3\) branched along a degree \(4\) hypersurface (the author even allow the double solid to have at most \(7\) ordinary double points). For a rationally connected \(3\)-fold, having an integral decomposition of diagonal is closedly related to the existence of a universal codimension \(2\) cycle in \(X \times J(X)\) which induces the identity on \(J(X)\) for the Abel-Jacobi map and the existence of a surjective with rationally connected general fiber Abel-Jacobi map from a family of algebraic cycles. Here \(J(X)\) is the intermediate Jacobian. Using the degeneration argument, the author also shows the non-existence of such Abel-Jacobi maps.