Polarized mixed Hodge structures: On irrationality of threefolds via degeneration (Q1066971)

The author considers smooth complex projective varieties X of dimension 3 with \(\pi_ 1(X)=0\), \(h^ i({\mathcal O}_ X)=0\) for \(i=1,2,3\) and \(H^ 3(X,{\mathbb{Z}})\) torsion free. A one-parameter degeneration \({\mathcal X}\to \Delta\) of such varieties gives rise to a degenerating family of abelian varieties with principal polarization over \(\Delta\), obtained by considering the intermediate jacobians of the fibers. By comparison with the analogous situation for curves one obtains necessary conditions for the generic fiber to be rational. The case is considered where the special fiber \(X_ 0\) is a union \(D_ 0\cup D_ 1\cup D_ 2\) of smooth divisors with normal crossings such that \(D_ 1(D_ i\cap D_ j)=0\) for \(i\neq j\). The special fiber of the family is then an extension of \(\oplus J^ 3(D_ i)\) by a torus \(\tau\). The character group of \(\tau\) is then equipped with a skew-symmetric integral bilinear form \(\psi\). In the analogous curve case, \(\psi\) is known. By showing in explicit examples that \(\psi\) is different from the curve case, one shows that the generic fiber \(X_ t\) is not rational. This works for cubic and quartic threefolds, complete intersections of type (2,3) in \({\mathbb{P}}^ 5\) or (2,2,2) in \({\mathbb{P}}^ 6\), the quartic double solid and hypersurfaces of bidegree (p,3), \(p\geq 2\), in \({\mathbb{P}}^ 1\times {\mathbb{P}}^ 3.\) The proof of proposition 1.9.2 of the paper contains a gap, which can be filled, as the author told me.