Deformations of Calabi-Yau hypersurfaces arising from deformations of toric varieties (Q1882585)

The infinitesimal deformations of Calabi-Yau manifolds are unobstructed so it is easy to calculate how many local deformations of a (nef and big) Calabi-Yau hypersurface \(X\) in a toric variety \({\mathbb P}_\Sigma\) there ought to be. Some of these deformations come from perturbing the polynomial that defines \(X\), but not all. In a few cases the ``missing'' or ``non-polynomial'' deformations had been constructed, but always as polynomial deformations in a different embedding. That approach is often not even applicable, and does not usually produce enough deformations when it is. The present paper solves the problem completely. The key is to deform the ambient variety \({\mathbb P}_\Sigma\). Not very much is known about infinitesimal deformations of toric varieties in general. Mavlyutov here constructs the deformations he needs and this construction is very interesting in its own right. A previous theorem of the author [Duke Math. J. 101, 85--116 (2000; Zbl 1023.14027)] supplies a toric birational morphism \(\pi: {\mathbb P}_\Sigma \to {\mathbb P}_{\Sigma_X}\) such that \(\pi_*[X]\) is ample and \(\pi^*\pi_*[X]=[X]\). The fan \(\Sigma\) is a subdivision of \(\Sigma_X\). The extra rays in this subdivision determine an open cover of \({\mathbb P}_\Sigma\), and also (via some cocyles written down by the author in [Compos. Math. 138, 289--336 (2003; Zbl 1117.14052)]) a direct sum decomposition of \(H^1(\Theta_X)\). One summand is identified with the polynomial deformations. The others lift to \(H^1(\Theta_{{\mathbb P}_\Sigma})\), and they do so in a way independent of \(X\), which suggests that the non-polynomial deformations come from deformations of \({\mathbb P}_\Sigma\). The next step is to construct global deformations (i.e. over \({\mathbb A}^1\)) of \({\mathbb P}_\Sigma\) from the open cover by changing the gluing maps. This is done in such a way as to induce global deformations of \(X\) also. Because everything has been explicitly described it is possible to compute the Kodaira-Spencer maps of the families over \({\mathbb A}^1\) and thus to compare the global deformations with the predicted infinitesimal deformations. In particular it follows that all the non-polynomial deformations of \(X\) do indeed arise in this way; and we see also which classes in \(H^1(\Theta_{{\mathbb P}_\Sigma})\) are realised.