Weak del Pezzo surfaces with irregularity (Q2472598)

Recall that a weak Fano variety \(X\) is one for which \(\omega^*_X\), the dual of the dualising sheaf, is nef and big. It follows at once from general (Kawamata-Viehweg) vanishing that \(H^i(X, {\mathcal O}_X)=0\) for \(i\geq 0\) if \(X\) is a smooth weak Fano variety over the complex numbers. However, since vanishing fails in general in positive characteristic, it is natural to ask whether this case of it still holds. For small dimension (\(2\) or \(3\)) this is known to be the case. On the other hand, there are non-normal del Pezzo surfaces with \(H^1(X,{\mathcal O}_X)\neq 0\), due to \textit{M. Reid} [Publ. Res. Inst. Math. Sci. 30, No. 5, 695--727 (1994; Zbl 0867.14015)]. The author's idea is to try to construct regular (not smooth) Fano varieties with nonvanishing \(H^1(X,{\mathcal O}_X)\) over nonperfect fields. To this end he studies twisted forms of Reid's surfaces. This does not quite yield the result sought for. The author finds himself obliged to weaken, very slightly, either the Fano condition or the regularity condition. The main theorem is that over every nonperfect field of characteristic~\(2\), there are weak del Pezzo surfaces that are regular, and there are normal del Pezzo surfaces whose formal singularities are factorial rational double points of type \(A_1\), in both cases with \(H^1({\mathcal O})\neq 0\). In the paper, the early sections collect general facts on twisted forms and recall Reid's construction, showing the the resulting surfaces are locally complete intersections and analysing their Picard groups and dualising sheaves. Then there is a discussion of curves of degree one, some but not all of which are Cartier divisors. Then it is shown that the cotangent sheaf mod torsion on such a surface \(Y\) is locally free of rank two (``a rather special situation'', says the author) and hence that the tangent sheaf is locally free. From this it is possible to construct a vector field \(\delta\in H^0(Y, \Theta_{Y/k})\), which gives enough information to construct the twisted forms. However, \(\delta\) has a zero: it is shown that this is inevitable and that the choice of \(\delta\) is best possible, and the resulting singularity is analysed. At the end of the paper the results are used to construct some Fano-Mori contractions. There is an investigation of the ampleness properties of line bundles on \(Y\), from which it emerges that \(Y\) can be neither a hypersurface nor a double cover of projective space, and is thus rather complicated.