Galois module structure of unramified covers (Q2480873)

The author considers the fundamental problem of Galois module theory (is \(O_N\) free over \({\mathbb Z}[G]\), when \(N/K\) is a tame \(G\)-extension?) in a higher-dimensional setting, that is, he considers a \(G\)-covering \(X \to Y\) with \(Y\) a projective algebraic variety over Spec(\({\mathbb Z}\)). Here the counterpart of \(O_N\) is a certain complex arising from the right derived functor of the global section functor, evaluated on \({\mathcal O}_X\). Another even more fundamental difference is that the extension is assumed to be unramified everywhere (not only tame). The main result states that for \(G\) of prime order \(p\) and under the assumptions \(p>\dim(Y)\) and that Vandiver's conjecture is true for \(p\), the fundamental problem has an affirmative answer in the geometric situation just described, that is, the stable projective Euler characteristic of the complex R\(\Gamma(X,{\mathcal O}_X)\) is zero. This main theorem is underpinned by results which give more detailed information (also for other Galois groups) but are more complicated to state. (A finicky comment: many authors use the double factorials occurring in Theorem 1.4 with a different meaning, but the author clearly explains his use of terminology.) It is very interesting and satisfactory to have a link between geometry and class groups of prime cyclotomic fields (Vandiver) as presented in this paper. The main steps in setting up the connection are: \(n\)-cubic structures, and multiextensions of multiplicative groups. The latter have to do with certain eigenspaces of the \(p\)-part of the class group of \({\mathbb Q}^+(\zeta_p)\), and this permits (as mentioned) fairly detailed results. (If we assume Vandiver's conjecture, all these eigenspaces are zero.) We do not try to explain the general theory of hypercubic structures here (referring to earlier papers of G. Pappas). Just to give an idea what happens: if \(B/A\) is a \(G\)-Galois extension of commutative rings, then the class of the projective rank one \({\mathbb Z}[G]\)-module is primitive, which means the existence of an isomorphism \(i_1^*A \otimes i_2*A \cong \Delta^*A\). Here the tensor product is taken over \(G\times G\), and \(i_1,i_2,\Delta: G\to G\times G\) send \(g\) to \((g,1)\), \((1,g)\), and \((g,g)\) respectively. The notion of a hypercubic structure can be seen as an axiomatisation of a generalisation of this phenomenon, with the number 2 (two copies of \(G\)) being replaced by some \(n\geq 3\). The existence of a hypercubic structure on a projective rank 1 module (line bundle, in geometric language) is a strong structural restriction on the module. It is a crucial point in the paper that determinants of cohomology along a suitably nice morphism can be endowed with a hypercubic structure. The paper concludes with a construction that shows the existence of unramified extensions with any group \(G\) and any dimension strictly larger than 1.