Invariant Brauer group of an abelian variety (Q2164436): Difference between revisions

The authors study the invariant Brauer group \(\operatorname{Br}_G(X)\) introduced by \textit{Y. Cao} [Compos. Math. 154, No. 4, 773--819 (2018; Zbl 1429.11115)] for the action of an algebraic \(k\)-group \(G\) on a \(k\)-variety \(X\), in the important special case where \(G=X=A\) is an abelian variety. In particular they address the question, also raised by Cao, of how \(\operatorname{Br}_A(A)\) compares with the algebraic Brauer group \(\operatorname{Br}_1(A)\). \par It turns out that \(\operatorname{Br}_A(A)\) is somewhat subtle even over an algebraically closed field. (As the authors point out, the algebraic Brauer group, which the kernel of the map \(\operatorname{Br}(X)\to \operatorname{Br}(X_s)\) to the Brauer group over the separable closure, is the more accessible part of \(\operatorname{Br}(X)\), so perhaps non-trivialities over a separably closed field are not so unexpected.) In particular, it is not related to \(\operatorname{Br}_1(A)\) in a very simple way: neither need be contained in the other. \par One consequence of the results of this paper, which also illustrates this, is that \(\operatorname{Br}_A(A)\) can be seen as something like an Arf invariant, in that it is the obstruction to an element \(\alpha\in \Hom(A,A')/2\) such that \(e(x,\alpha x)=0\) for every \(2\)-torsion point \(x\) to come from a symmetric morphism \(A\to A^\vee\) (here \(e\colon A[2]\times A^\vee[2]\to \mu_2\) is the Weil pairing). \par The more general results here are of two kinds. First, the authors obtain an estimate on the size of the \(p\)-free part of \(\operatorname{Br}_A(A)\), where \(k\) is an algebraically closed field of characteristic \(p\neq 2\) (possibly \(p=0\)). This, they show, is a finite elementary \(2\)-group, of order \(2^n\) with \(n\le {\operatorname{rank}}{\operatorname{End}}(A)-{\operatorname{rank}}{\operatorname{NS}} (A)\). They give some cases in which it is trivial: it is preserved by odd-degree isogenies. \par The other results concern the case of complex tori (not necessarily abelian varieties at all). Indeed, in every dimension \(g\ge 3\) there exist both algebraic and non-algebraic complex tori with invariant Brauer group of order~\(2\), and this can also happen for abelian surfaces of CM type, whether simple or not.