Analogues of the Brauer group for algebras with involution (Q1196385): Difference between revisions

Let \(X\) be a scheme with 2 a unit. The authors define a Brauer group \(\text{Br}^*(X)\) for Azumaya algebras over \(X\) with involution of the first kind and a group \(\text{Br}(X,\delta)\) for Azumaya algebras over \(X\) with involution of the second kind, where \(\pi:Y\to X\) is an étale cover of degree 2 and \(\delta\) is the nontrivial automorphism of \(Y/X\). There are natural injections \[ \text{Br}^*(X)\to H^ 0_{et}(X,\mu_ 2)\oplus H^ 2_{et}(X,\mu_ 2) \] and \(\text{Br}(X,\delta)\to H^ 2_{et}(X,{\mathcal G})\), where \({\mathcal G}\) is the étale sheaf on \(X\) obtained from invertible functions on \(Y\) of norm 1. The first map is an isomorphism precisely if the 2-torsion in the cohomological Brauer group \(H^ 2_{et}(X,\mathbb{G}_ m)\) is represented by classes of Azumaya algebras and the second is an isomorphism precisely if every class in \(\ker(N_{Y/X}:H^ 2_{et}(Y,\mathbb{G}_ m)\to H^ 2_{et}(X,\mathbb{G}_ m))\) is represented by an Azumaya algebra on \(Y\). To any quadratic space \(({\mathcal E},q)\) on \(X\), one associates its Clifford invariant, i.e., the class of its Clifford algebra \(C(q)\) in \(_ 2\text{Br}(X)\). The algebra \(C(q)\) is equipped with two involutions and the corresponding classes in \(\text{Br}^*(X)\) have the same component in \(H^ 2_{et}(X,\mu_ 2)\), which maps to the Clifford invariant of \(q\) under the natural surjection \(H^ 2_{et}(X,\mu_ 2)\to H^ 2_{et}(X,\mathbb{G}_ m)\) obtained from the Kummer sequence. Thus the component in \(H^ 2_{et}(X,\mu_ 2)\) may be considered as a refinement of the Clifford invariant. Let \(I_ 2(q)\) be the subgroup of the Witt group of \(X\) consisting of classes of forms of even rank and trivial discriminant. The Clifford invariant yields a homomorphism \(e_ 2:I_ 2(q)\to{_ 2\text{Br}}(X)\). However the refined Clifford invariant does not yield a homomorphism \(\tilde e_ 2:I_ 2(q)\to{_ 2\text{Br}}(X)\) which lifts \(e_ 2\). In another direction the authors give a positive answer to a question raised by the reviewer and the authors [J. Algebra 130, No. 1, 65-82 (1990; Zbl 0695.16003)] in connection with the complex \[ _ 2\text{Br}(R)\to{_ 2\text{Br}}(S)\to_ 2\text{Br}(R), \] where \(S\) is an étale quadratic \(R\)-algebra. The same question is also discussed by \textit{K.-H. Ulbrich} ``On the Teichmüller cocycle and a theorem of Eilenberg-Maclane'' (to appear)].