Real components of algebraic varieties and étale cohomology (Q802681)

The main theorem of this paper says that, of X/\({\mathbb{R}}\) is a smooth irreducible variety of dimension d, then, for any \(n>d\), the subgroup of \(H^ n_{et}({\mathbb{R}}(X),{\mathbb{Z}}/2)\) of classes unramified on X is isomorphic, via the signature map, to the group of locally constant \({\mathbb{Z}}/2\)-valued functions on X(\({\mathbb{R}})\) (i.e. \(({\mathbb{Z}}/2)^ s\) where s is the number of connected components of X(\({\mathbb{R}}))\). The definition of the signature map uses results of \textit{S. Bloch} and \textit{A. Ogus} [Ann. Sci. Éc. Norm. Supér., IV. Sér. 7(1974), 181-202 (1975; Zbl 0307.14008)]. Given \(x\in X({\mathbb{R}})\), the canonical surjection \({\mathcal O}_{X,x}\to {\mathbb{R}}\) yields for any unramified class \(\alpha\) an image in \(H^ n_{et}({\mathbb{R}},{\mathbb{Z}}/2)={\mathbb{Z}}/2\); this is the signature of \(\alpha\) at x. Actually, this may be formulated in the framework of real spectrum which is more convenient. The quadratic forms play a central role in the proof. The link between quadratic forms and étale cohomology appears by sending the Pfister form \(<1,-a_ 2>\otimes...\otimes <1,-\alpha_ n>\) on \(F={\mathbb{R}}(X)\) to the cup product \(\overline{a_ 1}\cup...\cup \overline{a_ n}\) of the classes of the \(a_ i's\) in \(F^*/(F^*)^ 2=H^ 1_{et}(F,{\mathbb{Z}}/2)\). The main ingredients are results about quadratic forms over fields contained in the survey by \textit{J. K. Arason}, \textit{R. Elman} and \textit{B. Jacob} [in Quadratic and Hermitian forms, Conf., Hamilton/Ont. 1983, CMS Conf. Proc. 4, 17-50 (1984; Zbl 0554.10011)] and Mahé's result on the separation of connected components by quadratic spaces [\textit{L. Mahé}, Math. Ann. 260, 191-210 (1982; Zbl 0507.14019)]. In the case where X(\({\mathbb{R}})\) is compact, the Stone-Weierstrass theorem and the injectivity of the signature imply that any unramified class (for \(n>d)\) can be written as \(\bar f\cup \overline{-1}\cup...\cup \overline{- 1}\) with f invertible on X(\({\mathbb{R}})\). This has consequences for surfaces: separation of connected components by the 2-torsion subgroup of the Brauer group when X is affine, lower bounds for the rank of étale cohomology in terms of the number of connected components such as \(2s- 1\leq \dim_{{\mathbb{Z}}/2}H^ 3_{et}(X,{\mathbb{Z}}/2),...\) This very interesting paper has suscited further investigations: another proof, without quadratic forms (\textit{C. Scheiderer}); a generalisation to singular varieties, obtained from a result of \textit{D. A. Cox} [Proc. Am. Math. Soc. 76, 17-22 (1979; Zbl 0442.14004)] about étale homotopy type (\textit{Colliot-Thélène}, in discussions with \textit{Mahé}); new comparison results, in the spirit of those coming from Cox, in a preprint ``Real and étale cohomology'' by \textit{C. Scheiderer}.