Lower \({\mathcal H}\)-cohomology of higher-dimensional quadrics (Q1901988)

Let \(F\) be a field of characteristic \(\neq 2\) and \(X\) an anisotropic quadric over \(F\). The kernel of the natural map \(H^n (F, \mathbb{Z}/2)\to H^n (F(X), \mathbb{Z}/2)\), where \(F(X)\) is the function field of \(X\), has been studied for \(n\leq 3\) by \textit{J. Kr. Arason} [J. Algebra 36, 448-491 (1975; Zbl 0314.12104)]\ and for \(n=4\) by \textit{B. Jacob} and \textit{M. Rost} [Invent. Math. 96, 551-570 (1989; Zbl 0685.10015)]\ and \textit{M. Szyjewski} [Leningr. Math. J. 2, No. 1, 179-198 (1991); translation from Algebra Anal. 2, No. 1, 213-234 (1990; Zbl 0719.11023)]. In this paper, we consider the cokernel of the map \[ H^n (F, \mathbb{Z}/2) @> \mu^{0,n}>> H^0 (X, {\mathcal H}^n (\mathbb{Z}/2)), \] where \(H^0 (X, {\mathcal H}^n (\mathbb{Z}/ 2)) \subseteq H^n (F (X), \mathbb{Z}/2)\) is the unramified cohomology of \(F(X)/F\). The main results are: Theorem 1. \(\mu^{0,2}\) is surjective, unless \(X\) is defined by a neighbour of an anisotropic 2-fold Pfister form. Theorem 2. If \(\dim X\geq 3\), then \(\mu^{0,3}\) is surjective, unless \(X\) is defined by a neighbour of an anisotropic 3-fold Pfister form or by an anisotropic Albert form (= a quadratic form in 6 variables of discriminant \(-1\)). If \(X\) is defined by a neighbour of the anisotropic 3-fold Pfister form \(\langle \langle a, b, c\rangle \rangle\), then \(\text{Coker } \mu^{0,3} \simeq \mathbb{Z}/2\) or 0 according as \((-1, a, b, c)\) is or is not 0 in \(H^4 F\). If \(X\) is defined by an anisotropic Albert form, then \(\text{Coker } \mu^{0,3} \simeq \mathbb{Z}/2\). As a by-product, we get information on the kernel of the cycle class map \(cl^2_X: CH^2 X/2\to H^4 (X_{\text{ét}}, \mathbb{Z}/2)\): Theorem 3. For a quadric \(X\), the cycle class map \(cl^2_X\) is injective, except in the following cases: (i) \(X\) is defined by a neighbour of an anisotropic 3-fold Pfister form \(\langle \langle a, b, c \rangle \rangle\) and \((-1, a, b, c) =0\in H^4 F\); (ii) \(X\) is defined by an anisotropic Albert form. In these two cases, \(\text{Ker }cl^2_X \simeq \mathbb{Z}/2\). The method of proof is to combine the Hochschild-Serre spectral sequence (for the covering \(X \otimes_F F_s\to X\), where \(F_s\) is a separable closure of \(F\)) and the Bloch-Ogus spectral sequence with computations of \(CH^1\) and \(CH^2\) of quadrics due to Szyjewski and Karpenko. Theorems 1, 2, and 3 generalize and amplify earlier results of \textit{J.-L. Colliot-Thélène} and \textit{R. Sujatha} on unramified \(H^3\) of real anisotropic quadrics [Unramified Witt groups of real anisotropic quadrics [\textit{W. Jacob} and \textit{A. Rosenberg} (ed.), \(K\)-theory and algebraic geometry: connections with quadratic forms and division algebras, Proc. Symp. Pure Math. 58, Part 2, 127-147 (1995; Zbl 0827.11022)]\ and of \textit{R. Sujatha} on anisotropic Pfister quadrics [Unramified cohomology and Witt group of anisotropic Pfister quadrics, preprint (1995)]. In particular, the part of theorem 2 concerning Pfister neighbours is due to Sujatha (ibid.). The proof we give in this paper is different from hers. The main application of theorems 1 and 2 is the following corollary, which relies on results of Arason, Rost and Merkurjev-Suslin: Corollary. a) The map \[ W(F)/ I^3 F\to W_{nr} (F (X))/ I^3_{nr} F(X) \] is surjective, unless \(X\) is defined by a neighbour of an anisotropic 2-fold Pfister form. b) If \(\dim X\geq 3\), then the map \[ W(F)/ I^4 F\to W_{nr} (F (X))/ I^4_{nr} F(X) \] is surjective, unless \(X\) is defined by a neighbour of an anisotropic 3-fold Pfister form or by an anisotropic Albert form. Here \(W_{nr} (F(X))\) denotes the unramified Witt ring of the extension \(F(X) /F\) and \(I^n_{nr} F(X)= W_{nr} (F (X)) \cap I^n F(X)\). In the case of an anisotropic Albert form the cokernel of \(W(F)/ I^4 F\to W_{nr} (F (X))/ I^4_{nr} F(X)\) is isomorphic to \(\mathbb{Z} /2\) and we exhibit an explicit generator. To our knowledge, this is the first example of a genuine unramified Witt class over the function field of a quadric to appear in the literature.