Hyperbolicity and near hyperbolicity of quadratic forms over function fields of quadrics (Q1627783)

Let \(F\) be a field of characteristic not 2, let \(p\) and \(q\) be anisotropic quadratic forms over \(F\) of dimension \(\ge 2\) and let \(F(p)\) denote the function field of the projective quadric \(p=0\). Let \(k\) be the dimension of the anisotropic part of the extension of \(q\) to \(F(p)\). The author proposes a conjecture: Let \(s\) be the unique non-negative integer such that \(2^s < \dim (p)\le 2^{s+1}\). Then \(\dim (q) =a\cdot 2^{s+1}+\varepsilon\) for some non-negative integer \(a\) and integer \(\varepsilon\) with \(|\varepsilon |\le k\). When \(\dim (q)\le 2^s\) then the Conjecture gives back Hoffmann's Separation Theorem, namely that \(k=\dim (q)\) and so \(q\) remains anisotropic over \(F(p)\). The conjecture is trivial if \(k > 2^s-2\). The main result of the paper is a proof of the conjecture for \(k < 2^{s-1}\). The case \(k=0\) yields a remarkable corollary: With the same notation, if \(q\) becomes hyperbolic over \(F(p)\) then \(\dim (q)\) is divisible by \(2^{s+1}\). The proof first reduces the conjecture to the problem of checking the parity of certain binomial coefficients. These coefficients can be found from the action of the Steenrod algebra on the total mod-2 Chow group of the quadric defined by \(q\). Then a result of \textit{A. Vishik} [Manuscr. Math. 122, No. 3, 365--374 (2007; Zbl 1154.14003)] (extended to odd characteristic by Fino) completes the proof. The author also proves the conjecture if \(k\le 7\) or \(\dim (q) \le 2^{s+2}+2^{s-1}\). In both these cases, and elsewhere, the author proves stronger statements at the level of the splitting pattern of \(q\), that is, the extensions of \(q\) to its Knebusch splitting tower.