On the \(u\)-invariant of function fields of curves over complete discretely valued fields (Q2346054)

Let \(K\) be a complete discretely valued field with residue field \(\kappa\), \(F\) a function field of one variable over \(K\) and \(\roman{char\;} \kappa \neq 2\). The \(u=u(K)\) invariant of \(K\) is the supremum of dimensions of anisotropic quadratic forms over \(K\). In [Invent. Math. 197, No. 1, 215--235 (2014; Zbl 1356.11018)] the authors gave an alternative proof that the \(u\)-invariant of function fields of all \(p\)-adic curves (including dyadic curves) is \(8\). In fact, more generally, they proved that if \(\roman{char\;}K=0\), \(\roman{char\;}\kappa=2\) and \(k\) is perfect, then \(u(F)\leq 8\), fact originally proved by \textit{D. B. Leep} [J. Reine Angew. Math. 679, 65--73 (2013; Zbl 1276.11050)]. If \([\kappa:\kappa^2]\) is infinite, there exist quadratic forms over \(F\) of arbitrarily large dimension. It remained an open question whether the \(u\)-invariant of \(F\) is finite if \(\roman{char\;}\kappa =2\) and \([\kappa:\kappa^2]\) is finite. The main objective of this paper is to give an affirmative answer to this question. In fact the authors prove that if \(\roman{char\;} K=0\), \(\roman{char\;} \kappa=2\) and \([\kappa:\kappa^2]\) is finite, then there exists an integer \(M\) depending only on \([\kappa:\kappa^2]\) such that for any finite extension \(F\) of \(K(t)\) we have \(u(F)\leq M\). The proof of the main result is based in showing the uniform \(p\)-boundedness of the Brauer group of any function field \(F\) over a complete discretely valued field with \(\roman{char\;}\kappa =p\), \(\roman{char\;} F\neq p\) and \([\kappa:\kappa^p]\) finite. To prove these results the authors use the patching techniques of Harbater-Hartmann-Krashen [\textit{D. Harbater} et al., Invent. Math. 178, No. 2, 231--263 (2009; Zbl 1259.12003)] and their previous article [loc. cit.]. In Section 1, the authors prove a lemma that is used along the paper. If \(l\) is a prime not equal to the characteristic, then \(K\) is said to be \((n,l)\)-uniformly bounded if there exists an integer \(N\) such that for any finite extension \(L\) of \(K\) and \(\alpha_1,\ldots,\alpha_m\in H^n(L,\mu_l^{\otimes n})\), where \(\mu_l\) is the Galois module of \(l\)-th roots of unity, there exists an extension \(E\) of \(L\) with \([E:L]\leq N\) and \(\alpha_i\otimes_L E=0\) for \(1\leq i\leq m\). Such \(N\) is called an \((n,l)\)-uniform bound for \(K\). The lemma states that if \(N\) is an \((n,l)\)-uniform bound, then \(N\) is also a \((d,l)\)-uniform bound for \(K\) for all \(d\geq n\). In Section 2 it is considered the ``bad characteristic'' case. The authors show that if \(p=\roman{char\;} \kappa\), then there is a \((2,p)\)-uniform bound for \(K(t)\) which depends only on \([\kappa:\kappa^p]\). In fact they prove that if \(\roman{char\;} K=0\) and \(\roman{char\;} \kappa=p>0\), \([\kappa:\kappa^p]=p^d\), and \(\zeta\) is a primitive \(p\)-th root of unity, then \([K(\zeta):K] p^{4d+4}\) is an \((n,p)\)-uniform bound for \(K(t)\) for all \(n\geq 2\). In Section 3, the authors consider the ``good characteristic'' case. It is proved that if a prime \(l\) not equal to \(\roman{char\;} \kappa\), \(N_1\) is a \((2, l)\)-uniform bound for \(\kappa(t)\) and \(N_2\) is a \((2,l)\) uniform bound for \(\kappa\), then \(l^3 (N_1!)(N_2!)\) is an \((n,l)\)-uniform bound for \(K(t)\) for all \(n\geq 2\). In Section 4 the main result is proved.