Springer's odd degree extension theorem for quadratic forms over semilocal rings (Q2243175)

A famous theorem by \textit{T. A. Springer} [C. R. Acad. Sci., Paris 234, 1517--1519 (1952; Zbl 0046.24303)] states that any anisotropic quadratic form over a field \(F\) of characteristic not \(2\) remains anisotropic over any odd degree extension of \(F\) (in fact, Springer's proof easily carries over to characteristic \(2\)). Before that, E.~Artin communicated a proof of this result to E.~Witt who posed the original question, but Artin's proof had never been published. There have been various efforts to prove versions of this result for quadratic forms over more general rings, for example, semilocal rings. In that case, results were obtained under various additional assumptions on the semilocal rings. In the present paper, the authors dispense with these assumptions on the semilocal ring itself. Let \(R\) be a semilocal ring. A quadratic space \((M,q)\) over \(R\) is a finitely generated projective \(R\)-module with a nonsingular quadratic form \(q\) on \(M\), where nonsingular means that \(q\) has trivial radical. A finite \(R\)-algebra is a commutative \(R\)-algebra that is finitely generated as \(R\)-module (but not necessarily projective). The main theorem of the paper states that if \((M,q)\) is an anisotropic quadratic space over \(R\), then it will stay anisotropic over any finite étale or one-generated \(R\)-algebra \(S\) of constant odd rank. There are various immediate corollaries. For example, for \(R\) and \(S\) as above, the Witt index of a quadratic space \((M,q)\) over \(R\) will stay unchanged when passing to \(S\). Also, if two quadratic spaces over \(R\) will be isometric over \(S\), then they are already isometric over \(R\). Furthermore, the natural restriction map from the Witt-Grothendieck ring (resp. Witt ring) of \(R\) to that of \(S\) will be injective. While most parts of the proofs use fairly elementary but clever methods, at some crucial points the authors invoke various deep results such as Deligne's trace homomorphism for the flat cohomology of group schemes, Demazure's Conjugacy Theorem for reductive \(R\)-group schemes, and a recent result by \textit{E. Bayer-Fluckiger} et al. [``On the Grothendieck-Serre conjecture for classical groups'', Preprint, \url{arXiv:1911.07666}] concerning embeddings of finite étale \(R\)-algebras into one-generated \(R\)-algebras.