Splitting probability invariant for Cordes schemes (Q2120903): Difference between revisions

A Cordes scheme is a triple \(S=(G,-1,D_G)\) where \(G\) is an elementary 2-group, \(-1\) is a distinguished element in \(G\), and \(D_G\) is a map from \(G\) to the set of subgroups of \(G\) satisfying \(a \in D_G(a)\) for any \(a \in G\) and \(b\in D_G(a) \Leftrightarrow -a \in D_G(-b)\) for any \(a,b \in G\) (\(-a\) standing for \((-1)\cdot a\)). This term was originally introduced in Zbl 0335.10024, and the motivation comes of course from quadratic forms over fields \(F\) of \(\operatorname{char}(F) \neq 2\), where \(G=F^\times/(F^\times)^2\) and \(D_G(a)\) is the value group of \(\langle 1,a \rangle\). Assuming \(|G|<\infty\), we define the splitting probability of \(S\) to be \(p_s(S)=\frac{1}{|G|^2} \sum_{a\in G} |D_G(-a)|\). Much of the paper is devoted to extending known results from [\textit{L. Jindal} and \textit{A. Khurana}, J. Algebra Appl. 20, No. 6, Article ID 2150098, 9 p. (2021; Zbl 1472.11113)], originally proved for fields, to Cordes schemes. The authors prove several other nice but technical results on this invariant. Reviewer's note: the quadratic form \(\langle 1,a \rangle\) is what is nowadays referred to as the 1-fold Pfister form \(\langle \! \langle -a \rangle \! \rangle\). Replacing \(a\) with \(-a\) would modify the property of \(D_G\) from \(b\in D_G(a) \text{ if and only if } -a \in D_G(-b)\) to \(b\in D_G(a) \text{ if and only if } a \in D_G(b)\), which perhaps would be more elegant.