Generalised quadratic forms and the \(u\)-invariant (Q1680280): Difference between revisions
From MaRDI portal
Latest revision as of 17:11, 14 July 2024
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Generalised quadratic forms and the \(u\)-invariant |
scientific article |
Statements
Generalised quadratic forms and the \(u\)-invariant (English)
0 references
15 November 2017
0 references
Let \(F\) be a field, let \(D\) be a central division \(F\)-algebra and let \(\sigma:D\to D\) be an involution. When \(\mathrm{char}\,F\neq 2\) or \(\sigma\) is unitary, the \(u\)-invariant of \(\varepsilon\)-hermitian forms over \((D,\sigma)\) is defined as the supremum of integers \(n\) such that there exists an \(n\)-dimensional anisotropic \(\varepsilon\)-hermitian form over \((D,\sigma)\). This clearly generalizes the classical \(u\)-invariant of a field. The paper under discussion considers the case where \(\mathrm{char}\,F=2\) and \(\sigma\) is of the first kind, i.e., \(F\)-linear. In this case, the fact that hermitian forms and (generalized) quadratic forms are inequivalent and the presence of singular quadratic forms promote the consideration of several variants of the \(u\)-invariant. The author introduces five such variations, defined in terms of hermitian forms (\(u^+(D)\)), direct hermitian forms (\(u^+_d(D)\)), alternating hermitian forms (\(u^-(D)\)), quadratic forms (\(u(D)\)) and nonsingular quadratic forms (\(\tilde{u}(D)\)). These turn out to be independent of the involution \(\sigma\), and one always has \(u^+(D)=u_d^+(D)+u^-(D)\). Numerous relations and bounds involving these invariants are obtained. For example, it is shown that \(u^+_d(D)\leq [F:F^2]/\deg D\), and equality holds if \(D\) factors as a tensor product of quaternion algebras. In addition, when \(D\) is quaternion, \(u^+(D)\leq [F:F^2]\) and \(u(D)\leq \frac{3}{2}[F:F^2]\). The invariants are further determined when \([F:F^2]=2\) or when \(F\) is a Kaplansky field, i.e., admits a unique quaternion algebra \(Q\) up to isomorphism. In the latter case, these are given by \(u^+(Q)=\frac{1}{2}[F:F^2]+1\), \(u^+_d(Q)=\frac{1}{2}[F:F^2]\), \(u^-(Q)=1\), \(u(Q)=\sup\{3,\frac{1}{2}[F:F^2]+1\}\) and \(\tilde{u}(Q)=3\). This is used to construct quaternion algebras \(Q\) over Kaplansky fields in which \(u^+(Q)\) and \(u(Q)\) are arbitrarily large. In contrast, over Kaplansky fields of characteristic \textit{not} \(2\), the \(1\)-Hermitian \(u\)-invariant of a quaternion algebra with an orthogonal involution cannot exceed \(3\); see [\textit{K. J. Becher} and \textit{M. G. Mahmoudi}, Bull. Belg. Math. Soc. - Simon Stevin 17, No. 1, 181--192 (2010; Zbl 1206.11042)].
0 references
central simple algebras
0 references
involutions
0 references
generalised quadratic forms
0 references
Hermitian forms
0 references
\(u\)-invariant
0 references
characteristic two
0 references
quaternion algebras
0 references