On totally decomposable algebras with involution in characteristic two. (Q908349): Difference between revisions

Let \(F\) be a field and \((A,\sigma)\) a (finite-dimensional associative) central simple algebra over \(F\) with involution of the first kind. By Merkurjev's theorem (and its earlier analogue in characteristic \(2\), due to Albert), then \(A\) is Brauer equivalent to a tensor product of central simple \(F\)-algebras of degree \(2\). We say that \(A\) is decomposable, if it is \(F\)-isomorphic to such a tensor product; \(A\) is called totally decomposable, if it is presentable as a tensor product of \(\sigma\)-invariant central simple \(F\)-algebras of degree \(2\). As shown by Albert, \(A\) is decomposable, provided that the degree \(\deg(A)\) is equal to \(4\). When \(\deg(A)>4\), this is not necessarily true, by examples obtained by Amitsur-Rowen-Tignol (if \(\text{char}(F)\neq 2\)) and Rowen, in characteristic \(2\) [see \textit{S. A. Amitsur, L. H. Rowen, J. P. Tignol}, Isr. J. Math. 33, 133-148 (1979; Zbl 0422.16010) and \textit{L. H. Rowen}, J. Algebra 90, 71-83 (1984; Zbl 0548.16020)]. The former paper also shows that \(A\) is decomposable if and only if there exists a finite square-central subset of \(A\) which satisfies some commuting conditions. The study of the total decomposability condition pursues similar goals. When \(\deg(A)=4\) and \(\sigma\) is symplectic, it turns out that \((A,\sigma)\) is totally decomposable [see \textit{L. H. Rowen}, Isr. J. Math. 29, 285-301 (1978; Zbl 0392.16011) and \textit{W. Streb}, J. Algebra 83, 20-25 (1983; Zbl 0523.16011)]. A necessary and sufficient condition that \((A,\sigma)\) be totally decomposable has been found by \textit{M. A. Knus, R. Parimala} and \textit{R. Sridharan} in the case where \(\deg(A)=4\) and \(\sigma\) is orthogonal [see J. Indian Math. Soc., New Ser. 57, No. 1-4, 143-151 (1991; Zbl 0824.16017)]. Furthermore, in characteristic different from \(2\), in both cases, the classification of orthogonal and of symplectic involutions on \(A\) is known, up-to conjugation. Note also that the total decomposability problem for \((A,\sigma)\) is solved, if \(A\) is split, \(\sigma\) is orthogonal and \(\deg(A)=2^n\). The paper under review studies the problems of decomposition and classification of central simple \(F\)-algebras with involution in characteristic two. It shows that \((A,\sigma)\) is totally decomposable if and only if it possesses a symmetric and self-centralizing subalgebra \(S=\Phi(A,\sigma)\), such that: (i) \(x^2\in F\), for every \(x\in S\), and (ii) \(\dim_FS=2^{r(S)}\), where \(r(S)=r_F(S)\) is the minimum rank of \(S\). The authors prove that if \((A,\sigma)\) is totally decomposable and \(\sigma \) is orthogonal, then \(\Phi(A,\sigma)\) is unique, up-to an isomorphism. In addition, they establish the existence of a natural associative bilinear form on \(\Phi (A,\sigma)\) that is related to a recently introduced bilinear Pfister form \(\mathfrak{Pf}(A,\sigma)\) by \textit{A. Dolphin} [see J. Pure Appl. Algebra 218, No. 10, 1900-1915 (2014; Zbl 1297.11020)]. The authors prove that \(\mathfrak{Pf}(A,\sigma)\) determines \((A,\sigma)\), up-to conjugation, and so give an affirmative answer to a question asked by Dolphin [loc. cit., 7.4].