Clifford algebras of hyperbolic involutions (Q5929727)

Let \(F\) be a field of characterictic not equal to \(2\), \(A\) a central simple \(F\)-algebra of degree \(\sqrt{\dim_FA}=n\) with an involution \(\sigma \), \(C_0(A)\) the even Clifford algebra of \(A\) and \(\overline\sigma\) the involution of \(C_0(A)\) canonically induced by \(\sigma\). When \(4\mid n\), \(\sigma\) is hyperbolic and \(A\) is not similar to \(F\), it is known that \((C_0(A),\overline\sigma)\) is isomorphic to the orthogonal sum \((S,\tau)\times(N,\pi)\), where \(S\) and \(N\) are central simple \(F\)-algebras similar to \(F\) and \(A\), respectively, and \(\tau\) is an isotropic involution of \(S\) [see \textit{J. Tits}, Invent. Math. 5, 19-41 (1968; Zbl 0155.05202) and \textit{H. P. Allen}, J. Algebra 10, 503-515 (1968; Zbl 0302.15031)]. In addition, it is known that \(\tau\) and \(\pi\) are simultaneously symplectic or orthogonal depending on whether \(n\equiv 4\pmod 8\) or \(8\mid n\) [see \textit{M.-A. Knus, A. S. Merkurjev, M. Rost} and \textit{J.-P. Tignol}, The book of involutions, Colloquium Publications 44, AMS, Providence, RI (1998; Zbl 0955.16001)]. The main theorem of the paper under review strengthens and supplements these results. The additional information is presented in the language of hyperbolic extensions (see Definition 3.1), a notion extending the one of Witt-equivalence for quadratic forms to simple algebras with involution. In particular, the paper shows that if \(8\mid n\), i.e. \(A\cong M_2(B)\), for some central simple \(F\)-algebra \(B\) of degree \(n/2\), then \(\pi\) is hyperbolic and \((S,\tau)\) is a hyperbolic extension of the exterior power \(F\)-algebra \(\lambda^{n/4}B:=B_{n/4}\) with its canonical involution \(\gamma\). Since \(B_{n/4}\) is a central simple \(F\)-algebra of degree \((n/2)!/((n/4)!^2)\) and is similar to \(F\), this enables one to associate with \(B_{n/4}\) a quadratic form \(q_B\) in \((n/2)!/((n/4)!^2)\) variables (with coefficients in \(F\)). This invariant of \(B\) is uniquely determined up to equivalence; the author shows that its anisotropic part can be a quadratic form in many variables, whence the involution \(\tau\) can be very far from being hyperbolic (in contrast to the case of \(n\equiv 4\pmod 8\)).