Hermitian analogue of a theorem of Springer (Q5949407)

Let \(A\) be a quaternion algebra over a field \(k\) of characteristic not \(2\), and let \(h\) be a Hermitian form over \(A\), relative to an involution of the first kind on \(A\). The authors show that if \(h\) is anisotropic, it remains anisotropic under scalar extensions of odd degree. The corresponding statement for quadratic forms is a classical theorem of Springer. A weaker form of Springer's theorem is known to hold for any central simple algebra \(A\): a Hermitian form which is not hyperbolic does not become hyperbolic under any scalar extension of odd degree [\textit{E. Bayer-Fluckiger} and \textit{H. W. Lenstra}, Am. J. Math. 112, 359-373 (1990; Zbl 0729.12006)]. The proof of the stronger form over quaternion algebras easily follows from two results of independent interest: (a) an excellence property for fields of conics: if \(h\) is a Hermitian form over a central simple \(k\)-algebra and \(C\) is a conic over \(k\), then the anisotropic part of \(h\otimes k(C)\) is extended from \(k\); (b) if a skew-Hermitian form over a quaternion algebra (relative to the conjugation involution) is hyperbolic over the function field of the corresponding conic, then it is hyperbolic. Result~(b) is also found (with essentially the same proof) in [\textit{I.~Dejaiffe}, ``Formes antihermitiennes devenant hyperboliques sur un corps de déploiment,'' C. R. Acad. Sci., Paris, Sér. I, Math. 332, 105-108 (2001; Zbl 1032.11014)]. By way of contrast, the authors show that the strong form of Springer's theorem does not hold for involutions of the second kind. They produce an example of a division algebra of odd degree with an involution \(\sigma\) of the second kind and a rank~\(2\) Hermitian form relative to \(\sigma\) which is anisotropic but becomes isotropic over a quadratic extension and over an odd degree extension of the base field.