Biquaternion algebras and quartic extensions (Q1314734)

Biquaternion algebras, i.e. tensor products of two quaternion algebras, are not necessarily cyclic. In this work, the authors give two criteria for a biquaternion algebra to be cyclic. The first one is expressed in terms of the splitting properties of the Albert form, which is a 6- dimensional quadratic form associated with any biquaternion algebra, hence is quadratic-form-theoretic. The second criterion is in terms of the corestriction of central simple algebras, hence is purely algebra- theoretic. The criteria, which are simple enough to permit explicit computations, are presented and proved separately, independently of each other. A substantial part of the paper is devoted to the study of quartic 2- extensions, i.e. field extensions \(L/F\) of degree 4 which contain an intermediate quadratic extension. Using their computation of the Witt kernel \(W(L/F)\) the authors give necessary and sufficient conditions for an \(n\)-Pfister form to have a quartic 2-extension with prescribed discriminant as a splitting field. Considering then differences of two Pfister forms (the Albert form is such a difference) they get similar conditions for biquaternion algebras. In the algebraic approach the Witt kernel is replaced by the relative Brauer group \(B(L/F)\). The cyclicity criteria mentioned above then are special cases of the splitting results. In a useful appendix, the authors give proofs of two results of Galois cohomology which are used in the paper. One of them is the fact that \(3\)- Pfister forms are classified by their Arason invariant. The results are known to the experts, however no reasonably self-contained proofs (in particular which do not rely on Merkurjev's main theorem) were yet in the literature.