Totally positive extensions and weakly isotropic forms (Q2494328): Difference between revisions

The author introduces the invariant \[ \widehat u(F)=\sup\{\dim \varphi\mid\varphi \text{ anisotropic, weakly isotropic quadratic form over } F\} \] and compares it to the \(u\)-invariant \(u(F)\) and the Hasse number \(\widetilde u(F)\). Clearly \(u\leq\widehat u\leq\widetilde u\), and the three invariants coincide if \(F\) is nonreal. If \(F\) is real the new invariant is related to totally positive extensions of \(F\) and to the ``pythagorean index'' \(\text{pind}(A):=\text{ind}(A\otimes \widetilde F)\) of central simple algebras \(A\) over \(F\) \((\widetilde F\) is the pythagorean closure of \(F)\). The two main results read: Theorem 4.3. Let \(\varphi\) be weakly isotropic over \(F\) and \(K=F(\varphi)\). Then for any central simple \(F\)-algebra \(A\), one has \[ \text{pind}(A\otimes K)=\text{pind}(A). \] Theorem 5.4. For any \(m\), \(n\in\mathbb{N}\) with \(2\leq m\leq n\), there is a real field \(K\) such that \(I^3K\) is torsion-free, \(u(K)=2m\), \(\widehat u(K)=2n\), and \(\widetilde u(K)= \infty\). -- The proof uses, of course, Merkurjev's construction. Some open questions and conjectures of this paper may also be of interest.