Dimensions of anisotropic indefinite quadratic forms. I (Q5949809)

The main concerns of this paper are the following invariants of a field \(F\): The \(u\)-invariant \(u(F)\), the Hasse number \(\widetilde u(F)\) and the ``symbol length'' \(\lambda(F)\). If \(F\) is nonreal and \(\lambda(F)\neq 0\) then B. Kahn proved that \(\widetilde u(F)=u(F)\geq 2\lambda(F)+2\) with equality if \(I^3 F=0\). The author intends to prove similar results for a real field \(F\). As always this case is much more difficult and technically involved. He succeeds under the additional assumptions: \(F\) is a SAP-field, \(I^3F\) is torsion-free and (w.l.o.g.) \(\lambda(F)\geq 2\). The realizable triples \((\lambda,u,\widetilde u)\) are then of the form \((n,2n,2n+2)\), \((n,2n+2,2n+2)\), \((n,2n,\infty)\), \((n,2 n+2,\infty)\) or \((\infty,\infty, \infty)\). For the construction of fields corresponding to a given triple of this shape he uses a variant of Merkurjev's method. Near the end of \S 2 he mentions the interesting (more general) Conjecture 2.13: Let \(F\) be real with \(I_t^3F=0\). Then \(u(F)\leq 2\lambda(F)+2\).