\(u=4\) and quadratic extensions (Q912893)

The \(u\)-invariant of a nonreal field \(L\) is the maximum dimension of an anisotropic quadratic form over \(L\). The authors' motivation is the conjecture that \(L\) and any quadratic extension of \(L\) always have the same \(u\)-invariant. They prove that \(u(L)=u(L(a^{1/2}))\) whenever \(L\) is nonreal and \(a\) is a rigid element in \(L\). Also if the cohomological 2-dimension of the group \(\mathrm{Gal}(L(2)/L)\) is 2, where \(L(2)\) is a quadratic closure of \(L\), and if \(L\) is uniformly linked, then \(u(L)=4\), and for any \(a\in R(L)\), the Kaplansky radical of \(L\), we also have \(u(L(a^{1/2}))=4\). One more difficult result assures that \(u(K)=4\) for any finite 2-extension \(K\) of \(L\) provided the above hypotheses are satisfied in addition to some structure requirements for \(\text{Gal}(L(2)/L)\). \{An amusing feature of the paper is that the only result stated as theorem is a quotation from Serre given without proof in the introduction. In their modesty the authors call their own results lemmas, remarks, examples and propositions.\}