On splitting of totally singular quadratic forms (Q2569860)

Let \(F\) be a field of characteristic 2. Every quadratic form \(\phi\) over \(F\) is an orthogonal sum \(\phi=\phi^{\prime}\perp ql(\phi)\), where \(\phi^{\prime}\) is nonsingular and \(ql(\phi)\) is totally singular. The form \(\phi\) is called totally singular if \(\phi=ql(\phi).\) In one of his earlier papers [Math. Z. 240, No. 4, 711--730 (2002; Zbl 1007.11017)] the author examined the standard splitting tower \((F_i,\phi_i)_{0\leq i\leq h=h(\phi)}\) of a nonzero quadratic form \(\phi\). He showed that there exists \(\epsilon\in\{0,...,h\}\) such that \(\phi_\epsilon \cong ql(\phi)_{F_\epsilon}.\) M. Knebusch proved (book in preparation) that up to \(F_\epsilon\) the standard splitting of \(\phi\) behaves as in characteristic different from 2 and in a different way for \(F_i\) if \(i\geq \epsilon\) as shown by \textit{D. Hoffmann} and the author in [Trans. Amer. Math. Soc. 356, No. 10, 4019--4053 (2004; Zbl 1116.11020)]. Thus in the paper under review the author concentrates on the splitting of totally singular forms. For such forms he introduces the analogue of a Pfister form, called in the paper quasi-Pfister forms, and examines their properties. He completely describes the standard splitting of quasi-Pfister neighbors proving that an anisotropic totally singular form \(\phi\) of dimension in \(]2^k,2^{k+1}]\) is a quasi-Pfister neighbor iff \(h(\phi)=k+1,\) which means the same that the splitting pattern of \(\phi\) equals \((\dim \phi-2^k,2^{k-1},...,2,1).\) The paper ends with a theorem which proves that for a totally singular form \(\phi\) one can make a choice on some quadratic form that appears in the standard splitting tower of \(\phi\). More precisely, if \(\phi=\psi\perp\psi^{\prime}\) with \(\dim \psi=2^m<\dim \phi\) for some \(m\geq 1\) and \((F_i,\phi_i)\) is the splitting tower of \(\phi\), then \(\phi_{\epsilon}\cong \psi_{F_{\epsilon}}\) for some \(0\leq \epsilon\leq h(\phi).\)