Spaces of similarities. II: Pfister factors (Q1239221)

Continue the notation from the previous review. The main topic of this paper is the question: Which quadratic forms \(q\) over the field \(F\) can admit large subspaces of similarities? Pfister factor conjecture. If \(\dim q = n\) and if there is a form \(\sigma\) with \(\sigma < Sim(q)\) and \(\dim \sigma = \rho(n)\), then \(q \cong \varphi \otimes \omega\), where \(\varphi\) is a Pfister form and \(\dim \omega\) is odd. This conjecture is reduced to the case \(\dim q = 2^m\) by decomposing \(q\) into unsplittable pieces admitting \(\sigma\), and applying the following uniqueness result. Theorem. If \(\dim \sigma = \rho(2^m)\) and \(\sigma < Sim(q)\) where \(\dim q = 2^m\), then the form \(q\) is uniquely determined, up to similarity, by the form \(\sigma\). A new construction of spaces of similarities is also presented. If \(C = C(\langle-a_1,\dots,-a_m\rangle)\) is a Clifford algebra, there is a natural trace-form \(B_1\) on \(C\) making \((C, B_1) \cong (\langle-a_1,\dots,-a_m\rangle)\), a Pfister form. The left and right regular representations can then be employed to find subspaces \(\sigma<Sim(C, B_1)\) with \(\dim \sigma = \rho(2^m)\).