A geometric characterisation of Desarguesian spreads (Q2408112)

A spread \(\mathcal S\) of a projective space \(\Pi\) is a non-trivial collection of subspaces of \(\Pi\) such that any point of \(\Pi\) is contained in precisely one element of \(\mathcal S\). An element \(E \in \mathcal S\) is called normal if for any other two elements \(F, G \in \mathcal S\) either \(G \subset \langle E, F\rangle\) or \(G \cap \langle E, F\rangle = \emptyset\). A spread is called normal (or geometric) if all of its elements are normal. By a theorem of \textit{R. Baer} [Arch. Math. 14, 73--83 (1963; Zbl 0113.25403)], every normal spread is either planar, i.e. there holds \(\Pi = \langle E, F\rangle\) for any two distinct elements \(E,F \in \mathcal S\), or else \(\Pi\) naturally carries the structure of a projective space over a skewfield \(L\) such that the elements of \(\mathcal S\) are precisely the \(L\)-points and the sets \(\langle E, F\rangle, E,F \in {\mathcal S}\), \(E \not= F\) are the \(L\)-lines. In the latter case, the spread is called Desarguesian. The authors prove the beautiful theorem that a spread is normal if it contains \(r+1 \geq 3\) normal elements such that any \(r\) of them generates \(\Pi\). The proof rests on the fact that a normal element \(E \in \mathcal S\) and any other element \(F \in \mathcal S\) generate a planar subspread, which turns out to define a Desarguesian translation plane. The authors formulate their result only for finite projective spaces, but their arguments are entirely geometric, and hence this restriction is not really necessary.