Ostrom-derivates (Q2493441)

A flock of a quadratic cone \(C\) in the finite 3-dimensional projective space PG\((3,q)\) over the field with \(q\) elements is a partition of \(C\) into its vertex and \(q\) plane conics. Embedding \(C\) in the Klein quadric and taking the conjugate planes, one obtains \(q^2+1\) points that correspond to a spread \(S\) of PG\((3,q)\). On its turn, \(S\) gives rise to a translation plane \(\pi\). The spread \(S\) now consists of \(q\) (base) reguli all containing one fixed line (which corresponds to the vertex of \(C\)). The paper under review points out that there are \(q\) more translation planes and spreads by replacement of these \(q\) reguli (Ostrom-derivates). The question is now: to how many isomorphism classes of translation planes does this give rise? Equivalently: what is the number of orbits on the base reguli of the full collineation group of the original plane \(\pi\)? In the paper under review, it is shown that there are always at least two non-isomorphic translation planes obtained by reguli-replacement, except in a few well-known cases (Desarguesian, Walker or Betten, Knuth). This is achieved as follows. Suppose that there is only one isomorphism class. Consider the skeleton of the spread \(S\) (these are the \(q+1\) spreads corresponding to the flocks obtained from the original one by (BLT-)derivation). The collineation group of \(S\) permutes the other members of the skeleton. Since all spreads are isomorphic, this leads to a doubly transitive group acting on the skeleton. Invoking the classification of the finite doubly transitive groups, the authors determine all possibilities (in fact, rule out the ``non-possibilities'').