The nonexistence of projective planes of order 12 with a collineation group of order 16 (Q598436): Difference between revisions

Let \((P,\mathcal G)\) be a projective plane of order~12. The author shows that a collineation group of \(P\) cannot have order~16. Combined with older results of various authors this proves that the order of such a collineation group divides 8 or 9. Now assume that \(H\) is a collineation group of order~16 of \(P\). Using results by \textit{Z. Janko} and \textit{T. Trung} [J.~Comb. Theory, Ser. A 32, 401--404 (1982; Zbl 0492.51014)], the author finds that \(H\) can only be cyclic or generalized quaternion. In both cases there exists a unique elements \(\varphi\) of order~2, and \(\varphi\) is an elation. Let \(P_{0}\) be the center and \(\ell_{0}\) the axis of \(\varphi\). Now let \(V\) and \(\mathcal B\) be the set of point, resp.~line orbits of \(\langle\varphi\rangle\) which are two element (i.e., from all points or lines which are not fixed by \(\varphi\)). For \(Q\in V\) and \(B\in{\mathcal B}\) define an incidence relation \(QIB\) if any point in \(Q\) has an incident line in \(B\). Then \((V,{\mathcal B},I)\) is a symmetric transversal design with \(k=12\) and \(\lambda=2\). The point groups are given by the lines through \(P_{0}\) and each has \(g=6\) elements. Dually for the line groups. Note that this construction works for any group of elations with common axis and center. Now \(H/\langle\varphi\rangle\) acts on \((V,{\mathcal B},I)\) as automorphism. Indeed, the author shows that this action is semiregular on both \(V\) and \({\mathcal B}\). Some four dozen cases for the special nature of this action are listed, and the effect on the incidence matrix of \((V,{\mathcal B},I)\) is studied. Then a computer is used to exclude all the cases. Unfortunately, no indication of the algorithms, nor a source for the code is given.