Large automorphism groups of 8-dimensional projective planes are Lie groups (Q1337123)

A projective plane is called topological, if its point and line space are endowed with topologies that render the geometric operations (joining and intersecting) continuous. The point space \(P\) of a topological projective plane is either connected or totally disconnected. If \(P\) is connected and compact, then by a deep result of R. Löwen the topological (covering) dimension \(d = \dim P\) may be 2, 4, 8, 16, or infinite. The finite values are attained by the classical (Moufang) planes over the fields of real and complex numbers, the skewfield of Hamilton's quaternions, and the real alternative algebra of Cayley's octonions. Large classes of non-classical planes are known. The group \(\Sigma\) of all automorphisms (continuous collineations) of a compact projective plane is a locally compact group of dimension at most \(5d - 2\). H. Salzmann has shown that for \(d \in \{2,4\}\), the point space \(P\) is a topological manifold, and \(\Sigma\) is a Lie group. This is also true for all sufficiently well known examples with \(d > 4\). At present, there is no hope that the question whether \(\Sigma\) is a Lie group can be settled in general. For the applications in the theory of compact projective planes, however, it has been proved that (the connected component of) \(\Sigma\) is a Lie group if \(\dim \Sigma\) is greater than some specified bound depending on \(d\). For the case \(d = 8\), the present paper improves these results considerably, it is shown: If \(d = 8\) and \(\dim \Sigma \geq 12\), then the connected component of \(\Sigma\) is a Lie group. Meanwhile, the author has applied similar techniques in her dissertation [`Kompakte 16-dimensionale Ebenen mit großen halbeinfachen Gruppen', Tübingen (1994)] and obtained: If \(d = 16\) and \(\dim \Sigma \geq 28\), then the connected component of \(\Sigma\) is a Lie group. In general, a locally compact group need not be a Lie group although its connected component is; in fact, the factor group modulo the connected component may be nondiscrete. The methods that are used in the present paper do not allow an extension of the result to the whole group \(\Sigma\). This is due to the fact that the question whether a connected locally compact group of finite dimension is a Lie group is decided in some totally disconnected subgroup of the center, while the extension of the connected component need not be central.