Stabilizers of collineation groups of smooth stable planes (Q1279720)

The author continues his investigation of smooth stable planes and of their collineation groups begun in [Result. Math. 31, 300-321 (1997; Zbl 0882.51007) and Forum Math. 10, 751-773 (1998; Zbl 0916.51014)]. A stable plane is a linear space in which the point set and the line set are endowed with topologies such that the unique line joining two different points depends continuously on these points and that the intersection point of two different intersecting lines depends continuously on these lines; moreover the set of all pairs of different intersecting lines is required to be open (so that the property of intersecting is stable). A particular class among these planes are the topological projective planes. A stable plane is called smooth if the point space and the line space are smooth manifolds and the operations of joining and intersection are even smooth. The line pencils then are homeomorphic to a sphere of a certain dimension \(l\); by topological reasons related to the theorem of Adams on multiplications on spheres, \(l\) can only take on the values \(1, 2, 4\) or \(8\). The point space and the line space then have dimension \(2l\). The classical examples are the projective planes \({\mathcal P}_2 {\mathbb{F}}\) for \({\mathbb{F}} \in \{{\mathbb{R}}, {\mathbb{C}}, {\mathbb{H}}, {\mathbb{O}}\}\) over the real numbers, the complex numbers, the quaternions and the octonions. Until recently, such planes have mainly been investigated without assuming smoothness, under corresponding topological assumptions. The profound reason why the theory of smooth stable planes is in some respect more difficult than that of just topological stable planes is related to the results of the present paper and deserves some explanation. The group \(\Gamma\) of continuous collineations of such a stable plane is a locally compact group and in many instances even a Lie group; in the smooth case, the latter is always true. One of the guiding ideas of the theory of stable planes is that the dimension of this group should be a measure for how close the plane in question is to the classical examples. It is a vast program initiated originally by H. Salzmann to make this precise in various ways. For compact connected topological projective planes, over the last fourty years such programs have been carried far. A key result is that for compact connected projective planes of dimension \(2l\), \(l \in \{1,2,4,8\}\) there is a critical dimension \(c_l\) depending on the dimension \(l\) of a line pencil such that non-classical compact connected planes of dimension \(2l\) with \(\dim \Gamma = c_l\) exist but \(\dim \Gamma > c_l\) forces the plane to be isomorphic to the classical plane \({\mathcal P}_2 {\mathbb{F}}\) of the same dimension. The author's results in the present paper and in other papers on this subject accord with the leitmotiv that such a critical dimension for the smooth case will be lower than in the topological case, in other words, the non-classical planes with the largest possible groups do not admit the structure of a smooth plane. Now these planes are at the same time the planes which are not too complicated to construct, the collineation group helping to understand them; thus non-classical smooth planes tend to be less accessible. In fact, the known examples mostly are somewhat weird and difficult to handle. In the present paper one of the main results in this sense concerns the stabilizer \(\Lambda\) of three concurrent lines in the collineations group \(\Gamma\) of a stable plane of dimension \(n = 2l\). Let \(d_{\text{class}}(n)\) denote the dimension of this stabilizer in the classical plane \({\mathcal P}_2 {\mathbb{F}}\) of dimension \(2l\). The author shows that if \(\dim \Lambda > d_{\text{class}}(n) - l\), then the plane is an almost projective translation plane (that is, a translation plane with a closed subset of the translation line removed). If the plane is a projective plane, then the author even shows that it is isomorphic to \({\mathcal P}_2 {\mathbb{F}}\). On the way to this result, the author proves a number of powerful theorems about the structure of the stabilizer of a flag or a point in \(\Gamma\). An important tool is the tangent translation plane \({\mathcal A}_p\) of a point \(p\) studied in the previous papers mentioned above. The point set of this translation plane is the tangent space of the point manifold in \(p\); the lines through the origin in this plane are the tangent spaces of the lines through \(p\) of the original stable plane, which are submanifolds. \({\mathcal A}_p\) turns out to be a topological plane. Now for topological translation planes of this kind (with point space \({\mathbb{R}}^{2l}\)), the reviewer has developed structure theorems for the stabilizers of two concurrent lines in the group of continuous collineations. The author can use these results by considering the derivation map \(\text{D}_p : \Gamma_p \to \text{Aut}({\mathcal A}_p)\). He shows using representation theory that a Levi subgroup \(\Psi\) of a connected Lie subgroup \(\Delta\) of \(\Gamma_p\) fixing a line through \(p\) is compact. Incidentally, his argument yields that the same is true for affine topological translation planes with point set \({\mathbb{R}}^{2l}\) (without any smoothness assumptions); this is an important contribution to the theory of such planes. Furthermore, the author can obtain information about semi-simple subgroups of point stabilizers. An important consequence is that the connected component of the stabilizer \(\Gamma_p\) of a point \(p\) is a linear group. This is used to show that the Moulton planes (the only non-classical \(2-\)dimensional compact projective planes with \(\dim \Gamma \geq 4\)) cannot be made smooth planes, thus confirming the leitmotiv explained above in the case of \(2\)-dimensional planes. In the meantime, further papers of the author on collineation groups of smooth stable planes have appeared [Beitr. Algebra Geom. 39, 121-133 (1998; Zbl 0893.51014); Geom. Dedicata 72, No. 3, 283-298 (1998; Zbl 0926.51019)]. The following hints on misprints and omissions might be helpful for the reader. In 3.5 and 3.7, the group \(\Delta\) is meant to be a connected Lie group contained in the stabilizer \(\Gamma_p\) of a point. In the table of the dimensions \(d_{\text{class}}(n)\) preceding Theorem 4.1, the entry for \(d_{\text{class}}(8)\) should read \(15\) instead of \(18\). For the dimension estimate at the bottom of p. 486 resulting in \(\dim \Delta \leq d_{\text{class}}(n) - l\) one needs the estimate \(\dim \Theta_3 \leq d_{\text{class}}(n) - 2l\) where \(\Theta_3\) is the stabilizer of a degenerate quadrangle in the projective closure of a tangent translation plane. The latter estimate is valid even for arbitrary compact connected projective planes and is well known by now, although in higher dimensions it was difficult to obtain. It may be found e.g. in the comprehensive monograph by \textit{H. Salzmann, D. Betten, T. Grundhöfer, R. Löwen, M. Stroppel} and the reviewer [Compact projective planes, De Gruyter Expositions in Mathematics 21, Berlin (1996; Zbl 0851.51003)], see 71.7(b), 83.17 and 83.23 there for planes of dimension at least \(4\). The only topological translation plane with point space \({\mathbb{R}}^2\) is the classical plane over \({\mathbb{R}}\), for which the estimate in question can be verified immediately. The result cited for \(16\)-dimensional planes concerns the stabilizer of a non-degenerate quadrangle, from which the dimension estimate for the stabilizer of a degenerate quadrangle can be obtained easily. At the end of the proof of Theorem 4.8, the reference [3], Lemma 4.5 is given for proving the fact that \(\alpha\) is a homology; a consequence of the quoted lemma, viz. [3], Corollary 4.8, expresses this fact more directly. The end of the proof of Theorem 4.10 contains some information about representations of the groups \(\text{Spin}_7 {\mathbb{R}}\) and \(\text{SU}_4 {\mathbb{C}}\) which is somewhat distorted; what the author really means and needs is the fact that these groups do not have nontrivial real representations \(\rho\) of dimension less than \(8\) such that \(\rho(-\text{id}) = -\text{id}\), analogously as in the proof of Theorem 4.8.