Maximal subgroups of the finite orthogonal and unitary groups stabilizing anisotropic subspaces (Q791657)

If \(f\) is a non-degenerate quadratic or Hermitian form on a vector space we call a subspace \(W\) anisotropic if the only zero of \(f\) in \(W\) is the zero vector. Suppose that \(G\) is one of the groups \(\mathrm{GL}_n(K)\), \(\mathrm{Sp}_n(K)\), \(\mathrm{O}_n(K)\), \(\mathrm{U}_n(K)\) acting on a vector space \(V\) of dimension \(n\geq 3\). Suppose, further, that \(V\) is the direct sum of subspaces \(V_1,V_2,\dots,V_s\) \((s\geq 2)\) all of the same dimension \(r\), and that if \(G\neq \mathrm{GL}_n(K)\) all the \(V_i\) are isomorphic, non-isotropic and pairwise conjugate (with respect to the form \(f\) associated with \(G)\). Let \(F\) be the stabilizer in \(G\) of \(\{V_1,V_2,\dots,V_s\}\). In [J. Algebra 68, 109--120 (1981; Zbl 0449.20049), ibid. 73, 350--375 (1981; Zbl 0467.20037), ibid. 76, 540--548 (1982; Zbl 0486.20029), J. Lond. Math. Soc., II. Ser. 25, 416--424 (1982; Zbl 0458.20042), Math. Z. 182, 193--203 (1983; Zbl 0488.20036)] \textit{O. King} has settled the question when \(F\) is a maximal subgroup of \(G\) except for two cases: (i) when \(G=\mathrm{GL}_n(K)\) and \(r=1\), (ii) when \(G\) is \(\mathrm{O}_n(K)\) or \(\mathrm{U}__n(K)\) and the \(V_i\) are anisotropic. To find techniques that will cope with the case of anisotropic \(V\) in general is probably very difficult; for the action of \(F\) on \(V\) is determined by the actions of the orthogonal or unitary group of the restrictions of \(f\) to the \(V_i\), and the groups of anisotropic forms depend crucially on \(K\) and are hard to elucidate. So here we settle King's excluded cases when \(K\) is a finite field \(\mathrm{GF}(q)\). If \(G\) is \(\mathrm{U}_n(q)\), with \(q\) necessarily a square, then \(F\) is always maximal in \(G\). So it is when \(G= \mathrm{GL}_n(q)\) except when \(q=2,3\) or \(4\). In both these cases \(r=1\). If \(G=\mathrm{O}_n(q)\) then one possibility is that \(r=2\). Then \(F\) is maximal in \(G\) except when \(q=3\) or \(q=2\) and \(n=4\). The exceptions to maximality are geometrically easily explained; usually \(F\) is properly contained in another of the groups under consideration, so the whole situation fits nearly together. The only other possibility is that \(r=1\), \(q\) is odd and \(G=\mathrm{O}_n(q)\). If \(n\neq 4\) then \(F\) is maximal in \(G\) if and only if \(q\) is a prime and \(q\equiv \pm 3\pmod 8\). This result is striking for another reason: \(| F| =2^nn!\), independent of \(q\), so that for large \(q\) our \(F\) is a very small maximal subgroup of \(\mathrm{O}_n(q)\). The subgroups containing \(F\) when \(q\) is a prime and \(q\equiv \pm 1\pmod 8\) are found. The existence of triads of desmic tetrahedra separates out the case \(n=4\). The proofs for the cases \(r=1\) use \textit{A. Wagner}'s classification of groups generated by reflections and homologies [Geom. Dedicata 7, 387--398 (1978; Zbl 0402.20036); ibid. 9, 239--253 (1980; Zbl 0443.51009); ibid. 10, 191--203 (1981; Zbl 0471.51014 and ibid. 10, 475-523 (1981; Zbl 0471.51015)]. The proof of the case \(r=2\) and \(G=\)\mathrm{O}_n(q)\( uses the action of orthogonal symmetries.\)