Groups generated by k-transvections (Q910851)

Let k be a field and let \(\Sigma\) be a class of abelian subgroups of the group G that generate G. By definition, subgroup U of G is called a \(\Sigma\)-subgroup if \(U=<U\cap \Sigma >\) where \(U\cap \Sigma =\{A\in \Sigma |\) \(A\leq U\}.\) Also \(\Sigma\) is called a class of k-root-subgroups of G if \(\Sigma\) satisfies: (1) if \(A,B\in \Sigma\), then one of the following holds: \((\alpha)[A,B]=1\), \((\beta)<A,B>\cong (P)SL_ 3(k)\) and A,B are full unipotent subgroups of \(<A,B>\), \((\gamma)<A,B>\) is special with \([a,B]=[b,A]=[A,B]\in \Sigma\) for all \(a\in A^{\#}\) and \(b\in B^{\#}\); and (2) G satisfies the maximality condition (ascending chain condition) for \(\Sigma\)-subgroups. If possibility (1) \((\gamma)\) never occurs, then \(\Sigma\) is said to be a class of k-transvections of G. A prototype for groups generated by a class of k-root subgroups ought to be Lie-type and classical groups over k (resp. over extension skewfields of k where k is the fixed field of some involutory antiautomorphism) with \(\Sigma\) the class of ``long-root subgroups''. The main result of this important paper is: Theorem: Let k be a field, and let G be a group generated by a class \(\Sigma\) of k-transvections. Suppose that there exist \(A\neq B\in \Sigma\) with \([A,B]=1\) and that one of the following two conditions is satisfied: (i) \(| k| >3\) and G has no nilpotent normal subgroup; or (ii) \(| k| \leq 3\), \(G'\) is simple, \(Z(G)=1\) and \(Z(<C_{\Sigma}(AB)>)\cap \Sigma \neq \{A,B\}\). Then one of the following holds: (1) \(G\cong PSp(2n,k)\), \(n\geq 2\) and \(\Sigma\) is the class of symplectic transvection groups; (2) \(G\cong PSU_ n(\bar k,f)\) where \(k\subset \bar k\) is a division ring, f a non-degenerate \(\sigma\)-hermitian form of Witt-index \(\geq 2\), \(\sigma\) an involutory antiautomorphism of \(\bar k\) with \(\bar k_{\sigma}=\{c\in \bar k|\) \(c^{\sigma}=c\}\cong k\). Further \(\Sigma\) is the uniquely determined class of ``unitary transvection groups''. Moreover, if \(\bar k\) is non-commutative, then char \(k\neq 2\) and \(\bar k\) is a quaternion algebra over k. (3) k is infinite of char 2. There exists a quaternion division ring \(\bar k\) over k, a finite dimensional left \(\bar k-\)vector space V, a non-degenerate pseudo-quadratic form q: \(V\to \bar k\) over k of Witt- index \(\geq 2\) with associated \(\sigma\)-hermitian form f, where \(\sigma\) is the standard antiautomorphism of \(\bar k\) (satisfying \(\bar k_{\sigma,1}=\{c+c^{\sigma}|\) \(c\in \bar k\}=k)\) such that (a) there exists a monomorphism \(\phi: G\to PO(V,q)\) with \(\phi(G)\leq PO(V,q)\); and (b) \(\phi(\Sigma)\) is the set (of images under \(P: O(V,q)\to PO(V,q))\) of orthogonal transvection groups: \(T_ p=\{t_ c: v\to v+f(v,p)cp|\) \(c\in k\}\) where \(0\neq p\in V\) with \(q(p)=0.\) (4) There exists a non-perfect field \(\bar k\) of char 2 with \(\bar k^ 2\subseteq k\subset \bar k\), a finite dimensional \(\bar k\)-vector space V, a non-degenerate quadratic form q: \(V\to \bar k\) of Witt-index \(\geq 2\) such that: (a) \(\dim_{\bar k^ 2}(k)=\dim_{\bar k}V^{\perp}\) (where \(V^{\perp}\) is defined with respect to the associated bilinear form f); and (b) there exists a monomorphism \(\phi: G\to PO(V,q)\) such that \(\phi(G)\leq PO(V,q)\) and \(\phi(\Sigma)\) is the set (of images under \(P: O(V,q)\to PO(V,q))\) of orthogonal transvection groups: \(T_ p=\{t_ c: w\to w+f(w,p)c(p+(1/c)b_ c)|\) \(0\neq c\in k\) and \(v_ c\in V^{\perp}\}\) where \(0\neq p\in V\) with \(q(p)=0.\) (5) k is infinite and there exists a Cayley division algebra K over k such that the set of subspaces of \(\Sigma\) is isomorphic to the uniquely determined polar space \(\pi\) of rank 3 over K (the planes of which are non-desarguesian). Moreover G is isomorphic to the uniquely determined non-trivial simple normal subgroup of Aut(\(\pi)\). (6) k is infinite: \((\Sigma,L,\subseteq)\) is a non-embeddable Moufang generalized quadrangle on which the elements of \(\Sigma\) act as transvection groups (i.e., pointwise fix all lines through a given point) where \(L=\{C_{\Sigma}(AB)|\) \(A\neq B\in \Sigma\) with \([A,B]=1\}.\) The proof of this theorem is mainly geometric.