Deformation theory and finite simple quotients of triangle groups. I. (Q461275)

For given integers \(k,\ell,m\geq 2\), a group \(G\) is said to be \((k,\ell,m)\)-generated if it can be generated by elements \(x\) and \(y\) of orders \(k\) and \(\ell\) such that \(xy\) has order \(m\), or equivalently, if \(G\) is a smooth quotient of the ordinary \((k,\ell,m)\) triangle group, which is the abstract group \(\Delta(k,\ell,m)\) with presentation \(\langle x,y,z\mid xyz=z^k=y^\ell=z^m=1\rangle\). If \(1/k+1/\ell+1/m>1\) then \(\Delta(k,\ell,m)\) is finite, while if \(1/k+1/\ell+1/m=1\) then it is infinite but soluble, and if \(1/k+1/\ell+1/m<1\) then it is infinite but insoluble. These are called the \textit{spherical}, \textit{Euclidean} and \textit{hyperbolic} cases, respectively. The hyperbolic triangle groups are co-compact Fuchsian groups (discrete subgroups of \(\text{PSL}(2,\mathbb R)\)) with fundamental region of small hyperbolic area, and as such, their finite smooth quotients give `large' groups of automorphisms of compact Riemann surfaces of genus \(g>1\). The most important hyperbolic case occurs when \((k,\ell,m)=(2,3,7)\), because this gives the largest such groups for given \(g\), namely \(84(g-1)\), and for this reason, the non-trivial quotients of \(\Delta(2,3,7)\) are known as \textit{Hurwitz groups}. Over the last 30 years or so, a lot of attention has been paid to proving or dis-proving the existence of \((k,\ell,m)\)-generating pairs for given finite simple groups, when \((k,\ell,m)=(2,3,7)\) or \((k,\ell)=(2,3)\) for example. Most approaches have been constructive, or probabilistic, or enumerative (using character theory and counting arguments on subgroup lattices). In this paper, a new approach is taken, based on the theory of representation varieties, via deformation theory. The authors weaken the condition on the (exact) orders of the generators, and say that \(G\) is a \((k,\ell,m)\)-group if it is a quotient of \(\Delta(k,\ell,m)\), not necessarily smooth. They study the representation variety \(\Hom(\Delta,\underline G)\) where \(\Delta=\Delta(k,\ell,m)\) is a hyperbolic triangle group and \(\underline G\) is a quasi-simple algebraic group over some infinite field \(F\). A key point is that infinitely many finite simple groups of the same type as \(\underline G\) are \((k,\ell,m)\)-groups if and only if there exists a representation \(\rho\colon\Delta\to\underline G\) which is not `locally rigid' and has Zariski dense image. Here the `type' is given by an irreducible Dynkin diagram, say \(X\), and the authors say that \(\Delta\) is \textit{saturated with finite quotients of type \(X\)} if there exist natural numbers \(p_0\) and \(e\) such that the finite simple group \(X(p^et)\) is a quotient of \(\Delta\) for all \(t\in N\) and all \(p>p_0\), and \(X(p^t)\) is a quotient of \(\Delta\) for all \(t\in N\) and all \(p\) in a set of primes of positive density. Then their main theorem states that \(\Delta\) is saturated with finite quotients of type \(X\) in almost all cases, and gives a list of possible exceptions. In particular, the assertion holds whenever \(1/k+1/\ell+1/m<1/2\) and \(X\neq A_1\). The approach they take is to compose a representation of \(\Delta\) to \(\text{SO}(3,\mathbb R)\) with the principal homomorphism from \(\text{SO}(3,\mathbb R)\) to a compact simple Lie group \(G\) of type \(X\), and then to deform the resulting (non-dense) homomorphism (from \(\Delta\to G\)). A disadvantage of this approach is a lack of good control on the two parameters \(p_0\) and \(e\) mentioned above. On the other hand, the authors are able to exploit the method further and make some improvements to their results, at the expense of some more technical detail, in a sequel [in part II, Groups Geom. Dyn. 8, No. 3, 811-836 (2014; Zbl 1312.20031)]. They also answer a question posed by Guralnick, showing that \(\Delta(2,3,7)\) is saturated with finite quotients of type \(X\) except possibly when \(X=A_r\) for \(1\leq r\leq 19\), or \(B_3\), or \(C_2\), or \(D_r\) for \(19\) small values of \(r\) (between \(4\) and \(43\)), or \(E_6\). In particular, it follows that infinitely many of the simple groups \(\text{PSp}_n(q)\) are Hurwitz groups, for any given \(n\neq 4\), and that \(\Delta(2,3,7)\) is saturated with finite quotients of type \(E_8\). A further new theorem is given in the case where the triple is `rigid' (in a sense involving dimensions of sub-varieties), thereby proving an earlier conjecture by the third author.