Geometric finiteness and rationality (Q1210358)

The paper under review answers positively one of the main questions raised in a paper of \textit{S. M. Gersten} and \textit{H. B. Short} [Ann. Math., II. Ser. 134, 125-158 (1991; Zbl 0744.20035)]. To state the result, let us recall briefly (and informally) a few definitions. (The paper is concise, and for background and definitions, the reader is referred to the paper of Gersten and Short.) Let \(A\) be a finite set and \(A^*\) the free monoid generated by \(A\). A subset of \(A^*\) is said to be a regular language if it is the set of words recognized by a finite state automaton. An automatic structure for a group \(G\) is given by a pair \((A,L)\) where \(A\) is a finite set which is a semi-group generator of \(G\), and \(L\) a regular language in \(A^*\), with the property that the question whether two elements of \(L\) representing elements of \(G\) at distance \(\leq 1\) in the Cayley graph can be decided by a finite state automaton. The notion of an automatic group is due to Thurston et al. A subset \(S\subset G\) is said to be \(L\)- rational if the complete inverse image of \(S\) in \(L\) is a regular subset. The paper considers geometrically finite groups, that is, discrete subgroups of isometries of the hyperbolic space \(H^ n\) such that the quotient by \(G\) of the Nielsen convex hull of the action is compact (for this, and other equivalent definitions, see Thurston's Princeton lecture notes). We can state now the main theorem of this paper: Let \(G\) be a geometrically finite group without parabolics, and let \((A,L)\) be any rational structure on \(G\) which is automatic. Then, a subgroup \(H\) of \(G\) is \(L\)-rational if and only if \(H\) is geometrically finite. In his proof of this theorem, the author establishes also the following result, which concerns negatively curved groups (i.e. hyperbolic groups in the sense of Gromov): Let \(G\) be a negatively curved group. A subgroup \(H\) of \(G\) is \(L\)-rational with respect to any automatic structure \((A,L)\) on \(G\) if and only if the embedding of \(H\) in \(G\) is a quasi- isometry (with respect to any Cayley-graph metric). Finally, the author makes new conjectures.