The limit sets of subgroups of lattices in \(\mathrm {PSL}(2,\mathbb R)^r\) (Q292051)

An important geometric invariant of groups acting on a symmetric space \(X\) of noncompact type is their limit set, the set of accumulation points on the boundary \(\partial X\) of an orbit in \(X\). When \(X\) is of higher rank there are many notions of boundary for \(X\), which are not equivalent to each other. The two which are of interest for the present paper, which deals with the case where \(X = (\mathbb H^2)^r\) is the symmetric space associated to the semisimple Lie group \(\mathrm{PSL}_2(\mathbb R)^r\), are the visual boundary which is homeomorphic to the \(2r-1\)-dimensional sphere, and the Furstenberg boundary which in this case is just the product of the boundaries \(\partial \mathbb H^2 = \mathbb S^1\) of the rank-1 factors and is thus homeomorphic to the \(r\)-dimensional torus. The limit set in the Furstenberg boundary is denoted by \(F_\Gamma\). The visual boundary has a stratification into \(\mathrm{PSL}_2(\mathbb R)^r\)-invariant subsets; the regular part is a dense open subset and the part of the limit set which lies therein is denoted by \(\mathcal L_\Gamma^{reg}\). It was proven by G. Link that the regular limit set is homeomorphic to the product with the Furstenberg boundary of a ``projective limit set''. The subgroups for which it is most natural to investigate these sets are finitely generated, infinite-index subgroups of lattices in \(\mathrm{PSL}_2(\mathbb R)^r\): it is well-known that the limit set of a lattice is always the full boundary, and there are not many known constructions of discrete subgroups. Let \(\Gamma \leq \Delta\) be such a subgroup. The simplest case is when \(\Gamma\) is Zariski-dense despite being of infinite covolume; the author proves that in this case the following dichotomy holds: if \(\Gamma\) has a discrete projection \(p_j(\Gamma)\) to one of the \(\mathrm{PSL}_2(\mathbb R)\)-factors then \(F_\Gamma\) is equal to the product of the limit set \(\mathcal L_{p_j(\Gamma)} \subset \partial \mathbb H^2\) (either the full boundary or a Cantor subset) with \((\partial H^2)^m\), where \(m\) is the number of nondiscrete projections of \(\Gamma\); if not, then \(F_\Gamma\) is the whole Furstenberg boundary. The description of the regular limit set in this case is completed by a result of \textit{Y. Bénoist} [Geom. Funct. Anal. 7, No. 1, 1--47 (1997; Zbl 0947.22003)] which states that the projective factor is open. If \(\Gamma\) is not Zariski-dense in \(\mathrm{PSL}_2(\mathbb R)^r\) and it is not elementary, then its Zariski-closure is a diagonal embedding \(\mathrm{PSL}_2(\mathbb R)^{r/k} \subset \mathrm{PSL}_2(\mathbb R)^r\) for some divisor \(k\) of \(r\) (which can be read off the extension of invariant trace fields associated to \(\Gamma\) and \(\Delta\)); the previous theorem then states that the limit set is either homeomorphic to the full Furstenberg boundary of \((\mathbb H^2)^{r/k}\), or to a product with the limit set of a discrete projection if one exists. On the other hand the result of Bénoist describing the regular limit set is not available in this case, and the author proves that the projective factor is of dimension \(r/k - 1\).