Ideal triangle groups, dented tori, and numerical analysis. (Q1826267)

The complex hyperbolic plane \(\mathbb{C}\mathbb{H}^2\) is a 2-complex-dimensional manifold which is negatively curved, Kähler, and also a symmetric space. It contains \(\mathbb{H}^2\) as a totally real, totally geodesic subspace. The paper under review is about the theory of deforming \(\text{Iso}(\mathbb{H}^2)\) representations into \(\text{Iso}(\mathbb{C}\mathbb{H}^2)\), where \(\text{Iso}(X)\) denotes the isometry group of the metric space \(X\). \textit{W. M. Goldman} and \textit{J. R. Parker} [J. Reine Angew. Math. 425, 71-86 (1992; Zbl 0739.53055)] defined and partially classified the complex hyperbolic ideal triangle groups which are representations \(\rho_s\colon\Gamma\to\text{Iso}(\mathbb{C}\mathbb{H}^2)\). Here \(\Gamma=\mathbb{Z}_2*\mathbb{Z}_2*\mathbb{Z}_2\). The representation \(\rho_s\) maps the standard generators of \(\Gamma\) to distinct, order-two, complex reflections, such that any product of two distinct generators is parabolic. Modulo conjugation, there is a one-parameter family \(\{\rho_s\mid s\in\mathbb{R}\}\) of such representations. The indexing parameter, \(s\), is the tangent of the angular invariant of the ideal triangle formed by the three complex lines fixed by the generators. Goldman and Parker proved that if \(|s|>\overline s\) then \(\rho_s\) is not a discrete embedding, but if \(|s|\leq\underline s\) then \(\rho_s\) is a discrete embedding. Here \(\underline s=(105/3)^{1/2}\) and \(\overline s=(125/3)^{1/2}\). That result is not sharp. Goldman and Parker conjectured that \(\rho_s\) remains a discrete embedding for \(|s|\in(\underline s,\overline s]\). The author proves a sharp version of this conjecture. That means: \(\rho_s\) is a discrete embedding if and only if \(g_s\) is not elliptic. Moreover, \(\rho_s\) is indiscrete if \(g_s\) is elliptic. Here \(g_s\) is the product of all three generators of \(\rho_s(\Gamma)\). The proof is going through some construction of a surface-like set \(\mathbb{Z}(s)\subset\partial\mathbb{C}\mathbb{H}^2\). Then is proved that the orbit \(\rho_s(\Gamma)\mathbb{Z}(s)\) consists of disjoint surface-like sets and it feeds into a variant of the Klein combination theorem to prove that \(\rho_s\) is a discrete embedding. The construction of \(\mathbb{Z}(s)\) is experimentally, using a computer. The author mentions that it was done during 350 computational hours, when run on Sparc Ultra 5.