Tessellations of homogeneous spaces of classical groups of real rank two (Q1430024)

A tessellation of a homogeneous space \(M=G/H\) is a covering by translates of the fundamental domain of a crystallographic group \(\Gamma\) (this means that the action is proper and discontinuous and that the quotient \(\Gamma\backslash M\) is compact; also, if the action is free then \(\Gamma\backslash M\) is said to be a Clifford--Klein form for \(M\)). The group \(G\) is supposed to be simple and connected with finite centre. The problem of the existence of a tessellation is not easy if \(M\) and \(H\) are noncompact and the real rank of \(G\) is strictly greater than \(1\). The case of compact \(G\) is trivial. If \(H\) is compact then \(M\) admits a tessellation due to Borel's theorem: any simple connected Lie group has a cocompact lattice. If rank\(\,G=1\) and neither \(M\) nor \(H\) are compact then \(M\) contains a compact subset \(C\) such that \(C\cap gC\neq\emptyset\) for all \(g\in G\); this is an obstruction to the existence of an infinite proper discontinuous group \(\Gamma\). For higher ranks the situation is much more complicated. However, there are several essential facts and observations which possibly could clarify it. The main tool is the Cartan projection. Let \(G=KAK\), where \(K\) is a maximal compact subgroup, \(A\cong{\mathbb R}^r\), \(r=\mathop{\text{ rank}}G\), be the Cartan decomposition of \(G\). Fix a Weyl chamber \(C\) of the real root system and set \(A^+=\exp(C)\). Then each \(g\in G\) has the decomposition \(g=k_1ak_2\), where \(k_1,k_2\in K\) and \(a\in A^+\) is unique; this \(a\) is the Cartan projection of \(g\). For subsets \(H_1,H_2\subseteq G\), \(H_1\sim H_2\) means that there exists a compact set \(C\subset G\) such that \(H_1\subseteq CH_2C\) and \(H_2\subseteq CH_1C\) (for example, \(A\sim G\)). This equivalence is quite transparent for vector groups and is very natural in the context of crystallographic groups (for instance, the Calabi-Marcus phenomenon might be formulated as follows: if \(H\sim G\) then \(M\) has no tessellation). Roughly speaking, the Cartan projection keeps this relation for subgroups of \(G\) [\textit{Y. Benoist}, Ann. Math. (2) 144, No.~2, 315--347 (1996; Zbl 0868.22013), \textit{T. Kobayashi}, J. Lie Theory 6, No.2, 147--163 (1996; Zbl 0863.22010)]. There is no complete classification of the homogeneous spaces admitting tessellations for rank\(\,G=2\). The authors obtained it for \(G=\text{ SU}(2,2n)\) and get many partial results. For a connected Lie group \(H\), set \(d(H)=\dim H-\dim K_H\), where \(K_H\) is a maximal compact subgroup of \(H\). If \(G= \text{ SU}(2,2n)\) and \(H\) is its closed connected subgroup then \(G/H\) has a tessellation if and only if \(d(H)=4n\) and either \(H\sim\text{ SU}(1,2n)\) or \(H\sim\text{ Sp}(1,n)\). There is also a more precise version. Let \(G=KAN\) be the Iwasawa decomposition. The homogeneous space \(G/H\) has a tessellation if and only if it is conjugate to a subgroup \(H'\) such that \(H'\cap AN\) is a deformation of the group \(\text{ SU}(1,2n)\cap AN\) or the group \(\text{ Sp}(1,n)\cap AN\). The deformations are explicitly described. Similar results were obtained by Oh and Witte for \(\text{ SO}(2,n)\) (to appear). The paper contains a description of the results of other authors, many illustrative examples and it is quite self-contained.