Geometric structures and representations of discrete groups
GGEOMETRIC STRUCTURES AND REPRESENTATIONS OFDISCRETE GROUPS
FANNY KASSEL
Abstract.
We describe recent links between two topics: geometric structureson manifolds in the sense of Ehresmann and Thurston, and dynamics “atinfinity” for representations of discrete groups into Lie groups. Introduction
The goal of this survey is to report on recent results relating geometric structureson manifolds to dynamical aspects of representations of discrete groups into Liegroups, thus linking geometric topology to group theory and dynamics.1.1.
Geometric structures.
The first topic of this survey is geometric structureson manifolds. Here is a concrete example as illustration (see Figure 1).
Example 1.1.
Consider a two-dimensional torus T .(1) We can view T as the quotient of the Euclidean plane X = R by Γ = Z ,which is a discrete subgroup of the isometry group G = O(2) (cid:110)R of X (acting by lin-ear isometries and translations). Viewing T this way provides it with a Riemannianmetric and a notion of parallel lines, length, angles, etc. We say T is endowed witha Euclidean (or flat ) structure , or a ( G, X )-structure with (
G, X ) = (O(2) (cid:110)R , R ).(2) Here is a slightly more involved way to view T : we can see it as the quotientof the affine plane X = R by the group Γ generated by the translation of vector ( )and the affine transformation with linear part ( ) and translational part ( ). Thisgroup Γ is now a discrete subgroup of the affine group G = GL(2 , R ) (cid:110) R . Viewing T this way still provides it with a notion of parallel lines and even of geodesic, butno longer with a notion of length or angle or speed of geodesic. We say T is endowedwith an affine structure , or a ( G, X )-structure with (
G, X ) = (GL(2 , R ) (cid:110) R , R ).(3) There are many ways to endow T with an affine structure. Here is a differentone: we can view T as the quotient of the open subset U = R (cid:114) { } of X = R bythe discrete subgroup Γ of G = GL(2 , R ) (cid:110) R generated by the homothety ( / ).This still makes T “locally look like” X = R , but now the image in T of an affinegeodesic of X pointing towards the origin is incomplete (it circles around in T withshorter and shorter period and disappears in a finite amount of time).As in Example 1.1, a key idea underlying a large part of modern geometry isthe existence of model geometries which various manifolds may locally carry. Bydefinition, a model geometry is a pair ( G, X ) where X is a manifold ( model space )and G a Lie group acting transitively on X ( group of symmetries ). In Example 1.1we encountered ( G, X ) = (O( n ) (cid:110) R n , R n ) and ( G, X ) = (GL( n, R ) (cid:110) R n , R n ),corresponding respectively to Euclidean geometry and affine geometry . Anotherimportant example is X = H n (the n -dimensional real hyperbolic space) and G =PO( n,
1) = O( n, / {± I } (its group of isometries), corresponding to hyperbolic Partially supported by the European Research Council under Starting Grant 715982 (DiGGeS). a r X i v : . [ m a t h . G T ] F e b F. KASSEL
Figure 1.
Tilings of X = R showing the three Γ-actions in Example 1.1 geometry . (For n = 2 we can see X as the upper half-plane and G , up to index two,as PSL(2 , R ) acting by homographies.) We refer to Table 1 for more examples.The idea that a manifold M locally carries the geometry ( G, X ) is formalizedby the notion of a (
G, X ) -structure on M : by definition, this is a maximal atlas ofcoordinate charts on M with values in X such that the transition maps are given byelements of G (see Figure 2). Note that this is quite similar to a manifold structureon M , but we now require the charts to take values in X rather than R n , and thetransition maps to be given by elements of G rather than diffeomorphisms of R n . g ∈ G XM Figure 2.
Charts defining a (
G, X )-structure on M Although general (
G, X )-structures may display pathological behavior (see [59]), inthis survey we will restrict to the two “simple” types of (
G, X )-structures appearingin Example 1.1, to which we shall give names to facilitate the discussion: • Type C (“complete”) : (
G, X )-structures that identify M with a quotientof X by a discrete subgroup Γ of G acting properly discontinuously; • Type U (“incomplete but still uniformizable”) : (
G, X )-structures thatidentify M with a quotient of some proper open subset U of X by a discretesubgroup Γ of G acting properly discontinuously.Setting V = X or U as appropriate, we then have coverings (cid:102) M (cid:39) (cid:101) V → V → Γ \V (cid:39) M (where (cid:101) denotes universal covers). The charts on M are obtained by takingpreimages in V ⊂ X of open subsets of M . Moreover, the basic theory of coveringgroups gives a natural group homomorphism hol : π ( M ) → G with image Γ andkernel π ( V ), called the holonomy .In this survey, we use the phrase geometric structures for ( G, X )-structures. Weshall not detail the rich historical aspects of geometric structures here; instead, werefer to the excellent surveys [57, 58, 59]. We just mention that the notion of modelgeometry has its origins in ideas of Lie and Klein, formulated in Klein’s 1872 Erlan-gen program. Influenced by these ideas and those of Poincar´e, Cartan and others,Ehresmann [51] initiated a general study of geometric structures in 1935. Later,geometric structures were greatly promoted by Thurston’s revolutionary work [107].1.2.
Classifying geometric structures.
The fundamental problem in the theoryof geometric structures is their classification, namely:
Problem A.
Given a manifold M , EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 3
Figure 3.
The limit set of a quasi-Fuchsian group in ∂ ∞ H (cid:39) C ∪ {∞} (1) Describe which model geometries ( G, X ) the manifold M may locally carry;(2) For a fixed model ( G, X ), describe all possible (
G, X )-structures on M .We refer to [57] for a detailed survey of Problem A with a focus on dimensionstwo and three, and to [83] for a special case.Problem A.(1) asks how the global topology of M determines the geometriesthat it may locally carry. This has been the object of deep results, among which: • the classical uniformization theorem : a closed Riemann surface may carry aEuclidean, a spherical, or a hyperbolic structure, depending on its genus; • Thurston’s hyperbolization theorem : a large class of 3-dimensional manifolds,defined in purely topological terms, may carry a hyperbolic structure; • more generally, Thurston’s geometrization program (now Perelman’s theorem):any closed orientable 3-dimensional manifold may be decomposed into pieces,each admitting one of eight model geometries (see [16]).Problem A.(2) asks to describe the deformation space of ( G, X )-structures on M .In the simple setting of Example 1.1, this space is already quite rich (see [96]). Forhyperbolic structures on a closed Riemann surface of genus ≥ Teichm¨uller theory .1.3.
Representations of discrete groups.
The second topic of this survey is rep-resentations (i.e. group homomorphisms) of discrete groups (i.e. countable groups)to Lie groups G , and their dynamics “at infinity”. We again start with an example. Example 1.2.
Let Γ = π ( S ) where S is a closed oriented Riemann surface of genus ≥
2. By the uniformization theorem, S carries a complete (“type C”) hyperbolicstructure, which yields a holonomy representation Γ → PSL(2 , R ) as in Section 1.1.Embedding PSL(2 , R ) into G = PSL(2 , C ), we obtain a representation ρ : Γ → G ,called Fuchsian , and an associated action of Γ on the hyperbolic space X = H andon its boundary at infinity ∂ ∞ H = (cid:98) C (the Riemann sphere). The limit set of ρ (Γ)in (cid:98) C is the set of accumulation points of ρ (Γ)-orbits of X ; it is a circle in the sphere (cid:98) C . Deforming ρ slightly yields a new representation ρ (cid:48) : Γ → G , called quasi-Fuchsian , which is still faithful, with discrete image, and whose limit set in (cid:98) C is stilla topological circle (now “wiggly”, see Figure 3). The action of ρ (cid:48) (Γ) is chaotic onthe limit set (e.g. all orbits are dense) and properly discontinuous on its complement.Example 1.2 plays a central role in the theory of Kleinian groups and in Thurston’sgeometrization program; it was extensively studied by Ahlfors, Beardon, Bers, Mar-den, Maskit, Minsky, Sullivan, Thurston, and many others.In this survey we report on various generalizations of Example 1.2, for represen-tations of discrete groups Γ into semisimple Lie groups G which are faithful (or withfinite kernel) and whose images are discrete subgroups of G . While in Example 1.2 F. KASSEL the group G = PSL(2 , C ) has real rank one (meaning that its Riemannian sym-metric space H has no flat region beyond geodesics), we also wish to consider thecase that G has higher real rank , e.g. PGL( d, R ) with d ≥
3. In general, semisimplegroups G tend to have very different behavior depending on whether their real rankis one or higher; for instance, the lattices of G (i.e. the discrete subgroups of finitecovolume for the Haar measure) may display some forms of flexibility in real rankone, but exhibit strong rigidity phenomena in higher real rank. Beyond lattices, thelandscape of discrete subgroups of G is somewhat understood in real rank one (atleast several important classes of discrete subgroups have been identified for theirgood geometric, topological, and dynamical properties, see Section 3.1), but it re-mains very mysterious in higher real rank. We shall explain some recent attemptsat understanding it better.One interesting aspect is that, even when G has higher real rank, discrete sub-groups of G of infinite covolume may be nonrigid and in fact admit large deformationspaces. In particular, as part of higher Teichm¨uller theory , there has recently beenan active and successful effort to find large deformation spaces of faithful and dis-crete representations of surface groups π ( S ) into higher-rank semisimple G whichshare some of the rich features of the Teichm¨uller space of S (see Sections 4.3 and 5,and [27, 110]). Such features also include dynamics “at infinity” as in Example 1.2,which are encompassed by a notion of Anosov representation [85] (see Section 4).1.4.
Flag varieties and boundary maps.
Let us be a bit more precise. Givena representation ρ : Γ → G , by dynamics ‘at infinity” we mean the dynamics of theaction of Γ via ρ on some flag varieties G/P (where P is a parabolic subgroup), seenas “boundaries” of G or of its Riemannian symmetric space G/K . In Example 1.2we considered a rank-one situation where G = PSL(2 , C ) and G/P = ∂ ∞ H = (cid:98) C .A typical higher-rank situation that we have in mind is G = PGL( d, R ) with d ≥ G/P = Gr i ( R d ) (the Grassmannian of i -planes in R d ) for some 1 ≤ i ≤ d − G/P .More recently, in the context of higher Teichm¨uller theory [26, 52, 85], it has provedimportant to study continuous equivariant boundary maps which embed the bound-ary ∂ ∞ Γ of a Gromov hyperbolic group Γ (i.e. the visual boundary of the Cayleygraph of Γ) into
G/P . Such boundary maps ξ : ∂ ∞ Γ → G/P define a closed invari-ant subset ξ ( ∂ ∞ Γ) of
G/P , the limit set , on which the dynamics of the action byΓ accurately reflects the intrinsic chaotic dynamics of Γ on ∂ ∞ Γ. These boundarymaps may be used to transfer the Anosov property of the intrinsic geodesic flowof Γ into some uniform contraction/expansion properties for a flow on a natural flatbundle associated to ρ and G/P (see Section 4). They may also define some opensubsets U of G/P on which the action of Γ is properly discontinuous, by removingan “extended limit set” L ρ (Γ) ⊃ ξ ( ∂ ∞ Γ) (see Sections 3, 5, 6); this generalizes thedomains of discontinuity in the Riemann sphere of Example 1.2.For finitely generated groups Γ that are not Gromov hyperbolic, one can stilldefine a boundary ∂ ∞ Γ in several natural settings, e.g. as the visual boundaryof some geodesic metric space on which Γ acts geometrically, and the approachconsidered in this survey can then be summarized by the following general problem.
Problem B.
Given a discrete group Γ with a boundary ∂ ∞ Γ, and a Lie group G with a boundary G/P , identify large (e.g. open in Hom(Γ , G )) classes of faithful and
EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 5 discrete representations ρ : Γ → G for which there exist continuous ρ -equivariantboundary maps ξ : ∂ ∞ Γ → G/P . Describe the dynamics of Γ on
G/P via ρ .1.5. Goal of the paper.
We survey recent results on (
G, X )-structures (Prob-lem A) and on representations of discrete groups (Problem B), making links betweenthe two topics. In one direction, we observe that various types of (
G, X )-structureshave holonomy representations that are interesting for Problem B. In the otherdirection, starting with representations that are interesting for Problem B (Anosovrepresentations), we survey recent constructions of associated (
G, X )-structures.These results tend to indicate some deep interactions between the geometry of(
G, X )-manifolds and the dynamics of their holonomy representations, which largelyremain to be explored. We hope that they will continue to stimulate the develop-ment of rich theories in the future.
Organization of the paper.
In Section 2 we briefly review the notion of a holo-nomy representation. In Section 3 we describe three important families of (
G, X )-structures for which boundary maps into flag varieties naturally appear. In Sec-tion 4 we define Anosov representations and give examples and characterizations.In Section 5 we summarize recent constructions of geometric structures associatedto Anosov representations. In Section 6 we discuss a situation in which the linksbetween geometric structures and Anosov representations are particularly tight, inthe context of convex projective geometry. In Section 7 we examine an instance of(
G, X )-structures for a nonreductive Lie group G , corresponding to affine manifoldsand giving rise to affine Anosov representations. We conclude with a few remarks. Acknowledgements.
I would like to heartily thank all the mathematicians whohelped, encouraged, and inspired me in the past ten years; the list is too longto include here. I am very grateful to all my coauthors, in particular those in-volved in the work discussed below: Jeffrey Danciger ( §
5, 6, 7), Fran¸cois Gu´eritaud( §
4, 5, 6, 7), Olivier Guichard ( §
4, 5), Rafael Potrie ( § §
4, 5).I warmly thank J.-P. Burelle, J. Danciger, O. Guichard, and S. Maloni for readingearlier versions of this text and making many valuable comments and suggestions,and R. Canary and W. Goldman for kindly answering my questions.2.
Holonomy representations
Let G be a real Lie group acting transitively, faithfully, analytically on a mani-fold X , as in Table 1. In Section 1.1 we defined holonomy representations for certaintypes of ( G, X )-structures. We now give a short review of the notion in general.Type of geometry
X G H
Real projective P n ( R ) PGL( n + 1 , R ) stab. in G of a line of R n +1 Affine R n Aff( R n ) = GL( n, R ) (cid:110)R n GL( n, R )Euclidean R n Isom( R n ) = O( n ) (cid:110) R n O( n )Real hyperbolic H n Isom( H n ) = PO( n,
1) O( n )Spherical S n Isom( S n ) = O( n + 1) O( n )Complex projective P n ( C ) PGL( n + 1 , C ) stab. in G of a line of C n +1 Table 1.
Some examples of model geometries (
G, X ), where X (cid:39) G/H
Let M be a ( G, X )-manifold, i.e. a manifold endowed with a (
G, X )-structure.Fix a basepoint m ∈ M and a chart ϕ : U → X with m ∈ U . We can lift any loop F. KASSEL on M starting at m to a path on X starting at ϕ ( m ), using successive charts of M which coincide on their intersections; the last chart in this analytic continuationprocess coincides, on an open set, with g · ϕ for some unique g ∈ G ; we set hol( γ ) := g where γ ∈ π ( M, m ) is the homotopy class of the loop (see Figure 4). This defines arepresentation hol : π ( M ) → G called the holonomy (see [57, 59] for details); it isunique modulo conjugation by G . This coincides with the notion from Section 1.1;in particular, if M (cid:39) Γ \V with V open in X and Γ discrete in G , and if V is simplyconnected, then hol : π ( M ) → Γ is just the natural identification of π ( M ) with Γ. M m U ϕ X g ∈ G Figure 4.
Construction of a holonomy representationWe shall define the deformation space
Def ( G,X ) ( M ) to be the quotient of the setof marked ( G, X )-structures on M (i.e. pairs ( M (cid:48) , f (cid:48) ) where M (cid:48) is a ( G, X )-manifoldand f : M → M (cid:48) a diffeomorphism) by the group of diffeomorphisms of M isoto-pic to the identity (acting by precomposition). The holonomy defines a map fromDef ( G,X ) ( M ) to the space Hom(Γ , G ) /G of representations of Γ to G modulo conju-gation by G . This map may be bijective in some cases, as in Example 2.1 below, butin general it is not. However, when M is closed, the so-called Ehresmann–Thurstonprinciple [107] states that the map is continuous (for the natural topologies on bothsides), open, with discrete fibers; in particular, the set of holonomy representationsof (
G, X )-structures on M is then stable under small deformations. Example 2.1.
Let (
G, X ) = (PO(2 , , H ) where PO(2 , (cid:39) PGL(2 , R ) is theisometry group of the real hyperbolic plane H . Let M = S be a closed oriented con-nected surface of genus g ≥
2. All (
G, X )-structures on S are complete. Their ho-lonomy representations are the Fuchsian (i.e. faithful and discrete) representationsfrom π ( S ) to G . The deformation space Def ( G,X ) ( S ) is the Teichm¨uller spaceTeich( S ). The holonomy defines a homeomorphism between Teich( S ) (cid:39) R g − andthe space of Fuchsian representations from π ( S ) to G modulo conjugation by G .3. Examples of ( G, X ) -structures and their holonomy representations In this section we introduce three important families of (
G, X )-structures, whichhave been much studied in the past few decades. We observe some structural stabil-ity for their holonomy representations, and the existence of continuous equivariantboundary maps together with expansion/contraction properties “at infinity”. Thesephenomena will be captured by the notion of an Anosov representation in Section 4.3.1.
Convex cocompact locally symmetric structures in rank one.
Let G bea real semisimple Lie group of real rank one with Riemannian symmetric space X = G/K (i.e. K is a maximal compact subgroup of G ). E.g. ( G, X ) = (PO( n, , H n )for n ≥
2. Convex cocompact groups are an important class of discrete subgroups Γof G which generalize the uniform lattices. They are special cases of geometricallyfinite groups, for which no cusps appear; see Bowditch [20, 21] for a general theory.By definition, a discrete subgroup Γ of G is convex cocompact if it preservesand acts with compact quotient on some nonempty convex subset C of X = G/K ; EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 7 equivalently, the complete (
G, X )-manifold (or orbifold) Γ \ X has a compact con-vex subset (namely Γ \C ) containing all the topology. Such a group Γ is alwaysfinitely generated. A representation ρ : Γ → G is called convex cocompact if itskernel is finite and its image is a convex cocompact subgroup of G .For instance, in Example 1.2 the quasi-Fuchsian representations are exactly theconvex cocompact representations from π ( S ) to G = PSL(2 , C ); modulo conjuga-tion, they are parametrized by Teich( S ) × Teich( S ) [14]. Another classical exampleof convex cocompact groups in rank-one G is Schottky groups, namely free groupsdefined by the so-called ping pong dynamics of their generators in ∂ ∞ X .Here ∂ ∞ X denotes the visual boundary of X , yielding the standard compacti-fication X = X (cid:116) ∂ ∞ X of X ; for X = H n we can see X in projective space as inExample 3.2.(1) below. The G -action on X extends continuously to X , and ∂ ∞ X identifies with G/P where P is a minimal parabolic subgroup of G .For a convex cocompact representation ρ : Γ → G , the existence of a cocompactinvariant convex set C implies (by the ˇSvarc–Milnor lemma or “fundamental ob-servation of geometric group theory”) that ρ is a quasi-isometric embedding . Thismeans that the points of any ρ (Γ)-orbit in X = G/K go to infinity at linear speedfor the word length function | · | : Γ → N : for any x ∈ X there exist C, C (cid:48) > d G/K ( x , ρ ( γ ) · x ) ≥ C | γ | − C (cid:48) for all γ ∈ Γ. (This property does notdepend on the choice of finite generating subset of Γ defining | · | .) A consequence“at infinity” is that any ρ -orbital map Γ → X extends to a ρ -equivariant embedding ξ : ∂ ∞ Γ → ∂ ∞ X (cid:39) G/P , where ∂ ∞ Γ is the boundary of the Gromov hyperbolicgroup Γ. The image of ξ is the limit set Λ ρ (Γ) of ρ (Γ) in ∂ ∞ X . The dynamics on ∂ ∞ X (cid:39) G/P is decomposed as in Example 1.2: the action of ρ (Γ) is “chaotic” onΛ ρ (Γ) (e.g. all orbits are dense if Γ is nonelementary), and properly discontinuous,with compact quotient, on the complement Ω ρ (Γ) = ∂ ∞ X (cid:114) Λ ρ (Γ) .Further dynamical properties were studied by Sullivan: by [103], the action of ρ (Γ) on ∂ ∞ X (cid:39) G/P is expanding at each point z ∈ Λ ρ (Γ) , i.e. there exist γ ∈ Γand
C > ρ ( γ ) multiplies all distances by ≥ C on a neighborhood of z in ∂ ∞ X (for some fixed auxiliary metric on ∂ ∞ X ). This implies that the group ρ (Γ) is structurally stable , i.e. there is a neighborhood of the natural inclusion inHom( ρ (Γ) , G ) consisting entirely of faithful representations. In fact, ρ admits aneighborhood consisting entirely of convex cocompact representations, by a variantof the Ehresmann–Thurston principle. For G = SL(2 , C ), a structurally stablesubgroup of G is either locally rigid or convex cocompact, by [104].3.2. Convex projective structures: divisible convex sets.
Let G be the pro-jective linear group PGL( d, R ) and X the projective space P ( R d ), for d ≥
2. Recallthat a subset of X = P ( R d ) is said to be convex if it is contained and convex insome affine chart, properly convex if its closure is convex, and strictly convex if itis properly convex and its boundary in X does not contain any nontrivial segment. Remark 3.1.
Any properly convex open subset Ω of X = P ( R d ) admits a well-behaved (complete, proper, Finsler) metric d Ω , the Hilbert metric , which is invariantunder the subgroup of G = PGL( d, R ) preserving Ω (see e.g. [13]). In particular,any discrete subgroup of G preserving Ω acts properly discontinuously on Ω.By definition, a convex projective structure on a manifold M is a ( G, X )-structureobtained by identifying M with Γ \ Ω for some properly convex open subset Ω of X F. KASSEL and some discrete subgroup Γ of G . When M is closed, i.e. when Γ acts with com-pact quotient, we say that Γ divides Ω. Such divisible convex sets
Ω are the objectsof a rich theory, see [13]. The following classical examples are called symmetric . Examples 3.2. (1) For d = n + 1 ≥
3, let (cid:104)· , ·(cid:105) n, be a symmetric bilinear form ofsignature ( n,
1) on R d , and Ω = { [ v ] ∈ P ( R d ) | (cid:104) v, v (cid:105) n, < } be the projective modelof the real hyperbolic space H n . It is a strictly convex open subset of X = P ( R d )(an ellipsoid), and any uniform lattice Γ of PO( n, ⊂ G = PGL( d, R ) divides Ω.(2) For d = n ( n + 1) /
2, let us see R d as the space Sym( n, R ) of symmetric n × n real matrices, and let Ω ⊂ P ( R d ) be the image of the set of positive definite ones.The set Ω is a properly convex open subset of X = P ( R d ); it is strictly convex if andonly if n = 2. The group GL( n, R ) acts on Sym( n, R ) by g · s := gs t g , which inducesan action of PGL( n, R ) on Ω. This action is transitive and the stabilizer of a point isPO( n ), hence Ω identifies with the Riemannian symmetric space PGL( n, R ) / PO( n ).In particular, any uniform lattice Γ of PGL( n, R ) divides Ω. (A similar constructionworks over the complex numbers, the quaternions, or the octonions: see [13].)Many nonsymmetric strictly examples were also constructed since the 1960s byvarious techniques; see [13, 35] for references. Remarkably, there exist irreducibledivisible convex sets Ω ⊂ P ( R d ) which are not symmetric and not strictly convex:the first examples were built by Benoist [12] for 4 ≤ d ≤
7. Ballas–Danciger–Lee [4]generalized Benoist’s construction for d = 4 to show that large families of nonhy-perbolic closed 3-manifolds admit convex projective structures. Choi–Lee–Marquis[36] recently built nonstrictly convex examples of a different flavor for 5 ≤ d ≤ strictly convex Ω, dynamics “at infinity” are relatively well understood: if Γdivides Ω, then Γ is Gromov hyperbolic [10] and, by cocompactness, any orbital mapΓ → Ω extends continuously to an equivariant homeomorphism from the boundary ∂ ∞ Γ of Γ to the boundary of Ω in X . This is similar to Section 3.1, except thatnow X itself is a flag variety G/P (see Table 1). The image of the boundary mapis again a limit set Λ Γ on which the action of Γ is “chaotic”, but Λ Γ is now partof a larger “extended limit set” L Γ , namely the union of all projective hyperplanestangent to Ω at points of Λ Γ . The space X (cid:39) G/P is the disjoint union of L Γ and Ω. The dynamics of Γ on X are further understood by considering the geodesicflow on Ω ⊂ X , defined using the Hilbert metric of Remark 3.1; for Ω = H n as inExample 3.2.(1), this is the usual geodesic flow. Benoist [10] proved that the inducedflow on Γ \ Ω is Anosov and topologically mixing; see [39] for further properties.For nonstrictly convex
Ω, the situation is less understood. Groups Γ dividing Ωare never Gromov hyperbolic [10]; for d = 4 they are relatively hyperbolic [12], butin general they might not be (e.g. if Ω is symmetric), and it is not obvious what typeof boundary ∂ ∞ Γ (defined independently of Ω) might be most useful in the contextof Problem B. The geodesic flow on Γ \ Ω is not Anosov, but Bray [22] proved it isstill topologically mixing for d = 4. Much of the dynamics remains to be explored.By Koszul [84], discrete subgroups of G dividing Ω are structurally stable;moreover, for a closed manifold M with fundamental group Γ = π ( M ), the setHom conv M (Γ , G ) of holonomy representations of convex ( G, X )-structures on M isopen in Hom(Γ , G ). It is also closed in Hom(Γ , G ) as soon as Γ does not contain aninfinite normal abelian subgroup, by Choi–Goldman [33] (for d = 3) and Benoist [11](in general). For d = 3, when M is a closed surface of genus g ≥
2, Goldman [56]showed that Hom conv M (Γ , G ) /G is homeomorphic to R g − , via an explicit parame-trization generalizing classical ( Fenchel–Nielsen ) coordinates on Teichm¨uller space.
EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 9
AdS quasi-Fuchsian representations.
We now discuss the Lorentzian coun-terparts of Example 1.2, which have been studied by Witten [111] and others assimple models for (2 + 1)-dimensional gravity. Let M = S × (0 ,
1) be as in Exam-ple 1.2. Instead of taking (
G, X ) = (PO(3 , , H ), we now take G = PO(2 ,
2) and X = AdS = { [ v ] ∈ P ( R ) | (cid:104) v, v (cid:105) , < } . In other words, we change the signature of the quadratic form defining X from(3 ,
1) (as in Example 3.2.(1)) to (2 , G -invariant metricfrom Riemannian to Lorentzian, and the topology of X from a ball to a solid torus.The space X = AdS is called the anti-de Sitter M = S × (0 ,
1) does not admit (
G, X )-structures of type C (see Sec-tion 1.1), but it admits some of type U, called globally hyperbolic maximal Cauchy-compact (GHMC). In general, a Lorentzian manifold is called globally hyperbolic ifit satisfies the intuitive property that “when moving towards the future one does notcome back to the past”; more precisely, there is a spacelike hypersurface (
Cauchyhypersurface ) meeting each inextendible causal curve exactly once. Here we alsorequire that the Cauchy surface be compact and that M be maximal (i.e. not iso-metrically embeddable into a larger globally hyperbolic Lorentzian 3-manifold).To describe the GHMC ( G, X )-structures on M , it is convenient to consider a dif-ferent model for AdS , which leads to beautiful links with 2-dimensional hyperbolicgeometry. Namely, we view R as the space M ( R ) of real 2 × (cid:104)· , ·(cid:105) , as minus the determinant. This induces an identification of X = AdS with G = PSL(2 , R ) sending [ v ] ∈ X to (cid:2) |(cid:104) v,v (cid:105)| ( v + v v + v v − v − v + v ) (cid:3) ∈ G , anda corresponding group isomorphism from the identity component G = PO(2 , of G acting on X = AdS , to G × G acting on G by right and left multipli-cation: ( g , g ) · g = g gg − . It also induces an identification of the boundary ∂X ⊂ P ( R ) with the projectivization of the set of rank-one matrices, hence with P ( R ) × P ( R ) (by taking the kernel and the image); the action of G on ∂X corresponds to the natural action of G × G on P ( R ) × P ( R ). AdS Ω Λ C Figure 5.
The sets Λ, Ω, C for an AdS quasi-Fuchsian representationWith these identifications, Mess [93] proved that all GHMC ( G, X )-structures on M = S × (0 ,
1) are obtained as follows. Let ( ρ L , ρ R ) be a pair of Fuchsian representa-tions from Γ = π ( M ) (cid:39) π ( S ) to G = PSL(2 , R ). The group ( ρ L , ρ R )(Γ) ⊂ G × G ⊂ G preserves a topological circle Λ in ∂X , namely the graph of the homeomorphismof P ( R ) conjugating the action of ρ L to that of ρ R . For any z ∈ Λ, the orthogonal z ⊥ of z for (cid:104)· , ·(cid:105) , is a projective hyperplane tangent to X at z . The complementΩ in P ( R ) of the union of all z ⊥ for z ∈ Λ is a convex open subset of P ( R ) con-tained in X (see Figure 5) which admits a Γ-invariant Cauchy surface. The action of Γ on Ω via ( ρ L , ρ R ) is properly discontinuous and the convex hull C of Λ in Ω(called the convex core ) has compact quotient by Γ. The quotient ( ρ L , ρ R )(Γ) \ Ω isdiffeomorphic to M = S × (0 , G, X )-structure on M .Such ( G, X )-structures, or their holonomy representations ρ = ( ρ L , ρ R ) : Γ → G × G ⊂ G , are often called AdS quasi-Fuchsian , by analogy with Example 1.2.Their deformation space is parametrized by Teich( S ) × Teich( S ), via ( ρ L , ρ R ) [93].Their geometry, especially the geometry of the convex core and the way it deter-mines ( ρ L , ρ R ), is the object of active current research (see [7, 19]). Generalizationshave recently been worked out in several directions (see [6, 8, 18] and Section 6.2).As in Section 3.1, the compactness of the convex core of an AdS quasi-Fuchsianmanifold implies that any orbital map Γ → Ω extends “at infinity” to an equivariantembedding ξ : ∂ ∞ Γ → ∂X with image Λ. Here ∂X is still a flag variety G/P , where P is the stabilizer in G = PO(2 ,
2) of an isotropic line of R for (cid:104)· , ·(cid:105) , . Although G has higher rank, the rank-one dynamics of Section 3.1 still appear through theproduct structure of G (cid:39) G × G acting on ∂X (cid:39) P ( R ) × P ( R ) (cid:39) ∂ ∞ H × ∂ ∞ H .4. Anosov representations
In this section we define and discuss Anosov representations. These are repre-sentations of Gromov hyperbolic groups into Lie groups G with strong dynamicalproperties, defined using continuous equivariant boundary maps. They were intro-duced by Labourie [85] and further investigated by Guichard–Wienhard [69]. Theyplay an important role in higher Teichm¨uller theory and in the study of Problem B.As we shall see in Section 4.5, most representations that appeared in Section 3 werein fact Anosov representations.4.1. The definition.
Let Γ be a Gromov hyperbolic group with boundary ∂ ∞ Γ(e.g. Γ a surface group and ∂ ∞ Γ a circle, or Γ a nonabelian free group and ∂ ∞ Γ aCantor set). The notion of an Anosov representation of Γ to a reductive Lie group G depends on the choice of a parabolic subgroup P of G up to conjugacy, i.e. on thechoice of a flag variety G/P (see Section 1.4). Here, for simplicity, we restrict to G = PGL( d, R ). We choose an integer i ∈ [1 , d −
1] and denote by P i the stabilizerin G of an i -plane of R d , so that G/P i identifies with the Grassmannian Gr i ( R d ).By definition, a representation ρ : Γ → PGL( d, R ) is P i -Anosov if there exist twocontinuous ρ -equivariant maps ξ i : ∂ ∞ Γ → Gr i ( R d ) and ξ d − i : ∂ ∞ Γ → Gr d − i ( R d )which are transverse (i.e. ξ i ( η ) + ξ d − i ( η (cid:48) ) = R d for all η (cid:54) = η (cid:48) in ∂ ∞ Γ) and satisfy auniform contraction/expansion condition analogous to that defining Anosov flows.Let us state this condition in the original case considered by Labourie [85], whereΓ = π ( M ) for some closed negatively-curved manifold M . We denote by (cid:102) M theuniversal cover of M , by T the unit tangent bundle, and by ( ϕ t ) t ∈ R the geodesicflow on either T ( M ) or T ( (cid:102) M ). Let E ρ = Γ \ ( T ( (cid:102) M ) × R d )be the natural flat vector bundle over T ( M ) = Γ \ T ( (cid:102) M ) associated to ρ , where Γacts on T ( (cid:102) M ) × R d by γ · (˜ x, v ) = ( γ · ˜ x, ρ ( γ ) · v ). The geodesic flow ( ϕ t ) t ∈ R on T ( M )lifts to a flow ( ψ t ) t ∈ R on E ρ , given by ψ t · [(˜ x, v )] = [( ϕ t · ˜ x, v )]. For each ˜ x ∈ T ( (cid:102) M ),the transversality of the boundary maps induces a decomposition R d = ξ i (˜ x + ) ⊕ ξ d − i (˜ x − ), where ˜ x ± = lim t →±∞ ϕ t · ˜ x are the forward and backward endpoints ofthe geodesic defined by ˜ x , and this defines a decomposition of the vector bundle EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 11 E ρ into the direct sum of two subbundles E ρi = { [(˜ x, v )] | v ∈ ξ i (˜ x + ) } and E ρd − i = { [(˜ x, v )] | v ∈ ξ d − i (˜ x − ) } . This decomposition is invariant under the flow ( ψ t ). Bydefinition, the representation ρ is P i -Anosov if the following condition is satisfied. Condition 4.1.
The flow ( ψ t ) t ∈ R uniformly contracts E ρi with respect to E ρd − i , i.e.there exist C, C (cid:48) > such that for any t ≥ , any x ∈ T ( M ) , and any nonzero w i ∈ E ρi ( x ) and w d − i ∈ E ρd − i ( x ) , (cid:107) ψ t · w i (cid:107) ϕ t · x (cid:107) ψ t · w d − i (cid:107) ϕ t · x ≤ e − Ct + C (cid:48) (cid:107) w i (cid:107) x (cid:107) w d − i (cid:107) x , where ( (cid:107) · (cid:107) x ) x ∈ T ( M ) is any fixed continuous family of norms on the fibers E ρ ( x ) . See [23] for an interpretation in terms of metric Anosov flows (or Smale flows).Condition 4.1 implies in particular that the boundary maps ξ i , ξ d − i are dynamics-preserving , in the sense that the image of the attracting fixed point in ∂ ∞ Γ of anyinfinite-order element γ ∈ Γ is an attracting fixed point in Gr i ( R d ) or Gr d − i ( R d ) of ρ ( γ ). Thus ξ i and ξ d − i are unique, by density of such fixed points in ∂ ∞ Γ.We note that P i -Anosov is equivalent to P d − i -Anosov , as the integers i and d − i play a similar role in the definition (up to reversing the flow, which switches contrac-tion and expansion). In particular, we may restrict to P i -Anosov for 1 ≤ i ≤ d/ T ( (cid:102) M ) by ∂ ∞ Γ (2) × R where ∂ ∞ Γ (2) is the space of pairs of distinct points in the Gromovboundary ∂ ∞ Γ of Γ, and the flow ϕ t by translation by t along the R factor. The workof Gromov [62] (see also [31, 92, 95]) yields an appropriate extension of the Γ-actionon ∂ ∞ Γ (2) to ∂ ∞ Γ (2) × R , which is properly discontinuous and cocompact. This leadsto a notion of an Anosov representation for any Gromov hyperbolic group Γ [69].4.2. Important properties and examples.
A fundamental observation moti-vating the study of Anosov representations is the following: if G is a semisimpleLie group of real rank one, then a representation ρ : Γ → G is Anosov if and onlyif it is convex cocompact in the sense of Section 3.1.Moreover, many important properties of convex cocompact representations intorank-one groups generalize to Anosov representations. For instance, Anosov repre-sentations ρ : Γ → G are quasi-isometric embeddings [69, 85]; in particular, theyhave finite kernel and discrete image. Also by [69, 85], any Anosov subgroup (i.e. theimage of any Anosov representation ρ : Γ → G ) is structurally stable; moreover, ρ admits a neighborhood in Hom(Γ , G ) consisting entirely of Anosov representations.This is due to the uniform hyperbolicity nature of the Anosov condition.Kapovich, Leeb, and Porti, in a series of papers (see [73, 77, 67]), have developeda detailed analogy between Anosov representations into higher-rank semisimple Liegroups and convex cocompact representations into rank-one simple groups, from thepoint of view of dynamics (e.g. extending the expansion property at the limit set ofSection 3.1 and other classical characterizations) and topology (e.g. compactifications).Here are some classical examples of Anosov representations in higher real rank. Examples 4.2.
Let Γ = π ( S ) where S is a closed orientable surface of genus ≥ d ≥
2, let τ d : PSL(2 , R ) → G = PGL( d, R ) be theirreducible representation (unique up to conjugation by G ). For any Fuchsianrepresentation ρ : Γ → PSL(2 , R ), the composition τ d ◦ ρ : Γ → G is P i -Anosov for all 1 ≤ i ≤ d −
1. Moreover, any representation in the connected component of τ d ◦ ρ in Hom(Γ , G ) is still P i -Anosov for all 1 ≤ i ≤ d −
1. These representationswere first studied by Hitchin [71] and are now known as
Hitchin representations .(2) (Burger–Iozzi–Labourie–Wienhard [25, 28]) If a representation of Γ into G =PSp(2 n, R ) ⊂ PGL(2 n, R ) (resp. G = PO(2 , q ) ⊂ PGL(2 + q, R )) is maximal , thenit is P n -Anosov (resp. P -Anosov).(3) (Barbot [5] for d = 3) Let d ≥
2. Any Fuchsian representation Γ → SL(2 , R ),composed with the standard embedding SL(2 , R ) (cid:44) → SL( d, R ) (given by the directsum of the standard action on R and the trivial action on R d − ), defines a P -Anosov representation Γ → G = PSL( d, R ).In (2), we say that ρ : Γ → G is maximal if it maximizes a topological invariant,the Toledo number T ( ρ ), defined for any simple G of Hermitian type. If X = G/K is the Riemannian symmetric space of G , then the imaginary part of the G -invariantHermitian form on X defines a real 2-form ω X , and by definition T ( ρ ) = π (cid:82) S f ∗ ω X where f : (cid:101) S → X is any ρ -equivariant smooth map. For G = PSL(2 , R ), thiscoincides with the Euler number of ρ . In general, T ( ρ ) takes discrete values and | T ( ρ ) | ≤ rank R ( G ) | χ ( S ) | where χ ( S ) is the Euler characteristic of S (see [27]).While (1) and (3) provide Anosov representations in two of the three connectedcomponents of Hom(Γ , PSL(3 , R )) for Γ = π ( S ), it is currently not known whetherAnosov representations appear in the third component.See [9, 24, 30, 69, 75] for higher-rank Anosov generalizations of Schottky groups.4.3. Higher Teichm¨uller spaces of Anosov representations.
Anosov repre-sentations play an important role in higher Teichm¨uller theory , a currently veryactive theory whose goal is to find deformation spaces of faithful and discrete rep-resentations of discrete groups Γ into higher-rank semisimple Lie groups G whichshare some of the remarkable properties of Teichm¨uller space. Although variousgroups Γ may be considered, the foundational case is when Γ = π ( S ) for someclosed connected surface S of genus ≥ S ) have been found for two types of higherTeichm¨uller spaces : the space of Hitchin representations of Γ into a real split simpleLie group G such as PGL( d, R ), modulo conjugation by G ; and the space of maximalrepresentations of Γ into a simple Lie group G of Hermitian type such as PSp(2 n, R )or PO(2 , q ), modulo conjugation by G . Both these spaces are unions of connectedcomponents of Hom(Γ , G ) /G , consisting entirely of Anosov representations (seeExamples 4.2.(1)–(2)). Similarities of these spaces to Teich( S ) include:(1) the proper discontinuity of the action of the mapping class group Mod( S ) [86];(2) for Hitchin representations to PGL( d, R ): the topology of R ( d − | χ ( S ) | [71];(3) good systems of coordinates generalizing those on Teich( S ) [17, 52, 56, 102, 112];(4) an analytic Mod( S )-invariant Riemannian metric ( pressure metric ) [23, 97];(5) versions of the collar lemma for associated locally symmetric spaces [29, 89].Other higher Teichm¨uller spaces of Anosov representations of π ( S ) are also beingexplored [70]. We refer to Section 5 for geometric structures associated to such spaces.4.4. Characterizations.
Various characterizations of Anosov representations havebeen developed in the past few years, by Labourie [85], Guichard–Wienhard [69],
EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 13
Kapovich–Leeb–Porti [73, 74, 75], Gu´eritaud–Guichard–Kassel–Wienhard [63], andothers. Here are some characterizations that do not involve any flow. They holdfor any reductive Lie group G , but for simplicity we state them for G = PGL( d, R ).For 1 ≤ i ≤ d and g ∈ GL( d, R ), we denote by µ i ( g ) (resp. λ i ( g )) the logarithm ofthe i -th singular value (resp. eigenvalue) of g . Theorem 4.3.
For a Gromov hyperbolic group Γ , a representation ρ : Γ → G =PGL( d, R ) , and an integer ≤ i ≤ d/ , the following are equivalent:(1) ρ is P i -Anosov (or equivalently P d − i -Anosov, see Section 4.1) ;(2) there exist continuous, ρ -equivariant, transverse, dynamics-preserving bound-ary maps ξ i : ∂ ∞ Γ → Gr i ( R d ) and ξ d − i : ∂ ∞ Γ → Gr d − i ( R d ) , and ( µ i − µ i +1 )( ρ ( γ )) → + ∞ as | γ | → + ∞ ;(3) there exist continuous, ρ -equivariant, transverse, dynamics-preserving bound-ary maps ξ i : ∂ ∞ Γ → Gr i ( R d ) and ξ d − i : ∂ ∞ Γ → Gr d − i ( R d ) , and ( λ i − λ i +1 )( ρ ( γ )) → + ∞ as (cid:96) Γ ( γ ) → + ∞ ;(4) there exist C, C (cid:48) > such that ( µ i − µ i +1 )( ρ ( γ )) ≥ C | γ | − C (cid:48) for all γ ∈ Γ ;(5) there exist C, C (cid:48) > such that ( λ i − λ i +1 )( ρ ( γ )) ≥ C (cid:96) Γ ( γ ) − C (cid:48) for all γ ∈ Γ . Here we denote by | · | : Γ → N the word length with respect to some fixed finitegenerating subset of Γ, and by (cid:96) Γ : Γ → N the translation length in the Cayleygraph of Γ for that subset, i.e. (cid:96) Γ ( γ ) = min β ∈ Γ | βγβ − | . In a Gromov hyperbolicgroup Γ the translation length (cid:96) Γ ( γ ) is known to differ only by at most a uniformadditive constant from the stable length | γ | ∞ = lim n → + ∞ | γ n | /n , and so we mayreplace (cid:96) Γ ( γ ) by | γ | ∞ in Conditions (3) and (5).The equivalence (1) ⇔ (2) is proved in [63] and [74], the equivalence (2) ⇔ (3) in[63], the equivalence (1) ⇔ (4) in [75] and [15], and the equivalence (4) ⇔ (5) in [79].Condition (4) is a refinement of the condition of being a quasi-isometric embed-ding, which for G = PGL( d, R ) is equivalent to the existence of C, C (cid:48) > (cid:112)(cid:80) k ( µ k − µ k +1 ) ( ρ ( γ )) ≥ C | γ |− C (cid:48) for all γ ∈ Γ. We refer to [63] (
CLI condition )or [75] (
Morse condition ) for further refinements satisfied by Anosov representations.By [75, 15], if Γ is any finitely generated group, then the existence of a repre-sentation ρ : Γ → PGL( d, R ) satisfying Condition (4) implies that Γ is Gromovhyperbolic. The analogue for (5) is more subtle: e.g. the Baumslag–Solitar groupBS(1 , , R ) satisfying Condition (5) for the stable length | · | ∞ , see [79].Kapovich–Leeb–Porti’s original proof [75] of (1) ⇔ (4) uses the geometry ofhigher-rank Riemannian symmetric spaces and asymptotic cones. Bochi–Potrie–Sambarino’s alternative proof [15] is based on an interpretation of (1) and (4) interms of partially hyperbolic dynamics, and more specifically of dominated split-tings for locally constant linear cocycles over certain subshifts. Pursuing this pointof view further, [79] shows that the equivalence (4) ⇔ (5) of Theorem 4.3 impliesthe equivalence between nonuniform hyperbolicity (i.e. all invariant measures arehyperbolic) and uniform hyperbolicity for a certain cocycle naturally associatedwith ρ on the space of biinfinite geodesics of Γ. In general in smooth dynamics,nonuniform hyperbolicity does not imply uniform hyperbolicity.4.5. Revisiting the examples of Section 3.
The boundary maps and dynamics“at infinity” that appeared in most examples of Section 3 are in fact explained bythe notion of an Anosov representation: • convex cocompact representations into rank-one simple Lie groups as in Sec-tion 3.1 are all Anosov (see Section 4.2); • if S is a closed orientable connected surface of genus ≥
2, then by Choi–Goldman [33, 56] the holonomy representations of convex projective structureson S as in Section 3.2 are exactly the Hitchin representations of π ( S ) intoPSL(3 , R ); they are all P -Anosov (Example 4.2.(1)); • for general d ≥
3, Benoist’s work [10] shows that if Γ is a discrete subgroupof PGL( d, R ) dividing a strictly convex open subset Ω of P ( R d ), then Γ isGromov hyperbolic and the inclusion Γ (cid:44) → PGL( d, R ) is P -Anosov; • Mess’s theory [93] implies that a representation ρ : π ( S ) → PO(2 , ⊂ PGL(4 , R ) is AdS quasi-Fuchsian if and only if it is P -Anosov.5. Geometric structures for Anosov representations
We just saw in Section 4.5 that various (
G, X )-structures described in Section 3give rise (via the holonomy) to Anosov representations; these (
G, X )-structures areof type C or type U (terminology of Section 1.1). In this section, we study the con-verse direction. Namely, given an Anosov representation ρ : Γ → G , we wish to find: • homogeneous spaces X = G/H on which Γ acts properly discontinuously via ρ ;this will yield ( G, X )-manifolds (or orbifolds) M = ρ (Γ) \ X of type C; • proper open subsets U ( domains of discontinuity ) of homogeneous spaces X = G/H on which Γ acts properly discontinuously via ρ ; this will yield ( G, X )-manifolds (or orbifolds) M = ρ (Γ) \U of type U.We discuss type U in Sections 5.1–5.2 and type C in Section 5.3. One motivationis to give a geometric meaning to the higher Teichm¨uller spaces of Section 4.3.5.1. Cocompact domains of discontinuity.
Domains of discontinuity with com-pact quotient have been constructed in several settings in the past ten years.Barbot [5] constructed such domains in the space X of flags of R , for the Anosovrepresentations to G = PSL(3 , R ) of Example 4.2.(3) and their small deformations.Guichard–Wienhard [69] developed a more general construction of cocompact do-mains of discontinuity in flag varieties X for Anosov representations into semisimpleLie groups G . Here is one of their main results. For p ≥ q ≥ i ≥
1, we denote by F p,qi the closed subspace of the Grassmannian Gr i ( R p + q ) consisting of i -planes that aretotally isotropic for the standard symmetric bilinear form (cid:104)· , ·(cid:105) p,q of signature ( p, q ). Theorem 5.1 (Guichard–Wienhard [69]) . Let G = PO( p, q ) with p ≥ q and X = F p,qq . For any P -Anosov representation ρ : Γ → G ⊂ PGL( R p + q ) with boundarymap ξ : ∂ ∞ Γ → F p,q ⊂ P ( R p + q ) , the group ρ (Γ) acts properly discontinuously withcompact quotient on U ρ := X (cid:114) L ρ , where L ρ := (cid:91) η ∈ ∂ ∞ Γ { W ∈ X = F p,qq | ξ ( η ) ∈ W } . We have U ρ (cid:54) = ∅ as soon as dim( ∂ ∞ Γ) < p − . The homotopy type of ρ (Γ) \U ρ isconstant as ρ varies continuously among P -Anosov representations of Γ in G . For q = 1, we recover the familiar picture of Section 3.1: the set L ρ is the limitset Λ ρ (Γ) ⊂ ∂ ∞ H p , and U ρ is the domain of discontinuity Ω ρ (Γ) = ∂ ∞ H p (cid:114) Λ ρ (Γ) .Guichard–Wienhard [69] used Theorem 5.1 to describe domains of discontinuityfor various families of Anosov representations into other semisimple Lie groups G . EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 15
Indeed, by [69] an Anosov representation ρ : Γ → G can always be composed witha representation of G into some PO( p, q ) so as to become P -Anosov in PO( p, q ).Kapovich–Leeb–Porti [76] developed a more systematic approach to the construc-tion of domains of discontinuity in flag varieties. They provided sufficient conditions(expressed in terms of a notion of balanced ideal in the Weyl group for the Bruhatorder) on triples ( G, P, Q ) consisting of a semisimple Lie group G and two parabolicsubgroups P and Q , so that P -Anosov representations into G admit cocompact do-mains of discontinuity in G/Q . These domains are obtained by removing an explicit“extended limit set” L ρ as in Theorem 5.1. The approach of Kapovich–Leeb–Portiis intrinsic: it does not rely on an embedding of G into some PO( p, q ).5.2. Geometric structures for Hitchin and maximal representations.
Let S be a closed orientable surface of genus ≥
2. Recall (Example 4.2) that Hitchin rep-resentations from Γ = π ( S ) to G = PSL( d, R ) are P i -Anosov for all 1 ≤ i ≤ d − G = PO(2 , q ) ⊂ PGL(2 + q, R ) are P -Anosov.For G = PSL(2 , R ) (cid:39) PO(2 , , Hitchin representations and maximal represen-tations of Γ to G both coincide with the Fuchsian representations; they are theholonomy representations of hyperbolic structures on S (Example 2.1). In the set-ting of higher Teichm¨uller theory (see Section 4.3), one could hope that Hitchinor maximal representations of Γ to higher-rank G might also parametrize certaingeometric structures on a manifold related to S . We saw in Section 4.5 that thisis indeed the case for G = PSL(3 , R ): Hitchin representations of Γ to PSL(3 , R )parametrize the convex projective structures on S [33, 56]. In an attempt to gen-eralize this picture, we now outline constructions of domains of discontinuity forHitchin representations to PSL( d, R ) with d >
3, and maximal representations toPO(2 , q ) with q >
1. By classical considerations of cohomological dimension, suchdomains cannot be both cocompact and contractible; in Sections 5.2.2–5.2.3 below,we will prefer to forgo compactness to favor the nice geometry of convex domains.5.2.1.
Hitchin representations for even d = 2 n . Let (
G, X ) = (PSL(2 n, R ) , P ( R n )).Hitchin representations to G do not preserve any properly convex open set Ω in X ,see [45, 113]. However, Guichard–Wienhard associated to them nonconvex ( G, X )-structures on a closed manifold: by [69], if ρ : Γ → G is Hitchin with boundary map ξ n : ∂ ∞ Γ → Gr n ( R n ), then X (cid:114) (cid:83) η ∈ ∂ ∞ Γ ξ n ( η ) is a cocompact domain of disconti-nuity for ρ , and the homotopy type of the quotient does not depend on ρ . So far thetopology and geometry of the quotient are understood only for n = 2 (see [110]).5.2.2. Hitchin representations for odd d = 2 n + 1 . Hitchin representations to G =PSL(2 n + 1 , R ) give rise to ( G, X )-manifolds for at least two choices of X .One choice is to take X to be the space of partial flags ( V ⊂ V n ) of R n +1 with V a line and V n a hyperplane: Guichard–Wienhard [69] again constructed explicitcocompact domains of discontinuity in X in this setting.Another choice is X = P ( R n +1 ): Hitchin representations in odd dimension arethe holonomies of convex projective manifolds, which are noncompact for n > Theorem 5.2 ([45, 113]) . For any Hitchin representation ρ : Γ → PSL(2 n + 1 , R ) ,there is a ρ (Γ) -invariant properly convex open subset Ω of P ( R n +1 ) and a nonemptyclosed convex subset C of Ω which has compact quotient by ρ (Γ) . More precisely, if ρ has boundary maps ξ : ∂ ∞ Γ → Gr ( R n +1 ) = P ( R n +1 ) and ξ n : ∂ ∞ Γ → Gr n ( R n +1 ), we may take Ω = P ( R n +1 ) (cid:114) (cid:83) η ∈ ∂ ∞ Γ ξ n ( η ) and C to be the convex hull of ξ ( ∂ ∞ Γ) in Ω. The group ρ (Γ) acts properly discontinuously on Ω(Remark 3.1), and so ρ (Γ) \ Ω is a convex projective manifold, with a compact convexcore ρ (Γ) \C . In other words, ρ (Γ) is convex cocompact in P ( R n +1 ), see Section 6.5.2.3. Maximal representations.
Maximal representations to G = PO(2 , q ) give riseto ( G, X )-manifolds for at least two choices of X .One choice is X = F ,q (also known as the space of photons in the Einstein uni-verse
Ein q ): Theorem 5.1 provides cocompact domains of discontinuity for ρ in X .Collier–Tholozan–Toulisse [37] recently studied the geometry of the associated quo-tient ( G, X )-manifolds, and showed that they fiber over S with fiber O( q ) / O( q − X = P ( R q ): by [44], maximal representations ρ : Γ → G are the holonomy representations of convex projective manifolds ρ (Γ) \ Ω, which arenoncompact for q > H ,q − = { [ v ] ∈ P ( R q ) | (cid:104) v, v (cid:105) ,q < } (see [44, 37]), which isa pseudo-Riemannian analogue of the real hyperbolic space in signature (2 , q − ρ (Γ) is H ,q − -convex cocompact in the sense of Section 6.2 below.5.3. Proper actions on full homogeneous spaces.
In Sections 5.1–5.2, wemainly considered compact spaces X = G/H (flag varieties); these spaces can-not admit proper actions by infinite discrete groups, but we saw that sometimesthey can contain domains of discontinuity U (cid:40) X , yielding ( G, X )-manifolds oftype U (terminology of Section 1.1).We now consider noncompact X = G/H . Then Anosov representations ρ : Γ → G may give proper actions of Γ on the whole of X = G/H , yielding (
G, X )-manifolds ρ (Γ) \ X of type C. When H is compact, this is not very interesting since all faithfuland discrete representations to G give proper actions on X . However, when H isnoncompact, it may be remarkably difficult in general to find such representationsgiving proper actions on X , which led to a rich literature (see [83] and [78, Intro]).One construction for proper actions on X was initiated by Guichard–Wienhard [69]and developed further in [64]. Starting from an Anosov representation ρ : Γ → G ,the idea is to embed G into some larger semisimple Lie group G (cid:48) so that X = G/H identifies with a G -orbit in some flag variety F (cid:48) of G (cid:48) , and then to find a cocompactdomain of discontinuity U ⊃ X for ρ in F (cid:48) by using a variant of Theorem 5.1. Theaction of ρ (Γ) on X is then properly discontinuous, and ρ (Γ) \ ( U ∩ X ) provides acompactification of ρ (Γ) \ X , which in many cases can be shown to be well-behaved.Here is one application of this construction; see [64] for other examples. Example 5.3 ([64]) . Let G = PO( p, q ) and H = O( p, q −
1) where p > q ≥ P q -Anosov representation ρ : Γ → G ⊂ PGL( p + q, R ), the group ρ (Γ) actsproperly discontinuously on X = H p,q − = { [ v ] ∈ P ( R p + q ) | (cid:104) v, v (cid:105) p,q < } (cid:39) G/H ,and for torsion-free Γ the complete (
G, X )-manifold ρ (Γ) \ X is topologically tame.By topologically tame we mean homeomorphic to the interior of a compact mani-fold with boundary. For other compactifications of quotients of homogeneous spacesby Anosov representations, yielding topological tameness, see [68, 72, 76].Another construction of complete ( G, X )-manifolds for Anosov representationsinto reductive Lie groups G was given in [63], based on a properness criterion ofBenoist [9] and Kobayashi [82]. For simplicity we discuss it for G = PGL( d, R ).As in Section 4.4, let µ i ( g ) be the logarithm of the i -th singular value of a matrix g ∈ GL( d, R ); this defines a map µ = ( µ , . . . , µ d ) : PGL( d, R ) → R d / R (1 , . . . , (cid:39) EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 17 R d − . The properness criterion of [9, 82] states that for two closed subgroups H, Γof G = PGL( d, R ), the action of Γ on G/H is properly discontinuous if and only ifthe set µ (Γ) “drifts away at infinity from µ ( H )”, in the sense that for any R > d R d − ( µ ( γ ) , µ ( H )) ≥ R for all but finitely many γ ∈ Γ. If Γ is the image of anAnosov representation, then we can apply the implication (1) ⇒ (2) of Theorem 4.3to see that the properness criterion is satisfied for many examples of H . Example 5.4 ([63]) . For i = 1 (resp. n ), the image of any P i -Anosov representationto G = PSL(2 n, R ) acts properly discontinuously on X = G/H for H = SL( n, C )(resp. SO( n + 1 , n − Convex cocompact projective structures
In Sections 3 and 4.5 we started from (
G, X )-structures to produce Anosov rep-resentations, and in Section 5 we started from Anosov representations to produce(
G, X )-structures. We now discuss a situation, in the setting of convex projectivegeometry, in which the links between (
G, X )-structures and Anosov representationsare particularly tight and go in both directions, yielding a better understanding ofboth sides. In Section 6.4 we will also encounter generalizations of Anosov represen-tations, for finitely generated groups that are not necessarily Gromov hyperbolic.6.1.
Convex cocompactness in higher real rank.
The results presented hereare part of a quest to generalize the notion of rank-one convex cocompactness ofSection 3.1 to higher real rank.The most natural generalization, in the setting of Riemannian symmetric spaces,turns out to be rather restrictive: Kleiner–Leeb [80] and Quint [98] proved that if G is a real simple Lie group of real rank ≥ K a maximal compact subgroupof G , then any Zariski-dense discrete subgroup of G acting with compact quotienton some nonempty convex subset of G/K is a uniform lattice in G .Meanwhile, we have seen in Section 4.2 that Anosov representations into higher-rank semisimple Lie groups G have strong dynamical properties which nicely gen-eralize those of rank-one convex cocompact representations (see [77, 67]). However,in general Anosov representations to G do not act with compact quotient on anynonempty convex subset of G/K , and it is not clear that Anosov representationsshould come with any geometric notion of convexity at all (see e.g. Section 5.2.1).In this section, we shall see that Anosov representations in fact do come withconvex structures. We shall introduce several generalizations of convex cocompact-ness to higher real rank (which we glimpsed in Sections 5.2.2–5.2.3) and relate themto Anosov representations. This is joint work with J. Danciger and F. Gu´eritaud.6.2.
Convex cocompactness in pseudo-Riemannian hyperbolic spaces.
Westart with a generalization of the hyperbolic quasi-Fuchsian manifolds of Exam-ple 1.2 or the AdS quasi-Fuchsian manifolds of Section 3.3, where we replace thereal hyperbolic space H or its Lorentzian analogue AdS by their general pseudo-Riemannian analogue in signature ( p, q −
1) for p, q ≥
1, namely X = H p,q − = (cid:8) [ v ] ∈ P ( R p + q ) | (cid:104) v, v (cid:105) p,q < (cid:9) . The symmetric bilinear form (cid:104)· , ·(cid:105) p,q of signature ( p, q ) induces a pseudo-Riemannianstructure of signature ( p, q −
1) on X , with isometry group G = PO( p, q ) andconstant negative sectional curvature (see [44, § Definition 6.1.
A discrete subgroup Γ of G = PO( p, q ) is H p,q − -convex cocompact if it preserves a properly convex open subset Ω of X = H p,q − ⊂ P ( R p + q ) and ifit acts with compact quotient on some closed convex subset C of Ω with nonemptyinterior, whose ideal boundary ∂ i C := C (cid:114) C = C∩ ∂X does not contain any nontrivialprojective segment. A representation ρ : Γ → G is H p,q − -convex cocompact if itskernel is finite and its image is an H p,q − -convex cocompact subgroup of G .Here C is the closure of C in P ( R p + q ) and ∂X the boundary of X = H p,q − in P ( R p + q ). For Γ , Ω , C as in Definition 6.1, the quotient Γ \ Ω is a (
G, X )-manifold (ororbifold) (see Remark 3.1), which we shall call convex cocompact ; the subset Γ \C is compact, convex, and contains all the topology, as in Sections 3.1 and 3.3.There is a rich world of examples of convex cocompact ( G, X )-manifolds, includ-ing direct generalizations of the quasi-Fuchsian manifolds of Sections 3.1 and 3.3(see [8, 44, 45]) but also more exotic examples where the fundamental group is notnecessarily realizable as a discrete subgroup of PO( p,
1) (see [44, 88]).The following result provides links with Anosov representations.
Theorem 6.2 ([44, 45]) . For p, q ≥ , let Γ be an infinite discrete group and ρ : Γ → G = PO( p, q ) ⊂ PGL( p + q, R ) a representation.(1) If ρ is H p,q − -convex cocompact, then Γ is Gromov hyperbolic and ρ is P -Anosov.(2) Conversely, if Γ is Gromov hyperbolic, if ρ is P -Anosov, and if ρ (Γ) pre-serves a properly convex open subset of P ( R p + q ) , then ρ is H p,q − -convexcocompact or H q,p − -convex cocompact.(3) If Γ is Gromov hyperbolic with connected boundary ∂ ∞ Γ and if ρ is P -Anosov, then ρ is H p,q − -convex cocompact or H q,p − -convex cocompact. In (2)–(3), “ ρ is H q,p − -convex cocompact” is understood after identifying PO( p, q )with PO( q, p ) and P ( R p,q ) (cid:114)H p,q − with H q,p − under multiplication of (cid:104)· , ·(cid:105) p,q by − q = 2 and Γ is a uniform lattice of PO( p,
1) is due to Barbot–M´erigot [8].The links between H p,q − -convex cocompactness and Anosov representations inTheorem 6.2 have several applications.Applications to ( G, X )-structures (see [44, 45]): • H p,q − -convex cocompactness is stable under small deformations, because beingAnosov is; thus the set of holonomy representations of convex cocompact ( G, X )-structures on a given manifold M is open in Hom( π ( M ) , G ). • Examples of convex cocompact (
G, X )-manifolds can be obtained using classicalfamilies of Anosov representations: e.g. Hitchin representations into PO( n +1 , n ) are H n +1 ,n − -convex cocompact for odd n and H n,n -convex cocompact for even n , andHitchin representations into PO( n + 1 , n + 1) are H n +1 ,n -convex cocompact. Max-imal representations into PO(2 , q ) are H ,q − -convex cocompact, see Section 5.2.3.Applications to Anosov representations: • New examples of Anosov representations can be constructed from convex cocom-pact (
G, X )-manifolds: e.g. this approach is used in [44] to prove that any Gromovhyperbolic right-angled Coxeter group in d generators admits P -Anosov represen-tations into PGL( d, R ). This provides a large new class of hyperbolic groups admit-ting Anosov representations; these groups can have arbitrary large cohomologicaldimension and exotic boundaries (see [44, § • For q = 2 and Γ a uniform lattice of PO( p, , Barbot [6] used convex cocompact EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 19 ( G, X )-structures to prove that the connected component T of Hom(Γ , PO( p, (cid:44) → PO( p, (cid:44) → PO( p,
2) consists entirely ofAnosov representations. This is interesting in the framework of Section 4.3.6.3.
Strong projective convex cocompactness.
We now consider a broader no-tion of convex cocompactness, not involving any quadratic form. Let d ≥ Definition 6.3.
A discrete subgroup Γ of G = PGL( d, R ) is strongly P ( R d ) -convexcocompact if it preserves a strictly convex open subset Ω of X = P ( R d ) with C boundary and if it acts with compact quotient on some nonempty closed convexsubset C of Ω. A representation ρ : Γ → G is strongly P ( R d ) -convex cocompact if itskernel is finite and its image is a strongly P ( R d )-convex cocompact subgroup of G .The action of Γ on Ω in Definition 6.3 is a special case of a class of geometricallyfinite actions introduced by Crampon–Marquis [40]. We use the adverb “strongly”to emphasize the strong regularity assumptions made on Ω. In Definition 6.3 wesay that the quotient Γ \ Ω is a strongly convex cocompact projective manifold (ororbifold); the subset Γ \C is again compact, convex, and contains all the topology.Strongly P ( R d )-convex cocompact representations include H p,q − -convex cocom-pact representations as in Section 6.2 (see [44]), and the natural inclusion of groupsdividing strictly convex open subsets of P ( R d ) as in Section 3.2. The followingresult generalizes Theorem 6.2, and improves on earlier results of [10, 40]. Theorem 6.4 ([45]) . Let Γ be an infinite discrete group and ρ : Γ → G =PGL( d, R ) a representation such that ρ (Γ) preserves a nonempty properly convexopen subset of X = P ( R d ) . Then ρ is strongly P ( R d ) -convex cocompact if and onlyif Γ is Gromov hyperbolic and ρ is P -Anosov. Another generalization of Theorem 6.2 was independently obtained by Zimmer[113]: it is similar to Theorem 6.4, but involves a slightly different notion of convexcocompactness and assumes ρ (Γ) to act irreducibly on P ( R d ).Applications of Theorem 6.4 include: • Examples of strongly convex cocompact projective manifolds using classical Anosovrepresentations (e.g. Hitchin representations into PSL(2 n +1 , R ) as in Section 5.2.2). • In certain cases, a better understanding of the set of Anosov representations of aGromov hyperbolic group Γ inside a given connected component of Hom(Γ , G ): e.g.for an irreducible hyperbolic right-angled Coxeter group Γ, it is proved in [46], usingTheorem 6.4 and Vinberg’s theory [109], that P -Anosov representations form thefull interior of the space of faithful and discrete representations of Γ as a reflectiongroup in G = PGL( d, R ).For a Gromov hyperbolic group Γ and a P -Anosov representation ρ : Γ → G =PGL( d, R ), the group ρ (Γ) does not always preserve a properly convex open subsetof X = P ( R d ): see Section 5.2.1. However, as observed in [113], ρ can always becomposed with the embedding ι : G (cid:44) → PGL( V ) described in Example 3.2.(2), for V = Sym( d, R ) (cid:39) R d ( d +1) ; then ι ◦ ρ (Γ) preserves a properly convex open subsetin P ( V ). The composition ι ◦ ρ : Γ → PGL( V ) is still P -Anosov by [69], and it isstrongly P ( V )-convex cocompact by Theorem 6.4. More generally, using [69], anyAnosov representation to any semisimple Lie group can always be composed withan embedding into some PGL( V ) so as to become strongly P ( V )-convex cocompact. Projective convex cocompactness in general.
We now introduce an evenbroader notion of convex cocompactness, where we remove the strong regularity as-sumptions on Ω in Definition 6.3. This yields a large class of convex projective man-ifolds, whose fundamental groups are not necessarily Gromov hyperbolic. Their ho-lonomy representations are generalizations of Anosov representations, sharing someof their desirable properties (Theorem 6.7). This shows that Anosov representationsare not the only way to successfully generalize rank-one convex cocompactness tohigher real rank.
Definition 6.5 ([45]) . A discrete subgroup Γ of G = PGL( d, R ) is P ( R d ) -convexcocompact if it preserves a properly convex open subset Ω of X = P ( R d ) and if itacts with compact quotient on some “large enough” closed convex subset C of Ω.A representation ρ : Γ → G is P ( R d ) -convex cocompact if its kernel is finite and itsimage is a P ( R d )-convex cocompact subgroup of G .In Definition 6.5, by “ C large enough” we mean that all accumulation points ofall Γ-orbits of Ω are contained in the boundary of C in X = P ( R d ). If we did notimpose this (even if we asked C to have nonempty interior), then the notion of P ( R d )-convex cocompactness would not be stable under small deformations: see [45, 47].In Definition 6.5 we call Γ \ Ω a convex cocompact projective manifold (or orbifold).The class of P ( R d )-convex cocompact representations includes all strongly P ( R d )-convex cocompact representations as in Section 6.3, hence all H p,q − -convex cocom-pact representations as in Section 6.2. In fact, the following holds. Proposition 6.6 ([45]) . Let Γ be an infinite discrete group. A representation ρ : Γ → G = PGL( d, R ) is strongly P ( R d ) -convex cocompact (Definition 6.3) if andonly if it is P ( R d ) -convex cocompact (Definition 6.5) and Γ is Gromov hyperbolic. This generalizes a result of Benoist [10] on divisible convex sets. Together withTheorem 6.4, Proposition 6.6 shows that P ( R d )-convex cocompact representationsare generalizations of Anosov representations, to a larger class of finitely generatedgroups Γ which are not necessarily Gromov hyperbolic. These representations stillenjoy the following good properties. Theorem 6.7 ([45]) . Let Γ be an infinite discrete group and ρ : Γ → G =PGL( d, R ) a P ( R d ) -convex cocompact representation. Then(1) ρ is a quasi-isometric embedding;(2) there is a neighborhood of ρ in Hom(Γ , G ) consisting entirely of faithful anddiscrete P ( R d ) -convex cocompact representations;(3) ρ is P (( R d ) ∗ ) -convex cocompact;(4) ρ induces a P ( R D ) -convex cocompact representation for any D ≥ d (bylifting ρ to a representation to SL ± ( d, R ) and composing it with the naturalinclusion SL ± ( d, R ) (cid:44) → SL ± ( D, R ) ). In order to prove (2), we show that the representations of Theorem 6.7 are exactlythe holonomy representations of compact convex projective manifolds with strictlyconvex boundary [45]; we can then apply the deformation theory of [38].Groups that are P ( R d )-convex cocompact but not strongly P ( R d )-convex cocom-pact include all groups dividing a properly convex, but not strictly convex, opensubset of X = P ( R d ) as in Section 3.2, as well as their small deformations inPGL( D, R ) for D ≥ d (Theorem 6.7.(2)–(4)). Such nontrivial deformations exist:e.g. for d = 4 we can always bend along tori or Klein bottle subgroups [12]. There EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 21 seems to be a rich world of examples beyond this, which is just starting to be un-veiled, see [45, 46, 47]. It would be interesting to understand the precise natureof the corresponding abstract groups Γ, and how the dynamics of P ( R d )-convexcocompact representations generalize that of Anosov representations.7. Complete affine structures
In Sections 3 to 6 we always considered semisimple, or more generally reductive,Lie groups G . We now discuss links between ( G, X )-structures and representationsof discrete groups into G in an important case where G is not reductive: namely G isthe group Aff( R d ) = GL( d, R ) (cid:110) R d of invertible affine transformations of X = R d .We shall see in Section 7.3 that for d = 3 the holonomy representations of cer-tain complete (i.e. type C in Section 1.1) ( G, X )-structures are characterized by auniform contraction condition, which is also an affine Anosov condition; we shallbriefly mention partial extensions to d >
3, which are currently being explored.7.1.
Brief overview: understanding complete affine manifolds.
Let (
G, X ) =(Aff( R d ) , R d ). This section is centered around complete affine manifolds , i.e. ( G, X )-manifolds of the form M = ρ (Γ) \ X where Γ (cid:39) π ( M ) is a discrete group and ρ : Γ → G a faithful representation through which Γ acts properly discontinuouslyand freely on X = R d . The study of such representations has a rich history throughthe interpretation of their images as affine crystallographic groups , i.e. symmetrygroups of affine tilings of R d , possibly with noncompact tiles; see [1] for a detailedsurvey. The compact and noncompact cases are quite different.For a compact complete affine manifold M , Auslander [3] conjectured that π ( M )must always be virtually (i.e. up to finite index) polycyclic. This extends a classicaltheorem of Bieberbach on affine Euclidean isometries. The conjecture is proved for d ≤ d ≥
7, despite partial results (see [1]).In contrast, in answer to a question of Milnor [90], there exist noncompact com-plete affine manifolds M for which π ( M ) is not virtually polycyclic. The first ex-amples were constructed by Margulis [90] for d = 3, with π ( M ) a nonabelian freegroup. In these examples the holonomy representation takes values in O(2 , (cid:110) R (this is always the case when π ( M ) is not virtually polycyclic [53]), hence M in-herits a flat Lorentzian structure. Such manifolds are called Margulis spacetimes .They have a rich geometry and have been much studied since the 1990s, most promi-nently by Charette, Drumm, Goldman, Labourie, and Margulis. In particular, thequestions of the topological tameness of Margulis spacetimes and of the existence ofnice fundamental domains in X = R (bounded by piecewise linear objects called crooked planes ) have received much attention: see e.g. [32, 34, 41, 42, 49, 50]. Seealso [2, 60, 101] for higher-dimensional analogues M with π ( M ) a free group.Following [41, 42] (see [100]), a convenient point of view for understanding Mar-gulis spacetimes is to regard them as “infinitesimal analogues” of complete AdS ma-nifolds. In order to describe this point of view, we first briefly discuss the AdS case.7.2. Complete AdS manifolds.
As in Section 3.3, let (
G, X ) = (PO(2 , , AdS ),and view X as the group G = PSL(2 , R ) and the identity component G of G as G × G acting on X (cid:39) G by right and left multiplication. We consider ( G, X )-manifolds of the form M = ρ (Γ) \ X where Γ (cid:39) π ( M ) is an infinite discrete groupand ρ = ( ρ L , ρ R ) : Γ → G × G ⊂ G a faithful representation through which Γ actsproperly discontinuously and freely on X . Not all faithful and discrete ρ = ( ρ L , ρ R ) yield properly discontinuous actions on X : e.g. if ρ L = ρ R , then ρ has a global fixedpoint, precluding properness. However, the following properness criteria hold. Wedenote by λ ( g ) := inf x ∈ H d H ( x, g · x ) ≥ g ∈ G in H . Theorem 7.1 ([78, 63]) . Let G = PO(2 , and G = PSL(2 , R ) . Consider a dis-crete group Γ and a representation ρ = ( ρ L , ρ R ) : Γ → G × G ⊂ G with ρ L convex co-compact. The following are equivalent, up to switching ρ L and ρ R in both (2) and (3) :(1) the action of Γ on X = AdS (cid:39) G via ρ is properly discontinuous;(2) there exists C < such that λ ( ρ R ( γ )) ≤ Cλ ( ρ L ( γ )) for all γ ∈ Γ ;(3) there is a ( ρ L , ρ R ) -equivariant Lipschitz map f : H → H with Lip( f ) < ;(4) Γ is Gromov hyperbolic and ρ : Γ → G ⊂ PGL(4 , R ) is P -Anosov. The equivalences (1) ⇔ (2) ⇔ (3), proved in [78], have been generalized in [65]to G = PO( n,
1) for any n ≥
2, allowing ρ L to be geometrically finite insteadof convex cocompact. These equivalences state that ρ = ( ρ L , ρ R ) acts properlydiscontinuously on X = AdS (cid:39) G if and only if, up to switching the two factors, ρ L is faithful and discrete and ρ R is “uniformly contracting” with respect to ρ L .The equivariant map f in (3) provides an explicit fibration in circles of ρ (Γ) \ X over the hyperbolic surface ρ L (Γ) \ H , see [65]. We refer to [43, 48, 65, 66, 87, 99]for many examples, to [106] for a classification in the compact AdS case, and to[65, 78] for links with Thurston’s asymmetric metric on Teichm¨uller space [108].The equivalences (1) ⇔ (2) ⇔ (4), proved in [63], generalize to G = PO( n, n, n, n + 2 , K )where K is R , C , or the quaternions. As an application [63], the set of holonomyrepresentations of complete ( G × G, G )-structures on a compact manifold M is openin the set of holonomy representations of all possible ( G × G, G )-structures on M .By [105], it is also closed, which gives evidence for an open conjecture stating thatall ( G × G, G )-structures on M should be complete (i.e. obtained as quotients of (cid:101) G ).7.3. Complete affine manifolds.
We now go back to (
G, X ) = (Aff( R d ) , R d ),looking for characterizations of holonomy representations of complete affine mani-folds, i.e. representations into G yielding properly discontinuous actions on X .We first note that any representation from a group Γ to the nonreductive Liegroup G = GL( d, R ) (cid:110) R d is of the form ρ = ( ρ L , u ) where ρ L : Γ → GL( d, R )(linear part) is a representation to GL( d, R ) and u : Γ → R d (translational part) a ρ L -cocycle, meaning u ( γ γ ) = u ( γ ) + ρ L ( γ ) · u ( γ ) for all γ , γ ∈ Γ.We focus on the case d = 3 and ρ L with values in O(2 , G = O(2 , (cid:39) PSL(2 , R )be the group of isometries of H . Its Lie algebra g (cid:39) R is the set of “infinitesimalisometries” of H , i.e. Killing vector fields on H . Here are some properness criteria. Theorem 7.2 ([60, 41]) . Let G = Aff( R ) and G = O(2 , (cid:39) PSL(2 , R ) . Considera discrete group Γ and a representation ρ = ( ρ L , u ) : Γ → G (cid:110) g ⊂ G with ρ L convexcocompact. The following are equivalent, up to replacing u by − u in both (2) and (3) :(1) the action of Γ via ρ = ( ρ L , u ) on X = R (cid:39) g is properly discontinuous;(2) there exists c < such that dd t | t =0 λ ( e u ( γ ) ρ L ( γ )) ≤ c λ ( ρ L ( γ )) for all γ ∈ Γ ;(3) there is a ( ρ L , u ) -equivariant vector field Y on H with “lipschitz” constant < . The equivalence (1) ⇔ (2) is a reinterpretation, based on [61], of a celebrated re-sult of Goldman–Labourie–Margulis [60]. The equivalence (1) ⇔ (3) is proved in [41]. EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 23
These equivalences are “infinitesimal versions” of the equivalences (1) ⇔ (2) ⇔ (3)of Theorem 7.1. Indeed, as explained in [41], we can see the ρ L -cocycle u : Γ → g asan “infinitesimal deformation” of the holonomy representation ρ L of the hyperbolicsurface (or orbifold) S = ρ L (Γ) \ H ; Condition (2) states that closed geodesics on S get uniformly shorter under this infinitesimal deformation. We can see a ( ρ L , u )-equivariant vector field Y on H as an “infinitesimal deformation” of the developingmap of the hyperbolic surface S ; Condition (3) states that any two points of H get uniformly closer compared to their initial distance. Thus Theorem 7.2 statesthat ρ = ( ρ L , u ) acts properly discontinuously on X = R (cid:39) g if and only if theinfinitesimal deformation u , up to replacing it by − u , is “uniformly contracting”.The vector field Y in (3) provides an explicit fibration in lines of the Margulisspacetime ρ (Γ) \ X over the hyperbolic surface S , and this can be used to define a geometric transition from complete AdS manifolds to Margulis spacetimes, see [41].In Theorem 7.1, the “uniform contraction” characterizing properness was in factan Anosov condition, encoding strong dynamics on a certain flag variety. It is nat-ural to expect that something similar should hold in the setting of Theorem 7.2.For this, a notion of affine Anosov representation into O(2 , (cid:110) R was recentlyintroduced by Ghosh [54] and extended to O( n + 1 , n ) (cid:110) R d ⊂ Aff( R d ) = G for any d = 2 n + 1 ≥ P n -Anosovrepresentation ρ L : Γ → O( n + 1 , n ) and a cocycle u : Γ → R d , the action of Γ on X = R d via ρ = ( ρ L , u ) is properly discontinuous if and only if ρ is affine Anosov.Theorem 7.2 was recently generalized in [43] as follows: for G = O( p, q ) with p, q ≥
1, consider a discrete group Γ, a faithful and discrete representation ρ L : Γ → G , and a ρ L -cocycle u : Γ → g ; then the action of Γ on g via ρ = ( ρ L , u ) : Γ → Aff( g )is properly discontinuous as soon as u satisfies a uniform contraction property inthe pseudo-Riemannian hyperbolic space H p,q − of Section 6.2. This allowed forthe construction in [43] of the first examples of irreducible complete affine manifolds M such that π ( M ) is neither virtually polycyclic nor virtually free: π ( M ) can infact be any right-angled Coxeter group. It would be interesting to understand thelinks with a notion of affine Anosov representation in this setting.8. Concluding remarks
By investigating the links between the geometry of (
G, X )-structures on mani-folds and the dynamics of their holonomy representations, we have discussed onlya small part of a very active area of research.We have described partial answers to Problem A for several types of modelgeometries (
G, X ). However, Problem A is still widely open in many contexts. Asan illustration, let us mention two major open conjectures on closed affine manifolds(in addition to the Auslander conjecture of Section 7.1): the Chern conjecture statesthat if a closed d -manifold M admits an (Aff( R d ) , R d )-structure, then its Eulercharacteristic must be zero; the Markus conjecture states that an (Aff( R d ) , R d )-structure on M is complete if and only if its holonomy representation takes valuesin SL( d, R ) (cid:110) R d . See [81] and references therein for recent progress on this.We have seen that Anosov representations from Gromov hyperbolic groups tosemisimple Lie groups provide a large class of representations answering Problem B.However, not much is known beyond them. One further class, generalizing Anosovrepresentations to finitely generated groups Γ which are not necessarily Gromov hyperbolic, is the class of P ( R d )-convex cocompact representations into PGL( d, R )of Section 6.4; it would be interesting to understand this class better in the frame-work of Problem B, see Section 6.4 and [45, Appendix]. As another generalizationof Anosov representations, it is natural to look for a class of representations ofrelatively hyperbolic groups into higher-rank semisimple Lie groups which wouldbear similarities to geometrically finite representations into rank-one groups, withcusps allowed: see [73, §
5] for a conjectural picture. Partial work in this directionhas been done in the convex projective setting, see [40].To conclude, here are two open questions which we find particularly interesting.
Structural stability.
Sullivan [104] proved that a structurally stable, nonrigidsubgroup of G = PSL(2 , C ) is always Gromov hyperbolic and convex cocompactin G . It is natural to ask if this may be extended to subgroups of higher-ranksemisimple Lie groups G such as PGL( d, R ) for d ≥
3, for instance with “convexcocompact” replaced by “Anosov”. In Section 6.4 we saw that there exist nonrigid,structurally stable subgroups of G = PGL( d, R ) which are not Gromov hyperbolic,namely groups that are P ( R d )-convex cocompact but not strongly P ( R d )-convexcocompact (Definitions 6.3 and 6.5). However, does a Gromov hyperbolic, nonrigid,structurally stable, discrete subgroup of G always satisfy some Anosov property? Abstract groups admitting Anosov representations.
Which linear hyper-bolic groups admit Anosov representations to some semisimple Lie group? Classicalexamples include surface groups, free groups, and more generally rank-one convexcocompact groups, see Section 4.2. By [44], all Gromov hyperbolic right-angledCoxeter groups (and all groups commensurable to them) admit Anosov representa-tions, see Section 6.2. On the other hand, if a hyperbolic group admits an Anosovrepresentation, then its Gromov flow (see Section 4.1) must satisfy strong dynami-cal properties, which may provide an obstruction: see the final remark of [23, § References [1] H. Abels,
Properly discontinuous groups of affine transformations: a survey , Geom. Dedi-cata 87 (2001), 309-333.[2] H. Abels, G. Margulis, G. Soifer,
The Auslander conjecture for dimension less than 7 ,arXiv:1211.2525.[3] L. Auslander,
The structure of compact locally affine manifolds , Topology 3 (1964), 131–139.[4] S. Ballas, J. Danciger, G.-S. Lee,
Convex projective structures on non-hyperbolic three-manifolds , to appear in Geom. Topol.[5] T. Barbot,
Three-dimensional Anosov flag manifolds , Geom. Topol. 14 (2010), 153–191.[6] ,
Deformations of Fuchsian AdS representations are quasi-Fuchsian , J. DifferentialGeom. 101 (2015), 1–46.[7] T. Barbot, F. Bonsante, J. Danciger, W. M. Goldman, F. Gu´eritaud, F. Kassel, K. Krasnov,J.-M. Schlenker, A. Zeghib,
Some open questions on anti-de Sitter geometry , arXiv:1205.6103.[8] T. Barbot, Q. M´erigot,
Anosov AdS representations are quasi-Fuchsian , Groups Geom.Dyn. 6 (2012), 441–483.[9] Y. Benoist,
Actions propres sur les espaces homog`enes r´eductifs , Ann. Math. 144 (1996),315–347.[10] ,
Convexes divisibles I , in
Algebraic groups and arithmetic , Tata Inst. Fund. Res.Stud. Math. 17 (2004), 339–374.[11] ,
Convexes divisibles III , Ann. Sci. ´Ec. Norm. Sup. 38 (2005), 793–832.[12] ,
Convexes divisibles IV , Invent. Math. 164 (2006), 249–278.[13] ,
A survey on divisible convex sets , Adv. Lect. Math. 6 (2008), 1–18.
EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 25 [14] L. Bers,
Simultaneous uniformization , Bull. Amer. Math. Society 66 (1960), 94–97.[15] J. Bochi, R. Potrie, A. Sambarino,
Anosov representations and dominated splittings , to ap-pear in J. Eur. Math. Soc.[16] F. Bonahon,
Geometric structures on 3-manifolds , Handbook of Geometric Topology, 93-164.[17] F. Bonahon, G. Dreyer,
Parametrizing Hitchin components , Duke Math. J. 163 (2014), 2935–2975.[18] F. Bonsante, K. Krasnov, J.-M. Schlenker,
Multi black holes and earthquakes on Riemannsurfaces with boundaries , Int. Math. Res. Not. 2011 (2011), 487–552.[19] F. Bonsante, J.-M. Schlenker,
Fixed points of compositions of earthquakes , Duke Math. J. 161(2012), 1011–1054.[20] B. H. Bowditch,
Geometrical finiteness for hyperbolic groups , J. Funct. Anal. 113 (1993), 245317.[21] B. H. Bowditch,
Spaces of geometrically finite representations , Ann. Acad. Sci. Fenn. Math.23 (1998), 389–414.[22] H. Bray,
Geodesic flow of nonstrictly convex Hilbert geometries , arXiv:1710.06938.[23] M. Bridgeman, R. D. Canary, F. Labourie, A. Sambarino,
The pressure metric for Anosovrepresentations , Geom. Funct. Anal. 25 (2015), 1089–1179.[24] J.-P. Burelle, N. Treib,
Schottky groups and maximal representations , to appear in Geom.Dedicata.[25] M. Burger, A. Iozzi, F. Labourie, A. Wienhard,
Maximal representations of surface groups:Symplectic Anosov structures , Pure Appl. Math. Q. 1 (2005), 555–601.[26] M. Burger, A. Iozzi, A. Wienhard,
Surface group representations with maximal Toledo in-variant , Ann. Math. 172 (2010), 517–566.[27] ,
Higher Teichm¨uller spaces: from
SL(2 , R ) to other Lie groups , Handbook of Te-ichm¨uller theory IV, 539–618, IRMA Lect. Math. Theor. Phys. 19, 2014.[28] , Maximal representations and Anosov structures , in preparation.[29] M. Burger, M. B. Pozzetti,
Maximal representations, non Archimedean Siegel spaces, andbuildings , Geom. Topol. 21 (2017), 3539–3599.[30] R. D. Canary, M. Lee, M. Stover,
Amalgam Anosov representations , with an appendix bythe authors and A. Sambarino, Geom. Topol. 21 (2017), 215–251.[31] C. Champetier,
Petite simplification dans les groupes hyperboliques , Ann. Fac. Sci. ToulouseMath. 3 (1994), 161–221.[32] V. Charette, T. Drumm, W. M. Goldman,
Proper affine deformations of two-generator Fuch-sian groups , Transform. Groups 21 (2016), p. 953–1002.[33] S. Choi, W. M. Goldman,
The deformation space of convex RP -structures on -orbifolds ,Amer. J. Math. 127 (2005), 1019–1102.[34] , Topological tameness of Margulis spacetimes , Amer. J. Math. 139 (2017), 297–345.[35] S. Choi, G.-S. Lee, L. Marquis,
Deformations of convex real projective structures on manifoldsand orbifolds , to appear in the Handbook of Group Actions.[36] ,
Convex projective generalized Dehn filling , arXiv:1611.02505.[37] B. Collier, N. Tholozan, J. Toulisse,
The geometry of maximal representations of surfacegroups into SO (2 , n ), arXiv:1702.08799.[38] D. Cooper, D. Long, S. Tillmann, Deforming convex projective manifolds , to appear in Geom.Topol.[39] M. Crampon,
The geodesic flow of Finsler and Hilbert geometries , Handbook of Hilbert Geometry, 2014.[40] M. Crampon, L. Marquis,
Finitude g´eom´etrique en g´eom´etrie de Hilbert , with an appendixby C. Vernicos, Ann. Inst. Fourier 64 (2014), 2299–2377.[41] J. Danciger, F. Gu´eritaud, F. Kassel,
Geometry and topology of complete Lorentz spacetimesof constant curvature , Ann. Sci. ´Ec. Norm. Sup´er. 49 (2016), 1–56.[42] ,
Margulis spacetimes via the arc complex , Invent. Math. 204 (2016), 133–193.[43] ,
Proper affine actions for right-angled Coxeter groups , preprint.[44] ,
Convex cocompactness in pseudo-Riemannian hyperbolic spaces , Geom. Dedicata 192(2018), 87–126, special issue
Geometries: A Celebration of Bill Goldman’s 60th Birthday .[45] ,
Convex cocompact actions in real projective geometry , arXiv:1704.08711.[46] ,
Projective convex cocompactness for right-angled Coxeter groups , preprint.[47] ,
Examples and non-examples of convex cocompact actions in real projective space , preprint.[48] B. Deroin, N. Tholozan,
Dominating surface group representations by Fuchsian ones , Int.Math. Res. Not. 2016 (2016), 4145–4166.[49] T. Drumm,
Fundamental polyhedra for Margulis space-times , Topology 21 (1992), 677–683. [50] T. Drumm, W. M. Goldman,
The geometry of crooked planes , Topology 38 (1999), 323–352.[51] C. Ehresmann,
Vari´et´es localement projectives , Enseign. Math. 35 (1937), 317–333.[52] V. V. Fock, A. B. Goncharov,
Moduli spaces of local systems and higher Teichm¨uller theory ,Publ. Math. Inst. Hautes ´Etudes Sci. 103 (2006), 1–211.[53] D. Fried, W. M. Goldman,
Three-dimensional affine crystallographic groups , Adv. Math. 47(1983), 1-49.[54] S. Ghosh,
Anosov structure on Margulis Space Time , Groups Geom. Dyn. 11 (2017), 739–775.[55] S. Ghosh, N. Treib,
Affine Anosov representations and proper actions , arXiv:1711.09712.[56] W. M. Goldman,
Convex real projective structures on surfaces , J. Differential Geom. 31(1990), 791–845.[57] ,
Locally homogeneous geometric manifolds , Proceedings of the ICM 2010.[58] ,
Flat affine, projective and conformal structures on manifolds: a historical perspec-tive , to appear in
Geometry in History , Springer-Verlag, 2018.[59] ,
Geometric structures on manifolds , in preparation for the AMS, Graduate Studies in Math.[60] W. M. Goldman, F. Labourie, G. A. Margulis,
Proper affine actions and geodesic flows ofhyperbolic surfaces , Ann. of Math. 170 (2009), 1051–1083.[61] W. M. Goldman, G. A. Margulis,
Flat Lorentz -manifolds and cocompact Fuchsian groups , in Crystallographic groups and their generalizations , 135–145, Contemp. Math. 262, AMS, 2000.[62] M. Gromov,
Hyperbolic groups , in
Essays in group theory , 75–263, Mathematical SciencesResearch Institute Publications, volume 8, Springer, New York, 1987.[63] F. Gu´eritaud, O. Guichard, F. Kassel, A. Wienhard,
Anosov representations and properactions , Geom. Topol. 121 (2017), 485–584.[64] F. Gu´eritaud, O. Guichard, F. Kassel, A. Wienhard,
Compactification of certain Clifford-Klein forms of reductive homogeneous spaces , Michigan Math. J. 66 (2017), 49–84.[65] F. Gu´eritaud, F. Kassel,
Maximally stretched laminations on geometrically finite hyperbolicmanifolds , Geom. Topol. 21 (2017), 693–840.[66] F. Gu´eritaud, F. Kassel, M. Wolff,
Compact anti-de Sitter 3-manifolds and folded hyperbolicstructures on surfaces , Pacific J. Math. 275 (2015), 325–359.[67] O. Guichard,
Groupes convexes-cocompacts en rang sup´erieur [d’apr`es Labourie, Kapovich,Leeb, Porti, ...] , S´eminaire Bourbaki, October 2017.[68] O. Guichard, F. Kassel, A. Wienhard,
Tameness of Riemannian locally symmetric spacesarising from Anosov representations , arXiv:1508.04759.[69] O. Guichard, A. Wienhard,
Anosov representations : Domains of discontinuity and applica-tions , Invent. Math. 190 (2012), 357–438.[70] ,
Positivity and higher Teichm¨uller theory , Proceedings of the ECM 2016.[71] N. J. Hitchin,
Lie groups and Teichm¨uller space , Topology 31 (1992), 449–473.[72] M. Kapovich, B. Leeb,
Finsler bordifications of symmetric and certain locally symmetricspaces , arXiv:1505.03593.[73] ,
Discrete isometry groups of symmetric spaces , arXiv:1703.02160.[74] M. Kapovich, B. Leeb, J. Porti,
Morse actions of discrete groups on symmetric spaces ,arXiv:1403.7671.[75] ,
A Morse Lemma for quasigeodesics in symmetric spaces and euclidean buildings ,arXiv:1411.4176.[76] ,
Dynamics on flag manifolds: domains of proper discontinuity and cocompactness ,Geom. Topol. 22 (2018), 157–234.[77] ,
Some recent results on Anosov representations , Transf. Groups 21 (2016), 1105–1121.[78] F. Kassel,
Quotients compacts d’espaces homog`enes r´eels ou p -adiques , PhD thesis,Univ. Paris-Sud, 2009.[79] F. Kassel, R. Potrie, Eigenvalue gaps for hyperbolic groups and semigroups , preprint.[80] B. Kleiner, B. Leeb,
Rigidity of invariant convex sets in symmetric spaces , Invent. Math. 163(2006), 657–676.[81] B. Klingler,
Chern’s conjecture for special affine manifolds , Ann. Math. 186 (2017), 69–95.[82] T. Kobayashi,
Criterion for proper actions on homogeneous spaces of reductive groups , J.Lie Theory 6 (1996), 147–163.[83] T. Kobayashi, T. Yoshino,
Compact Clifford–Klein forms of symmetric spaces — Revisited ,Pure Appl. Math. Q. 1 (2005), 591–663.[84] J.-L. Koszul,
D´eformation des connexions localement plates , Ann. Inst. Fourier 18 (1968),103–114.
EOMETRIC STRUCTURES AND REPRESENTATIONS OF DISCRETE GROUPS 27 [85] F. Labourie,
Anosov flows, surface groups and curves in projective space , Invent. Math. 165(2006), 51–114.[86] ,
Cross Ratios, Anosov Representations and the Energy Functional on Teichm¨ullerSpace , Ann. Sci. ´Ec. Norm. Sup´er. 41 (2008), 439–471.[87] G. Lakeland, C. J. Leininger,
Strict contractions and exotic SO ( d, quotients , to appearin J. Lond. Math. Soc.[88] G.-S. Lee, L. Marquis, Anti-de Sitter strictly GHC-regular groups which are not lattices , toappear in Trans. Amer. Math. Soc.[89] G.-S. Lee, T. Zhang,
Collar lemma for Hitchin representations , Geom. Topol. 21 (2017),2243–2280.[90] G. A. Margulis,
Complete affine locally flat manifolds with a free fundamental group , J.Soviet Math. 1934 (1987), 129–139, transl. from Zap. Naucha. Sem. LOMI (1984).[91] G. A. Margulis,
Discrete subgroups of semisimple Lie groups , Ergebnisse der Mathematikund ihrer Grenzgebiete (3) 17, Springer-Verlag, Berlin, 1991.[92] F. Math´eus,
Flot g´eod´esique et groupes hyperboliques d’apr`es M. Gromov , S´emin. Th´eor.Spectr. G´eom. 9 (1991), 67–87.[93] G. Mess,
Lorentz spacetimes of constant curvature (1990), Geom. Dedicata 126 (2007), 3–45.[94] J. Milnor,
On fundamental groups of complete affinely flat manifolds , Adv. Math. 25 (1977),178–187.[95] I. Mineyev,
Flows and joins of metric spaces , Geom. Topol. 9 (2005), 403–482.[96] T. Nagano, K. Yagi,
The affine structures on the real two torus 1 , Bull. Amer. Math. Soc. 79(1973), 1251–1253.[97] M. Pollicott, R. Sharp,
Higher Teichm¨uller theory for surface groups and shifts of finite type , preprint.[98] J.-F. Quint,
Groupes convexes cocompacts en rang sup´erieur , Geom. Dedi. 113 (2005), 1–19.[99] F. Salein,
Vari´et´es anti-de Sitter de dimension 3 exotiques , Ann. Inst. Fourier 50 (2000),257–284.[100] J.-M. Schlenker,
Vari´et´es lorentziennes plates vues comme limites de vari´et´es anti-de Sitter ,S´eminaire Bourbaki, June 2015, Ast´erisque 380 (2016), 475–497.[101] I. Smilga,
Proper affine actions: a sufficient criterion , arXiv:1612.08942.[102] T. Strubel,
Fenchel–Nielsen coordinates for maximal representations , Geom. Dedicata 176(2015), 45–86.[103] D. Sullivan,
The density at infinity of a discrete group of hyperbolic motions , Publ. Math.Inst. Hautes ´Etudes Sci. 50 (1979), 419–450.[104] D. Sullivan,
Quasiconformal homeomorphisms and dynamics II: Structural stability implieshyperbolicity for Kleinian groups , Acta Math. 155 (1985), 243–260.[105] N. Tholozan,
Sur la compl´etude de certaines vari´et´es pseudo-riemanniennes localementsym´etriques , Ann. Inst. Fourier 65 (2015), 1921–1952.[106] ,
Dominating surface group representations and deforming closed AdS 3-manifolds ,Geom. Topol. 21 (2017), 193–214.[107] W. P. Thurston,
The geometry and topology of three-manifolds , lecture notes (1980).[108] ,
Minimal stretch maps between hyperbolic surfaces , preprint (1986), arXiv:9801039.[109] E. B. Vinberg,
Discrete linear groups generated by reflections , Math. USSR Izv. 5 (1971),1083–1119.[110] A. Wienhard,
An invitation to higher Teichm¨uller theory , Proceedings of the ICM 2018.[111] E. Witten, (2 + 1) -dimensional gravity as an exactly soluble system , Nuclear Physics B 311(1988), 46–78.[112] T. Zhang,
Degeneration of Hitchin representations along internal sequences , Geom. Func.Anal. 25 (2015), 1588–1645.[113] A. Zimmer,
Projective Anosov representations, convex cocompact actions, and rigidity ,arXiv:1704.08582.
CNRS and Institut des Hautes ´Etudes Scientifiques, Laboratoire Alexander Grothendieck,35 route de Chartres, 91440 Bures-sur-Yvette, France
E-mail address ::