On separated families of Anosov representations

Joaquín Lejtreger, Joaquín Lema · 2026

Read On separated families of Anosov representations by Joaquín Lejtreger, Joaquín Lema (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-18994.txt.

Introduction

Lejtreger and Lema define ways to split sets of Anosov representations. Along a diverging run of such families, the critical exponent nears a combinatorial invariant read from the spectral data of a finite graph. Their method yields bounds on the Thurston asymmetric metric. As an application, they study degenerations of convex projective structures on a pair of pants, building on a McMullen example.

Catalog record

CREATORJoaquín Lejtreger, Joaquín Lema
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE107850
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.18994

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

On separated families of Anosov representations Joaquín Lejtreger, Joaquín Lema Abstract . We introduce different notions of separation for families of Anosov representations. We show that, along a diverging sequence of such families, the critical exponent is asymptotic to a combinatorial invariant computable from the spectral data of a finite graph. Our method allows us to derive bounds on the Thurston asymmetric metric of [ 10 ] . As an application, we study specific degenerations of convex projective structures on a pair of pants, generalizing an example of McMullen. 1. Introduction The goal of this article is to estimate the critical exponent of diverging families of representations of a hyperbolic group \Gamma into a semisimple Lie group G , under a dynamical condition on a chosen generating set that we call strong separation . Our motivation comes from an example of McMullen [ 21 ] , recently reviewed and expanded by Courtois-Guilloux [ 15 ] and Dang-Mehmeti [ 12 ] , who compute asymptotics for the critical exponent for diverging Schottky representations acting on \mathbb{H}^{n} . Our framework unifies and generalizes these computations to higher rank. Furthermore, our techniques also provide control along strongly separated sequences of the Thurston asymmetric metric introduced in [ 10 ] . As an application, we study degenerations of convex projective structures on a pair of pants, whose associated holonomies define strongly separated representations of the free group into SL_{3}(\mathbb{R}) . 1.1. The setup Given \Gamma a hyperbolic group, G a semisimple Lie group with non-compact factors, \mathfrak{a}^{+} a Weyl chamber, and \Theta a set of simple roots, we will be concerned with the study of \Theta -Anosov representations. This notion was introduced in [ 18 ] and [ 14 ] , and can be thought of as a higher-rank analog of convex cocompact representations. Such representations come equipped with a pair of continuous, dynamics-preserving, and equivariant maps (\xi,\xi^{op}):\partial\Gamma\to\mathcal{F}_{\Theta}\times\mathcal{F}^{op}_{\Theta}, where \mathcal{F}_{\Theta} and \mathcal{F}^{op}_{\Theta} are the opposite partial flag manifolds associated to \Theta . We consider the Cartan projection \kappa:G\to\mathfrak{a}^{+} , and the Jordan projection \lambda:G\rightarrow\mathfrak{a}^{+} . In SL_{n}(\mathbb{R}) , the Cartan projection is the vector containing the log of the singular values of a matrix in order, and the Jordan projection is the vector containing the log of the moduli of the eigenvalues of a matrix in order. There are many length functions assigned to a given Anosov representation, and each corresponds to a different notion of growth in the symmetric space. Given a functional \varphi:\mathfrak{a}\to\mathbb{R} , we consider its associated length function L^{\varphi}_{\rho}:\Gamma\rightarrow\mathbb{R} via L^{\varphi}_{\rho}(\gamma)=\varphi(\lambda(\rho(\gamma))). For a representation \rho:\Gamma\to G , we define the limit cone \mathcal{L}_{\rho} of \rho , as the smallest closed convex cone of \mathfrak{a}^{+} containing \lambda(\rho(\Gamma)) . We will say that \varphi\in\mathfrak{a}^{*} is positive in the limit cone of \rho if \varphi|_{\mathcal{L}_{\rho}-\{0\}}>0 . This condition implies that the length function is positive, and one can define the critical exponent h_{\rho}(\varphi):=\limsup_{t\to\infty}\frac{1}{t}\log\#\{[\gamma]\in[\Gamma]:L^{\varphi}_{\rho}(\gamma)\leq t\}\in[0,\infty], where [\Gamma] is the set of conjugacy classes of \Gamma . To estimate these critical exponents, it will be necessary to code the limit map and apply techniques from thermodynamic formalism. For this purpose, we use the notion of strong Markov codings , defined in [ 9 ] and [ 8 ] . Given a symmetric set of generators S for \Gamma , a strong Markov coding is a finite directed graph \mathcal{G}=(\mathcal{V},\mathcal{E}) equipped with a labeling of the edges \pi:\mathcal{E}\to S by elements of S . Such labels allow us to associate with any finite directed path a geodesic segment in \Gamma , and similarly, to any bi-infinite directed path, a geodesic in the group. Collecting the labels traversed by a bi-infinite path in \mathcal{G} , we obtain a dynamical system X_{\mathcal{G}}\subset S^{\mathbb{Z}} equipped with the shift map, called a sofic shift . Such a sofic shift codes the boundary via a map p:X_{\mathcal{G}}\rightarrow\partial\Gamma , sending the sequence of labels of a bi-infinite geodesic path to its forward endpoint on \partial\Gamma . Using this language, the work of Bochi-Potrie-Sambarino [ 4 ] provides a criterion for certifying Anosov representations adapted to a given strong Markov coding. More precisely, one can prove that a representation \rho:\Gamma\rightarrow G is \Theta -Anosov if and only if we can assign for every vertex v\in\mathcal{V} , an open set M_{v} in the partial flag \mathcal{F}_{\Theta} , called a multicone, in such a way that, for each e=(v_{0},v_{1})\in\mathcal{E} , \rho(\pi(e))^{-1}M_{v_{1}}\subseteq M_{v_{0}}. Such multicones serve as a “coarse” Markov partition of the limit set. One can show that \xi\circ p(\underline{x})\in M_{v_{0}} , for every \underline{x}\in X_{\mathcal{G}} induced by a path starting at v_{0} (see Theorem 3.7 ). Every element in the image of a \Theta- Anosov representation acts in the partial manifold \mathcal{F}_{\Theta} with north-south dynamics, in the sense that for every \gamma\in\Gamma of infinite order, we get the property that \rho(\gamma)^{n} acting on \mathcal{F}_{\Theta} converges uniformly in compact sets to \xi(\gamma_{+}) outside a compact set C_{\rho(\gamma)}^{-} , called the Schubert locus. The separation conditions we impose require these north-south dynamics of the generators to be aligned with the multicone partition. Informally, \rho is \varepsilon -separated with respect to \mathcal{G} and \Theta if we can choose an invariant multicone family \{M_{v}\}_{v\in\mathcal{V}} in \mathcal{F}_{\Theta} in such a way that M_{v} is \varepsilon -away from the Schubert loci of the repelling flags of generators corresponding to incoming edges. It is \varepsilon -strongly separated if, in addition, each generator \rho(s) exhibits uniform north-south behavior in \mathcal{F}_{\Theta} : its attracting and repelling flags are 2\varepsilon -apart, and its action contracts the \varepsilon -complement of the repelling Schubert locus into the \varepsilon -ball around the attracting flag. See Definition 4.1 for a detailed definition. Notice that the strong separation condition resembles the usual definition of a “ping-pong” system, except that admissible words are governed by the strong Markov coding, taking into account relations in \Gamma . To verify that a representation is separated, we only need to check finitely many conditions, which makes this notion easy to verify in specific examples, as we will see, for instance, in the proof of Theorem 5.7 . The key remark is that, for a given \varepsilon>0 , a representation being \varepsilon -strongly separated allows us to identify regions of the limit set where the Busemann cocycle is well-behaved. For example, in rank one, these conditions imply the existence of C>0 (only depending on \varepsilon ) such that for any convex-cocompact representation \rho that is strongly separated, we get the following lower bound for the Busemann function: d(o,\rho(g)^{-1}o)-C\leq b_{x}(o,\rho(g)^{-1}o), where S denotes the set of generators, g,h\in S , and x belongs to the subset of the limit set codified by infinite words starting with h , provided gh is part of the language of the Markov coding. See Lemma 4.2 for the higher rank version of this inequality. 1.2. Main results Having established the setup, we proceed to state the main results. Given a choice of a Weyl chamber \mathfrak{a}^{+} , we denote the space \mathfrak{a}_{\Theta}=\bigcap_{\alpha\in\Pi\setminus\Theta}\ker\alpha. Two facts drive the argument of our main theorems: first, a characterization of h_{\rho}(\varphi) as the unique zero of the function s\to P(-s\varphi\circ B) , where B is an \mathfrak{a}_{\Theta} -valued potential obtained from the Iwasawa cocycle (see Proposition 3.11 ). Second, the strong separation condition bounds

Read the complete text