Higher Teichmüller Theory: Anosov Representations¶
Job: define \(P\)-Anosov representations \(\rho:\Gamma\to G\) of a word-hyperbolic group \(\Gamma\) into a semisimple Lie group \(G\) relative to a parabolic \(P\subset G\) (Labourie, generalized by Guichard–Wienhard) via a dynamics-preserving transverse boundary map \(\xi_\rho:\partial\Gamma\to G/P\) with contraction/expansion, and state the foundational theorems: discreteness + quasi-isometric embedding (Labourie), openness, structural stability (Guichard–Wienhard), and the flavors (\(B\)-Anosov, \(P_k\)-Anosov, uniformly Morse / relatively Anosov).
Core¶
Setup¶
Let \(\Gamma\) be a word-hyperbolic group with Gromov boundary \(\partial\Gamma\) (a compact metrizable space on which \(\Gamma\) acts by homeomorphisms; for \(\Gamma=\pi_1\Sigma_g\) this is \(S^1\)). Let \(G\) be a connected semisimple Lie group of non-compact type, \(K\subset G\) maximal compact, and fix a parabolic subgroup \(P\subset G\). Let \(G/P\) be the generalized flag variety; for \(G=\mathrm{SL}_n\) and \(P=P_k\) (stabilizer of a \(k\)-plane), \(G/P_k=\mathrm{Gr}(k,n)\); for \(P=B\) a minimal parabolic (Borel), \(G/B\) is the full flag variety.
A representation \(\rho:\Gamma\to G\) induces an action on \(G/P\). The Anosov condition asks for a \(\rho\)-equivariant boundary map that is dynamics-preserving and transverse, with a contraction/expansion estimate along the Cartan decomposition of \(G\).
\(P\)-Anosov representations¶
A continuous map \(\xi:\partial\Gamma\to G/P\) is transverse if for any distinct \(a,b\in\partial\Gamma\), the pair \((\xi(a),\xi(b))\) lies in the open \(G\)-orbit of \(G/P\times G/P\) (the unique open orbit corresponding to opposite parabolics); it is dynamics-preserving if for any sequence \(\gamma_n\in\Gamma\) with \(\gamma_n\to a\in\partial\Gamma\) and \(\gamma_n^{-1}\to b\in\partial\Gamma\), the limit \(\rho(\gamma_n)\xi\) contracts \(G/P\setminus\{\xi(a)\}\) to \(\xi(b)\) (in the appropriate partial flag sense).
Definition (Labourie; Guichard–Wienhard). A representation \(\rho:\Gamma\to G\) is \(P\)-Anosov if there exists a continuous \(\rho\)-equivariant transverse dynamics-preserving map \(\xi_\rho:\partial\Gamma\to G/P\) such that, writing \(\rho(\gamma)=k_1\exp(H(\gamma))k_2\) in the Cartan decomposition \(G=K\exp(\mathfrak a^+)K\) with \(H(\gamma)\in\mathfrak a^+\), the dominant Cartan component \(\alpha_1(H(\gamma))\to+\infty\) uniformly as \(\gamma\to\infty\) in \(\Gamma\) (the contraction/expansion along the \(P\)-direction). Equivalently (Guichard–Wienhard), \(\rho\) induces a proper dynamics-preserving embedding of the geodesic flow of \(\Gamma\) into the \(G\)-frame bundle over \(G/P\), Anosov in the dynamical systems sense.
The foundational theorems¶
| Theorem (reference) | Statement |
|---|---|
| Discreteness + QI-embedding (Labourie 2006, for \(\Gamma=\pi_1\Sigma\); Guichard–Wienhard general) | \(P\)-Anosov \(\rho:\Gamma\to G\) is discrete and a quasi-isometric embedding \(\Gamma\hookrightarrow G\) (with word metric on \(\Gamma\), left-invariant Riemannian on \(G\)). In particular \(\rho\) has finite kernel and is injective up to finite index. |
| Openness (Labourie; Guichard–Wienhard) | The set of \(P\)-Anosov representations is open in \(\mathrm{Hom}(\Gamma,G)\). |
| Structural stability (Guichard–Wienhard) | \(P\)-Anosov representations are structurally stable: nearby representations are \(P\)-Anosov and conjugate on the boundary (the boundary map varies continuously); in particular a higher Teichmüller component is contained in the Anosov locus. |
| Morse lemma (Kapovich–Leeb–Porti; Schrader–Wyatt) | The image of a \(P\)-Anosov orbit in \(G/K\) is Morse: quasi-geodesics at bounded distance stay close, the shadowing property holds; generalizes the classical Morse lemma for rank-one. |
| Boundary map uniqueness | \(\xi_\rho\) is uniquely determined by \(\rho\) (when transverse + dynamics-preserving); it is Hölder continuous. |
| Convex cocompact recovery | for \(G=\mathrm{PSL}_2(\mathbb{C})\) and \(P=B\) the minimal parabolic, \(P\)-Anosov \(\Leftrightarrow\) convex cocompact Kleinian representation. |
The openness + structural stability theorem is the engine of higher Teichmüller theory: it shows that the Anosov property is open in the character variety, hence any higher Teichmüller component (built from Hitchin / maximal paradigms) that contains one Anosov representation consists entirely of Anosov representations, and the discreteness + QI-embedding is uniform across the component.
Anosov flavors¶
| Flavor | Parabolic \(P\) | Target \(G/P\) | Setting |
|---|---|---|---|
| \(B\)-Anosov | minimal parabolic \(B\) | full flag \(G/B\) | strongest; Labourie's original for Hitchin into \(\mathrm{PSL}_n(\mathbb{R})\) |
| \(P_k\)-Anosov (\(\mathrm{SL}_n\)) | maximal parabolic \(P_k\) | Grassmannian \(\mathrm{Gr}(k,n)\) | for \(\mathrm{SL}_n\); the boundary map gives a curve \(\partial\Gamma\to\mathrm{Gr}(k,n)\) |
| Uniformly Morse | (no fixed \(P\); Morse property in \(G/K\)) | geodesics in \(G/K\) | Kapovich–Leeb–Porti; independent of parabolic choice |
| Relatively \(P\)-Anosov | \(P\), for \(\Gamma\) relatively hyperbolic | relative boundary \(\partial\Gamma\) | relatively hyperbolic groups; peripheral non-Anosov directions |
| Partially \(P\)-Anosov | a non-minimal \(P\) | partial flag \(G/P\) | maximal representations of Hermitian type; only some Cartan directions contract |
The implication chain: \(B\)-Anosov \(\Rightarrow\) \(P\)-Anosov for all \(P\supset B\) (more parabolics = weaker condition); for \(\mathrm{SL}_n\), \(B\)-Anosov \(\Leftrightarrow\) \(P_k\)-Anosov for all \(k=1,\dots,n-1\). The Morse property is implied by \(P\)-Anosov for any \(P\) and is the parabolic-independent formulation of "Anosov-like."
Relation to convex cocompact Kleinian groups¶
For \(G=\mathrm{PSL}_2(\mathbb{C})\) acting on \(\mathbb{H}^3\) with \(G/B=\mathbb{CP}^1\) the Riemann sphere, the \(B\)-Anosov condition for \(\rho:\Gamma\to\mathrm{PSL}_2(\mathbb{C})\) (with \(\Gamma\) word-hyperbolic) is equivalent to \(\rho\) being convex cocompact (discrete, faithful, with a convex hull quotient of finite volume / compact limit set). The boundary map \(\xi_\rho:\partial\Gamma\to\mathbb{CP}^1\) is the limit map of the convex cocompact representation. This is the original motivating case: Anosov representations are the higher-rank generalization of convex cocompact Kleinian groups, replacing \(\mathrm{PSL}_2(\mathbb{C})\) by a general semisimple \(G\) and \(\mathbb{CP}^1\) by \(G/P\).
See it¶
The Hitchin representation into \(\mathrm{PSL}_n(\mathbb{R})\) is \(B\)-Anosov, and the boundary curve is the Frenet lift (Labourie). Let \(\rho:\pi_1\Sigma\to\mathrm{PSL}_n(\mathbb{R})\) be a Hitchin representation (see Hitchin & Maximal). Labourie constructs a unique continuous \(\rho\)-equivariant map
that is a Frenet curve: for any distinct \(x_1,\dots,x_n\in S^1\), the osculating flag \((\xi_\rho(x_1),\dots,\xi_\rho(x_n))\) is a full flag in \(\mathbb{R}^n\) in general position. Concretely, for a Fuchsian representation (the irreducible \(\mathrm{PSL}_2(\mathbb{R})\hookrightarrow\mathrm{PSL}_n(\mathbb{R})\)), the limit curve is the Veronese embedding \(S^1\to\mathbb{RP}^{n-1}\): identify \(S^1\) with \(\mathbb{RP}^1\) and map \([x:y]\mapsto[x^{n-1}:x^{n-2}y:\dots:y^{n-1}]\), the rational normal curve of degree \(n-1\). For a general Hitchin representation the boundary curve is a deformation of the Veronese curve, still Frenet, and the Frenet property is equivalent to \(B\)-Anosov (Labourie): the transversality of the osculating flags is exactly the transversality in \(G/B\times G/B\), and the Frenet osculation encodes the contraction along the principal \(\mathrm{SL}_2\) direction. This is the prototypical "compute the boundary map" computation in higher Teichmüller theory: a Hitchin representation is detected by the geometry of its limit curve in projective space, and the Frenet property is the visible signature of the Anosov dynamics.
Check¶
Prove that \(B\)-Anosov for \(\mathrm{SL}_n\) implies \(P_k\)-Anosov for every \(k\), and explain where the implication can fail in the reverse direction.
The full flag \(G/B\) maps \(G\)-equivariantly to every partial flag \(G/P_k\) (a maximal parabolic is \(P_k\supset B\)). Given a \(B\)-Anosov boundary map \(\xi_B:\partial\Gamma\to G/B\), compose with the projection \(G/B\to G/P_k\) to get \(\xi_k:\partial\Gamma\to G/P_k\); transversality descends (the open orbit in \(G/B\times G/B\) maps into the open orbit of \(G/P_k\times G/P_k\)), dynamics-preservation and the contraction estimate hold because the Cartan contraction along the simple root \(\alpha_1\) implies contraction along the simple root \(\alpha_k\) (the dominant root contracts all simple roots). Hence \(P_k\)-Anosov. The reverse fails: a \(P_k\)-Anosov representation may not contract the other simple roots, so the full flag boundary map need not exist. Example: representations into \(\mathrm{SL}_n\) that are \(P_1\)-Anosov (a curve in \(\mathbb{RP}^{n-1}\)) but not \(B\)-Anosov (no compatible flag of higher osculating spaces) — these exist; the Frenet condition is genuinely stronger than a single transverse curve.
Sketch why the openness of the \(P\)-Anosov condition plus structural stability implies that a higher Teichmüller component is a smooth manifold, and where the closedness of Anosov would have been needed but is absent.
Openness: the Anosov estimate (uniform expansion of the dominant Cartan component) is an open condition in \(\mathrm{Hom}(\Gamma,G)\) because it is a finite set of strict inequalities on the Cartan projections \(\alpha_i(H(\rho(\gamma)))\) for a finite generating set of \(\Gamma\) (the contraction is uniform over the geodesic flow, and stability of hyperbolic dynamical systems under perturbation gives the open condition). Structural stability: a nearby representation is conjugate on the boundary, so the boundary map varies continuously and the higher Teichmüller component (a connected set containing an Anosov rep) is entirely in the Anosov locus. Smoothness: the character variety is smooth at reductive representations, and Anosov representations are reductive (discrete faithful into a reductive \(G\)), so the component is a smooth manifold of the expected dimension. The closedness is not needed: a higher Teichmüller component is connected and contains an Anosov point (by the Hitchin / maximal construction), so openness propagates Anosov across the whole component — the failure of Anosov to be closed is precisely why one cannot characterize higher Teichmüller spaces as "the Anosov locus" (which would be a union of components, not a single one).
Pitfalls¶
- Anosov is open, not closed: the locus of \(P\)-Anosov representations is open in \(\mathrm{Rep}(\Gamma,G)\) but not closed; the boundary of a higher Teichmüller component contains non-Anosov (degenerate) representations. Treating Anosov as a closed condition leads to wrong compactness claims.
- \(B\)-Anosov \(\neq\) \(P\)-Anosov in general: the implication is one-way (\(B\)-Anosov \(\Rightarrow\) \(P\)-Anosov for \(P\supset B\)); a \(P_k\)-Anosov representation into \(\mathrm{SL}_n\) need not be \(B\)-Anosov, and the boundary map into \(\mathrm{Gr}(k,n)\) need not lift to the full flag. The flavor matters for which properties (e.g. Frenet) hold.
- Relatively Anosov requires relatively hyperbolic \(\Gamma\): the relative Anosov condition is defined for \(\Gamma\) relatively hyperbolic (with peripheral structure); applying it to an arbitrary \(\Gamma\) without the relative hyperbolicity hypothesis is a category error — the boundary \(\partial\Gamma\) is not well-behaved.
- Convex cocompact \(\neq\) Anosov for \(\mathrm{PSL}_2(\mathbb{R})\) surface groups without faithfulness: \(B\)-Anosov into \(\mathrm{PSL}_2(\mathbb{C})\) recovers convex cocompact Kleinian reps, which include representations of free groups; for surface groups one additionally needs faithfulness to land in Teichmüller space (the Anosov property alone gives discreteness + QI-embedding, hence faithfulness up to finite kernel, but the orientation / Euler class condition is extra).
Refs¶
- Labourie, Anosov flows, surface groups and curves in projective space (Invent. 2006)
- Guichard–Wienhard, Anosov representations: characterizations and applications (preprint, 2012 / Amer. J. Math. 2015)
- Kapovich–Leeb–Porti, Morse actions of word-hyperbolic groups on symmetric spaces (Morse lemma)
- Schrader–Wyatt, Anosov representations and the Morse property
- Burger–Iozzi–Wienhard, Surface group representations with maximal Toledo invariant (partially Anosov for Hermitian)
Next¶
Hitchin & Maximal Components · Frontiers · Higher Teichmüller index · Foundations · Representation theory · Algebraic topology · Differential geometry