[Paper Review] Finitude g\'eom\'etrique en g\'eom\'etrie de Hilbert + an erratum/addendum

Blayac, Pierre-Louis, Crampon, Micka\"el+1 more|arXiv (Cornell University)|Feb 24, 2012

Mathematics and Applications36 references26 citations

TL;DR

This paper establishes that the standard definition of geometrical finiteness in Gromov-hyperbolic spaces does not suffice for Hilbert geometries defined on strictly convex domains with C¹ boundary. By constructing explicit counterexamples, the authors show that a strengthened definition is required to ensure equivalence with geometric properties of the quotient orbifold, extending Bowditch’s hyperbolic geometry results to Hilbert geometries. An erratum corrects critical flaws in the original proof, particularly in the convergence argument involving projective transformations and horoballs.

ABSTRACT

The paper is divided in 2 parts. The first part is the original paper of the second and third authors arXiv:1202.5442v2. The second part is an erratum/addendum written in english and concatenated at the end of the former paper. In the erratum/addentum, we amend Theorems 1.3 and 1.11 of arXiv:1202.5442v2: Finitude g\'eom\'etrique en g\'eom\'etrie de Hilbert. We seize the opportunity to show that in round Hilbert geometry, geometrical finiteness (gf) is equivalent to cusp-uniform action and to fill some small gaps that appear in two other proofs of arXiv:1202.5442v2.

Motivation & Objective

• To resolve the discrepancy between the standard Gromov-hyperbolic definition of geometrical finiteness and its expected equivalence with quotient orbifold geometry in Hilbert geometries.
• To identify why the classical definition fails in Hilbert geometries by constructing explicit counterexamples where the group action is not geometrically finite despite satisfying the standard boundary condition.
• To provide a corrected and strengthened definition of geometrical finiteness that ensures equivalence with topological, volume, and dynamical properties of the quotient.
• To fix errors in the original proof of the main theorem, particularly in the convergence of sequences of convex sets under projective transformations.
• To demonstrate that certain representations of SL₂(R) into SL₅(R) yield non-geometrically finite actions on round convex domains in RP⁴, even when the action is cocompact on the convex core.

Proposed method

• Constructing a counterexample using a noncocompact lattice Γ ⊂ SL₂(R) and its irreducible representation ρ: SL₂(R) → SL₅(R), preserving a round convex domain Ω ⊂ RP⁴.
• Analyzing the action of ρ(Γ) on the convex core C(ΛΓ) and showing that while C is Gromov-hyperbolic and has finite volume quotient, the group is not geometrically finite due to parabolic subgroups not conjugate into O₄,₁(R).
• Using projective geometry and the theory of Busemann functions to study horoballs and conical limit points in the Hilbert metric.
• Applying Benoist’s compactness lemma ([Ben03, Lem. 2.8]) to extract a sequence of projective transformations (gₙ) such that gₙ(Ω) → Ω∞, with Ω∞ ∩ S an ellipsoid.
• Recentering the geometry via a one-parameter subgroup (γₜ) of hyperbolic automorphisms preserving a subspace S and the point u ∈ ∂Ω, to analyze the behavior near parabolic fixed points.
• Correcting the flawed convergence argument by replacing the sequence (γₖₙ) with a carefully chosen (gₙ) from Benoist’s lemma, ensuring that the image of the convex core converges to a set contained in S.

Experimental results

Research questions

• RQ1 Why does the standard Gromov-hyperbolic definition of geometrical finiteness fail to capture the correct geometric structure in Hilbert geometries?
• RQ2 Can a group act cocompactly on the convex core and have finite volume quotient while failing to be geometrically finite in the Hilbert setting?
• RQ3 What conditions ensure that a group action on a Hilbert geometry is geometrically finite, especially when parabolic subgroups are not conjugate into the standard parabolic subgroup O₄,₁(R)?
• RQ4 How can the original proof of the equivalence between geometrical finiteness and finite volume of the convex core be corrected when the convergence of (γₖₙ(Ω) ∩ Eₑₓₜ) to Eₑₓₜ fails?
• RQ5 Are there representations of SL₂(R) into SL₅(R) that preserve convex domains in RP⁴ where the image group acts cocompactly on the convex core but is not geometrically finite?

Key findings

• The standard definition of geometrical finiteness based on boundary group action is insufficient for Hilbert geometries; a strengthened condition is required to ensure equivalence with topological and volume properties of the quotient orbifold.
• A counterexample exists where ρ(Γ) acts cocompactly on the convex core and has finite-volume quotient, yet is not geometrically finite because maximal parabolic subgroups are not conjugate into O₄,₁(R).
• The original proof’s claim that (γₖₙ(Ω) ∩ Eₑₓₜ) converges to Eₑₓₜ is incorrect; the convergence fails when restricted to the cone Co, and the sequence may converge to a convex set with empty interior.
• The corrected proof uses Benoist’s lemma to extract a sequence (gₙ) such that gₙ(Ω) → Ω∞, with Ω∞ ∩ S an ellipsoid, and shows that the image of the convex core under gₙ converges to a subset of S.
• The corrected argument establishes that if the Hilbert distance from a sequence of points to the sides of a triangle tends to infinity, then the limit point must lie in S, contradicting the assumption unless the limit is in S.
• The erratum shows that (VF)₁ ⇒ (GF) and ((gf)&(Hyp)) ⇒ (GF) are false in general, with counterexamples arising from irreducible representations of SL₂(R) in SL₅(R) preserving round convex domains in RP⁴.