alphaFeedback

WorksMathematics

Strong stable sets and asymptotic fineness

AI contributions
  • Claude
View details

Abstract

We investigate the strong stable sets of the geodesic flow on hyperbolic surfaces and their relation to the injectivity radius. We prove that the strong stable set of a vector does not coincide with its stable horocycle if and only if the injectivity radius along the associated geodesic ray is not bounded away from zero and the ray does not diverge through a cusp. In that case, the strong stable set is an uncountable union of stable horocycles. In previous works, we considered the particular cases where the ray meets arbitrarily short closed geodesics or closed horocycles. Here we complete the study by treating the third possible case, where the ray meets closed hypercycles of arbitrarily small length.

Introduction

We consider the geodesic flow gtg_t on the unit tangent bundle T1ST^1S of a hyperbolic surface SS. We endow T1ST^1S with the Sasaki distance dSad_{Sa} induced by the hyperbolic metric of SS. For u∈T1Su \in T^1S, the stable horocycle HussH^{ss}_u is defined as the projection of the stable horocycle of any lift of uu to T1HT^1\mathbb{H}, whereas the strong stable set WussW^{ss}_u is the set

Wuss={v∈T1S:lim⁡t→+∞dSa(gtu,gtv)=0}.W^{ss}_u = \{v \in T^1S : \lim_{t \to+\infty} d_{Sa}(g_tu,g_tv)=0\}.

Following Bellis [6], we define the asymptotic fineness of the geodesic ray u(R+)u(\mathbb{R}^+) as

lim inf⁡t≥0Inj⁡(u(t)),\liminf_{t \ge0} \operatorname{Inj}(u(t)),

where Inj⁡(x)\operatorname{Inj}(x) denotes the injectivity radius of SS at x∈Sx \in S. Our main result establishes a link between the asymptotic fineness and strong stable sets.

Theorem 1. Let SS be a hyperbolic surface and u∈T1Su \in T^1S. Then Huss⊊WussH^{ss}_u \subsetneq W^{ss}_u if and only if

lim inf⁡t≥0Inj⁡(u(t))=0\liminf_{t \ge0} \operatorname{Inj}(u(t)) = 0

and the geodesic ray u(R+)u(\mathbb{R}^+) does not diverge through a cusp. In that case, the set WussW^{ss}_u is an uncountable union of stable horocycles HvissH^{ss}_{v_i}.

There are several ways for the asymptotic fineness of u(R+)u(\mathbb{R}^+) to vanish. Bellis [5] and the authors together with S. Herrero [BCDHV26] study the case where u(R+)u(\mathbb{R}^+) meets a sequence of arbitrarily short closed geodesics. In [2], the authors study the case where u(R+)u(\mathbb{R}^+) meets a sequence of arbitrarily short closed horocycles. The contribution of this article is to study a third possible case, which is not always covered by the previous works: the ray u(R+)u(\mathbb{R}^+) meets closed hypercycles of arbitrarily small length.

Theorem 2. Let SS be a hyperbolic surface and u∈T1Su \in T^1S. If the geodesic ray u(R+)u(\mathbb{R}^+) meets an infinite sequence of closed hypercycles whose lengths tend to zero, then

  1. Huss⊊WussH^{ss}_u \subsetneq W^{ss}_u,

  2. the set WussW^{ss}_u is an uncountable union of stable horocycles HvissH^{ss}_{v_i}.

These three cases are exhaustive: if the asymptotic fineness of u(R+)u(\mathbb{R}^{+}) vanishes and u(R+)u(\mathbb{R}^{+}) does not diverge through a cusp, then at least one of them occurs (Proposition 3). However, they are not mutually exclusive.

The proof of Theorem 2 follows the strategy of [BCDHV26] and [2]. We adapt to closed hypercycles the notion of winding used in the previous works. One novelty is the use of the hyperbolic band model, in which the hypercycles around a given geodesic become horizontal lines, to obtain the necessary estimates.

The paper is organized as follows. In § 2 we recall the facts of hyperbolic geometry that will be used. In § 3 we prove the trichotomy (Proposition 3), we deduce Theorem 1 from Theorem 2 and the previous works, and we give an example of the third case. In § 4 we introduce the winding around a closed hypercycle and we prove the Key Proposition (Proposition 5), which controls the distance between a geodesic ray and its winding. § 5 is devoted to the proof of Theorem 2, where we construct uncountably many vectors in Wssu∖HssuW^{ssu}\setminus H^{ssu}.

Preliminaries

Throughout the paper we use the upper half-plane model H={z=x+iy∈C:y>0}\mathbb{H}=\{z=x+iy\in\mathbb{C}:y>0\} equipped with the metric

ds2=dx2+dy2y2.ds^{2}=\frac{dx^{2}+dy^{2}}{y^{2}}.

The Riemannian distance induced by this metric is denoted by dd. We denote the unit tangent bundle of H\mathbb{H} by T1HT^{1}\mathbb{H}, and the basepoint projection by π:T1H→H\pi:T^{1}\mathbb{H}\to\mathbb{H}.

Horocycles and boundary at infinity

The boundary at infinity of H\mathbb{H} is ∂H=R∪{∞}\partial\mathbb{H}=\mathbb{R}\cup\{\infty\}. Every geodesic ray in H\mathbb{H} converges to a unique point of ∂H\partial\mathbb{H}, and two rays define the same point at infinity if and only if they remain at bounded distance from each other.

Let ξ∈∂H\xi\in\partial\mathbb{H}, and let (c(t))t≥0(c(t))_{t\geq0} be a geodesic ray converging to ξ\xi. The Busemann function based at ξ\xi is defined by

Bξ(x,y)=lim⁡t→+∞(d(x,c(t))−d(y,c(t))),x,y∈H.B_{\xi}(x,y)=\lim_{t\to+\infty}\bigl(d(x,c(t))-d(y,c(t))\bigr),\qquad x,y\in\mathbb{H}.

This limit exists and does not depend on the choice of the geodesic ray cc asymptotic to ξ\xi. For fixed x∈Hx\in\mathbb{H}, the level sets of the map y↦Bξ(y,x)y\mapsto B_{\xi}(y,x) are called horocycles centered at ξ\xi. Geometrically, horocycles are circles tangent to the real line or horizontal lines.

For u∈T1Hu\in T^{1}\mathbb{H}, we denote by u(+∞)∈∂Hu(+\infty)\in\partial\mathbb{H} its forward endpoint. The stable horocycle through uu is the subset

Hssu={v∈T1H:v(+∞)=u(+∞) and Bu(+∞)(π(v),π(u))=0}.H^{ss}u=\{v\in T^{1}\mathbb{H}:v(+\infty)=u(+\infty)\text{ and }B_{u(+\infty)}(\pi(v),\pi(u))=0\}.

We will also denote

Hwsu={v∈T1H:v(+∞)=u(+∞)}.H^{ws}u=\{v\in T^{1}\mathbb{H}:v(+\infty)=u(+\infty)\}.

The following lemma can be proved by elementary computations using the formula for the hyperbolic distance [4 Theorem 7.2.1].

Lemma 1. Let u,v∈T1Hu,v\in T^{1}\mathbb{H} such that v∈Hwsuv\in H^{ws}u. Then d(v(t),u(t))d(v(t),u(t)) is non-increasing in t∈Rt\in\mathbb{R}.

Distance on the unit tangent bundle

We consider the distance d1d_1 on the unit tangent bundle of T1HT^1\mathbb{H} defined as follows [1]:

d1(v~,w~):=d(v~(0),w~(0))+d(v~(1),w~(1)),v~,w~∈T1H.d_1(\widetilde{v},\widetilde{w}) := d(\widetilde{v}(0),\widetilde{w}(0)) + d(\widetilde{v}(1),\widetilde{w}(1)), \qquad\widetilde{v},\widetilde{w} \in T^1\mathbb{H}.

Proposition 1. The distance d1d_1 is equivalent to the Sasaki distance on the unit tangent bundle T1HT^1\mathbb{H}.

Fuchsian groups

Throughout the paper, Γ\Gamma denotes a torsion-free Fuchsian group acting properly discontinuously by isometries on H\mathbb{H}. We consider the quotient surface S=Γ\HS = \Gamma\backslash\mathbb{H} and its unit tangent bundle T1S=Γ\T1HT^1S = \Gamma\backslash T^1\mathbb{H}. We use a tilde to denote lifts to the universal cover: given u∈T1Su \in T^1S, we write u~∈T1H\widetilde{u} \in T^1\mathbb{H} for a unit tangent vector projecting onto uu under the canonical projection T1H→T1ST^1\mathbb{H} \to T^1S.

The sets HssuH^{ssu} and HwsuH^{wsu} are Γ\Gamma-equivariant and therefore descend to well-defined subsets in T1ST^1S, which we still denote by HssuH^{ssu} and HwsuH^{wsu}.

On the unit tangent bundle T1ST^1S of a hyperbolic surface S=Γ\HS = \Gamma\backslash\mathbb{H}, the distance d1d_1 descends as follows:

d1(u,v)=inf⁡γ∈Γ{d1(u~,γv~)}.d_1(u,v) = \inf_{\gamma\in\Gamma}\{d_1(\widetilde{u},\gamma\widetilde{v})\}.

where u~\widetilde{u} and v~\widetilde{v} are lifts of uu and vv to T1HT^1\mathbb{H}. This distance is again equivalent to the Sasaki metric on T1ST^1S.

The strong-stable set of a vector u∈T1Su \in T^1S is defined as

Wssu={v∈T1S:lim⁡t→+∞d1(gtu,gtv)=0}.W^{ssu} = \{v \in T^1S : \lim_{t\to+\infty} d_1(g_tu,g_tv) = 0\}.

We remark that, since the distance d1d_1 is equivalent to the Sasaki distance, they give the same notion of strong-stable set.

Hypercycles and hyperbolic isometries

Let γ\gamma be a hyperbolic isometry of H\mathbb{H}. It fixes exactly two points in ∂H\partial\mathbb{H}, the repelling fixed point γ−\gamma^- and the attracting fixed point γ+\gamma^+. The geodesic cc joining γ−\gamma^- to γ+\gamma^+ is the axis of γ\gamma, and γ\gamma acts on cc as a translation of length

ℓ(γ)=min⁡z∈Hd(z,γz).\ell(\gamma) = \min_{z\in\mathbb{H}} d(z,\gamma z).

The displacement of a point z∈Hz \in\mathbb{H} by γ\gamma only depends on its distance to the axis [4 Theorem 7.35.1]:

sinh⁡(d(z,γz)2)=cosh⁡(d(z,c))sinh⁡(ℓ(γ)2).(1)\sinh\left(\frac{d(z,\gamma z)}{2}\right) = \cosh(d(z,c))\sinh\left(\frac{\ell(\gamma)}{2}\right). \tag*{(1)}

In particular, the displacement is minimal along the axis and increases with the distance to the axis.

For r>0r > 0, the set of points of H\mathbb{H} at distance rr from cc consists of two curves, one on each side of cc, called hypercycles of γ\gamma. Each of them is invariant under γ\gamma. In the upper half-plane model, they are arcs of Euclidean circles, or Euclidean half-lines, joining γ−\gamma^- to γ+\gamma^+. Given a point q∈Hq \in\mathbb{H}, we denote by H(γ,q)\mathcal{H}(\gamma,q) the hypercycle of γ\gamma through qq.

Proposition 2. Let c1,c2c_1, c_2 be two geodesics in H\mathbb{H} such that d(c1,c2)>0d(c_1,c_2) > 0. Then there exist unique p=c1(t1)p = c_1(t_1), q=c2(t2)q = c_2(t_2) realizing the distance between c1c_1 and c2c_2. Moreover, there is an hypercycle at distance d(c1,c2)d(c_1,c_2) from c1c_1 that is tangent to c2c_2 at qq.

If γ\gamma belongs to Γ\Gamma, the hypercycle H(γ,q)\mathcal{H}(\gamma,q) projects to a closed curve in S=Γ\HS = \Gamma\backslash\mathbb{H}, which we call a closed hypercycle. A direct computation shows that the arc of H(γ,q)\mathcal{H}(\gamma,q) between qq and γq\gamma q has length

ℓ(γ)cosh⁡d(q,c).(2)\ell(\gamma)\cosh d(q,c). \tag*{(2)}

If γ\gamma is primitive in Γ\Gamma, this is the length of the closed hypercycle of γ\gamma.

The band model

For some computations it is convenient to use the band model

B={x+iy∈C:−π2<y<π2},ds2=dx2+dy2cos⁡2(y).\mathbb{B} = \left\{x+iy \in\mathbb{C} : -\frac{\pi}{2} < y < \frac{\pi}{2}\right\}, \qquad ds^2 = \frac{dx^2+dy^2}{\cos^2(y)}.

The map z↦iezz \mapsto ie^z is an isometry from B\mathbb{B} onto H\mathbb{H}, which sends the real line {y=0}\{y=0\} onto the imaginary axis. The boundary at infinity of B\mathbb{B} consists of the two lines {y=±π/2}\{y=\pm\pi/2\} together with the two points ±∞\pm\infty. We will use the following facts.

  1. The real line {y=0}\{y=0\} is a geodesic. For λ>0\lambda>0, the translation z↦z+λz\mapsto z+\lambda is a hyperbolic isometry with axis {y=0}\{y=0\} and translation length λ\lambda.

  1. The vertical segments {x=k}\{x=k\}, k∈Rk\in\mathbb{R}, are geodesics orthogonal to {y=0}\{y=0\}.

  1. The horizontal lines {y=k}\{y=k\}, k∈(−π/2,π/2)∖{0}k\in(-\pi/2,\pi/2)\setminus\{0\}, are hypercycles. More precisely, for x∈Rx\in\mathbb{R} and y∈(−π/2,π/2)y\in(-\pi/2,\pi/2),

cosh⁡d(x,x+iy)=1cos⁡(y).\cosh d(x,x+iy)=\frac{1}{\cos(y)}.
  1. Equations of the geodesics relative to {y=0}\{y=0\} (see Figure 1). Every geodesic other than {y=0}\{y=0\} is of one of the following three types.

Geodesics of the band model

Figure 1. Geodesics of the band model B\mathbb{B}. The labels refer to the items of Section 2.5. The geodesics of item (4a) are tangent to the hypercycles {y=±y0}\{y=\pm y_0\} at their points closest to {y=0}\{y=0\}, and the vertical geodesic through one of these points is the common perpendicular. The geodesics of item (4c) are asymptotic to {y=0}\{y=0\}.

(a) Ultraparallel geodesics. A geodesic at positive distance from {y=0}\{y=0\} has an equation of the form

cosh⁡(x−x0)=αsin⁡(y),x0∈R, ∣α∣>1.\cosh(x-x_0)=\alpha\sin(y), \qquad x_0\in\mathbb{R},\ |\alpha|>1.

Its endpoints are x0±r+iϵπ/2x_0\pm r+i\epsilon\pi/2, where ϵ\epsilon is the sign of α\alpha and cosh⁡r=∣α∣\cosh r=|\alpha|, and its point closest to {y=0}\{y=0\} is x0+iy0x_0+iy_0, where sin⁡y0=1/α\sin y_0=1/\alpha.

(b) Intersecting geodesics. A geodesic intersecting {y=0}\{y=0\} has an equation of the form

sinh⁡(x−x0)=βsin⁡(y),x0∈R, β∈R.\sinh(x-x_0)=\beta\sin(y), \qquad x_0\in\mathbb{R},\ \beta\in\mathbb{R}.

Its endpoints are x0+s+iπ/2x_0+s+i\pi/2 and x0−s−iπ/2x_0-s-i\pi/2, where sinh⁡s=β\sinh s=\beta. It meets {y=0}\{y=0\} at x0x_0, and the angle θ∈(0,π)\theta\in(0,\pi) from the positive direction of {y=0}\{y=0\} to the geodesic satisfies cot⁡θ=β\cot\theta=\beta. The case β=0\beta=0 gives the geodesics of item (2).

(c) Asymptotic geodesics. A geodesic asymptotic to {y=0}\{y=0\} has an equation of the form

sin⁡(y)=ϵe−δ(x−x0),x0∈R, ϵ,δ∈{−1,1}.\sin(y)=\epsilon e^{-\delta(x-x_0)}, \qquad x_0\in\mathbb{R},\ \epsilon,\delta\in\{-1,1\}.

Its endpoints are x0+iϵπ/2x_0+i\epsilon\pi/2 and +∞+\infty if δ=1\delta=1, or −∞-\infty if δ=−1\delta=-1.

Vanishing asymptotic fineness

Given x∈Sx \in S, we denote by Inj⁡(x)\operatorname{Inj}(x) the injectivity radius of SS at xx. The next proposition reduces the study of the vectors whose geodesic ray has vanishing injectivity radius, without diverging through a cusp, to three cases. Each case is governed by the geometry of the thin parts visited by the ray: short closed geodesics, cusps, or short geodesic loops enclosed by hypercycles.

Proposition 3. Let S=Γ\HS = \Gamma\backslash\mathbb{H} be a hyperbolic surface and let u∈T1Su \in T^1S. Assume that the geodesic ray u(R+)u(\mathbb{R}^{+}) does not diverge through a cusp and that its injectivity radius is not bounded below, that is,

inf⁡t≥0Inj⁡(u(t))=0.\inf_{t \ge0} \operatorname{Inj}(u(t)) = 0.

Then at least one of the following holds:

  1. u(R+)u(\mathbb{R}^{+}) crosses an infinite sequence of closed geodesics whose lengths tend to 00,

  2. u(R+)u(\mathbb{R}^{+}) intersects an infinite sequence of closed horocycles whose lengths tend to 00,

  3. u(R+)u(\mathbb{R}^{+}) intersects an infinite sequence of closed hypercycles whose lengths tend to 00 but it does not meet the associated closed geodesics.

Proof. Since inf⁡t≥0Inj⁡(u(t))=0\inf_{t \ge0} \operatorname{Inj}(u(t)) = 0, there exists a sequence of times tn≥0t_n \ge0 with Inj⁡(u(tn))→0\operatorname{Inj}(u(t_n)) \to0.

Fix a lift u~∈T1H\widetilde{u} \in T^1\mathbb{H} of uu. The injectivity radius at u(tn)u(t_n) equals half the length of the shortest geodesic loop based at u(tn)u(t_n), so there exists γn∈Γ∖{id⁡}\gamma_n \in\Gamma\setminus\{\operatorname{id}\} such that

εn:=d(u~(tn),γnu~(tn))=2Inj⁡(u(tn))⟶0.\varepsilon_n := d\bigl(\widetilde{u}(t_n),\gamma_n\widetilde{u}(t_n)\bigr) = 2\operatorname{Inj}(u(t_n)) \longrightarrow0.

Since Γ\Gamma is torsion free, each γn\gamma_n is either parabolic or hyperbolic. After passing to a subsequence, either all the γn\gamma_n are parabolic, or all of them are hyperbolic.

Parabolic case. Assume that every γn\gamma_n is parabolic. Let xn∈∂Hx_n \in\partial\mathbb{H} be its fixed point and let H~n\widetilde{H}_n be the horocycle centered at xnx_n passing through u~(tn)\widetilde{u}(t_n). Denote by ℓ(H~n,γn)\ell(\widetilde{H}_n,\gamma_n) the length of the horocyclic arc of H~n\widetilde{H}_n between u~(tn)\widetilde{u}(t_n) and γnu~(tn)\gamma_n\widetilde{u}(t_n). A direct computation in the upper half-plane model shows that

sinh⁡(εn2)=ℓ(H~n,γn)2,\sinh\left(\frac{\varepsilon_n}{2}\right) = \frac{\ell(\widetilde{H}_n,\gamma_n)}{2},

hence ℓ(H~n,γn)→0\ell(\widetilde{H}_n,\gamma_n) \to0. The horocycle H~n\widetilde{H}_n projects to a closed horocycle of SS of length at most ℓ(H~n,γn)\ell(\widetilde{H}_n,\gamma_n) passing through u(tn)u(t_n). So u(R+)u(\mathbb{R}^{+}) intersects a sequence of closed horocycles whose lengths tend to 00. As u(R+)u(\mathbb{R}^{+}) does not diverge through a cusp, this shows that uu is cusp-recurrent in the sense of [2], and we are in case (2).

Hyperbolic case. Assume that every γn\gamma_n is hyperbolic. Let cnc_n be its axis and ℓ(γn)\ell(\gamma_n) its translation length. Clearly, ℓ(γn)≤d(u~(tn),γnu~(tn))=εn\ell(\gamma_n) \le d\bigl(\widetilde{u}(t_n),\gamma_n\widetilde{u}(t_n)\bigr) = \varepsilon_n tends to 00 when n→∞n \to\infty.

We distinguish two subcases.

Subcase (i): u(R+)u(\mathbb{R}^{+}) meets the closed geodesic obtained by projecting cnc_n onto SS for infinitely many nn. Passing to a subsequence, this holds for every nn. These closed geodesics have length at most ℓ(γn)\ell(\gamma_n). Since ℓ(γn)→0\ell(\gamma_n) \to0, we are in case (1).

Subcase (ii): u(R+)u(\mathbb{R}^{+}) meets the closed geodesic obtained by projecting cnc_n onto SS for at most finitely many nn. Passing to a subsequence, u(R+)u(\mathbb{R}^{+}) does not meet these closed geodesics for any nn. The hypercycle H(γn,u~(tn))\mathcal{H}(\gamma_n,\widetilde{u}(t_n)) projects to a closed hypercycle in SS passing through u(tn)u(t_n). Let γn′\gamma'_n be a generator of the stabilizer of cnc_n in Γ\Gamma, so that ℓ(γn′)≤ℓ(γn)\ell(\gamma'_n) \le\ell(\gamma_n). By (2.2), the length of the projection of H(γn,u~(tn))\mathcal{H}(\gamma_n,\widetilde{u}(t_n)) is

ℓ(γn′)cosh⁡(d(u~(tn),cn)).(3)\ell(\gamma'_n)\cosh\bigl(d(\widetilde{u}(t_n),c_n)\bigr). \tag*{(3)}

Using the inequality x≤2sinh⁡(x/2)x \le2\sinh(x/2) and (1), we obtain

ℓ(γn′)cosh⁡(d(u~(tn),cn))≤ℓ(γn)cosh⁡(d(u~(tn),cn))≤2sinh⁡(εn2),\ell(\gamma'_n)\cosh(d(\widetilde{u}(t_n),c_n)) \le\ell(\gamma_n)\cosh(d(\widetilde{u}(t_n),c_n)) \le2\sinh\left(\frac{\varepsilon_n}{2}\right),

which tends to 00 when n→+∞n \to+\infty. Hence we are in case (3).

We now deduce Theorem 1 from Proposition 3, Theorem 2 and the previous works.

Proof of Theorem 1. Let SS be a hyperbolic surface and u∈T1Su \in T^1S. Since Inj⁡\operatorname{Inj} is continuous and positive on SS, the asymptotic fineness of u(R+)u(\mathbb{R}^{+}) vanishes if and only if inf⁡t≥0Inj⁡(u(t))=0\inf_{t\ge0}\operatorname{Inj}(u(t))=0.

Assume first that the asymptotic fineness of u(R+)u(\mathbb{R}^{+}) is positive. Then Bellis proved that Huss=WussH^{ss}_u=W^{ss}_u [5 Théorème D]. If the geodesic ray u(R+)u(\mathbb{R}^{+}) diverges through a cusp, then HussH^{ss}_u is a closed horocycle around that cusp and the equality Huss=WussH^{ss}_u=W^{ss}_u is clear.

Assume now that the asymptotic fineness of u(R+)u(\mathbb{R}^{+}) vanishes and that u(R+)u(\mathbb{R}^{+}) does not diverge through a cusp. By Proposition 3, at least one of the following cases occurs.

  1. The ray u(R+)u(\mathbb{R}^{+}) crosses closed geodesics whose lengths tend to 00. The conclusion follows from [BCDHV26, Main Theorem].

  1. The ray u(R+)u(\mathbb{R}^{+}) intersects closed horocycles whose lengths tend to 00, so uu is cusp-recurrent and the conclusion follows from [2 Theorem 2].

  1. The ray u(R+)u(\mathbb{R}^{+}) intersects closed hypercycles whose lengths tend to 00. The conclusion follows from Theorem 2.

In all cases, Huss⊊WussH^{ss}_u \subsetneq W^{ss}_u and WussW^{ss}_u is an uncountable union of stable horocycles.

The following example shows that case (3) of Proposition 3 can occur while cases (1) and (2) do not.

Example 3. There exist a hyperbolic surface SS without cusps and a vector u∈T1Su \in T^1S such that

  • lim inf⁡t≥0Inj⁡(u(t))=0\liminf_{t\ge0}\operatorname{Inj}(u(t))=0,

  • for every sequence of closed geodesics (ck)(c_k) with ℓ(ck)→0\ell(c_k)\to0, we have d(ck,u(R+))→+∞d(c_k,u(\mathbb{R}^{+}))\to+\infty.

Let ℓn\ell_n be a sequence of positive numbers converging to 00 and let D>0D>0. Let PnP_n be the pair of pants with boundary geodesic components of lengths ℓn\ell_n, DD and DD [Bus92, Chapter 3]. For n≥0n\ge0, PnP_n is glued to Pn+1P_{n+1} through a geodesic boundary of length DD. Then funnels are attached to the remaining boundary components. We thus obtain a complete hyperbolic surface with no cusps called a flute surface.

Let αn\alpha_n be the boundary component of PnP_n of length ℓn\ell_n. Let knk_n be a sequence of integers. We consider a vector u∈T1Su\in T^1S whose geodesic ray u(R+)u(\mathbb{R}^{+}) has the following behavior:

  • it starts in P0P_0, that is, u(0)∈P0u(0)\in P_0,

  • it visits the pairs of pants P0,P1,P2,…P_0,P_1,P_2,\ldots in this order,

  • for every n≥1n\ge1, in PnP_n the ray turns knk_n times around αn\alpha_n and does not turn around the other two boundary geodesics.

By choosing the numbers of turns knk_n appropriately, we can guarantee that the distance between u(R+)u(\mathbb{R}^{+}) and αn\alpha_n tends to infinity, while u(R+)u(\mathbb{R}^{+}) visits points of the collar of αn\alpha_n where the injectivity radius is small, so that the injectivity radius along the ray vanishes.

To conclude one shows that there exists a constant ε0>0\varepsilon_0>0 such that every closed geodesic of SS distinct from the geodesics αn\alpha_n has length at least $\varepsilon_0.

Winding around a closed hypercycle

Definition 4. Let γ\gamma be an hyperbolic isometry of H\mathbb{H} with axis cc and let α:R→H\alpha:\mathbb{R}\to\mathbb{H} be a unit speed parametrization of a hypercycle of cc. The orientation of α\alpha is positive with respect to γ\gamma if there exists D>0D>0 such that for all t∈Rt\in\mathbb{R},

γα(t)=α(t+D).\gamma\alpha(t)=\alpha(t+D).

Definition 5. Let γ\gamma be an hyperbolic isometry of H\mathbb{H} with axis cc and let u~∈T1H\tilde{u}\in T^{1}\mathbb{H} such that d(c,u~(R))>0d(c,\tilde{u}(\mathbb{R}))>0. Let q=u~(t1)q=\tilde{u}(t_{1}) be the point in u~(R)\tilde{u}(\mathbb{R}) that realizes the distance to cc. Let α:R→H\alpha:\mathbb{R}\to\mathbb{H} be the positive unit speed parametrization of the hypercycle of cc through qq such that α(0)=q\alpha(0)=q. We say that u~\tilde{u} is in good position with respect to γ\gamma if t1>0t_{1}>0 and

du~dt(t1)=dαdt(0).\frac{d\tilde{u}}{dt}(t_{1})=\frac{d\alpha}{dt}(0).

Definition 6. Let u~∈T1H\tilde{u}\in T^{1}\mathbb{H} be in good position with respect to a hyperbolic isometry γ\gamma. We define the winding of u~\tilde{u} around γ\gamma (see Figure 2), denoted by Wind⁡γ(u~)∈T1H\operatorname{Wind}_{\gamma}(\tilde{u})\in T^{1}\mathbb{H}, as the unique vector satisfying

  • Wind⁡γ(u~)(0)=u~(0)\operatorname{Wind}_{\gamma}(\tilde{u})(0)=\tilde{u}(0).

  • Wind⁡γ(u~)(+∞)=γu~(+∞)\operatorname{Wind}_{\gamma}(\tilde{u})(+\infty)=\gamma\tilde{u}(+\infty).

The winding of a vector around a hyperbolic isometry in the disk model

Figure 2. The winding Wind⁡γ(u~)\operatorname{Wind}_{\gamma}(\tilde{u}) of u~\tilde{u} around γ\gamma, in the disk model.

Definition 7. The winding time associated to γ\gamma is the real number τγ,u~\tau_{\gamma,\tilde{u}} defined by

τγ,u~=Bu~(+∞)(γ−1u~(0),u~(0)).\tau_{\gamma,\tilde{u}}=B_{\tilde{u}(+\infty)}(\gamma^{-1}\tilde{u}(0),\tilde{u}(0)).

Proposition 4 (bound of the winding time). Let u~∈T1H\tilde{u}\in T^{1}\mathbb{H} be in good position with respect to a hyperbolic isometry γ\gamma. Let q=u~(t1)q=\tilde{u}(t_{1}) be the point in u~(R)\tilde{u}(\mathbb{R}) that realizes the distance to the axis cc of γ\gamma. Then

∣τγ,u~∣≤d(q,γq).\lvert\tau_{\gamma,\tilde{u}}\rvert\leq d(q,\gamma q).

Proof. As qq belongs to u~(R+)\tilde{u}(\mathbb{R}^{+}), we have

Bu~(+∞)(q,u~(0))=−d(q,u~(0)).B_{\tilde{u}(+\infty)}(q,\tilde{u}(0))=-d(q,\tilde{u}(0)).

Thus

Bu~(+∞)(γ−1u~(0),u~(0))=Bu~(+∞)(γ−1u~(0),γ−1q)+Bu~(+∞)(γ−1q,q)+Bu~(+∞)(q,u~(0))≤d(γ−1u~(0),γ−1q)+d(γ−1q,q)−d(q,u~(0))=d(q,γq).\begin{aligned} B_{\tilde u(+\infty)}(\gamma^{-1}\tilde u(0),\tilde u(0)) \\ &= B_{\tilde u(+\infty)}(\gamma^{-1}\tilde u(0),\gamma^{-1}q)+B_{\tilde u(+\infty)}(\gamma^{-1}q,q)+B_{\tilde u(+\infty)}(q,\tilde u(0)) \\ &\le d(\gamma^{-1}\tilde u(0),\gamma^{-1}q)+d(\gamma^{-1}q,q)-d(q,\tilde u(0)) \\ &= d(q,\gamma q). \end{aligned}

For the other inequality, we observe that γu~(+∞)\gamma\tilde u(+\infty) is between u~(+∞)\tilde u(+\infty) and the attracting fixed point xγ+x_\gamma^+ of γ\gamma. Then the geodesic ray [u~(0),γu~(+∞))[\tilde u(0),\gamma\tilde u(+\infty)) is closer than u~(R+)\tilde u(\mathbb{R}^+) to the axis cc of γ\gamma. Let q′q' be the point of [u~(0),γu~(+∞))[\tilde u(0),\gamma\tilde u(+\infty)) closest to cc. We have d(q′,c)≤d(q,c)d(q',c)\le d(q,c). Then d(q′,γq′)≤d(q,γq)d(q',\gamma q')\le d(q,\gamma q).

Since q′q' is in the geodesic ray [u~(0),γu~(+∞))[\tilde u(0),\gamma\tilde u(+\infty)), we have

Bu~(+∞)(γ−1u~(0),γ−1q′)=Bγu~(+∞)(u~(0),q′)=d(u~(0),q′).B_{\tilde u(+\infty)}(\gamma^{-1}\tilde u(0),\gamma^{-1}q')=B_{\gamma\tilde u(+\infty)}(\tilde u(0),q')=d(\tilde u(0),q').

Therefore

Bu~(+∞)(γ−1u~(0),u~(0))=Bu~(+∞)(γ−1u~(0),γ−1q′)+Bu~(+∞)(γ−1q′,q′)+Bu~(+∞)(q′,u~(0))≥d(u~(0),q′)−d(γ−1q′,q′)−d(q′,u~(0))=−d(q′,γq′)≥−d(q,γq).\begin{aligned} B_{\tilde u(+\infty)}(\gamma^{-1}\tilde u(0),\tilde u(0)) \\ &= B_{\tilde u(+\infty)}(\gamma^{-1}\tilde u(0),\gamma^{-1}q')+B_{\tilde u(+\infty)}(\gamma^{-1}q',q')+B_{\tilde u(+\infty)}(q',\tilde u(0)) \\ &\ge d(\tilde u(0),q')-d(\gamma^{-1}q',q')-d(q',\tilde u(0)) \\ &=-d(q',\gamma q')\ge-d(q,\gamma q). \end{aligned}

Proposition 5 (Key Proposition). Let u~∈T1H\tilde u\in T^1\mathbb{H} be in good position with respect to a hyperbolic isometry γ\gamma. Let qq be the point in u~(R)\tilde u(\mathbb{R}) closest to the axis of γ\gamma. Assume that d(q,γq)≤1d(q,\gamma q)\le1. Then, for every t≥0t\ge0,

min⁡{d1(gt+τγ,u~(Wind⁡γ(u~)),gtu~),d1(gt+τγ,u~(Wind⁡γ(u~)),γgtu~)}≤17d(q,γq).\min\{d_1(g_{t+\tau_{\gamma,\tilde u}}(\operatorname{Wind}_{\gamma}(\tilde u)),g_t\tilde u),d_1(g_{t+\tau_{\gamma,\tilde u}}(\operatorname{Wind}_{\gamma}(\tilde u)),\gamma g_t\tilde u)\}\le17d(q,\gamma q).

Proof. For simplicity, we denote v~=Wind⁡γ(u~)\tilde v=\operatorname{Wind}_{\gamma}(\tilde u) and τ=τγ,u~\tau=\tau_{\gamma,\tilde u}. Let q=u~(t1)q=\tilde u(t_1), t1≥0t_1\ge0, be the point that realizes the distance of u~(R)\tilde u(\mathbb{R}) to the axis of γ\gamma. We observe that u~(0)\tilde u(0) and γu~(+∞)\gamma\tilde u(+\infty) lie in the same half-plane in H∖(γ−,γ+)\mathbb{H}\setminus(\gamma^-,\gamma^+). Hence, the geodesic generated by v~\tilde v does not intersect the axis of γ\gamma. Let q′=v~(t1′)q'=\tilde v(t'_1), t1′∈Rt'_1\in\mathbb{R}, be the point that realizes the distance of v~(R)\tilde v(\mathbb{R}) to the axis of γ\gamma. The first step is to control the distance between qq and q′q'.

To do the next computations we use the band model B\mathbb{B} (see § 2.5). We can assume that the axis of γ\gamma is the geodesic y=0y=0 and that γ\gamma has the form γ(z)=z+λ\gamma(z)=z+\lambda, where λ>0\lambda>0 is the translation length of γ\gamma.

The geodesics at positive distance from y=0y=0 are those of item (4a). Without loss of generality, we can assume that the equation of u~(R)\tilde u(\mathbb{R}) has x0=0x_0=0 and α>1\alpha>1. Hence,

u~(R)={(x,y)∈R×(−π/2,π/2)∣cosh⁡(x)=αsin⁡(y)},\tilde u(\mathbb{R})=\{(x,y)\in\mathbb{R}\times(-\pi/2,\pi/2)\mid\cosh(x)=\alpha\sin(y)\},

which lies in the upper half-strip y>0y>0. The endpoints at infinity of u~(R)\tilde u(\mathbb{R}) are

u~(−∞)=−r+π/2i,u~(+∞)=r+π/2i,\tilde u(-\infty)=-r+\pi/2i,\qquad\tilde u(+\infty)=r+\pi/2i,

where r>0r>0 satisfies cosh⁡(r)=α\cosh(r)=\alpha. The point qq realizing the distance to the axis y=0y=0 is q=y0iq=y_0i with α=1/sin⁡(y0)\alpha=1/\sin(y_0). The initial point u~(0)\tilde u(0) lies between u~(−∞)\tilde u(-\infty) and qq, so it has negative real part.

The geodesic v~(R)\tilde v(\mathbb{R}) is also at positive distance from the axis y=0y=0 and in the upper half-strip. Let

cosh⁡(x−x0′)=α′sin⁡(y),x0′∈R, α′>1,\cosh(x-x'_0)=\alpha'\sin(y),\qquad x'_0\in\mathbb{R},\ \alpha'>1,

be the equation of v~(R)\tilde{v}(\mathbb{R}). The endpoints of v~(R)\tilde{v}(\mathbb{R}) are

v~(−∞)=x0′−r′+π/2i,v~(+∞)=x0′+r′+π/2i,\tilde{v}(-\infty)=x'_0-r'+\pi/2i,\qquad\tilde{v}(+\infty)=x'_0+r'+\pi/2i,

where r′>0r'>0 satisfies cosh⁡(r′)=α′\cosh(r')=\alpha'. The point q′q' realizing the distance to the axis is q′=x0′+y0′iq'=x'_0+y'_0i with α′=1/sin⁡(y0′)\alpha'=1/\sin(y'_0).

By definition, v~(+∞)=γu~(+∞)\tilde{v}(+\infty)=\gamma\tilde{u}(+\infty), which implies r+λ=x0′+r′r+\lambda=x'_0+r'. The point v~(−∞)\tilde{v}(-\infty) must lie between −r+π/2i-r+\pi/2i and π/2i\pi/2i.

From the position of the endpoints of u~(R)\tilde{u}(\mathbb{R}) and v~(R)\tilde{v}(\mathbb{R}), we obtain the following estimates

(1) 0<y0′<y00<y'_0<y_0,

(2) 0<r′−r<λ/20<r'-r<\lambda/2,

(3) 0<x0′<λ0<x'_0<\lambda.

Consider the function f(t)=arcsin⁡(1/cosh⁡(t))f(t)=\arcsin(1/\cosh(t)) defined for t∈Rt\in\mathbb{R}. We have f(r)=y0f(r)=y_0 and f(r′)=y0′f(r')=y'_0. An easy computation shows that ff is 1-Lipschitz, hence

∣y0−y0′∣≤∣r−r′∣<λ2.|y_0-y'_0|\le|r-r'|<\frac{\lambda}{2}.

Introducing q′′=x0′+y0iq''=x'_0+y_0i, we have

d(q,q′)≤d(q,q′′)+d(q′′,q′).d(q,q')\le d(q,q'')+d(q'',q').

First, we notice that γq=λ+y0i\gamma q=\lambda+y_0i. Since 0<x0′<λ0<x'_0<\lambda, the point q′′=x0′+y0iq''=x'_0+y_0i belongs to the hypercycle segment between qq and γq\gamma q. This implies that d(q,q′′)≤d(q,γq)d(q,q'')\le d(q,\gamma q).

The points q′′q'' and q′q' lie on the vertical geodesic {x=x0′}\{x=x'_0\} (item (2)). Hence d(q′′,q′)d(q'',q') is the length of the vertical segment joining them and, since 0<y0′<y00<y'_0<y_0,

d(q′′,q′)=∫y0′y0dtcos⁡(t)≤y0−y0′cos⁡(y0)≤λ/2cos⁡(y0).d(q'',q')=\int_{y'_0}^{y_0}\frac{dt}{\cos(t)}\le\frac{y_0-y'_0}{\cos(y_0)}\le\frac{\lambda/2}{\cos(y_0)}.

By item (3), cosh⁡(d(0,q))=1/cos⁡(y0)\cosh(d(0,q))=1/\cos(y_0). Since d(0,q)d(0,q) is the distance from qq to the axis of γ\gamma, (1) gives

sinh⁡(d(q,γq)2)=cosh⁡(d(0,q))sinh⁡(λ2).\sinh\left(\frac{d(q,\gamma q)}{2}\right)=\cosh(d(0,q))\sinh\left(\frac{\lambda}{2}\right).

These together imply

d(q′′,q′)≤λ/2cos⁡(y0)≤sinh⁡(λ/2)cosh⁡(d(0,q))=sinh⁡(d(q,γq)2).d(q'',q')\le\frac{\lambda/2}{\cos(y_0)}\le\sinh(\lambda/2)\cosh(d(0,q))=\sinh\left(\frac{d(q,\gamma q)}{2}\right).

If 0≤x≤10\le x\le1 then sinh⁡(x/2)≤x\sinh(x/2)\le x. Since d(q,γq)≤1d(q,\gamma q)\le1, we obtain

d(q,q′)≤2d(q,γq).d(q,q')\le2d(q,\gamma q).

The rest of the proof follows from the bound of the distance between qq and q′q'. Recall that q=u~(t1)q=\tilde{u}(t_1) and q′=v~(t1′)q'=\tilde{v}(t'_1). By the triangular inequality,

∣t1−t1′∣≤d(q,q′)≤2d(q,γq).|t_1-t'_1|\le d(q,q')\le2d(q,\gamma q).

We now bound the distance between the basepoints of the geodesics of u~\tilde{u} and v~\tilde{v}.

Case 1. If 0≤t≤t10\le t\le t_1, then

d(v~(t),u~(t))≤4d(q,γq).d(\tilde{v}(t),\tilde{u}(t))\le4d(q,\gamma q).

Since the distance in H\mathbb{H} between two geodesic rays starting at the same point is increasing, it follows

∀ 0≤t≤t1,d(v~(t),u~(t))≤d(v~(t1),u~(t1)).\forall\,0\le t\le t_1,\qquad d(\tilde{v}(t),\tilde{u}(t))\le d(\tilde{v}(t_1),\tilde{u}(t_1)).

Using the bounds established above, one obtains

d(v~(t1),u~(t1))≤∣t1−t1′∣+d(v~(t1′),u~(t1))≤2d(q′,q)≤4d(q,γq),d(\tilde{v}(t_1),\tilde{u}(t_1))\le|t_1-t'_1|+d(\tilde{v}(t'_1),\tilde{u}(t_1))\le2d(q',q)\le4d(q,\gamma q),

which implies the statement.

Case 2: If t≥t1t \ge t_1, then

d(v~(t),γu~(t))≤5d(q,γq).d(\tilde{v}(t),\gamma\tilde{u}(t)) \le5d(q,\gamma q).

Since v~\tilde{v} and γu~\gamma\tilde{u} have the same point at infinity, the distance d(v~(t),γu~(t))d(\tilde{v}(t),\gamma\tilde{u}(t)) is decreasing (Lemma 1), and hence for t≥t1t \ge t_1,

d(v~(t),γu~(t))≤d(v~(t1),γu~(t1)).d(\tilde{v}(t),\gamma\tilde{u}(t)) \le d(\tilde{v}(t_1),\gamma\tilde{u}(t_1)).

Moreover,

d(v~(t1),γu~(t1))≤∣t1−t1′∣+d(v~(t1′),u~(t1′))≤∣t1−t1′∣+d(q′,q)+d(q,γq)d(\tilde{v}(t_1),\gamma\tilde{u}(t_1)) \le|t_1-t_1'|+d(\tilde{v}(t_1'),\tilde{u}(t_1')) \le|t_1-t_1'|+d(q',q)+d(q,\gamma q)

By the bounds proved above, the first two terms on the right hand side are bounded by 2d(q,γq)2d(q,\gamma q). We conclude that, for t≥t1t \ge t_1, d(v~(t),γu~(t))≤5d(q,γq)d(\tilde{v}(t),\gamma\tilde{u}(t)) \le5d(q,\gamma q).

Let us now compute the bounds for the distance d1d_1. Recall that

d1(gtv~,gtu~)=d(v~(t),u~(t))+d(v~(t+1),u~(t+1)).d_1(g_t\tilde{v},g_t\tilde{u})=d(\tilde{v}(t),\tilde{u}(t))+d(\tilde{v}(t+1),\tilde{u}(t+1)).

We can distinguish three scenarios:

(a) If t≤t1−1t \le t_1-1, by the first case, we have that both

d(v~(t),u~(t))andd(v~(t+1),u~(t+1))d(\tilde{v}(t),\tilde{u}(t)) \quad\text{and} \quad d(\tilde{v}(t+1),\tilde{u}(t+1))

are bounded by 4d(q,γq)4d(q,\gamma q). Then

d1(gtv~,gtu~)≤8d(q,γq).d_1(g_t\tilde{v},g_t\tilde{u}) \le8d(q,\gamma q).

(b) If t≥t1t \ge t_1, by the second case, we can bound both

d(v~(t),γu~(t))andd(v~(t+1),γu~(t+1))d(\tilde{v}(t),\gamma\tilde{u}(t)) \quad\text{and} \quad d(\tilde{v}(t+1),\gamma\tilde{u}(t+1))

by 5d(q,γq)5d(q,\gamma q). Then

d1(gtv~,γgtu~)≤10d(q,γq).d_1(g_t\tilde{v},\gamma g_t\tilde{u}) \le10d(q,\gamma q).

(c) Let t1−1<t<t1t_1-1<t<t_1. We have

d1(gtv~,gtu~)=d(v~(t),u~(t))+d(v~(t+1),u~(t+1))≤d(v~(t),u~(t))+d(v~(t+1),γu~(t+1))+d(γu~(t+1),u~(t+1)).\begin{aligned} d_1(g_t\tilde{v},g_t\tilde{u})=d(\tilde{v}(t),\tilde{u}(t))+d(\tilde{v}(t+1),\tilde{u}(t+1)) \\ &\le d(\tilde{v}(t),\tilde{u}(t))+d(\tilde{v}(t+1),\gamma\tilde{u}(t+1))+d(\gamma\tilde{u}(t+1),\tilde{u}(t+1)). \end{aligned}

The first and the second terms in the last line are bounded by 4d(q,γq)4d(q,\gamma q) and 5d(q,γq)5d(q,\gamma q) by Cases 1 and 2 above, respectively.

In order to bound the third term we apply (1). Writing z=u~(t+1)z=\tilde{u}(t+1) and cc for the axis of γ\gamma, we obtain

sinh⁡(d(z,γz)2)=cosh⁡(d(z,c))sinh⁡(ℓ(γ)2),\sinh\left(\frac{d(z,\gamma z)}{2}\right)=\cosh(d(z,c))\sinh\left(\frac{\ell(\gamma)}{2}\right),
sinh⁡(d(q,γq)2)=cosh⁡(d(q,c))sinh⁡(ℓ(γ)2).\sinh\left(\frac{d(q,\gamma q)}{2}\right)=\cosh(d(q,c))\sinh\left(\frac{\ell(\gamma)}{2}\right).

We notice that

d(z,c)≤d(z,q)+d(q,c)=t+1−t1+d(q,c)≤1+d(q,c).d(z,c)\le d(z,q)+d(q,c)=t+1-t_1+d(q,c)\le1+d(q,c).

We obtain

sinh⁡(d(z,γz)2)≤esinh⁡(d(q,γq)2).\sinh\left(\frac{d(z,\gamma z)}{2}\right)\le e\sinh\left(\frac{d(q,\gamma q)}{2}\right).

Since d(q,γq)≤1d(q,\gamma q)\le1, we have

d(z,γz)≤2sinh⁡(d(z,γz)2)≤2esinh⁡(d(q,γq)2)≤2ed(q,γq)≤6d(q,γq).d(z,\gamma z)\le2\sinh\left(\frac{d(z,\gamma z)}{2}\right)\le2e\sinh\left(\frac{d(q,\gamma q)}{2}\right)\le2ed(q,\gamma q)\le6d(q,\gamma q).

We get

∀ t1−1<t<t1,d1(gtv~,gtu~)≤15d(q,γq).\forall\ t_1-1<t<t_1,\quad d_1(g_t\tilde{v},g_t\tilde{u})\le15d(q,\gamma q).

Finally, we observe that

d1(gt+τv⃗,gtu~)≤d1(gt+τv⃗,gtv⃗)+d1(gtv⃗,gtu~)=2∣τ∣+d1(gtv⃗,gtu~)≤2d(q,γq)+d1(gtv⃗,gtu~),d_1(g_{t+\tau}\vec{v},g_t\tilde{u}) \le d_1(g_{t+\tau}\vec{v},g_t\vec{v}) + d_1(g_t\vec{v},g_t\tilde{u}) \\ = 2|\tau| + d_1(g_t\vec{v},g_t\tilde{u}) \\ \le2d(q,\gamma q) + d_1(g_t\vec{v},g_t\tilde{u}),

thanks to Proposition 4. Similarly we have

d1(gt+τv⃗,γgtu~)≤2d(q,γq)+d1(gtv⃗,γgtu~).d_1(g_{t+\tau}\vec{v},\gamma g_t\tilde{u}) \le2d(q,\gamma q) + d_1(g_t\vec{v},\gamma g_t\tilde{u}).

Together with the previous bounds, these imply the statement. □\square

Proof of Theorem 2

Rays intersecting arbitrarily short hypercycles

Let S=Γ\HS=\Gamma\backslash\mathbb{H} be a hyperbolic surface and let u∈T1Su\in T^1S be such that the geodesic ray u(R+)u(\mathbb{R}^+) meets closed hypercycles of arbitrarily small length. Notice that each closed hypercycle has a closed geodesic in its free homotopy class, which is shorter. The case where u(R+)u(\mathbb{R}^+) meets infinitely many of these closed geodesics falls in the first case of Proposition 3 and the conclusion of Theorem 2 follows from [BCDHV26]. Therefore, we may assume that u(R+)u(\mathbb{R}^+) does not meet the closed geodesics associated to these hypercycles.

Recall that H(γ,q)\mathcal{H}(\gamma,q) denotes the hypercycle of γ\gamma through qq (Section 2.4). If γ\gamma is in good position with respect to u~\tilde{u} and qq is the point of u~(R)\tilde{u}(\mathbb{R}) closest to the axis of γ\gamma, then u~(R)\tilde{u}(\mathbb{R}) is tangent to H(γ,q)\mathcal{H}(\gamma,q) at qq.

Proposition 6. Let S=Γ\HS=\Gamma\backslash\mathbb{H} be a hyperbolic surface and let u∈T1Su\in T^1S be such that u(R+)u(\mathbb{R}^+) meets closed hypercycles of arbitrarily small length without meeting the associated closed geodesics. Then, up to conjugating Γ\Gamma by an isometry of H\mathbb{H}, there

exist a lift u~\tilde{u} of uu with u~(0)=i\tilde{u}(0)=i and u~(+∞)=∞\tilde{u}(+\infty)=\infty, and a sequence (γn)n≥0(\gamma_n)_{n\geq0} of hyperbolic isometries in Γ\Gamma such that

  1. u~\tilde{u} is in good position with respect to each γn\gamma_n,

  2. the point qn=u~(tn)q_n=\tilde{u}(t_n) of u~(R)\tilde{u}(\mathbb{R}) closest to the axis of γn\gamma_n is such that (tn)n≥0(t_n)_{n\geq0} is increasing and tends to +∞+\infty,

  3. 0<γn−<γn+<γn+1−0<\gamma_n^-<\gamma_n^+<\gamma_{n+1}^- for all n≥0n\geq0,

  4. the hypercycles H(γn,qn)\mathcal{H}(\gamma_n,q_n) are pairwise disjoint,

  5. d(qn,γnqn)d(q_n,\gamma_nq_n) tends to 00.

Proof. By hypothesis, there are times sn≥0s_n\geq0 such that u(sn)u(s_n) belong to a hypercycle whose length tend to 00 when nn tends to +∞+\infty.

Up to conjugating the group Γ\Gamma by an isometry of H\mathbb{H}, we can assume that uu lifts to u~∈T1H\tilde{u}\in T^1\mathbb{H} such that u~(0)=i\tilde{u}(0)=i and u~(+∞)=∞\tilde{u}(+\infty)=\infty. For every nn, there exists a primitive hyperbolic isometry γn\gamma_n such that the hypercycle H(γn,u~(sn))\mathcal{H}(\gamma_n,\tilde{u}(s_n)) of γn\gamma_n through u~(sn)\tilde{u}(s_n) project to the closed hypercycles whose length tend to 00. In particular, εn:=d(u~(sn),γnu~(sn))\varepsilon_n:=d(\tilde{u}(s_n),\gamma_n\tilde{u}(s_n)) is smaller than the length of the hypercycle, so it also tends to 00.

Since ℓ(γn)≤εn→0\ell(\gamma_n)\leq\varepsilon_n\to0, up to passing to a subsequence, the isometries γn\gamma_n can be taken to be pairwise distinct. It can also be assumed that their fixed points are different from 00 and ∞\infty.

Let cnc_n be the axis of γn\gamma_n. By hypothesis u~(R+)\tilde{u}(\mathbb{R}^+) does not intersect cnc_n. Since the fixed points of γn\gamma_n are different from 00 and ∞\infty, the geodesics u~(R)\tilde{u}(\mathbb{R}) and cnc_n are not asymptotic. Let qn=u~(tn)q_n=\tilde{u}(t_n) be the point of u~(R)\tilde{u}(\mathbb{R}) closest to cnc_n. By (2.1), the displacement d(z,γnz)d(z,\gamma_nz) is smaller the closer zz lies to cnc_n. In particular,

d(qn,γnqn)≤d(u~(sn),γnu~(sn))=εn⟶0.d(q_n,\gamma_nq_n)\leq d(\tilde{u}(s_n),\gamma_n\tilde{u}(s_n))=\varepsilon_n\longrightarrow0.

We claim that tn>0t_n>0 for nn large enough. Otherwise, if tn≤0t_n\leq0, then tn≤0≤snt_n\leq0\leq s_n, so u~(0)=i\tilde{u}(0)=i is closer to cnc_n than u~(sn)\tilde{u}(s_n). By the property of the displacement function, we would get

d(i,γni)≤d(u~(sn),γnu~(sn))=εn,d(i,\gamma_n i)\leq d(\tilde{u}(s_n),\gamma_n\tilde{u}(s_n))=\varepsilon_n,

which if it happened for infinitely many nn would contradict the discreteness of Γ\Gamma. Up to taking a subsequence, we assume tn>0t_n>0. Moreover, qnq_n does not lie on cnc_n, thus u~(R)\tilde{u}(\mathbb{R}) stays at positive distance from cnc_n.

Observe that u~(R)\tilde{u}(\mathbb{R}) is tangent to the hypercycle of γn\gamma_n through qnq_n. Replacing γn\gamma_n by γn−1\gamma_n^{-1} if necessary, the orientation of u~(R)\tilde{u}(\mathbb{R}) agrees with the one induced by γn\gamma_n on the hypercycle through qnq_n. Together with tn>0t_n>0, this shows that u~\tilde{u} is in good position with respect to γn\gamma_n.

Moreover, tn→+∞t_n\to+\infty. Otherwise, up to passing to a subsequence, the points qnq_n stay in a compact subset of H\mathbb{H}, and d(qn,γnqn)→0d(q_n,\gamma_nq_n)\to0 with pairwise distinct γn\gamma_n would contradict the discreteness of Γ\Gamma. Up to passing to a subsequence, we may assume that (tn)n≥0(t_n)_{n\geq0} is increasing.

The axis cnc_n does not meet u~(R)\tilde{u}(\mathbb{R}), hence it is contained in one of the two half-planes {x>0}\{x>0\} or {x<0}\{x<0\}. Up to taking a subsequence and conjugating again the group Γ\Gamma by the reflection z↦−z‾z\mapsto-\overline{z}, which fixes u~(R)\tilde{u}(\mathbb{R}) pointwise and preserves good position, we may assume that they all lie in {x>0}\{x>0\}. Then γn−>0\gamma_n^->0 and γn+>0\gamma_n^+>0, and good position forces the attracting fixed point to be on the right of the repelling one, so that 0<γn−<γn+0<\gamma_n^-<\gamma_n^+.

The geodesic realizing the distance between u~(R)\tilde{u}(\mathbb{R}) and cnc_n is orthogonal to both, and it passes through the point qnq_n. Hence it is a Euclidean half-circle centered at 00. One of the end points xnx_n of this geodesic is between γn−\gamma_n^- and γn+\gamma_n^+, so γn−<xn<γn+\gamma_n^-<x_n<\gamma_n^+. By definition, qn=u~(tn)=ietnq_n=\tilde{u}(t_n)=ie^{t_n}, then xn=etnx_n=e^{t_n} tends to +∞+\infty, and so does γn+\gamma_n^+. We claim that γn−\gamma_n^- also tends to +∞+\infty. Otherwise, up to passing to a subsequence, we may assume that γn−\gamma_n^- converges to some ξ∈R\xi\in\mathbb{R}. Then the axes cnc_n converge to the geodesic with endpoints ξ\xi and +∞+\infty. Picking a point zz on this geodesic, it is accumulated by points znz_n on cnc_n. Since ℓ(γn)\ell(\gamma_n) tends to 00, we have d(z,γnz)→0d(z,\gamma_nz) \to0, which contradicts the fact that Γ\Gamma is discrete. Hence, we must have lim⁡nγn=+∞\lim_{n}\gamma_n=+\infty.

The hypercycle H(γn,qn)\mathcal{H}(\gamma_n,q_n) is an arc of the Euclidean circle through γn−\gamma_n^-, γn+\gamma_n^+ and qnq_n, which is tangent to the imaginary axis at qnq_n (see Figure 3). Then the Euclidean distance between 00 and H(γn,qn)\mathcal{H}(\gamma_n,q_n) tends to +∞+\infty. For this we can observe that this distance is greater than the distance from 00 to the Euclidean circle of radius γn−\gamma_n^- centered at γn−+γn−i\gamma_n^-+\gamma_n^-i. This allows us to extract a subsequence such that the hypercycles H(γn,qn)\mathcal{H}(\gamma_n,q_n) are pairwise disjoint and γn+<γn+1−\gamma_n^+<\gamma_{n+1}^- for all n≥0n\geq0. □

The geodesic $\tilde{u}(\mathbb{R})$, the axis $c_n$ of $\gamma_n$, and the hypercycle $\mathcal{H}(\gamma_n,q_n)$ tangent to $\tilde{u}(\mathbb{R})$ at $q_n$

Figure 3. The geodesic u~(R)\tilde{u}(\mathbb{R}), the axis cnc_n of γn\gamma_n, and the hypercycle H(γn,qn)\mathcal{H}(\gamma_n,q_n) tangent to u~(R)\tilde{u}(\mathbb{R}) at qnq_n.

From now on we fix the vector u∈T1Su\in T^1S, its lift u~\widetilde{u} with u~(0)=i\widetilde{u}(0)=i and u~(+∞)=∞\widetilde{u}(+\infty)=\infty, and the sequence (γn)n≥0(\gamma_n)_{n\geq0} of hyperbolic isometries given by Proposition 6.

Construction of the vector wαw_\alpha in Hssu‾\overline{H^{ss}u}

Let (αn)n≥0(\alpha_n)_{n\geq0} be a subsequence of (γn)n≥0(\gamma_n)_{n\geq0}. We write βn=α0⋯αn∈Γ\beta_n=\alpha_0\cdots\alpha_n\in\Gamma, and we denote by αn−\alpha_n^-, αn+\alpha_n^+ the repelling and attracting fixed points of αn\alpha_n, so that 0<αn−<αn+<αn+1−0<\alpha_n^-<\alpha_n^+<\alpha_{n+1}^-. For each n≥0n\geq0, let qnαq_n^\alpha be the point of u~(R)\widetilde{u}(\mathbb{R}) closest to the axis of αn\alpha_n, and set

δnα=d(qnα,αnqnα).\delta_n^\alpha=d(q_n^\alpha,\alpha_nq_n^\alpha).

Since (αn)(\alpha_n) is a subsequence of (γn)(\gamma_n), we have δnα→0\delta_n^\alpha\to0. We introduce the vectors v~n∈T1H\widetilde{v}_n\in T^1\mathbb{H} defined by

v~n(0)=i,v~n(+∞)=βn(∞).\widetilde{v}_n(0)=i,\qquad\widetilde{v}_n(+\infty)=\beta_n(\infty).

Lemma 2. The sequence (v~n)n≥0(\widetilde{v}_n)_{n\geq0} converges to a vector v~α∈T1H\widetilde{v}_\alpha\in T^1\mathbb{H} defined by v~α(0)=i\widetilde{v}_\alpha(0)=i and v~α(+∞)=lim⁡n→+∞βn(∞)\widetilde{v}_\alpha(+\infty)=\lim_{n\to+\infty}\beta_n(\infty).

Proof. Since every v~n\widetilde{v}_n is based at ii, it suffices to prove that the sequence (βn(∞))n≥0⊂∂H(\beta_n(\infty))_{n\geq0}\subset\partial\mathbb{H} converges. Each αi\alpha_i is a hyperbolic isometry with repelling and attracting fixed points $0<\alpha_i^-<\alpha_i^+$. Since ∞\infty lies outside [αi−,αi+][\alpha_i^-,\alpha_i^+], it is attracted by αi\alpha_i toward αi+\alpha_i^+, so that αi+<αi∞<+∞\alpha_i^+<\alpha_i\infty<+\infty. Moreover, αi\alpha_i preserves the order on the interval (αi+,+∞)(\alpha_i^+,+\infty), hence αi+<x<y\alpha_i^+<x<y implies αi+<αix<αiy\alpha_i^+<\alpha_ix<\alpha_iy. Using the nesting αi+<αi+1−\alpha_i^+<\alpha_{i+1}^-, we obtain for every $n\geq 1

0<αn−1αn∞<αn−1∞.0<\alpha_{n-1}\alpha_n\infty<\alpha_{n-1}\infty.

Applying α0,…,αn−2\alpha_0,\ldots,\alpha_{n-2} successively, which preserve the order on the relevant intervals, yields

0<βn(∞)<βn−1(∞).0<\beta_n(\infty)<\beta_{n-1}(\infty).

Hence (βn(∞))n≥0(\beta_n(\infty))_{n\geq0} is a decreasing sequence of positive numbers, and therefore converges. □

The ray βn−1v~n(R+)\beta_n^{-1}\widetilde{v}_n(\mathbb{R}^+) issues from βn−1i\beta_n^{-1}i and points to βn−1(βn∞)=∞\beta_n^{-1}(\beta_n\infty)=\infty; it is therefore the vertical geodesic ray starting at βn−1i\beta_n^{-1}i. The next lemma places this ray in good position with respect to αn+1\alpha_{n+1} and compares its displacement with δn+1α\delta_{n+1}^\alpha. This is the geometric point where the present case departs from [BCDHV26]: the ray does not cross the axis of αn+1\alpha_{n+1} but remains tangent to one of its hypercycles.

Lemma 3. For every n≥0n\geq0, the vector βn−1v~n\beta_n^{-1}\widetilde{v}_n is in good position with respect to αn+1\alpha_{n+1}. Moreover, if qn+1′q_{n+1}' denotes the point of βn−1v~n(R)\beta_n^{-1}\widetilde{v}_n(\mathbb{R}) closest to the axis of αn+1\alpha_{n+1}, then

d(qn+1′,αn+1qn+1′)≤δn+1α.d(q_{n+1}',\alpha_{n+1}q_{n+1}')\leq\delta_{n+1}^\alpha.

Proof. For n≥0n\geq0, let RnR_n be the closed region of the first quadrant bounded by the segment of the imaginary axis between 00 and qnαq_n^\alpha, the arc of the hypercycle H(αn,qnα)\mathcal{H}(\alpha_n,q_n^\alpha) between qnαq_n^\alpha and αn−\alpha_n^-, and the segment [0,αn−][0,\alpha_n^-] of ∂H\partial\mathbb{H}. Since the hypercycles H(αn,qnα)\mathcal{H}(\alpha_n,q_n^\alpha) are pairwise disjoint and qnα=u~(tnα)q_n^\alpha=\widetilde{u}(t_n^\alpha) with (tnα)(t_n^\alpha) increasing, we have Rn⊂Rn+1R_n\subset R_{n+1} for all n≥0n \ge0. Moreover, since H(αn,qnα)\mathcal{H}(\alpha_n,q_n^\alpha) is invariant under αn\alpha_n and αˉn−1\bar{\alpha}_n^{-1} moves the points of ∂H\partial\mathcal{H} toward its attracting fixed point αˉn\bar{\alpha}_n, one checks that αˉn−1(Rn)⊂Rn\bar{\alpha}_n^{-1}(R_n) \subset R_n.

We prove by induction that βn−1i∈Rn\beta_n^{-1}i \in R_n for all n≥0n \ge0. For n=0n=0 we have α0−1(i)∈α0−1(R0)⊂R0\alpha_0^{-1}(i) \in\alpha_0^{-1}(R_0) \subset R_0, because i∈R0i \in R_0. Assume βn−1i∈Rn\beta_n^{-1}i \in R_n. Then

βn+1−1i=αn+1−1βn−1i∈αn+1−1(Rn)⊂αn+1−1(Rn+1)⊂Rn+1.\beta_{n+1}^{-1}i=\alpha_{n+1}^{-1}\beta_n^{-1}i\in\alpha_{n+1}^{-1}(R_n)\subset\alpha_{n+1}^{-1}(R_{n+1})\subset R_{n+1}.

Since βn−1i∈Rn⊂Rn+1\beta_n^{-1}i \in R_n \subset R_{n+1}, the vertical ray βn−1v~n(R+)\beta_n^{-1}\widetilde v_n(\mathbb{R}^+) lies on the left of αˉn\bar\alpha_n, hence it stays at positive distance from the axis of αn+1\alpha_{n+1}, and is tangent to a hypercycle of αn+1\alpha_{n+1} enclosed by H(αn+1,qn+1α)\mathcal{H}(\alpha_{n+1},q_{n+1}^\alpha). The orientation of βn−1v~n(R+)\beta_n^{-1}\widetilde v_n(\mathbb{R}^+) agrees with the one induced by αn+1\alpha_{n+1}, so βn−1v~n\beta_n^{-1}\widetilde v_n is in good position with respect to αn+1\alpha_{n+1}. Finally, the point qn+1′q'_{n+1} realizing the distance to the axis lies on a hypercycle enclosed by H(αn+1,qn+1α)\mathcal{H}(\alpha_{n+1},q_{n+1}^\alpha), hence closer to the axis than qn+1αq_{n+1}^\alpha. As the displacement function decreases with the distance to the axis, we conclude

d(qn+1′,αn+1qn+1′)≤d(qn+1α,αn+1qn+1α)=δn+1α.d(q'_{n+1},\alpha_{n+1}q'_{n+1})\le d(q_{n+1}^\alpha,\alpha_{n+1}q_{n+1}^\alpha)=\delta_{n+1}^\alpha.

By Lemma 3, the winding Wind⁡αn+1(βn−1v~n)\operatorname{Wind}_{\alpha_{n+1}}(\beta_n^{-1}\widetilde v_n) is defined, and it is the vector based at βn−1i\beta_n^{-1}i pointing at αn+1∞\alpha_{n+1}\infty. On the other hand, βn−1v~n+1\beta_n^{-1}\widetilde v_{n+1} is based at βn−1i\beta_n^{-1}i and points at βn−1(βn+1∞)=αn+1∞\beta_n^{-1}(\beta_{n+1}\infty)=\alpha_{n+1}\infty. Hence, for all n≥0n\ge0,

βn−1v~n+1=Wind⁡αn+1(βn−1v~n).\beta_n^{-1}\widetilde v_{n+1}=\operatorname{Wind}_{\alpha_{n+1}}(\beta_n^{-1}\widetilde v_n).

By the definition of the winding time (Definition 7) applied to αn+1\alpha_{n+1} and βn−1v~n\beta_n^{-1}\widetilde v_n, together with Proposition 4 and Lemma 3,

∣B∞(αn+1−1βn−1i,βn−1i)∣=∣ταn+1,βn−1v~n∣≤d(qn+1′,αn+1qn+1′)≤δn+1α.(4)\left|B_\infty(\alpha_{n+1}^{-1}\beta_n^{-1}i,\beta_n^{-1}i)\right| =\left|\tau_{\alpha_{n+1},\beta_n^{-1}\widetilde v_n}\right| \le d(q'_{n+1},\alpha_{n+1}q'_{n+1})\le\delta_{n+1}^\alpha. \tag*{(4)}

We define

rn=B∞(βn−1i,i).r_n=B_\infty(\beta_n^{-1}i,i).

By construction grnβn−1v~n∈Hssu ~,g_{r_n}\beta_n^{-1}\widetilde v_n\in H^{ssu}\widetilde{\ }, so its projection satisfies grnvn∈Hssug_{r_n}v_n\in H^{ssu}.

Proposition 7. Assume that ∑n≥0δnα<+∞\sum_{n\ge0}\delta_n^\alpha<+\infty. Then the sequence (rn)n≥0(r_n)_{n\ge0} converges to a real number rαr_\alpha, and ∣rn∣≤∑i≥0δiα|r_n|\le\sum_{i\ge0}\delta_i^\alpha for all n≥0n\ge0; hence ∣rα∣≤∑i≥0δiα|r_\alpha|\le\sum_{i\ge0}\delta_i^\alpha.

Proof. By definition of rnr_n and (5.1),

∣rn+1−rn∣=∣B∞(αn+1−1βn−1i,βn−1i)∣=∣ταn+1,βn−1v~n∣≤δn+1α.|r_{n+1}-r_n|=\left|B_\infty(\alpha_{n+1}^{-1}\beta_n^{-1}i,\beta_n^{-1}i)\right|=\left|\tau_{\alpha_{n+1},\beta_n^{-1}\widetilde v_n}\right|\le\delta_{n+1}^\alpha.

Hence ∣rm−rn∣≤∑i=n+1mδiα|r_m-r_n|\le\sum_{i=n+1}^{m}\delta_i^\alpha, so (rn)(r_n) is a Cauchy sequence and converges. Moreover

∣rn∣≤∣r0∣+∑i=0n−1∣ri+1−ri∣≤∑i=0n−1δiα≤∑i≥0δiα,|r_n|\le|r_0|+\sum_{i=0}^{n-1}|r_{i+1}-r_i|\le\sum_{i=0}^{n-1}\delta_i^\alpha\le\sum_{i\ge0}\delta_i^\alpha,

using ∣r0∣=∣τα0,u~∣≤δ0α|r_0|=|\tau_{\alpha_0,\widetilde u}|\le\delta_0^\alpha (Proposition 4). Letting n→+∞n\to+\infty gives the last inequality.

Construction of the vector wαw_\alpha in WssuW^{ssu}

Throughout this subsection we fix a subsequence (αn)n≥0(\alpha_n)_{n\ge0} of (γn)n≥0(\gamma_n)_{n\ge0} such that δnα≤1\delta_n^\alpha\le1 for all nn and ∑n≥0δnα<+∞\sum_{n\ge0}\delta_n^\alpha<+\infty, and we write Δ=∑n≥0δnα\Delta=\sum_{n\ge0}\delta_n^\alpha.

By Proposition 7 the winding times rnr_n converge to rαr_\alpha, and by Lemma 2 the vectors v~n\widetilde v_n converge to v~α\widetilde v_\alpha. We set, for n≥0n\ge0,

w~n=grnv~n∈T1H,w~α=grαv~α=lim⁡n→+∞w~n∈T1H.\widetilde w_n=g_{r_n}\widetilde v_n\in T^1\mathbb{H},\qquad \widetilde w_\alpha=g_{r_\alpha}\widetilde v_\alpha=\lim_{n\to+\infty}\widetilde w_n\in T^1\mathbb{H}.

We denote by vnv_n, vαv_\alpha, wnw_n, wαw_\alpha the projections to T1ST^1S of v~n\widetilde v_n, v~α\widetilde v_\alpha, w~n\widetilde w_n, w~α\widetilde w_\alpha, respectively. We remark that, since βn−1w~n=grnβn−1v~n∈Hssu ~,\beta_n^{-1}\widetilde w_n=g_{r_n}\beta_n^{-1}\widetilde v_n\in H^{ssu}\widetilde{\ }, the vectors wnw_n lie in HssuH^{ssu} and wnw_n converges to wαw_\alpha.

Our goal is to show that in fact wα∈Wssuw_\alpha\in W^{ssu}. For this we show that the distance between gtwng_t w_n and gtug_t u tends to zero uniformly in nn and then we take the limit in nn.

Proposition 8. For every ε>0\varepsilon> 0 there exists a time T>0T > 0 such that

d1(gtu,gtwn)<εfor all t>T and all n≥0.d_1(gtu,gtw_n) < \varepsilon\qquad\text{for all } t > T \text{ and all } n \ge0.

Proof. Fix ε>0\varepsilon> 0. Since ∑kδkα<+∞\sum_k \delta_k^\alpha< +\infty, we may fix NN with 17∑k>Nδkα<ε/217\sum_{k>N}\delta_k^\alpha< \varepsilon/2. We will use that, if n>Nn > N, then

d1(gtu,gtwn)≤d1(gtu,gtwN)+∑k=Nn−1d1(gtwk+1,gtwk).(5)d_1(gtu,gtw_n) \le d_1(gtu,gtw_N) + \sum_{k=N}^{n-1}d_1(gtw_{k+1},gtw_k). \tag*{(5)}

Recall that βk−1v~k+1=Wind⁡αk+1(βk−1v~k)\beta_k^{-1}\widetilde{v}_{k+1}=\operatorname{Wind}_{\alpha_{k+1}}(\beta_k^{-1}\widetilde{v}_k) and ταk+1,βk−1=rk+1−rk\tau_{\alpha_{k+1},\beta_k^{-1}}=r_{k+1}-r_k, and that, by Lemma 3, the point of βk−1v~k(R)\beta_k^{-1}\widetilde{v}_k(\mathbb{R}) closest to the axis of αk+1\alpha_{k+1} has displacement at most δk+1α≤1\delta_{k+1}^\alpha\le1. Applying the Key Proposition 5 to βk−1v~k\beta_k^{-1}\widetilde{v}_k and αk+1\alpha_{k+1} and then projecting to T1ST^1S, we obtain

d1(gs+rk+1−rkvk+1,gsvk)≤17δk+1α,∀s≥0.d_1(g_{s+r_{k+1}-r_k}v_{k+1},g_sv_k)\le17\delta_{k+1}^\alpha,\qquad\forall s\ge0.

Substituting s=t+rks=t+r_k gives

d1(gtwk+1,gtwk)=d1(gt+rk+1vk+1,gt+rkvk)≤17δk+1αd_1(gtw_{k+1},gtw_k)=d_1(g_{t+r_{k+1}}v_{k+1},g_{t+r_k}v_k)\le17\delta_{k+1}^\alpha

for all t≥−rkt\ge-r_k. Since ∣rk∣≤Δ|r_k|\le\Delta, this holds for all t≥Δt\ge\Delta in particular. Therefore the second term in (5.2) is bounded by 17∑k=Nn−1δk+1α<ε/217\sum_{k=N}^{n-1}\delta_{k+1}^\alpha<\varepsilon/2 for all t≥Δt\ge\Delta.

To bound the first term in (5.2) or to treat the case n≤Nn\le N, we may observe again that wi∈Hssuw_i\in H^{ss}u, so d1(gtwi,gtu)→0d_1(gtw_i,gtu)\to0 as t→+∞t\to+\infty. Hence there exists T≥ΔT\ge\Delta such that d1(gtwi,gtu)<ε/2d_1(gtw_i,gtu)<\varepsilon/2 for all t>Tt>T. This concludes the proof. □

Since wα=lim⁡n→+∞wnw_\alpha=\lim_{n\to+\infty}w_n, letting n→+∞n\to+\infty in Proposition 8 gives, for every ε>0\varepsilon>0 and every t>Tt>T,

d1(gtwα,gtu)=lim⁡n→+∞d1(gtwn,gtu)≤ε.d_1(gtw_\alpha,gtu)=\lim_{n\to+\infty}d_1(gtw_n,gtu)\le\varepsilon.

Hence lim⁡t→+∞d1(gtwα,gtu)=0\lim_{t\to+\infty}d_1(gtw_\alpha,gtu)=0, that is, wα∈Wssuw_\alpha\in W^{ss}u.

Construction of the vector wαw_\alpha in Wssu∖HssuW^{ss}u\setminus H^{ss}u

We denote by Σ\Sigma the set of subsequences α=(αn)n≥0\alpha=(\alpha_n)_{n\ge0} of (γn)n≥0(\gamma_n)_{n\ge0} such that

δnα≤1for all n≥0and∑n≥0δnα<+∞.\delta_n^\alpha\le1\quad\text{for all }n\ge0 \qquad\text{and}\qquad \sum_{n\ge0}\delta_n^\alpha<+\infty.

By Proposition 8, every α∈Σ\alpha\in\Sigma satisfies wα∈Wssuw_\alpha\in W^{ss}u. Since δn=d(qn,γnqn)→0\delta_n=d(q_n,\gamma_nq_n)\to0, the set Σ\Sigma is nonempty. For α∈Σ\alpha\in\Sigma we set

ξ(α)=w~α(+∞)=lim⁡n→+∞βn(∞)∈R,\xi(\alpha)=\widetilde{w}_\alpha(+\infty)=\lim_{n\to+\infty}\beta_n(\infty)\in\mathbb{R},

which is a positive number by Lemma 2.

Proposition 9. The image ξ(Σ)\xi(\Sigma) is uncountable.

Proof. We use the following property, which follows from the dynamics of the αn\alpha_n as in the proof of Lemma 2: for every α∈Σ\alpha\in\Sigma and every n≥0n\ge0,

αnαn+1⋯αm(∞)>αn+for all m≥n,\alpha_n\alpha_{n+1}\cdots\alpha_m(\infty)>\alpha_n^+ \qquad\text{for all }m\ge n,

hence

αn−1−1⋯α0−1(ξ(α))≥αn+.(6)\alpha_{n-1}^{-1}\cdots\alpha_0^{-1}(\xi(\alpha))\ge\alpha_n^+. \tag*{(6)}

Assume, for contradiction, that ξ(Σ)\xi(\Sigma) is countable and enumerate it as {x0,x1,…}\{x_0,x_1,\ldots\}. We construct α′=(αn′)∈Σ\alpha'=(\alpha'_n)\in\Sigma with ξ(α′)≠xi\xi(\alpha')\ne x_i for all ii.

Choose α0′=γk0\alpha'_0=\gamma_{k_0} with δ0′<1\delta'_0<1 and α0′+>x0\alpha'_0{}^+>x_0. Inductively, having chosen α0′,…,αn′\alpha'_0,\ldots,\alpha'_n, pick αn+1′=γkn+1\alpha'_{n+1}=\gamma_{k_{n+1}} with kn+1>knk_{n+1}>k_n such that

δn+1′<12n+1andαn+1′+>αn′−1⋯α0′−1(xn+1).\delta'_{n+1}<\frac{1}{2^{n+1}} \qquad\text{and}\qquad \alpha'_{n+1}{}^+>\alpha_n'{}^{-1}\cdots\alpha_0'{}^{-1}(x_{n+1}).

This is possible since γk+→+∞\gamma_k^+ \to+\infty and δk→0\delta_k \to0. Since ∑n≥0δnα′<∑n≥02−n<+∞\sum_{n\ge0}\delta_n^{\alpha'} < \sum_{n\ge0}2^{-n} < +\infty and δnα′≤1\delta_n^{\alpha'} \le1, the resulting sequence α′\alpha' belongs to Σ\Sigma.

By (5.3) and the choice of αn′\alpha'_n, we have for n≥1n \ge1

αn−1′−1⋯α0′−1(ξ(α′))≥αn′+>αn−1′+⋯α0′+(xn),\alpha_{n-1}^{\prime-1}\cdots\alpha_0^{\prime-1}\bigl(\xi(\alpha')\bigr) \ge\alpha_n^{\prime+} > \alpha_{n-1}^{\prime+}\cdots\alpha_0^{\prime+}(x_n),

whence ξ(α′)≠xn\xi(\alpha') \ne x_n. Finally ξ(α′)>α0′+>x0\xi(\alpha') > \alpha_0^{\prime+} > x_0, so ξ(α′)≠x0\xi(\alpha') \ne x_0. This contradicts ξ(α′)∈ξ(Σ)={xi}\xi(\alpha') \in\xi(\Sigma)=\{x_i\}. □

Corollary 1. There exists a subsequence α∈Σ\alpha\in\Sigma such that wα∉Huwsw_\alpha\notin H^{ws}_u. Moreover, WussW^{ss}_u contains an uncountable union of pairwise distinct stable horocycles.

Proof. If w∈T1Sw \in T^1S belongs to HuwsH^{ws}_u, then w~(+∞)∈Γu~(+∞)=Γ∞\widetilde w(+\infty) \in\Gamma\widetilde u(+\infty)=\Gamma_\infty for any lifts. Since Γ\Gamma is countable and ξ(Σ)\xi(\Sigma) is uncountable by Proposition 9, the set ξ(Σ)\xi(\Sigma) meets uncountably many Γ\Gamma-orbits in ∂H\partial\mathbb{H}. Hence there exists α∈Σ\alpha\in\Sigma with ξ(α)=w~α(+∞)∉Γ∞\xi(\alpha)=\widetilde w_\alpha(+\infty) \notin\Gamma_\infty, that is, wα∉Huwsw_\alpha\notin H^{ws}_u. By Proposition 8, all these wαw_\alpha lie in WussW^{ss}_u, and their forward endpoints lie in distinct Γ\Gamma-orbits, so the stable horocycles Hwαss⊂WussH^{ss}_{w_\alpha} \subset W^{ss}_u are pairwise distinct and uncountable in number. □

Declaration on the use of AI

During the preparation of this work the authors used Claude (Anthropic) to assist with the writing and the proofreading of the manuscript. The mathematical content and the arguments are entirely the authors’ own. The authors have reviewed all the text and take full responsibility for the content of the article.

References

IMERL, CMAT, Universidad de la República, IFUMI, IRL CNRS 2030, Uruguay

Email address: [email protected]

IRMAR, Université de Rennes, France and IFUMI, IRL CNRS 2030

Email address: [email protected]

References

  1. [1]Werner Ballmann. Lectures on spaces of nonpositive curvature, volume 25 of DMV Seminar. Birkhäuser Verlag, Basel, 1995. With an appendix by Misha Brin.DOI
  2. [2]Sergi Burniol Clotet and Françoise Dal’Bo. On the non-expansiveness of the geodesic flow on surfaces with cusps, 2026. Preprint, arXiv:2603.24310.arxiv.org/abs/2603.24310
  3. [4]Alan F. Beardon. The Geometry of Discrete Groups, volume 91 of Graduate Texts in Mathematics. Springer, New York, 1983.DOI
  4. [5]Alexandre Bellis. Étude topologique du flot horocyclique : le cas des surfaces géométriquement infinies. PhD thesis, Université de Rennes, May 2018.DOI
  5. [6]Alexandre Bellis. On the links between horocyclic and geodesic orbits on geometrically infinite surfaces. Journal de l’École polytechnique. Mathématiques, 5:443–454, 2018.DOI

Paper details

vlnc:W_rc3m2eg9
v1 · latestHistory (1)

published · Fresh — added to the Hub within the last 30 days

added ·

Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

CC BY 4.0

Original source

  • Mathematics

Claude · View statement →

Looking for related papers…
Contents