General properties of groups of isometries in metric spaces

Definitions

Let GG be a group of isometries in a metric space XX, we say that the action of GG is proper if the action map

ϕ:G×XX×X,(g,x)(x,g.x),\phi: G\times X \to X \times X, \qquad (g, x) \mapsto (x, g.x),

is a proper map (i.e. the preimage ϕ1(C)\phi^{-1}(C) of any compact set CX×XC \subset X \times X is compact). Equivalently, ϕ\phi is proper if and only if, for any compact set KXK \subset X, the set

{gGg(K)K}\left\{ g \in G\mid g(K) \cap K \neq \emptyset \right\}

is compact. For discrete subgroups, this definition is equivalent to requiring that the set @eq-proper-action-equiv (Private) is finite, as in the definition of Scott @scott1983 (Private) pg. 86.

Discrete subgroups of isometries

The following proposition resumes the basic properties of groups of isometries acting on a metric space XX, the tool for the proof will be Arzela-Ascoli theorem. Recall a family of continuous functions F\mathcal{F} is equicontinuous if for all ϵ>0\epsilon > 0 exists δ>0\delta > 0 such that

d(x,y)<δd(f(x),f(y))<ϵ,d(x,y) < \delta \Rightarrow d(f(x), f(y)) < \epsilon,

independently of the value of xx and yy.

::: {#thm-arzela-ascoli} Let FC(X,Y)\mathcal{F} \subset \mathcal{C}(X, Y) be a family of maps between metric spaces, then F\mathcal{F} is relatively compact in the topology of compact convergence (i.e. the topology of uniform convergence on compact sets) if and only if F\mathcal{F} is equicontinuous and F(x)\mathcal{F}(x) is relatively compact in YY for any xXx\in X. :::

Note that families of isometries of the metric space XX are trivially equicontinuous,

::: {#prp-1 .prp} If ΓX\Gamma \subset X is a discrete subgroup of isometries of the complete metric space (X,d)(X, d) with the Heine-Borel property (each closed bounded set in XX is compact), then:

  1. The action of Γ\Gamma is proper.
  2. For any xXx \in X, the orbit Γx\Gamma x has no accumulation point in XX. :::

The source of the proof for the first part can be found here.

::: proof Suppose the existence of a compact set KXK\subset X and an infinite sequence of distinct elements γnΓ\gamma_n \in \Gamma such that γn(K)K\gamma_n (K) \cap K \neq \emptyset. Let pKp \in K and R=2diam(K)R = 2 \operatorname{diam}(K), then γn(p)\gamma_n(p) is contained in the closed ball B(p,R)\overline{B}(p, R), hence for any xXx \in X,

d(γn(x),p)d(γn(x),γn(p))+d(γn(p),p)=d(x,p)+d(γn(p),p)d(x,p)+R.\begin{align*} d(\gamma_n(x), p) &\leq d(\gamma_n(x), \gamma_n(p)) + d(\gamma_n(p), p)\\ &= d(x, p) + d(\gamma_n(p), p) \\ &\leq d(x, p) + R. \end{align*}

Since XX has the Heine-Borel property, the inequality above implies Γx\Gamma x is relatively compact for all xx and since Γ\Gamma is a group of isometries, it is also equicontinuous, therefore by the Arzela-Azcoli theorem, Γ\Gamma is relatively compact in the topology of uniform convergence on compact sets, hence we can suppose there is a continuous function, in fact an isometry, γ:XX\gamma: X \to X, such that γnγ\gamma_n \to \gamma uniformly on compact sets, hence γn1γn+1e\gamma_n^{-1}\gamma_{n+1} \to e, uniformly on compact sets, but since the topology of Γ\Gamma is discrete, γn=γn+1\gamma_n = \gamma_{n+1} for nn sufficiently large, a contradiction.

For the second part assume towards a contradiction the existence of a limit point yy for the orbit Γx\Gamma x, hence there is an infinite sequence {γn}Γ\{\gamma_n\} \subset \Gamma such that γnxy\gamma_n x \to y. If KK is the compact set {x}{γnx}{y}\{x\} \cup \{\gamma_n x\} \cup \{y\}, then γn(K)K0\gamma_n(K)\cap K \neq 0 for all γn\gamma_n in the sequence, a contradiction. :::

In particular, complete Riemannian manifolds have the Heine-Borel property and therefore discrete subgroups of isometries act properly.

Critical exponent of groups of isometries

Let Γ\Gamma be a discrete subgroup of isometries of the complete metric space XX with the Heine-Borel property and let x,yXx, y \in X be any pair of points, we can form the Poincare series of the group, defined as,

P(s;x,y)=γΓesd(x,γy),sR.P(s; x, y) = \sum_{\gamma \in \Gamma} e^{-s d(x, \gamma y)}, \qquad s \in \mathbb{R}.

The critical exponent of the group is the number δΓ\delta_\Gamma such that the series is convergent for s>δΓs > \delta_\Gamma. It can be shown that δΓ\delta_\Gamma is independent of the pair of points chosen in XX, in fact, it is well known that

δΓ=lim supRlog(#{γΓγxB(x,R)})R.\delta_\Gamma = \limsup_{R \to \infty} \frac{\log (\#\{\gamma \in \Gamma \mid \gamma x \in B(x, R)\})}{R}.

If XX is also endowed with a measure mm, we can define the volume entropy of XX as,

δX=lim supRlogm(B(x,R))R.\delta_X = \limsup_{R \to \infty} \frac{\log m(B(x,R))}{R}.

The following facts are well known for manifolds of negative curvature, however they are valid for more general metric spaces.

::: prp-deltag-deltax Let XX be a complete metric space with the Heine-Borel property equipped with a measure mm compatible with the metric, in the sense that for any measurable set AXA \subset X and any isometry ff, we have m(f(A))=m(A)m(f(A)) = m(A). If Γ\Gamma is a discrete group of isometries, then

δΓδX,\delta_\Gamma \leq \delta_X,

moreover, if Γ\Gamma is co-compact (the quotient space ΓX\Gamma \setminus X is compact), then δΓ=δX\delta_\Gamma = \delta_X. :::

The following proof can be found in @quint. (Private)

::: proof Since Γ\Gamma is discrete and XX has the Heine-Borel property, the action of Γ\Gamma is proper, hence no orbit Γx\Gamma x has accumulation points, therefore there exists a real number s>0s > 0 such that for every\ γΓ\gamma \in \Gamma, if B(γx,s)B(x,s)B(\gamma x, s) \cap B(x, s) \neq \emptyset then γx=x\gamma x = x. Let nn be the number of elements of Γ\Gamma that fix xx, by the triangle inequality,

#{γΓγxB(x,r)}m(B(x,r))nm(B(x,r+s)),\#\{\gamma \in \Gamma \mid \gamma x \in B(x, r)\} m(B(x,r)) \leq n m(B(x, r+s)),

taking logarithms, Quint deduces δΓδX\delta_\Gamma \leq \delta_X. If Γ\Gamma is co-compact, there is a compact fundamental region FXF \subset X such that Γ(F)=X\Gamma(F) = X. If s>0s > 0 is sufficiently large for FB(x,s)F \subset B(x, s), then X=γΓB(γx,s)X = \cup_{\gamma \in \Gamma}B(\gamma x, s), since each γ\gamma is an isometry and γ(B(x,s))=B(γx,s)\gamma (B(x, s)) = B(\gamma x, s). Thus for any r>0r > 0, the triangle inequality implies

B(x,r){γΓγxB(x,r+s)}B(γx,s),B(x,r) \subset \cup_{\{\gamma \in \Gamma \mid\gamma x\in B(x, r+s)\}} B(\gamma x, s),

hence,

m(B(x,r))#{γΓγxB(x,r+s)}m(B(x,s)),m(B(x, r)) \leq \#\{\gamma \in \Gamma \mid \gamma x \in B(x, r+s)\} m(B(x, s)),

where we have used that the measure is compatible with the metric. Taking logs, the last inequality implies δXδΓ\delta_X \leq \delta_\Gamma. :::

The visual boundary in a metric space

Let (X,d)(X, d) be a complete metric space, a geodesic in XX is a continuous function γ:RX\gamma: \mathbb{R} \to X such that

d(γ(s),γ(t))=st.d(\gamma(s), \gamma(t)) = |s - t|.

Given any two geodesics γ1,γ2\gamma_1, \gamma_2, we define the equivalence relation γ1γ2\gamma_1 \sim \gamma_2 if and only if there is a constant c>0c > 0, such that d(γ1(t),γ2(t))cd(\gamma_1(t), \gamma_2(t)) \leq c for all t0t \geq 0. We denote the set of equivalence classes as X()X(\infty) and call it the visual boundary of XX. Let X=XX()\overline{X} = X \cup X(\infty), we call X\overline{X} the completion of XX, if ξX()\xi \in X(\infty) and γ\gamma is a geodesic in the equivalence class, we also denote ξ\xi as γ()\gamma(\infty) and call ξ\xi the end of γ\gamma. Let us introduce a topology on X\overline{X} in the following manner: Let pXp \in X, ξX()\xi \in X(\infty), ΘR\Theta \in \mathbb{R}, and assume there is a geodesic γ\gamma such that γ(0)=p\gamma(0) = p and γ()=ξ\gamma(\infty) = \xi, we define the cone set

C(p,ξ,Θ)={xXp(x,ξ)<Θ}.C(p, \xi, \Theta) = \{x \in \overline{X} \mid \angle_p(x, \xi) < \Theta\}.

References

  1. Quint, J.-F. n.d. “An Overview of Patterson-Sullivan Theory.”
  2. Scott, Peter. 1983. “The Geometries of 3-Manifolds.” Bulletin of the London Mathematical Society 15 (5): 401–87. https://doi.org/10.1112/blms/15.5.401.