Hyperbolic Groups Presentation

We start with hyperbolic 22-space, a real Riemannian 22-mainfold denoted by H2\mathbb{H}^2 which has different isometrically diffeomorphic models.

Hyperbolic Models Revisited

Hyperboloid Model I2\mathbb{I}^2

Consider following bi-linear form in R3\mathbb{R}^{3} :

xy=x1y1+x2y2x3y3\langle x | y \rangle = x_ 1 y_ 1 + x_ 2 y_ 2 - x_ 3 y_ 3

and the upper fold of the hyperboloid associated to \langle \cdot | \cdot \rangle:

I2={xR3:xx=1,x3>0}\mathrm{I} ^ { 2 } = \{ x \in \mathbb{R}^3 : \langle x | x \rangle = - 1 , x _ 3 > 0 \}

I2\mathrm{I}^ 2 is a smooth 22-manifold with tangent space:

TxI2={yR3:xy=0}={x}T _ { x } \mathrm{I} ^ { 2 } = \{ y \in \mathbb{R} ^ 3 : \langle x | y \rangle = 0 \} = \{ x \} ^ { \perp }

Since xx=1\langle x | x \rangle = - 1, the restriction of \langle \cdot | \cdot \rangle to {x}\{ x \} ^ { \perp } is positive-definite.
Then it is a scalar product on TxI2T _ { x } \mathrm{I} ^ { 2 } and hence gives a metric struture.

We shall denote by I2\mathbb{I} _ 2 the manifold I2\mathrm{I} _ 2 endowed with this metric structure.

Disk Model D2\mathbb{D}^2

Disk model D2\mathbb{D}^2 comes from steographic projection of I2\mathbb{I}^2.

Dash lines indicate boundary of this projection.
Projection from Hyperboloid

Half-Plane Model Π2\mathbb{\Pi}^2

Half-space model comes from a Möbius transformation of D2\mathbb{D}^2.

It is a geometric inversion explained in the left picture: OPOP=r2OP\cdot OP^{\prime }=r^{2}.

Inversion from Disk 2dimensionalExm

Isometry of H2\mathbb{H}^2

We admit following non-trival proposition

A sufficient condition for a diffeomorphism of H2\mathbb{H}^{2} to preserve lengths is that it preserves angles.

We remark that

  • it is trivial that length preserving implies angle preseving
  • Möbius transformation is conformal, hence maps geodesics to geodesics in hyperbolic space
  • this proposition is powerful and streamlines proofs in elementary hyperbolic geometry.

Now we land back to H2\mathbb{H}^{2} and claim:

In disk model D2\mathbb{D}^2 and half-plane model Π2\mathbb{\Pi}^2, geodesics are orthocircles, which are lines/circles perpendicular to the boundary.

Because half-plane Π2\mathbb{\Pi}^2 is conformal to disk model D2\mathbb{D}^2, we only need to prove this for disk model D2\mathbb{D}^2.

  • Diameters are fixed set of reflections hence are geodesics, they are perpendicular to boundary.
  • Diameters are all geodesics passing the origin by uniquess of geodesics.
  • After Möbius transformations that fix D2\mathbb{D}^2 we get all orthocircles from diameters.

Here we play a game to illustrate geodesics in disk model D2\mathbb{D}^2:
Spaceship in disk model D2\mathbb{D}^2.

Use Mathematica to interact with triangles in disk model D2\mathbb{D}^2.

δ\delta-Hyperbolicity

δ\delta-hyperbolicity is a property shared by geometrical trees and hyperbolic plane:

Triangles are thin.

A triangle is δ\delta-thin if each side is in the δ\delta-neighborhood of other two sides.

A thin triangle

In the definition of δ\delta-hyperbolicity, the actual value δ\delta is not important.

But we remark:

  • In geometric trees, δ=0\delta=0, since ACABBC\overline{AC} \subset \overline{AB} \cup \overline{BC}.

Triangle in a tree

  • What should the actual value of δ\delta of hyperbolic plane?
    We can calculate this if we have time at the end.

Hyperbolic group

We claim without detailed verfication that:

δ\delta-hyperbolicity is invariant under quasi-isometry.

We apply this to get a definition:

  • A finitely generated group is called hyperbolic if any of its Cayley graphs (for a finite generating set) is δ\delta-hyperbolicity.
  • Well-defined since different finite generators give quasi-isometric Cayley graphs

Surface groups are hyperbolic

Surface groups are defined to be fundamental groups of hyperbolic surfaces.

  • To define a hyperbolic surfaces / mainfold, we need to work with charts.
  • This claim says that

    Hyperbolic surface has a hyperbolic fundamental group.

Add extra structure to manifold

A mainfold has chart in Rn\mathbb{R}^n with smooth coordinate transformation.
To generalize, we

  1. allow charts to locate at a space form XX
  2. restrict coordinate transformation to be restriction of orientation-preserving isometry group I+(X)\mathcal{I}^+(X).
  3. This defines a (X,I+(X))(X, \mathcal{I}^+(X))-manifold.

Example: torus is flat

We call an (X,I+(X))(X, \mathcal{I}^+(X))-manifold is

  • hyperbolic, if X=HnX=\mathbb{H}^n
  • elliptic, if X=SnX=S^n
  • flat, if X=RnX=\mathbb{R}^n

We shall classify all surfaces soon.

Quotient manifold and Švarc–Milnor lemma

  1. A (X,I+(X))(X, \mathcal{I}^+(X))-manifold MX/π1(M)M\cong X/\pi_ 1 (M) as Riemmanian manifold
  2. π1(M)\pi_ 1(M) is quasi-isometric to XX
  1. It follows

    Hyperbolic surface has a hyperbolic fundamental group.

  2. We then want to find out hyperbolic surfaces, that is to say,
    a geometric classification of surfaces.

Geometric classification of surfaces

We always assume closed, oriented and connected surface / 22-manifold.

Elliptic surface

22-Sphere is the only elliptic surface since every element of I+(S2)=SO(3)\mathcal{I}^{+}(S^2) = SO(3) has fixed points, which forces elliptic surfaces as quotient manifold of S2S^2 to be trivial.

Flat surface

We have shown torus in Mathematica

Hyperbolic surface

Surface with genus bigger than 2, i.e., connected sum of nn-torus (with n2n \ge 2) are hyperbolic.
Prove by JavaScipt Code.

Bonus (if we have time)

  • Actual value of δ\delta of hyperbolic plane is tanh1(1/2)=log(1+2)\tanh^{-1}(1/\sqrt2)=\log(1+\sqrt2).

    • In disk model, move triangle to contain the origin then vertices are on radius, endpoints of three radius form an ideal triangle containing the original triangle.
    • Any ideal triangle is enough to conclude since Möbius transformation is 3-transitive.
    • We only need to cover one side of ideal triangle by the same reason above, to simplify put it in half plane model.
    • Finally our problem reduces to finding distance of a midle-point to another side. Use reflection.
  • A snake game on 22-torus T2\mathrm{T}^2

Appendix and notes (out of presentation)

Formal definition of (X,G)(X,G)-structure

Let AA be a connected, simply connected, oriented nn-dimensional manifold(n2n\ge 2), and let GG be a group of diffeomorphisms of AA onto itself; we shall say a differentiable nn-manifold MΜ is endowed with an (X,G)(X,G)-structure if we are given an open covering {Ui}\{U_ i\} of MM and a set of differentiable open mappings {φi}\{\varphi_ i\} (with φi:UiX\varphi_ i : U_ i \mapsto X) such that

  1. φi:Uiφi(Ui)\varphi_ i : U_ i \mapsto \varphi_ i(U_ i) is a diffeomorphism;
  2. if UiUjU_ i \cap U_ j \ne \emptyset then the restriction of φiφj1\varphi_ i \circ \varphi_ {j}^{-1} to each connected component of φj(UiUj)\varphi_ j(U_ i \cap U_ j) is the restriction of an element of GG.

{(Ui,φi)}\{(U_ i , \varphi_ i )\} will be called an atlas defining the (X,G)(X,G)-structure.

Classfication using complex analysis

This classfication corresponds to uniformization theorem about simply connected Riemannian surfaces.

  • Any surface has isothermal coordinate.
  • Isothermal coordinate gives a complex structure on this surface.
  • Universal covering of a Riemannian surface is a simply connected Riemannian surface.
  • There are only three possible simply connected Riemannian surfaces: C^\hat{\mathbb{C}}, C\mathbb{C} and D\mathbb{D}.
  • These three possibilities correspond to elliptic, flat and hyperbolic surfaces.
  • Details are in Dynamics in One Complex variable, chapter 1, section 2, third edition by John Milnor.

Notes

  • Uniquess of geometric classification comes from Gauss-Bonet theorem and trianglization of surfaces.
  • Another proof of hyperbolic surfaces uses pant decomposition of nn-torus Tn\mathrm{T}^n (with n2n \ge 2). See Reimannian Geometry, section 3.L.4 compact surfaces, third edition, by Sylvestre Gallot, Dominique Hulin and Jacques Lafontaine.