Introduction to Teichmuller Spaces

Jing Tao
Notes by Serena Yuan

1. Riemann Surfaces
Definition 1.1. A conformal structure is an atlas on a manifold such that the
differentials of the transition maps lie in R+ SO(n).

Definition 1.2. A Riemann surface is a 2-dimensional manifold together with

a conformal structure; or, equivalently, a 1-dimensional complex manifold.

Figure 1: Examples of Riemann Surfaces

1.1 Riemanns Goal

Riemanns goal was to classify all Riemann surfaces up to isomorphism; i.e. up to

biholomorphic maps.
There are two types of invariants:
discrete invariants, which arise from topology (for example, genus)

continuous invariants (called moduli ), which come from deforming a conformal

structure.

1
 Figure 2: Conformal Deformation

1.2 Riemanns Idea

Riemanns idea was that the space of all closed Riemann surfaces up to isomor-
phism is a manifold, a geometric and topological object:

M = {closed Riemann surfaces}/

[
= Mg ,
g0

where Mg = {genus g Riemann surfaces}/ is a connected component of M . Now

the goal is to understand the topology and geometry of each Mg .

2. Uniformization
We will now investigate why genus is the only discrete invariant. Given a Riemann
surface Xg , its conformal structure lifts to its universal cover, Xg . Uniformization
Theorem says:
C if g = 0


Xg := C if g = 1

H2 if g > 2

Remarks.

i. Each of C, C, H2 has a distinct natural conformal structure.

ii. For g=0, Xg

= C so M0 = {C}.

iii. Each of C, C, H2 admits a Riemannian metric of constant curvature, which is

compatible with its natural conformal structure.

2
 C C H2
1 0 -1

So Xg admits a metric of constant , and we can identify

Mg = {genus g Riemann surfaces with constant curvature }/isometry

(For g=1, we need to normalize area as well.)

3. Teichmuller Space
We fix a topological surface S of genus g.

Definition 3.1. A marked Riemann surface (X, f ) is a Riemann surface X to-

gether with a homemorphism f : S X. Two marked surfaces (X, f ) (Y, g) are
equivalent if gf 1 : X Y is isotopic to an isomorphism.

Definition 3.2. We define the Teichmuler Space

Tg = {(X, f )}/

For g 2, Tg is also the set of marked hyperbolic surface (X, f ), where the equivalent
relation is given by isotopy to an isometry.

There is a natural forgetful map Tg Mg by sending (X, f ) 7 X. We note that

(X, f ) and (X, g) are equivalent in Mg if and only if exists an element h Homeo+ (S)
such that f = gh1 , where h well-defined up to isotopy. This introduces:

Definition 3.3. The mapping class group is

g = Homeo+ (S)/Homeo0 (S),

where Home0 (S) is the connected component of the identity.

We define an action of g y Tg by (X, f ) 7 (X, f h1 ). By the above discussion,
Tg /g = Mg .

5. Topology on Tg

3
 Teichmuller space Tg is naturally a manifold homeomorphic to R6gg , and g acts
properly discontinuously on Tg . Thus, Mg is an orbifold with 1orb (Mg ) = g .
We are able to see the topology in two ways:
By Representation theory:

Tg , Hom(1 (S), P SL2 (R))/P SL2 (R) = char2 (1 (S)),

where the image of Tg is the open subset of discrete and faithful representations. A
simple counting argument shows

dim(g ) = dim char2 (G) = (2g 1) 3 3 = 6g 6.

By Fenchel-Nielson Coordinates:
Example 5.1. Dehn Twist: We define an element D g , where is a simple
closed curve on S.

Figure 3: Dehn Twist

Example 5.2. Fenchel-Nielson coordinates on T1,1 (The Teichmuller space of the

once-punctured torus):
Given the once-punctured torus S. Fix , on S, will be a pants decomposition
of S and a seam. Let (X, f ) T1,1 . As shown in Figure 4, then the map f
identifies with a curve (also called) in X. Let ` = `X () be the length of the
unique geodesic in X in the homotopy class of .
As seen on the right side of the figure, in hyperbolic geometry, there exists a
unique arc that intersects perpendicularly on both sides. Let be the arc in
between the foots of the of . Now let 0 = . This is a closed curve which differs
from the image of in X by some power of Dehn twist along , i.e. 0 = Dn ().
We define
= n` + `x ()

4
 Definition 5.3 (FN Coordinates). The Fenchel-Nielsen coordinates relative
to the curves (, ) is
T1,1 R+ R, X 7 (`, )

Figure 4: FN on T1,1

In general (for higher-dimensional cases), we need to fix a pants decomposition

= {1 , ..., 3g3 } on S and a set of 3g 3 seams. Then the FN coordinates relative
to is
3g3
Tg R+ R3g3
X 7 (`1 , ..., `3g3 , 1 , ..., 3g3 )

6. Teichmuller Metric
(Or how to compare conformal structures)

If two points in Teichmuller space (X, f ) 6= (Y, g), then gf 1 : X Y is not

homotopic to a conformal map. Our goal is to quantify how far gf 1 is from being
conformal.
Let h : X Y be an orientation-preserving diffeomorphism. For p X,, we have

(dh)p : Tp X Tf (p) Y

(dh)p is R-linear, but not necessarily C-linear. There is a decomposition

!
a 0
(dh)p = R S,
0 b

where R and S are rotations, and a, b > 0.

5
 Definition 6.1. The dilatation at p as

max{a, b}
Kp = 1
min{a, b}

Definition 6.2. The dilatation of h is

Kh = supp Kp 1

We have:
(i) (dh)p is C-linear iff a = b iff Kp = 1
(ii) h is conformal iff Kh = 1.
Definition 6.4. h is a quasi-conformal map if Kh < . This holds automati-
cally if X is compact.
Definition 6.3 (Teichmuller Distance). The define the Teichmuller Dis-
tance is
1
dT ((X, f ), (Y, g)) = log inf1 Kh
2 hgf

where inf hgf 1 Kh is the smallest dilatation of a quasi-conformal map preserving the
marking.
Lemma. dT is a metric.

Figure 5: Ex. of extremal map h

6
 Example.
Consider !
2 0
h=
0 21
We see Kh = 4. h turns out to be the unique extremal map. This means that any
map h0 h has bigger dilatation, Kh0 > Kh . Hence dT (X, Y ) = log(4)
2
.

Definition 6.5 (Quadratic Differential). A quadratic differential on

X Tg , is q : T X C. Locally, q has the form q = q(z)dz 2 where q(z) is
holomorphic.
Remark. q has 4g 4 zeroes counted with multiplicity.
p If p not a zero of q, q(0) 6= 0 in local coordinates, then we can
Definition 6.6.
take a branch of q(z) and integrate to obtain a natural coordinates for q:
Z p
= q(z)dz, q = d 2

The transition of natural coordinates (or the change of charts between natural
coordinates) includes translations and possible sign flip, since d 2 = (d 0 )2 so 0 =
+ c.
So defines a (singular) flat Euclidean metric |d|2 on X (singularities come
from the zeros of q). Conversely, a collection of natural coordinates determines a
quadratic differential.
Example.
If we take X from the previous example, then let q = dz 2 .
Let QD = {quadratic differentials on X}. By Riemann-Roch, QD is a complex
vector space of dimC = 3g 3. Also, QD(X) = Tx (Tg ) = Tx (Mg ).
Definition 6.7. We define an L1 norm on QD(X). Let q = q(z)dz 2 . Let
Z
||q||1 = |q(z)|dzdz

This is just the area of X in the (singular) flat metric.

Definition 6.8. For a point X Tg and q QD(X), denote the open unit ball
by QD1 (X) = {||q|| < 1}.
Definition 6.9 (Teichmuller Map).
For X Tg and q QD1 (X), let

1 + ||q||
K= 1.
1 ||q||

7
Set = u+ iv to be a natural coordinate for q, and define a new natural coordinate
by 0 = Ku + i 1K v. This new coordinate 0 determines a surface Yq Tg and a
hq
canonical map X Yq , called a Teichmuller map.
Theorem 6.10. We have
(i) hg is the unique extremal map in its homotopy class.
(ii) QD1 (X) Tg such that q 7 Yq is a homeomorphism.
Consequences.
(i) dT is complete.
t t
(ii) t 7 e 2 u + ie 2 v defines a bi-infinite geodesic line in this metric.
(iii) Any X, Y Tg is connected by one and only one segment of such a line.
Remarks.
(i) (T, dT ) = (H2 , hyperbolic metric) but for g 2, (Tg , dT ) is not hyperbolic in any
sense. (Masur, Masur-Wolf, Minsky)
(ii) Geodesic rays do not always converge in the Thurston boundary. (Lenzhen)
(iii) (Masur-Minsky, Rafi) gave a combinatorial descriptions of Teichmuller geodesics.

7. Weil-Petersson Metric
(or L2 -norm on QD(X))
A point X Tg is a hyperbolic surface. Write the hyperbolic metric in local
coordinates as ds2 = (z)|dz|2 . For q1 , q2 Tg , define a Hermetian inner prodcut on
QD(X) by
Z
q1 (z)q2 (z)
h(q1 , q2 ) = dzdz
X (z)
Remarks.
(Tg , h) is a Kahler manifold, that is Tg has three natural structures that are all
compatible with each other:
a complex structure
a Riemannian structure, the associated Riemannian metric called the Weil-
Peterssoon metric is gp = Real(h)
and a symplectic structure, the associated WPsymplectic form (i.e. a closed
(1, 1) form) is = Im(h).
Theorem 7.1 (Walperts Formula).
Choose a set of FN coordinates on Tg
3g3
: Tg R+ R3g3

8
 X 7 (`1 , ..., `3g3 , 1 , ..., 3g3 )
Then the WP sympletic form is
3g3
1X
= d`i di
2 i=1

Example.
For T1,1 , its natural complex structure is H2 . For y large, xy , ` xy , therefore

1
= d` d (dx dy),
y3
thus
1
gwp (dx2 + dy 2 )
y3
when y is large.

Figure 6: T1,1 Ex. of Walperts Formula

We that that the arc length of the imaginary axis y13 2 |dz| < . This implies
R

that gwp is incomplete.

Also, wp y for y large, so gwp has negative Gaussian curvature with
sup = . But wp is bounded away from 0.
Remarks.
(i) In general, the WP metric is always incomplete.

9
(ii) It always has negative sectional curvature, but for dimC (Tg ) > 2, sup wp = 0
and inf wp = (Huang).
(ii) (Brock) showed (Tg , gwp ) is quasi-isomorphic to a pants graph.

8. Thurston Metric
(or how to compare hyperbolic structures)

Definition 8.1.
A map h : X Y is a Kh -Lipschitz map
d(h(x), h(y)) Kh d(x, y)
Definition 8.2. For X, Y Tg , define
L(X, Y ) = inf1 Kh
hgf

where h is a Lipschitz homeomorphism.

Lemma (Thurston). L(X, Y ) 1 and is not necessarily symmetric.
Definition 8.3 (Thurston distance). The Thurston distance is dL (X, Y ) =
logL(X, Y ) which by the preceding lemma is an asymmetric metric.
It is also complete.
Theorem 8.4 (Thurston).

`Y ()
L(X, Y ) = sup ,
`X ()
where ranges over all simple close curve on S.
Lemma. If is a simple closed curve which is a short curve on X or dual to a
short curve on X, then
+ `y ()
L(X, Y )  max .
`x ()
+
( is = up to additive error)
We do some examples of finding the Thurston distance between points in T1,1 .
On i, the length of is i, and the length of is 1/y on yi, thus
+
dL (yi, i)  log(y).

On the other hand, by the collar lemma, the length of the blue curve is log(y),
hence
+
dL (i, yi )  log(log(y)).

10
 Figure 7: lengths on T1,1

On 1 + yi, the length of the blue curve is log(y) + y1 , hence

+ 1 + 1
dL (yi, 1 + yi)  log(1 + ) .
y log y y log y
Now give a large integer n, let y log y = n, so d(yi , n + yi )  1. We see that
dL (i, yi ) + d(yi , n + yi ) + d(n + yi , n + i)  log n  dL (i, n + i).
9. Description of Geodesics
We can give the following description of geodesics X, Y Tg :
Definition 9.1. A map h : X Y is called T extremal if Kh = L(X, Y ).
Theorem 9.2 (Thurston). The set h extremal {stretch locus of h} is a geodesic
lamination (X, Y ), called the maximally-stretched lamination.
Remarks.
(i) Env(X, Y ) = {geodesics from X to Y } =6 but |Env(X, Y )| can be infinite. Each
element of Env(X, Y ) must stretch (X, Y ) maximally.
(ii) Elements in Env(X, Y ) do not necessarily fellow-travel, the reversal a geodesic
from X to Y may not be a geodesic from Y to X, even after reparametrization
(Lenzhen-Raf-T)
(iii) From the coarse perspective, the shadow map from Tg to the curve complex
Tg C(S)

11
defined by sending X to a short curve on X sends every Thurston geodesic to a
reparametrized quasi-geodesic in C(S) (LRT). The same statement is not true if we
replace S by a proper subsurface of S.
Open Questions.
1. Are there preferred geodesics in Env(X, Y )?
2. Is there a combinatorial description (in the sense of Rafi) of a Thurston geodesic?
Is there a distance formula?
3. What does Env(X, Y ) look like? In T1,1 , Env(X, Y ) is the intersection of two
cones; a complete understanding is in progress (Dumas-Lenzhen-Rafi-Tao).

12