MAPPING CLASS GROUP NOTES

WADE BLOOMQUIST

These notes look to introduce the mapping class groups of surfaces, explicitly
compute examples of low complexity, and introduce the basics on projective
representations that will be helpful going forward when looking at quantum
representations of mapping class groups. The vast majority of the material
is modelled off how it is presented in A Primer on Mapping Class Groups by
Benson Farb and Dan Margalit. Unfortunately many accompanying pictures
given in lecture are going to be left out due to laziness.

1. Fixing Notation
Let Σ be a compact connected oriented surface potentially with punctures.
Note that when the surface has punctures then it is no longer compact, but it
is necessary to start with an un-punctured compact surface first. This is also
what is called a surface of finite type.
• g will be the genus of Σ, determined by connect summing g tori to the
2−sphere.
• b will be the number of boundary components, determined by removing
b open disks with disjoint closure
• n will be the number of punctures determiend by removing n points
from the interior of the surface
We will denote a surface by Σbg,n , where the indices b and n may be excluded
if they are zero. We will also use the notation S 2 for the 2−sphere, D2 for the
disk, and T 2 for the torus (meaning the genus 1 surface).
The main invariant (up to homeomorphism) of surfaces is the Euler Char-
acteristic, χ(Σ) defined as
χ(Σbg,n ) := 2 − 2g − (b + n).
The classification of surfaces tells us that Σ is determined by any 3 of
g, b, n, χ(Σ).

2. Mapping Class Group Basics

Definition 1. Homeo+ (Σ, ∂Σ) is the group of orientation preserving self-
homeomorphisms of Σ which restrict to the identity on ∂Σ.
1
2 WADE BLOOMQUIST

Definition 2.
MCG(Σ) := π0 (Homeo+ (Σ, ∂Σ)).
In words, the mapping class group of Σ is the group of isotopy classes of
Homeo+ (Σ, ∂Σ), where isotopies must fix the boundary pointwise.
We will always take f to be a mapping class which is the isotopy class of φ,
meaning f = [φ] ∈ M CG(Σ).

2.1. Punctures, Marked Points, and Boundary Components. If the

n puncutes of Σ are instead thought of marked points on the surface, then
MCG(Σ) is the group of self-homeomorphisms that leave the set of marked
points invariant modulo isotopies which also leave the marked points invariant
When looking at Σg,n it is tempting to use that Σg,n is homeomorphic to
the interior of Σng , but one needs to be careful. In particular, based on our
definition we must restrict to the identity on the boundary, and so boundary
components cannot be permuted, while punctures need only be fixed set-wise.

3. Alexander’s Trick
Theorem 3. MCG(D2 ) is trivial.
Reworded: Any oreientation preserving self-homeomorphism φ of D2 , that
is the identity on ∂D2 , is isotopic to the identity through homeomorphisms
that are the identity on ∂D2 .
Proof. Identify D2 as {(x, y) ∈ R2 : x2 + y 2 ≤ 1}. Let φ : D2 → D2 be the
identity on ∂D2 . Now define
(
x
(1 − t)φ( 1−t ) 0 ≤ |x| < 1 − t
F (x, t) :
x 1 − t ≤ |x| ≤ 1
This is best seen in Figure 1. 
We also have
Theorem 4. MCG(Σ10,1 ) is trivial.
Proof. The same homotopy as above can be used. This is seen in the vertical
line rising from the tip of the cone corresponding to the puncture. 
Theorem 5. MCG(Σ0,1 ) and MCG(S 2 ) are trivial.
Proof. Sketch: We can just use stereographic projection (and if we are careful
abound a point being fixed in the S 2 case) then we can use the straight line
homotopy in R2 . 
 MAPPING CLASS GROUP NOTES 3

Figure 1. The level sets of F as the support of the cone in the cylinder

4. The 3−punctured Sphere

Lemma 6. Any two essential simple proper arcs with the same end points in
Σ0,3 are isotopic.
Proof. Let α and β be two arcs connecting the same marked points. In general
position α and β can be taken to intersect transversely.
Now the third disjoint point can be used in applying sterographic projection
to the plane.
Now if α and β interesect, then there exists an innermost disk bounded by
subarcs of α and β. Then α can be isotoped across this disk to reduce the
intersection number until α and β and disjoint away from their end points.
Now cut Σ0,3 along α ∪ β. This gives a disk disjoint union a once punctured
disk (this uses the classification of surfaces). Thus we have that α and β
originally bounded an embedded disk and so are isotopic. 
Theorem 7. MCG(Σ0,3 ) ∼ = S3 , where S3 is the permutation group on 3 ele-
ments.
Proof. Define ψ : MCG(Σ0,3 ) → S3 which sends φ to the permutation induced
on the 3 marked points. This is clearly a surjective homomorphism.
Now assume that φ fixes the 3 marked points, meaning the induced per-
mutation is the identity. Now let α be an arc connecting two of the marked
points. By assumption φ(α) is an arc between the same two marked points (as
they are fixed by φ). Thus we have that α and φ(α) are isotopic. Then using
a face from point set topology, we have that φ is isotopic to a map which fixes
α point wise, and as such we may assume that φ fixes α pointwise.
4 WADE BLOOMQUIST

Now cut Σ0,3 along α to obtain a disk with a marked points (and two
marked points on the boundary, but that is not relevant). Then we can apply
Alexander’s trick to the induced map φ̂ on the disk to get that φ is homotopic
to the identity. Thus we have that ψ is injective, and so an isomorphism. 
5. The Annulus
Let A denote Σ20,0 the annulus.
Theorem 8. MCG(A) ∼ =Z
Proof. The universal cover of A is the infinite strip, Ã = R × [0, 1], as seen in
figure 2.

Figure 2. A poorly drawn figure of the universal cover of the annulus

Now any φ : A → A has a preferred lift φ̃ which fixes the origin. Now let
φ̃1 : R → R be the restriction to R × {1}. This is a lift to R of the identity
map, and thus an integer translation.
Now define ρ : MCG(A) → Z by ρ(f ) = φ̃1 (0), or just φ̃1 if we identify Z
with the group of integer translations. This again is clearly a homomorphism.
We need to show that ρ is surjective. Take the linear transformation of R2
given by  
1 n
M= .
0 1
 MAPPING CLASS GROUP NOTES 5

We see that this preserves à and is equivariant. Thus we descend to a home-

omorphism φn : A → A, and
ρ([φn ]) = n
by definition.
Now we must show that ρ is injective. Let [φ] = f ∈ ker(ρ), with preferred
lift φ̃. Then as ρ(f ) = 0 we know that φ̃ acts as the identiy on ∂ Ã. Now we
claim that the staright line homotopy form φ̃ to Id : Ã → Ã is equivariant,
meaning φ̃(τ · x) = τ · φ̃(x) for any deck transformation τ and any x ∈ Ã. We
have that
φ̃(τ · x) = φ∗ (τ ) · φ̃(x).
but we have that φ fixes ∂A pointwise so φ∗ is the identity. Thus we have that
φ̃ is homotopic to the identity by the straight line homotopy and this descends
to a homotopy of A that fixes A pointwise, making ρ injective, and thus an
isomorphism.


6. The Torus
Theorem 9. T2 ∼
= SL(2, Z).
Proof. Define σ : MCG(T 2 ) → SL(2, Z) which sends φ to the induced map
φ∗ : H1 (T 2 , Z) → H1 (T 2 , Z).
As φ is invertible φ∗ is an automorphism, and H1 (T 2 , Z) ∼= Z × Z. So φ 7→ φ∗
∼
induces a map into Aut(Z × Z) = GL(2, Z).
Now we have that algebraic intersection numbers in T 2 correspond to deter-
minants in GL(2, Z) and these are preserved by orientation preserving home-
omorphisms, meaning the map is into SL(2, Z).
We look to show that σ is surjective. Take M ∈ SL(2, Z). This induces an
orientation preserving linear homeomorphism of R2 that is equivariant with
respect to deck transformations (meaning Z × Z), and thus this descends to a
linear homemorphism φM of T 2 ∼ = R2 /Z2 . Now identifiying primitive vectors in
Z×Z with homotopy classes of simple closed curves gives us that σ([φM ]) = M .
Now we must show that σ is injective. Suppose that
 
1 0
σ(f ) = ∈ SL(2, Z).
0 1
and let α and β be the meridian and longitude of the torus (the (1, 0) and
(0, 1) curves). The φ(α) is isotopic to α and φ(β) is isotopic to β. Up to
isotopy φ fixes α pointwise, and preserves the two sides of α. Now cut along
alpha to give an annulus. This induces a homeomorphism φ̄ : A → A. Then
β and φ̄(β) are arcs in A. As they are isotopic we knwo that ρ(f¯) = 0, using
6 WADE BLOOMQUIST

the map from the previous proof. As this map is injective we have that f¯ is
¯ and thus so is φ. This gives us that σ is injective,
the identity, and so is phi,
and thus an isomorphism. 

A similar argument can be used to show that MCG(Σ1,1 ) ∼ = SL(2, Z) just

by being careful to say that each linear homeomorphism of R2 can be taken
to fix the origin.

7. The 4 Punctured Sphere

7.1. The Hyperelliptic Involution. Define the map
i : T2 → T2
which rotates the square about the center, where the torus is viewed as the
square with top and bottom identified and with the left and right edges iden-
tified.
This map has 4 fixed points. These are the center of the square, the vertex
in the corner, and the midpoint of the two edges. This tells us that T 2 /i is
Σ0,4 .
Lemma 10. The hyperelliptic involution induces a bijection between the simple
closed curves in T 2 and the simple closed curves in Σ0,4 .
Proof. In T 2 the simple closed curves are given by (p, q) curves. Let α and β
be two simple closed curves in T 2 which intersect once. Then identify α with
(1, 0) ∈ Z × Z and β with (0, 1) ∈ Z × Z. Let (p, q) be primitive in Z × Z. A
simple closed curve γ is called a (p, q) curve if (int(γ, β), int(γ, α)) = ±(p, q).
Now take p parallel copies of α and twist by a 2π
q
twist along β. Up to homotopy
we may assume i projects α and β to simple closed curves ᾱ and β̄ in Σ0,4
which intersect in not one, but now two points.
Now we can similarly take p copies of ᾱ and twist along β̄ by πq to get (p, q)
curves on Σ0,4 .
Now let γ be an arbitrary simple closed curve in Σ0,4 . Up to homotopy
we may assume γ is in minimal position with respect to ᾱ. Now cut Σ0, 4
along β̄ to get two twice punctured disks, with ᾱ and γ giving collections of
disjoint arcs. Now minimal position of γ with resepct to α ensures that each
arc is essential. Thus using the argument preceeding the discussion of the
three punctued sphere we have that these arcs must be freely homotopic and
thus γ is (p, q) curve as constructed.
Finally we have that the pre-image of a (p, q) curve is a (2p, 2q) curve on T 2
which is just two copies of a (p, q) curve.

 MAPPING CLASS GROUP NOTES 7

7.2. The 4 Punctured Sphere.

Theorem 11. MCG(Σ0,4 ) ∼
= PSL(2, Z) n (Z/2Z × Z/2Z)
Proof. First we define a homomorphism σ̄ : MCG(Σ0,4 ) → PSL(2, Z), and then
we will construct a right inverse which has a kernel isomorphic to Z/2Z×Z/2Z
to prove our claim.
For [φ] ∈ MCG(Σ0,4 ) there are two lifts of φ to Homeo+ (T 2 ), namely φ̃ and
iφ̃. Now define
σ̄(f ) = σ([φ̃])
.
Then we note that σ(i) = −Id, and we have our desired homomorphism
into PSL(2, Z).
Now we need a right inverse of σ̄. An element of PSL(2, Z) gives an orien-
tation preserving linear homemorphism of T 2 well defined up to hyperelliptic
involution. Now each map commutes with i and thus induces an orientation
preserving homeomorphism of Σ0,4 , so we have a map
PSL(2, Z) → MCG(Σ0,4 ),
and this map is cearly a right inverse so σ̄.
Now let i1 and i2 be the maps which are rotation by π along the axes
determined by the equaters seperating the marked ponts, as seen in figure 3.
Now that i1 and i2 generate a Z/2Z × Z/2Z subgroup fo MCG(Σ0,4 ).

Figure 3. The axis of rotation for i1 and i2

We have that i1 and i2 lift to rotation by π in one of the factors of T 2 when

viewed as S 1 × S 1 , and are not seen by σ̄. Now we claim that hi1 , i2 i = ker(σ̄).
8 WADE BLOOMQUIST

Take f ∈ ker σ̄. So f˜ ∈ Homeo+ (T 2 ) acts by ±Id on H1 (T 2 , Z), and so acts

trivially on the simple closed curves in T 2 . Now our lemma gives us that f
also acts trivially on homotopy classes of simple closed curves in Σ0,4 .
Thus we have f fixes ᾱ and β̄ so we can precompose f with k ∈ hi1 , i2 i so
that f k fixes the 4 punctures. Then say that f k = [φ] so up to isotopy φ fixes
ᾱ and β̄ pointwise. So we hav ethat φ fixes the parked points, and cutting
along ᾱ and β̄ allows us to use Alexander’s trick. 

8. Dehn Twists
For any surface Σ, let α be a simple closed curve. Let N be a regular annular
neighborhood of α, with φ : A → N . Now define Tα : Σ → Σ by
(
φ ◦ T ◦ φ−1 (x) x ∈ N
Tα (x) = .
x x∈S−N
Where T is the map ρ−1 (1) given in proof of the mapping class group of the
annulus.
Theorem 12 (Humphries). The mapping class group of the closed genus g
surface is generated by Dehn twists about the 2g + 1 curves seen in figure 4.

Figure 4. The Humphries Generators

The proof of this is more than we will be able to cover. A more managable
proof is that the collection of all Dehn twists generates the mapping class
group, but we also won’t cover that. Dropping down to finitely many Dehn
twists is difficult, and Humphries also proved that there are no generating sets
of fewer than 2g + 1 Dehn twists (but you can generate with only elements if
they don’t need to be Dehn twists).

9. The Capping Homomorphism

Let Σ0 be the surface obtained from Σ by capping the boundary component
β (where β is the curve parallel to the boundary component), with a once punc-
tured disk where the puncture is called p0 . Now let MCG(Σ, {p1 , ..pk }) be the
subgroup of MCG(Σ) fixing p1 , ..., pk , and similarly for MCG(Σ0 , {p0 , p1 , ..., pk }).
 MAPPING CLASS GROUP NOTES 9

Now define cap : MCG(Σ, {p1 , ..., pk }) → MCG(Σ0 , {p0 , ..., pk }) by the in-
duced homomorphism. Then the following sequence is exact:
1 → hTβ i → MCG(Σ, {p1 , ..., pk }) → MCG(Σ0 , {p0 , ..., pk }) → 1.
9.1. Pair of Pants.
Theorem 13. MCG(Σ30,0 ) ∼
=Z×Z×Z
This is just repeated use of the capping homomorphism as well as earlier
result on the 3 puncured sphere.
9.2. The Torus with One Boundary Component.
Theorem 14. MCG(Σ11 ) ∼ ˜
= SL(2, Z)
Let SL(2,˜ Z) = ha, b : aba = babi, and recall that SL(2, Z) = ha, b : aba =
bab, (ab)6 = 1i. Now using the capping homomorphism we have
1 Z ˜ Z)
SL(2, SL(2, Z) 1

1 hTβ i MCG(Σ11 ) MCG(Σ1,1 ) 1

and the 5−Lemma gives the desired isomorphism.

10. Projective Representations

Let V be a complex vector space, and define
PGL(V ) := GL(V )/C∗ Id.
Definition 15. Let G be a group. A projective representation of G is a ho-
momorphism
ρ : G → PGL(V ).
Put another way this is a collection of operators index by G, meaning an
operator ρ(g) for each g ∈ G, and
ρ(gh) = c(g, h)ρ(g)ρ(h),
where c(g, h) is a constant depending on g and h. This description won’t be
focused on here, but it tells us that there is group cohomology lurking in the
background.
Given a linear representation r : G → GL(V ) we can always define a pro-
jective representation using the quotient map π
r π
G GL(V ) PGL(V )
π◦r
10 WADE BLOOMQUIST

Now given a projective representation ρ : G → PGL(V ), can you find such

an r? Put another way, can projective representations be lifted to lienar
representations.
10.1. Central Extensions. A central extension of G is a short exact sequence
1 → A → E → G → 1.
Also said as E is an extension of G by A. So if A ≤ Z(E) we say this is a
central extension.
Now define H ≤ G × GL(V ) as
H := {(g, A) : π(A) = ρ(g).
Then
1 → {(e, cId) : c ∈ C∗ } → H → G → 1
where φ : H → G by (g, A) 7→ g. Then we note that ker(φ) = {(e, cId) : c ∈
C∗ } and so H is a central extension of G.
Now define σ : H → GL(V ) by (g, A) 7→ A, then this is a linear representa-
tion and π(σ(g, A)) = ρ(g) = ρ(φ(g)). And so we almost were able to find a
lift, but at the cost of passing to a central extension.
11. Quantum Representations
Theorem 16. A modular tensor category gives rise to a projective represen-
tation of the mapping class group for any surface.
In short
• Associate a vector space V (Σ)
• Give a projective action of Dehn twists