Three manifold

SHUMIN JIANG

July 2022
Contents

1 Generalities on three manifolds and constructions 2

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Definitions and basic examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Constructing 3-manifold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.1 Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.2 Mapping torus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3.3 Connected sum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.4 Constructing 3-manifolds from surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 An example of mapping torus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4.1 The fundamental group of the mapping torus π1 (M T2 ) . . . . . . . . . . . . . . . . . 6

2 Examples of polyhedral manifolds 8

2.1 The construction of Poincaré homology via dodecahedron . . . . . . . . . . . . . . . . . . . . 8
2.2 The fundamental group of the Poincaré homology sphere . . . . . . . . . . . . . . . . . . . . . 10
2.2.1 Alternative method to compute the fundamental group . . . . . . . . . . . . . . . . . 11
2.2.2 Homology groups of the homology sphere . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.3 Other quotient spaces of platonic solids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3.1 Seifert–Weber space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

3 Heegaard splitting 15
3.1 3 dimensional Lens space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.2 Definition and examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

1
Chapter 1

Generalities on three manifolds and

constructions

1.1 Introduction
Since all the surfaces(2-manifolds), both orientable and non-orientable, are classified. This project started
with basic construction of 3-manifolds from surfaces, in the hope that we can built a classification of 3-
manifolds based on the classification of surfaces. But by simply constructing using products of 2-manifolds
and 1-manifolds, or mapping torus of a 2-manifold are not going to represent all 3-manifolds. That’s how
chapter 3 provides a more efficient tool - Heegaard splitting.
Chapter 2 is an introduction to polyhedral manifolds. It starts with the famous Poincaré homology
sphere. The chapter includes the construction, homology groups and fundamental groups. We found out
that the computational method used on the homology sphere can be generalized to all other 3 dimensional
polyhedral manifolds, and it is connected to the classification of all compact 3-manifolds.
Chapter 3 explains how polyhedral manifolds are related to classification theory of compact 3-manifolds
through a process called ’Heegaard splitting’. It was first introduced by Poul Heegaard in 1898. The claim
was all compact 3-manifolds can be splited into two so called handlebodies. To get back to the original 3-
manifold, simply identify the boundaries of the handlebodies - compact surfaces, with homeomorphisms.[1]

1.2 Definitions and basic examples

Definition 1. A topological n-manifold is a second countable Hausdorff space which is locally homeomorphic
to Rn , meaning there exists an open neighbourhood around each point on the manifold that is homeomorphic
to Rn .
From now on, manifold means topological manifold.
Definition 2. A manifold with boundary is a manifold with interior points locally homeomorphic to Rn , and
boundary points homeomorphic to Rn−1 × [0, ∞), which is half a n-dim Euclidean space.
Example 1. A one circle S 1 := {(x, y) | x2 + y 2 = 1} is a 1-manifold. It’s fundamental group is isomorphic
to Z, the homology groups are H0 ≃ H1 ≃ Z. The circle is the real line quotient by the action generated by
f := f (x) = x + 1.
Example 2. S 1 × S 1 is a product of two one manifolds, it is a 2-manifold. It’s a one-holed torus as shown
in the figure 1.1. It’s fundamental group is isomorphic to Z2 , homology groups are H0 ≃ H2 ≃ Z, H1 ≃ Z2 .
The universal cover of the 2-torus is R2 with a nice action generated by ⟨g, f ⟩, g(y) = y + 1 and f (x) = x + 1.
x is the x-direction, y is the y-direction.
Example 3. S 3 , the 3-sphere is defined to be {(x0 , x1 , x2 , x3 ) | x20 + x21 + x22 + x23 = 1}. It’s fundamental
group is trivial. The homology groups are H0 ≃ H3 ≃ Z, and H1 ≃ H2 ≃ 0. The universal cover is itself
since it’s simply-connected.

2
Example 4. The mapping torus of the one-holed torus is a 3-manifold constructed from a surface.
Example 5. 3-torus T3 := S 1 × S 1 × S 1 , is a triple product of three 1-manifolds. it’s fundamental group
is isomorphic to Z3 as shown later in chapter 2 with a different representation of the 3-torus. Its homology
groups are H0 ≃ H3 ≃ Z, H1 ≃ H2 ≃ Z3 .

1.3 Constructing 3-manifold

3-manifolds are harder to be given a complete list of or classified as families
of them. They are at the limit of our imagination, not as easy to visualise as
2-manifold, also not unreachable as 4-manifolds. That’s what makes them one
of the most interesting objects to study. Constructing a 3-manifold from lower
dimensional manifolds is one of the approaches.

1.3.1 Products
Figure 1.1: S 1 × S 1 Theorem 1. The Cartesian product with product topology of a n-manifold N
and a m-manifold M is a (m + n)-manifold M N .
Proof. • Prove for any point in MN, there exists an open neighbourhood homeomorphic to
Rm+n .
For a point x ∈ N and a point y ∈ M , there exist two open neighbourhoods Vx ≃ Rn ⊂ N and
Vy ≃ Rm ⊂ M , then (Vx , Vy ) is an open neighbourhood ≃ Rmn around point (x, y) in MN.
• Second countability follows from mapping torus claim 1, which is left as an exercise for the reader.
• The space is Hausdorff. Let (a,b), (c,d) be two distinct points in (m + n)-manifold called Z, which
means either a ̸= c or b ̸= d. a and c ∈ X, b and d ∈ Y , with X as the n-manifold, Y as the m-manifold.
Case 1
Without loss of generality, let a = c. Let U be open neighbourhood around a or c, Vb , Vd be disjoint
open neighbourhoods in Y, then U × Vb and U × Vd are disjoint.
Case 2
Let a ̸= c, Ua , Uc be disjoint open neighbourhoods in X, again Ua × Vb and Uc × Vd are disjoint.

1.3.2 Mapping torus

Definition 3. Given a topological space X and a homeomorphism f : X → X, the mapping torus M X of X
is X × [0, 1]/(x, 0) ∼ (f (x), 1).
Theorem 2. The mapping torus of an n-manifold is an (n+1)-manifold.
Proof. Let X be an n-manifold, X × [0, 1] is an (n+1)-manifold with boundary. Let f : X −→ X be a
homeomorphism, there exist open neighbourhoods Vx × [0, ε) and Vf (x) × (δ, 1] around x × {0} and f (x) × {1}
with f (Vx ) = Vf (x) , 0 < ε < δ < 1. (It’s open in a sense that it’s open in the (n+1)-manifold with boundary,
since we are taking the chart here.) Without loss of generality, assume x ̸= f (x), Vx and Vf (x) are disjoint,
no particular reason, but makes things easier in the diagram, see the figure below. By construction, exist
two homeomorphisms g : Vx × [0, ε) −→ {Rn+1 |xn+1 ⩾ 0} and f −1 ◦ g : Vf (x) × (δ, 1] −→ {Rn+1 |xn+1 ⩾ 0}.
Hence Vx × [0, ε) ∪ Vf (x) × (δ, 1]/Vx × {0} ∼ Vf (x) × {1} is homeomorphic to Rn+1 . As shown in the figure
1.2a, the red lines denotes open brackets, blue lines denotes closed brackets, gluing two blue lines gives an
open neighbourhood with just the red boundaries.
M X is Hausdorff.
Case 0: (See figure 1.2b in blue) Let x × {t1 } and x × {t2 } for x ∈ X and t1 , t2 ∈ (0, 1). At least one of
x ̸= y or t1 ̸= t2 is true. This case is trivial.
Case 1: (See figure 1.2b in green) If x = y, without loss of generality, let x×{t1 } and x×{0} for t1 ∈ (0, 1).
There exists a small enough ϵ and an open neighbourhood Vx around x such that Vx ×(t1 −ϵ, t1 +ϵ) is disjoint

3
 (a) Homeomorphic to Rn (b) Hausdorff

Figure 1.2

to Vx ×[0, ϵ) and f (Vx )×(1−ϵ, 1]. Hence Vx ×(t1 −ϵ, t1 +ϵ) is disjoint to Vx ×[0, ϵ)⊔f (Vx )×(1−ϵ, 1]/V1 ×{0} ∼
f (Vx ) × {1}.
Case 2: (See figure 1.2b in red)
Take two points x ̸= y ∈ X and their disjoint open neighbourhoods Vx and Vy , this is possible since X is
Hausdorff, then f (Vx ) and f (Vy ) are open in X. Because f (x) ̸= f (y), we can shrink the neighbourhood of
f (Vx ) and f (Vy ) such that they are disjoint, call it Ux and Uy . Then f −1 (Ux ) and f −1 (Uy ) is also open and
disjoint given they are the subsets of Vx and Vy . Perform the same quotient as before, f −1 (Ux ) × [0, ε) ∪
Ux × (δ, 1]/f −1 (Ux ) × {0} ∼ Ux × {1} and f −1 (Uy ) × [0, ε) ∪ Uy × (δ, 1]/f −1 (Uy ) × {0} ∼ Uy × {1} are
disjoint open neighbourhoods. Hence the mapping torus is Hausdorff.

Claim 1. The product of two second-countable spaces is second-countable.

Proof. Suppose X and Y are both second countable. There exists countable bases Ui and Vj for X and Y
respectively, hence Ui × Vj is countable. Since for any x ∈ X and y ∈ Y , (x, y) ∈ Ui × Vj , as a result Ui × Vj
is a countable basis for X × Y . (See figure 1.2a) [2]

Claim 2. If space X is second-countable and the equivalent relation is open then the quotient space X/ ∼ is
second-countable.
Second-countable can be proved by proving the equivalent relation is open, meaning the projection
map π : X × [0, 1] −→ X × [0, 1]/ ∼ is open.

1.3.3 Connected sum

Definition 4. Let M1 and M2 be two n-manifolds and B1n ⊂ M1 , B2n ⊂ M2 are two open balls embedded
in the manifolds. The boundary of the open balls are homeomorphic to S n−1 . Two inclusion map i1 :
S n−1 → ∂B1n and i2 : S n−1 → ∂B2n . Take equivalence relation ∼:= x1 ∼ x2 if and only if there exist y such
that i1 (y) = i2 (y). And the connected sum M1 #M2 is defined as (M1 \B̊1n ) ⊔ (M2 \B̊2n )/ ∼. i2 ◦ i−1
1 is a
homeomorphism. [3]
Theorem 3. Connected sum of any 3-manifolds is a 3-manifold.

Proof. For any point on the manifold, there exists an open neighbourhood homeomorphic to R3 :
Case 1: For any point ∈ (M1 \B13 ) ⊔ (M2 \B23 ), this is trivial.
Case 2: Take the point i1 (y) ∈ ∂B13 and i2 (y) ∈ ∂B23 such that i1 (y) = i2 (y). Take two open neighbour-
hoods ≃ R2 × [0, ∞) around point y: V1 ⊂ (M1 \B̊13 ) and V2 ⊂ (M2 \B̊23 ), such that i1 (V1 ) = i2 (V2 ). Then
V1 ⊔ V2 / ∼ is homeomorphic to R3 .
Hausdorff and second-countability is similar to that of mapping torus.

4
1.3.4 Constructing 3-manifolds from surfaces
A natural question is, can we classify all 3-manifolds by building up on the classification of surfaces?
Definition 5. A surface is a 2-manifold.

Definition 6. A genus g surface is a surface homeomorphic to connected sum of g 1-tori for g ∈ Z+ . When
g = 0, the surface is defined to be homeomorphic to S 2 .
Theorem 4. Classification of orientable surfaces:
A compact, connected and orientable surface are homeomorphic a genus g surface for some g ⩾ 0.
Definition 7. A submanifold is a subset N ⊂ M , such that there exist an open neighbourhood Ux for any
point x ∈ M , and a homeomorphism ψ : Ux → V where V ∈ Rm . ψ(x) = 0, ψ(Ux ∩ N ) = 0m−n × Rn . n is
called the dimension of the submanifold. [3]
Definition 8. A subsurface is a 2 dimensional submanifold.
Proposition 1. Any sub-surface of an orientable surface is orientable.

Proof. Assume there is a subsurface Z ⊂ S, since Z is a submanifold, there exist charts (ϕZi , VZi ∩ Um ). For
all VZi ∩ VZj ∩ Um ̸= ∅, ϕZi ◦ ϕ−1
Zj : ϕZj (VZi ∩ VZj ∩ Um ) → ϕZi (VZi ∩ VZj ∩ Um ) is a orientation-preserving
map. See definition in the notes [3].
There is an easier definition of non-orientability for a specific type of manifolds.
Claim 3. For a manifold M in the form of M = Rn /Γ (where Γ is a group acting nicely on Rn ), if there
exists at least one element of Γ that doesn’t preserve the orientation of Rn , then M is not orientable.
Note that this is a necessary but not sufficient condition. Try example RP2 .
Example 6. Prove that the Möbius band defined as R2 /Γ, where Γ is defined as ⟨γ1 , γ2 ⟩, with
 γ1 ((x,
 y)) =
1 0
(x+1, 1−y), and γ2 ((x, y)) = (x, y+1) is non-orientable. This is because the det(Dγ1 ) = det = −1,
0 −1
which is negative. D stands for Jacobian.
something about non orientable, and refer to the paper about cutting polygons
Ways to construct 3-manifolds
• The mapping torus of any surfaces is a 3-manifold.
•The product of a surface and a 1-manifold is a 3-manifold.

1.4 An example of mapping torus

Mapping tori are harder to comprehend compared to products, hence an example is given below.
Example 7. Consider M T2 as the mapping torus of a 2-torus via a homeomorphism defined by

f : T2 −→ T2
(x, y) 7−→ (x + y, y)
 
1 1
or a matrix representation of the homeomorphism is . Compute the homology groups of M T2 .
0 1
To solve the problem, we need to introduce Mayer-Vietoris Sequence first.
Definition 9. Mayer-Vietoris Sequence
Φ Ψ ∂
· · · → Hn (A ∩ B) −
→ Hn (A) ⊕ Hn (B) −
→ Hn (X) −
→ Hn−1 (A ∩ B)

A variation of Mayer-Vietoris Sequence:

5
 f∗ −g∗ ∗ i f∗ −g∗
· · · → Hn (X) −−−−→ Hn (Y ) −→ Hn (Z) → Hn−1 (X) −−−−→ Hn−1 (Y ) → · · ·

where f and g are maps from X to Y, and space Z is X × I ⊔ Y /(x, 0) ∼ f (x) and (x, 1) ∼ g(x). To
prove the sequence is a long exact sequence, see Allen Hatcher book P 151. (By doing a relative homology
(X × I, X × ∂I)).

Sol 1. The question is an example for the sequence above when g is an identity map, and f is the homeo-
morphism via a matrix.
f∗ −1∗ ∗ i f∗ −1∗
· · · → H1 (X) −−−−→ H1 (X) −→ H1 (M T2) → H0 (X) −−−−→ H0 (X) → · · ·
f∗ −1∗ ∗ i f∗ −1∗
···→ Z2 −−−−→ Z2 −→ H1 (M T2 ) −
→ Z −−−−→ Z → · · ·
A B C D

with names of the maps under the arrows. The image of D is trivial since for all [a] ∈ H0 (X), (f∗ −
1∗ )[a] = f∗ [a] − 1∗ [a] = [b] − [c]. Since the space X is connected, all the points are in the same class, the
image is trivial.
 Similarly, f∗ − 1∗ acts on Z2 as matrix transformation acts on a 2×1 vector. The resulting
0 1
matrix is . The rank of the resulting matrix is 1, and the degree of the map is 1 (because it’s a
0 0
homeomorphism) so the image of A is Z.
Given the information that Im(A) = Z and Im(D) = 0 and by the first isomorphism theorem and the
exactness of the sequence, Z2 /Ker(A) = Im(A) = Ker(B), Z2 /Ker(B) = Z2 /Im(A) = Im(B) = Ker(C)
and so on. From here we can work out that H1 (M T2 ) = Z2 . All the equal signs stand for isomorphism.
Conclusion: H0 ≃ H3 ≃ Z, H1 ≃ H2 ≃ Z2 .

1.4.1 The fundamental group of the mapping torus π1 (M T2 )

By a proposition, if M T2 ≃ X/Γ with X simply-connected, and the group of Γ acts nicely on X, then for
[x] ∈ X/Γ, π1 (X/Γ, [x]) ≃ Γ.

Definition 10. An action of group Γ on topological space X is said to be nice if for any point x ∈ X, there
exists an open neighbourhood Ux such that for all γ ∈ Γ that’s not the identity, γ · Ux ∩ Ux = ∅.
Proving R3 is the universal cover of M T2 .
Parametrise T2 with coodinates (x(mod 1) , y(mod 1) ) for x, y ∈ R. Let the map f1 : T2 × R → T2 × R/Γ1
with Γ1 generated by ⟨g⟩ such that g := (x(mod 1) , y(mod 1) , z) → ((x + y)(mod 1) , y(mod 1) , z + 1).
1. T2 × R/Γ1 ≃ M T2 by construction.
2. The action is nice and can be proved by taking open Bδ3 such that δ ≪ y(mod 1) /2. Hence f1 is a
covering map.
Definition 11. A topological space is said to be locally path-connected if for all x ∈ X and a neighbourhood
U, there exists a path-connected neighbourhood V ⊆ U which contains x.
Since the map is a projection map, and Γ1 acts nicely on path-connected and locally path-connected
space T2 × R, the map is a covering map. (Another theorem that I’m not going to prove here.) [4]
Let map f2 : R3 → T2 × R ≃ R3 /Γ2 and an action Γ2 := ⟨h1 , h2 ⟩ with generators h1 : (x, y, z) → (x + 1, y, z)
and h2 : (x, y, z) → (x, y + 1, z). The action is clearly nice by taking an open neighbourhood B 3 of radius
< 1/4 for any point ∈ T2 × R.

Definition 12. A topological space is said semi-locally simply connected if for every x ∈ X, there exist an
open neighbourhood U such that the homomorphism π1 (U, x) → π1 (X, x) induced by the inclusion is trivial.
In another word, the generators of π1 (U, x) is trivial if written in terms of generators of π1 (X, x).

Theorem 5. Let X be a topological space, if X is locally path-connected and semi-locally simply connected.
And let f1 : Y → X and f2 : Z → Y be two covering maps, then the composition f2 ◦ f1 is a covering map.

6
 This theorem will not be proved here, see the proof in this paper [5].

Statement 1. All topological manifolds are locally path-connected and semi-locally simply connected.
Proof. • Semi-locally simply connected
Let X be a n-manifold, and let the chart (homeomorphism) k : U → B n the map induced by the chart
: π1 (U, x) → π1 (B n , k(x)) is an isomorphism. Since π1 (B n , k(x)) is trivial, π1 (U, x) is also trivial.
• Locally path-connected
There exists an open V ⊆ U that V is homeomorphic to B n with map p. Since B n is path-connected,
that is for any two points x, y in B n , there exists a continuous map γ : [0, 1] → B n . p−1 ◦ γ is also continuous.
Hence X is locally path-connected.

Combine theorem 5 and statement 1, f1 ◦ f2 is a covering map. Since R3 is simply-connected, by the

uniqueness of universal cover, R3 is a universal cover of M T2 .
Let Γ be an action generated by three generators: γ1 : (x, y, z) → (x + 1, y, z), γ2 : (x, y, z) → (x, y + 1, z)
and γ3 : (x, y, z) → (x + y, y, z + 1). Prove that R3 /Γ ≃ M T2 .
Define an operation ∗ on actions. Γ1 ∗ Γ2 := {γ : γ = ⟨γ1 , γ2 ⟩, f or all γ1 ∈ Γ1 and γ2 ∈ Γ2 }.
The group of action Γ is isomorphic to the group of the composition of actions Γ1 ∗ Γ2 . Here we don’t
want to work out the presentation of the group of each action, instead, we can check if for an arbitrary
x ∈ R3 , we get the same set of points by applying to the action of Γ or Γ1 ∗ Γ2 . It’s easy to check by
comparing generators.
Γ is a nice action on R3 .
√
2
3 min{ y 2 +1,1}
Proof. Let x ∈ R , take open Bϵ3 with ϵ ≪ 2 , then by construction, γ · Bϵ3 ∩ Bϵ3 = ∅ for γ ̸= e.
Work out the presentation of Γ.
• γ1 and γ2 commute.
• γ1 and γ3 commute.
• γ1−1 γ2−1 γ3 γ2 γ3−1 = 1.
Exercise: Check the 3 relations are the only relations.
Proof.
Conclusion: π1 (R3 /Γ) ≃ Γ ≃ ⟨γ1 , γ2 , γ3 | γ1 γ2 γ1−1 γ2−1 = γ1 γ3 γ1−1 γ3−1 = γ1−1 γ2−1 γ3 γ2 γ3−1 = 1⟩.
 
1 n
Example 8. The same question as above, but with matrix with n as an positive integer.
0 1
The first homology group is the same, but they have different fundamental groups, hence they are different
manifolds.

7
Chapter 2

Examples of polyhedral manifolds

The statement of Poincaré conjecture is initially made in 1904. Poincaré claimed in 1900 that the
homology group can tell 3-manifold apart from the 3-sphere, but then he found a counterexample to his
claim in 1904, which now is called the Poincaré homology sphere. It’s the first homology sphere discovered
in human history, which has the same homology groups as the 3-sphere, but non-trivial fundamental group.
This chapter introduces one of the construction of the Poincaré homology sphere and describes a method to
derive it’s fundamental group and extend the method further to other polyhedral manifolds. [6]
Statement 2. Statement of Poincaré conjecture:
Every compact, 3-manifold with trivial fundamental group is topologically equivalent to S 3 .

2.1 The construction of Poincaré homology via dodecahedron

A regular dodecahedron D is a solid in R3 with faces as the 12 identical regular pentagons. The 12 faces can
be divided into 6 equivalent classes by identifying 2 faces of opposite sides. For example, the front pentagon
with the pentagon in dotted line at the back.
Construct a line passing through the centre of the front and the back pentagon and treat that line as the
axis of rotation. Rotate the front pentagon clock-wise by π/5 and glue the corresponding edges (coloured in
red and blue), then the faces fall into place. (See figure 2.1a)
This identification results in 10 equivalent classes of edges and five equivalent classes of vertices. Each
class is represented by a different colour. (See figure 2.1b)
D/ ∼ with ∼ described as above is what we called a Poincaré homology sphere denoted as P .
Proposition 2. D/ ∼ is a 3-manifold.
Proof. • Find an open neighbourhood around any point in P such that it’s homeomorphic to
R3 . There are four cases based on where is the point.

(a) Edge and vertex classes (b) Rotation

Figure 2.1

8
 Case 0: When the point is inside the dodecahedron, this is a trivial case.
Case 1: When the point is on one of the faces.
Assume x ∈ DF , F means the point is on the face. The open neighbourhood of x is homeomorphic
to A = [0, ∞] × R2 . Take x as the point which is equivalent to x in P and take the open neighbourhood
of x which is also homeomorphic to B = [0, ∞] × R2 . Construct a homeomorphic map f : R2 → R2 , and
A ⊔ B/0 × x ∼ 0 × f (x) for x ∈ R2 is homeomorphic to R3 .
Case 2: When the point is on an edge.

Figure 2.2: Case 2

As shown in left figure 2.2, three edges of the pentagon are equivalent to each other, and for a point
x ∈ DE , the neighbourhood of x is homeomorphic to [0, ∞) × [0, ∞) × R and so is all the points that
are equivalent to x in P. See figure 2.2 second picture for the cross sections, by identifying (construct
homeomorphisms) 0 × [0, ∞) × R for each neighbourhood, it’s clear that the resulting space is homeomorphic
to R3 .
Case 3: When the point is one of the vertices.

Figure 2.3: Cutting a dodecahedron

The four tetrahedrons are cut from the dodecahedron in figure 2.3. From left to right in figure 2.4,
correspond to figure 2.3 first picture from top left cut, top right cut, bottom right, bottom left.

Figure 2.4: Four courners of dodecahedron

9
 Similar to case 2, it’s possible to consider the gluings of faces of tetrahedrons shown in figure 4 indicated
by the colour of edges, and show that it’s B 3 ≃ R3 .
Another clever way to prove it glues to B 3 , is to prove after gluing, the shaded green region becomes a
2-sphere. We know that the shaded region becomes a surface. To prove it’s a 2-sphere, we can use the fact
that the sphere has Euler-characteristic 2, and the triangles in green form a triangulation of the surface.
There are 6 edges (1-6), 4 vertices (A-D), and 4 faces. 4-6+4=2.
Exercise: Prove that with the boundary as the 2-sphere, it’s a 3-ball.
Proof. In case 3, taking the open neighbourhood of an open ball of radius ϵ for all ϵ, the boundary of the
neighbourhood is always homeomorphic to S 2 . By taking smaller and smaller ϵ, eventually the boundary
is a point. Hence by blow up the point, we can deduce the original space we bounded with a 2-sphere is a
3-ball.
• Hausdorff
• Second-countable

Counter example The proof ‘vertices locally R3 ‘ fails’.

Icosahedron see figure 2.5
An alternative method of proving the quotient space of the polyhedron is a
manifold is to count the equivalent classes of cells on the polyhedron directly.
If the Euler characteristic is 0, it is a manifold, otherwise, it’s not. To see a
complete proof, see Thurston’s book: 3-dimensional Geometry and Topology.
[7]
Exercise left for the reader: Check for the dodecahedron, the Euler
characteristic is 0.
In the case of the icosahedron. There’s one equivalent class of vertex, 10
Figure 2.5 classes of faces, 6 classes of edges. The Euler characteristic is 3 cells - 2 cells +
1 cells + 0 cells = 1 - 10 + 6 - 1 = - 4.

2.2 The fundamental group of the Poincaré homology sphere

By a theorem suggested by Poincaré, the fundamental group of the quotient space of polyhedron is ⟨the
generators of the faces | the relations of the edges⟩.

Figure 2.6: 3-torus

First, in the case of a 3-torus T3 :

The generators of faces
As shown in figure 2.6, there are three classes of faces, {A, A′ }, {B, B ′ } and {C, C ′ }. The purple, orange
and pink lines connecting the centre of each pair of faces intersect at O. The claim is Π1 (T 3 , O) has only
generators purple (Ã) A → A′ , orange (C̃) C → C ′ or (c → a) and pink (B̃) B → B ′ or (b → d), notice
that the alphabet is specifically ordered. But notice our starting point is ’O’, so the real presentation of
the orange generator is O to c, then a to O, considering a and c are the same point. Similarly for other
generators.

10
 The relations between the generators
Considering the trivial loops around each edge, they can be written as a linear combination of the three
generators.
Take the loop around the green generator for example. (follow the black loop and arrow in figure 2.6)
starting from face B, going one quarter and land on face C, which then reappears on face C ′ and going one
quarter again to face B. Reappearing on face B ′ , going faces C ′ , appears on face C, and ended up on face
B’, which is the same point as the starting point.
This complete cycle can be represented by B ∩ C −→ C ′ ∩ B −→ B ′ ∩ C ′ −→ C ∩ B ′ , which tells
you the location of the edges it’s going around. Question is can we write the cycle in forms of the three
generators. And yes! Deform the loop and you get bO → Oc → aO → Ob → dO → Oa → cO → Od, moving
bO to the end, and it’s B̃ C̃ B̃ −1 C̃ −1 , with left operation. The linear combination can be read off with the
representation of edges.
Conclusion: The fundamental group of T3 is ⟨Ã, B̃, C̃ | B̃ C̃ B̃ −1 C̃ −1 = B̃ ÃB̃ −1 Ã−1 = ÃC̃ Ã−1 C̃ −1 = 1⟩.

Figure 2.7: Fundamental group of dodecahedron

Apply the same method as in 3-torus to dodecahedron:

All the front faces are named from A to F, and A’ to F’ are the identified pair. The generators are Ã, B̃, C̃,
D̃, Ẽ, F̃ . The ten classes of edges give ten relations. The relation from the red edge: A∩B → B ′ ∩E → E ′ ∩A′ ,
and in terms of generators: Ã−1 Ẽ B̃ = 1.

Red A ∩ B → B ′ ∩ E → E ′ ∩ A′ Ã−1 Ẽ B̃ = 1
Dark green A ∩ C → C ′ ∩ F → F ′ ∩ A′ Ã−1 F̃ C̃ = 1
Blue A ∩ D → D ′ ∩ B → B ′ ∩ A′ Ã−1 B̃ D̃ = 1
Orange A ∩ E → E ′ ∩ C → C ′ ∩ A′ Ã−1 C̃ Ẽ = 1
Brown A ∩ F → F ′ ∩ D → D ′ ∩ A′ Ã−1 D̃F̃ = 1
Purple F ∩ D′ → D ∩ E → E ′ ∩ F ′ F̃ −1 Ẽ D̃−1 = 1
Yellow E ∩ C ′ → C ∩ D → D′ ∩ E ′ Ẽ −1 D̃C̃ −1 = 1
Light green D ∩ B ′ → B ∩ C → C ′ ∩ D′ D̃−1 C̃ B̃ −1 = 1
Blue green B ∩ E′ → E ∩ F → F ′ ∩ B′ B̃ −1 F̃ Ẽ −1 = 1
Black B ∩ F → F ′ ∩ C → C ′ ∩ B′ B̃ −1 C̃ F̃ = 1

Conclusion: The fundamental group of the homology sphere is the six generators with the ten relations in
the table.

2.2.1 Alternative method to compute the fundamental group

The second approach is to apply Seifert–Van Kampen theorem.
Theorem 6. Seifert-Van Kampen

11
 Let X be a path-connected space, and let A, B be two open and path-connected subsets such that X = A∪B
and A∩B is also path-connected. Let x0 be a base point in A∩B, consider two injection maps iA : A∩B → A
and iB : A∩B → B. The two injections introduce two inclusion maps: i∗A : A∩B ,→ A and i∗B : A∩B ,→ B.
The fundamental group of X: π1 (X, x0 ) ≃ π1 (A, x0 ) ∗π1 (A∩B,x0 ) π1 (B, x0 ). ∗ is the amalgamated product.
The proof of Van Kampen theorem is left as an exercise for readers.

T3 :
Presentation of π1 (A, x0 ), π1 (B, x0 ) and π1 (A ∩ B, x0 ).
split T3 into two open path-connected subsets A and B, such
that A∩B is a path-connected space. After deformation retraction,
A and B are shown in figure 2.8. A is homeomorphic to a thick-
ened 1-skeleton of the cube with identifications on the 1-skeletons
inherited from T3 . B is homeomorphic to three solid tubes go-
ing through the centre with gluing indicated by the colour. The
fundamental groups of both A and B is isomorphic to the free group of generator 3. Figure 2.8
Let x0 be a base point in A ∩ B. π1 (A, x0 ) = ⟨a, b, c⟩ and π1 (B, x0 ) = ⟨d, e, f ⟩.
It needs some imagination to see A ∩ B is a genus 3 surface. (See figure 2.9.)
Hence π1 (A ∩ B, x0 ) = ⟨X1 , X2 , X3 , Y1 , Y2 , Y3 | [X1 , Y1 ] · [X2 , Y2 ] · [X3 , Y3 ] = 1⟩. (Note
that [α, β] = αβα−1 β −1 is called a commutator of α and β.)
Presentation of inclusion maps iA∗ and iB∗
i∗A (A ∩ B) is exactly the 1-skeleton of A. i∗B (A ∩ B) is the surface of B, shown
in figure 2.10.
iA∗ (X1 ) = a, iA∗ (X2 ) = b, iA∗ (X3 ) = c, iB∗ (X1 ) = d, iB∗ (X2 ) = e, iB∗ (X3 ) = f
iA∗ (Y1 ) = [a, b], iA∗ (Y2 ) = [b, c], iA∗ (Y3 ) = [a, c], iB∗ (Y1,2,3 ) = trivial
Figure 2.9 π1 (T3 ) ≃ ⟨a, b, c, d, e, f | ad−1 = be−1 = cf −1 = [a, b] = [a, c] = [b, c] = 1⟩
≃ ⟨a, b, c | [a, b] = [a, c] = [b, c] = 1⟩
Poincaré homology sphere:
Apply the same splitting of the space as with T3 .
π1 (A, x0 ) is a group with 6 generators:
• Generators represented by 3 loops around three adjacent pen-
tagons looping back to the same point.
• Generators represented by 3 loops starting from the base point,
end at one of the other points in the same vertex equivalent class.
π1 (B, x0 ) = ⟨X1 , X2 , X3 , X4 , X5 , X6 ⟩
π1 (A ∩ B, x0 ) = ⟨Y1 , · · ·, Y6 , Z1 , · · ·, Z6 | [Y1 , Z1 ] · · · [Y6 , Z6 ] = 1⟩. Figure 2.10: Inclusion
Presentation of inclusion maps
i∗A (Zi ) = loops around pentagons.
i∗A (Yi ) = loops from one face to the opposite face. (The started point and the end point are in the same
equivalent class.)
i∗B (Zi ) = 0, i∗B (Yi ) = Xi (See figure 2.11a)
By naming the ten different classes of edges and carefully place their orientation as shown in the figure
2.11b, it’s fundamental group is ⟨hceab, b−1 i−1 c−1 g −1 d, d−1 e−1 ijh−1 , f −1 ib, f −1 ed, f −1 c−1 h−1 , X1 , · · ·, X6 |
hceab = b−1 i−1 c−1 g −1 d = d−1 e−1 ijh−1 = a−1 gh−1 f −1 i = f b−1 jg −1 e−1 = j −1 a−1 df −1 c−1 = 1, X1 =
f −1 ed, X2 = f −1 c−1 h−1 , X6 = f −1 ib, X3 = b−1 a−1 d, X4 = d−1 gh−1 , X5 = hj −1 b⟩.
This devises a general method to
• Find a Heegaard splitting of all the polyhrdron manifolds with faces pairwise identified.
• Find a representation of the fundamental group from the Heegaard splitting.

2.2.2 Homology groups of the homology sphere

By adding the commutativity to the six relations, it’s easy to see that all generators are trivial, hence
the first homology group is trivial.

12
 (a) A ∩ B of dodecahedron (b) Indexed dodecahedron

Figure 2.11

The second homology group can be derived from the first homology group, to do that, we need the
concept of cohomology.
Definition 13. Cohomology:
Theorem 7. Poincaré duality

2.3 Other quotient spaces of platonic solids

2.3.1 Seifert–Weber space
Construction
There are three types of quotient spaces of dodecahedron by identifying opposite faces with a rotation,
and all the spaces are proven to be 3-manifolds.
1
The Poincaré homology sphere is an identification with a 10 turn. The
3
Seifert-Weber space is an identification with a 10 turn.
As shown in figure 2.12, there are 6 equivalent classes of edges, 1
equivalent class of vertices and 6 classes of faces. It’s Euler-characteristic
is 1 − 6 + 6 − 1 = 0, which confirms that the quotient space is a manifold.

Fundamental group
Similarly as before, the fundamental group is the six generators from the
six classes of faces with 6 relations coming from the 6 classes of edges.
⟨A, B, C, D, E, F | BC −1 E −1 F A−1 = A−1 CD−1 F −1 B = B −1 D−1 E Figure 2.12: Seifert-Weber space
−1 −1 −1 −1 −1 −1 −1 −1 −1 −1 −1 −1
A F = A EF C D = A DE B C = D B E C F = 1⟩

Homology groups
The first homology group is Z35 . Don’t know how to prove that yet.

Remark 1. The third possible quotient space from dodecahedron is by

5
having a 10 turn, which is the projective space RP3 .
A table of all quotient spaces of polyhedron manifolds listing the amount of rotation before gluing opposite
(parallel) faces, number of equivalent classes of cells, Euler-characteristic calculated from number of cells
and whether or not it’s a manifold.

13
Platonic solids Rotation 3-cells 2-cells 1-cells 0-cells Euler-Characteristic Manifold
Tetrahedron N/A N/A N/A N/A N/A N/A N/A
Cube 0 1 3 3 1 0 Yes
1/4 1 3 4 2 0 Yes
Octahedron 1/6 1 4 4 1 0 Yes
3/6 Projective space
Dodecahedron 1/10 1 6 10 5 0 Yes
3/10 1 6 6 1 0 Yes
5/10 Projective space
Icosahedron 1/6 1 10 6 1 -4 No
3/6 Projective space

14
Chapter 3

Heegaard splitting

3.1 3 dimensional Lens space

Lens spaces are the only 3-manifolds with finite, cyclic, non-trivial fundamental group. The reverse
direction of this statement (that a 3-manifold with such a fundamental group must be a lens space) is closely
related to the Poincaré conjecture.
In the 3-manifold case, a lens space can be visualized as the result of gluing two solid tori together by a
homeomorphism of their boundaries
Definition 14. A 3-d lens space L(p; q) is a quotient of S 3 by Z/pZ action, with p, q coprime.
S 3 := {(x1 , x2 , x3 , x4 ) | |x1 |2 + |x2 |2 + |x3 |2 + |x4 |2 = 1}, which can be represented by a point (z1 , z2 ) in
C . The action is generated by homeomorphism ϕ : (z1 , z2 ) 7→ (e2πi/p · z1 , e2πiq/p · z2 ).
2

Proposition 3. The action is nice.

Proof. gcd(p, q) = 1 guarantees ϕ(z1 , z2 ) ̸= (z1 , z2 ) for ϕ ̸= I. Take Br such that r ≪ min1≤q≤p {dEucli |(e2πi/p ·
z1 , e2πiq/p · z2 ) − (z1 , z2 )|}. Hence ϕ(Br ) are all disjoint for ϕ ̸= I.
Remark 2. The Z/pZ action is free when p > 1 because (z1 , z2 ) = (e2πi/p · z1 , e2πiq/p · z2 ) if and only if
p = 1.

3.2 Definition and examples

Definition 15. The 3-dimensional handle is B 2 × [0, 1].
Definition 16. The 3 dimensional handlebodies Hg can be built by attaching 3-dimensional handles to
the 3-ball B 3 with homeomorphism Φi . Let disjoint sets U0 , U1 ∈ ∂B 3 homeomorphic to B 2 such that
Φi := Ui → B 2 × {i}. The number of handles attached to the 3-ball is the genus of the handlebody.
Example 9. The H1 are homeomorphic to the ’solid’ 2-torus B 2 × S 1 . By attaching more handles, it has
the boundary of genus g surface.
Definition 17. If a manifold can be written as Hg1 Hg2 / ∼, for ∼. The the manifold is said to have a
`
Heegaard splitting.
Note that Heegaard Splitting for a manifold is not unique. See examples below.
Example 10. Find Heegaard splitting for S 3 and T3 .
S3 ≃ H
Theorem 8. Any 3-manifold can be triangulated.
In 1951, Moise first proved that all 3 topological manifolds are triangulable. After that, Bing(1957) and
Hamilton(1974) came up with alternative proofs.
This theorem will not be proved here, because I don’t understand the proof at all... But, we can apply
the theorem and prove, given a triangulation of the 3-manifold, there exists a Heegaard splitting.

15
Bibliography

[1] Scharlemann, M. (2007). Heegaard Splitting of 3-manifolds. Santa Barbara. Available at:
https://web.math.ucsb.edu/ mgscharl/papers/MorningsideNotes.pdf

[2] Tu,L.W.(2010) An introduction to manifolds. Second edition. New York: Springer Science+Business
Media, pp 76-77
[3] Ghazouani, S. Algebraic Topology, Imperial College London, Lecture given: Jan-Mar 2022
[4] Maxim, L.G. (2017) ’Week 8: Regular Covers as Group Actions’. Math 751: Introductory Topology I.
Available at: https://people.math.wisc.edu/ maxim/751f14w8.pdf (Accessed: 29 September 2022).
[5] Ethan, J.(2008) Fundamental groups and covering spaces. Available at:
http://www.math.uchicago.edu/ may/VIGRE/VIGRE2008/REUPapers/Jerzak.pdf (Accessed: 29
September 2022).

[6] Wikipedia (2022) Poincaré conjecture. Available at: https://en.wikipedia.org/wiki/Poincar

[7] Thurston, W.P.(1997) Three-Dimensional Geometry. USA: Princeton Academic Press

16