Introduction to Algebraic Topology

Mitschri der Vorlesung von

Dr. M. Michalogiorgaki

Tobias Berner
Universität Zürich
Frühjahrssemester 2009
Contents

1 Topology 4
1.1 Topological spaces and continuous functions . . . . . . . . . . . . . . . . . 4
1.1.1 Topological spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.2 Continuous functions . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 Conectedness & Compactness . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.1 Connected spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.2 Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 Algebraic topology 16
2.1 Fundamental group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.1.1 Path homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 e fundamental group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3 Covering spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Liing properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5 e fundamental group of the circle and applications . . . . . . . . . . . . 27
2.5.1 e Fundamental eorem of Algebra . . . . . . . . . . . . . . . . 28
2.5.2 Deformation retracts and homotopy type . . . . . . . . . . . . . . 29
2.6 Seifert von Kampen eorem . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.6.1 Direct sums of abelian groups . . . . . . . . . . . . . . . . . . . . . 35
2.7 Free abelian groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.8 Free products of groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.8.1 Free groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.8.2 e Seifert-van Kampen theorem . . . . . . . . . . . . . . . . . . . 43
2.9 CW complexes (cell complexes) . . . . . . . . . . . . . . . . . . . . . . . . 47
2.10 Surfaces (two-dimensional manifolds) . . . . . . . . . . . . . . . . . . . . 49
2.10.1 Fundamental group of surfaces . . . . . . . . . . . . . . . . . . . . 49
2.10.2 Homology of surfaces . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.10.3 Classi cation of surfaces . . . . . . . . . . . . . . . . . . . . . . . . 51
2.11 Knot theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.12 Classi cation of covering spaces . . . . . . . . . . . . . . . . . . . . . . . . 58

Index 66

2
 CONTENTS

Literature
• Muncres: Topology 2nd edition

• Jänich: Topologie

• Massey: Algebraic Topology: An Introduction

• Stöcker, Zieschang: Algebraische Topologie

• Lickorish: An introduction to knot theory

Introduction

In calculus, you have studied Rn , n ∈ N, as well as functions f ∶ Rk → Rℓ , k, ℓ ∈ N. You

have studied notions such as neighbourhood of a point x ∈ Rn , as well as convergence and
continuity of a function f at a point x ∈ Rk . For this study, you have used the Euclidean
metric.
For instance if f ∶ R → R, we say that f is continuous at x ∈ R if ∀ε > 0 ∃δ > 0 such that
if ∣x − x0 ∣ < δ then ∣f (x) − f (x0 )∣ < ε.
We used the metric d ∶ R × R → R, with (x, y) ↦ ∣x − √ y∣. In general the Euclidean metric
d ∶ Rn × Rn → R with (x1 , . . . , xn ), (y1 , . . . , yn ) ↦ ∑ni=1 (xi − yi )2 .

In topology, we study notions such as neighbourhood of a point x ∈ X , convergence,

continuity for a general set X . For this study a metric is not necessary. What we will use
is open sets in X .

In algebraic topology, we use abstract algebra to study topological properties.

3
1 Topology

1.1. Topological spaces and continuous functions

1.1.1. Topological spaces

Consider a set X and P(X) ∶= {U ∣ U ⊆ X}.

De nition 1.1 A topology on X is a collection T ⊂ P(X) such that

1. ∅, X ∈ T .

2. If Ui ∈ T ∀i ∈ I , then ⋃i∈I Ui ∈ T .

3. If Ui ∈ T , i ∈ {1, . . . , n}, then ⋂i∈{1,...,n} Ui ∈ T .

(X, T ) is called a topological space (sometimes we will only write X ).

U ⊆ X is called open, if U ∈ T .
U ⊆ X is called closed, if X/U ∈ T .

Example 1.2

1. X some set.

• Td = {U ∣ U ⊆ X} discrete topology.
• Tt = {∅, X} trivial topology.

2. X = {x1 , x2 , x3 }. en the three collections

• {∅, X}
• {∅, {x1 }, {x1 , x2 }, X}
• {∅, {x2 }, {x1 , x2 }, {x2 , x3 }, X}

are topologies on X .

3. X is a set. en the collections

• Tf = {U ⊆ X ∣ X/U is nite or X/U = X} nite topology.

• Tc = {U ⊆ X ∣ X/U is countable or X/U = X} countable topology.

are topologies on X . [Homework]

4
 1.1 Topological spaces and continuous functions

De nition 1.3 Suppose that T and T ′ are topologies on X . If T ′ ⊇ T , then T ′ is called

ner than T . If T ′ ⊋ T , then T ′ is called strictly ner than T . If T ′ ⊇ T or T ⊇ T ′ ,
then T and T ′ are comparable.

De nition 1.4 Let (X, T ) be a topological space and A ⊆ X .

e interior of A is
intA ∶= ⋃ U .
U ∈T ,U ⊆A

e closure of A is
A ∶= ⋂ U.
X/U ∈T , A⊆U

Basis for a topology

De nition 1.5 A basis B for a topological space (X, T ) is a collection B of open sub-
setes X (i.e. B ⊆ T ), such that ∀U ∈ T ∃{Bi }i∈I , Bi ∈ B , ∀i ∈ I , with U = ⋃i∈I Bi .

Remark 1.6 A basis is not unique.

Properties: B basis for (X, T ) then B has the following properties
1. ∀x ∈ X ∃B ∈ B with x ∈ B .
P X ∈ T and B is a basis Ô⇒ X = ⋃i∈I Bi , Bi ∈ B , i ∈ I . So x ∈ Bj for
some j ∈ I .
2. If x ∈ B1 ∩ B2 , B1 , B2 ∈ B then ∃B3 ∈ B such that x ∈ B3 ⊆ B1 ∩ B2 .
P B1 , B2 ∈ B ⊆ T Ô⇒ B1 , B2 ∈ T Ô⇒ B1 ∩ B2 ∈ T Ô⇒ B1 ∩ B2 =
⋃i∈I Bi , …
Conversely: If a collection B of subsets of X satis es properties 1. and 2. then there is a
unique topology T for which B is a basis.
T is called the topology generated by B and it consists of all unions of elements of B .
P We prove that T is a topology
1. X ∈ T ?
Property 1 Ô⇒ if x ∈ X then ∃Bx ∈ B with x ∈ Bx ⊆ X . erefore X = ⋃x∈X Bx ,
i.e. X ∈ T .
∅∈T?
∅ is the empty union of elements in B, so ∅ ∈ T .
2. If Ui ∈ T ∀i ∈ I , does ⋃i∈I Ui ∈ T ?
Ui ∈ T Ô⇒ Ui = ⋃j<inJ Bij . So ⋃i∈I Ui = ⋃i∈I ⋃j∈J Bij ∈ T .
3. If Ui ∈ T for i ∈ {1, . . . , n}, is ⋂i∈{1,...,n} Ui ∈ T ?
We will show that if U1 , U2 ∈ T then U1 ∩ U2 ∈ T .
U1 = ⋃λ∈Λ Bλ , U2 = Uk∈K Bk . Consider x ∈ U1 ∩ U2 . en x ∈ Bλx and x ∈ Bkx ,
for some λx ∈ Λ and kx ∈ K . us x ∈ Bλx ∩ Bkx Ô⇒ [by property 2.] ∃Bx ∈ B
with x ∈ Bx ⊆ Bλx ∩ Bkx ⊆ U1 ∩ U2 .
erefore U1 ∩ U2 = ⋃x∈U1 ∩U2 Bx ∈ T .

5
Topology

Example 1.10

1. X = R, B = {(a, b) ∣ a, b ∈ R, a < b}. B satis es properites 1. and 2. therefore it

generates a topology on R. is is the standard topology on R.

2. X = R, B = {(a, b) ∣ a, b ∈ Q, a < b}
B satis es properties 1. and 2.
Claim [Homework]: e topology T generated by B is in fact the standard topol-
ogy on R.

Lemma 1.11 Let B, B ′ be bases for the topologies T and T ′ on X . en the following
statements are equivalent:

1. T ′ ⊇ T .

2. ∀x ∈ X , ∀B ∈ B with x ∈ B there is B ′ ∈ B ′ such that x ∈ B ′ ⊆ B .

P

1.⇒2.
Consider x ∈ X and B ∈ B with x ∈ B . B ∈ T ⊆ T ′ Ô⇒ B ∈ T ′ Ô⇒ B =
⋃λ∈Λ Bλ′ , Bλ′ ∈ B ′ .
So x ∈ B , therefore x ∈ Bλ′ , for some λx ∈ Λ. i.e. we have Bλ′ x ∈ B ′ , such that
x ∈ Bλ′ x ⊆ B .

2.⇒1.
Consider U ∈ T and x ∈ U . en x ∈ U = ⋃λ∈Λ Bλ , Bλ ∈ B . at is x ∈ Bλx , for
some λx ∈ Λ. 2. implies that there exists Bλ′ x ∈ B ′ such that x ∈ Bλ′ x ⊆ Bλx .
en U ⊆ ⋃x∈U Bλ′ x ⊆ U Ô⇒ U = ⋃x∈U Bλ′ x
Ô⇒ U ∈ T ′ . erefore T ⊆ T ′ .

Product topology
Let (X, Tx ), (Y, TY ) be two topological spaces. e product topology is the topology
with basis the collection B = {U × V ∣ U ∈ TX , V ∈ TY }.

Metric topology
X is a set.

De nition 1.13 A metric on this set is a function d ∶ X × X → R with

1. d(x, y) ≥ 0, and d(x, y) = 0 iff x = y .

2. d(x, z) ≤ d(x, y) + d(y, z) ∀x, y, z ∈ X .

3. d(x, y) = d(y, x), ∀x, y ∈ X .

We call the set Bd (x, ε) = {y ∈ X ∣ d(y, x) < ε} the ε-ball centered at x.

{Bd (x, ε)}x∈X,ε>0 is a basis for a topology on X , the metric topology.

6
 1.1 Topological spaces and continuous functions

Example 1.14 R with the standard topology (Tstand ).

Consider R × R = R2 .
Claim: e product topology on R2 and the metric topology are the same.
For this, one has to show

• Tpr ⊆ Tm
By lemma 1.11 one has to show ∀x ∈ R2 , ∀Bpr ∈ Bpr , x ∈ Bpr , ∃Bm ∈ Bm such
that x ∈ Bm ⊆ Bpr .

• Tm ⊆ Tpr
…

Subspace topology

De nition 1.15 (X, T ) is a topological space and Y ⊆ X . e collection TY ∶= {U ∩

Y ∣ U ∈ T } is a topology on Y , called the subspace topology.

Example 1.16 Y = [0, 1] ∪ {2} ⊆ X = R with the standard topology.

en the sets

• (a, b), with a, b ∈ [0, 1],

• [0, b), with b ∈ [0, 1],

• (a, 1], with a ∈ [0, 1],

• {2},

• [0, 1]

are open sets in the subspace Y .

1.1.2. Continuous functions

De nition 1.17 Let (X, TX ), (Y, TY ) be topological spaces and f ∶ X → Y . f is

called continuous if f −1 (V ) ∈ TX ∀V ∈ TY .

Claim f ∶ R → R (with standard topology). e ε–δ de nition of continuity is equivalent

to the de nition above.
P

“⇐” Suppose that f ∶ R → R is continuous with the de nition above.

Consider x0 ∈ R and ε > 0. en V = (f (x0 ) − ε, f (x0 ) + ε) ∈ Tstand , so
f −1 (V ) ∈ Tstand by our de nition.
Now x0 ∈ f −1 (V ), so there exists (a, b) ∈ Bstand with x0 ∈ (a, b) ⊆ f −1 (V ).
Take δ = min{x0 − a, b − x0 }. Clearly δ > 0. en ∣x − x0 ∣ < δ Ô⇒ x ∈ (a, b) ⊆
f −1 (V ) Ô⇒ f (x) ∈ V
Ô⇒ f (x) ∈ (f (x) − ε, f (x) + ε) Ô⇒ ∣f (x) − f (x0 )∣ < ε.

7
Topology

eorem 1.19 Let X, Y be topological spaces and f ∶ X → Y . e following are equiva-

lent

1. f is continuous.

2. for every closed subset of Y the inverse image of it is a closed subset of X .

P
1.⇒2.
Let B be closed in Y , then X/f −1 (B) = f −1 (Y /B) which is open in X , i.e.
f −1 (B) is closed in X .
2.⇒1.
…
Lemma 1.21 (the pasting lemma)
Let X, Y be topological spaces and A, B closed subsets of X with X = A ∪ B . Let
f ∶ A → Y , g ∶ B → Y be continuous with f (x) = g(x) ∀x ∈ A ∩ B . en h ∶ X → Y
with
f (x) x ∈ A
h(x) = {
g(x) x ∈ B
is continuous.
P Let C be a closed subset of Y . en h−1 (C) = f −1 (C) ∪ g −1 (C) where f −1 (C)
is closed in A and g −1 (C) is closed in B .

(a) f −1 (C) is closed in A Ô⇒ f −1 (C) = A ∩ G, where G closed in X .

(b) g −1 (C) is closed in B Ô⇒ g −1 (C) = B ∩ H , where H closed in X .

(a) Ô⇒ f −1 (C) is closed in X ,

(b) Ô⇒ g −1 (C) is closed in X .
Ô⇒ h−1 (C) = f −1 (C) ∪ g −1 (C) is closed in X .

De nition 1.23 Let X, Y be topological spaces. f ∶ X → Y is called a homeomorphism

if f is bijective, and f, f −1 are continuous.

Bijective correspondence not only between X and Y but also between the collection of
open sets in X and the collection of open sets Y .
us any property of X that is expressed in terms of its open subsets yields via f the
corresponding properties for Y . Such a property is called a topological property.

Quotient topology
Example 1.24 Torus

De nition 1.25 Let X, Y be topological spaces, p ∶ X → Y a surjective map. e map

p is called a quotient map if a subset U of Y is open in Y , if and only if p−1 (U ) is open
in X .

8
 1.2 Conectedness & Compactness

De nition 1.26 Let X be a topological space, Y be some set and p ∶ X → Y a surjective

map. e quotient topology on Y induced by p is de ned as follows:
A subset U ⊆ Y is open if and only if p−1 (U ) ⊆ X is open.

e fact that this is a topology follows from p−1 (∅) = ∅, p−1 (Y ) = X , p−1 (⋃a∈J Ua ) =
⋃a∈J p−1 (Ua ), p−1 (⋂ni=1 Ui ) = ⋂ni=1 p−1 (Ui ).

Remark 1.27 e quotient topology on Y is the nest topology that makes p continuous.
Example 1.28

1. X = {(x, y) ∈ R2 ∣ 0 ≤ x ≤ 2π, 0 ≤ y ≤ 1} = [0, 2π] × [0, 1] ⊆ R2 . Y =

{(x, y, z) ∈ R3 ∣ x2 + y 2 = 1, 0 ≤ z ≤ 1}.
f ∶ X → Y , with (x, y) ↦ (cos x, sin x, y).
f is surjective. Using f we can de ne the quotient topology in Y .

2. p ∶ R → {x1 , x2 , c3 }
⎧
⎪ x x>0
⎪
⎪ 1
p(x) = ⎨ x2 x<0
⎪
⎪
⎪
⎩ x3 x=0
e quotient topology on {x1 , x2 , x3 } induced by p is {∅, {x1 }, {x2 }, {x1 , x2 }, {x1 , x2 , x3 }}

1.2. Conectedness & Compactness

In calculus you have studies functions f ∶ [a, b] → R and you have proved three theorems
for continuous functions f ∶ [a, b] → R.

1. Intermediate Value eorem (IVT),

2. Maximus Value eorem (MVT),

3. Uniform Continuity eorem (UCT).

ese theorems rely on the continuity of f and some topological properties of [a, b] ⊂
R. In particular IVT relies on the connectedness of [a, b]. MVT and VCT rely on the
compactness of [a, b].

1.2.1. Connected spaces

De nition 1.29 Consider (X, T ) topological space. A separation of X is a pair U, V

where U, V ∈ T , U, V ≠ ∅, U ∩ V = ∅, and X = U ∪ V . X is called connected if there
is no separation of X .

Remark 1.30 Connectedness is a topological property.

Example 1.31

1. R, Tstand ,
Q = ((−∞, a) ∩ Q) ∪ ((a, ∞) ∩ Q), a ∈ R ∖ Q.

2. R, [a, b], [a, b), (a, b], (a, b) are connected.

9
Topology

Claim: X is connected iff the only subsets of X that are both open and closed are ∅ and
X.
P

“⇒” Suppose, that A ⊆ X , A ≠ ∅, A ≠ X , and A ∈ TX , X ∖ A ∈ TX . en X =

A ∪ (X ∖ A) Ô⇒ A, X ∖ A is separation of X , contradiction.

“⇐” If U, V is a separation of X . en U ∈ TX , X ∖ U = V ∈ Tx , U, X ∖ U ≠ ∅,

U ∩ (X ∖ U ) = ∅. V ∈ TX , X ∖ U ∈ TX , U ≠ ∅, U ≠ X , contradiction.
erefore X is connected.

Lemma 1.33 If C, D are a separation of X and Y is a connected subspace of X , then

Y ⊆ C or Y ⊆ D.
P C ∩ Y , D ∩ Y are open subsets of Y . In addition (C ∩ Y ) ∩ (D ∩ Y ) = ∅. If
C ∩ Y and D ∩ Y were both nonempty, then they would form a separation of Y . But Y
is connected. So C ∩ D = ∅, or D ∩ Y = ∅.
Ô⇒ Y ⊆ C or Y ⊆ D.

eorem 1.35 Let {Ai }i∈I be connected subspaces of X and p ∈ ∩i∈I Ai . en ∪i∈I Ai is
connected subspace of X .

P Assume that Ui∈I Ai is not connected, and ∪i∈I Ai = C ∪ D, C, D a separation

of ∪i∈I Ai . en p ∈ C ∪ D and WLOG we can assume that p ∈ C . Ai is a connected
subspace.
Ô⇒ [Lemma] Ai ⊆ C or Ai ⊆ D for any i ∈ I . p ∈ ∩i∈I Ai .
Ô⇒ p ∈ Ai ∀i ∈ I .
Ô⇒ Ai ⊆ C ∀i ∈ I .
Ô⇒ ∪i∈I Ai ⊆ C .
Ô⇒ D = ∅. Contradiction.

eorem 1.37 Let A be a connected subspace of X and A ⊆ B ⊆ A. en B is also

connected.

P Assume that C, D is a separation of B . By Lemma A ⊆ C or A ⊆ D. WLOG

A ⊆ C.
A ⊆ C Ô⇒ A ⊆ C .
D ⊆ B ⊆ A ⊆ C Ô⇒ D ⊆ C .
Claim: D ∩ C = ∅.
Proof: e closure of C in B is C ∩ B , where C is the closure of C in X . C = C ∩ B =
C ∩ (C ∪ D) = (C ∩ C) ∪ (C ∩ D) = C ∪ (C ∩ D).
Ô⇒ C ∩ D = ∅.

We have that D ⊆ C and we showed that D ∩ C = ∅. erefore D = ∅, contradiction.

ere is no separation of B , i.e. B is connected.

eorem 1.39 Let f ∶ X → Y be a continuous function between topological spaces X and

Y . If X is connected, then f (X) is connected.

10
 1.2 Conectedness & Compactness

P Consider f̃ ∶ X → f (X), x ↦ f (x). en f̃ is continuous. Indeed, if U open

in f (X), then U = f (X) ∩ V for some V open in Y . f̃1 (U ) = f̃−1 (f (X) ∩ V ) =
f̃−1 (f (X)) ∩ f̃−1 (V ) = X ∩ f −1 (V ) = f −1 (V ) open in X .
Suppose that f (X) = C ∪ D, C, D separation of f (X). en f̃−1 (C), f̃−1 (D) are open
subsets of X , disjoint and nonempty and X = f̃−1 (C) ∪ f̃−1 (D). Contradcition (X
connected).

eorem 1.41 A nite (cartesian) product of connected spaces (in the product topology) is
connected.

P We start by proving that if X and Y are connected, then X × Y is connected.

Consider a × b in X × Y , X × b. X × b is connected. Similarly x × Y is connected for
any x ∈ X . As a result, Tx = (X × b) ∪ (x × Y ) is also connected (by previous theorem,
since x × b ∈ (X × b) ∩ (x × Y )).
erefore, ∪x∈X Tx is connected, as the union of connected subspaces, Tx with common
point a × b. Now X × Y = ∪x∈X Tx , therefore X × Y is connected.
By induction, the general proof follows.

eorem 1.43 (IVT)

Let f ∶ X → R be a continuous function and X be a connected space. If a, b ∈ X and
r ∈ (f (a), f (b)), then ∃c ∈ X with f (c) = r.

P A = f (X) ∩ (−∞, r), B = f (X) ∩ (r, ∞). A, B are open in f (X), A ∩ B = ∅,
A ≠ ∅, B ≠ ∅ (as f (a) ∈ A, f (b) ∈ B ).
If r ∉ f (X), then f (X) = A ∪ B , i.e. A, B is a separation of f (X). However f (X) is
connected, because X is connected and f is continuous. I.e. such a separation cannot
exist. I.e. r ∈ f (X), that is, ∃c ∈ X such that f (c) = r.
Remark 1.45 One can prove an even more generalised form of IVT, where instead of R
one considers any ordered set.

De nition 1.46 Let X be a topological space and x, y ∈ X . A path in X from x to y

is a continuous map f ∶ [a, b] → X with f (a) = x and f (b) = y . X is called path
connected, if every pair of points in X can be joined by a path in x.

Claim: If X is path connected, then X is connected.

P Suppose X = C ∪ D, C, D a separation of X . Consider f ∶ [a, b] → X a
path. f ([a, b]) is connected, so f ([a, b]) ⊆ C or f ([a, b]) ⊆ D. is implies that
there is no path in X joining a point in C to a point in D. Ô⇒ X not path connected,
contradiction.
Example 1.48 (Topologist’s sine curve S )
Consider S = {(x, sin x1 ) ∣ 0 < x ≤ 1} ⊆ R2 .
f ∶ (0, 1] → R2 , x ↦ (x, sin x1 ). f is continuous, (0, 1] is connected, therefore S =
f ((0, 1]) is connected.
S = S ∪ {0 × [−1, 1]}. S is connected Ô⇒ S is connected.

Claim: S is not path connected.

Proof: Suppopse f ∶ [a, c] → S continuous with f (a) = (0, 0), and f (c) ∈ S . f −1 (0 ×
[−1, 1]) is closed in [a, c] therefore it has a largest element b. en f (b) ∈ 0 × [−1, 1]

11
Topology

and f ((b, c]) ∈ S . Let f (t) = (x(t), y(t)), t ∈ [a, c]. en x(a) = 0, x(t) > 0 and
y(t) = sin x(t)
1
for t > b. Given n ∈ N, choose un with 0 < un < x(b + n1 ) such that
sin u1n = (±1)n . By the IVT there exists tn ∈ (b, b + n1 ) with x(tn ) = un . en tn → b
but y(tn ) does not converge. Contradicting the continuity of f .
Remarks on yesterday’s example
• If X = R2 and S ⊆ X , then S = {limits of convergent sequences of points in S}.
More generally, this is true if X is rst countable, e.g. X is a metric space.
• If f ∶ [a, c] → S continuous. Of t ∈ [a, c] ∀n ∈ N, with tn → b, b ∈ [a, c], then
f (tn ) → f (b).
In general, continuous implies sequentially continuous.

De nition 1.49 Given a topological space X , de ne an equivalence relation ∼ on X ,

by setting x ∼ y (x, y ∈ X ) if there exists a connected subspace of X with x, y ∈ Y . e
equivalence classes are called the connected component of X .
De ne another equivalence relation ∼ by setting x ∼ y (x, y ∈ X ) if there is a path in X
from x to y . e equivalence classes are called path components of X .

1.2.2. Compactness

De nition 1.50 A topological space X is called compact if for any open covering
{Uλ }λ∈Λ of X , i.e. Uλ open ∀λ ∈ Λ and X = ∪λ∈Λ Uλ , there exist nitely many
λ1 , . . . , λn ∈ Λ such that X = ∪ni=1 Uλi .

Compactness is a topological property.

Example 1.51
1. R is not compact, R = ⋃r∈R (r − 21 , r + 12 ), or R = ⋃n∈N (n, n + 2).

2. (a, b], a, b ∈ R is not compact. (a, b] = ⋃n∈N (a + n1 , b].

3. Subspaces of X with nitely many points are obviously compact.
4. X = {0} ∪ { n1 ∣ n ∈ N} is compact. Consider {Uλ }λ∈Λ an open covering of X and
consider λ0 ∈ Λ with 0 ∈ Uλ0 . en there exists Nλ0 sucht that n1 ∈ Uλ0 ∀n ≥ Nλ0 .
Nλ
For each n ∈ {1, . . . , Nλ0 − 1} choose Uλn containing it. en ⋃i=00 Uλi = X .

Order topology

De nition 1.52 A relation C on a set X is an order relation if it has the following prop-
erties:

1. If x, y ∈ X , x ≠ y then xCY or yCx.

2. For no x ∈ X does xCx hold.

3. If xCy , yCz then xCz , x, y, z ∈ X .

12
 1.2 Conectedness & Compactness

Example 1.53 X = R, C = <.

If X is a set with an order relation < and a, b ∈ X , a < b.

(a, b) = {x ∈ X ∣ a < x < b},

(a, b] = {x ∈ X ∣ a < x ≤ b},
[a, b) = {x ∈ X ∣ a ≤ x < b},
[a, b] = {x ∈ X ∣ a ≤ x ≤ b}.

De nition 1.54 X is a set with an order relation and more than two elements. Let B
be the collection of all sets of the following types:

• (a, b), a, b ∈ X ,

• (a, b0 ], a ∈ X , b0 the largest element (if any) in X ,

• [a0 , b), b ∈ X , a0 the smallest element (if any) in X ,

B is a basis for a topology, and we call the topology generated by B the order topology.

Example 1.55 R2 with the dictionary order, i.e. a × b < c × d if a < c or a = c and b < d.
R × R has no largest or smallest element. [Graph .. open sets ... ]

An ordered set X has the least upper bound property, if every subset of X that is bounded
above has a least upper bound.

eorem 1.56 Let X an ordered set with the least upper bound property. en any closed
interval [a, b] in X is compact.

[Without proof]
Corollary 1.57 In R (Tstand = Tordered ), [a, b] is compact for any a, b ∈ R.

eorem 1.58 A closed subspace Y of a compact space X is compact.

P Consider {Uλ }λ∈Λ an open covering of Y . en Uλ = Y ∩ Uλ′ for some Uλ′ open
in X , ∀λ ∈ Λ. Denote U ′ = {Uλ′ }λ∈Λ . Now U ′′ = U ′ ∪ {X ∖ Y } is an open covering of
X . X is compact, so a nite subcollection of U ′′ covers X . If this contains X ∖Y , discard
it. If not I leave the subcollection unchanged. What we obtain is a nite subcollection of
U ′ , say {Uλ′ }i∈{1,...,n} with Y ⊆ ⋃ni=1 Uλ′ i . is implies, that {Uλi }i∈{1,...,n} is a nite
subcovering of Y .

eorem 1.60 X compact topological spaces. Y topological space and f ∶ X → Y contin-

uous. en f (X) is compact.

P Let U = {Uλ }λ∈Λ be an open covering of f (X). en ∀λ ∈ Λ Uλ = Uλ′ ∩ f (X)
for some Uλ′ open in Y . Denote U ′ = {Uλ′ }λ∈Λ . {f −1 (Uλ′ )}λ∈Λ is an open cover-
ing of X . X is compact, hence it has a nite subcovering {f −1 (Uλ′ i )}i∈{1,...,n} . en
{Uλi }i∈{1,...,n} is a nite covering of X .

13
Topology

eorem 1.62 e product of nitely many compact spaces is compact.

[Proof omitted]

eorem 1.63 (Extreme value theorem)

X is a compact topological space and Y is an ordered set in the order topology. If f ∶ X → Y
is continuous then ∃c, d ∈ X with f (c) ≤ f (x) ≤ f (d) ∀x ∈ X .

P X is compact and f is continuous, so f (X) is compact. De ne (−∞, y) ∶= {a ∈

Y ∣ a < y} and note that if Y has no smallest element, then (−∞, y) = ⋃ a∈Y (a, y) and
a<y
if Y has a smallest element a0 , then (−∞, y) = [a0 , y).
(−∞, y) is an open set in Y .
If f (X) has no largest element, then {(−∞, y) ∩ f (X)}y∈f (X) is open covering of
f (X). Since f (x) is compact, there exist yi , i ∈ {1, . . . , n}, such that {(−∞, yi ) ∩
f (X)}i∈{1,...,n} covers f (X). If yj = max{yi }i∈{1,...,n} , then yj ∈ f (X), but yj ∉
{(−∞, yi ) ∩ f (X)}, i ∈ {1, . . . , n}, contradiction, since {(−∞, yi ) ∩ f (X)}i∈{1,...,n}
is an open covering of f (X).
Similarly we can prove, that f (X) has a smallest element.f

Hausdorff spaces
Terminology: X topological space, x ∈ X , U open set in X with x ∈ U , then U is called
a open neighbourhood of x.

De nition 1.65 A topological space X is called Hausdorff if for every pair x1 , x2 ∈

X with x1 ≠ x2 there exist open neighbourhoods U1 , U2 of x1 , x2 respectively with
U1 ∩ U2 = ∅.

Facts:
• Every ordered set with the order topology is a Hausdorff space.
• If X, Y are Hausdorff spaces, then X × Y is Hausdorff.
• A subspace of a Hausdorff space is Hausdorff.
• Every nite point set in a Hausdorff space is closed.

eorem 1.66 Every compact subspace of a Hausdorff space is closed.

P X Hausdorff space, Y ⊆ X compact Ô⇒ Y is closed.

We will prove, that X ∖ Y is open.
Consider x ∈ X ∖Y . ∀y ∈ Y chose Uy , Vy open in X with x ∈ Uy , y ∈ Vy and Uy ∩Vy = ∅
(this can be done, because X is Hausdorff). en {Y ∩ Vy }y∈Y is an open covering
of Y . Y is compact, so ∃{Y ∩ Vyi }i∈{1,...,n} a nite subcovery of Y . Note that Y ⊆
Vy1 ∪ ⋅ ⋅ ⋅ ∪ Vyn =∶ V . Ux ∶= Uy1 ∩ ⋅ ⋅ ⋅ ∩ Uyn . en Ux ∩ V = ∅. Indeed, if z ∈ Ux ∩ V ,
then z ∈ Uyi ∀i ∈ {1, . . . , n} and z ∈ Vyj for some j ∈ {1, . . . , n} but Vyi ∩ Uyj = ∅. So
Ux ∩ V = ∅. We have constructed Ux open in X with x ∈ Ux and Ux ∩ Y ⊆ Ux ∩ V = ∅,
i.e. Ux ⊆ X ∖ Y .
So X ∖ Y = ∪x∈X∖Y Ux and X ∖ Y is open.

14
 1.2 Conectedness & Compactness

Remark 1.68 (X, T ) topological space, x ∈ X . Some authors (e.g. Munkres) de ne: A
neighbourhood of X is a set U ∈ T such that x ∈ U . Other authors (e.g. Janich) de ne:
A neighbourhood of X is a set U ⊆ X such that ∃V ∈ T with x ∈ V ⊆ U .

eorem 1.69 Let f ∶ X → Y be a continuous bijective function . If X is compact, and Y

is Hausdorff, then f is a homeomorphism.

P We have to prove that f −1 is continuous. Equivalently tat if U is open in X , then

(f −1 )−1 (U ) is open in Y . Equivalently, that if U is closed in X , then f (X) is closed in
Y.
Indeed, U is closed in X and X compact, so f (U ) ⊂ Y is compact. Y is Hausdorff,
therefore f (U ) is closed in Y .

Local connectedness and local path connectedness

De nition 1.71 X is called locally (path) connected at x ∈ X if for every open neigh-
bourhood U of x there is a (path) connected open neighbourhood V of x with V ⊆ U .
If Y is locally (path) connected at every x ∈ X , then X is called locally (path) connected.

15
2 Algebraic topology

Determining whether two spaces are homeomorphic and studying continuous functions
between topological spaces are two of the central problems in topology.
To show that X, Y are homemorphic, we need to construct f ∶ X → Y bijective, contin-
uous, with f −1 continuous.
To show that X, Y are not homeomorphic, we can for instance show that X has a topo-
logical property (e.g. connectendes or compactness) but Y does not have this property.
Any continuous map between topological spaces induces a homeomorphism between
there fundamental groups.
x0 ∈ X ↝ π1 (X, x0 ) fundamental group. We will show that if X ≃ Y then their funda-
mental group is isomorphic.

2.1. Fundamental group

2.1.1. Path homotopy

Notation: I ∶= [0, 1].

De nition 2.1 Let f ∶ X → Y , f ′ ∶ X → Y be continuous maps. f, f ′ are called

homeotopic if there exists a continuous map F ∶ X × I → Y with F (x, 0) = f (x)
∀x ∈ X and F (x, 1) = f ′ (x) ∀x ∈ X .
F is called a homotopy between f and f ′ .
If f, f ′ are homotopic, we write f ≃ f ′ .
If f is homotopic to a constant map then f is called nullhomotopic.

We focus on the special case X = [a, b].

en if f ∶ [a, b] → Y is a continuous function, f is called a path. f (a) is called the
initial point , and f (b) the nal point .
WLOG we can assume, that the domain of f is [0, 1].

De nition 2.2 Let f ∶ I → X , f ′ ∶ I → X two paths in X . f, f ′ are called path

homotopic if f (0) = f ′ (0), f (1) = f ′ (1) and there exists continuous map F ∶ I × I →
X , with F (x, 0) = f (x) ∀x ∈ I and F (x, 1) = f ′ (x) ∀x ∈ I and F (0, t) = f (0)
∀t ∈ I , F (1, t) = f (1) ∀t ∈ I .
If f, f ′ are path homotopic, we write f ≃p f ′ .

Claim: ≃ and ≃p are equivalence relations.

P We prove this for ≃p .
1. f ≃p f :
F (x, t) ∶= f (x).

16
 2.1 Fundamental group

2. f ≃p f ′ Ô⇒ f ′ ≃p f :
If F is a path homotopy from f to f ′ , then G(x, t) ∶= F (x, 1 − t) is a path homo-
topy from f ′ to f .
3. f ≃p f ′ , f ′ ≃p f ′′ Ô⇒ f ≃p f ′′ :
If F is a path homotopy from f to f ′ and F ′ is a path homotopy from f ′ to f ′′ ,
then
F (x, 2t) t ∈ [0, 1/2]
G(x, t) ∶= { ′
F (x, 2t − 1) t ∈ [1/2, 1]
is a path homotopy from f to f ′′ . G(x, 1/2) = F (x, 1) = F ′ (x, 0) = f ′ (x). i.e.
G is well de ned. Is G continuous?
G is continuous in I ×[0, 1/2] and G is continuous in [1/2, 1] Ô⇒ [pasting lemma
(lemma 1.21] G is continuous.

F (0, 2t) t ∈ [0, 1/2] f (0) = f ′ (0)

G(0, t) = { ={ ′
F ′ (0, 2t − 1) t ∈ [1/2, 1] f (0) = f ′′ (0)
G(1, t) = f (1),
G(x, 0) = f (x) ∀x ∈ I (since G(x, 0) = F (x, 0) = f (x) ∀x ∈ I .
G(x, 1) = f ′′ (x) ∀x ∈ I (since G(x, 1) = F ′ (x, 1) = f ′′ (x) ∀x ∈ I .
If f is a path, we denote its homotopy class by [f ].

Example 2.4
1. Let f ∶ I → R2 , g ∶ I → R2 , paths with f (0) = g(0) and f (1) = g(1). en
F ∶ I × I → R2 with (x, t) ↦ (1 − t)f (x) + tg(x) is a path homotopy between f
and g .
F is called a linear homotopy.
2. f ∶ I → R2 , s ↦ (cos(πs), sin(πs)), g ∶ I → R2 , t ↦ (cos(πt), 3 sin(πt)).
f (0) = (1, 0) = g(0), and f (1) = (−1, 0) = g(1).
f, g are path homotopic. Indeed, consider the linear homotopy F described in 1.
3. f ∶ I → R2 ∖ {(0, 0)} with s ↦ (cos(πs), sin(πs)), and g ∶ I → R2 ∖ {(0, 0)}
with t ↦ (cos(πt), sin(πt).
Are f, g path homotopic? Yes. Prove using the linear homotopy.
4. f ∶ I → R2 ∖ {(0, 0)}, with s ↦ (cos(πs), sin(πs)), and g ∶ I → R2 ∖ {(0, 0)}
with t ↦ (cos(πt), −3 sin(πt)).
e linear homotopy is not a path homotopy between f and g .
Inf fact, there exists no path homotopy between f and g .

We de ne the following operation:

De nition 2.5 Let f ∶ I → X path in X with f (0) = x0 , f (1) = x1 , x0 , x1 ∈ X .

g ∶ I → X path in X with g(0) = x1 , g(1) = x2 , x2 ∈ X .
We de ne the product f ⋅ g of f, g as the path h with
f (2s) s ∈ [0, 1/2]
h(s) = {
g(2s − 1) s ∈ [1/2, 1].

17
Algebraic topology

Remark 2.6 h is well de ned, since f (1) = g(0) = h(1/2).

h is continuous by the pasting lemma (lemma 1.21).
e product f ⋅ g induces a well de ned product on path homotopy equivalce classes
de ned by [f ] ⋅ [g] = [f ⋅ g].
Indeed, if f, f ′ paths in X with [f ] = [f ′ ], and g, g ′ paths in X with [g] = [g ′ ]. Consider
path homotopies F, G between f and f ′ , g and g ′ respectively. en
F (2s, t) t ∈ [0, 1/2]
H(s, t) ∶= {
G(2s − 1, t) s ∈ [1/2, 1]
is a path homotopy between f ⋅ g and f ′ ⋅ g ′ (homework).
erefore [f ⋅ g] = [f ′ ⋅ g ′ ], i.e. [f ] ⋅ [g] = [f ′ ] ⋅ [g ′ ] and the induced product is well-
de ned.

Reparametrisation
De ne a reparametrisation of a path f to be a composition f ○ φ, where φ ∶ I → I
continuous map, such that φ(0) = 0, and φ(1) = 1.
Reparametrising a path preserves its homotopy class (i.e. [f ○φ] = [f ]). Indeed, consider
the path homotopy f ○ φt whereφt (s) ∶= (1 − t)φ(s) + ts.
φ0 (s) = φ(s) ↝ f ○ φ0 = f ○ φ, φ1 (s) = s ↝ f ○ φ1 = f .
Note that (1 − t)φ(s) + ts lies between φ(s) and s, hence (1 − t)φ(s) + ts lies in I , so
f ○ φt is de ned.

2.2. e fundamental group

We restrict our attention to paths f ∶ I → X with f (0) = f (1). Call x0 ∶= f (0) = f (1).
Such paths are called loops in X at the basepoint x0 . e set {[f ] ∣ f ∶ I → X, f (0) =
f (1) = x0 } is denoted by π1 (X, x0 ).
If [f ], [g] ∈ π1 (X, x0 ), then f, g are loops in X at the basepoint x0 , so f ⋅ g is de ned.
e induced multiplication [f ] ⋅ [g] = [f ⋅ g] is well-de ned.
Proposition 2.7 (π1 (X, x0 ), ⋅) is a group.
P
1. associativity: [f ] ⋅ ([g] ⋅ [h]) = ([f ] ⋅ [g]) ⋅ [h], [f ], [g], [h] ∈ π1 (X, x0 ).
⇐⇒ [f ] ⋅ [g ⋅ h] = [f ⋅ g] ⋅ [h]
⇐⇒ [f ⋅ (g ⋅ h)] = [(f ⋅ g) ⋅ h].
Claim: f ⋅(g ⋅h) is a reparametrisation of (f ⋅g)⋅h by the piecewise linear function
φ: where
⎧
⎪ 1
t t ∈ [0, 1/2]
⎪
⎪ 2 1
φ(t) = ⎨ t − 4 t ∈ [1/2, 3/4]
⎪
⎪
⎪
⎩ 2t − 1 t ∈ [3/4, 1]
Proof:
⎧
⎪ f (2t) t ∈ [0, 1/2]
f (2t) t ∈ [0, 1/2] ⎪
⎪
(f ⋅(g⋅h))(t) = { = ⎨ g(4t − 2) t ∈ [1/2, 3/4]
(g ⋅ h)(2t − 1) t ∈ [1/2, 1] ⎪
⎪
⎪
⎩ h(4t − 3) t ∈ [3/4, 1]
Analogously
⎧
⎪ f (4s) s ∈ [0, 1/4]
(f ⋅ g)(2s) s ∈ [0, 1/2] ⎪
⎪
((f ⋅ g) ⋅ h)(s) = { = ⎨ g(4s − 1) s ∈ [1/4, 1/2]
h(2s − 1) s ∈ [1/2, 1] ⎪
⎪
⎪
⎩ h(2s − 1) s ∈ [1/2, 1]

18
 2.2 e fundamental group

So if one easily checks, that ((f ⋅ g) ⋅ h)(φ(t)) = (f ⋅ (g ⋅ h))(t).

2. Identity: Consider c ∶ I → X with c(s) = x0 ∀s ∈ I . en f ⋅ c is a reparametri-

sation of f via φ:
2t t ∈ [0, 1/2]
φ(t) = {
1 t ∈ [1/2, 1]

I.e. f ⋅ c = f ○ φ.
[f ] = [f ○ φ] = [f ⋅ c].
[f ] = [f ] ⋅ [c].

3. inverse: Consider [f ] ∈ π1 (X, x0 ) and f (s) = f (1 − s).

Claim: f ⋅ f ≃ c ≃ f ⋅ f .
Proof: Consider the homotopy
⎧
⎪ f (2s) s ∈ [0, (1 − t)/2]
⎪
⎪
H(s, t) = ⎨ f (1 − t) [(1 − t)/2, (1 + t)/2]
⎪
⎪
⎪
⎩ f (2 − 2s) [(1 + t)/2, 1]
H(0, t) = f (0) = x0 ,
H(1, t) = f (0) = x0 .
f (2s) s ∈ [0, 1/2]
H(s, 0) = { =f ⋅f
f (2 − 2s) s ∈ [1/2, 1]
H(s, 1) = x0 .

H is a homotopy between f ⋅ f and c i.e. [f ⋅ f ] = [c],

i.e. [f ] ⋅ [f ] = [c].
Analogously [f ] ⋅ [f ] = [c].

Ô⇒ [f ] is the inverse of [f ].

De nition 2.9 π1 (X, x0 ) is called the fundamental group of X at x0 .

Example 2.10 Consider X ⊆ Rn convex. x0 ∈ X .

π1 (X, x0 ) = {[f ] ∣ f ∶ I → X path, f (0) = f (1) = x0 }.
If f0 , f1 are two paths in X at the basepoint x0 , then ft (s) = (1 − t)f0 (s) + tf1 (s) is a
linear homotopy between f0 and f1 .
Let x0 , x1 ∈ X , π1 (X, x0 ), π1 (X, x1 ).
Proposition 2.11 If x0 , x1 are in the same path component of X , then π1 (X, x0 ), π1 (X, x1 )
are isomorphic.
P x0 , x1 ∈ X . Consider h ∶ I → X a path with h(0) = x0 , h(1) = x1 and
h ∶ I → X , h(s) = h(1 − s).
De ne ̂ h ∶ M1 (X, x1 ) → π1 (X, x0 ) where [f ] ↦ [h ⋅ f ⋅ f ].

• ̂
h is well-de ned: If [f ] = [g] ∈ π1 (X, x1 ) and ft is a homotopy between f and g
then h ⋅ ft ⋅ h is a homotopy between h ⋅ f ⋅ h and h ⋅ g ⋅ g i.e. [h ⋅ f ⋅ h] = [h ⋅ g ⋅ h].

• ̂
h is a homomorphism: ̂h([f ] ⋅ [g])̂
h([f ⋅ g]) = [h ⋅ f ⋅ g ⋅ h] = [h ⋅ f ⋅ h ⋅ h ⋅ g ⋅ h] =
̂ ̂
[hf h][hgh] = h([f ])h([g]).

19
Algebraic topology

h is bijective: ̂
• ̂ h = (̂
h)−1 .
(̂ ̂
hh)[f ] = ̂ ̂
h(h([f ])) = ̂ h([hf h]) = ⋅ ⋅ ⋅ = [f ].

De nition 2.13 X is simply connected if it is path connected and π1 (X) is trivial.

Proposition 2.14 X is simply connected iff there is a unique homotopy class of paths

connecting any two points in X .
P Path connectedness is the existence of paths connecting every pair of points in
X . So we only need to check uniqueness.
“⇒” Suppose π1 (X) = 0. Let x0 , x1 ∈ X . If f, g are two paths from x0 to x1 , then
f ≃ f gg ≃ g . Ô⇒ [f ] = [g].
“⇐” If there is a unique homotopy class of paths connecting any two points, then there
is a unique homotopy class of paths connecting a point to itself. Ô⇒ π1 (X) = 0.

Induced homomorphisms
X, Y topological spaces, x0 ∈ X , y0 ∈ Y and h ∶ X → Y a continuous map with
h(x0 ) = y0 . We will denote this by h ∶ (X, x0 ) → (Y, y0 ).
If f ∶ I → X is aloop in X , based at x0 , then h ○ f is a loop in Y based at y0 .

De nition 2.16 Let h ∶ (X, x0 ) → (Y, y0 ) continuous. De ne h∗ ∶ π1 (X, x0 ) →

π1 (Y, y0 ) where [f ] ↦ [h ○ f ]. h∗ is called the homomorphism induced by h relative
to x0 .

Remark 2.17 h∗ is well-de ned.

If f, g loops in X based at x0 with [f ] = [g]. Consider H a homotopy from f to g . en
h ○ H is a homotopy from h ⋅ f to h ⋅ g .
I.e. [h ⋅ f ] = [h ⋅ g] ⇐⇒ h∗ ([f ]) = h∗ ([g]).
Remark 2.18 h∗ is a group homomorphism.
Just calculate and compare the following two expressions:

h∗ ([f ] ⋅ [g]) = h∗ ([f ⋅ g]) = [h ○ (f ⋅ g)]

h∗ ([f ]) ⋅ h∗ ([g]) = [h ○ f ] ⋅ h[⋅g] = [(h ○ f ) ⋅ (h ○ g)].

Remark 2.19 h∗ depends on the basepoint x0 . So strictly speaking we should have writ-
ten (hx0 )∗ .

Properties of induces homomorphisms

1. If h ∶ (X, x0 ) → (Y, y0 ), g ∶ (Y, y0 ) → (Z, z0 ) continuous, then (g○h)∗ = g∗ ○h∗ .

2. If 1X ∶ (X, x0 ) → (X, x0 ), then (1x )∗ = 1π1 (X,x0 ) .

Proposition 2.20 If H ∶ (X, x0 ) → (Y, y0 ) a homoeomorphism,t hen h∗ ∶ π1 (X, x0 ) →

π1 (Y, y0 ) is a grou isomorphism.

20
 2.3 Covering spaces

P Consider h−1 (Y, y0 ) → (X, x0 ) and (h−1 )∗ ∶ π1 (Y, y0 ) → π1 (X, x0 )

(h−1 )∗ ○ h∗ = (h−1 ○ h)∗ = 1∗ = 1.
h∗ ○ (h−1 )∗ = (h ○ h1 )∗ = 1∗ = 1.
π1 gives a covariant functor from the category with objects topological spases with base-
point and morphism basepoint preserving continuous maps to the category with objects
groups and morphisms group homomorphisms.

A category D consists of

1. a collection of objects Ob(D),

2. sets of morphisms Mor(X, Y ) for each X, Y ∈ Ob(D) with a distinguished iden-

tity morphism 1X in Mor(X, X),

3. a composition of morphisms ○ ∶ Mor(X, Y ) × Mor(Y, Z(→ Mor(X, Z) for each

tripel X, Y, Z ∈ Ob(D) with the properties f ○ 1 = 1 ○ f = f and (f ○ g) ○ h =
f ○ (g ○ h).

Example 2.22

1. Ob(D) = {G ∣ G group}.
Mor(G, H) = {f ∶ G → H ∣ f group homomorphism}.

2. Ob(D) = {(X, x0 ) ∣ X topological space, x0 basepoint}.

Mor((X, x0 ), (Y, y0 )) = {f ∶ (X, x0 ) → (Y, y0 ) ∣ f continuous}.

A coveriant functor F from a categoriy C to a category D assigns to each X ∈ Ob(C)

an F (X) ∈ Ob(D) to each f ∈ Mor(X, Y ) a F (f ) such that F (1X ) = 1F (X) and
F (f ○ g) = F (f ) ○ F (g).
Example 2.23 π1 is a covariant functor from category C2 in example 2 to the category C1
in example 1.
π1
(X, x0 ) ∈ Ob(C2 ) ↝ π1 (X, x0 ) ∈ Ob(C1 ).
π1
f ∈ Mor((X, x0 ), (Y, y0 )) ↝ f∗ ∈ Mor(π1 (X, x0 ), (Y, y0 ))
(1(X,x0 ) )∗ = 1π1 (X,x0 ) .
(f ○ g)∗ = f∗ ○ g∗ .

2.3. Covering spaces

• useful for computing π1 ,

• algebraic features of π1 can be translated into geometric features of the spaces.

De nition 2.24 Consider p ∶ X ̃ → X a continuous and surjective map. If there exists

′
an open cover {Ua }a∈A of X such that ∀a ∈ A p−1 (Ua ) = ∪b∈Ba Vab , where Vab ∩ Vab =
′ ′
∅ for any b, b ∈ Ba , b ≠ b . p ∶ Va → Ua is a homeomorphism ∀b ∈ Ba and Va open
b b
̃ ∀b ∈ Ba , then p is called a covering map. X
in X ̃ is called a covering space of X .

Example 2.25

21
Algebraic topology

1. S1 = {(x, y) ∈ R2 ∣ x2 + y 2 = 1}. p ∶ R → S1 where s ↦ (cos 2πs, sin 2πs).

p is a covering map, p is continuous (calculus), and p is surjective.
√
U1 ∶= {(x, y) ∈ S1 ∣ x > −√ 2/2}
U2 ∶= {(x, y) ∈ S1 ∣ x < 2/2}.

.
− 21 1
2

{U1 , U2 } is an open cover of S1

p−1 (U1 ) = ∪n∈Z Vn , where Vn ∶= (n − 3/8, n + 3/8).
Vn is open in R ∀n ∈ Z.
Vn ∩ Vn = ∅, for any n, m ∈ Z, n ≠ m.
p ∶ Vn → U1 is a homeomorphism ∀n ∈ Z.
We can make the analogous construction for U2 . So p is a covering map.
Remark 2.26 You can prove that p is a covering by using any open cover of S1 by
two open subsets of S1 (≠ S1 ).
2. S1 = {z ∈ C ∣ ∣z∣ = 1}, n ∈ N, n ≥ 1.
pn ∶ S1 → S1 , z ↦ z n . pn is a covering map (problem sheet 5).
f ∶ X → Y is called an embedding if f ∶ X → f (X) is a homeomorphism.

Consider the solid torus S1 × D2 where D2 = {(x, y) ∈ R2 ∣ x2 + y 2 ≤ 1}, and its

boundary ∂(S1 × D2 ) = S1 × S1 .
Consider f ∶ S1 → ∂(S1 × D2 ) an embedding such that f (S1 ) wraps around the
rst S1 three times.
Lastly consider the projection π ∶ S1 × D2 → S1 × {(0, 0)} and restrict it to f (S1 ).
3. f ∶ R+ → S1 , where s ↦ (cos 2πs, sin 2πs) is not a covering map.
Consider an open cover {Ua }a∈A of S1 . en ∃U ∈ {Ua }a∈A with x ∈ U . p−1 (U ) =
V0 ∪ (∪b∈Ba Vab ).
p ∶ V0 → U is not a homeomorphism.

eorem 2.27 If p ∶ X̃ → X is a covering map, X0 ⊆ X , and X ̃0 = p−1 (X0 ), then

̃ ̃
p0 ∶ X0 → X0 obtained by restricting p to X0 is a covering map.

P p0 is continuous (restricting the domain or the range of a continuous function

gives a continuous function). p0 is surjective.
Consider an open cover {Ua }a∈A of X with the properties in the de nition of covering
map.

22
 2.3 Covering spaces

en ∀a ∈ A p−1 (Ua ) = ∪b∈Ba Vab .

If x0 ∈ X0 , then there exists a ∈ A with x0 ∈ Ua . Now Ua ∩ X0 is an open set in X0 .
p−1 −1 −1 −1 ̃ ̃
0 (Ua ∩ X0 ) = p0 (Ua ) ∩ p0 (X0 ) = p (Ua ) ∩ X0 = (∪b∈Ba Va ) ∩ X0 = ∪b∈Ba (Va ∩
b b

X̃0 ).
We have
̃0 open in X
• Vab ∩ X ̃0 ,

̃0 ) ∩ (V b′ ∩ X
• (Vab ∩ X ̃0 ) = ∅ for any b, b′ ∈ Ba , b ≠ b′ .
a

̃0 → Ua ∩ X0 is a homeomorphism.
• p0 ∶ Vab ∩ X
̃0 → X0 is
{Ua ∩ X0 }a∈A is an open cover of X0 with the desired properties Ô⇒ p0 ∶ X
a covering map.

eorem 2.29 If p ∶ X̃ → X , p′ ∶ X
̃ ′ → X ′ are covering maps, then p × p′ ∶ X
̃×X
̃′ →
′
X × X is a covering map.

Example 2.30

1. Consider p ∶ R → S1 , where s ↦ (cos 2πs, sin 2πs).

e map p × p ∶ R × R → S1 × S1 is a covering map.

2. X0 = (S1 × p(0)) ∪ (p(0) × S1 ).

X̃0 = (p × p)−1 (X0 ) = (R × Z) ∪ (Z × R).
̃0 → X0 is a covering map (by theorem).
(p × p)0 ∶ X
Two circles S1 with a one common point is called wedge of two circles (notation
S1 ∨ S1 ).

3. Other covering spaces of S1 ∨ S1 .

b a

. x

̃ as in the following picture

Consider X

a b a

x1 . x2

en p ∶ X ̃ → S1 ∨ S1 , where x1 ↦ x and x2 ↦ x. p maps each edge of X

̃ to the
edge of X with the same label by a map that is a homeomorphism and preserves
the orientation.

23
Algebraic topology

4. Consider p × 1R+ ∶ R × R+ → S1 × R+ (where p ∶ R → S1 : s ↦ (cos 2πs, sin 2πs))

and f ∶ S1 × R+ → R2 ∖ {(0, 0)}, where (x, t) ↦ t ⋅ x.
f is a homeomorphism.
f ○ (p × 1R+ ) ∶ R × R+ → R2 ∖ {(0, 0)} is a covering map.

P p × p′ is continuous (problem sheet 3). p × p′ is surjective (if (x, x′ ) ∈ X × X ′ ,

then x ∈ X and x′ ∈ X ′ Ô⇒ [p, p′ is surjective] ∃̃ x∈X ̃, x
̃′ ∈ X̃ ′ , such that p(̃
x) = x
′
and p (̃ ′ ′
x ) = x . Ô⇒ ∃(̃ x, x ′ ̃ ̃ ′ ′
̃ ) ∈ X × X such that (p × p )(̃ ̃ ) = (x, x′ )).
x, x′

Consider {Ua }a∈A , {Uc′ }c∈C open covers of X and X ′ as in the de nition of covering
maps.
If (x, x′ ) ∈ X × X ′ then x ∈ X and x′ ∈ X ′ i.e. ∃U ∈ {Ua }a∈A with x ∈ U and
∃U ′ ∈ {Uc }c∈C with x′ ∈ U ′ .
p−1 (U ) = ∪b∈B Vb , Vb open Vb ∩ Vb′ = ∅ for b ≠ b′ and p ∶ V ∶b → U homeomorphism
∀b ∈ B . (p′ )−1 (U ′ ) = ∪d∈D Vd′ , Vd′ open, Vd′ ∩ Vd′′ = ∅ for d ≠ d′ and p′ ∶ Vd′ → U ′
homeomorphism ∀d ∈ D.
(x, x′ ) ∈ U × U ′ , U × U ′ is open in X × X ′ ,
(p × p′ )−1 (U × U ′ ) = ∪ b∈B (Vb × Vd′ ), where Vb × Vd′ open in X × X ′ , Vb × Vd′ disjoint,
d∈D
p × p′ ∶ Vb × Vd′ → U × U ′ is a homeomorphism.
I.e. {Ua × Uc′ } a∈A is the open cover of X × X ′ . at shows that p × p′ is a covering map.
c∈C

2.4. Liing properties

We will discuss two properties of covering spaces.

̃ → X be a map. If f ∶ Y → X is continuous a liing of f is a

De nition 2.32 Let p ∶ X
map f ∶ Y → X such that p ○ f̃ = f .
̃ ̃

Example 2.33 p ∶ R → S1 , s ↦ (cos 2πs, sin 2πs), f ∶ I → S1 the path f (s) =

(cos πs, sin πs).
en f̃ ∶ I → R, f̃(s) = 2s is a liing of f .
Proposition 2.34 (path liing property)
̃ → X be a covering map with p(̃
Let p ∶ X x0 ) = x0 . If f ∶ I → X is a path with f (0) = x0 ,
then ∃! liing of f to a path f̃ ∶ I → X
̃ with f̃(0) = x̃0 .
P

1. Consider the open cover {Ua }a∈A of X (as in the de ntion of covering map). Let
t ∈ (0, 1), and f t) ∈ X Ô⇒ f (t) ∈ Ua for some a ∈ A. f ∶ I → X is continuous
Ô⇒ ∃(at , bt ) ⊆ (0, 1) with t ∈ (at , bt ) and f ([at , bt ]) ⊆ Ua .

Analogously for t = 0 ∃[0, b0 ) ⊆ [0, 1] with f ([0, b0 ]) ⊆ Ua , and for t = 1

∃(a0 , 1] ⊆ [0, 1] with f ([a0 , 1]) ⊆ Ua .
I is compact, and {(at , bt )}t∈(0,1) together with [0, b0 ) and (b1 , 1] is an open
cover of I . We can choose a nite subcover, say [0, b0 ), (a0 , 1], (a1 , b1 ), . . . , (am , bm )}.
Consider {ai , bi }i∈{0,...,m} and orther its elements. (Possibly) rename the ele-
ments as follows: 0 < t1 ≤ ⋅ ⋅ ⋅ ≤ t2m+2 < 1. this gives a subdivision 0 = s0 < s1 <
⋅ ⋅ ⋅ < sn = 1 of I with the property, that ∀i ∈ {0, . . . , n − 1} f ([si , si+1 ]) ⊆ Ua for
some a ∈ A.

24
 2.4 Liing properties

2. De ne f̃(0) = X ̃0 . Suppose that f̃ is de ned in [0, si ]. De ne f̃ in [si , si+1 ] as

follows: f ([si , si+1 ]) ⊆ U for some U in {Ua }a∈A .
Let p−1 (U ) = ∪b∈B Vb . Now (p ○ f̃)(si ) = f (si ) ∈ U .
Ô⇒ f̃(si ) ∈ p−1 (U ) = ∪b∈B Vb
Ô⇒ f̃(Si ) ∈ V0 for some V0 ∈ {Vb }b∈B .
De ne f̃(s) = (p∣V )−1 (f (s)), s ∈ [si , si+1 ].
0

f̃is continuous in [si , si+1 ], since p∣V0 is a homeomorphism. us we have de ned
f̃ ∶ [0, 1] → X
̃ continuous with f̃(0) = x ̃0 .
̃ ̃ ̃
3. Suppose f̃ is another liing of f with f̃(0) = x0 . en f̃(0) = f̃(0) = x0 .
̃ ̃
Suppose that f̃(s) = f̃(s) in [0, si ]. p ○ f̃([si , si+1 ]) = f ([si , si+1 ]) ⊆ Ua for
some a ∈ A.
̃
Ô⇒ f̃([si , si+1 ]) ⊆ p−1 (Ua ) = ∪b∈B Vb .
̃ ̃
f̃([si , si+1 ]) is connected and f̃(si ) = f̃(si ) ∈ V0 .
̃
Ô⇒ f̃([si , si+1 ]) ⊆ V0 .
̃
For s ∈ S , p ○ f̃(s) = f (s)
̃
Ô⇒ f̃(s) ∈ p−1 (f (s)).

̃ ̃
Also f̃ ∈ V0 . So f̃(s) = (p∣V0 )−1 (f (s)) = f̃(s).
̃
I.e. f̃(s) = f̃(s).
Proposition 2.36 Let p ∶ X ̃ → X be a covering map with p(̃ x0 ) = x0 . Let F ∶ I × I → X
be continuous with F (0, 0) = x0 en ∃! liing of F to a continuous map F̃ ∶ I × I → X ̃
with F̃(0, 0) = x
̃0 . If F is a path homotopy then F̃ is a path homotopy.
P Consider {Ua }a∈A an open cover of X with the properties as in the de nition of
covering map.
1. De ne F̃((0, 0)) = x ̃0 .
Extend F̃ to I × {0} and {0} × I using the proposition from last time. Extend̃to
I × I , as follows: Choose subdivision s0 < s1 < ⋅ ⋅ ⋅ < sm , t0 < t1 < ⋅ ⋅ ⋅ < tn of
I with the property that for each such rectangle Ii × Jj = [si−1 , si ] × [tj−1 , tj ],
F (Ii ×Jj ) ⊆ Ua for some a ∈ A. De ne F̃ rst in I1 ×Ij , then I2 ×J1 ,. . . , Im ×J1 ,
then I1 × J2 , I2 × J2 and so on.
Consider Ii0 × Jj0 and suppose that F̃ is de ned on the union A of the rectangles
Ii × Jj with j < j0 or j = j0 and i < i0 .
Consider C = A ∩ (Ii0 × Jj0 ). Choose U ∈ {Ua }a∈A with F (Ii0 × Jj0 ) ⊆ U .
p−1 (U )) ∪b∈B Vb , F̃ is de ned in C , C is connected. So, F̃(C) is connected, i.e.
∃V0 such that F̃(C) ⊆ V0 .
If p0 = p∣x0 ∶ V0 → U , then p0 ○ F̃(x) = p ○ F̃(x) = F (x) Ô⇒ F (x) =
(p−1
0 )(F (x)).
De ne F̃(x) = p−1 (F (x)) ∀x ∈ Ii × Jj .
0 0 0

F̃ is continuous, according to the pasting lemma.

2. Check that at every step, there is a unique way to de ne F̃.

25
Algebraic topology

3. If F is a path homotopy, then F ({0} × I) = x0 . Also, F̃({0} × I) ⊆ p−1 ({x0 }).

p−1 ({x0 }) has the discrete topology as a subspace of X̃ . Anlose F̃({0} × I) is
connected. us F̃({0} × I) = {̃ x0 }.
Similarly F̃({1} × I) is a set with one point.
us, F̃ is a path homotopy.

Proposition 2.38 (Homotopy liing property)

p∶X̃ → X covering map with p(̃ x0 ) = x0 . Consider f, g the paths in X from x0 to x1
and f̃, ̃ ̃ with f̃(0) = ̃
g their liings to paths in X ̃0 . If f ≃ g , then f̃ ≃ ̃
g (0) = x g.
P Consider F the path homotopy between f and g . en F ((0, 0)) = x0 . Let
F̃ ∶ I × I → X ̃ the liing of f to X
̃ with F̃((0, 0)) = x ̃0 . en F ({0} × I) = {̃ x0 } and
̃
F ({1} × I) = {̃ x1 }.
F̃∣I×{0 } is a path i X ̃0 that is a liing of F∣I×{0} . I.e. F̃I×{0 } = f̃(uniqueness
̃ starting at x
of path liings).
F̃∣I×{1} = ̃ g.
So f̃(1) = ̃ ̃1 and F̃ ∶ I × I → X
g (1) = x ̃ is a path homotopy from f̃ to ̃ g.

De nition 2.40 Let p ∶ X ̃ → X be a covering map with p(̃ x0 ) = x0 . Given a loop f in

̃ ̃ with f̃(0) = x
X based x0 , let f be the liing of f in X ̃0 . De ne

ϕ ∶ π1 (X, x0 ) → p−1 ({x0 }),

[f ] ↦ f̃(1).

ϕ is called the liing correspondence derived from the map p

Claim: ϕ is well-de ned.

P If [f ] = [f ′ ], then f, f ′ are path homotopic. We can li this homotopy to a path
homotopy between f̃ and f̃′ Ô⇒ f̃(1) = f̃′ (1).

eorem 2.42
̃ → X be covering map with p(̃
1. If p ∶ X ̃ is path connected, then ϕ is
x0 ) = x0 . If X
surjective.
̃ is simply connected, then ϕ is bijective.
2. If X

P

̃1 ∈ p−1 ({x0 }). en since X

1. Let x ̃ is path connected, there is a path f̃ ∶ I → X
̃
̃ ̃
̃0 and f (1) = x
with f (0) = x ̃
̃1 . en f = p ○ f is a loop in X based at x0 , since
f (0) = p(̃ x1 ) = x0 and ϕ([f ]) = f̃(1) = x
x0 ) = x0 , f( 1) = p(̃ ̃1 .

2. We only need to check that ϕ is injective. Indeed consider ä[f ], [g] ∈ π1 (X, x0 )
with ϕ([f ]) = ϕ([g]). en f̃(1) = ̃ g (1), where f̃, ̃
g are the liings of f, g with
̃
f (0) = ̃ ̃0 . Since X is simply connected, we have that [f̃] = [̃
g (0) = x ̃ g ], i.e.
there is a path homotopy F̃ between f̃ and ̃g . en F = p ○ F̃ is a path homotopy
between f and g , i..e [f ] = [g].

26
 2.5 e fundamental group of the circle and applications

2.5. e fundamental group of the circle and applications

Remark 2.44 S1 is path connected, so we will write π1 (S1 ).

eorem 2.45
(π1 (S1 ), ⋅) ≃ (Z, +).

P Let p ∶ R → S1 , p(t) = (cos 2πt, sin 2πt). Denote p(0) = (1, 0) = x0 . en
p−1 ({x0 }) = Z. Consider ϕ ∶ π1 (S1 ) → Z, where [f ] ↦ f̃(1).
By exercise 1 of sheet 3 R is simply connected. By the second point of the theorem ϕ is
bijective.
It remains to show that ϕ is a group homomorphism,
i.e. ϕ([f ] ⋅ [g]) = ϕ([f ]) + ϕ([g]).
⇐⇒ ϕ([f ⋅ g]) = ϕ([f ]) + ϕ([g])
⇐⇒ f̃ ⋅ g(1) = f̃(1) + ̃ g (1),
where f̃ is the li of f with f̃(0) = 0 ∈ R and ̃ g the li of g with ̃ g (0) = 0 ∈ R and f̃ ⋅g
the li of f ⋅ g where f̃ ⋅ g(0) = 0 ∈ R.
Let ̃g ′ ∶ I → R be the path ̃ g ′ (s) = ̃g (s) + f̃(1). en (p ○ ̃ g ′ )(s) = p(̃g (s) + f̃(1)) =
(p ○ ̃g )(s) = g(s).
Ô⇒ ̃ g ′ is a liing of g with ̃ g ′ (0) = f̃(1).
Furthermore, p((f̃⋅ ̃ g )(s)) = (f ⋅ g)(s) Ô⇒ f̃⋅ ̃
′
g ′ is a liing of f ⋅ g with f̃⋅ ̃
g ′ (0) = 0.
(f̃⋅ ̃
g )(+) = f̃(1) + ̃
′
g (1)
Ô⇒ f̃ ⋅ g(1) = f̃(1) + ̃ g (1).

Retractions

De nition 2.47 A retraction of a X onto A ⊂ X is a continuous map r ∶ X → A with

r(a) = a ∀a ∈ A.
If such a map exists A is called a retract of X .

Claim: If A is a retract of X and i ∶ A ↪ X is the inclusion map, then i∗ is injective.

P r ○ i ∶ A → A. r ○ i = 1A
Ô⇒ (r ○ i)∗ = (1A )∗ .
Ô⇒ r∗ ○ i∗ = 1π1 (A,x0 )
Ô⇒ i∗ is injective.
Proposition 2.49 ere is no retraction of D2 onto S1 .
P If S1 was a retract of D2 and i ∶ S1 → D2 the inclusion. en i∗ ∶ π1 (S1 ) →
π1 (D2 ) would be injective. But π1 (S1 ) ≃ Z and π1 (D)2 is trivial as a convex subset of
R2 .
Lemma 2.51 Let h ∶ S1 → X be a continuous map. e following are equivalent:

1. h is nullhomotopic.

2. ere exists a continuous map h ∶ D2 → X with h∣S1 = h.

3. h∗ is trivial.

P

27
Algebraic topology

1. “1.⇒2.”: Consider the homotopy H ∶ S1 × I → X between h and the constant

map c. Let q ∶ S1 × I → D2 with (x, t) ↦ t ⋅ x.
q is continuous, q is open and surjective.
2.

3.

Vorlesung von Di 17. März fehlt noch

Remark 2.53 Ha the corollary: If S2 = A1 ∪ A2 ∪ A3 , A1 closed ∀i ∈ {1, 2, 3}, then
∃x ∈ S2 , i ∈ {1, 2, 3} such that {x, −x} ⊆ Ai .

Inscribe a sphere in the tetrahedron. Project each face of the tetrahedral radially onto the
sphere. en one obtains four closed sets Ai , i ∈ {1, . . . , 4} with S2 = A1 ∪A2 ∪A3 ∪A4 ,
but none of the Ai contains a pair of antipodal points.

2.5.1. e Fundamental eorem of Algebra

(Any non-constant polynomial in C[x] has a root in C).
• proof in algebra,

• proof in complex analysis (corollary of Liouville’s theorem),

• proof in topology.

eorem 2.54 (Fundamental eorem of Algebra)

0 + ⋅ ⋅ ⋅ + a1 x0 + a0 = 0.
If an xn + ⋅ ⋅ ⋅ + a1 + a0 ∈ C[x], n > 0, then ∃x0 ∈ C with an xn

P Let an xn + ⋅ ⋅ ⋅ + a1 x + a0 ∈ C[x] and n > 0.

1. We can assume an = 1 (an (xn + aan−1 n
xn−1 + ⋅ ⋅ ⋅ + aan0 )).
We can assume ∑n−1 i=0 ∣ai ∣ < 1. Indeed choose c ∈ R>0 and set x = cy is gives
cn y n + ⋅ ⋅ ⋅ + a1 cy + a0 = cn (y n + an−1
c
y n−1 + ⋅ ⋅ ⋅ + can0 ). Choose c large enough
such that ∣ c ∣ + ⋅ ⋅ ⋅ + ∣ cn ∣ < 1. Now if y0 is a root of y n + an−1
an−1 a0
c
y n−1 + ⋅ ⋅ ⋅ + can0 ,
then cy0 is a root of x + ⋅ ⋅ ⋅ + a0 .
n

2. Consider f ∶ S1 → S1 , z ↦ z n and p ∶ I → S1 , s ↦ (cos 2πs, sin 2πs) = e2πis .

en f∗ ∶ π1 (S1 , (1, 0)) → π1 (S1 , (1, 0)). f∗ ([p]) = [f ○ p], where f ○ p(s) =
(cos 2πns, sin 2πns).
is shows that f∗ is injective. Also if ι ∶ S1 → R2 ∖ {(0, 0)} (inclusion), then ι∗
is injective, since S1 is a retract of R2 ∖ {(0, 0)}.
I.e. ι∗ ○ f∗ = (ι ○ f )∗ is injective, which implies (Lemma ??) that ι ○ f is not
nullhomotopic (a).

3. Suppose, that our polynomial has no root in D2 . Consider h ∶ D2 → R2 ∖{(0, 0)},

z ↦ z n + an−1 z n−1 + ⋅ ⋅ ⋅ + a1 z + a0 . By lemma ?? ↝ h∣S1 is nullhomotopic (b).
Furthermore: F ∶ S1 × I → R2 ∖ {(0, 0)} where (z, t) ↦ z n + t(an−1 z n−1 + ⋅ ⋅ ⋅ +
a1 z + a0 ). is is a homotopy between h∣S1 and ι ○ f (c).

28
 2.5 e fundamental group of the circle and applications

∣F (z, t)∣ = ∣z n + t(an−1 z n−1 + ⋅ ⋅ ⋅ + a0 )∣ ≥ ∣z n ∣ − ∣t∣∣an−1 z n−1 + ⋅ ⋅ ⋅ + a0 ∣ ≥

1 − ∣t∣ ∑n−1
i=0 ∣ai ∣ > 0.

(a),(b),(c) contradiction, i.e. our polynomial has a root in D2 .

Exercise: Any root of such a polynomial z n + ⋅ ⋅ ⋅ + a0 with ∑ ∣ai ∣ < 1 is in D2 .

2.5.2. Deformation retracts and homotopy type

Covering spaces ↝ fundamental group computation (e.g. π1 (S1 )).

De nition 2.56 Let A ⊆ X , A is called a deformation retract of X , if there is a contin-

uous map H ∶ X × I < raX with H(x, 0) = x ∀x, H(x, 1) ∈ A ∀x ∈ X , H(a, t) = a
∀a ∈ A, ∀i ∈ I . e homotopy H is called a deformation retraction of X onto A.

Remark 2.57
1. r ∶ X → A, r(x) = H(x, 1) is a retraction of X onto A.
2. H is a homotopy between 1X and ι ○ r, where ι ∶ A ↪ X inclusion map.
Example 2.58
1. {(x, y, z) ∈ R3 ∣ (x, y) ≠ (0, 0), z = 0} =∶ A is a deformation retract of
{(x, y, z) ∈ R3 ∣ (x, y) ≠ (0, 0)} =∶ X .
H ∶ X × I → X , H(x, y, z, t) ∶= (x, y, (1 − t)z).
• H is continuous,
• H(x, y, z, 0) = (x, y, z) ∀(x, y, z) ∈ X ,
• H(x, y, z, 1) = (x, y, 0) ∈ A ∀(x, y, z) ∈ X ,
• if (x, y, z) ∈ A, then z = 0 and thus H(x, y, z, t) = (x, y, 0) ∈ A.
2. Let p, q ∈ R2 , p ≠ q . en S1 ∨ S1 is a deformation retract of R2 ∖ {p, q}.

.p q

3. p, q ∈ R2 , p ≠ q , R2 ∖ {p, q} deform retracts onto the theta space:

(Homeomorph to S1 ∪ (0 × [−1, 1])).

4. Möbius band

29
Algebraic topology

eorem 2.59 Let A be a deformation retract of X , x0 ∈ A, and ι ∶ (A, x0 ) → (X, x0 )

the inclusion map. en ι∗ ∶ (A, x0 ) → π1 (X, x0 ) is an isomorphism.

P Let r ∶ (X, x0 ) → (A, x0 ), r(x) = H(x, 1) en r ○ ι ∶ (A, x0 ) → (A, x0 ) is

the id map 1A and (r ○ ι)∗ = (1A )∗ . Ô⇒ r∗ ○ ι∗ = 1π1 (A,x0 ) .
Furthermore ι∗ ○ r∗ = (ι ○ r)∗ . ι ○ r ∶ (X, x0 ) → (X, x0 ) and there is a homotopy H
between 1X and i ○ r. H(x0 , t) = x0 ∀t ∈ I .
If f ∶ I → X loop , f (0) = f (1) = x0 , then H ○(f ×1I )) ∶ I ×I → X is a path homotopy
between 1X ○ f and (ι ○ r) ○ f . (path homotopy: H ○ (f × 1I )(0, t) = H(x0 , t) = x0
∀t ∈ I , H ○ (f × 1I )(1, t) = H(x0 , t) = x0 ∀t ∈ I ).
So [1X ○ f ] = [(ι ○ r) ○ f ] Ô⇒ (1∗ (f ) = (ι ○ r)∗ (f ).
Ô⇒ 1π1 (X,x0 ) = (ι ○ r)∗ Ô⇒ ι∗ r∗ = 1π1 (X,x0 ) .
Corollary 2.61 If ι ∶ Sn → Rn+1 ∖{0} the inclusion map, x0 ∈ Sn , then ι∗ ∶ π1 (Sn , x0 ) →
π1 (Rn+1 ∖ {0}) is isomorphism.
P De ne H ∶ (Rn+1 ∖ {0}) × I → Rn+1 ∖ {0}, H(x, t) ∶= (1 − t)x + t ∥x∥ x
.

• H is continuous,

• H(x, 0) = x ∀x ∈ Rn+1 ∖ {0} =∶ X ,

• H(x, 1) = x
∥x∥
∈ Sn =∶ A.

• If x ∈ Sn , then H(x, t) = (1 − t)x + tx = x ∀t ∈ I .

H is a deformation retraction of Rn+1 ∖ {0} onto Sn . Now use theorem.

Example 2.63 We can look at example 2.58 again:
m Cor
1. π1 (X) ≃ π1 (A) ≃ π1 (S1 ) ≃ Z.
m
2.,3. “ gure eight“ space is deformaton retraction of R2 ∖{p, q} Ô⇒ π1 (A1 ) ≃ π1 (R2 ∖
{p, q}),
m
“theta” space is deformation retraction of R2 ∖ {p, q} Ô⇒ π1 (A2 ) ≃ π1 (R2 ∖
{p, q}).
Ô⇒ π1 (A1 ) ≃ π1 (A2 ).
Comment: “ gure eight” and “theta” space have isomorphic fundamental group,
but none is a deformation retract of the other (exercise).

De nition 2.64 Let f ∶ X → Y be a continuous map. If there exists a continuous map

g ∶ Y → X such that g ○ f ∶ X → X is homotopic to 1X ∶ X → X and f ○ g ∶ Y → Y
is homotopic to 1Y ∶ Y → Y , then f is called a homotopy equivalence.
g is called homotopy inverse of f . X ,Y are called homotopy equivalent .
If X, Y are homotopy equivalent, we say that X, Y have the same homotopy type.

Check: Homotopy equivalence is an equivalence relation.

Remark 2.65 Homotopy equivalence is more general than deformation retraction.

• Suppose that H ∶ X × I → X is a deformation retraction of X onto A. Let ι ∶ A →

X the inclusion map and r ∶ X → A, r(x) = H(x, 1). en r ○ ι ∶ A → A is the
id map 1A , and H is a homotopy between ι ○ r and 1X .

30
 2.5 e fundamental group of the circle and applications

Lemma 2.66 Let h, k ∶ X → Y homotopic continuous maps with h(x0 ) = y0 , k(x0 ) =

y1 . If H ∶ X × Y → Y is the homotopy between h and k and if a(t) = H(x0 , t), then
k∗ = ̂ a ○ h∗ .

π1 (X, x0 )
h∗
/ π1 (Y, y0 )
MMM
MMM
M
k∗ MMM
̂a
& 
π1 (Y1 , y1 )

P …der Beweis ist noch sehr fehlerha, noch fehlerhaer als gewohnt :P …
Let f ∶ I → X a loop based at x0 .
We want to show that k∗ ([f ]) = ̂a ○ h∗ ([f ])
⇐⇒ [k ○ f ] = ̂
a([h ○ f ])
⇐⇒ [k ○ f ] = [̂
a][h ○ f ][a]
⇐⇒ [a][k ○ f ] = [h ○ f ][a]
⇐⇒ [a(k ○ f )] = [(h ○ f )a].

h(x) = H(x, 0) ∀x ∈ X , so h(f (s)) = H(f (s), 0) = (H ○f0 )(s) where f0 ∶ I → X ×I ,

f0 (s) = (f (s), 0). I.e. h ○ f = H ○ f0 .
k(x) = H(x, 1) ∀x ∈ X , so k(f (s)) = H(f (s), 1) = (H ○f1 )(s) where f1 ∶ I → X ×I ,
f1 (s) = (f (s), 1). I.e. k ○ f = ○f1 .
Consider c ∶ I → X × I , c(t) = (x0 , t). en H(c(t)) = a(t).
F ∶ I × I → X × I , F (s, t) ∶= (f (s), t).
G0 γ1 is homotopic to γ0 B1 , call G the path homotopy between them.
F (β0 (s), 0), s ∈ [0, 12 ]
en F ○G ∶ I×I → X×I , continuous, (F ○G)(s, 0) = F (G(s, 0)) = { =
F (1, γ1 (s)) s ∈ [ 21 , 1]
(f ○ β0 (s), 0) s ∈ [0, 21 ] f (s) s ∈ [0, 12 ]
{ ={ 0 .
(f (1), γ1 (s)) s ∈ [ 2 , 1]
1
(x0 , s) s ∈ [ 12 , 1]
(F ○ G)(s, 1) = ⋅ ⋅ ⋅ = (c ⋅ f1 )(s), (F ○ G)(0, t) = F ((0, 0)) = (x0 , 0), (F ○ G)(1, t) =
F ((1, 1)) = (x0 , 1)
Ô⇒ (F ○ G is a path homotopy from c ⋅ f1 to c ⋅ f1 .
Lastly, H ○ (F ○ G) is a path homotopy from H(f0 ⋅ c) = H(f0 ) ⋅ H(c) = (h ○ f ) ⋅ a to
H(c ⋅ f1 ) = H(c) ⋅ H(f1 ) = a ⋅ (k ○ f )
Corollary 2.68 Let h, k ∶ (X, x0 ) → (Y, y0 ) homotopic continuous maps. If H ∶ X ×
I → Y is the homotopy between h and k and H(x0 , t) = y0 ∀t ∈ [0, 1], then k∗ = h∗ .

eorem 2.69 Let f ∶ X → Y homotopy equivalence with f (x0 ) = y0 . en f∗ ∶

π1 (X, x0 ) → π1 (Y, y0 ) is an isomorphism.

P Let g ∶ Y → X be a homotopy inverse of f , and let g(y0 ) = x1 and f (x1 ) = y1 .

(X, x0 )
f
/ (Y, y0 ) g
/ (X, x1 ) f
/ (Y, y1 )

(fx0 )∗ (fx1 )∗
π1 (X, x0 ) / π1 (Y, y0 ) g∗
/ π1 (X, x1 ) / π1 (Y, y1 )

g ○ fx0 is homotopy to 1X Ô⇒[Lemma] there is a path a ∶ I → X , such that (g ○ fx0 )∗ =

̂
a ○ (1X )∗ = ̂a.

31
Algebraic topology

̂
a is a group isomorphism, so g∗ ○ f∗ is a group isomorphism.
Ô⇒ g∗ is surjective (1).
Similarly (fx1 ○ g)∗ is a group isomorphism.
⇐⇒ (fx1 )∗ ○ g∗ is a group isomorphism.
Ô⇒ g∗ is injective (2).

(1),(2) ↝ g∗ is a group isomorphism, ↝ (fx0 )∗ = (g∗ )−1 ○ ̂

a is a group isomorphism.

eorem 2.71 X, Y have the same homotopy type iff X, Y are homeomorphic to to defor-
mation retracts of a space Z .

[Proof: hard, we omit it]

eorem 2.72 π1 (S1 ∨ S1 ) is not abelian.

P X ∶= S1 ∨ S1 .

B . x0 A

Consider the following covering space

B0

A2

A1
B−2 B−1 B2
. A0

A−1

A−2

p∶X ̃ → X , p maps each Ai homeomorphically onto A ∀i ∈ Z ∖ {0}. and Bi homeo-

morphically to B ∀i ∈ Z ∖ {0}. p maps all tangency points to X0 . p wraps A0 around
A, B0 around B .
p is a covering map (check). Consider f̃ ∶ I → X ̃ , f (s) = (s, 0). ̃g∶I →X ̃ , g(s) = (0, s)
̃
and f = p ○ f , g = p ○ ̃ g.
Also consier f ⋅ g and g ⋅ f and li these two paths to paths f̃ ⋅ g , g̃ ̃ both starting
⋅ f in X
̃ ̃
at (0, 0). f ⋅ g is a path in X that goes from (0, 0) to (1, 0) along x-axis and then once
around B1 . g̃ ̃ that goes from (0, 0) to (0, 1) along the y -axis and then
⋅ f is a path in X
once around A1 .
f̃ ⋅ g(1) ≠ g̃ ⋅ f (1). So f ⋅ g and g ⋅ f are not path-homotopic. [f ⋅ g] ≠ [g ⋅ f ] i.e.
[f ] ⋅ [g] ≠ [g] ⋅ [f ]
Ô⇒ π1 (S1 ∨ S1 ) is not abelian.

32
 2.5 e fundamental group of the circle and applications

eorem 2.74 π1 (X × Y, x0 × y0 ) ≃ π1 (X, x0 ) × π1 (Y, y0 ).

P Consider the projections p ∶ X ×Y → X , x×y ↦ x and q ∶ X ×Y → Y , x×y ↦ y ,

and p∗ ∶ π1 (X × Y, x0 × y0 ) → π1 (X1 , x0 and q∗ ∶ π1 (X × Y, x0 × y0 ) → π1 (Y, y0 ).
De ne Φ ∶ π1 (X × Y, x0 × y0 ) → π1 (X, x0 ) × π1 (Y, y0 ) , [f ] ↦ (p∗ ([f ]), q∗ ([f ])).
Claim: Φ is a group homomorphism.
Proof: Φ([f ][g]) = Φ([f g]) = (p∗ ([f g]), q∗ ([f g])) = ([p ○ (f g)], [g ○ (f g)]) =
([(p ○ f ][p ○ g], [g ○ f ][q ○ g)]).
Φ([f ])Φ([g]) = ([p ○ f ], [q ○ f ])([p ○ g][q ○ g]) = ([p ○ f ][p ○ g], [g ○ f ][q ○ g]).

Claim: Φ is surjective.
Proof: Consider g ∶ I → X Loop based at x0 , and h ∶ I → Y loop based at y0 . De-
ne f ∶ I → X × Y , f (s) = (g(s), h(s)). en f is a loop based at (x0 , y0 ) and
Φ([f ]) = (p∗ ([f ]), q∗ ([f ])) = ([g], [h]).

Claim: Φ is injective.
Proof: Suppose that f ∶ I → X × Y is a loop based at (x0 , y0 ) with Φ([f ]) is the
identity element in π1 (X, x0 ) × π1 (Y, y0 ). en p ○ f is path homotopic to cx0 via a
path homotopy G and q ○ f is path homotopic to cy0 via a path homotopy H . De ne
F ∶ I × I → X × Y by F (s, t) ∶= (G(s, t), H(s, t)). F is a path homotopy from f to
c(x0 ,y0 ) . Ô⇒ [f ] trivial.
Corollary 2.76 Consider the torus T ≃ S1 × S1 . Ô⇒ π1 (T) ≃ Z × Z.

De nition 2.77

1. Let {Xi }i∈I an indexed collection of sets. e disjoint union of these sets is
⊔i∈I Xi = ∪i∈I Xi × {i}.
2. Suppose that Xi is a topological space ∀i ∈ I . e disjoint union topology on
⊔i∈I Xi is the nest topology, such that all of the following maps are continuous:
φi ∶ Xi → ⊔i∈I Xi , x ↦ x × {i}.
Explicitly, U is open in ⊔i∈I Xi iff φ−1 (U ) is open ∀i ∈ I .

3. Let X, Y be topological spaces, A ⊆ X , and f ∶ A → Y continuous. e ad-

junction space X ∪f Y (adjunction of X to Y along f (along A)) is the quotient
space
X ⊔YÒ ,
∼
where a ∼ f (a) ∀a ∈ A. f is called the attaching map.

Remark 2.78 If Xi ∩ Xj = ∅ for any i, j ∈ I , i ≠ j , then ⊔i∈I Xi ≃ ∪i∈I Xi .

Example 2.79 X1 , X2 ⊂ {R, T, S2 }, X1 ⊔ X2 .

De nition 2.80 e wedge sum of X and Y is

X ∨ Y ∶= X ⊔ Y Ò x0 ∼ y0 ,

for x0 ∈ X , y0 ∈ Y .

33
Algebraic topology

If Xi are topological spaces, xi ∈ Xi , i ∈ I , then

∨i∈I Xi = ⊔i∈I Xi Ò {x ∼ x , i, j ∈ I}.

i j

De nition 2.81 Let X, Y n-manifolds. Consider U, V , open discs D2 in X, Y respec-

tively, and f ∶ ∂U → ∂V a homeomorphism. e connected sum of X and Y is

X#Y ∶= (X ∖ U ) ⊔ (Y ∖ V )Ò ∼,

where x ∼ y if x ∈ ∂U , y ∈ ∂V and f (x) = y .

Remark 2.82 Although this construction involves the choice of the discs U, V the result-
ing space is unique up to homeomorphism.

Claim: Y ∶= S1 ∨ S1 is a retract of X ∶= T#T.

P (hier bräuchte es ne Graphik…)
Consider f ∶ X → Y , where f maps the dotted circle to y , and f restricted to S1 ∨S1 ⊆ X
is a homeomorphism h, and f 1 ∶ 1 everywhere else. Further consider r, where r retracts
Y onto S1 ∨ S1 by mapping each cross-sectional circle to the point where it intersets
S1 ∨ S1 . Use h−1 to map S1 ∨ S1 in Y to S1 ∨ S1 in X .
↝ retraction of T#T onto S1 ∨ S1 .
Corollary 2.84 π1 (T#T) is not abelian.
P If ι ∶ S1 ∨S1 → T#T the inclusion map, then ι∗ is injective (follows from claim).

ι∗ ∶ π1 (S1 ∨ S1 ) → π1 (T#T).

So π1 (T#T) is not abelian.

Corollary 2.86 T, T#T are not homotopy equivalent.
P obvious.
Lemma 2.88 If P = (0, . . . , 0, 1) ∈ Sn ⊂ Rn+1 , then Sn ∖ {P } is homeomorphic to Rn .
P De ne f ∶ Sn ∖ {P } → Rn (stereographic projection) , where (x1 , . . . , xn+1 ) ↦
1
1−xn+1
(x1 , . . . , xn ).
f is obviously continuous and furthermore g ∶ Rn → Sn ∖ {P } where (y1 , . . . , yn ) ↦
( 1+∥y∥
2
2 y1 , . . . , 1+∥y∥2 yn , 1− 1+∥y∥2 ) is continuous and f ○g = 1Rn and g○f = idSn ∖{P } .
2 2

Proposition 2.90 π1 (Sn ) = 0 ∀n ∈ N, n ≥ 2.

P Consider f ∶ I → Sn a loop based at x0 ∈ Sn . Let x ∈ Sn ∖ {x0 } and B a small
open ball in Sn ∖ {x0 } with x ∈ X .
en f −1 (B) ⊆ (0, 1) and f 1− (B) is open., So f −1 (B) = ∪i∈I (ai , bi ).
f −1 ({x}) is a closed subset of the compact space I , i.e. f −1 ({x}) is compact.
f −1 ({x}) ⊆ f −1 (B) = ∪i∈I (ai , bi ), Ô⇒ ∃(a1 , b1 ), . . . , (an , bn ) such that f −1 ({x}) ⊆
(a1 , b1 ) ∪ ⋅ ⋅ ⋅ ∪ (an , bn ).
Consider fi = f∣[ai ,bi ] . en f ((ai , bi )) ⊆ B and f ([ai , bi ]) = f ((ai , bi )) ⊆ f ((ai , bi )) ⊆
B.
f (ai ), f (bi ) ∈ ∂B .
Since n ≥ 2, choose gi ∶ I → ∂B path, such that gi (0) = f (ai ) and gi (1) = f (bi ).
Furthermore, B is convex, so π1 (B) = 0. Ô⇒ fi ≃ gi .

34
 2.6 Seifert von Kampen eorem

Repeating this process for all intervals [ai , bi ], i ∈ {1, . . . , n}, we obtain a loop g ∶ I →
Sn , homotopic to f , and with the property that g(I) ⊆ Sn ∖ {x}, [f ] = [g].
e lemma implies that [g] = [cx0 ], so [f ] = [cx0 ] and π1 (Sn , x0 ) = 0.
Corollary 2.92 R2 and Rn are not homeomorphic for any n ≠ 2.
P Suppose that f ∶ R2 → Rn a homeomorphism (n ≠ 2). en

• n = 1: R2 ∖ {0} ≃ R ∖ {f (0)}, but R2 ∖ {0} is path connected, but R ∖ {f (0} is

not.

• Recall π1 (R2 ∖ {P }) ≃ π1 (S1 ) ≃ Z, π1 (Rn ∖ {f (p)}) ≃ π1 (Sn−1 ) = 0 for n ≥ 3.

f∣R2 ∖{P } ∶ R2 ∖ {P } → Rn ∖ {f (P )} homeomorphism, contradiction.

Recall:
space π1 using
X ⊆ Rn convex, x0 ∈ X 0 de nition
S1 Z covering spaces
R2 ∖ {(0, 0)} Z deformation retracts
R3 ∖ {z -axis} Z deformation retracts
S1 ∨ S1 π1 (S1 ∨ S1 ) ≃ π1 (R2 ∖ {p, q}) deformation retracts
θ-space π1 (θ) ≃ π1 (R2 ∖ {p, q}) deformation retracts
Möbius-band M Z deformation retracts
S1 ∨ S1 not abelian covering spaces
T#T not abelian retraction
Sn , n ∈ N, n ≥ 2 0
T Z×Z

2.6. Seifert von Kampen eorem

2.6.1. Direct sums of abelian groups

De nition 2.94 Let G be an abelian group and {Ga }a∈A a family of subgroups of G.
We say that the groups Ga generates G if every element x of G can be written as a nite
sum of elements of the groups Ga .
Since G is abelian, we can always write this sum in the form X = Xa1 + ⋅ ⋅ ⋅ + Xan ,
Xa ∈ Ga , ai ≠ aj , if i ≠ j .
In this case we oen write x = ∑a∈A Xa .
If the groups Ga generate G, then G is called the sum of Ga and we write G = ∑a∈A Ga
or G = G1 + ⋅ ⋅ ⋅ + Gn .
If Ga generates G and ∀x ∈ G ∃!A-tuple (xa )a∈A with xa = 0, for all but nitely
many a in A, and G = ∑a∈A Xa then G is called the direct sum of Ga , and we write
G = ⊕a∈A Ga , or G = G1 ⊕ ⋅ ⋅ ⋅ ⊕ Gn .

Example 2.95

1. R2 , G1 = {(x, 0) ∣ x ∈ R}, G2 = {(0, y) ∣ y ∈ R}, R2 = G1 ⊕ G2 .

2. R∞ : the set of all sequences of real numbers that are eventually zero. R∞ is a group
under coordinate addition. Gn = {(0, . . . , 0, xn , 0, . . . ∣ xn ∈ R} are subgroups of
R∞ ∀n ∈ N and R∞ = ⊕n∈N Gn .

35
Algebraic topology

Lemma 2.96 (Extension condition)

Let G be an abelian group and {Ga }a∈A be a family of subgroups. If G = ⊕a∈A Ga , then
G stais es the following condition:

(∗) If H is any abelian group and ha ∶ Ga → H is a family of homomorphisms, then

there exists a homomorphism h ∶ G → H , such that h∣Ga = ha ∀a ∈ A.

Furthermore, this is unique. Conversely if Ga generate G and (∗) holds, then G =

⊕a∈A Ga .
P Suppose that G = ⊕a∈A Ga . Consider H abelian group., and ha ∶ Ga → H family
of homomorphisms. De ne h ∶ G → H in the following way: If x ∈ G, then x = ∑n i=1 xai
and h(x) = ∑n i=1 hai (Xai ). h is well-de ned, because there is a unique way of writing x
as ∑ni=1 xai . h is a homomorphism and h∣Ga = ha .
Suppose that h′ ∶ G → H with h′Ga = ha , ∀a ∈ A. en h′ (x) = h′ (∑n i=1 xai ) =
∑i=1 h′ (xai ) = ∑i=1 hai (xai ) = h(x).
n n

Converse: Suppose that x = ∑a∈A xa = ∑a∈A ya for some x ∈ G. Choose b ∈ A and

let H = Gb . De ne ha ∶ Ga → H by ha = 1Gb if a = b, and ha trivial if a ≠ b. Let
h ∶ G → H be the extension of the homomorphisms ha . en h(x) = ∑a∈A h(xa ) =
∑a∈A ha (xa ) = Xb , h(x) = ∑a∈A h(ya ) = ∑a∈A ha (ya ) = Yb .
Ô⇒ Xb = Yb , b ∈ A,
Ô⇒ Xa = Ya ∀a ∈ A.

Corollary 2.98 Let G = G1 ⊕ G2 and G1 = ⊕a∈A Ha , and G2 = ⊕b∈B Hb , where the

index sets A,B are disjoint. en G = ⊕γ∈A∪B Hγ .
[Proof: Problem Set 7, Exercise 4.a)]

Corollary 2.99 If G = G1 ⊕ G2 , then GÒ G ≃ G1 .

2
[Proof: Problem Set 7, Exercise 4.b)]

Given a family of abelian groups {Ga }a∈A , nd a group G that contains subgroups G′a ≃
Ga ∀a ∈ A and G = ⊕a∈A G′a .

De nition 2.100 Let {Ga }a∈A be a family of abelian groups, G be an abelian group,
and ia ∶ Ga → G a family of monomorphisms such that G = ⊕a∈A ιa (Ga ). en G is
called the external direct sum of the groups Ga , relative to monomorphisms ia .

Notation: ≤: Subgroup.

eorem 2.101 If {Ga }a∈A is family of abelian groups, then there exists an abelian group
G, and a family of monomorphisms ia ∶ Ga → G, such that G = ⊕a∈A ia (Ga ).

P Consider the product ∏a∈A Ga (with coordinate wise addition), which is an
abelian group. Let G = {(xa )a∈A ∣ xa = 0 for all but nitely many indices in A} ≤ ∏ Ga .
Given β ∈ A, de ne iβ ∶ Gβ → G, by letting iβ (x) the tuple with x as its β -th coordinate
and 0 in every other coordinate. iβ is a monomorphism (follows from the de nition),
and if x ∈ G, then x has nitely many non zero coordinates, i.e. x = (x1 , . . . , xn , 0, . . . )
and x = i1 (x1 ) + ⋅ ⋅ ⋅ + in (xn ). Furthermore, this expression is unique.

36
 2.7 Free abelian groups

Lemma 2.103 (extension condition)

Let {Ga }a∈A be a family of abelian groups, G be an abelian group, ia ∶ Ga → G family
of homomorphisms. If ia is a monomorphism ∀a ∈ A, and G = ⊕a∈A ia (Ga ), then the
following condition holds:
(∗) Given any abelian group H and any family of homomorphisms, hA ∶ Ga → H ,
then there exists a homomorphism h ∶ G → H , with h ○ ia = ha ∀a ∈ A.
Futhermore h is unique.
Conversely if ia (Ga ) generate G and (∗) holds, then ia is a monomorphism ∀a ∈ A,
and G = ⊕a∈A ia (Ga ).
[Proof: Problem Set 7, Exercise 4.c)]

eorem 2.104 (Uniqueness of direct sums)

Let {Ga }a∈A be a family of abelian groups, G, G′ be abelian groups, ia ∶ Ga → G,
i′a ∶ Ga → G′ families of monomorphisms, such that G = ⊕a∈A ia (Ga ) and G′ =
⊕a∈A i′a (Ga ). en ∃!φ ∶ G → G′ isomorphism, such that φ ○ ia = i′a ∀a ∈ A.

P
• Lemma 2.103 (H = G′ ) Ô⇒ ∃!φ ∶ G → G′ , such that φ ○ ia = i′a ∀a ∈ A.
• Lemma 2.103 (H = G) Ô⇒ ∃!ψ ∶ G′ → G, such that ψ ○ i′a = ia ∀a ∈ A.
Now (Lemma 2.103) ψ ○ φ ∶ G → G satis es (ψ ○ φ) ○ ia = ia ∀a ∈ A.
Ô⇒ ψ ○ φ = 1G , analogously φ ○ ψ = 1G′ .

2.7. Free abelian groups

De nition 2.106 Let G be an abelian group and {xa }a∈A a family of elements in G.
Let Ga = ⟨xa ⟩, i.e. the subgroup generated by xa . If Ga generate G, we also say that
xa generate G.
If Ga = ⟨xa ⟩ is in nite cyclic ∀a ∈ A, and G = ⊕a∈A ⟨xa ⟩, then G is called a free abelian
group with basis {xa }a∈A .

Lemma 2.107 (extension condition)

Let G be an abelian group and {xa }a∈A a family of elements of G that generates G. en
G is a free abelian group with basis {xa }a∈A iff for any abelian group H and any family
{ya }a∈A of elements in H , there is a homomorphism h ∶ G → H , such that h(xa ) = ya
∀a ∈ A. In such a case h is unique.
P
“⇒” Given H , {ya }a∈A de ne homomorphisms ha ∶ Ga → H , with ha (xa ) = ya .
(Ga is cyclic, so the condition h(xa ) = ya determines uniquely a homomorphism
ha ∶ Ga → H ). By lemma 2.96 ∃h ∶ G → H homomorphism, with h∣Ga = ha
∀a ∈ A. Furthermore h is unique.
“⇐” Suppose that for some β ∈ A, ⟨xβ ⟩ is nite. en for H = Z and ya = 1 ∀a ∈ A,
there is no homomorphism h ∶ G → H with h(xa ) = 1 ∀a ∈ A.
By lemma 2.96, we get that G = ⊕a∈A Ga , i.e. G is a free abelian group with basis
{xa }a∈A .

37
Algebraic topology

Remark 2.109 Let G be free abelian with basis {x1 , . . . , xn }. G = ⟨x1 ⟩ ⊕ ⋅ ⋅ ⋅ ⊕ ⟨xn ⟩ ≃
Z ⊕ ⋅ ⋅ ⋅ ⊕ Z.
Proposition 2.110 If G is free abelian with basis {x1 , . . . , xn }, then n is uniquely deter-
mined by G.
P G ≃ Z ⊕ ⋅ ⋅ ⋅ ⊕ Z ≥ 2Z ⊕ ⋅ ⋅ ⋅ ⊕ 2Z,
Z ⊕ ⋅ ⋅ ⋅ ⊕ ZÒ
2Z ⊕ ⋅ ⋅ ⋅ ⊕ 2Z ≃ Z2 ⊕ ⋅ ⋅ ⋅ ⊕ Z2 .
∣Z2 ⊕ ⋅ ⋅ ⋅ ⊕ Z2 ∣ = 2n .

De nition 2.112 If G is a free abelian group with basis {x1 , . . . , xn }, then n is called
the rank of G.

2.8. Free products of groups

De nition 2.113 Let G be a group and {Ga }a∈A be a family of subgroups of G. We

say that Ga generate G, if ∀x ∈ G ∃(x1 , . . . , xn ) with xi element of some Ga ∀i ∈
{1, . . . , n}, and x = x1 . . . xn .
Such a sequence is called a word of length n in the groups Ga that represents x.

If xi , xi+1 ∈ Ga for some a ∈ A, we group them together to obtain the word

(x1 , x2 , . . . , xi−1 , xi xi+1 , xi+2 , . . . , xn )

of length n − 1. If xi = 1, we delete xi from our sequence.

Repeating these reduction operations, we obtain (y1 , . . . , ym ) representing x, such that
no group Ga contains yi and yi+1 and, yi ≠ 1 ∀i ∈ {1, . . . , m}. We call (y1 , . . . , ym ) a
reduced word.
Convention: ∅ is a reduced word representing 1 ∈ G.

De nition 2.114 Let G be a group and {Ga }a∈A a family of subgroup that generates
G and Ga ∩ Gb = {1} ∀a, b ∈ A, a ≠ b. If ∀x ∈ G there is only one reduced word in
the group {Ga }a∈A , that represents that element, then G is called the free product of
the groups {Ga }a∈A .
G = ∗a∈A Ga , or G = G1 ∗ ⋅ ⋅ ⋅ ∗ Gn .

Example 2.115 G1 = {1, x1 }, G2 = {1, x2 }. x ∈ G = G1 ∗ G2 . Elements of G are for

example x1 , x2 , x1 x2 , x2 x1 , x1 x2 x1 , x2 x1 x2 , . . .
Example 2.116 G = {φ ∶ {0, 1, 2} → {0, 1, 2} ∣ φ bijection}. φ1 ∶ {0, 1, 2} → {0, 1, 2},
φ1 (2) = 2, φ1 (1) = 0, φ1 (0) = 1
φ2 ∶ {0, 1, 2} → {0, 1, 2}, φ2 (0) = 0, φ2 (1) = 2, φ2 (2) = 1.

∣⟨φ1 ⟩∣ = 2, ∣⟨φ2 ⟩∣ = 2,
⟨φ1 ⟩, ⟨φ2 ⟩ generate G, but B ≠ ⟨φ1 ⟩ ∗ ⟨φ2 ⟩.
Indeed the reduced words (φ1 , φ2 , φ1 ) and (φ2 , φ1 , φ2 ) represent the same element in
G.

38
 2.8 Free products of groups

Claim: Suppose that Ga generate G. Ga ∩ Gb = {1} ∀a, b ∈ A with a ≠ b. If the

representation of 1 by the empty word is unique then G = ∗a∈A Ga .
P Let x ∈ G and suppose that (x1 , . . . , xn ), (y1 , . . . , ym ) are two reduced words
representing x, where xi ∈ Gai , yi ∈ Gbj , ai ∈ A, i ∈ {1, . . . , n}, bj ∈ A, j ∈ {1, . . . , m}.
−1
x1 . . . xn = x = y1 . . . ym Ô⇒ x1 . . . xn ym . . . y1−1 .
Ô⇒ an = bm (i.e. xn , ym ∈ Gan = Gbm ).
−1 −1
We obtain (x1 , . . . , xn−1 , xn ym , yn−1 , . . . , y1−1 ).
−1
Now we must have xn ym = 1 Ô⇒ xn = ym .
Continue this process, to conclue that n = m and xi = yi , ∀i ∈ {1, . . . , n}.
Lemma 2.118 (extension condition / universal mapping property) Let G be a group,
{Ga }a∈A be a family of subgroups. If G = ∗a∈A Ga , then it satis es the following ex-
tension condition: Given any group H and any family of homomorphisms ha ∶ Ga → H
thre exists a homomorphism h ∶ G → H , such that h∣Ga = hA . In addition h is a unique.
P De ne h ∶ G → H as follows:

• h(1) = 1,

• If x ∈ G, x ≠ 1, let (x1 , . . . , xn ) be the (unique) reduced word representing x. Set

h(x) = ha1 (x1 ) . . . han (xn ), where ai is the index for which xi ∈ Gai .

en

• h is well-de ned,

• h∣Ga = ha ∀a ∈ A,

• h is a group homomorphism (problem sheet 8).

In fact, uniqueness of h follows from the fact that h must statisfy h(x) = h(x1 , . . . , xn ) =
h(x1 ) . . . h(xn ) = ha1 (x1 ) . . . han (xn ).

External free product

De nition 2.120 Let {Ga }a∈A be a family of groups. Suppose that G s a group and
{ia ∶ Ga → G}a∈A a family of monomorphisms with G = ∗a∈A ia (Ga ), then G is
called the external free product of the groups Ga relative to the onomorphisms ia .

Does such a group G exist?

eorem 2.121 Let {Ga }a∈A be a family of groups. ere exists a group G and a family
of monomorphisms ia ∶ Ga → G, such that G = ∗a∈A ia (Ga ).

Lemma 2.122 (extension condition)

{Ga }a∈A family of groups, G group. ia ∶ Ga → G family of homomorphisms. If ia is a
monomorphism ∀a ∈ A and G = ∗a∈A ia (Ga ), then the following extension condition
holds: Given any group H and a family of homomorphisms ha ∶ Ga → H , ∃ ∶ G → H
homomorphism such that h ○ ia = ha ∀a ∈ A. In addition h is unique.
[Proof ommited]

39
Algebraic topology

eorem 2.123 (uniqueness of free product)

Let {Ga }a∈A be a family of gropus, G, G′ be groups, {ia ∶ Ga → G}a∈A , {i′a ∶ Ga →
G′ }a∈A families of monomorphism such that {ia (Ga )}a∈A , {i′a (Ga )} generate G, G′
repsectively.
If G, G′ satisfy the extension condition, then ∃! isomorphism φ ∶ G → G′ such that φ○ia =
i′a ∀a ∈ A.

[Proof: analogous to the proof of uniqueness of direct sums (problem sheet 8)]

Lemma 2.124 Let {Ga }a∈A be a family of groups, G be a gropu and ia ∶ Ga → G be

a family of homomorphisms. If ia (Ga ) generate G, and the extension condition holds,
then ia is a monoomorphism ∀a ∈ A, and G = ∗a∈A Ga .
P Consider b ∈ A. Set H = Gb , ha ∶ Ga → H the identity homomorphism 1Gb if
a = b, and ha ∶ Ga → H the trivial homomorphism, if a ≠ b.
Let h ∶ G → H be the homomorphism given by the extension condition.
en h ○ ib = hb Ô⇒ h ○ ib = 1Gb . Ô⇒ ib is 1:1.
I.e. ia is a monomorphism ∀a ∈ A.
Existence theorem Ô⇒ there exists a group G′ and {i′a ∶ Ga → G′ }a∈A a family of
monomorphisms such that G′ = ∗a∈A i′a (Ga ). G, G′ both satisfy the extension condition
and are generated by {ia (Ga )}a∈A , respectively {i′a (Ga )}a∈A .
Uniqueness theorem Ô⇒ there is an isomorphism φ ∶ G → G′ with φ ○ ia = i′a . G′ =
∗a∈A i′a (Ga ) .
Ô⇒ G = ∗a∈A ia (Ga ).
Corollary 2.126 If G = G1 ∗ G2 , , G1 = ∗a∈A Ha , G2 = ∗b∈B Hb , A ∩ B = ∅, then
G = ∗γ∈A∪B Hγ .
N normal subgroup of G: N ⊴ G.

eorem 2.127 Let G = G1 ∗ G2 , Ni ⊴ Gi , i ∈ {1, 2}. If N is the smallest normal

subgroup of G that contains N1 , and N ∶ , then

GÒ ≃ (G1 Ò ) ∗ (G2 Ò ) .
N N1 N2

P Consider the inclusion homomorphism i ∶ G1 → G1 ∗ G2 , i′ ∶ G2 → G1 ∗ G2 ,

and the projection homomorphism p ∶ G1 ∗ G2 → G1 ∗ G2/N , g ↦ gN .
Let n1 ∈ N1 . en (p ○ i)(n1 ) = n1 N = N . is implies that N1 ≤ ker(p ○ i) and p ○ i
induces (p ○ i)′ ∶ G1 Ò N1 → G1 ∗ G2 Ò N .
Analogously, (p ○ i′ )′ ; G2 Ò N2 → G1 ∗ G2 Ò N .
We will apply lemma 2.124 for the homomorphisms i1 , i2 .

• Check that the extension condition holds.

Let h1 ∶ G1 /N1 → H1 , h2 ∶ G2 /N2 → H arbitrary homomorphisms and p1 ∶
G1 → G1 /N1 , p2 ; G2 → G2 /N2 the projection homomorphism. en h1 ○ p1 ∶
G → H , h2 ○ p2 ∶ G2 → H homomorphisms. e extension condition for G1 ∗ G2
implies that ∃h ∶ G1 ∗G2 → H , with hGi = hi ○pi , i ∈ {1, 2}. If ni ∈ Ni , i ∈ {1, 2},
then h(ni ) = 1H . I.e. Ni ≤ ker h ∀i ∈ {1, 2}.
Ô⇒ N ≤ ker h.
is implies that h ∶ G1 ∗G2 → H induces a homomorphism h′ ∶ G1 ∗G2 /N → H

40
 2.8 Free products of groups

with h′ ○ i1 = h1 , and h′ ○ i2 = h2 .
(h′ ○ i1 (g1 N1 ) = h′ ○ (p ○ i)′ (g1 N1 ) = h′ ((p ○ i)(g1 )) = h′ (p(i(g1 ))) =
h′ (p(g1 )) = h′ (g1 N ) = h(g1 ) = (h1 ○ p1 )(g1 ) = h1 (g1 N1 )).
• Check that i1 (G1 /N1 ), i2 (G2 /N2 ) generate G1 ∗ G2 /N .
i1 (G1 /N1 ) = (p ○ i)′ (G1 /N1 ) = (p ○ i)(G1 ) = p(G1 ) = G1 /N , and similarly
i2 (G2 /N2 ) = G2 /N2 .
(Lemma 2.124) Ô⇒ i1 , i2 are monomorphisms and (G1 ∗ G2 )Ò N = i1 (G1 /N1 )∗
i2 (G2 /N2 ).
Corollary 2.129 If N is the smallest normal subgroup of G1 ∗ G2 that contains G1 , then
G1 ∗ G2 Ò ≃ G2 .
N
P N1 = G1 , N2 = {1G2 }.
G group, {Ga }a∈A family of groups, {ia ∶ Ga → G}a∈A family of homomorphisms.
en the following statements are equivalent:
1. (ia is a monomorphism ∀a ∈ A and) G = ∗a∈A ia (Ga ) .

2. Given any group H and any family of homomorphisms ha ∶ Ga → H , ∃! homo-

morphism h ∶ G → H with h ○ ia = ha , ∀a ∈ A.

2.8.1. Free groups

De nition 2.131 Let G be a group and {xa }a∈A a family of G with ⟨xa ⟩ in nite cylclic
∀a ∈ A. If G = ∗a∈A ⟨xa ⟩, then G is called a free group and {xa }a∈A is called a system
of free generators.
In this case, if x ∈ G ∖ {1}, then x can be written uniquely as x = (xai )n1 ⋅ (xak )nk
where ai ≠ ai ≠ ai+1 , ∀i ∈ {1, . . . , k − 1} and n1 ∈ Z ∖ {0} ∀i ∈ {1, . . . , k}.

Free groups are characterized by the following:

Lemma 2.132 (exension condition)
Let G be a group and {xa }a∈A be a family of elmeents in G. If G is a free group with
{xa }a∈A system of free generators, then G satisi es
(∗) Given any group H and any family {ya }a∈A of elements of H there exists a homo-
morphism h ∶ G → H with h(xa ) = ya ∀a ∈ A.
In addition h is unique. Conversely, if {xa }a∈A generates G, and (∗) holds, then G is a
free group with system of free generators {xa }a∈A .
P Lemma 2.118.
For the converse, consider b ∈ A. en there exists a homomorphism hb ∶ G → Z with
hb (xb ) = 1, and hb (xa ) = 0 ∀a ∈ A, a ≠ b. (H = Z, {ya }a∈A = {0, 1}). is implies,
that ⟨xb ⟩ is in nite cyclic. erefore ⟨xa ⟩ in nite cyclic ∀a ∈ A.
en lemma 2.124.

eorem 2.134 Let G = G1 ∗ G2 , G1 , G2 are free groups with {xa }a∈A , {xa }a∈B re-
sepective free sytems of generators with A ∩ B = ∅, then G is a fre group with {xa }a∈A∪B
free system of generators.

41
Algebraic topology

De nition 2.135 Let {xa }a∈A be an arbitrary indexed family and Ga be the set of all
a , n ∈ Z. De ne xa ⋅ xa = xa
symbols xn n m n+m
. is makes our set Ga a group.
e external free product of the groups Ga is called the free group on the elements xa .

a of Ga with their images ia (xa ) in G.

Notation: We will identify the elements xn n

Relation between free groups and free abelian groups.

eorem 2.136 If G is a free group and {xa }a∈A is a system of free generators, then
GÒ
[G, G] is the free abelian group with basis the {[xa ]}a∈A (where [xa ] is the coset
xa [G, G]).

P We wil show, that 1. {[xa ]}a∈A generate GÒ [G, G] and 2. if H is an abelian
group and {ya }a∈A is a family of elements in H , then there is a homomorphism h′ ∶
GÒ ′
[G, G] → H , with h ([xa ]) = ya ∀a ∈ A. Lemma 2.107 will them directly imply
the proof.
1. {xa }a∈A generate G ↝ {[xa ]}a∈A generate GÒ [G, G].

2. Consider an abelian group H and {ya }a∈A a family of elements in H . G free Ô⇒

∃h ∶ G → H homomorphism with h(xa ) = ya ∀a ∈ A.

H abelian Ô⇒ [G, G] ≤ ker h. Ô⇒ there exists an induced morphism h′ ∶

GÒ ′
[G, G] → H with h ([xa ]) = ya ∀a ∈ A.
Corollary 2.138 If G is a free group with n free generators, then an system of free gen-
erators of G has n elements.
P G Ò [G, G] free abelian group with basis {[xi ]}i=1,...,n .

De nition 2.140 If G is a free group, then

rank(G) ∶= rank(G Ò [G, G]).

Properties:
G free abelian group G free group
H ≤ G → H , is free abelian H ≤ G ↝ H is free.
rank(G) = n, and H ≤ G Ô⇒ Consider the free group G and two ele-
rank(H) ≤ n. ments {x1 , x2 }, [G, G] ≤ G is freely gen-
erated by [xn 1 , x2 ], m, n ∈ Z ∖ {0}. In
n

fact a free group with rank greater than 1

has subgroups of all countable orders.

Groups up to isomorphism:
G free abelian, then the isomorphism type is determined by the cardinality of the basis.
G free, then the isomorphism type is determined by the cardinality of systems of gener-
ators.
G nitely generated abelian group, then G ≃ H ⊕ T , where H is free-ablian and T is
torsion subgroup, so the isomorphismtype is de ned by the rank of H and elementary
divisors.

42
 2.8 Free products of groups

Group representation Let G be a group and {xa }a∈A be a family that generates G.
Consider the free group F on {xa }a∈A . De ne a epimorhpishm h ∶ F → G, xa ↦ xa .
G ≃ F Ò ker h.

N ∶= ker h. N is called the relation subgroup. If n ∈ N , then n is called a relation on F .

N ⊴ F and we can specify N by specifying a family {rb }b∈B of elements in F such that
N is the smalest normal subgroup containing {rb }b∈B . (Exercise 4, problem sheet 8).
Such a family {rb }b∈B is called a complete set of relations.

De nition 2.141 If G is a group, a presentation of G is family of generators {xa }a∈A

of generators for G together with a complete set of relations {rb }b∈B for G. If both
{xa }a∈A and {rb }b∈B are nite, then G is called nitely presented. If {xa }a∈A is nite,
then G is called nitely generated.

A presentation determines G uniquelly up to isomorphism. Given two different presen-

tations, it is hard to decide whether they determine isomorphic groups (Unsolvability of
the isomorphism problem for groups).
Example 2.142
1. G cyclic group of order n. Let x ∈ G be generator. ⟨x ∣ xn ⟩.

2. D2n dihedral group. ⟨r, s ∣ rn , s2 , (rs)2 ⟩.

3. G ∶= ⟨a, b ∣ a3 b−2 ⟩, G′ ∶= ⟨x, y ∣ xyxy −1 x−1 y −1 ⟩. Is G ≃ G′ ?

F = ⟨a, b⟩, F ′ = ⟨x, y⟩. De ne h ∶ F → G′ group homomorphism with h(a) = xy
and h(b) = xyx.

h(a3 b−2 = (h(a))3 (h(b))−2 = xyxyxyx−1 y −1 x−1 y −1 x−1 = xyxxyxx−1 y −1 x−1 x−1 y −1 x−1 =
1G .
If N is the smallest normal subgroup containing a3 b−2 then (exercise 4, problem
sheet 8) N = ⟨ga3 b−2 g −1 , g ∈ F ⟩.
en N ≤ ker h.
h′ ∶ F Ò N → G′ , i.e. h′ ∶ G → G′ .
De ne g ∶ F ′ → G group homomorphism, with g(x) ∶= a−1 b, and g(y) ∶= b−1 a2 .
Do the same as above, then, g induces a homomorphism g ′ ∶ G′ → G.
Check that h′ ○ g = 1G′ ∶ G′ → G′ and g ′ ○ h′ = 1G ∶ G → G.

2.8.2. e Seifert-van Kampen theorem

Let X be a topological space, x0 ∈ X , and X = ∪α∈A Aα , where Aα open path-connected
subset of X with x0 ∈ Aα ∀α ∈ A. en we can consider the homomorphism jα ∶
π1 (Aα ) → π1 (X) induced by the inclusion Aα ↪ X (for brevity we write π1 (Aα )
instead of π1 (Aα , x0 ) and we write π1 (X) instead of π1 (X, x0 )). ese homomomor-
phism extend to a homomorphism Φ ∶ ∗α∈A π1 (Uα ) → π1 (X) (extension condition for
free product).
e Seifert-van Kampen theorem tells us, that Φ is “very oen” surjective, but in general
not injective.
Consider iαβ ∶ π1 (Aα ∩Aβ ) → π1 (Aα ) the homomorphism induced by Aα ∩Aβ ↪ Aα .
en jα ○ iαβ = jβ ○ iβα . Let ω ∈ π1 (Aα ∩ Aβ ). en Φ(iαβ (ω) ⋅ (iβα (ω))−1 ) =

43
Algebraic topology

Φ(iαβ (ω))(Φ(iβα (ω)))−1 = (jα ○ iαβ )(w) ⋅ (jβ ○ iβα (w))−1 = 1π1 (X) . I.e. iαβ (ω) ⋅
i−1
βα ∈ ker Φ.

eorem 2.143 (Seifert-van Kampen)

If X = ∪α∈A Aα . Aα open and path connected with x0 ∈ Aα , ∀α ∈ A, and Aα ∩ Aβ is
path connected for any pair {α, β} in A, then Φ is surjective. If furthermore Aα ∩ Aβ ∩ Aγ
is path-connected, for any triple {α, β, γ} in A, then ker Φ = N , where N is the small-
est normal subgroup containing all the elements of the form iαβ (ω)i−1 βα , so Φ induces an
isomorphism ∗α π1 (Aα )Ò N ≃ π1 (X).

P

• surjectivity of Φ ∶ ∗α π1 (Aα ) → π1 (X).

Let f be a loop in π1 (X) based at x0 f is continuous and Aα is open ∀α ∈ A, so
∀s ∈ I there is an open interval VS ⊂ I with s ∈ Vs and f (V s ) ⊆ Aα for some
α ∈ A. We have I = ∪s∈I Vs and I is compact, so it is covered by nitely many of the
Vs . en endpoints of these intervalls de ne a partition 0 = s0 < s1 < s2 < ⋅ ⋅ ⋅ <
sm = 1 of I such that for each i there is some subset Aα with f ([si−1 , si ] ⊆ Aα .
Denote the Aα containing f ([si−1 , si ]) byAi and the path f∣[si−1 ,si ] = fi . en
f = f1 ⋅ ⋅ ⋅ ⋅ ⋅ fm , fi path in Ai .
Now Ai ∩ Ai+1 is path-connected, so we can choose a path gi in Ai ∩ Ai+1 from
x0 to f (si ).
Consider the loop (f1 ⋅ g1 ) ⋅ (g1 ⋅ f2 ⋅ g2 )⋯(gm−1 ⋅ fm ), which is homotopic to f .
is is a composition of loops based at x0 each lying in a single Ai . Hence [f ] =
[f1 ⋅ g1 ] ⋅ [g1 ⋅ f2 ⋅ g2 ]⋯[gm−1 ⋅ fm ] ∈ imΦ.

Notation: A factorization of [f ] ∈ π1 (X) is a fromal product [f1 ]⋯[fk ], where

each fi is a loop in some Aα based at x0 , and f is homotopic to f1 ⋯fk .
A factorization of [f ] is a word in ∗α∈A π1 (Aα ), possibly unreduced, with Φ([f1 ]⋯[fk ]) =
[f ].
Two factorizatons of [f ] are equivalent, if they are related by a sequence of the fol-
lowing two kinds of moves or their inverses: 1. Combine adjacent terms [fi ]⋅[fi+1 ]
into [fi ⋅ fi+1 ] if [fi ], [fi+1 ] lie in the same π1 (Aα ), and 2. regard the term
[fi ] ∈ π1 (Aα ) as lying in π1 (Aβ ), if fi is a loop in Aα ∩ Aβ .
Let Q ∶= ∗α∈A π1 (Aα )Ò N . e rst mmove, doesn’t change the element of ∗α∈A π1 (Aα )
de ned by the factorization. e second move doesn’t change the image of this el-
ment in Q.

• ker Φ.
Recall that N ≤ ker Φ. If I show that any two factorizations of f are equivalent,
this will imply that Φ′ ∶ Q → π1 (X) is injective, hence ker Φ = N .
Let [f1 ]⋯[fk ] and [f1′ ]⋯[fℓ′ ] be two factorizations of [f ] and let F ∶ I × I → X a
homotopy from f1 ⋯fk to f1′ ⋯fℓ′ . en there exists partions 0 = s0 < ⋅ ⋅ ⋅ < sm = 1
and 0 = t0 < ⋅ ⋅ ⋅ < tn = 1, such that each rectangle Rij = [si−1 , si ] × [tj−1 , tj ] is
mapped by f into a single Aα , which we label by Aij . ese partitions are obtained
by

44
 2.8 Free products of groups

– covering I × I with nitely many rectangles [a, b] × [c, d] each mapped into
a single Aα ,
– partioning I × I by the union of all horizontal and vertical lines containing
edgese of the rectangle above.
We can also assume that the s-partition subdivides the partitions given by the prod-
ucts f1 ⋯fk , f1′ ⋯fℓ′ . We may also perturb the vertical sides of the rectangles Rij ,
so that each point in I × I lies in at most three Rij ’s. Relable the new rectangles
Rmn ordering them as in the picture

R9 R10 R11 R12

R5 R6 R7 R8

R1 R2 R3 R4
.

Consider γr , r ∈ {0, 1, . . . , mn}, the path in I × I from the le edge {0} × I to
te right edge {1} × I that separates R1 , . . . , Rr from Rr+1 , . . . , Rmn . We will call
the corners of the rectangles vertices. For each vertex v with F (v) ≠ x0 let gv be
a path from x0 to F (v) in the intersection of the two or three Aij ’s correspond-
ing to the rectangles containing v . Insert into F∣γr the appropriate paths g v ⋅ gv
at successive vertices (same idea as in the rst part of the proof). is gives us a
factorization of [F∣γr ] by regarding the loop corresponding to a horizontal or ver-
tical segment between two adjacent vertices as lying in the Aij for either of the
rectangles containing the segment.
Diufferent choices of Aij give equivalent factorizations. Also the factorizations
associated to successive paths γr , γr+1 are equivalent, because pushing γr to γr+1
across Rr+1 changes F∣γr to F∣γr+1 by the homotopy within Aij corresponding to
Rr+1 .
We can arrange that the factorization associtated to γ0 is equivalent to [f1 ]⋯[fk ]
and by choosing gv for each vertex v along I × {0} ⊆ I × I , to lie not just in the
the two Aij ’s corresponding to the rectangles containing v , but also to to lie in the
Aα for the fi containing v in its domain.
Similarly, the factorization associated to γmn is equivalent to [f1′ ]⋯[fℓ′ ]. is gives
us that [f1 ]⋯[fk ], [f1′ ]⋯[fℓ′ ] are equivalent.
Corollary 2.145 Let X = U ∪V , U, V open, U, V, U ∩V path connected. Let x0 ∈ U ∩V .
en Φ induces an isomorphism π1 (X) ≃ π1 (U ) ∗ π1 (V )Ò N , where N is the normal
subgroup generated by all elements i1 (ω) ⋅ i2 (ω)−1 , where i1 ∶ π1 (U ∩ V ) → π1 (U ),
i2 ∶ π1 (U ∩ V ) → π1 (V ) morphisms induced by the inclusion and ω ∈ π1 (U ∩ V ).
[Proof: Immediate from Seifert-van Kampen theorem].
Corollary 2.146 Consider the assumption of corollary 2.145. If in addition U ∩ V is
simply connected, then Φ induces an isomorphism π1 (X) ≃ π1 (U ) ∗ π1 (V ).
[Proof: Immediate from corollary 2.145].
Example 2.147
1. Let X be a theta space. Recall π1 (X) is not abelian. (S1 ∨ S1 and theta space are
deformation retracts of R2 ∖ {p, q} therfore π(S1 ∨ S1 ) ≃ π1 (X)).

45
Algebraic topology

a b. c

U ∶= X ∖{c}, V ∶= X ∖{a}. U, V are open and path connected. U ∩V = X ∖{a, c}

is path connected, in addition U ∩ V is contractible.
Ô⇒ π1 (X) ≃ π1 (U ) ∗ π1 (V )Ò N . By corollary 2.146 π1 (X) ≃ π1 (U )∗π1 (V ).
U, V have the homotopy type of S1 Ô⇒ π1 (U ) ≃ Z, π1 (V ) ≃ Z.
Ô⇒ π1 (X) ≃ Z ∗ Z.
2. Let X be the wedge of two circles, i.e. X = S1 ∨ S1 . U ∶= X ∖ {a}, V ∶= X ∖ {c}.
π1 (X) ≃ Z ∗ Z.
Remark 2.148
1. Necessity of path connectedness of Aα ∩ Aβ .
Consider

S.1

And U analogously by re ection on the vertical through the center. en U, V

open, path connected, but U ∩ V is not path connected the morphism induced by
Φ is not surjective, π1 (U ) ∗ π1 (V ) → π1 (S1 ).
2. Necessity of path-connectedness of Aα ∩ Aβ ∩ Aγ .
Consider the theta space

a b. c

Aa ∶= X ∖ {a}, Ab ∶= X ∖ {b}, Ac ∶= X ∖ {c}, then Aa ∩ Ab ∩ Ac = X ∖ {a, b, c}

not path connected, π1 (X) ≃ Z ∗ Z∗.
Example 2.149 (e shrinking wedge of circles)
Consider Cn the circle with center (1/n, 0) and radius 1/n in R2 . Let X = ∪n∈N Cn with
the subspace topology.

46
 2.9 CW complexes (cell complexes)

Consider the retractions rn ∶ X → Cn collapsing all Ci ’s except Cn to (0, 0). Each rn

induces a surjection (rn )∗ ∶ π1 (X) → π1 (Cn ) (basepoint (0, 0)). e product of these
surjections (i.e. ρ[f ]) = (r1 [f ], r2 [f ], . . . , )) gives a homomorphism ρ ∶ π1 (X) →
∏∞ Z (=direct product of countably in nite many copies of Z).
Claim: ρ is surjective (Ô⇒ π1 (X) is uncountable).
Proof: For any sequence of integers (Kn )n∈N we can construct a loop f ∶ I → X based
at (0, 0) and going Kn times around the circle Cn in time [1 − 1/n, 1 − 1/(n + 1)].
f is indeed continuous:

• ∀t ∈ [0, 1), f is continuous at t ✓.

• f is contiuous at 1 ∈ I ⇐⇒ (∀ neighbourhoods V of f (1) = (0, 0) ∃ neighbour-

hood U of 1 such that f (U ) ⊆ V ).
V contains all but nitely many of the circles Cn , and this proves continuity of f
at 1.

2.9. CW complexes (cell complexes)

We have seen, that the torus T can be constructed from

a
R
b b
.
a

e interior of the rectangle can be though of as an open disc or a 2-cel attached to the
union of two circles a, b.
e union of the two circles can be thought of as obtained from their point of intersection
P by attachin to it two open arcs, 1-cells.
In general construct a space X by the following procedure:

1. Start with a discrete set X 0 , whose points are regarded as 0-cells.

2. Inductively construct the n-skeleton X n from the (n − 1)-skeleton X n−1 via con-
tinuous maps φα ∶ S n−1 → X n−1 .
n−1
Xn = X ∐ enα Ò ,
α ∼

where ∼ the equivalence relation generated by x ∼ φα (x), x ∈ ∂en

α.

3. Either stop this inductive process aer nitely many steps, setting X = X n , or
continue inde nitely, setting X = ∪n∈N X n . In latter case, X is given the weak
topology, i.e. A ⊂ X is open, iff A ∩ X n is open in X n ∀n.

A space constructed in this way is called a CW-complex, where C stands for closure-
niteness: the closure of each cell intersects only nitely many other cells, W stands for
weak-topology.
Example 2.150 S2 .

47
Algebraic topology

Suppose we attach a collection e2α of 2-cells to a path connected space X

Y ∶= X ∐ eα Ò , 2
α ∼

where x ∼ φα (x) for x ∈ ∂e2α .

If s0 is a base-point of S1 , then φα determines a loop in X based at φα (s0 ).
Proposition 2.151 If i ∶ X → Y is the inclusion map, then i∗ ∶ π1 (X, x0 ) → π1 (Y, x0 )
is surjective and ker Φ∗ = N , i.e. π1 (Y, x0 ) ≅ π1 (X, x0 )Ò N .
X path-connected space, Y space constructed from X by attaching a family of 2-cells
e2α via maps φα ∶ S1 → X . If s0 is a basepoint of S1 , then φa determines a loop in X
based at φα (s0 ). Let x0 ∈ X a basepoint and let γα ∶ I → X a path with γ(0) = x0 and
γ(1) = φα (s0 ).
en γα ∶ φα γα is a loop in X based at x0 , γα φα γα is nullhomotopic in Y ∀α.
Let N be the normal subgroup of π1 (X, x0 ) generated by all elements γα φα γα .
If i∗ ∶ π1 (X, x0 ) < raπ1 (Y, x0 ) is the homomorphism induced by i ∶ X ↪ Y , then
N ≤ ker i∗ .
Proposition 2.152 If i ∶ X → Y is the inclusion map, then i∗ ∶ π1 (X, x0 ) → π1 (Y, x0 )
is surjective, and ker i∗ = N .
P Consider the space Z obtained from Y by attaching rectangular strips Sα = I ×I
with I ×{0} is attached along γα , I ×{1} is not attached to anything, {0}×I are identi ed
for all α, {1} × I is attached along an arc in e2α .
Z deformation retracts onto Y .
A ∶= Z ∖ {yα }α , where yα is a point in e2α not in the arc along which Sα is attached.
B ∶= Z ∖ X .
We apply the Seifert-van Kampen theorem for Z = A ∪ B . A = Z ∖ {yα }α deformation
retracts onto X . B is contractible. A, B statisfy te conditions of the Seifert-van Kampen
theorem.
Ô⇒ π1 (Y ) ≃ π1 (Z) ≃ π1 (A)Ò N , where N is the normal subgroup generated by the
image of π1 (A ∩ B) → π1 (A).
Consider the cover Aα = A ∩ B ∖ ∪β≠α e2β .
Aα deformation retracts onto a circle in e2α ∖ {yα }, so π1 (Aα ) ≃ Z and π1 (Aα ) is gen-
erated by a loop homotopic to γα φα γα . So π1 (A ∩ B) is generated by loops homotopic
to γα φα γα for all α.
Example 2.154 T has a cell-structure with one 0-cell, two 1-cells and one 2-cell. Let X =
X 1 be the 1-skeleton of T , X 1 = S1 ∨ S1 and attach to X 1 one 2-cell e21 via φ1 ∶ S1 → X
with φ1 (S1 ) = a ⋅ b ⋅ a ⋅ b, to produce Y = T .
a
R
b b
.
a

It follows from the proof, that π1 (T ) ≃ π1 (X )Ò N , where N is the normal subgroup

1

of π1 (X 1 ) generated by φ1 .
Now π1 (X 1 ) ≃ Z∗Z, with generators [a], [b], so π1 (T ) = ⟨[a], [b] ∣ [a][b][a]−1 [b]−1 ⟩.
α ∶= [a], β ∶= [b]. Let F = ⟨α, β⟩ be the free group on two generators α, β and N be the
smallest normal subgropu containing [α, β]. en N ≤ [F, F ]. Furthermore F Ò N is
abelian, then [F, F ] ≤ N . Ô⇒ N = [F, F ], and ⟨α, β ∣ [α, β]⟩ = F Ò N = F Ò [F, F ] =
Z ⊕ Z.

48
 2.10 Surfaces (two-dimensional manifolds)

2.10. Surfaces (two-dimensional manifolds)

2.10.1. Fundamental group of surfaces

Tn ∶= T# . . . #T n-fold Torus.
n
Example 2.155 T#T has CW-cell structure one 0-cell, four 1-cells, one 2-cell.
Example 2.156 T#T#T
Example 2.157 n-fold torus: 4n-gon with labelling (a1 b1 a1 b1 )(a2 b2 a2 b2 ) . . . (an bn an bn )

Proposition 2.158

π1 (Tn ) ≃ ⟨α1 , β1 , . . . , αn , βn ∣ [α1 , β1 ][α2 , β2 ] . . . [αn , βn ]⟩.

P Completly analogous to the case n = 1, i.e. T. Tn has a cell structure with one
0-cell, 2n 1-cells and one 2-cell. e attaching map φ1 ∶ S1 → (Tn )1 of the 2-cell e21
determines a loop in Tn with [φ1 ] = [a1 ][b1 ][a1 ]−1 [b1 ]−1 . . . [an ][bn ][an ]−1 [bn ]−1 .

e real projective plane RP2 Notation: P ∶= RP2 .

RP2 ∶= R ∖ {0}Ò ∼,
3

where x ∼ y ∶⇐⇒ ∃λ ≠ 0 with x = λy .

We can also look at it as
S2 Ò
∼
where p, q ∈ S , p ∼ q ∶⇐⇒ p = −q .
2

RP2 can also be descried as the quotient of D2 with antipodal points of ∂D2 identi ed.
As U = {(x, y, z) ∈ S2 ∣ z ≥ 0} ≃ D2 .
Proposition 2.160 π1 (RP2 ) ≃ ⟨α ∣ α2 ⟩ ≃ Z2 .
P Method 1: Using the cell structure of RP2 given by the picture
a

Method 2: p ∶ S 2 → RP2 is a covering map, since π1 (S2 ) = 0, ϕ ∶ π1 (RP2 , x) → p−1 (x)

x ∈ RP2 is bijective.
Consider Pn ∶= P# . . . #P.
P2 = P#P. has cell structure

b c
b a
. a

Proposition 2.162 π1 (Pn ) ≃ ⟨α1 , . . . , αn ∣ α12 . . . αn

2
⟩.
P Using cell-structure of Pn .
Remark 2.164

49
Algebraic topology

1. For every group G, there exists a CW-complex XG with π1 (XG ) ≃ G: Consider

⟨gα ∣rβ ⟩ a representation of G and take Vα S1α . en attach 2-cells e2β along the
loops speci ed by hte relations rβ .

2. RP2 = P

• P cannot be embedded in R3 (i.e. there does not exist a map f ∶ P → R3 ,

such that f ∶ P → f (P ) is a homeomorphism).
• P can be immersed in R3 (i.e. there exists a differentiable map f ∶ P → R3
with Dp f ∶ Tp P → Tf (p) R3 is injective ∀p ∈ P .

Question: Can we deduce, that Tn , Tm are not homotopy equivalent, if n ≠ m?

Answer: Yes, for T and T2 . In general not yet. (It is not a tirivial matter to compare two
group presentations).
We turn to study π1 Ò [π , π ].
1 1

2.10.2. Homology of surfaces

Let X be a path connected space, x0 , x1 ∈ X , a ∶ I → X path from a(0) = x0 to
a(1) = x1 . en we can construct an isomorphism

â ∶ π1 (X, x0 ) → π1 (X, x1 ),

where [f ] ↦ [a][f ][a] and

âab ∶ π1 (X, x0 )Ò [π (X, x ), π (X, x )] → π1 (X, x1 )Ò [π (X, x ), π (X, x )].

1 0 1 0 1 1 1 1

If b ∶ I → X is a path in X with b(0) = x0 and b(1) = x1 and g = ab, then

ĝab ∶ π1 (X, x0 )Ò [π (X, x ), π (X, x )] → π1 (X, x1 )Ò [π (X, x ), π (X, x )],

1 0 1 0 1 0 1 0

where

[f ][π1 (X, x0 ), π1 (X, x0 )] ↦ [g][f ][g][π1 (X, x0 ), π1 (X, x0 )]

= [f ][π1 (X, x0 ), π1 (X, x0 )].

So
ĝab = 1π1 (X,x0 )Ò .
[π1 (X,x0 ),π1 (X,x0 )]

at is, the isomorphism âab is independent of the choice of path a.

De nition 2.165 If X is a path connected space and x0 ∈ X , let

H1 (X) ∶= π1 (X, x0 )Ò [π (X, x ), π (X, x )].

1 0 1 0

H1 (X) is called the rst homology group of X .

[We omit the basepoint, since there is in fact a unique isomorphism between the corre-
sponding groups of different basepoints].
One can de ne homology groups Hn (x) ∀n ∈ N.

50
 2.10 Surfaces (two-dimensional manifolds)

Lemma 2.166 Let F be a group and N ⊴ F . en

F Ò NÒ F Ò [F, F ]Ò
[F Ò N, F Ò N ] ≃ {n[F, F ] ∣ n ∈ N }.
[Proof omitted].

Let F be a free group with free generators α1 , . . . , αn . Let x ∈ F and N be the smallest
normal subgroup containing x. Finally, let G = F Ò N . F Ò [F, F ] is free abelian
gropu with basis α1 [F, F ], . . . , αn [F, F ]. N Ò [F, F ] is the subgroup generated by
x[F, F ].
e lemma implies that G Ò [G, G] is isomorphic to the quotient of a free abelian group
with basis {α1 [F, F ], . . . , αn [F, F ]} by the subgroup ⟨x[F, F ]⟩.

eorem 2.167 H0 (Tn ) is a free abelian group of rank 2n ≃ Z ⊕ ⋅ ⋅ ⋅ ⊕ Z.

2n

P π1 (Tn ) = ⟨α1 , β1 , . . . , αn , βn ∣ [α1 , β1 ]⋯[αn , βn ]⟩.

H1 (Tn ) = π1 (Tn ) Ò [π1 (Tn ), π1 (Tn ) is the quotient of a free abelian group of rank
2n by the group generated by

[α1 , β1 ]⋯[αn , βn ][π1 (Tn ), π1 (Tn )] = 1π1 (Tn )Ò .

[π1 (Tn ),π1 (Tn )]

I.e. H1 (Tn ) ≃ Z ⊕ ⋅ ⋅ ⋅ ⊕ Z.

eorem 2.169 H1 (Pn ) has a torsion subgroup T (Pn ) of order 2 and H1 (Pn ) Ò
T (P(n) is a free abelian group of order n − 1.

P π1 (Pn ) = ⟨α1 , . . . , αn ∣ α12 . . . αn2 ⟩. H1 (Pn ) = π1 (Pn ) Ò [π1 ([Pn ), π1 (Pn )]
is the quotient of a free abelian group of rank n (basis {α1 , . . . , αn }) by the subgroup
generated by α12 . . . αn2 [π1 (Pn ), π1 (Pn )]. Since we ompute in an abelian group we can
use additive notation and write 2α1 + ⋅ ⋅ ⋅ + 2αn + [. . . ]. Change the basis {α1 , . . . , αn }
to the basis {α1 , . . . , αn−1 , . . . , α1 + ⋅ ⋅ ⋅ + αn }. is shows that H1 (Pn ) is isomorphic
to the quotient of the free abelian group with basis {α1 , . . . , αn−1 , α1 + ⋅ ⋅ ⋅ + αn } by the
subgroup ⟨2(α1 + ⋅ ⋅ ⋅ + αn )⟩. So H1 (Pn ) = Z ⊕ ⋅ ⋅ ⋅ ⊕ Z ⊕Z2 .
n−1

eorem 2.171 e surfaces S2 , T, T2 , …, P, P2 , …are topologically distinct (i.e. any

two of these have different homotopy type).

[Proof: Follows immediatly from the previous theorems].

2.10.3. Classi cation of surfaces

eorem 2.172 (Classi cation of closed connected surfaces)

Let S be a closed surface (i.e. compact without boundary). en S is either homeomorphic
to S2 , or to Tn for some n ∈ N, or to Pm for some m ∈ N.

Ideas of the proof:

51
Algebraic topology

Triangulation of a closed surface

De nition 2.173 A triangulation of a closed surface S is a nite family of closed sub-

setes {T1 , . . . , Tn } that cover S together with homeomorphisms φi ∶ Ti′ → Ti where
each Ti′ is a triangle in R2 . e subsets Ti are called triangles. e subsetes of Ti that
are images of vertices and edges are called vertices, respectively edges. We require that
if Ti , Tj are distinct, i, j ∈ {1, . . . , n}, then Ti ∩ Tj = ∅, or Ti ∩ Tj is vertex or Ti ∩ Tj
is a common edge.

Example 2.174 Torus:

2 3 1

7 7

1. 2 3

eorem 2.175 (T. Radó, 1925)

Any closed surface admits a triangulation.

[Proof: Using the Jordan curve theorem].

Cutting and pasting polygonal regions in R2

eorem 2.176 If S is a closed surface, then S is homeomorphic to a space obtained from

a polygonal region in R2 by gluing its edges together in pairs.

eorem 2.177 If X is the quotient space obtained from a polygonal region in R2 by gluing
its edges together, then X is homeomorphic either to S2 or to Tn , for some n ∈ N or to Pm
for some m ∈ N.

2.11. Knot theory

First attempt to de ne the notion of a knot A knot is a simple closed curve in R3 (i.e.
∃γ ∶ S1 → R3 continuous and injective, further, as S1 is compact and R3 Hausdorff, we
have f is closed) ⇐⇒ γ ∶ S1 → γ(S1 ) is a homeomorphism ⇐⇒ γ is an embedding of
S1 in R3 .
Problem: “wild” knots are allowed in this de nition.

52
 2.11 Knot theory

De nition 2.178 A link L of m components is a subset of R3 consisting of m disjoint,

piecewise linear, simple closed curves. A link of one component is a knot K .

Each of the curves is a union of nitely many line segments attached end to end.
Example 2.179
1. e unknot U :

(U is the only knot bounding a disc embedded in R3 ).

2. e trivial knot of m components:

3. Hopf Link

4. Trefoil knot
5. Borromean rings

De nition 2.180 Too links L1 , L2 in R3 (respectively in S3 = R3 ∪ {∞}) are called

equivalent if there is an orientation preserving piecewise linear homeomorphism h ∶
R3 → R3 (respectively h ∶ S3 → S3 ) with h(L1 ) = L2 .

Question: Given two links L1 , L2 how can we decide if they are equivalent.
• If L1 , L2 are equivalent, then try to deform one into the other.
• If L1 and L2 are not equivalent, use invariants to distinguish them.
If L1 , L2 are equivalent, ten R3 ∖ L1 and R3 ∖ L2 are homeomorphic. e homeomor-
phism type of R3 ∖ L is such an invariant.
How strong an invariant is it?

eorem 2.181 (Gordon-Luecue, 1989)

e knots K1 , K2 are equivalent ⇐⇒ S3 ∖ K1 and S3 ∖ K2 are homeomorphic.

[Proof: Very difficult; omited].

Note: e theorem says, that the homeomorphism type of S3 ∖K is a complete invariant.

53
Algebraic topology

Remark 2.182 e theorem does not hold for links (with more than one component).

Determining the hoeomorphism type of S3 ∖L is not so simple. We turn to study π1 (S3 ∖

L).
at is π1 (S3 ∖ L) ≠ π1 (S3 ∖ L2 ) Ô⇒ S3 ∖ L1 ≠ S3 ∖ L2 Ô⇒ L1 and L2 are not
equivalent.

De nition 2.183 If L is a link in S3 , then π1 (S3 ∖ L) is called the group of the link L.
It is sometimes denoted by π1 (L).

Remark 2.184 e inclusion R3 ↪ S3 induces an inclusion i ∶ R3 ∖ L ↪ S3 ∖ L with

with i∗ ∶ π1 (R3 ∖ L) → π1 (S3 ∖ L) an isomorphism. So it is therefore equivalent to
consider L ⊂ R3 or L ⊂ S3 .

Torus knots Consider T ⊂ R3 (rotat C1 = {(x, y, z) ∈ R3 ∣ y = 0, (x − 1)2 + z 2 = 19 }

about the z -axis). We want to study knots on T.
T can be thought of as the quotient space of {(x, y) ∈ R2 ∣ 0 ≤ x ≤ 1, 0 ≤ y ≤ 1} with
opposite edges identi ed or as the quotient space obtained from R2 by identifying two
points (x, y), (x′ , y ′ ) if x − x′ ∈ Z and y − y ′ ∈ Z.
Let q ∶ R2 → T be the identi cation map described and L = {(x, y) ∈ R2 ∣ ∣y =
m
n
x, m, n ∈∈ Z, 1 < m < n, (m, n) = 1}. Take T(m,n) ∶= q(L) ⊆ T.
is is a simple closed curve on T that wraps m times around a line through the hole in
the torus and n-times around a circle inside T. T(m,n) is called an (m, n)-torus knot.
T(m,n) can be given by the following parametrisation:

x = (2 + cos(nφ/m)) cos φ,
y = (2 + cos(nφ/m)) sin φ,
z = sin(nφ/m),

with φ ∈ [0, 2mn]. is lies on the torus given by (r − 2)2 + z 2 = 1 in cylindrical coor-
dinates.

Remark 2.185

1. T(1,n) is equivalent to U the unknot.

2. T(m,n) , T(n,m) are equivalent.

3. (m, n) ≠ 1: T(m,n) torus link.

We want to compute π1 (S3 ∖ K), where K is the unknot U , or K is a torus knot.

Decomposition of S3 A ∶= {(x1 , x2 , x3 , x4 ) ∈ S3 ∣ x21 + x22 ≤ x23 + x24 } and B ∶=

{(x1 , x2 , x3 , x4 ) ∈ S4 ∣ x21 + x22 ≥ x23 + x24 }.
en A, B are closed subsets of S3 and S3 = A ∪ B . Further A ∩ B = {(x1 , x2 , x3 , x4 ) ∈
S3 ∣ x21 + x22 = x23 + x24 }.
So A ∩ B is the cartesian products of the circle {(x1 , x2 ) ∈ R2 ∣ x21 + x22 = 1/2} with the
circle {(x3 , x4 ) ∈ R2 ∣ x23 + x24 = 1/2}, that is A ∩ B is a torus.

54
 2.11 Knot theory

Now let D2 = {(x1 , x2 ) ∈ R2 ∣ x21 +x22 ≤ 1/2} and S1 = {(x3 , x4 ) ∈ R3 ∣ x23 +x24 = 1/2},
and de ne f ∶ D2 × S1 → A by
√ √ √ √
(x1 , x2 , x3 , x4 ) ↦ (x1 , x2 , 2x3 1 − (x21 + x22 ), 2x4 1 − (x21 + x22 )).

f is obviously continuous and injective. One can easily verify, that it is also surjective.
Further (as D2 × S1 compact, A Hausdorff) f homeomorphism.
Analogously, we can prove that B is homeomorphic to D2 × S1 .

S3 = A ∪ B .

We can now compute:

1. π1 (S3 ∖ U ):
Take U to be the core circle of the A torus, i.e. U = {(x1 , x2 , x3 , x4 ) ∈ A ∣ x1 =
x2 = 0}. e boundary A ∩ B of A is a deformation retract of A ∖ U . So B is a
deformation retract of S3 ∖ U = (A ∪ B) ∖ U .
Ô⇒ π1 (S3 ∖ U ) ≃ π1 (B) ≃ Z.

2. π1 (S3 ∖ K) where K is an (m, n) torus knot T(m,n) :

Consider K as a subset of A ∩ B in S3 . S3 ∖ K = (A ∖ K) ∪ (B ∖ K). (A ∖ K),
(B ∖ K), (A ∖ K) ∩ (B ∖ K) path connected. However (A ∖ K), (B ∖ K),
(A ∖ K) ∩ (B ∖ K) are not open subsets so we need to modify them in order to
apply the Seifert-van Kamepen theorem.
Choose ε > 0, such that there exists a tubular neighbourhood N of K in S3 with
radius ε. S3 ∖ N is a deformation retract of S3 ∖ K . en consider 21 ε open
neighbourhoods A′ , B ′ of A and B respectively.
A′ , B ′ are homeomorphic to the product of S1 with an open disk. And A′ ∩ B ′ is
homemorphic to to (A ∩ B) × (− 12 ε, − 21 ε). Lastly A, B are deformation retracts
of A′ and B ′ respectively.
We can apply the Seifert-van Kampen theorem for S3 ∖ N = (A′ ∖ N ) ∪ (B ′ ∖ N )
to compute π1 (S3 ∖ N ) ≃ π1 (S3 ∖ K).

• A′ ∖ N deformation retracts onto the core circle of the torus A i..e π1 (A′ ∖
N ) ≃ Z, B ′ ∖ N deformation retracts onto the wire cricle of B , i.e. π1 (B ′ ∖
N ) ≃ Z.
• (A′ ∖N )∩(B ′ ∖N ) = (A′ ∩B ′ )∖N has the homotopy type of (A∩B)∖K .
So π1 ((A′ ∖ N ) ∩ (B ′ ∖ N )) ≃ π1 ((A ∩ B) ∖ K) ≃ Z.
• π1 (A′ ∩ B ′ ∖ N ) = ⟨γ⟩ → π1 (A′ ∖ N ) = ⟨α⟩, π2 (A′ ∩ B ′ ∖ N ) = ⟨γ⟩ →
π1 (B ′ ∖ N ) = ⟨β⟩.
i1 (γ) = am , i2 (γ) = β n .
• Applying the Seifert-van Kamen theorem gives

Proposition 2.186 π1 (S3 ∖ T(m,n) ) = ⟨α, β ∣ αm β −n ⟩.

[Proof: Follows from the above analysis].
Proposition 2.187 If the knots T(m,n) , T(m′ ,n′ ) are equivalent, then m = m′ and n = n′ .
Furthermore T(m,n) and U are not equivalen.t

55
Algebraic topology

P (Schreier, 1923)

Consider the element αm = bn =∶ z ∈ π1 (S3 ∖ T(m,n) =∶ G, and the subgroup N of G
generated by z . z ∈ Z(G), i.e. zg = gz∀g ∈ G. So N ≤ Z(G).
en N ⊴ G and we can consider the quotient G Ò N . G Ò N has the following
presentation: ⟨αN, βN ∣ (αN )m , (βN )n ⟩. us G Ò N ≃ Zm ∗ Zn , which in turn
implies that Z(G Ò N ) = {1GÒN }.
Sine Z(G) Ò N ≤ Z(G Ò N ), it follows that N = Z(G), so G Ò Z(G) ≃ Zm ∗ Zn .
Now if T(m,n) , T(m′ ,n′ ) are equivalent, π( T(m,n) ) ≃ π1 (T(m′ ,n′ ) ).
G Ò Z(G) ≃ G′ Ò Z(G′ ) Ô⇒ Zm ∗ Zn ≃ Zm′ ∗ Zn′ .
Ô⇒ m = m′ and n = n′ .
Lastly π1 (U ) ≃ Z, that is π1 (U ) Ò Z(π1 (U )) ≃ Z Ò Z ≠ Zm ∗ Zn for any m, n ∈ Z
with (m, n) = 1

Wirtinger presentation Procedure for writing a presentation of the group of a knot K

starting from a diagram for K ⊂ R3 .

Label the arcs by a1 , . . . , an , so that each ai is ocnnected to ai−1 and ai + 1 (mod n).
Assume that the arcs are oriented compatibly with their labellling.
Draw an arrow xi passing under each ai in a right-le direction. Each xi represents a
loop in R3 ∖ K as follows. Suppose that the black board is the xy -plane P . Consider as
basepoint ∗ the point (0, 0, 1). e loop consists of a segment from ∗ to the tail of xi ,
then the arrow, then a segment from the head of xi to ∗.
At each crossing, there is a ceratain relation:

. .

In total there are n relations.

eorem 2.189 π1 (R3 ∖ K) = ⟨x1 , . . . , xm ∣ r1 , . . . , rn ⟩.

Example 2.190 Trefoil

π1 (trefoil) = ⟨x1 , x2 , x3 ∣ x3 x1 = x2 x3 , x2 x3 = x1 x2 , x1 x2 = x3 x1 ⟩ = ⟨x1 , x2 ∣

x1 x2 x1 = x2 x1 x2 ⟩.

56
 2.11 Knot theory

Remark 2.191 In example 3 of group presentations we proved that ⟨x1 , x2 ∣ x1 x2 x1 =

x2 x1 x2 ⟩ = ⟨α, β ∣ α3 = β 2 ⟩.
P (of theorem 2.189)
We will write f instead of [f ].
We will apply the Seifert-van Kampen theorem for R3 ∖ K . Assume that K lies on the
xy -plane except where it goes down by distance ε > 0, at each crossing. Set the basepoint
at (0, 0, 1).

. z=0
z = −2ε/3

Let A = {(x, y, z) ∈ R3 ∣ z > −2ε/3} − K . en A ≃homeo open 3-dim Ball ∖{n un-
knotted arcs with endpoints on the boundery of the ball}.

So π1 (A) = ⟨x1 , . . . , xn ⟩.
For each crossin
ak
. .
ai ai+1
or

let Yℓ be an open rectangular box around it at level −2ε < z < −ε/3, and set Bℓ =
(Yℓ ∖ K) ∪ { open neighbourhood of an arc from a point on ∂Yℓ to ∗ }.
en Bℓ ≃homeo open 3-dim ball ∖{ one unknotted arc with endpoints on the boundary}.
So π1 (Bℓ ) = ⟨yℓ ⟩.
Finally C = open neighbourhood of (R3 ∖ A ∖ ∪n ℓ=1 Bℓ ) ∪ ( open neighbourhood of an
arc to ∗ ).

R3 ∖ K = A ∪ B1 ∪ ⋅ ⋅ ⋅ ∪ Bℓ ∪ C .

π1 (A ∪ B1 ) =?.
A ∩ B1 =?.
i1 ∶ π1 (A ∩ B1 ) → π1 (A), c1 ↦ xk xi x−1 k , b1 ↦ xi+1 ,
i2 ∶ π1 (A ∩ B1 ) → π1 (B1 ), c1 ↦ y1 , c1 ↦ y1 .
So we have i1 (c1 )(i2 (c1 ))−1 = xk x1 x−1 −1
k y1 , i1 (b1 )(i2 (b1 ))
1−
= xi+1 y1−1 .
So π1 (A∪B1 ) = ⟨x1 , . . . , xn , y1 ∣ xi+1 y1 , xk xi xk y1 ⟩ = ⟨x1 , . . . , xn ∣ xk xi x−1
−1 −1 −1 −1
k xi+1 ⟩.
Repeat this process for each Bi to get π1 (A ∪ B1 ∪ ⋅ ⋅ ⋅ ∪ Bn ) = ⟨x1 , . . . , xn ∣ r1 , . . . , rn ⟩.
Since C and C ∩ (A ∪ B1 ∪ ⋅ ⋅ ⋅ ∪ Bn ) are simply connected, applying the Seifert-van
Kampen theorem we get that π1 (K) = ⟨x1 , . . . , xn ∣ r1 , . . . , rn ⟩.

Study π1 (S3 ∖ K) (≃ π1 (R3 ∖ K)) to show that any one of the ri can be ommited.

Let A′ = A ∪ {∞}. C ′ = C ∪ Bn ∪ {∞}. en π1 (A′ ) ≃ π1 (A) and π1 (A′ ∪ B1 ∪ ⋅ ⋅ ⋅ ∪

Bn−1 ) = ⟨x1 , . . . , xn ∣ r1 , . . . , rn−1 ⟩.
S3 ∖ K = A′ ∪ B1 ∪ ⋅ ⋅ ⋅ ∪ Bn−1 ∪ C ′ .
Now we can compute that π1 (C ′ ) = ⟨yn ⟩ and also that π1 (C ′ ∩(A′ ∪B1 ∪⋅ ⋅ ⋅∪Bn−1 )) =
⟨bn ⟩.

57
Algebraic topology

π1 (C ′ ∩ (A′ ∪ B1 ∪ ⋅ ⋅ ⋅ ∪ Bn−1 )) → π1 (C ′ ), bn ↦ yn .
π1 (C ′ ∩ (A′ ∪ B1 ∪ ⋅ ⋅ ⋅ ∪ Bn−1 )) → π1 (A′ ∪ B1 ∪ ⋅ ⋅ ⋅ ∪ Bn−1 ), bn ↦ xi+1 .

Applying the Seifert-van Kampen theorem we obtain that ⟨x1 , . . . , xn ∣ r1 , . . . , rn−1 ⟩)π1 (S3 ∖
K) = π1 (R3 ∖ K).
Remark 2.193 e above reasoning applies also to compute π1 (S3 ∖L) where L is a link
and π1 (S3 ∖ L) = ⟨x1 , . . . , xn ∣ r1 , . . . , rm ⟩.
Example 2.194

1.

2.

eorem 2.195 e group of a knot is not a complete knot invariant.

2.12. Classi cation of covering spaces

p∶X̃ → X covering map with p(̃ ̃ x

x0 ) = x0 , p∗ ∶ π1 (X, ̃0 ) → π1 (X, x0 ).
̃ ̃0 )) ≤ π1 (X, x0 ) to deduce several statements about
We will study H ∶= p∗ (π1 (X, x
covering maps.
̃ → X be a covering map with p(̃
Lemma 2.196 Let p ∶ X x0 ) = x0 . en
̃ x
1. p∗ ∶ π1 (X, ̃0 ) → π1 (X, x0 ) is a monomorphism.

2. e liing correspondance ϕ ∶ π1 (X, x0 ) → p−1 (x0 ) where [f ] ↦ f̃(1) induces

an injective map Φ ∶ π1 (X, x0 ) Ò H → p−1 (x0 ), where by π1 (X, x0 ) Ò H we
̃ is path connected, then Φ is
denote all the right cosets of H in π1 (X, x0 ). If X
bijective.
̃ based at
3. If f ∶ I → X is a loop based at x0 , then [f ] ∈ H iff f lis to a loop in X
̃0 .
x

P

1. Immediate, it follows from homotopy liing property.

2. If f ∶ I → X , g ∶ I → X are loops based at x0 , then let f̃ ∶ I → X ̃, ̃ g∶I →X ̃

̃
be their lis with f (0) = ̃ g (0) = x̃0 . If [f ] ∈ H[g], then [f ] = [g], where
h ∈ H ∶= p∗ (π1 (X, ̃ x ̃0 )) i.e. h = p ○ ̃
h for some loop ̃ h∶I →X ̃ based at x ̃0 . Now
̃
h⋅̃ g is de ned and it is a li of h ⋅ g . Since [f ] = [h ⋅ g], f (1) = ̃
̃ h⋅̃
g (1).
Ô⇒ f̃(1) = ̃ g (1).
Ô⇒ ϕ([f ]) = ϕ([g]). at shows Φ is well-de ned.
Check that Φ is injective: If ϕ([f ] = ϕ([g]), then f̃(1) = ̃ g (1) and we can de ne
f̃⋅g loop based at x ̃0 . erefore [(f̃⋅̃ g )⋅g] = [f̃]. If F̃ is a path homotopy between
(f̃⋅ ̃
g ) and f̃, then F ∶= p ○ F̃ is a path homotopy between p((f̃⋅ ̃ g ) and p(f̃).
g) ⋅ ̃
̃
at is [(p(f ⋅ ̃ g )) ⋅ g] = [f ].
Ô⇒ [f ] ∈ H[g].
Ô⇒ Φ is injective.
̃ is path connected, then ϕ is surjective. So Φ is also surjective.
If X

58
 2.12 Classi cation of covering spaces

3. Φ is injective, hence ϕ([f ]) = ϕ([g]) ⇐⇒ [f ] ∈ H[g]. For g the constant loop

̃0 ⇐⇒ f̃(1) = x
based at x0 this gives [f ] ∈ H ⇐⇒ ϕ([f ]) = x ̃0 .
̃a is path com-
Lemma 2.198 Let X be path connected and locally path connected. If X
̃ , then p ̃ ∶ X
ponent of X ̃a → X is a covering map.
∣Xa

P X ̃ is locally homeomorphic to X , so X ̃ is also locally path connected. X ̃a is a

̃ ̃ ̃
path component of X , so Xa is open. Covering maps are open maps, so p(Xa ) is open.
Claim: p(X ̃a ) is closed.
Proof: Let x ∈ p(X ̃a ) and U be a path connected open neighbourhood of x such that
−1
p (U ) = ∪α Vα , where Vα pairwise disjoint open, p∣Vα ∶ Vα → U homemorphism ∀α.
Since U contains a point of p(X ̃a ), Vα ∩ X
̃a ≠ ∅ for some α. Now Vα is pat hconnected,
hence Vα ⊆ X ̃α . en p(Vα ) = U ⊆ p(X ̃a ), i.e. x ∈ p(X̃a ). So p(X̃a ) = p(X ̃a ).
p(X a ) ⊆ X is both open and closed
Ô⇒ p(X ̃a ) = X
Ô⇒ p∣X̃a is surjective.
Now let x ∈ X and chose an open neighbourhood U of x as before. If Vα ∩ X ̃a ≠ ∅, then
Vα ⊆ X ̃a . erefore (p ̃ )−1 (U ) is the union of those Vα ’s that intersect X̃α . Each of
∣Xa
these is ope with p∣Vα ∶ Vα → U is a homeomorphism.
p ∶ X → X covering map. We restrict to spaces X thate are locally path connected. en
X is also locally path connected. We can also assume that X is path connected. Lastly
we assume, that X ̃ is also path connected. So, we can determine all covering spaces of
a locally path connected space X by determining all path connected covering spaces of
path components of X .
From onw on, whenever we write p ∶ X ̃ → X is a covering map, we imply that X, X ̃ are
locally path connected, and path connected.
Lemma 2.200 (liing lemma) Let p ∶ X ̃ → X covering map with p(̃ x0 ) = x0 , and
f ∶ Y → X continuos. With Y path connected and locally path connected. f (y0 ) = x0 .
en f can be lied to a continous map f̃ ∶ Y → X ̃ with f (y0 ) = x
̃0 iff f∗ (π1 (Y, y0 )) ⊆
̃ ̃0 )). Also if such a li exists, it is unique.
p∗ (π1 (X, x
P

“⇒” If f̃ exists, then f∗ (π1 (Y, y0 )) = (p ○ f̃)∗ (π1 (Y, y0 )) = p∗ (f̃∗ (π1 (Y, y0 ))) ⊆
̃ x
p∗ (π1 (X, ̃0 )).

“⇐” Given y1 ∈ Y , choose a path α ∶ I → Y with α(0) = y0 , α(1) = y1 . Li f ○α ∶ I →

X to f̃○α∶I →X ̃ with f̃ ̃0 . De ne f̃ ∶ Y → X
○ α(0) = x ̃ by f̃(y1 ) = f̃ ○ α(1).

– f̃ is well-de ned:
Let β ∶ I → Y be a path with β(0) = y0 and β(1) = y1 . Li f ○ β ∶ I → X
to a path f̃ ○β ∶ I → X ̃ with f̃ ○ β(0) = f̃○ α(1). en f̃ ○ α ⋅ f̃○ β is a
li of the loop f ⋅ (αβ). By assumption f∗ (π1 (Y, y0 )) ⊆ p∗ (π1 (X,̃ x ̃0 )). So
[f ○ (αβ)] ∈ p∗ (π1 (X, ̃ x̃0 )). Lemma 2.196 implies that its li f̃ ○ α ⋅ f̃○β
is a loop based at x̃0 . So f̃○ α(1) = f̃ ○ β(1).
– f̃ is constinuous:
Let y1 ∈ Y and N an open neighbourhood of f̃(y1 ). We have to show that
there is an open neighbourhood W of y1 with f̃(W ) ⊆ N . Start by choos-
ing a path connected open neighbourhood U of f (y1 ) such that p−1 (U ) =
∪α Vα , Vα disjoint open sets. p∣Vα ∶ Vα → U homeomorhpism ∀α. f is

59
Algebraic topology

continuous at y1 and Y is locally path connected, so there is a path con-

nected open neighbourhood W of y1 with f (W ) ⊆ U . We will show that
f̃(W ) ⊆ V0 , where V0 is the subset containing f (y0 ). Let y ∈ W , and choose
a path b ∶ I → W with b(0) = y1 and b(1) = y . since f is well-de ned f̃(y)
can be obtained by taking the path αβ from y0 to y , liing f ○ (aβ) to a path
f̃○ (aβ) ∶ I → X ̃ with f ̃ ○ (aβ)(0) = X ̃0 and setting f̃(y) = f ̃○ (αβ)(1).
̃ ̃
Now f ○ α is a li of f ○ β with f ○ α(0) = x ̃0 . Since f ○ β(I) ⊆ U , the path
(p∣V0 )−1 ○ f ○ β is a li of f ○ β with ((p∣V0 )−1 ○ f ○ β)(0) = f (y1 ). en
(f̃ ○ α)((p∣V0 )−1 ○f ○β) is a li of f ○(αβ) with (f̃○α)((p∣V0 )−1 ○f ○β)(0) =
̃0 . But (f̃
x ○ α)((p∣V0 )−1 ○ f ○ β)(1) = ((p∣V0 )−1 ○ f ○ β)(1) ∈ V0 , hence
f̃(W ) ⊆ V0 .

Let y1 ∈ Y and α ∶ U → Y , with α(0) = y0 , and α(1) = y1 . Consider the path f ○ α ∶

I → X and li it to a path f̃ ○α ∶ I → X ̃ with f̃○ α(v) = x̃0 . en ỹ1 = f̃(α(1) and
f̃(α(1)) = f̃ ○ α(1) since f̃ ○ α is a li of f ○ α with f̃(α(0)) = f̃(y0 ) = x0 . So we see
that if such an f̃ exists, it is unique.
Example 2.202 Space that is path connected, but not locally path connected.

.
0 11 1 1
43 2

And consider a point (0, y), where y > 0.

Example 2.203 Space that is locally path connected, but not path connected.

Equivalent covering spaces

De nition 2.204 Let p ∶ X ̃ → X , p′ ∶ X

̃ ′ → X be covering maps, p, p′ are called
equivalent if there exists a homeomorphism h ∶ X̃ →X ̃ ′ with p = p′ ○ h. h is called
equivalence between the covering spaces.

eorem 2.205 Let p ∶ X ̃ → X , p′ ∶ X ̃ ′ → X be covering maps with p(̃

x0 ) = p′ (̃
x′0 ) =
x0 .
ere is an equivalence h ∶ X ̃ → X ̃ ′ with h(̃ ̃′0 ⇐⇒ p∗ (π1 (X,
x0 ) = x ̃ x ̃0 ) =
′ ̃ ′ ′
̃0 )). If such an h exists, then it is unique.
p∗ (π1 (X , x

P

“⇒” h homeomorphism Ô⇒ h∗ (π1 (X, ̃ x ̃′, x

̃0 )) = π1 (X ̃′0 ).
′ ̃
Ô⇒ p∗ (h∗ (π1 (X, x ′ ̃ ′ ′
̃0 ))) = p∗ (π1 (X , x ̃0 )).
̃ x
Ô⇒ p∗ (π1 (X, ̃ x
̃′0 )) = p′∗ (π1 (X, ̃′0 )).

60
 2.12 Classi cation of covering spaces

⇐” – p′ covering map, p continous, X ̃ path connected and locally path connected

̃
p∗ (π1 (X, x ′ ̃ ′ ′
̃0 )) = p∗ (π1 (X , x̃0 )).
Ô⇒ ∃!h ∶ X ̃→X ̃ ′ with p′ ○ h = p and h(̃ ̃′0 .
x0 ) = x
– p covering map, p′ continuos, X ̃ ′ path connected and locally path conected
′ ̃ ′ ′ ̃
̃0 )) = p∗ (π1 (X, x
p∗ (π1 (X , x ̃0 )).
Ô⇒ ∃!k ∶ X ̃′ → X ̃ continous with p ○ k = p′ and k(̃ x′0 ) = x
̃0 .
̃ ̃ ′
– k ○ h ∶ X → X is a li of p since p ○ k ○ h = p ○ h = p iwth k ○ h(̃ ̃0 .
x0 ) = x
̃→X
1X̃ ∶ X ̃ is another such liing, so by uniqueness of lis k ○ h = 1 ̃ .
X
– Analogously, we can prove that h ○ k = 1X̃ ′ .

What about equivalences h ∶ X ̃→X

̃ ′ that do not necessarily satisfy h(̃ ̃′0 ?
x0 ) = x
̃
Lemma 2.207 Let p ∶ X → X be a covering map and x ̃0 , x −1
̃1 ∈ p (x0 ) and Hi ∶=
̃ x
p∗ (π1 (X, ̃i )), i ∈ {0, 1}.

1. If γ ∶ I → X ̃ is a path with γ(0) = x

̃0 and γ(1) = x
̃1 , and α ∶= p ○ γ , then
−1
[α]H1 [α] = H0 .

2. Conversly given x̃′0 and a subgroup H ≤ π1 (X, x0 ) conjugate to H0 , there exists

̃1 ∈ p−1 (x0 ) with H1 = H .
a basepoint x

P

1. “⊆” If [h] ∈ H1 , then [h] = p∗ ([̃

h]) for some ̃ h∶I→X ̃ loop based at x
̃1 . Let
̃ ̃ ̃
k = (γ ⋅ h) ⋅ γ . en p∗ ([k]) = p∗ ([γ ⋅ h ⋅ γ]) = [α ⋅ h ⋅ α] = [α][h][α]−1 .
at is [α][h][α]−1 = p∗ ([̃ ̃ x
k]) ∈ p∗ (π1 (X, ̃0 )) = H0 .
“⊇” Apply the above for γ and α = p ○ γ . We that that [α]H0 [α]−1 ⊆ H1 . Ô⇒
H0 ⊆ [a]H1 [a]−1 .
̃ and H ≤ π1 (X, x0 ) conjugate to H0 . en H0 = [α]H[α]−1 for som
̃0 ∈ X
2. Let x
̃ with γ(0) = x
eα ∶ I → X based at x0 . et let γ be the li of α to X ̃0 . Let
−1
γ(1) = x̃1 . en by 1. Ô⇒ H0 = [α]H1 [α] .

eorem 2.209 Let p ∶ X ̃ → X , p′ ∶ X

̃ ′ → X be covering maps with p(̃
x0 ) = p′ (̃
x′0 ) =
′ ̃
x0 . p, p are quivalent ⇐⇒ H0 ∶= p∗ (π1 (X, x ′ ′ ̃ ′ ′
̃0 )) and H0 ∶= p∗ (π1 (X , x
̃0 )) are con-
jugate.

P
̃→X
“⇒” If h ∶ X ̃ ′ is an equivalence, let x
̃′1 = h(̃ ̃ ′, x
x0 ) and H1′ = p∗ (π1 (X ̃′1 )). By
′ ′ ′
the previous theroem H0 = H1 . And by the lemma H0 , H1 are conjugate. Ô⇒
H0 , H0′ are conjugate.

“⇐” If H0 , H0′ are conjugtae, then according to the lemma there exists x̃1 ′ ∈ ̃ ′ with
X
H1′ = H0 . en by previous theorem there is an equivalence h ∶ X ̃ → ̃
X ′ with
h(̃ ̃1 0.
x0 ) = x

Example 2.211 X = S1 , x0 ∈ S1 , π1 (S1 , x0 ) ≃ Z.

Subgroups of Z: nZ, n ∈ N.
Covering spaces of S1 : p ∶ R → S1 where t ↦ (cos 2πt, sin 2πt).
p∗ (π1 (R)) is trivial. at is this covering space of S1 corresponds to the trivial subgroup

61
Algebraic topology

of π1 (S1 ).
An other covering map we saw was pn ∶ S1 → S1 , z ↦ z n . en (pn )∗ (π1 (S1 )) = nZ.
̃ → S1 , that is not equivalent to any of the
Question: Is there an a covering map p ∶ X
above.
̃ ≤ π1 (S1 ).
No, as p∗ (π1 (X))

e universal covering space

De nition 2.212 Let p ∶ X ̃ → X be a covering map. If X

̃ is simply connected, then X
̃
is called a universal covering space of X .

Remark 2.213 p∗ (π1 (X)) ̃ trivial Ô⇒ any two universal covering spaces of X are equiv-
alent. We can therefore speak of the universal covering space of X .
Lemma 2.214 Let p, q, r be continuous maps with p = r ○ q . If p, r are covering maps,
then q is also a covering map.
P Let x0 ∈ X , z0 ∈ p(x0 ), y0 = q(x0 ).
Claim 1: q is surjective.
Let y ∈ Y . Choose a path α ̃ ∶ I → Y with α ̃(0) = y0 and α ̃(1) = y . en α = r ○ α̃ is a
path in Z iwth α(0) = r(̃ α(0)) = r(y0 ) = r(q(x0 )) = p(x0 ) = z0 . Let α ̃
̃ be the li of
α to a path in X with α ̃
̃(0) = x0 . en q ○ α ̃ ̃
̃(0) = y0 . By
̃ is a li of α to Y with q ○ α
uniqueness of path liings q ○ α ̃
̃=α ̃
̃. So q ○ α
̃(1) = α ̃(1) = y Ô⇒ y = q(α ̃
̃(1)). Ô⇒ q
is surjective.
Claim 2: Given y ∈ Y , there exists n open neighbourhood V of y with q −1 (V ) = ∪γ Wγ ,
where Wγ pairwise disjoint open sets with q∣Wγ ∶ Wγ → V homeomorphism ∀γ . So, let
y ∈ Y . Set z ∶= r(y). Since p, r are covering maps we can nd an open path-connected
neighbourhood of z such that p−1 (U ) = ∪α Uα , Uα disjoint open with p∣Uα ∶ Uα → U
homeomorphism ∀α. r−1 (U ) = ∪β Vβ , Vβ open, r∣Vβ ∶ Vβ → U homeomorphism ∀β .
Call V the member of the family Vβ that contains the point y . q(Uα ) ⊆ r−1 (U ) ∀α
and Uα is connected, hence q(Uα ) ⊆ Vβ for some β . Also q −1 (V ) ⊆ ∪α Uα . Indeed let
x ∈ q −1 (V ) then q(x) ∈ V Ô⇒ r(q(x)) ∈ r(V ) = U Ô⇒ p(x) ∈ U Ô⇒ x ∈ p−1 (U )
Ô⇒ x ∈ ∪α Uα .
erefore q −1 (V ) is the union of those Uα ’s with q(Uα ) ⊆ V .
In fact, q∣Uα ∶ Uα → V is a homeomorphism for each such α.
Since p∣Uα , r∣V are homeomorphisms and q∣Uα = (r∣V )−1 ⋅ p∣Uα .
Call this family of Uα ’s Wγ .

eorem 2.216 Let p ∶ X ̃ → X be a covering map, and X ̃ simply connected. If r ∶ Y → X

̃
is a covering map. en there exists a covering map q ∶ X → Y with r ○ q = p.

P Let x0 ∈ X . Choose x ̃0 ∈ X ̃ and g0 ∈ Y with p(̃x0 ) = x0 , r(y0 ) = x0 .

̃ x
p∗ (π1 (X, ̃0 )) ⊆ r∗ (π1 (Y, y0 )). So we can apply the liing lemma for the covering
map r ∶ Y → X . at is ∃!q ∶ X ̃ → Y continuous with q(̃ x0 ) = y0 and r ○ q = p. p, r are
covering maps Ô⇒ q is covering map.
Remark 2.218 is theorem justi es the use of the term the universal covering space.
Question: Does every space X have a universal covering space?
Answer: No.
Lemma 2.219 Let p ∶ X ̃ → X be a covering map with p(̃ x0 )x0 and X̃ is simply con-
nected. en x0 has an open neighbourhood U such that the map i∗ induced by the
inclusion i ∶ (U, x0 ) ↪ (X, x0 ) is trivial.

62
 2.12 Classi cation of covering spaces

P Let U be an open neighbourhood of x0 with p−1 (U ) = ∪α Uα , Uα pairwise dis-

joint open sets, p∣Uα ∶ Uα → U homeomorphism ∀α. Let U0 be the Uα , that contains x ̃0 .
̃ −1 ̃
Consider f ∶ I → U a loop based at x0 . And let f ∶= (p∣U0 ) (f ) be the li of f in X .
f̃ is a loop based at x ̃ x
̃0 . π1 (X, ̃0 ) trivial Ô⇒ ∃ a path homotopy F̃ between f̃ and the
constant loop at x̃0 . en F = p ○ F̃ is a path homotopy in X between p ○ f̃ = f and the
constant path at p(̃ x0 ) = x0 . at is, i∗ is trivial.
Example 2.221 X = ∪n∈N Cn (Hawaiian earrings). Let rn ∶ X → Cn mapping every Ci
with i ≠ n to (0, 0) with rnCn = 1Cn .
Let U be an open neighbourhood of x0 ∈ X . Choose n large enough such Cn ⊆ U .
j ∶ Cn ↪ X , k ∶ Cn ↪ U , i ∶ U ↪ X .
rn ○ j = 1Cn Ô⇒ (rn )∗ ○ j∗ = 1π( Cn ,x0 ) Ô⇒ j∗ is injective.
Now j∗ = i∗ ○ k∗ , hence i∗ ○ k∗ ∶ π1 (Cn , x0 ) → π1 (X, x0 ) is injective. I.e. i∗ is not
trivial.
Lemma Ô⇒ e shrinking of circles has no universal covering space.

Existence of covering spaces

De nition 2.222 X is called semilocally simply connected if ∀x ∈ X ∃ open neighbour-

hood U of x with i∗ ∶ π1 (U, x) → π1 (X, x), induced by the inclusion i ∶ (U, x) ↪
(X, x), is trivial.

Remark 2.223
1. If U is an open neighbourhood of x with i∗ ∶ π1 (U, x) → π1 (X, x) trivial and
V is an open neighbourhood of x with V ⊆ U , then obviously j∗ ∶ π1 (V, x) →
π1 (X, x) is trivial.
2. If X is locally simply connected, i.e. ∀x ∈ X and ∀ open neighbourhoods U of x
there exists a simply connected open neighbourhood V of x with V ⊆ U , then X
is semilocally simply connected.
Semilocal simply connectedness is in fact a necessary and sufficient condition for the
corresopondance
1∶1
“(Covering maps of X )Ò ∼ → conjugacy classes of subgroups of π1 (X, x0 )”
to be surjective.

eorem 2.224 Let X be path connected, locally path and semilocally simply connected.
̃ → X covering map with p∗ (π1 (X.̃
Let x0 ∈ X and H ≤ π1 (X, x0 ). en ∃p ∶ X ̃ x0 )) =
H.

Corollary 2.225 X has a universal covering space iff X is parth connected, locally path
connected and semilocally simply connected.
P (of the theorem)
We do this in seven steps:
Step 1 Construction of X ̃:
De ne P ∶= {γ ∶ I → Y path with γ(0) = x0 }. For α, β ∈ P de ne a ∼ β if α(1) = β(1)
and αβ ∈ H . Now de ne X ̃ ∶= P Ò ∼ = {γ # ∣ γ ∶ I → X path with γ(0) = x0 }, and
̃
de ne p ∶ X → X by p(γ ) = γ(1).
#

Note:

63
Algebraic topology

• p is surjective, since X is path connected.

We will de ne a topology on X ̃ so that p ∶ X

̃ → X is a covering map.
Remark 2.227 If [α] = [β], then α# = β # . Furhter, if α# = β # and δ ∶ I → X with
δ(0) = α(1), then (αδ)# = (βδ)#
Step 2 Let α ∈ P and U a path connected neighbourhood of α(1). De ne B(U, α) ∶=
{(αδ)# ∣ δ ∶ I → U path with δ(0) = α(1)}. en α# ∈ B(U, α).
Claim: e sets B(U, α) form a basis for a topology on X ̃.
Proof: We rst prove that if β # ∈ B(U, α), then α# ∈ B(U, β) and B(U, α) = B(U, β).
For this consider β # ∈ B(U, α). en β # = (αδ)# , for some δ ∶ I → U with δ(0) =
α(1) and (βδ)# = (αδδ)# = α# . So α# = (βδ)# ∈ B(U, β). Not let (βγ)# ∈
B(U, β). en (βγ)# = ((αδ)γ)# = (α(δγ))# ∈ B(U, α). at is B(U, β) ⊆
B(U, α), and by analogous argument B(U, α) = B(U, β).
̃ . Indeed:
Now we can prove that the sets B(U, α) form a basis for a topology on X
̃ . en choose a path connected open neighbourhood U of α(1). en
• Let α# ∈ X
α# ∈ B(U, α).

• Let α# ∈ B(U, α1 )∩B(U, α2 ). Choose a path connected open neighbourhood V

of α(1) with V ⊆ U1 ∩ U2 . en B(V, α) ⊆ B(U1 , α) ∩ B(U2 , α) = B(U1 , α1 ) ∩
B(U2 , α2 ).
̃.
erefore the sets B(U, α) form a basis for a topology on X
Step 3 p is open and continuous:
p is open: We will prove that p(B(U, α)) = U , which implies p is open.

“⊇” Let x ∈ U and choose a path δ ∶ I → U with δ(0) = x, δ(1) = x. en (αδ)# ∈
B(U, α) and p((αδ)# ) = (αδ)(1) = δ(1) = x. I.e. x ∈ p(B(U, α)), i.e. U ⊆
p(B(U, α)).

“⊆” p(B(U, α)) ⊆ U (follows form the de nition of p and B(U, α).

p is continuos: Let α# ∈ X ̃ and W an open neighbourhood of p(α# ). Choose a path

connected open neighbourhood U of p(α# ) = α(1) with U ⊆ W . en B(U, α) is an
open neighbourhood of α# in X ̃ with p(B(U, α)) = U ⊆ W .
Step 4 ∀x ∈ X ∃ an open neighbourhood U of x with p−1 (U ) = ∪α Vα where Vα pair-
wise disjoint open sets, and p∣Vα ∶ Vα → U homemomorphism.
Let x ∈ X . Choose a path connected open neighbourhood U of X such that i∗ ∶ π1 (U, x) →
π1 (X, x) is trivial.
Claim 1: p−1 (U ) = ∪α∶I→X,α(0)=x0 ,α(1)=x B(U, α).
Proof:

“⊇” p(B(U, α)) = U Ô⇒ B(U, α) ⊆ p−1 (U ) Ô⇒ ∪α∶I→X,... B(U, α) ⊆ p−1 (U ).

“⊆” Let β # ∈ p−1 (U ). en p(β # ) ∈ U , i.e. β(1) ∈ U . Choose a path δ ∈ U from
x to β(1) and let α = βδ . en [β] = [αδ] Ô⇒ β # = (αδ)# ∈ B(U, α). I.e.
p−1 (U ) ⊆ ∪α∶I→X,... B(U, α).

Claim 2: Distinct sets B(U, α) are disjoint.

Proof: Suppose β # ∈ B(U, α) ∩ B(U, α2 ). en B(U, α1 ) = B(U, β) = B(U, α2 ).
Claim 3: p∣B(U,α) ∶ B(U, α) → U is bijective.
Proof: We have shown that p(B(U, α)) = U , i.e. p is surjective. To check injectiv-
ity suppose p((αδ1 )# ) = p((αδ2 )# ). δi ∶ I → U , δi (0) = α(1), i ∈ {1, 2}. en

64
 2.12 Classi cation of covering spaces

(αδ1 )(1) = (αδ2 )(1) Ô⇒ δ1 (1) = δ2 (1). So we can de ne δ1 δ 2 ∶ I → U (loop based

at δ1 (0) = δ2 (0) = δ 2 (1) = x). Since i∗ ∶ π1 (U, x) → π1 (X, x) is trivial, there is a path
homotopy in X between δ1 δ 2 and the constant loop at x. [αδ1 ] = [αδ2 ]. Now p∣B(U,α)
is bijective, continuous and open. Ô⇒ p∣B(U,α) homeomorphism.
Step 5 Liing a path in X to a path in X ̃.
̃
Let e0 be the equivalence class in X of the constant path at x0 . en p(e0 ) = x0 . Let
α ∶ I → X path with α(0) = x0 . We want to compute its li to a path α ̃ ∶ I → X ̃
with α ̃(0) = e0 and show that α ̃(1) = α . Given c ∈ [0, 1], let αc ∶ I → X the
#

path de ned by αc (t) ∶= α(ct), t ∈ I . De ne α ̃ ∶ I → X ̃ by α ̃(c) = (αc )# . en

p(̃α(c)) = (p(αc ) ) = αc (1) = α(c). Ô⇒ p ○ α
#
̃ = α and α ̃(0) = (α0 )# = e0 ,
̃(1) = (α1 ) = α .
α # #

Claim: α ̃ is continuous.
Proof: omitted.
Step 6 p ∶ X̃ → X is convering map.
p is surjective (step 1), p stais es the covering condition (step 4). X is path and locally
path connected by assumption, X ̃ is path connectec (step 5). (X̃ is locally path connected
as X is).
Step 7 H = p∗ (π1 (X,̃ e0 )):
Let α ∶ I → X be a loop based at x0 and α ̃ ∶ I → X ̃ its li with α̃(0) = e0 . en
[α] ∈ p∗ (π1 (X,̃ e0 )) ⇐⇒ α ̃(1) = α ̃(0) = e0 ⇐⇒ α# = e0 . ⇐⇒ α ∼ constant path at
x0 ⇐⇒ [αcx0 ] ∈ H ⇐⇒ [α] ∈ H .

65
Index

Symbols embedding . . . . . . . . . . . . . . . . . . . . . . . . . . 22
equivalence . . . . . . . . . . . . . . . . . . . . . . . . . 60
ε-ball . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 equivalent . . . . . . . . . . . . . . . . . . . . . . . 53, 60
external direct sum . . . . . . . . . . . . . . . . . . 36
A external free product . . . . . . . . . . . . . . . . . 39
adjunction space . . . . . . . . . . . . . . . . . . . . . 33 F
attaching map . . . . . . . . . . . . . . . . . . . . . . . 33
nal point . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
B ner . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
nite topology . . . . . . . . . . . . . . . . . . . . . . . . 4
basepoint . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 nitely generated . . . . . . . . . . . . . . . . . . . . 43
basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 nitely presented . . . . . . . . . . . . . . . . . . . . 43
rst homology group . . . . . . . . . . . . . . . . . 50
C
free abelian group . . . . . . . . . . . . . . . . . . . . 37
category . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 free group . . . . . . . . . . . . . . . . . . . . . . . . . . 41 f
closed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 free product . . . . . . . . . . . . . . . . . . . . . . . . . 38
closure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 fundamental group . . . . . . . . . . . . . . . . . . 19
compact . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 G
comparable . . . . . . . . . . . . . . . . . . . . . . . . . . 5
connected . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 generate . . . . . . . . . . . . . . . . . . . . . . . . . 35, 38
connected component . . . . . . . . . . . . . . . . 12 group of the link . . . . . . . . . . . . . . . . . . . . . 54
connected sum . . . . . . . . . . . . . . . . . . . . . . 34
continuous . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 H
countable topology . . . . . . . . . . . . . . . . . . . . 4
Hausdorff . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
coveriant functor . . . . . . . . . . . . . . . . . . . . 21
homeomorphism . . . . . . . . . . . . . . . . . . . . . 8
covering map . . . . . . . . . . . . . . . . . . . . . . . . 21
homeotopic . . . . . . . . . . . . . . . . . . . . . . . . . 16
covering space . . . . . . . . . . . . . . . . . . . . . . . 21
homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
D homotopy equivalence . . . . . . . . . . . . . . . 30
homotopy equivalent . . . . . . . . . . . . . . . . . 30
deformation retract . . . . . . . . . . . . . . . . . . 29 homotopy inverse . . . . . . . . . . . . . . . . . . . . 30
deformation retraction . . . . . . . . . . . . . . . 29 homotopy type . . . . . . . . . . . . . . . . . . . . . . 30
direct sum . . . . . . . . . . . . . . . . . . . . . . . . . . 35
I
external . . . . . . . . . . . . . . . . . . . . . . . . 36
discrete topology . . . . . . . . . . . . . . . . . . . . . 4 induced homomorphism . . . . . . . . . . . . . 20
disjoint union . . . . . . . . . . . . . . . . . . . . . . . 33 initial point . . . . . . . . . . . . . . . . . . . . . . . . . 16
disjoint union topology . . . . . . . . . . . . . . . 33 interior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
E K

edges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 knot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

66
 INDEX

L S

least upper bound property . . . . . . . . . . . 13 semilocally simply connected . . . . . . . . . 63

length . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 separation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
liing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 simply connected . . . . . . . . . . . . . . . . . . . . 20
liing correspondence . . . . . . . . . . . . . . . . 26 standard topology . . . . . . . . . . . . . . . . . . . . 6
link . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 stereographic projection . . . . . . . . . . . . . . 34
locally connected . . . . . . . . . . . . . . . . . . . . 15 strictly ner . . . . . . . . . . . . . . . . . . . . . . . . . . 5
locally path connected . . . . . . . . . . . . . . . . 15 subset
loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 closed . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
open . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
M subspace topology . . . . . . . . . . . . . . . . . . . . 7
sum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
metric topology . . . . . . . . . . . . . . . . . . . . . . 6 T

topological property . . . . . . . . . . . . . . . . . . 8
N
topological space . . . . . . . . . . . . . . . . . . . . . 4
topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
nullhomotopic . . . . . . . . . . . . . . . . . . . . . . 16
countable . . . . . . . . . . . . . . . . . . . . . . . . 4
discrete . . . . . . . . . . . . . . . . . . . . . . . . . 4
O
nite . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
metric . . . . . . . . . . . . . . . . . . . . . . . . . . 6
open . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
product . . . . . . . . . . . . . . . . . . . . . . . . . 6
open neighbourhood . . . . . . . . . . . . . . . . . 14
subspace . . . . . . . . . . . . . . . . . . . . . . . . 7
order relation . . . . . . . . . . . . . . . . . . . . . . . 12
trivial . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
order topology . . . . . . . . . . . . . . . . . . . . . . 13
triangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
triangulation . . . . . . . . . . . . . . . . . . . . . . . . 52
P
trivial topology . . . . . . . . . . . . . . . . . . . . . . . 4
path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11, 16 U
path components . . . . . . . . . . . . . . . . . . . . 12
path connected . . . . . . . . . . . . . . . . . . . . . . 11 universal covering space . . . . . . . . . . . . . . 62
path homotopic . . . . . . . . . . . . . . . . . . . . . 16
presentation . . . . . . . . . . . . . . . . . . . . . . . . . 43 V
product . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
product topology . . . . . . . . . . . . . . . . . . . . . 6 vertices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

Q W

wedge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
quotient map . . . . . . . . . . . . . . . . . . . . . . . . . 8
wedge sum . . . . . . . . . . . . . . . . . . . . . . . . . . 33
quotient topology . . . . . . . . . . . . . . . . . . . . . 9
word . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
R

rank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
reduced word . . . . . . . . . . . . . . . . . . . . . . . 38
relation subgroup . . . . . . . . . . . . . . . . . . . . 43
reparametrisation . . . . . . . . . . . . . . . . . . . . 18
retract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
retraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

67