Master

Algebraic Topology
Gilles Castel
March 25, 2023

Contents
Lecture 1: Introduction
di 01 okt 10:33
• First part: fundamental groups. We follow Munkres:
– Chap 9 ‘Fundamental group’

– Chap 11 ‘The Seifert-Van Kampen thm’ and
– Chap 12/13 ‘Classification covering spaces’
• Second part: Homology groups, via cudi, chapter of a book.
• 10 problems (solve them during the semester) No feedback during the

semester, but asking questions is allowed. Working together is allowed.
• Exam (completely open book) First 4 questions: 1h30 prep. Last one 30
min, no prep
1. Theoretical question (open book, explain the proof, . . . ) /4
2. New problem (comparable to one of the 10 problems) /4
3. Explain your solution n-th problem solved at home /4
4. Explain your solution m-th problem solved at home /4
5. 4 small questions /4
After the exam, hand in your solutions of the other problems. After a
quick look, the points of 3. and 4. can be ±1 (in extreme cases ±2)
• No exercise classes for this course!
1
Chapter 0
Introduction
0.1 What is algebraic topology?

Functor from category of topological spaces to the category of groups.
• Category: set of spaces and morphisms.
• Functor: X ⇝ GX and f : X → Y ⇝ f∗ : GX → GY such that

– (f ◦ g)∗ = f∗ ◦ g∗
– (1X )∗ = 1GX
Two systems we’ll discuss:
• fundamental groups
• homology groups
Example. Suppose we have a functor. If GX ∼ ̸ GY , then X and Y are

=
not homeomorphic. If ‘shadows’ are different, then objects themselves are
different too.
Proof. Suppose X and Y are homeomorphic. Then ∃f : X → Y and
g : Y → X, maps (maps are always continuous in this course), such that
g ◦ f = 1X and f ◦ g = 1Y . Then f∗ : GX → GY and g∗ : GY → GX
such that (g ◦ f )∗ = (1X )∗ and (f ◦ g)∗ = (1Y )∗ . Using the rules discussed
previously, we get
g∗ ◦ f∗ = 1GX f∗ ◦ g∗ = 1GY ,
which means that f∗ : GX → GY is an isomorphism.
0.2 Fundamental group

Pick a base point x0 and consider it fixed. (The fundamental group will not
depend on it. We assume all spaces are path connected) X ⇝ π(X).
2
• A loop based at x0 ∈ X is a map f : I = [0, 1] → X, f (0) = f (1) = x0 .

• Loops are equivalent if one can be deformed in the other in a continuous
way, with the base point fixed.
• The fundamental group consists of equivalent classes of loops.
Example. Let X = B 2 (2 dimensional disk). Then π(B 2 ) = 1, because

every loop is equivalent to the ‘constant’ loop.
Example. Let X = S 1 and pick x0 on the circle. Two options:

• The loop is trivial equivalent to the constant loop
• The loop goes around the circle.
• The loop goes around the circle, twice.
• The loop goes around the circle, clockwise, once

• ...
π(S 1 ) ∼
= Z (proof will follow).
The composition of loops is simply pasting them. In the case of the circle,
the loop −1 ◦ the loop 2 is the loop 1.
Suppose α : I → X and f : X → Y . Then we define
f∗ [α] = [f ◦ α].
Theorem 1 (Fixed point theorem of Brouwer). Any continuous map from a

rectangle to itself has at least one fixed point.
Proof. Suppose there is no fixed point, so f (x) ̸= x for all x ∈ B 2 . Then

we can construct map r : B 2 → S1 as follows: take the intersection of the
boundary and half ray between f (x) and x.
r(x)
B 2 f (x)
x
Figure 1: Proof of the Brouwer fixed point theorem
If x lies on the boundary, we have the identity map. This map is
CHAPTER 0. INTRODUCTION 3
continuous. Then we have S 1 → B 2 → S 1 , via the inclusion and r. Looking

at the fundamental groups:
π(S 1 ) = Z → π(B 2 ) = 1 → π(S 1 ) = Z.
The map from π(S 1 ) → π(S 1 ) is the identity map, but the first map maps
everything on 1.
CHAPTER 0. INTRODUCTION 4
Chapter 9
Fundamental group
9.51 Homotopy of paths

Definition 1 (Homotopy). Let f, g : X → Y be maps (so continuous). Then
a homotopy between f and g is a continuous map H : X × I → Y such
that
• H(x, 0) = f (x), H(x, 1) = g(x)
• For all t ∈ I, define ft : X → Y : x 7→ H(x, t)

We say that f is homotopic with g, we write f ≃ g. If g is a constant map,
we say that f is null homotopic.
Definition 2 (Path homotopy). Let f, g : I → X be two paths such that

f (0) = g(0) = x0 and f (1) = g(1) = x1 . Then H : I × I → X is a path
homotopy between f and g, if and only if
• H(s, 0) = f (s) and H(s, 1) = g(s) (homotopy between maps)
• H(0, t) = x0 and H(1, t) = x1 (start and end points fixed)

Notation: f ≃p g.
Lemma 1. ≃ and ≃p are equivalence relations.
Proof.
• Reflective: F (x, t) = f (x)

• Symmetric: G(x, t) = H(x, 1 − t)
• Transitive: Suppose f ≃ g and g ≃ h, with H1 , H2 resp.
(
H1 (x, 2t) 0 ≤ t ≤ 21
H(x, t) = .
H2 (x, 2t − 1) 12 ≤ t ≤ 1
5
f
X
Figure 9.1: General example of a homotopy.
Example (Trivial, but important). Let C ⊂ Rn be a convex subset.

• Any two maps f, g : X → C are homotopic.
• Any two paths f, g : I → C with f (0) = g(0) and g(1) = f (1) are
path homotopic.
Choose H : X × I → C : (x, t) → H(x, t) = (1 − t)f (x) + tg(x).
Product of paths
Let f : I → X, g : I → X be paths, f (1) = g(0). Define
(
f (2s) 0 ≤ s ≤ 21
f ∗ g : I → X : s 7→
g(2s − 1) 12 ≤ s ≤ 1.
Remark. If f is path homotopic to f ′ and g path homotopic to g ′ (which

means that f (1) = f ′ (1) = g(0) = g ′ (0)), then f ∗ g ≃p f ′ ∗ g ′ .
So we can define [f ] ∗ [g] := [f ∗ g] with [f ] := {g : I → X | g ≃p f }
Theorem 2.
1. [f ] ∗ ([g] ∗ [h]) is defined iff ([f ] ∗ [g]) ∗ [h] is defined and in that case,
they are equal.
2. Let ex denote the constant path ex : I → X : s 7→ x, x ∈ X. If

f (0) = x0 and f (1) = x1 then [ex0 ] ∗ [f ] = [f ] and [f ] ∗ [ex1 ] = [f ].
3. Let f : I → X : s 7→ f (1−s). Then [f ]∗[f ] = [ex0 ] and [f ]∗[f ] = [ex1 ]
Proof. First, two observations

• Suppose f ≃p g via homotopy H, f, g : I → X. Let k : X → Y .
CHAPTER 9. FUNDAMENTAL GROUP 6

Then k ◦ f ≃p k ◦ g using k ◦ H.
• If f ∗ g (not necessarily path homotopic). Then k ◦ (f ∗ g) = (k ◦ f ) ∗
(k ◦ g).
Now, the proof
2. Take e0 : I → I : s 7→ 0. Take i : I → I : s 7→ s Then e0 ∗ i is a
path from 0 to 1 ∈ I. The path i is also such a path. Because I is
a convex subset, e0 ∗ i and i are path homotopic, e0 ∗ i ≃p i. Using
one of our observations, we find that
f ◦ (e0 ∗ i) ≃p f ◦ i
(f ◦ e0 ) ∗ (f ◦ i) ≃p f
ex0 ∗ f ≃p f
[ex0 ] ∗ [f ] = [f ].
3. Note that i ∗ i ≃p e0 . Now, applying the same rules, we get
f ◦ (i ∗ i) ≃p f ◦ e0
f ∗ f ≃p e x 0
[f ] ∗ [f ] = [ex0 ].
1. Remark: Only defined if f (1) = g(0), g(1) = h(0). Note that f ∗ (g ∗

h) ̸= (f ∗ g) ∗ h. The trajectory is the same, but the speed is not.
Assume the product is defined. Suppose [a, b], [c, d] are intervals in
R. Then there is a unique positive (positive slope), linear map from
[a, b] → [c, d]. For any a, b ∈ [0, 1) with 0 < a < b < 1, we define a
path
ka,b : [0, 1] −→ X
lin. f
[0, a] −−→ [0, 1] −
→X
lin. g
[a, b] −−→ [0, 1] −
→X
lin. h
[b, 0] −−→ [0, 1] −
→X
Then f ∗ (g ∗ h) = k 21 , 34 and (f ∗ g) ∗ h = k 41 , 12
Let γ be that path γ : I → I with the following graphs:

0 a b 1
Figure 9.2: proof path
Note that γ ≃p i. Now, using the fact that composition of positive

linear maps is positive linear.
kc,d ◦ γ ≃p kc,d ◦ i
ka,b ≃p kc,d ,
which is what we wanted to show.
9.52 Fundamental group

Definition 3. Let X be a space and x0 ∈ X, then the fundamental group
of X based at x0 is
π(X, x0 ) = {[f ] | f : I → X, f (0) = f (1) = x0 }.
(Also π1 (X, x0 ) is used, first homotopy group of X based at x0 )

For [f ], [g] ∈ π(X, x0 ), [f ] ∗ [g] is always defined, [ex0 ] is an identity
element, ∗ is associative and [f ]−1 = [f ]. This makes (π(X, x0 ), ∗) a group.
Example. If C ⊂ Rn , convex then π(X, x0 ) = 1. E.g. π(B 2 , x0 ) = 1.
Remark. All groups are a fundamental group of some space.

Question: how does the group depend on the base point?
Theorem 3 (52.1). Let X be a space, x0 , x1 ∈ X and α : I → X a path

Lecture 2: Covering spaces
from x0 to x1 . Then
α̂ : π(X, x0 ) −→ π(x, x1 )
[f ] 7−→ [α] ∗ [f ] ∗ [α].
is an isomorphisms of groups. Note however that this isomorphism depends

on α.
Proof. Let [f ], [g] ∈ π1 (X, x0 ). Then
α̂([f ] ∗ [g]) = [α] ∗ [f ] ∗ [g] ∗ [α]

= [α] ∗ [f ] ∗ [α] ∗ [α] ∗ [g] ∗ [α]
= α̂[f ] ∗ α̂[g].
We can also construct the inverse, proving that these groups are isomor-
phic.
X
x0 f
x1
Figure 9.3: Construction of the group homomorphism
Remark. If f : (x, x0 ) → (Y, y0 ) is a map of pointed topology spaces (f :

X → Y continuous and f (x0 ) = y0 ). Then
f∗ : π(X, x0 ) → π(Y, y0 ) : [γ] 7→ [f ◦ γ]
is a morphism of groups, because of the two ‘rules’ discussed previously,

with
(f ◦ g)∗ = f∗ ◦ g∗ (1X )∗ = 1π(X,x0 ) .

di 08 okt 10:28
Definition 4. Let X be a topological space, then X is simply connected iff
X is path connected and π1 (X, x0 ) = 1 for some x0 ∈ X

Remark. If trivial for one base point, it’s trivial for all base points.
Example. Any convex subset C ⊂ Rn is simply connected.
Example (Wrong proof of π(S 2 ) being trivial). Let f be a path from [0, 1] →
S 2 . Then pick y0 ̸∈ Im(f ). Then S 2 \ {y0 } ≈ R2 . Then use R2 .
This is wrong because we cannot always find y0 ̸∈ Im(f ). Space filling
loops! We’ll see the correct proof later on.
Lemma 2. Suppose X is simply connected and α, β : I → X two paths

with same start and end points. Then α ≃p β.
Proof. Simply connected implies loops are homotopic? Consider α ∗ β ≃p

ex0 , since the space is simply connected.
([α] ∗ [β]) ∗ [β] = [ex0 ] ∗ [β] = [β]

[α] ∗ ([β] ∗ [β]) = [α] ∗ [ex0 ] = [α].
And therefore α ≃p β. (Note: make sure end and start point matchs when
using ∗)
9.53 Covering Spaces

Definition 5 (Evenly covered). Let p : E → B, surjective map (soScontinu-
ous). Let U ⊂ B open. Then U is evenly covered iff p−1 (U ) = α∈I Vα .
with
• Vα open in E
• Vα ∩ Vβ = ∅ if α ̸= β
• p|Vα : Vα → U is a homeomorphism.
Remark. If U ′ ⊂ U , also open and U is evenly covered, then also U ′ .
Definition 6. Let p : E → B be a surjective map. Then p is a covering

projection iff ∀b ∈ B, ∃U ⊂ B open, containing b such that U is evenly
covered by p. Then (E, p) is called a covering space.
Example. Let S 1 = {z ∈ C | |z| = 1}. Take p : R → S 1 : t 7→ e2πit . Note

that R is an easier space than S 1 , and so will be π1 (1 vs Z).

Vα
p−1 (U )
p
U
Figure 9.4: Evenly covered
b
S1
V1 V2
R
Figure 9.5: Example of a covering space
There are also other covering spaces of p. For example, p′ : S 1 → S 1 :

z 7→ z 3 .
S1 S1
p
Figure 9.6: Second example of a covering space
Here we have three copies for each point. We say that the covering has
3 sheets. Note that this is independent of which point we take. This is
always the case! We can show that these are the only coverings of S 1 : R
and z 7→ z n .

Proposition 1. A covering map is always a open map.
Proof. Exercise.
Proposition 2. For any b ∈ B, p−1 (b) is a discrete subset of E. (No

accumulation point)
Proof. Indeed for any α ∈ I, Vα ∩ p−1 (b) is exactly one point.
Remark. A covering is always local homeomorphism. But: there are sur-

jective local homeomorphism which are not covering maps. A covering
map is more than a surjective local homeomorphism.
For example, p : R+ 1
0 → S : t 7→ e
2πit
. Consider the inverse image of a
neighbourhood around 1. When we restrict p to the part around 0, it is no
longer a homeomorphism (we don’t get the whole neighbourhood around
one.)
Creating new covering spaces out of old ones.

• Suppose p : E → B is a covering and B0 ⊂ B is a subspace with the
subspace topology. Let E0 = p−1 (B0 ) and p0 = p|E0 . Then (E0 , p0 ) is a
covering of B0 .
• Suppose that (E, p) is a covering of B and (E ′ , p′ ) is a covering of B ′ , then
(E × E ′ , p × p′ ) is a covering of B × B ′ .
Example. Let T 2 = S 1 × S 1 .
• p : R2 → S 1 × S 1 : (t, s) 7→ (eait , ebis ).

• p′ : R × S 1 → T 2 : (t, z) 7→ (eait , z n )
• p′ : S 1 × S 1 → T 2 : (z1 , z2 ) 7→ (z1n , z2m )
These are the only types of coverings of the torus. We’ll prove this later
on.

(1, 1)
B0 = S 1 × {1} ∪ {1} × S 1
Figure 9.7: Covering of the torus. B0 is the figure 8 space: two connected
circles.
9.54 Fundamental group of the circle (and more)

Given f , when can f be ‘lifted’ to E? I.e. when does there exist an f˜ : X → E
such that p ◦ f˜ = f ? In this section, we’ll only consider X = [0, 1], X = [0, 1]2 .
E
f˜
f
X B
Lemma 3 (Important result). Suppose (E, p) is a covering of B, b0 ∈ B

e0 ∈ p−1 (b0 ). Suppose that f : I → B is a path starting at b0 . Then there
exists a unique lift f˜ : I → E of f with f˜(0) = e0 .
Proof. For any b of B, we choose an open Ub such that Ub is evenly covered

by p. Then {f −1 (Ub ) | b ∈ B} is an open cover of I, which is compact.
There is a number δ > 0 such that any subset of I of diameter ≤ δ is con-
tained entirely in one of these opens f −1 (Ub ). (Lebesgue number lemma).
Now, we divide the interval into pieces 0 = s0 < s1 < . . . < sn = 1 such
that |si+1 − si | ≤ δ. For any i, we have that f ([si , si+1 ]) ⊂ Ub for some b.

E
e0
f˜(si )
f˜
p
B
f Ub
b0
I
s0 sn
si+1
si
We now construct f˜ by induction on [0, si ]
• f˜(0) = e0
• Assume f˜ has been defined on [0, si ]. Let U be an open such that
f [si , si+1 ] ⊂ Ub .
There is exactly one slice Vα in p−1 (Ub ) containing f˜(si ). We define
∀s ∈ [si , si+1 ] : f˜(s) = (p|Vα )−1 ◦ f (s). By the pasting lemma, f˜ is
continuous.
• In this way, we can construct f˜ on the whole of I.
Uniqueness works in exactly the same way, by induction.
Lemma 4 (54.2). (E, p) is a covering of B, b0 ∈ B, e0 ∈ E, with p(e0 ) = b0 .

Suppose F : I × I → B is a continuous map with f (0, 0) = b0 , then there
is a unique F̃ : I × I → E. Moreover, if F is a path homotopy, then also F̃
is a path homotopy.
Proof. Same as in the one dimensional case.

E
F̃
e0
p
F
B
b0
‘Moreover, if F is a path homotopy, then also F̃ is a path homotopy’:

Easy explanation in the book. Another explanation:
B
F
b0
• F (0, t) = b0 , F (1, t) = b1
• F̃ (0, ·) : I → E : t 7→ F̃ (0, t) is a path starting at e0 , and is a lift of

F (0, ·) (the constant path at b0 ).
The path k : I → E : t 7→ e0 is also a lift of the constant path at b0
starting at e0 . Then k = F̃ (0, ·), as the lift is unique.
Theorem 4 (54.3). Let (E, p) be a covering of B, b0 ∈ B, e0 ∈ E with

p(e0 ) = b0 . Let f, g be two paths in B starting in b0 s.t. f ≃p g (so f and
g end at the same point). Let f˜, g̃ be the unique lifts of f, g starting at e0 .
Then f˜ ≃p g̃, and so f˜(1) = g̃(1).
Proof. F : I × I → B is a path homotopy between f and g. Then F̃ :

I × I → E with F̃ (0, 0) = e0 . Then F̃ is a path homotopy, by the previous
result, between F̃ (·, 0) and F̃ (·, 1). Note that p ◦ F̃ (t, 0) = F (t, 0) = f (t)

and p ◦ F̃ (t, 1) = F (t, 1) = g(t). By uniqueness F̃ (·, 0) = f˜, F̃ (·, 1) = g̃.
We’ve shown that homotopy from below lifts to above. The converse is easy.
Now we’re ready to discuss the relation between groups and covering spaces.
Definition 7. Let (E, p) be a covering of B. b0 ∈ B, e0 ∈ E and p(e0 ) = b0 .

Then the lifting correspondence is the map
ϕ : π(B, b0 ) −→ p−1 (b0 )

[f ] 7−→ f˜(1), where f˜ is the unique lift of f , starting at e0 .
This is well-defined because [f ] = [g] ⇒ f˜ ≃p g̃ ⇒ f˜(1) = g̃(1). This ϕ

depends on the choice of e0 .
e0 e1
E
f
b0 B
f
Figure 9.8: Lifting Correspondence
Theorem 5 (54.4). If E is path connected, then ϕ is a surjective map. If

E is simply connected, then ϕ is a bijective map.
Proof. Suppose E is path connected, and let e0 , e1 ∈ p−1 (b0 ). Consider a

path f˜ : I → E with f˜(0) = e0 and f˜(1) = e1 . This is possible because
E is path connected. Let f = p ◦ f˜ : I → B with f (0) = p(e0 ) = b0 and
f (1) = p(e1 ) = b0 , so f is a loop based at b0 . So f is a loop at b0 and its
unique lift to E starting at e0 is f˜. Hence ϕ[f ] = f˜(1) = e1 , which shows
that ϕ is surjective.
Now assume that E is simply connected (group is trivial). Consider
[f ], [g] ∈ π(B0 ) with ϕ[f ] = ϕ[g]. This implies f˜(1) = g̃(1). These start at
e0 . It follows from Lemma 52.3 that f˜ ≃p g̃.
Example. Take the circle and the real line as covering space. Then p−1 (1) =
Z. So we know that as a set π(S 1 ) is countable. Therefore, p ◦ f˜ ≃p p ◦ g̃.
This implies that f ≃p g, and therefore [f ] = [g].

Theorem 6. π1 (S 1 , 1) ∼
= (Z, +).
Proof. Take b0 = 1 and e0 = 0 and p : R → S 1 : t 7→ e2πit . Then

p−1 (b0 ) = Z. And since, R is simply connected, we have that ϕ : π(S, 1) →
Z : [f ] 7→ f˜(1) is a bijection.
Now we’ll show that it’s a morphism. Let [f ] and [g] elements of the
fundamental group of S 1 and assume that ϕ[f ] = f˜(1) = m and ϕ[g] =
g̃(1) = n.
We’re going to prove that ϕ([f ] ∗ [g]) = ϕ([f ]) + ϕ([g]) = m + n. Define
g̃˜ : I → R : t 7→ g̃(t) + m. Then p ◦ g̃˜ = p ◦ g̃ = g, as p(s + m) = p(s) for all
m. Now, look at f˜ ∗ g̃. ˜ This is a lift of p ◦ (f˜ ∗ g̃)
˜ = (p ◦ f˜) ∗ (p ◦ g̃)
˜ = f ∗ g,
which starts at 0. Hence, ϕ([f ] ∗ [g]) = ϕ([f ∗ g]) = the end point of f˜ ∗ g̃, ˜
˜
so g̃(1) = g̃(1) + m = n + m.
The following lemma makes the fact that the covering space is simpler than
the space itself exact.
Lemma 5 (54.6). Let (E, p) be a covering of B. b0 ∈ B, e0 ∈ E and

p(e0 ) = b0 . Then
1. p∗ : π(E, e0 ) → π(B, b0 ) is a monomorphism (injective).

2! Let H = p∗ (π1 (E, e0 )). The lifting correspondence induces a well
defined mapa
Φ : == H⧹π1 (B, b0 ) −→ p−1 (b0 )

H ∗ [f ] 7−→ ϕ[f ],
so ϕ is constant on right cosets. Dividing by H makes Φ always

bijective, even when E is not simply connected.
3. Let f be a loop based at b0 , then f˜ is a loop at e0 iff [f ] ∈ H.
a different notation than the book
Proof. 1. Let f˜ : I → E be a loop at e0 and assume that p∗ [f˜] = 1.

(Then we’d like to show that f itself is trivial.) This implies p ◦ f˜ ≃p
eb0 . This implies that f˜ ≃p ẽb0 = ee0 , or [f˜] = 1.
2. We have to prove two things:

Well defined H ∗ [f ] = H ∗ [g] ⇒ ϕ(f ) = ϕ(g).
Assume [f ] ∈ H ∗ [g], or H ∗ [f ] = H ∗ [g]. This means that
[f ] = [h] ∗ [g], were h = p ◦ h̃ for some loop h̃ at e0 . In other
words [f ] = [h ∗ g], or f ≃p h ∗ g. Let f˜ be the unique lift of
f starting at e0 . Let g̃ be the unique lift of g starting at e0 .
Then h̃ ∗ g̃ (which is allowed, h̃ is a loop) the unique lift of h ∗ g
starting at e0 .
f˜(1) = ϕ(f ) = ϕ(h ∗ g) = (h̃ ∗ g̃)(1) = g̃(1) = ϕ(g). If the cosets
are the same, then the end points of the lifts are also the same.

Lecture 3: Retractions
Injective H ∗ [f ] = H ∗ [g] ⇐ ϕ(f ) = ϕ(g). The end points of f and

g are the same
Now consider h̃ = f˜ ∗ g̃. Then [h̃] ∗ [g̃] = [f˜] ∗ [g̃] ∗ [g̃] = [f˜]. By
applying p∗ , [h] ∗ [g] = [f ].
3. Trivial. Exercise. Apply 2. with the constant path.
Remark. k : X → Y induces a morphism k∗ , we’ve proved that earlier.

Here we only showed injectiveness.
di 15 okt 10:30
9.55 Retractions and Fixed points

Definition 8 (Retract, retraction). Let A ⊂ X, then A is a retract of X iff
there exists a map r : X → A such that r|A = 1|A , i.e. r(a) = a for all
a ∈ A. The map r is called a retraction
Example. Let X be the figure 8 space, and denote the right circle by A.
Then it’s easy to see that there exists a retract from the whole space to A
by mapping the left circle onto the right.
Figure 9.9: Figure 8 space
Lemma 6 (55.1). If A is a retract of X, then i : A → X : a 7→ a induces a

monomorphism i∗ : π(A, a0 ) → π(X, a0 ) with a0 ∈ A.
Proof. Let r : X → A be a retraction. Then r ◦ i = 1A .

i r
(A, a0 ) →
− (X, x0 ) −
→ (A, a0 ).
i
∗ ∗ r
π(A, a0 ) −→ π(X, x0 ) −→ π(A, a0 ).
As r ◦ i = 1A , we get that r∗ ◦ i∗ = (r ◦ i)∗ = (1A )∗ = 1π(A,a0 ) . So i∗ is
injective, r∗ is surjective, which completes the proof.

Example (Theorem 55.2). Let S 1 be the boundary of B 2 . Then S 1 is not

a retract of B 2 . There is a surjective map from B 2 to S 1 , but not one that
is the identity on S 1 .
Proof. Suppose S 1 is a retract. Then i∗ : π(S 1 , x0 ) → π(B 2 , x0 ) is injec-

tive, but i∗ : Z → 1.
Theorem 7 (55.6, Brouwer fixed point theorem). For any map f : B 2 → B 2 ,

there exists at least one fixed point.
Proof. Look at the proof of the first lecture. Now that we’ve actually
proven that π(S1 ) = Z and π(B2 ) = 1, the proof is complete.
Example. Let A be a 3 × 3 matrix with strict positive real entries. Then

A has a positive real eigenvalue.
Proof. Let B = {(x1 , x2 , x3 )T ∈ R3 | 0 ≤ x1 , x2 , x3 ≤ 1 ∧ x21 + x22 + x23 = 1},

an octant of a 2-sphere. Note that B ≈ B 2 , a disk. Now, define f : B →
Ax
B : x 7→ ∥Ax∥ . Note that this maps vectors from B to vectors of B, as A
has positive entries. Note that f is continuous. By Brouwer fixed point
theorem, there exists x0 ∈ B, such that f (x0 ) = x0 , or Ax0 = ∥Ax0 ∥x0 .
Remark. Section 56 gives a proof of the theorem of algebra using funda-

mental proofs. Interesting, but not part of this course.
Remark. We’ll skip the Borsak-Ulam theorem for now.
9.58 Deformation retracts and homotopy type

Lemma 7. Suppose h, k : (X, x0 ) → (Y, y0 ) and assume H : X × I : Y is a
homotopy with
• H(x, 0) = h(x), H(x, 1) = k(x) (definition of homotopy)
• H(x0 , t) = y0 , for all t ∈ I
Then h∗ = k∗ : π(X, x0 ) → π1 (Y, y0 ).
Proof. We have to show that for all f : I → X with f (0) = f (1) = x0 that
h ◦ f ≃p k ◦ f , i.e. h∗ [f ] = k∗ [f ].
H
G:I ×I X ×I Y
.
(s, t) (f (s), t) H(f (s), t)
• Then G is continuous.

• G(s, 0) = H(f (s), 0) = (h ◦ f )(s)

• G(s, 1) = H(f (s), 1) = (k ◦ f )(s)
• G(0, t) = H(f (0), t) = H(x0 , t) = y0

• G(1, t) = H(f (1), t) = H(x0 , t) = y0
So G is a homotopy, and a path homotopy between the two loops.
Definition 9 (Deformation retract). Let A ⊂ X, then A is a deformation

retract of X, iff there exists
• r : X → A, such that r(a) = a for all a ∈ A. (normal retract)
• homotopy H : X × I → X such that
– H(x, 0) = x
– H(x, 1) = r(x)
– H(a, t) = a for all a ∈ A
This means that 1X is homotopic to i◦r via a homotopy leaving A invariant.
Example. Let S 1 ⊂ R2 \ {(0, 0)}. Then S 1 is a deformation retract of

x
R2 \ {(0, 0)}. Using homotopy H : R20 × I → R20 : x 7→ (1 − t)x + t ∥x∥ .
n n
(The same for S and R0 )
x′
S1 R2 \ {(0, 0)}
Figure 9.10: Example of a deformation retract
Example. Consider the figure 8 space. Claim: A is not a deformation

retract of X. We’ll prove this later on.

Figure 9.11: Example of a deformation rectract
Example. Consider the torus and a circle on the torus. Then it is a retract,
but not a deformation retract.
Theorem 8. If A is a deformation retract of X, then i : A → X induces

an isomorphism i∗ . I.e. If you have a deformation retract, it’s not only
injective but also surjective.
Proof. Let i : A → X be the inclusion and r : X → A be the deformation

retraction using H. Then r ◦ i = 1A , which gives r∗ ◦ i∗ = 1π(A,a0 ) .
Now, i ◦ r ≃p 1X using the homotopy of the previous lemma, i.e. H
with H(a0 , t) = a0 . Call h = i ◦ r, k = 1X , and using the previous lemma,
(i ◦ r)∗ = (1X )∗ : π(X, x0 ) → π(X, x0 ), which shows that i∗ ◦ r∗ = 1π(X,x0 ) .
We conclude that both i∗ and r∗ are isomorphisms.
Remark. This means that the fundamental group of R20 is the same as the
one of S 1 , which is Z.
Example. The fundamental group of the figure 8 space and the θ-space are
isomorphic. These spaces are not deformations of each other, but we can
show that they are deformation retracts of R2 \ {p, q}. We say that these
spaces are of the same homotopy type.
R2 \ {p, q} R2 \ {p, q}
Figure 9.12: The figure eight and theta space seen as a deformation retract
of R2 \ {p, q}

Definition 10. Let X, Y be two spaces, then X and Y are said to be of the
same homotopy type iff there exists f : X → Y and g : Y → X such that
g ◦ f ≃ 1X and f ◦ g ≃ 1Y . We say that f, g are homotopy equivalences
and are homotopy inverses of each other.
Remark. This is an equivalence relation.

We’ll prove that spaces of the same homotopy type have the same funda-
mental group. For that, we’ll prove the previous lemma in a more general form,
not preserving the base point.
Lemma 8 (58.4). Suppose h, k : X → Y with h(x0 ) = y0 and k(x0 ) = y1 .

Assume that h ≃ k via a homotopy H : X × I → X, (H(x, 0) = h(x),
H(x, 1) = k(x)). Then α : I → X : s 7→ H(x0 , s) is a path starting in y0
and ending in y1 such that the following diagram commutes
π(Y, y0 ) ∋ [g]
h∗
π(X, x0 ) α̂
k∗
π(Y, y1 ) ∋ [α] ∗ [g] ∗ [α]
Proof. We need to show that α̂(h∗ [f ]) = k∗ [f ], or [α] ∗ [h ◦ f ] ∗ [α] = [k ◦ f ],

or [h ◦ f ] ∗ [α] = [α] ∗ [k ◦ f ]. We’ll prove that these paths are homotopic.
Using the picture, we see that β0 ∗ γ2 ≃p γ1 ∗ β1 , because they are loops in
a path connected space, I ×I. Therefore, F ◦(β0 ∗γ2 ) ≃p F ◦(γ1 ∗β1 ). This
is f0 ∗ c ≃p c ∗ f1 . Now, if we apply H, we get H ◦ (f0 ∗ c) ≃p H ◦ (c ∗ f1 ),
so (h ◦ f ) ∗ α ≃p α ∗ (k ◦ f ), which implies that [h ◦ f ] ∗ [α] = [α] ∗ [k ◦ f ].
I ×I X ×I Y
F (s, t) = (f (s), t) H
β1
(x0 , 1) f1 y1
γ1 γ2 α
c
f0 y0
β0
(x0 , 0)
Figure 9.13: Proof of the Lemma

Theorem 9. Let : X → Y be a homotopy equivalence, with f (x0 ) = y0 .

Then f∗ : π(X, x0 ) → π(Y, y0 ) is an isomorphism.
Proof. Let g be a homotopy inverse of f .

f g f
(X, x0 ) (Y, y0 ) (X, x1 ) (Y, y1 ) · · ·
f∗,x0 g∗,x0
π(X, x0 ) π(Y, y0 ) π(X, x1 )
α̂
1π(X,x0 ) =(1X )∗
π(X, x0 )
.
g∗,x0 f∗,x1
π(Y, y0 ) π(X, x1 ) π(Y, y1 )
β̂
1π(Y,y0 ) =(1Y )∗
π(Y, x0 )
From the first diagram, gy0 ,∗ ◦ fx0 ,∗ is an isomorphism, gy0 ,∗ is surjective.

The second diagram gives that fx1 ,x ◦ gy0 ,∗ is an isomorphism, so gy0 ,∗
is injective, so gy0 ,∗ is an isomorphism. Now composing, we find that
gy−1
0 ,∗
◦ (gy0 ,∗ ◦ fx0 ,∗ ) = fx0 ,∗ is an isomorphism.
9.59 The fundamental group of S n (and more)

Theorem 10 (59.1). Let X = U ∪ V , where U, V are open subsets of X,
such that U ∩ V is path connected. Let i : U → X and j : V → X denote
the natural inclusions and consider x0 ∈ U ∩ V . Then the images of i∗ and
j∗ generate the whole group π(X, x0 ). In other words: any loop based at
x0 can be written as a product of loops inside U and V .
Proof. Let [f ] ∈ π(X, x0 ) denote f : I → X is a loop based at x0 .

Claim: there exists a subdevision of [0, 1] such that f [ai , ai+1 ] lies en-
tirely inside U or V and f (ai ) ∈ U ∩ V . Proof of the claim: Lebesgue
number lemma says that such a subdivision bi exists. Now assume bj is
such that f (bj ) ̸∈ U ∩ V , for 0 < j < m. Then either f (bj ) ∈ U \ V ,
or f (bj ) ∈ V \ U . The first one would imply that f ([bj−1 , bj ]) ⊂ U and
f ([bj , bj+1 ]) ⊂ U . So f [bj−1 , bj+1 ] ⊂ U , so we can discard bj . Same for the
second possibility.
Let αi be a path from x0 to f (ai ) and α0 the constant path t 7→ x0 ,
inside U ∩ V (which is possible, as it is path connected). Now define
p.l.m. f
fi : I → X : I −−−→ [ai−1 , ai ] −
→ X.

Then [f ] = [f1 ] ∗ [f2 ] ∗ · · · ∗ [fn ]. Note that all fi have images inside U or
V . Now,
[f ] = [a0 ] ∗ [f1 ] ∗ [α1 ] ∗ [α1 ] ∗ [f2 ] ∗ [α2 ] ∗ [α2 ] ∗ [f3 ] ∗ · · · ∗ [αn−1 ] ∗ [fn ] ∗ [αn ]
= [α0 ∗ (f1 ∗ α1 )] ∗ [α1 ∗ (f2 ∗ α2 )] ∗ · · · .
Every path of the form αi−1 ∗ (fi ∗ αi ) is a loop based at x0 lying entirely
inside U or V . This means that
[f ] ∈ grp{i∗ (π(U, x0 )), j∗ (π(V, x0 ))}.
f2 V
U f (a2 ) f3
α2
x0 = f (a0 ) = f (a4 ) f (a1 )
α1
f1
α0
α3
f (a3 )
f4
Figure 9.14: Proof of Theorem 59.1
Corollary. Let n ≥ 2, then π(S n , x0 ) = 1
Proof. Consider S n and N, S the north and south pole. Let U = S n \ {N }

and V = S n \ {S}. Then U, V ≈ Rn and U ∩ V is path connected, which
is easy to prove as it is simply homeo to Rn with points removed. Then
π(S n , x0 ) is generated by i∗ (π(U, x0 )) and j∗ (π(V, x0 )), which both are
trivial. This proof doesn’t work for S 1 because then the intersection is not
path connected anymore!
9.60 Fundamental group of some surfaces

Definition 11 (Surface). Surface: compact two-dimensional topological man-

ifold.
Theorem 11. Let X be a space and x0 ∈ X. Let Y be a space and y0 ∈ Y .

Then π(X × Y, (x0 , y0 )) ∼
= π(X, x0 ) × π(Y, y0 ).
Proof. Exercise. Idea: Let f : I → X × Y be a loop based at (x0 , y0 ).

Then f (s) = (g(s), h(s)) where g is a loop in X based at x0 , similar for h,
and conversely.
Example. π1 (T 2 , x0 ) = π1 (S 1 )×π1 (S 1 ) = Z2 . We know that π(S 2 , x0 ) = 1,

so the torus and the two sphere are not homeomorphic to each other, they
aren’t even homotopically equivalent.
2
Example. RP2 = S∼ Then p : S 2 → RP2 , which is by definition continuous
by definition of the topology on the projective plane.
p a a
e1 e0
a ≃p a
Figure 9.15: Example: projective plane.
This means that (S, p) is a covering of the projective plane. The lifting
correspondence says that
Φ : π(RP2 , x0 ) → p−1 (x0 ) = {x˜0 , −x˜0 }
is a isomorphism. Therefore, π1 (RP2 , x0 ) is a group with 2 elements, so

Z2 .
This means, there exist loops which we cannot deform to the trivial
loop, but when going around twice, they do deform to the trivial loop.
E.g. consider the loop a. This is not homotopic equivalent with the trivial
loop, as e1 ̸= e0 . (Or also you can see it because α = α.) But pasting the
loop it twice, we see that is possible. This means that the fundamental
group of the projective space is different from all the one we’ve seen before.
Example. T 2 is the torus. T 2 #T 2 is the connected sum of two tori (Remove

small disc of both tori and glue together), in Dutch: ’tweeling zwemband’.
This space has yet another fundamental group.

Lecture 4: Seifert-Van Kampen theorem
Example. Figure eight space: fundamental group is not abelian. Indeed,

[b ∗ a] ̸= [a ∗ b].
p
˜
b̃
ã ˜
ã
e0 b̃
b a
b0
Figure 9.16: Figure 8 space is not Abelian
Example. Tweeling zwemband. The space retracts to the figure 8 situation,

which shows that the group of the tweeling zwemband has a nonabelian
component.
Figure 9.17: Tweeling zwemband: T #T

di 22 okt 10:26

Chapter 11
Seifert-Van Kampen theorem
Note. This doesn’t follow the book very well.
Definition 12. A free group on a set X consists of a group Fx and a map:

i : X → FX such that the following holds: For any group G and any map
f : X → G, there exists a unique morphism of groups ϕ : FX → G such
that
i
X Fx
f
∃!ϕ .
Note. The free group of a set is unique. Suppose i : X → FX and j : X →

′
FX are free groups.
i j ′
X FX X FX
j i
∃ϕ ∃ψ .
′
FX FX
Then
i
X FX
i .
ψ◦ϕ
FX
Then by uniqueness, ψ ◦ ϕ is 1FX , and likewise for ϕ ◦ ψ.
Note. The free group on a set always exists. You can construct it using
“irreducible words"
Example. Consider X = {a, b}. An example of a word is aaba−1 baa−1 bbb−1 a.
27
This is not a irreducible word. The reduced form is aaba−1 bba = a2 ba−1 b2 a.
Then FX is the set of irreducible words.
Example. If X = {a}, then Fx = {az | z ∈ Z} ∼

= (Z, +). Exercise: check
that Z satisfies the universal property.
Example. If X = ∅, then FX = 1.
Definition 13 (Free product of a collection of groups). Let Gi with i ∈ I,

be a set of groups. Then the free product of these groups denoted by
∗i∈I Gi is a group G together with morphisms ji : Gi → G such that the
following universal property holds: Given any group H and a collection of
morphisms fi : Gi → H, then there exists a unique morphism f : G → H,
such that for all i ∈ I, the following diagram commutes:
ji
Gi G
fi .
∃!f
Note. Again, ∗i∈I Gi is unique.
Example. Construction is similar to the construction of a free group. Let

I = {1, 2} and G1 = G, G2 = H. Then G ∗ H. Elements of G ∗ H
are “words” of the form g1 h1 g2 h2 g3 , g1 h1 g2 h2 , or h1 g1 h2 g2 h3 g3 or h1 g1 h2 ,
. . . with gj ∈ G, hj ∈ H.
Note. G ∗ H is always infinite and nonabelian if G ̸= 1 ̸= H. Even if G, H

are very small, for example Z2 ∗ Z2 = {1, t} ∗ {1, s}. Then ts ̸= st and the
order of ts is infinite.
Note. Z ∗ Z = Fa,b . In general: FX = ∗x∈X Z
11.70 The Seifert-Van Kampen theorem

Theorem 12 (70.1). Let X = U ∪ V where U, V, U ∩ V are open and
path connected.a Let x0 ∈ U ∩ V . For any group H and 2 morphisms
Φ1 : π(U, x0 ) → H and Φ2 : π(V, x0 ) → H such that that Φ1 ◦ i1 and
Φ2 ◦ i2 , there exists exactly one Φ : π(X, x0 ) → H making the diagram
CHAPTER 11. SEIFERT-VAN KAMPEN THEOREM 28

commute
π(U, x0 )
i1 Φ1
j1
i Φ
π(U ∩ V, x0 ) π(x, x0 ) H .
i2 Φ2
j2
π(V, x0 )
i1 , i2 , i, j1 , j2 are induced by inclusions.

a Note that U, V should also be path connected!
Proof. Not covered in this class. You can have a look in the book, but not
insightful.
Theorem 13 (70.2, Seifert-Van Kampen (classical version)). Let X = U ∪ V

as before (U, V, U ∩ V , path connected) and x0 ∈ U ∩ V . Let j : π(U, x0 ) ∗
π(V, x0 ) → π(X, x0 ) (induced by j1 and j2 ). On elements of π(U, x0 ) it
acts like j1 , on elements of π(V, x0 ) it acts like j2 .
G1
f1
f
G1 ∗ G2 H .
f2
G2
Then j is surjectivea and the kernel of j is the normal subgroup of π(U, x0 )∗

π(U, x0 ) generated by all elements of the form i1 (g)−1 i2 (g), were g ∈ π(U ∩
V, x0 ).
a This is the only place where algebraic topology is used. We’ve proved this last week.
The groups U and V generate the whole group. The rest of this theorem follows from
the previous theorem.
Proof. • j is surjective (last week)

• Let N be the normal subgroup generated by i1 (g)−1 i2 (g). Then we
claim that N ⊂ ker(j) This means we have to show that i1 (g)−1 i2 (g) ∈
ker j. j(i1 (g)) = j1 (i1 (g)) by definition of j. Looking at the dia-
gram, we find that j1 (i1 (g)) = j2 (i2 (g)) = i(g) = j(i2 (g)). There-
fore j(i1 (g)−1 i2 (g)) = 1, which proves that elements of the form
i1 (g)−1 i2 (g) are in the kernel.
• Since N ⊂ ker j, there is an induced morphism
k : (π1 (U, x0 ) ∗ π1 (V, x0 ))/N −→ π1 (X, x0 )

gN 7−→ j(g).

To prove that N = ker j, we have to show that k is injective. Because

this would mean that we’ve divided out the whole kernel of j.
Now we’re ready to use the previous theorem. Let H = (π(U ) ∗
π(V ))/N Repeating the diagram:
π(U, x0 )
i1 Φ1
j1
i Φ
π(U ∩ V, x0 ) π(X, x0 ) H .
i2 Φ2 k
j2
π(V, x0 )
Now, we define Φ1 : π(U, x0 ) → H : g 7→ gN , and Φ2 : π(V, x0 ) →

H : g 7→ gN . For the theorem to work, we needed that Φ1 ◦
i1 = Φ2 ◦ i2 . This is indeed the case: let g ∈ π(U ∩ V ). Then
Φ1 (i1 (g)) = i1 (g)N and Φ2 (i2 (g)) = i2 (g)N and i1 (g)N = i2 (g)N
because i1 (g)−1 i2 (g) ∈ N .
The conditions of the previous theorem are satisfied, so there exists
a Φ such that the diagram commutes.
Note that we also have k : H → π(X). We claim that Φ ◦ k = 1H ,
which would mean that k is injective, concluding the proof. It’s
enough to prove that Φ◦k(gN ) = gN for all g ∈ π(U ) and ∀g ∈ π(V ),
as these g’s generate the product of the groups. If a map is the
identity on the generators, it is the identity on the whole group.
Let g ∈ π(U ). Then (Φ ◦ k)(gN ) = Φ(k(gN )) = Φ(j(g)), per defi-
nition of k. On elements of π(U ), j ≡ j1 , so Φ(j(g)) = Φ(j1 (g)) =
Φ1 (g) by looking at the diagram, and per definition of Φ1 , we find
that Φ(g) = gN . So we’ve proven that (Φ◦k)(gN ) = gN . This means
that N is the kernel, so we’ve proved that k is an isomorphism.
Corollary. Suppose U ∩V is simply connected, so π1 (U ∩V, x0 ) is the trivial

group. In this case N = ker j = 1, hence π(U, x0 ) ∗ π(V, x0 ) → π(X, x0 ) is
an isomorphism.
Corollary. Suppose U is simply connected. Then π(X, x0 ) ∼ = π(V, x0 )/N

where N is the normal subgroup generated by the image of i2 : π(U ∩V ) →
π(V, x0 ).
Example. Let X be the figure 8 space.

u v
a
deformation
U Z
b
deformation Z
V
deformation p 1
U ∩V
Figure 11.1: figure eight free product group
Conclusion: π1 (X, p) ∼
= Z ∗ Z = F{[a],[b]}
Example. By induction. Let Wn be the wedge of n circles. Then π(Wn , p) =

Fn . This also holds for a wedge of infinite circles. (But be careful when
choosing topology)
Figure 11.2: Wedge of circles
Example. Fundamental group of the torus.

p b p
torus without inside
U a b
V = T 2 \ {q} = π(V ) = F2
a q a U = disc π(U ) = 1
x0 U ∩V = π1 (U ∩ V, x0 ) = Z
γ1
p b p
Define V, U as in the figure above.
π(T 2 \ {q}, x0 ) −→ π(T 2 \ {q}, p)

[f ] 7−→ [γ1 ] ∗ [f ] ∗ [γ1 ].
0 ) ∼ π(V,p)
Then π(X, x0 ) = π(V,x
N = γ̂(N ) . With N the normal subgroup generated
by the image of i2 : π(U ∩ V, x0 ) → π(T 2 \ {q}, x0 ).
i2 : π(U ∩ V, x0 ) ∼
= Z −→ π(T 2 \ {q}, 0)
⟨[f1 ] ∗ [f2 ] ∗ [f3 ] ∗ [f4 ]⟩ 7−→ ⟨[f1 ] ∗ [f2 ] ∗ [f3 ] ∗ [f4 ]⟩ .
p b p
γ2
γ3
f2
a f1 q f3 a
f4 γ4
γ1
p b p
Now, defining γi as in the picture, we get
γ̂([f1 ] ∗ [f2 ] ∗ [f3 ] ∗ [f4 ]) = [γ1 ] ∗ [f1 ] ∗ [γ2 ] ∗ [γ2 ] ∗ [f2 ] ∗ [γ3 ]
| {z } | {z }
[a] [b]
∗ [γ3 ] ∗ [f3 ] ∗ [γ4 ] ∗ [γ4 ] ∗ [f4 ] ∗ [γ1 ] .

| {z } | {z }
[a]−1 [b]−1

Therefore,
π(V, p)
π(X, p) =
⟨⟨[a][b][a−1 ][b−1 ]⟩⟩
F{[a],[b]}
=
⟨⟨[a] ∗ [b] ∗ [a]−1 ∗ [b]−1 ⟩⟩
= ⟨α, β | αβ = βα⟩ .
Example. Fundamental group of a poolring for twins.
b b c
a a a d
U
b q
d b d
a c
b c
c c c
a
d b a a d
Figure 11.3: Fundamental group of a poolring for twins
Same idea. (Note, we’ve renamed the edges in the figure on the right.)
V = X \ {a}. π1 (V, x0 ) = F{[a],[b],[c],[d]} , π(U ∩ V, x0 ) = Z. Conclusion
π(X, x0 ) = ⟨α, β, γ, δ | [α, β][γ, δ] = 1⟩
In this way, we can calculate the fundamental group of any surface,
e.g. projective space (Z2 ), klein bottle ( α, β | αβα−1 β = 1 , ‘just read

the boundary’), . . .
End of Chapter 11.

Chapter 12
Classification of covering
spaces
Note. This can be chapter 13 in some books
Lemma 9 (74.1! (General Lifting lemma)). Let p : E → B be a covering, Y

a space. Assume B, E, Y are path connected, and locally path connected.a
Let f : Y → B, y0 ∈ Y, b0 = f (y0 ). Let e0 ∈ E such that p(e0 ) = b0 .
Then ∃f˜ : Y → E with f˜(y0 ) = e0 and p ◦ f˜ = f
(E, e0 )
f˜
p .
f
(Y, y0 ) (B, b0 )
iff f∗ (π(Y, y0 ) ⊂ p∗ π(E, e0 ). If f˜ exists then it is unique.

a From now on, all spaces are locally path connected: Every neighbourhood contains
an open that is path connected
Example. Take Y = [0, 1]. Then f is a path, then we showed that every
map can be lifted. And indeed, the condition holds: f∗ (π(Y, y0 )) = 1, the
trivial group, which is a subgroup of all groups.
Proof. ⇒ Suppose f˜ exists. Then p ◦ f˜ = f , so (p ◦ f˜)∗ π(Y, y0 ) =

π(Y, y0 ). The left hand side is of course p∗ (f˜∗ (π(Y, y0 )) ⊂ p∗ (π(E, e0 )),
so p∗ (π(E, e0 )) ⊂ f∗ (π(Y, y0 )).
⇐ First, we’ll show the uniqueness. Suppose f˜ exists.
34
f˜(y)
e0 E
f˜
p
α
I
f
y b0 f (y)
y0
B
Figure 12.1: general lifting lemma
p ◦ (f˜ ◦ α) = f ◦ α, so f˜ ◦ α is the unique lift of f ◦ α starting at e0 .

Hence f˜(y) the endpoint of the unique lift of f ◦ α to E starting at e0 .
This also shows how you can define f˜: choose a path α from y0 to y.
Lift f ◦ α to a path starting at e0 . Define f˜(y) = the end point of this lift.
Problem: is this well defined? A second problem: is f˜ continuous?
• Well defined:
γ = lift of f ◦ α
e0 γ e1
E
˜
δ =f ◦β
f˜
p
f
y b0 f ◦ α f (y)
y0 α
f ◦β B
β
β
Figure 12.2: well defined general lifting lemma
As [α] ∗ [β] ∈ π(Y, y0 )
f∗ ([α] ∗ [β]) = ([f ◦ α] ∗ [f ◦ β]) ∈ f∗ (π1 (Y, y0 )),
CHAPTER 12. CLASSIFICATION OF COVERING SPACES 35

Lecture 5: The general lifting Lemma
which is by assumption a subgroup of p∗ (π(E, e0 )) = H.

And now, by Lemma 54.6 (c), a loop in the base space lifts to a loop
in E if the loop is in H. This lift is of course just γ ∗ δ, so the end
points in the drawing should be connected! This means that δ is the
lift of f ◦ β starting at e0 , so the endpoint of the lift of f ◦ β is the
endpoint of the lift of f ◦ α. Therefore f˜(y) is well defined.
• Note that we didn’t use the locally path connectedness yet. We’ll
need this for continuity. To be continued . . .

di 05 nov 10:22
Recap:
Lemma 10 (74.1! (General Lifting lemma)). Let p : E → B be a covering, Y

a space. Assume B, E, Y are path connected, and locally path connected.a
Let f : Y → B, y0 ∈ Y, b0 = f (y0 ). Let e0 ∈ E such that p(e0 ) = b0 .
Then ∃f˜ : Y → E with f˜(y0 ) = e0 and p ◦ f˜ = f
(E, e0 )
f˜
p .
f
(Y, y0 ) (B, b0 )
iff f∗ (π(Y, y0 ) ⊂ p∗ π(E, e0 ). If f˜ exists then it is unique.

a From now on, all spaces are locally path connected: Every neighbourhood contains
an open that is path connected
Proof. We prove that f˜ is continuous.
• Choose a neighborhood of f˜(y1 ), say N .

• Take U , a path connected open neighbourhood of f (y1 ) which is
evenly covered, such that the slice p−1 (U ) containing f˜(y1 ) is com-
pletely contained in N .
Can we do this? The inverse image of U is a pile of pancakes. One of
these pancakes contains f˜(y1 ). Then, because N is a neighborhood
of f˜(y1 ), we can shrink the pancake such that it is contained in N
• Choose a path connected open which contains y1 such that f (W ) ⊂
U . We can do this because of continuity of f .
• Take y ∈ W . Take a path β in W from y1 to y. (Here we use that W
is path connected) Now consider p|V and defined Then α ∗ β is path
from y0 to y, f ◦ (α ∗ β) = (f ◦ α) ∗ (f ◦ β). Then f]◦ α ∗ (p−1 |V ◦ f ◦ β)
is the lift of f ◦ (α ∗ β) starting at y0 . So by definitie of f˜, we have

that f˜(y) is the endpoint of that lift, which belongs to V ⊂ N . This

means that f˜(W ) ⊂ N , which proves continuity.
E
N
V
e0
f˜(y1 )
f˜
U
W
y y1 f b0 f (y1 ) f (y)
y0 α β
Figure 12.3: Proof of the continuity of the general lifting lemma
This is a powerful lemma: ‘There exists a lift f˜ such that . . . ’ is pure

topology, and f∗ π(Y, y0 ) ⊂ p∗ π(E, e0 ) is pure algebra. It’s very unusual that
two statements like this are completely equivalent.
Recap:
Lemma 11 (General lifting lemma, short statement). Short statement:
(E, e0 )
f˜
p
f
(Y, y0 ) (B, b0 )
∃!f˜ ⇔ f∗ π(Y, y0 ) ⊂ p∗ π(E, e0 ).
Definition 14. Let (E, p) and (E ′ , p′ ) be two coverings of a space B. An

equivalence between (E, p) and (E ′ , p′ ) is a homeomorphism h : E → E ′
such that
E h E′
p
p′
B
′
is commutative. p ◦ h = p.

Theorem 14 (74.2). Let p : (E, e0 ) → (B, b0 ) and p′ : (E ′ , e′0 ) → (B, b0 )

be coverings, and let H0 = p∗ π(E, e0 ) and H0′ = p′∗ π(E ′ , e′0 ) ≤ π(B, b0 ).
Then there exists an equivalence h : (E, p) → (E ′ , p′ ) with h(e0 ) = e′0 iff
H0 = H0′ . Not isomorphic, but really the same as a subgroup of π(B, b0 ).
in that case, h is unique.
Proof. ⇒ Suppose h exists. Then
h
(E, e0 ) (E ′ , e′0 )
p
p′
(B, b0 )
Therefore p∗ π(E, e0 ) = p′∗ (h∗ π(E, e0 )). Since h is a homeomorphism,

it induces an isomorphism, so p′∗ (h∗ π(E, e0 )) = p ∗′ (π(E ′ , e′0 )
⇐
(E ′ , e′0 )
k
p′
p
(E, e0 ) (B, b0 )
By the previous lemma, there exists a unique k iff p∗ π(E, e0 ) ⊂

p′∗ π(E ′ , e′0 ) or equivalently H0 ⊂ H0′ , which is the case. Reversing
the roles, we get
(E, e0 )
l p
p′
(E ′ , e′0 ) (B, b0 )
for the same reasoning, l exists. Now, composing the diagrams
(E, e0 ) (E ′ , e′0 )
l◦k k◦l
p p′
′
p p
(E, e0 ) (B, b0 ) (E ′ , e′0 ) (B, b0 )
But placing the identity in place of l ◦ k or k ◦ l, this diagram also

commutes! By unicity, we have that l ◦ k = 1E and k ◦ l = 1E ′ .
Therefore, k and l are homeomorphism k(e0 ) = e′0 .
Uniqueness is trivial, because of the general lifting theorem.
Note that this doesn’t answer the question ‘is there a equivalence between
two coverings’, it only answers the question ‘is there an equivalence between two
coverings mapping e0 → e′0 ’. So now, we seek to understand the dependence of
H0 on the base point.

Lemma 12. Let (E, p) be a covering of B. Let e0 , e1 ∈ p−1 (b0 ). Let

H0 = p∗ π(E, e0 ), H1 = p∗ π(E, e1 ).
• Let γ be a path from e0 to e1 and let p ◦ γ = α be the induced loop
at b0 . Then H0 = [α] ∗ H1 ∗ [α]−1 , so H0 and H1 are conjugate inside
π(B, b0 ).
• Let H be a subgroup of π(B, b0 ) which is conjugate to H0 , then there
is a point e ∈ p−1 (b0 ) such that H = p∗ π(E, e).
So a covering space induces a conjugacy class of a subgroup of π(B, b0 ).
Proof. • Let [h] ∈ H1 , so this means that h = p ◦ h̃, where h̃ is a

loop based at e1 . Then (γ ∗ h̃) ∗ γ is a loop based at e0 . This
means that the path class [p((γ ∗ h̃) ∗ γ)] ∈ H0 . This means that
[p ◦ γ] ∗ [h] ∗ [p ◦ γ] ∈ H0 , or [α] ∗ [h] ∗ [α]−1 ∈ H0 . So we showed that
if we take any element of H1 and we conjugate it with α, we end up
in H0 , so [α] ∗ H1 ∗ [α]−1 ⊂ H0
For the other inclusion, consider γ going from e1 → e0 . The same
argument shows that [α]−1 ∗ H0 ∗ [α] ⊂ H1 , or H0 ⊂ [α] ∗ H1 ∗ [α]−1 .
This proves that H0 = [α] ∗ H1 ∗ [α]−1 .
• Take H = [β] ∗ H0 ∗ [β]−1 for some [β] ∈ π(B, b0 ). So H0 = [β]−1 ∗

H ∗ [β]. Take α = β. Then H0 = [α] ∗ H ∗ [α]−1 , where α, β are
loops based at b0 . Let γ be the unique lift of α starting at e0 . Take
e = γ(1), the end point of γ. (So p(e) = b0 ) From the first bullet
point, it follows that p∗ π(E, e) = H ′ satisfies H0 = [α] ∗ H ′ ∗ [α]−1 .
So we have both H0 = [α] ∗ H ∗ [α]−1 = H0 = [α] ∗ H ′ [α]−1 . This
implies that H ′ = H.
This completely answers the question: ‘When are two covering spaces equiv-
alent’:
Corollary (Theorem 74.4). Let (E, p) and (E ′ , p′ ) be two coverings, e0 ∈

E, e′0 ∈ E ′ with p(e0 ) = p(e′0 ) = b0 . Let H0 = p∗ π(E, e0 ), H0′ = p′∗ π(E ′ , e′0 ).
Then (E, p) and (E ′ , p′ ) are equivalent iff H0 and H0′ are conjugate inside
π(B, b0 ).
Question: can we reach every possible subgroup? Answer: yes, in some
conditions.
12.75 Universal covering space

Definition 15. Let B be a path connected and locally path connected space.
A covering space (E, p) of B is called the a universal covering space iff E
is simply connected, so π(E, e0 ) = 1.
a upto equivalence

Remark. Any two universal coverings are equivalent. Even more, we can
choose any base point we want.
h(e0 )=e′0
(E, e0 ) (E ′ , e′0 )
p
p′
(B, b0 )
h exists because the groups of (E, e0 ) and (E ′ , e′0 ) are trivial.
Remark. We don’t cover Lemma 75.1, and we only cover point (a) from
75.2
Lemma 13 (75.2). Suppose
X
q
p Y
r
Z
If p and r are covering maps, then also q is a covering map. (Also: if q and
p are covering maps, then so is r. Not the case for q, r ⇒ p!)
Proof. • q is a surjective map. Choose a base point in x0 , and call

y0 = q(x0 ), z0 = r(y0 ). Certainly, y0 lies in the image of q. Now,
take y ∈ Y , and choose a path α̃ from y0 to y. Now, denote by α
the projection of α̃, a path from z0 to r(y). Let α̃ ˜ be the unique lift
of α to X starting at x0 . This is defined as we assume that p is a
covering map. Then q ◦ α̃ ˜ is a path starting at q(α̃(0))
˜ = q(x0 ) = y0 .
Moreover, q ◦ α̃ ˜ is a lift of α = r ◦ α̃ to Y . Indeed consider the
˜ = p ◦ α̃
projection, r ◦ q ◦ α̃ ˜ = α. Of course, α̃ is also a lift from α
starting at y0 . Since r is assumed to be a covering, and lifts of paths
are unique, we get that q ◦ α̃ ˜ = α̃, so the end points are the same:
˜
q(α̃(1)) = α̃(1) = y, so y lies in the image of q.
The only fact we’ve used is that q is a continuous map, so that q ◦ α̃ ˜
is again a path.
• Now we show that every point of y has a neighbourhood that is

evenly covered. Choose y ∈ Y and project it down to Z. r(y) has a
neighbourhood U that is evenly covered by p, and also by r. Now we
can shrink it so that is is evenly covered by both covering maps. We
can also choose it to be path connected.
So p−1 (U ) = α∈I Uα , and r−1 (U ) = β∈J Vβ . Let V be the slice
S S
containing Y . Then we claim that V will be evenly covered by U .

S
Consider a Uα . Then q(Uα ) is connected and contained in β∈J Vβ ,
but all these Vβ are disjoint, so there is exactly one Vβ such that
q(Uα ) ⊂ Vβ .
Now, let I ′ = {α | q(Uα ) ⊂ V }. For any α ∈ I ′ , we have the diagram
Uα
q
p V
r
U
As r and pSis a homeomorphism, q is also a homeomorphism. Hence

q −1 (V ) = α∈I ′ Uα , and q|Uα : Uα → V is a homeomorphism.
This means that q is a covering projection.
Why is this useful? Because now we can say why the universal covering
space is a universal covering space.
Theorem 15 (57.3). Let (E, p) be a universal covering of B. Let (X, r) be

a another covering of B. Then there exists a map q : E → X such that
r ◦ q = p and q is a covering map.
E
q
p X
r
B
Every covering space is itself covered by the universal covering space, if it

exists.
Proof. Drawing the diagram differently,
X
q
r
p
E B
Choose e0 , x0 mapped to b0 ∈ B. Then π(E, e0 ) = 1 ⊂ r∗ π(X, x0 ).

Then there exists a map q by the general lifting lemma. So q makes the
diagram commutative. By the previous result, q is a covering map.
Covering transformations
Definition 16 (Group of covering transformation). Let (E, p) be a covering

of B. We define
C(E, p, B) = {h : E → E | h is an equivalence of covering spaces}
Elements of this set are homeomorphism h such that p ◦ h = p. The com-

position of two such elements is again such an elements, same for inverse.
This means that C is a group, the group of covering transformations, also
called Deck-transformations.
Example. Consider the covering space R → S 1 : t 7→ e2πit . For any z ∈ Z,

there is a map hz : R → R : r 7→ r + z, which is a covering transformation.
Indeed e2πit = e2πi(t+z) . Claim: these are the only covering transforma-
tions. Conclusion: C(R, p, S 1 ) = (Z, +)
Proof. Suppose h : R → R is another covering transformation. We cer-

tainly have h(0) = z for some z ∈ Z. Therefore, h(0) = hz (0), from this
follows immediately that h ≡ hz .
Why? ‘If two covering transformations agree in one point, the agree
everywhere.’ Indeed, h1 , h2 ∈ C(E, p, B) and h1 (e) = h2 (e) ⇒ h1 ≡ h2 ,
because
E
h1 andh2
p
E B
and, h1 and h2 are both lifts of p and there is a unique lift when fixing
the base point, so h1 and h2 agree.
Goal: what is the structure of the group C(E, p, B) in terms of fundamental
groups? Let (E, p) be a covering of B. p(e0 ) = b0 , H0 = p∗ π(E, e0 ). Remember:
Φ : π(B, b0 )/H0 −→ p−1 (b0 )

H0 ∗ [α] 7−→ α̃(t)
is a bijection, where α̃ is the unique lift of α starting at e0 .

Now consider ψ : C(E, p, B) → p−1 (b0 ) : h 7→ h(e0 ). ψ is injective. Reason:
same as before, if they agree on one point, these are the same. In general ψ will
not be surjective.

Nπ(B,b0 ) (H0 )
Lemma 14. Im ψ = Φ , where
H0
Nπ(B,b0 ) (H0 ) = {[α] ∈ π(B, b0 ) | [α] ∗ H0 ∗ [α]−1 = H0 },
which is the largest subgrouop of π(B, b0 ) in which H0 is normal. ‘Nor-

maliser’.
We’ll later show that this group has the same structure as C(E, p, B).
Proof. Consider H0 ∗ [α].a Then Φ(H0 ∗ [α]) = α̃(1), where α̃ is the lift of α
starting at e0 . Let’s denote α̃(1) = e1 . Question: which of these elements

lie in the image of ψ.

e1 lies in the image of ψ iff there exists a covering transformation h
sending e0 to e1 , which is equivalent to H0 = H1 = p∗ π(E, e1 ). On the
other hand, we also know that H0 = [α] ∗ H1 ∗ [α]−1 . Conclusion: e1 lies
in the image of ψ iff H0 = [α] ∗ H0 ∗ [α]−1 , iff [α] ∈ Nπ(B,b0 ) (H0 ).
a Inthe book, they use [α] ∗ H0 . This is not wrong, as for elements in the normalizer,
left and right cosets are the same, so writing [α] ∗ H0 is allowed. But in general, we
write H0 ∗ [α].
This means we have the following situation:
ψ Ψ−1 Nπ(B,b0 ) (H0 )

C(E, p, B) Im ψ ⊂ p−1 (b0 )
H0
h h(e0 ) = e1 H0 ∗ [α], α = p ◦ γ
Theorem 16. The map Φ−1 ◦ ψ : C(E, p, B) → Nπ(B,b0 ) (H0 )/H0 is an

isomorphism of groups.
Proof. Let h, k ∈ C(E, p, B) with h(e0 ) = e1 and k(e0 ) = e2 .
e1
γ
e0
δ e2 e3
Then
(Φ−1 ◦ ψ)(h) = H0 ∗ [α] α = p ◦ γ

(Φ−1 ◦ ψ)(k) = H0 ∗ [β] β = p ◦ δ.
Then call h(k(e0 )) = h(e2 ) = e3 . Claim: (h ◦ δ)(0) = h(e1 ) = e1 .

(h ◦ δ)(1) = h(e2 ) = e4 . Then ε := γ ∗ (h ◦ δ) is a path from e0 to e3 . This
implies that (Φ−1 ◦ ψ)(h ◦ k) = H0 ∗ [p ◦ ε]. Then p ◦ ε = (p ◦ γ) ∗ (p ◦ h ◦ δ) =
α ∗ (p ◦ δ) = α ∗ β. This means that (Φ−1 ◦ ψ)(h ◦ k) = H0 ∗ [α ∗ β].
This shows that this is indeed a morphism.
What can we do with this? Some covering spaces are nice: e.g. when the
normalizer is the entire group. (Which is the case when abelian)
Definition 17. A covering space (E, p) of B is called regular if H0 is normal

in π(B, b0 ), where b0 ∈ B, p(e0 ) = b0 , H0 = p∗ π(E, e0 ).
This is the case iff for every e1 , e2 ∈ p−1 (b0 ), there exists an h ∈ C(E, p, B)
such that h(e1 ) = e2 .

h
(E, e1 ) (E, e2 )
p p
B
This h exists iff H1 = H2 . H1 = p∗ π(E, e1 ) and H2 = p∗ π(E, e2 ). If H0 is
normal. Then H1 and H2 are the same as H0 , because they are conjugate to
H0 . So for H1 , H2 the h exists. Conversely: exercise.
In this case, (Φ−1 ◦ ψ) : C(E, p, B) → π(B,bH0
0)
is an isomorphism. Special
case: Let (E, p) be the universal covering space, so π(E, e0 ) = 1, or H0 = 1. In
this case, Φ−1 ◦ ψ : C(E, p, B) → π(B, b0 ) is an isomorphism.
Example. C(R, p, S 1 ) ∼
=Z∼
= π(S 1 , b0 ).
p
Example. Consider S 2 − → RP 2 . C(S 2 , p, RP 2 ) = {1S 2 , −1S 2 }, because
2
take a point e0 ∈ S , then under a covering transformations, it can only be
mapped to e0 or to −e0 . So these are the only covering transformations.
And indeed, π(RP 2 ) = Z2
2
Note that RP 2 ∼= S∼ , where ∼ is in a sense defined by the groupactions
of 1S 2 and −1S 2 . More on this next week.
Lemma 15 (75.4). If B admits a universal covering space (E, p), then any
element b of B has a neighbourhood U such that
i8 : π1 (U, b) −→ π1 (B, b)
[α] 7−→ [1].
is trivial. So the loop doesn’t have to be trivial in E, but is is in B.
Proof. Take U evenly covered. α̃ ≃p eE in E, so p ◦ α̃ ≃p p ◦ ee in B so

α ≃p eb , so i∗ [α] = [eb ].
So this condition is needed for a universal covering spaces. And fun
fact: this is sufficient. This is proved in section 77, but we don’t need to
know this.

Lecture 6
E
α̃
αU B
Figure 12.4: lemma universal covering space
We call this property semi-locally simply connected. The loops themself

don’t have to be simply connected, but in the bigger space they are.
Lecture 6
di 12 nov 10:27
Recap: p : E → B and C(E, p, B) = {h : E → E | h homeo, p ◦ h = h}.
Nπ(B,b0 ) (H0 )
C(E, p, B) ∼
= H0 , where H0 = p∗ π(E, e0 ).
When p : E → B is regular (H0 is normal in π(B, b0 )), then
π(B, b0 )
C(E, p, B) ∼
= .
H0
In particular when E is simply connected,
C(E, p, B) ∼
= π1 (B, b0 ).
p : E → B is rigular iff (for all e1 , e2 ∈ E such that p(e1 ) = p(e2 ) ⇒ ∃h ∈

C(E, p, B) : h(e1 ) = e2 )
Now let’s talk about group actions. Let X be a space and G ≤ Homeo(X).
G acts on X: gx = g(x). We can consider X/G, the quotient space consisting
of all orbits of this action with the quotient topology. Let π : X → X/G denote
the natural projection: x 7→ Gx. So O ⊂ X/G is open iff π −1 (O) is open in X
Definition 18. G acts properly discontinuously (nice) on X iff for all x ∈ X,

there exists neighborhood U (x) open such that gU ∩ U ̸= 0 ⇒ g = 1, or in
other words, g1 U ∩ g2 U ̸= ∅ ⇒ g1 = g2 .
Remark. In the literature: many definition of non-equivalent properly dis-

continuous.

Lecture 6
Remark. If G acts properly discontinuously, then G acts freely, i.e. gx =

x ⇒ g = 1, i.e. there are no fixed points.
Theorem 17 (76.5, important!). Let G ≤ Homeo(X). Then π : X → X/G

is a covering map iff G acts properly discontinuously. Moreover, when π
is a covering, then it is a regular covering and G is a the group of covering
transformations.
a not isomorphic to, it is
Proof. Remark that π is an open S map. Indeed, let U ⊂ X be open. Then

π(U ) is open, since π −1 (π(U )) = g∈G g(U ) is open, as g is homeo.
Step 1. Assume G acts properly discontinuously on X. We have to show

that π is a covering map. Let x = π(x) be an element of X/G.
Let U (x) be an open such that gU ∩ U = ∅ if g ̸= 1 (definition of
properly discontinuous action) π is anSopen map, so V = π(U ) is
an open containing x, and π −1 (V ) = g∈G gU . We want to show
that V is evenly covered. Take as slices gU , which are disjoint since
g1 U ∩ g2 U = ∅ if g1 ̸= g2 . Consider π : U → V
• Continuous
• Bijective (surjective, injective)
• Open,
which means it’s an homeomorphism. Of course, π : gU → V is also

a homomorphism, as it is a composition of g −1 and π, both of which
are homeo. So we have a covering map
Step 3. Suppose π is a covering.
• Certainly, G is a subgroup of the group of covering transforma-
tions.
• Let h ∈ C(x, π, X/G) and x1 ∈ X. Then πh(x1 ) = π(x1 ), so
the orbit of h(x1 ) is the orbit of x1 , so there exists a g ∈ G such
that g(x1 ) = h(x1 ). Both of them are covering transformations,
g(x1 ) = h(x1 ). From previous results, we conclude that g = h.
Therefore G = C(X, π, X/G).
• Let x1 , x2 ∈ X with π(x1 ) = π(x2 ), then there exists a g ∈ G
such that g(x1 ) = x2 , so π a regular covering.
Step 2. Reverse implication Suppose X → X/G is a regular covering.
We have to show that the action is properly discontinuous. Let x ∈
X. Take x = π(x), and let V be a pathF connected open containing x,
which is evenly covered. π −1 (V ) = α∈I Uα , and π|Uα : Uα → V is
homeo. Let U be the slice Uα which contains x. Then U is an open
containing x, and we claim this is ‘the good one’ for the definition of
properly discontinuous. Suppose u ∈ g(U ) ∩ U , so u is an element of
U and also an element of the form gu′ for some u′ ∈ U . This implies

Lecture 6
π(u′ ) = π(gu′ ) = π(u). Since π|U is a homeomorphism, u′ = u. This

shows that g(u) = u. By Step 3, g is a covering transformation.
Therefore g is the identity map, because that’s the only covering
transformation that has a fixed point.
Corollary. Let X be simply connected and G ≤ Homeo(X) acting properly

discontiously. Then π : X → X/G is a regular covering and π1 (X/G) =
C(x, π, X/G) = G.
Remark (Exercise, Thm 76.6). If p : X → B is a regular covering and

G = C(E, p, B), then B is homeomorphic to X/G.
If B admits a universal covering space, then B = E/G where E is
the universal covering space space and G ∼ = π(B, b0 ). So if a space has a
universal covering space, it’s always of the form E/G.
Example. Let α : R2 → R2 : (x, y) 7→ (x + 1, y). Let γ : R2 → R2 : (x, y) 7→

(x, y + 1). Let G = grp{α, γ}, an Abelian group. Every element of G is of
′ ′
the form αk γ ℓ (x, y) = (x + k, y + ℓ), so αk γ ℓ = αk γ ℓ ⇔ k = k ′ , ℓ = ℓ′ , so
G∼ = Z2 .
G acts properly discontinuously on R2 . Therefore π1 (R2 /Z2 , x) ∼ = Z2 .
2 2
R /Z is the torus.
Example. Take α : R2 → R2 : (x, y) 7→ (x + 1, y). Take β : R2 → R2 :

(x, y) → (−x, y + 1). Note that β 2 = γ 2 . Claim: α ◦ β = β ◦ α−1 . Note
that α−1 β = βα, αβ −1 = β −1 α−1 and α−1 β −1 = β −1 α, so we can always
move α to the front, which means that
G = grp{α, β} = {αk β l | k, l ∈ Z}.
note
αk β ℓ (x, y) = ((−1)ℓ x + k, y + ℓ).
′ ′
Therefore αk β l = αk β l ⇔ k = k ′ ∧ l = l′ . As a set G = Z2 , but not
isomorphic.
Claim: G acts properly discontinuously on R2 . Indeed consider B(x, 41 ).
So
π1 (R2 /G) = G = α, β | βα = α−1 β .


Lecture 6
This is the Klein bottle!
Remark. These two spaces are the only possible spaces you can create using
affine actions. Projective plane? Not properly discontinuously?

Chapter 13
Singular Homology
This is the most general homology.

• A point is a 0-simplex
• A line segment is a 1-simplex
• A triangle (including interior) is a 2-simplex
• A filled in tetrahedron is a 3-simplex.
Definition 19 (Simplex). In general a p-simplex is the convex hull of p + 1

points in general position in Rn (affine independent)
Proposition 3 (1.1). Let x0 , . . . , xp ∈ Rn , (p ≥ 1) then the following are

equivalent:
• The vectors x1 − x0 , . . . , xp − x0 are linearly independent

Pp Pp Pp P
• If i=0 si xi = i=0 ri xi and i=0 si = ri , then si = ri for all
i = 0, . . . , p.
Pp Pp Pp P
Proof. • Assume that i=0 si xi = i=0 ri xi and i=0 si = ri .
Then we can multiply the last expression by x0 . Then
p
X p
X
si (xi − x0 ) = ri (xi − x0 )
i=0 0
Xp p
X
si (xi − x0 ) = ri (xi − x0 ).
i=1 1
Therefore for all i = 1, . . . , p ⇒ si = ri and therefore also s0 = r0 .

Pp Pp Pp
• Assume i=1 ri (xi − x0 ) = 0. Then i=1 ri xi = i=1 ri x0 . This
implies that r0 = 0, r1 = . . . = rp = 0 and s1 = . . . = sp = 0.
49
Lecture 6
Proposition 4. Let x0 , . . . , xp satisfy the conditions of the proposition.

Then any point x in the convex hull of these points has a unique rep-
resentation of the form
X X
x= t i xi ti ≥ 0 and ti = 1.
This convex hull is the p-simplex spanned by x0 , . . . , xp .
Proof. Exercise. Hint: consider p = 0, p = 1, and do the rest by induction.
Definition 20. The standard p-simplex σp is the (ordered) p-simplex in

Rp+1 spanned by
x0 = (1, 0, 0, . . . , 0)
x1 = (0, 1, 0, . . . , 0)
xp = (0, 0, . . . , 0, 1).
X
σp = {(t0 , . . . , tp ) | ti > 0 and ti = 0}.
Remark. Any p-simplex is homeomorphic to the standard p-simplex.
h : σp −→ p-simplex
X
(t0 , . . . , tp ) 7−→ t i xi .
This is a homomorphism because this is a continuous map between Haus-

dorff, compact spaces.
Definition 21. A singular p-simplex in a topological space X is a continuous

map
ϕ : σp −→ X.
In the case of p = 2, we get the following:
Note that in the case of p = 0, ϕ : σ0 = {x0 } → X : x0 7→ ϕ(x0 ) ∈ X. So
CHAPTER 13. SINGULAR HOMOLOGY 50

Lecture 6
zero-simplices are identified with points of X.

In the case of p = 1, we get σ1 = I, so a singular 1-simplex is a path in X.
For the moment, we don’t assume spaces are connected, etc.
Definition 22. Let X be a topological space. Then Sn (X) is the free abelian
group on all singular n-simplices. This is a giant group. ByPthe definition
of a free abelian group, elements of Sn (X) are of the form nϕ ϕ, where
• the sum is finite,

• nϕ ∈ Z,
• ϕ : σn → X.
Let ϕ : σp → X be a singular p-simplex, and i ∈ {0, . . . , p}. Then the i-th
boundary of ϕ, denoted by ∂i ϕ is the singular p − 1 simplex.
∂i ϕ : σp−1 → X : (t0 , . . . tp−1 ) 7→ ϕ(t0 , t1 , . . . , ti−1 , 0, ti , . . . , tp−1 ).
This is ϕ restricted to the face opposite to xi .
x2
ϕ ∂σ2
∂ 1 σ2 ∂0 σ2
− +
x0 σ2
+ ∂2 σ2
x1
Figure 13.1: Boundary operator
Now, define
X X
∂ : Sn (X) → Sn−1 (X) : nϕ ϕ 7→ nϕ (∂ϕ),
where X
∂ϕ = (−1)i ∂i ϕ ∈ Sn−1 (X).
Then certainly, ∂ is a morphism of abelian groups. In the case of the figure, we
get
∂ϕ = ∂0 ϕ − ∂1 ϕ + ∂2 ϕ,
which really does goes around the simplex in the correct direction. The (−1)i
fixes the direction of the simplices.
Example. Let ϕ : σ1 → X be a 1-simplex. Then ∂ϕ = ϕ(x1 ) − ϕ(x0 ).
Example. Let ϕ be a two simplex. Then ∂(∂ϕ) = 0.

Lecture 7
Remark. We define S−n = 0 for n ≥ 0. Then ∂ : S0 (X) → S−1 (X) = 0 :

ψ 7→ 0.
Proposition 5 (1.3). ∂ ◦ ∂ = 0.
Proof. Exactly the same thing as in the example.

Let X be a topological space.
∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂
−
→ Sn+1 (X) −
→ Sn (X) −
→ Sn−1 (X) −
→ ··· −
→ S1 (X) −
→ S0 (X) −
→0−
→0−
→ ··· .
We do know that ∂ ◦ ∂ = 0. Suppose we start with α ∈ Sn+1 . Then ∂α belongs

to the kernel of the “next” ∂. This shows that the image of ∂ is a subset of the
kernel of the “next” ∂. This holds at any spot. We call this a chain complex of
abelian groups. Define
Zn (X) = the n-cycles of X = kernel of ∂ : Sn → Sn−1

Bn (X) = the n-boundaries of X = the image of ∂Sn+1 → Sn .
We claimed that Bn ⊂ Zn .
Zn (X)
Definition 23. The n-th singular homology group of X is Hn (X) = Bn (X) .
Two cycles are the same if they differ with a boundary.
A cycle is more or less a loop. Then two cycles are the same when they differ
with a boundary, in this case the boundary of the triangles indicated with the
dashed lines.
Figure 13.2: Intuition Homology

Lecture 7
Lecture 7
di 19 nov 10:27
Consider the more general situation.
Definition 24. Chain complex:

∂n+1 n ∂
C∗ : · · · → Cn+1 −−−→ Cn −→ Cn−1 → . . . ,
where Cn abelian group, ∂n group morphism and ∂n ◦ ∂n+1 = 0
Definition 25. A chain map, Φ = {ϕi }i∈Z : C∗ → D∗ and each ϕn : Cn →

Dn .
∂n+1 n ∂
C∗ : · · · →Cn+1 −−−→ Cn −→ Cn−1 → . . .
↓ ϕn+1 ↓ ϕn
′
∂n+1 n ∂′
D∗ : · · · →Dn+1 −−−→ Dn −→ Dn−1 → . . .
.
where ϕn ◦ ∂n = ∂n′ ◦ ϕn
Define Zn = Ker ∂n ≤ Cn and Bn = Im ∂n+1 ≤ Cn and Bn ≤ Zn .
Then we define
Zn (C∗ )
Hn (C∗ ) = .
Bn (C∗ )
Definition 26. A sequence is exact if not only Bn ≤ Zn , but Bn = Zn . In

other words: the homology group at n is the trivial group. The homology
group measures the ‘exactness’ of a chain complex.
Given a chain map Φ, it wil induce
Φ∗ : Hn (C∗ ) −→ Hn (D∗ )
z + Bn (C∗ ) 7−→ ϕn (z) + Bn (D∗ ).
We claim that this is well-defined. Indeed, first we check that ϕn (z) is a cycle.
n ∂ ϕn−1
Indeed, using the commutativity of the diagram, z −→ 0 −−−→ 0 implies that
′
∂n
ϕn (z) −→ 0. Also, a boundary stays a boundary by the same reasoning.
We apply this in connection with a continuous map f : X → Y . Consider
the following:

Lecture 7
X Y
ϕ f
f# (ϕ) = f ◦ ϕ
Then we can extend f# to a map from Sn (X) → Sn (Y ). Then
∂ ∂
Sn+1 (X) ϕ ∈ Sn (X) ∂ϕ ∈ Sn−1
f# f# f#
∂ ∂
Sn+1 (Y ) f# (ϕ) ∈ Sn (Y ) Sn−1
Now, the question is, is ∂f# (ϕ) = f# (∂ϕ). It’s enough to check this for ∂i :
(f# (∂i ϕ))(t0 , . . . , tn−1 ) = f (ϕ(t0 , . . . , ti−1 , 0, ti , . . . , tn−1 ))

∂i (f# ϕ)(t0 , . . . , tn−1 ) = f (ϕ(t0 , . . . , ti−1 , 0, ti , . . . , tn−1 )),
which proves that we can interchange f# and δ.

So f induces a map on Homology:
f∗ : Hn (X) −→ Hn (Y )
z + Bn (x) 7−→ f# (z) + Bn (Y ),
or, using different notation:
f∗ : Hn (X) −→ Hn (Y )
⟨z⟩ 7−→ ⟨f# (z)⟩ .
It’s clear that
(f ◦ g)∗ = f∗ ◦ g∗ (1X )∗ = 1Hn (X) ∀n ∈ N,
so this is a functor.
Theorem 18. If f : X → Y is a homeomorphism, then f∗ is an isomorphism

for all n
Example. What are the homology groups of the space consisting of one
point, X = {x0 }. Note that ∀n ∈ N, there is exactly one singular n-
simplex, ϕn : σn → X : (t0 , . . . , tn ) 7→ x0 . Sn is the free abelian group on
ϕn ∼
= Z for all n.
∂ : S0 (X) → S−1 (X) : ϕ0 7→ 0.

Lecture 7
for n > 0, we have
∂ : Sn (X) → Sn−1 (X)

(
X X ϕn−1 n even
ϕn 7→ (−1)i ∂i ϕn = (−1)i ϕn−1 =
0 n odd.
So all odd maps are 0 and if odd, then it maps the generator of Sn to
the generator of Sn−1 :
0 1Z 0 1Z 0 0 0
S2n+3 S2n+2 S2n+1 S2n S2n−1 S1 S0 0
Z Z Z Z Z Z Z 0
Now,
Z0 (X) Z ∼
H0 (X) = = =Z
B0 (X) ⟨0⟩
Z
H2n+1 (X) = ∼=0
Z
0
H2n (X) = ∼ = 0.
0
Example. The other extreme example is computing H0 of an arbitrary

space, X. This is always possible. We’ll assume that X is path connected.
∂ ∂
S1 (X) S0 (X) 0
P clear that S0 (X) = Z0 (X), which consists of elements of the form

It’s
ni xi , where xi ∈ X, a zero P
simplex. P
Define α : Z0 (X) → Z : ni xi 7→ ni . Then α is a morphism of
abelian groups. Also Im α = Z (as zx0 7→ z). This would imply that
Z0 ∼ Z0 ∼
ker α = Z. Claim: B0 (X) = ker α, which would imply that B0 = H0 = Z.
Proof:
• B0 ⊂ ker α. Take ϕ : σ1 → X. Then ∂ϕ ∈ S0 and ∂ϕ = ∂0 ϕ −
∂1 ϕ = ϕ(1) − ϕ(0). Then α(∂ϕ) = 1 − 1 = 0. Since B0 consist of
linear combinations of things like ∂ϕ, we get that α(B0 (ϕ)) = 0, so
B0 ⊂ ker α.
P
• Take any element ni xi ∈ ker α. Then we have to showP that this
element is a boundary. Belonging to the kernel means that ni = 0.
Choose any point x0 ∈ X, fix one. For each i, choose a singular one
simplex, a path ϕ : σ1 → X, with starting point x0 and end point xi .
Here we use the assumption that X is path-connected. Then
X X X X X X
∂( ni ϕi ) := ni ∂ϕi = ni (xi −x) = ni xi − ni x = ni xi ,

Lecture 7
so this element,
P which we’ve assumed lied in the kernel is also a
boundary, i.e. ni xi ∈ B0 (X).
What happens if our space is not path-connected?
Suppose C∗ is a chain complex and there exists an I such that for all n,
M
Cn = Cni ,
i∈I
∂i , where for all i ∈ I, ∂i : Cni → Cn−1

i
L
and ∂ : Cn → Cn−1 is of the form ∂ = .
If this is the case, we have that ∂i ◦ ∂i = 0. Then it’s easy to show that
M
Hn (C∗ ) = Hn (C∗i ).
F
Let X be a space such that X = α∈A Xα , where Xα are the path compo-
nents of X. Now, let ϕ : σp → X be any singular p-simplex.
L Then ϕ(σp ) ⊂ Xα
for exactly one α. It’s not so hard to see that Sn (X) = Sn (Xα ). Moreover,
∂(Sn (Xα )) ⊂ Sn−1 (Xα ), which means that ∂ is a direct sum. From this, it will
follow that M
Hn (X) = Hn (Xα ).
This shows that
H0 (Xα ) ∼
L L
Proposition 6. H0 (X) = α = αZ
Theorem 19. Let X ⊂ Rn be a convex subset and assume that X ̸= ∅.

Then H0 (X) ∼ = Z, because a convex set is path connected, and Hp (X) = 0
for all p > 0.
Proof. We will define a map going from Sn (X) → Sn+1 (X). Fix x ∈ X.
Take ϕ : σn → X, and we’ll build up a θ : σn+1 → X.
Now consider the purple point, we can write is as a convex combination
of the blue and the red point:

t1 t2 tp+1
t0 (1, 0, . . . , 0) +(1 − t0 ) 0, , ,..., .
| {z } 1 − t0 1 − t0 1 − t0
| {z }
Then we define:

t1 t2 tp+1
θ(t0 , t1 , . . . , tn+1 ) = t0 x + (1 − t0 )ϕ , ,...,
1 − t0 1 − t0 1 − t0
and if t0 = 1, we simply define θ to be x.

We need to prove that θ is continuous.
∥θ(t0 , t1 , . . . , tn+1 ) − x∥ = ∥(1 − t0 )ϕ(. . .) + t0 x − x∥

= ∥(t − t0 )ϕ(. . .) − (1 − t0 )x∥
≤ |(1 − t0 )|(∥ϕ(. . .)∥ + ∥x∥).
Now, the second term is a constant. As σn is compact, ϕ(σn ) is also

Lecture 7
compact, so ∥ϕ(σn )∥ ≤ M . So we have
0 t →0
≤ |(1 − t0 )|(M + C) −− −→ 0.
θ x
v0
originally: v0
v2
ϕ
v1
originally: v0
This means we have a map T : Sn (X) → Sn+1 (X), which is a mor-

phism, as we defined it for all generators, T (ϕ) = θ. Have a look at
the drawing and conclude that ∂0 (T ϕ) = ϕ. Also ∂1 T ϕ = T ∂0 ϕ and
∂2 T ϕ = T ∂1 ϕ. In general, we claim that ∂i (T (ϕ)) = T (∂i−1 ϕ), for any
n-simplex with n > 0. Forgetting about t0 = 1 (not a problem, consider
this case separatly)
(∂i T (ϕ))(t0 , . . . , tn ) = θ(t0 , t1 , . . . , ti−1 , 0, ti , . . . , tn )

t1 ti−1 ti tn
= (1 − t0 )ϕ ,..., , 0, ,..., + t0 x
1 − t0 1 − t0 1 − t0 1 − t0
= T (∂i−1 ϕ)(t0 , t1 , . . . , tn ).
This does not work if we’re working with a 0-simplex. (Because then
∂i−1 becomes zero, and the only homomorphisms that we have are the
identity, so . . . )
Now, consider the total boundary,
n
X
∂T ϕ = ∂0 T ϕ + (−1)i ∂i T ϕ
i=1
n
X
=ϕ+ (−1)i T ∂i−1 ϕ
i=1
n−1
X
=ϕ+ (−1)j+1 T ∂j ϕ
j=0
 
n−1
X
=ϕ+T  (−1)j+1 ∂j ϕ
j=0
= ϕ − T (∂ϕ).

Lecture 7
Have a look at the drawing to see this. Conclusion: ϕ = (∂T + T ∂)ϕ.

For all n > 0. Let ⟨z⟩ ∈ Hn (X), with z ∈ Zn (X). Then z = (∂T +T ∂)z,
or z = ∂T z + T ∂z. Because z is a cycle, z = ∂(T z), so z is not only a
cycle, but also a boundary. Hence ⟨z⟩ = ⟨0⟩, so Hn (X) = 0.
Remark. The appearance of ∂T + T ∂ is a rather general phenomenon.

Suppose we have two chain complexes.
∂ ∂
Cn+1 Cn Cn−1
gn+1 fn+1 gn fn gn+1 fn−1
T T
∂ ∂
Dn+1 Dn Dn−1
If f − g = ∂T + T ∂ for some T for some T : Cn → Dn+1 for all n,
where f and g are chain maps. Then f∗ = g∗ : Hn (C∗ ) → Hn (D∗ ). When
f − g = ∂T + T ∂ is satisfied, we say that f and g are chain homotopic.
This indeed has something to do with homotopic maps.
If you have two maps that are homotopic, then f# and g# will be chain
homotopic maps.
There is a proof of the theorem in the book: long, not difficult. There
is a better proof, see problem 8.
Idea of the proof:

Lecture 7
g∼f g
ϕ
Tϕ
P
≈ σ3
∂T ϕ T ∂ϕ
g# (ϕ) f# (ϕ)
This goes from two dimensional stuff to three dimensional stuff. But the
prism on the left is not a simplex. But we can divide it into simplices. Now, we
look at ∂T ϕ + T ∂ϕ, and we see that (being careful with signs) is g# (ϕ) − f# (ϕ).
Conclusion: If f ∼ g : X → Y , then for all n ≥ 0, we have that f∗ = g∗ :
Hn (X) → Hn (Y ).
Theorem 20. If X and Y have the same homotopy type, then they also
have isomorphic homology groups.
Proof. Let f : X → Y be the homotopy equivalence and g : Y → X be the

homotopy inverse, so f ◦ g ≃ 1Y and g ◦ f ≃ 1X . Then f∗ ◦ g∗ = 1Hn (Y )
and f∗ ◦ g∗ = 1Hn (X) . This means that f∗ is an isomorphism with inverse
g∗ .
As we saw with homotopy groups, retracts and deformation retracts can be
useful.
Recall:
Definition 27. Let A ⊂ X. Then A is a retract of X iff there exists a map

r : X → A such that r|A = 1A .

Lecture 9
We make the definition of deformation retract bit more general
Definition 28 (deformation retract (version 2)). A is a deformation retract

iff there is a retract r : X → A such that r ≃ 1X . So there exists rt such
that r0 = r and r1 = 1. a
a And not necessarily rt |A = 1A for all t.
Remark. We sometimes call this a weak deformation retract
Proposition 7 (1.12). • Let A be a retract of X. Then i∗ : Hn (A) →

Hn (X) (with i : A → X) is injective and moreover, i∗ (Hn (A)) is a
direct summandof Hn (X)! So there exists a subgroup B ≤ Hn (X)
such that Hn (X) = i∗ Hn (A) ⊕ B.
• If A is a deformation retract of X, then i∗ is an isomorphism.
Proof. • The last bullet point is trivial. This follows from induced
maps from homotopic maps.
• Let r : X → A be a retraction. Then we have
i r
A X A
r◦i=1A
So (r ◦ i)∗ = (1A )∗ = 1Hn (A) for all n. Therefore, i∗ is injective.

(Note that this proof holds for all functors!)
Claim: Hn (X) = Im(i∗ ) ⊕ ker r∗ .
– Let α ∈ Im(i∗ ) ∩ ker r∗ . We show that α = 0. So α = i∗ (β) and

r∗ (α) = 0. This means that r∗ (i∗ (β)) = 0, but r∗ ◦ i∗ = 1, so
β = 0. So α = i∗ β = 0.
– We show that any element can be written as a sum of elements
of both groups. Let γ ∈ Hn (X). Then
γ = i∗ (r∗ (γ)) +γ − i∗ (r∗ (γ)),

| {z }
∈Im i∗
and
r∗ (γ − i∗ r∗ γ) = r∗ γ − r∗ i∗ r∗ γ
= r∗ γ − r∗ γ = 0.
Lecture 8: Ill
di 03 dec 23:02
Lecture 9
di 10 dec 10:31
Recap:

Lecture 9
inj
0 → H1 (S 1 ) = Z −−→ H0 (U ∩ V ).
and ⟨c + d⟩ 7→ x − y.
x y V
U
Figure 13.3: homology of circle
For S n . Then we had

∆,∼
= r
0 → Hn (S n ) = Z −−−→ Hn−1 (U ∩ V ) −→
∗
(S n−1 ) = Z.
Corollary. There is no retraction of Dn = B n = n-dimensional disc to

S n−1 .
Proof. Suppose there is a retract. Then
1S n−1
S n−1 S n+1
i r
Dn
(1S n )∗ =1Hn−1
Z = Hn−1 (S n−1 ) Hn−1 (S n−1 ) =Z
i∗ r∗
Hn−1 (Dn ) = 0
Corollary (Brouwers fixed point theorem). Let n ≥ 1. Then any continuous

map f : Dn → Dn has a fixed point. (Note: this time, for any dimension!)

Lecture 9
Maps from spheres to spheres

Let f : S n → S n . Then f induces a map f∗ : Hn (S n ) → Hn (S n ), so f∗ : Z → Z,
or we can write f∗ : ⟨α⟩ 7→ m ⟨α⟩, where α is the generator.
Definition 29. The degree of f is m. We write d(f ) = m.
Proposition 8. Few properties of the degree:

1. d(Id) = 1
2. d(f ◦ g) = d(f )d(g)
3. f constant, or if f is not surjective, then d(f ) = 0.

4. If f ≃ g, then d(f ) = d(g), because they induce the same map
(Problem 8) (Deep theorem: f ≃ g iff d(f ) = d(g).)
5. If f is a homotopy equivalence, then d(f ) = ±1. (We have a homo-
topy inverse, and f∗ ◦ g∗ = 1?, so the only possibilities are both 1 or
−1.
Proof. We prove 3. Suppose p ̸∈ Im f . Then

f
Sn Sn
′
f
i
S n \ {p}
Then
f∗
Hn (S n ) Hn (S n )
f∗′
i∗
Hn (S n \ {p}) = 0
where f ′ is f restricted to the image. We see that f∗ = 0.
Theorem 21. Let S n = {(x1 , . . . , xn+1 ) ∈ Rn |

P 2
xi = 1} and f : S n →
n
S
f : (x1 , x2 , . . . , xn+1 ) 7→ (−x1 , x2 , . . . , xn+1 ).
Then d(f ) = −1.
Proof. By induction on n.
Base case, n = 1. S 1 = {(x1 , x2 ) | x21 + x22 = 1}.

Lecture 9
z
c
x y
d
z′
Figure 13.4: base case of theorem
Note that f (x) = y, f (y) = x. Denote U = S \ {z ′ }, V = S \ {z}. And

f (U ) ⊂ U and f (V ) ⊂ V .
Then
∆
0 H1 (S) H0 (U ∩ V )
f∗ (f |U ∩V )∗
∆
0 H1 (S) H0 (U ∩ V )
TODO: right arrows tails

Then
(f |U ∩V )∗ (x − y) = f (x) − f (y) = y − x = −(x − y).
So ∆ is injective, we have that f∗ (⟨α⟩) = − ⟨α⟩, so d(f ) = −1.
Now, assume that the theorem holds for S n−1 .

Define U , V in a similar way. f (U ) ⊂ U and f (V ) ⊂ V . Again, we
have a commutative diagram.
Question: is it true that f∗ (⟨α⟩) is just − ⟨α⟩?
Now, we know that the f∗ on the right is simply ·(−1). Then, going
the whole way around,
f∗ (⟨α⟩) = ∆−1 (i∗ (−1(r∗ (∆(α)))).
This is
−∆−1 (i∗ r∗ ∆(⟨α⟩)).
As i∗ and r∗ are inverses, we have that d(f ) = −1.
Suppose f : S n → S n . Then there is a map Σf : S n+1 → S n+1 (the
suspension of f ), given by

Lecture 9
S n+1 Σf S n+1
f
Sn Sn
Σf
Σf
Figure 13.5: definition of suspension
Using formulas:
S n+1 = {(x, t) : x ∈ Rn+1 , t ∈ R, ∥x∥2 + t2 = 1}.
Then
Σf : S n+1 −→ S n+1
(
(x, t) if x = 0, north and south pole .
(x, t) 7−→ x
∥x∥f ( ∥x∥ ), t if x ̸= 0
Not too difficult to prove that Σf is continuous. Claim: d(f ) = d(Σf ).

Proof. Exactly the same technique as we used to prove the previous theo-
rem.
Remark. Suspension is a general technique. Let X be a topological space.

Then ΣX, the suspension of X is defined as follows:
[−1, 1] × X
Figure 13.6: definition of suspension general
ΣX = X × [−1, 1]/ ∼ with
(x, t) ∼ (x′ , t′ ) ⇔ t = t′ = 1 or t = t′ = −1.

Lecture 9
Check for yourself: ΣS n = S n+1 (pointy sphere).
Corollary. If fi (x1 , . . . , xn+1 ) = (x1 , . . . , −xi , . . . , xn+1 ). Then d(f ) = −1.
Proof. Let h : S n → S n be the map that exchanges the first coordinate and
the ith coordinate. Then h is a homeomorphism: h−1 = h, so h ◦ h = 1S n ,
so d(h) = ±1 Then fi = h ◦ f ◦ h. So
d(fi ) = d(h)2 d(f ) = (±1)2 (−1) = −1.
There is a big difference between even and odd dimensional spheres:
Corollary. Let A : S n → S n defined by
A(x1 , . . . , xn+1 ) = (−x1 , . . . , −xn+1 ).a
Then d(A) = (−1)n+1 .

a Note the error in the book. A should have n + 1 inputs. TODO
Proof.
d(A) = d(f1 )d(f2 ) · · · d(fn+1 ) = (−1)n+1 .
So for even dimensional spheres, d(A) = −1 and for odd dimensional sphere,
d(A) = 1. This is a big difference.
Theorem 22. Let f : S n → S n and g : S n → S n . When f (x) ̸= g(x) for

all x ∈ S n . Then f ≃ A ◦ g.
f (x)
Sn
Ag(x)
Figure 13.7: proof prop 123

Lecture 9
Proof. The line does not go through the center, as f (x) ̸= g(x). Define a
homotopy,
(1 − t)f (x) + tAg(x)
(x, t) 7→ .
∥(1 − t)f (x) + tAg(x)∥ =
̸ 0
This is a homotopy between f and Ag
Not in the text:
Corollary. If |d(f )| =
̸ |d(g)|, then there is a point x such that f (x) = g(x).
Proof. The previous theorem says that d(f ) = d(A)d(g), so |d(f )| = |d(g)|.
You can even make this better by making a distinction between odd and
even dimensional spheres.
Corollary. If |d(f )| =
̸ 1, then f has a fixed point, i.e. an x such that f (x) =
x = Id(x).
Proof. If |d(f )| =
̸ d(Id) = 1, then . . .
Theorem 23. Let f : S 2n → S 2n . Then there exists an x in S 2n such that

f (x) = x or f (x) = −x.
Proof. Suppose there is no x such that f (x) = x and f (x) = −x.
• Since f (x) ̸= x for all x ∈ X, we have that f ≃ A.

• Since f (x) ̸= −x = A(x) for all x ∈ X, we have that f ≃ A ◦ A = 1.
From this it follows that
d(f ) = (−1)2n+1 = −1 d(f ) = 1,
which is a contradiction.
Remark. This does not hold for odd dimensional spheres. For example,
consider S1 , and make a small rotation, then no point is mapped onto
itself or onto its antipodal point. Doing the same for S2 , we see that when
we do a small rotation, the north and south pole are switched.
Remark. It’s harder to show that for odd dimensional spheres, there always
exist such a map.
Now, we prove the hairy ball theorem, which only holds in even dimensional
spheres.
Corollary. There is no continuous map f : S 2n → S 2n such that x and f (x)

are orthogonal for all x.

Lecture 9
Proof. By the previous theorem, there is a point for which f (x) = ±x, and
x ̸⊥ ±x.
Corollary. There exists no nonzero vector field ϕ on S 2n .
Proof. Suppose ϕ does exists. Then, map the tangent vector to a point on
the sphere:
ϕ(x)
ψ : S 2n → S 2n : x 7→ .
∥ϕ(x)∥
x
f (x)
Figure 13.8: hairy ball theorem
Remark. For odd dimensional spheres, there always exists a such a vector
field. We even know how many linearly independent vector fields exist!
Torus
Goal: Hn (T 2 ).

Lecture 9
a
V
c1
p
b c2 b
Figure 13.9: homology of torus
and , and U ∩ V = B 2 \ {p}.
• Choose U = T 2 \ {p} Then U retracts to the figure eight space (boundary

of square) Last week, we saw that H0 (U ) = Z and H1 (U ) = Z ⊕ Z, with
generators a and b, and Hn (U ) = 0 for n ≥ 2.
• V = B 2 . As B 2 is path connected, H0 (U ) = Z and Hn (U ) = 0 for all
n≥1
• U ∩V = B 2 \{p}, which retracts to a circle, so H0 (U ∩V ) = H1 (U ∩V ) = Z.
H2 (U ) ⊕ H2 (U ) → H2 (T 2 ) → H1 (U ∩ V ) → H1 (U ) ⊕ H1 (U ) → H1 (T 2 ) → H0 (U ∩ V ) → H0 (U ) ⊕ H0 (U ) → H0 (T 2 )
So we have
→ 0 ⊕ 0 → H2 (T 2 ) → Z → Z2 ⊕ 0 → H1 (T 2 ) → Z → Z ⊕ Z → Z → 0.
Now, Z → Z ⊕ Z is injective, as the kernel of Z ⊕ Z → Z is non-trivial. (1

gets mapped to something non-trivial) So we can shorten the sequence
g∗
→ 0 ⊕ 0 → H2 (T 2 ) → Z −→ Z2 ⊕ 0 → H1 (T 2 ) → 0.
Now, we want to know how g∗ works. We need a generator of Z and look at its
embedding? Then Z = ⟨c1 + c2 + c3 + c4 ⟩. Then
i r
U ∩V U figure 8
ir r∗
H1 (U ∩ V ) H1 (U ) H1 (figure 8)
⟨c1 + c2 + c3 + c4 ⟩ ⟨c1 + c2 + c3 + c4 ⟩ ⟨r ◦ c1 + r ◦ c2 + r ◦ c3 + r ◦ c4 ⟩

a+b+a+b

Lecture 9
Now, as these are all cycles, we have that this is

⟨a⟩ + b + ⟨a⟩ + ⟨b⟩ .
Now, using Problem 6, we find that this is
− ⟨a⟩ − ⟨b⟩ + ⟨a⟩ + ⟨b⟩ = 0.
Therefore, g∗ is the zero map.
0
→ 0 ⊕ 0 → H2 (T 2 ) → Z −
→ Z2 ⊕ 0 → H1 (T 2 ) → 0,
so the kernel of Z2 ⊕ 0 → H1 (T 2 ) is an injective map. As H1 (T 2 ) → 0, this
implies that Z2 ⊕ 0 → H1 (T 2 ) is a surjective map. Therefore, H1 (T 2 ) = Z2 .
Now, H2 (T 2 ) → H1 (U ∩ V ) is surjective, and it is also injective, so H2 (T 2 ) = Z.
Now, for n ≥ 3. Then
Hn (U ) ⊕ Hn (V ) → Hn (T 2 ) → Hn−1 (U ∩ V )..
This simplifies to
0 ⊕ 0 → Hn (T 2 ) → 0.,
so Hn (T 2 ) = 0. Conclusion:
H0 (T 2 ) = Z H1 (T 2 ) = Z2 H2 (T 2 ) = Z Hn (T 2 ) = 0 for all n >= 3.
Remark. H2 (S 1 × S 1 ) ̸= H2 (S 1 ) × H2 (S 1 ).
n
Remark. Hn (T k ) = Z(k ) , where n

k = 0 when k > n.
Klein bottle
a
V
c1
p
b c2 b
Figure 13.10: klein bottle homology

Lecture 9
• U = K \ {p}, which retracts to the figure eight space (boundary of the

square) So H0 = Z and H1 = Z 2 = grp(⟨a⟩ , ⟨b⟩), and Hn = 0 for n ≥ 2.
• V = B 2 So H0 = Z and Hn = 0 for n ≥ 1.
• U ∩ V = B 2 \ {p}, which retracts to a circle. So H0 = Z and H1 = Z =
grp(⟨c1 + c2 + c3 + c4 ⟩)
g∗
→ H2 (U ) ⊕ H2 (V ) → H2 (K) → H1 (U ∩ V ) −→ H1 (U ) ⊕ H1 (V ) → H1 (K) → H0 (U ∩ V ) → H0 (U ) ⊕ H0 (V ) → H0 (K) → 0
Klein bottle is path connected, so
g∗
→ 0 ⊕ 0 → H2 (K) → Z −→ Z2 ⊕ 0 → H1 (K) → Z → Z ⊕ Z → Z → 0
We can cut of the sequence because all spaces are path connected.
g∗
→ 0 ⊕ 0 → H2 (K) → Z −→ Z2 ⊕ 0 → H1 (K) → 0
Now, we have to understand what g∗ is
Then
i r
U ∩V U figure 8
ir r∗
H1 (U ∩ V ) H1 (U ) H1 (figure 8)
⟨c1 + c2 + c3 + c4 ⟩ ⟨c1 + c2 + c3 + c4 ⟩ ⟨r ◦ c1 + r ◦ c2 + r ◦ c3 + r ◦ c4 ⟩

a+b+a+b
So we get
− ⟨a⟩ − ⟨b⟩ + ⟨a⟩ − ⟨b⟩ = −2 ⟨b⟩ .
So g∗ : Z → Z2 : 1 7→ (0, −2), or z → (0, −2z). So, the image of g∗ =
0 × 2Z ≤ Z × Z. The kernel of g∗ is {0}. So g∗ is injective. And 0 × 2Z is the
kernel of h∗
g∗ h
→ 0 ⊕ 0 → H2 (K) → Z −→ Z2 −→
∗
H1 (K) → 0
Now, the zero on the right implies that h∗ is surjective. so (H1 (U ) ⊕ 0)/ Ker h∗ ,
Z2
which is 0×2Z , which is Z ⊕ Z2 .
Now, on the left. As g∗ is injective, so H2 (K) → H1 (U ∩ V ) = Z is the zero
map. On the other hand, we have zeros on the left, so H2 (K) → H1 (U ∩ V ) = Z
is also injective. The only way the zero map can be injective is when the spaces
are zero. Therefore, H2 (K) = 0.
Higher dimensional: trivial.
Let n ≥ 3
0 ⊕ 0 = Hn (U ) ⊕ Hn (V ) → Hn (K) → Hn−1 (U ∩ V ) = 0,

Lecture 10: Ham sandwich theorem
so Hn (K) = 0 for all n ≥ 3.

Conclusion:
H0 (K) = Z H1 (K) = Z ⊕ Z2 H2 (K) = 0 H≥3 (K) = 0.
Remark. Fact: if X is an n-dimensional manifold, which is non orientable,

then H2 is always zero. If it’s orientable, then H2 (X) = Z.
π1 (X)
Remark. Fact: The H1 (X) = [π1 (X),π 1 (X)]
. Abelianized version of funda-
mental group. Remember, we did compute the fundamental group of the
klein bottle! π1 (K) = α, β | βα = α−1 β . Making this group commuta-

tive, we have
H1 (K) = α, β | βα = α−1 β, αβ = βα .

From this is follows that α−1 β = αβ, so α−1 = α, so α2 = 1. Therefore,

we can rewrite this as
H1 (K) = α, β | αβ = βα, α2 = 1 = Z2 ⊕ Z.

Another way is to actually compute commutators.
[α, β] = α−1 β −1 αβ = α−1 α−1 β −1 β −1 = α−2 .
So, we mod out α2 .
Remark (Exam). Hn (RP 2 ). The answer should be:

• H0 = Z (path connected)
• H1 = Z2 (abelianized version)
• H2 = 0 (orientable)
• Hn≥3 = 0

di 17 dec 10:26
1 1
Theorem 24. Let f : S → S such that f is antipode preserving: f (Ax) =
Af (x). Then d(f ) is odd.
Proof. Picture:

c f A◦γ
f (a)
f ◦c
A(a) = b a
d γ
Af (a) = f (Aa) = f (b)
∼ Z = ⟨α⟩ = ⟨c + d⟩. Note

What is f∗ : H1 (S 1 ) → H1 (S 1 )? H1 (S 1 ) =
that d = A ◦ c.
Now, Let’s look at the image of c under f . The only thing we know, is
that it starts at f (a) and ends at f (b) = Af (a). Let γ be the path that is
the half circle, counterclockwise from f (b) to f (a). Then f ◦c+γ ∈ Z1 (S 1 ).
Indeed, ∂(f ◦ c + γ) = f (b) − f (a) + f (a) − f (b) = 0. Then ⟨f ◦ c + γ⟩ ∈
H1 (S 1 ). This has to be equal to m ⟨α⟩ for some m ∈ Z.
Now, A∗ : H1 (S 1 ) → H1 (S 1 ). We know that d(A∗ ) = (−1)2 = 1, so
this is just the identity map:
A∗ (⟨α⟩) = A∗ ⟨c + d⟩ = ⟨A ◦ c + A ◦ d⟩ = ⟨d + c⟩ = ⟨α⟩ .
So
m ⟨α⟩ = A∗ ⟨f ◦ c + γ⟩ = ⟨A ◦ f ◦ c + A ◦ γ⟩ .
Now, A ◦ γ is the path from f (a) to f (b), going counter clockwise. Now,
we assume A ◦ f = f ◦ A, so
m ⟨α⟩ = ⟨f ◦ d + A ◦ γ⟩ .
Now, also
2m ⟨a⟩ = ⟨f ◦ c + γ⟩ + ⟨f ◦ d + A ◦ γ⟩
= ⟨f ◦ c + f ◦ d + γ + A ◦ γ⟩
γ + A ◦ γ is a one cycle, and f ◦ c + f ◦ d is also a one cycle
= ⟨f ◦ c + f ◦ d⟩ + ⟨γ + A ◦ γ⟩
= f∗ ⟨c + d⟩ + ⟨α⟩
= f∗ ⟨α⟩ + ⟨α⟩ .
So
f∗ ⟨α⟩ = (2m − 1) ⟨α⟩ ,
for some integer m.
TODO: In
problem:
Remark. Exercise: even maps have even degrees: f (Ax) = f (x). say that one
cycles

Remark. We don’t use −x for Ax because it could be misleading: this −

has nothing to do with the − from the abelian homology group.
Theorem 25. Let f : S n → S n such that f (Ax) = Af (x). Then d(f ) is

odd.
Proof. We don’t prove this.
Theorem 26. There is no continuous antipodal preserving map from S 2 →

S1.
Proof. Suppose there is such a map.
S2 S1
f
S1
f
S2 S1
i
f |S 1
S1
Then we have
f∗
H1 (S 2 ) = 0 H1 (S 1 ) = Z
i∗
·(2k+1)
1
H1 (S ) = Z
But (f∗ ◦ i∗ )(1) = 0, but (f |S 1 )∗ (1) = 2k + 1, which is odd, and certainly

not 0. This works for all higher dimensional spheres to S 1 .
Theorem 27 (Borsuk-Ulam theorem). There are two antipodal points on the

earth such that both the temperature and air pressure are the same. If f :
S 2 → R2 is continuous. Then there exists x ∈ S 2 such that f (x) = f (−x).
Proof. Suppose f (x) ̸= f (−x) for all x ∈ S 2 . Then we define
f (x) − f (−x)
g : S 2 → R2 : x 7→ ∈ S1.
∥f (x) − f (−x)∥
Then g(−x) = −g(x), which is a contradiction with the previous theorem.

Remark. Also holds for S n → Rn .
Theorem 28 (Ham Sandwich theorem). Suppose you give me two pieces of

break and 1 slice of ham. Then it is possible to divide both the pieces of
bread and the slice of ham in equal pieces by 1 straight cut of knife.
Proof. Consider for each v ∈ S 2 a plane Pv ⊂ R3 . Pv ⊥ v and Pv cuts the

slice of ham exactly in two. We define the “upper side” of the plane to be
the half to which v is pointing to.
If you have some weird ham which which you can cut in multiple places
in half, then you take the middel of the line segment. This makes it unique.
Note that Pv = P−v .
b1 S2
v
h
b2
Now, consider
f : S 2 → R2 : v 7→ (f1 (v), f2 (v)).
Then f1 (v) is the volume of bread b1 above Pv . Then f2 (v) is the volume
of bread b2 above Pv .
Now, you should believe that f1 and f2 are continuous. (Proving this
precisely needs measure theory etc.) So, now, we can use the Borsak
Ulam theorem. So there exists a v ∈ S 2 such that f (v) = f (−v). So
f1 (v) = f1 (−v), so volume of bread b1 above Pv is the volume of bread b1
below Pv , and similar for f2 . This proves the Ham sandwich theorem.
Last result.
Theorem 29. Let A1 , A2 , A3 be three closed subsets of S 2 such that S 2 =

A1 ∪ A2 ∪ A3 . Then at least one of the Ai contains a pair of antipodal
points.
Proof. Let f : S 2 → R2 mapping x 7→ (d(x, A1 ), d(x, A2 )).
d(x, Ai ) = min d(x, y).

y∈Ai
Then, there exists a point x ∈ S 2 such that f (x) = f (−x). So
d(x, A1 ) = d(−x, A1 ) d(x, A2 ) = d(−x, A2 ).

Possibility 1 d(x, A1 ) = 0. Then x ∈ A1 and −x ∈ A1 .

Possibility 2 d(x, A2 ) = 0. Then x ∈ A2 and −x ∈ A2 .
Possibility 3 d(x, A2 ) ̸= 0 ̸= d(x, A1 ): x, −x ̸∈ A1 and x, −x ̸∈ A2 . As

x ∈ S 2 , x, −x ∈ A3 .
Remark. This also works for 2 (Just take A3 = ∅), but this doesn’t work
for 4.
A4
A2
A3
A1
Figure 13.11: Example with four elements
Remark. Time on schedule is when you have to enter the room. Hand over
problems after exam.

Master

Uploaded by

Copyright:

Available Formats

Master

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Master

Uploaded by

Copyright:

Available Formats

Algebraic Topology

March 25, 2023

– Chap 9 ‘Fundamental group’

• 10 problems (solve them during the semester) No feedback during the

0.1 What is algebraic topology?

• Functor: X ⇝ GX and f : X → Y ⇝ f∗ : GX → GY such that

Example. Suppose we have a functor. If GX ∼ ̸ GY , then X and Y are

which means that f∗ : GX → GY is an isomorphism.

0.2 Fundamental group

• A loop based at x0 ∈ X is a map f : I = [0, 1] → X, f (0) = f (1) = x0 .

Example. Let X = B 2 (2 dimensional disk). Then π(B 2 ) = 1, because

Example. Let X = S 1 and pick x0 on the circle. Two options:

• The loop goes around the circle, clockwise, once

Theorem 1 (Fixed point theorem of Brouwer). Any continuous map from a

Proof. Suppose there is no fixed point, so f (x) ̸= x for all x ∈ B 2 . Then

Figure 1: Proof of the Brouwer fixed point theorem

If x lies on the boundary, we have the identity map. This map is

continuous. Then we have S 1 → B 2 → S 1 , via the inclusion and r. Looking

π(S 1 ) = Z → π(B 2 ) = 1 → π(S 1 ) = Z.

9.51 Homotopy of paths

• For all t ∈ I, define ft : X → Y : x 7→ H(x, t)

Definition 2 (Path homotopy). Let f, g : I → X be two paths such that

• H(0, t) = x0 and H(1, t) = x1 (start and end points fixed)

Lemma 1. ≃ and ≃p are equivalence relations.

• Reflective: F (x, t) = f (x)

Figure 9.1: General example of a homotopy.

Example (Trivial, but important). Let C ⊂ Rn be a convex subset.

Remark. If f is path homotopic to f ′ and g path homotopic to g ′ (which

2. Let ex denote the constant path ex : I → X : s 7→ x, x ∈ X. If

Proof. First, two observations

CHAPTER 9. FUNDAMENTAL GROUP 6

3. Note that i ∗ i ≃p e0 . Now, applying the same rules, we get

1. Remark: Only defined if f (1) = g(0), g(1) = h(0). Note that f ∗ (g ∗

CHAPTER 9. FUNDAMENTAL GROUP 7

Figure 9.2: proof path

Note that γ ≃p i. Now, using the fact that composition of positive

which is what we wanted to show.

9.52 Fundamental group

π(X, x0 ) = {[f ] | f : I → X, f (0) = f (1) = x0 }.

(Also π1 (X, x0 ) is used, first homotopy group of X based at x0 )

Example. If C ⊂ Rn , convex then π(X, x0 ) = 1. E.g. π(B 2 , x0 ) = 1.

Remark. All groups are a fundamental group of some space.

Theorem 3 (52.1). Let X be a space, x0 , x1 ∈ X and α : I → X a path

CHAPTER 9. FUNDAMENTAL GROUP 8

is an isomorphisms of groups. Note however that this isomorphism depends

Proof. Let [f ], [g] ∈ π1 (X, x0 ). Then

α̂([f ] ∗ [g]) = [α] ∗ [f ] ∗ [g] ∗ [α]

Figure 9.3: Construction of the group homomorphism

Remark. If f : (x, x0 ) → (Y, y0 ) is a map of pointed topology spaces (f :

f∗ : π(X, x0 ) → π(Y, y0 ) : [γ] 7→ [f ◦ γ]

is a morphism of groups, because of the two ‘rules’ discussed previously,

Lecture 2: Covering spaces

CHAPTER 9. FUNDAMENTAL GROUP 9

Example. Any convex subset C ⊂ Rn is simply connected.

Lemma 2. Suppose X is simply connected and α, β : I → X two paths

Proof. Simply connected implies loops are homotopic? Consider α ∗ β ≃p

([α] ∗ [β]) ∗ [β] = [ex0 ] ∗ [β] = [β]

9.53 Covering Spaces

Remark. If U ′ ⊂ U , also open and U is evenly covered, then also U ′ .

Definition 6. Let p : E → B be a surjective map. Then p is a covering