Algebraic Topology I (MAT431)

Lecture Notes
Preface
This series of lecture notes has been prepared for aiding students who took the BRAC University
course Algebraic Topology I (MAT431) in Summer 2021 semester. These notes were typeset under
the supervision of mathematician Dr. Syed Hasibul Hassan Chowdhury. The video lectures can
be found here. The main goal of this typeset is to have an organized digital version of the notes,
which is easier to share and handle. If you see any mistakes or typos, please send me an email at
atonuroychowdhury@gmail.com

Atonu Roy Chowdhury

References:
• Topology, by James R. Munkres

• Elements of Algebraic Topology, by James R. Munkres

• A Basic Course in Algebraic Topology, by William S. Massey

ii
Contents

Preface ii

0 Topology Review 4
0.1 Euclidean Space Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.2 Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
0.3 Bases and Countability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
0.4 Hausdorff Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
0.5 Continuity and Homeomorphism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
0.6 Quotient Topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
0.7 Compactness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
0.8 Quotient Topology Continued . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
0.9 Open Equivalence Relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
0.10 An Alternate Approach to Quotient Topology . . . . . . . . . . . . . . . . . . . . . . . 18

1 Lecture 1 22
1.1 Homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2 Lecture 2 33
2.1 The Fundamental Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2 Path Connectedness and Simply Connectedness . . . . . . . . . . . . . . . . . . . . . . 33
2.3 Induced Homomorphism and Its Properties . . . . . . . . . . . . . . . . . . . . . . . . . 36

3 Lecture 3 39
3.1 Covering Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

4 Lecture 4 44
4.1 Lifting Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.2 Lifting Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

5 Lecture 5 52
5.1 Retraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2 Fixed Point Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3 Fundamental Theorem of Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

6 Lecture 6 60
6.1 Simplices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
6.2 Simplicial Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
6.3 Abstract Simplicial Complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.4 Abelian Group Essentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

7 Lecture 7 69
7.1 Notion of Orientation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.2 Homology Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
7.3 Homology Group Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

A Category Theory Basics 79

A.1 What is a Category? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
A.2 Functor and Functorial Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

3
0 Topology Review
§0.1 Euclidean Space Rn
Before embarking on the concept of general topological space, let us look at the Euclidean space Rn .
Rn is equipped with the notion of distance between 2 points p and q.

Definition 0.1.1 (Distance). Let the coordinates of p and q be (p1 , p2 , .....pn ) and (q 1 , q 2 , ......q n ),
respectively. The distance between p and q is given by
" #1
X
n 2

d(p, q) = (pi − q i )2
i=1

Definition 0.1.2 (Open ball). An open ball B(p, r) in Rn with center p ∈ Rn and radius r > 0 is
defined as the set
B(p, r) = {x ∈ Rn : d(x, p) < r}

A set equipped with the notion of distance between its elements is called a metric space1 . Thus the
Euclidean space Rn is a metric space. And we can talk about open balls in Rn using this metric. We
can define open sets in Rn using open balls B(p, r) defined above.

Definition 0.1.3 (Open Set in Rn ). A set U in Rn is said to be open if for every p in U , there is
an open ball B(p, r) such that B(p, r) ⊆ U .

Proposition 0.1.1
The union of an arbitrary collection of {Uα } of open sets is open. The intersection of finite
collection of open sets is open.

Proof. Trivial. ■

Example 0.1.1


The intervals − n1 , n1 , n = 1, 2, 3, .... are all open in R but their intersection
\  1 1
− , = {0}
n n
n∈N

is not open.

The metric d in Rn allows us to define open sets in Rn . In other words, given a subset of Rn , we can
tell if it is open or not. This situation is a special case called metric topology in Rn .

§0.2 Topology

1
There are some properties that a metric (distance) function should have. We won’t go into much details

4
0 Topology Review 5

Definition 0.2.1 (Topology). A topology on a set S is a collection T of subsets of S containing

both the empty set ∅ and the S such that T is closed under arbitrary union and finite intersection.
In other words,
S
• If Uα ∈ T for all α in an index set A, then Uα ∈ T
α∈A

T
n
• If Ui ∈ T for i ∈ {1, 2, ..., n}, then Ui ∈ T
i=1

The elements of T are called open sets.

Definition 0.2.2 (Topological Space). The pair (S, T ) consisting of a set S together with a topology
T on S is called a topological space.

Abuse of Notation. We shall often say “S is a topological space” in short. But there is always a
topology T on S, which we recall when necessary.

Definition 0.2.3 (Neighborhood). A neighbourhood of a point p ∈ S is called an open set U

containing p.

Definition 0.2.4 (Closed Set). The complement of an open set is called a closed set.

Proposition 0.2.1
The union of a finite collection of closed sets is closed. The intersection of an arbitrary collection
of closed sets is closed.

Proof. Let {Fi }ni=1 be a finite collection of closed sets. Then, {S \ Fi }ni=1 is a finite collection of open
Tn
sets. The intersection of a finite collection of open sets is open, therefore (S \ Fi ) is open. By De
i=1
Morgan’s law, !
\
n [
n [
n
(S \ Fi ) = S \ Fi is open =⇒ Fi is closed
i=1 i=1 i=1

Therefore, the union of a finite collection of closed sets is closed.

Now, let {Fα }α∈A be an arbitrary collection of closed sets with A being an index set. Then {S \Fα }α∈A
is an arbitrary collection
S of open sets. We know that the union of an arbitrary collection of open sets
is open, therefore α∈A (S \ Fα ) is open. By De Morgan’s law,
!
[ \ \
(S \ Fα ) = S \ Fα is open =⇒ Fα is closed
α∈A α∈A α∈A

Therefore, the intersection of an arbitrary collection of closed sets is closed. ■

Definition 0.2.5 (Subspace Topology). Let (S, T ) be a topological space and A a subset of S.
Define TA to be the collection of subsets

TA = {U ∩ A | U ∈ T }

TA is called the subspace topology of A in S.

5
0 Topology Review 6

It is not hard to see that TA satisfies the conditions of a Topology. Firstly, TA contains both ∅ and A.
For these, taking U = ∅ and U = S, respecitvely, suffices. By the distributive property of union and
intersection ! !
[ [ \
n \n
(Uα ∩ A) = Uα ∩ A and (Ui ∩ A) = Ui ∩ A
α α i=1 i=1
which shows that TA is closed under arbitrary union and finite intersection. So T is a Topology
indeed.

Example 0.2.1
 1

Consider the subset A = [0,
 1] of

R. In the
subspace topology, the half-open interval 0, 2 is an
open subset of A, because 0, 12 = − 12 , 21 ∩ [0, 1]

Lemma 0.2.2
Let Y be a subspace of X (that is Y has the subspace topology inherited from X). If U is open
in Y and Y is open in X, then U is open in X.

Proof. Since U is open in Y , U = Y ∩ V for some V open in X. Both Y and V are open in X, hence
Y ∩ V = U is also open in X. ■

§0.3 Bases and Countability

Definition 0.3.1 (Basis and Basic Open Sets). A subcollection B of a topology T is a basis for T
if given an open set U and a point p in U , there is an open set B ∈ B such that p ∈ B ⊆ u. An
element of B is called a basic open set.

Example 0.3.1
The collection of all open balls B(p, r) in Rn with p ∈ Rn and r > 0 is a basis for the standard
topology (metric topology) on Rn .

Proposition 0.3.1
A collection B of open sets of S is a basis if and only if every open set in S is a union of sets in B.

Proof. (⇒) We are given a collection of B of open sets of S that is a basis. U is any open set in S.
Also, let p ∈ U . STherefore, there is a basic open set Bp ∈ B such that p ∈ Bp ⊆ U . Hence, one can
show that U = Bp .
p∈U

S set in S is a union of open sets in B. Now, given an open set U and a

(⇐) Suppose, every open
point p ∈ U , since U = Bα , there is a Bα ∈ B, such that p ∈ Bα ⊆ U . Hence B is a basis. ■
Bα ∈B

We say that a point in Rn is rational if all of its coordinates are rational numbers. Let Q be the set
of rational numbers and Q+ the set of positive rational numbers.

Lemma 0.3.2
Every open set in Rn contains a rational point.

Y
n
Proof. An open set U in Rn contains an open ball B(p, r) which, in turn, contains an open cube Ii
  i=1
where Ii is the open interval pi − √rn , pi + √rn . Here is a visual example for n = 2.

6
0 Topology Review 7

 
p1 , p 2 + √r
2
 
p1 − √r ,
2
p2

 
r p1 + √r , p2
2

( )
p p1 , p 2

 
p1 , p 2 − √r
2

   
Figure 1: B(p, r) contains p1 − √r , p1
n
+ √r
n
× p2 − √r , p2
n
+ √r
n

Now back to general n. For each i, let q i be a rational number in Ii . Then (q 1 , q 2 , ..., q n ) is a rational
Y
n
point in Ii ⊆ B(p, r). Therefore, every open set contains a rational point. ■
i=1

Proposition 0.3.3
The collection BQ of all open balls in Rn with rational centers and rational radii is a basis for Rn .

Proof. Given an open set U in Rn and p ∈ U , there is an open ball B(p, r′ ) with positive real radius
r′ such that p ∈ B(p, r′ ) ⊆ U . Take a rational number r ∈ (0, r′ ). Then we have

p ∈ B(p, r) ⊆ B(p, r′ ) ⊆ U


By Lemma 0.3.2 , there is a rational point in the smaller ball B p, 2r .


Claim — p ∈ B q, 2r ⊆ B(p, r)



Proof. Since d(p, q) < 2r , we have p ∈ B q, 2r . Next, if x ∈ B q, 2r , then by triangle inequality
r r
d(x, p) ≤ d(x, q) + d(q, p) < + =r
2 2
Therefore, x ∈ B(p, r).

7
0 Topology Review 8

B(q, r/2)

r p

B(p, r/2)
B(p, r)

So, p ∈ B q, 2r and B q, 2r ⊆ B(p, r). □


As a result, p ∈ B q, 2r ⊆ B(p, r) ⊆ B (p, r′ ) ⊆ U . Hence we proved,
 r
p ∈ B q, ⊆U
2
In other words, the collection BQ of open balls with rational centers and rational radii is a basis for
Rn . ■

Both the sets Q and Q+ are countable. Since the centers of the open balls in BQ are indexed by Qn ,
a countable set, and the radii are indexed by Q+ , also a countable set, the collection BQ is countable.

Definition 0.3.2 (Second Countable). A topological space is said to be second countable if it has
a countable basis.

Proposition 0.3.3 shows that Rn with its standard topology is second countable.

Definition 0.3.3 (Neighborhood Basis). Let S be a topological space and p be a point in S. A basis
of neighbourhoods or a neighbourhood basis at p is a collection B = {Bα } of neighbourhoods
of p such that for any neighbourhood U of p there is a Bα ∈ B such that p ∈ Bα ⊆ U .

Definition 0.3.4 (First Countable). A topological space S is first countable if it has a countable
basis of neighbourhoods at every point p ∈ S.

Example 0.3.2


∞
For p ∈ Rn , let B p, n1 be the open ball of center p and radius 1
n in Rn . Then B p, n1 n=1
is
a neighbourhood basis at p. Thus Rn is first countable.

An important note: An uncountable discrete topological space is first countable but not second
countable. A second countable topological space is always first countable.

§0.4 Hausdorff Space

8
0 Topology Review 9

Definition 0.4.1 (Hausdorff Space). A topological space S is Hausdorff if given any 2 distinct
points x, y in S there exist disjoint open sets U, V such that x ∈ U and y ∈ V .

x y
U

Figure 2: Here S is a Hausdorff space, U and V are disjoint open sets containing x and y respectively.

Proposition 0.4.1
Every singleton set (a one-point set) in a Hausdorff space S is closed.

Proof. Let x ∈ S. We want to prove that {x} is closed, i.e. S \ {x} is open.
Let y ∈ S \ {x}. Since S is Hausdorff, we can find disjoint open sets Uy and Vy such that x ∈ Uy
and y ∈ Vy . No such Vy contains x. Therefore
[
S \ {x} = Vy
y∈S\{x}

So S \ {x} is union of open sets, hence open. So {x} is closed. ■

Example 0.4.1
The Euclidean space Rn (equipped with standard/ metric topology) is Hausdorff, for given distinct
points x, y in Rn , if ϵ = 12 d(x, y), then the open balls B(x, ϵ) and B(y, ϵ) will be disjoint.

B(x, ϵ) B(y, ϵ)
x y

In a similar manner, one can show that every metric space is Hausdorff.

Lemma 0.4.2
Let A be a subspace of X. If X is a Hausdorff space, then so is A.

Proof. Take x, y ∈ A ⊆ X with x 6= y. As X is Hausdorff, we can find disjoint open sets U and V in
X, such that U 3 x and V 3 y. x ∈ A and x ∈ U , so x ∈ A ∩ U . Similarly, y ∈ A ∩ V .
Now, both A ∩ U and A ∩ V are open in A, with respect to the subspace topology. Furthermore,
(A ∩ U ) ∩ (A ∩ V ) = A ∩ (U ∩ V ) = ∅. Therefore, for x, y ∈ A we’ve found disjoint open sets A ∩ U
and A ∩ V , containing x and y respectively. So A is Hausdorff. ■

9
0 Topology Review 10

§0.5 Continuity and Homeomorphism

Definition 0.5.1 (Continuous Maps). Let f : X → Y be a map of topological spaces. f is said to

be continuous if for each open subset V of Y , the set f −1 (V ) is an open subset of X.

Proposition 0.5.1
f : X → Y is continuous if and only if for every closed subset B of Y , the set f −1 (B) will be
closed in X.

Proof. (⇒) Suppose f is continuous. B is closed, so Y \ B is open in Y . Therefore, by the continuity

of f , f −1 (Y \ B) = X \ f −1 (B) is open in X, so f −1 (B) is closed.
(⇐) Suppose f −1 (B) is closed in X for any closed B ⊆ Y . Take any open set U in Y . Choose
B = Y \ U . Then by the assumption f −1 (Y \ U ) = X \ f −1 (U ) is closed in X. This gives us f −1 (U )
is open. So f is continuous. ■

Definition 0.5.2 (Homeomorphism). Let X and Y be topological spaces; let f : X → Y be a

bijection. If both f and the inverse function f −1 : Y → X are continuous, then f is called a
homeomorphism.

Example 0.5.1
The function f : R → R given by f (x) = 3x + 1 is a homeomorphism. We define g : R → R by
g(y) = 13 (y − 1). Then we have

f (g(y)) = y and g (f (x)) = x ∀ x, y ∈ R

This proves g = f −1 . It is easy to see that both f and g are continuous functions. Therefore f is
a homeomorphism.

However, a bijective function can be continuous without being a homeomorphism.

Example 0.5.2
Let S 1 denote the unit circle in R2 ; that is S 1 = {(x, y) ∈ R2 | x2 + y 2 = 1}, considered as a
subspacea of the space R2 . Let f : [0, 1) → S 1 be the

f (t) = (cos 2πt, sin 2πt)

It is left as an exercise for the reader to show that f is a continuous bijective function. But the
function f −1 is not continuous.



f 0, 14

f
0 1 p V
1
4

S1


U = 0, 14 is an open set in [0, 1) according to the subspace topology. We want to show that f (U )
is not open in S 1 . That would prove the discontinuity of f −1 .

10
0 Topology Review 11

Let p be the point f (0). And p ∈ f (U ). We need to find an open set of S 1 in subspace topology
containing p = f (0) and contained in f (U ) to show that f (U ) is open in S 1 , i.e we have to find
an open set in V of R2 such that f (0) = p ∈ V ∩ S 1 ⊆ f (U ). But it is impossible as is evident
from the figure above. No matter what V we choose, some part of V ∩ S 1 would lie outside f (U ).
a
Subset of R2 equipped with subspace topology.

Lemma 0.5.2 (Pasting Lemma)

Let X = A ∪ B, where A and B are closed in X. Let f : A → Y and g : B → Y be continuous. If
f (x) = g(x) for every x ∈ A ∩ B, then f and g combine to give a continuous function h : X → Y
defined by (
f (x) if x ∈ A
h(x) =
g(x) if x ∈ B

Proof. Let C be a closed subset of Y . Now,

h−1 (C) = f −1 (C) ∪ g −1 (C)

Since f is continuous, f −1 (C) is closed in A, hence closed in X. Similarly, g −1 (C) is closed in X. So

h−1 (C) is the union of two closed sets in X, hence it is closed in X. Therefore, h is continuous. ■

Lemma 0.5.3
Let X, Y be topological spaces and f : X → Y . Then f is continuous if and only if for every
x ∈ X and each neighborhood V of f (x), there is a neighborhood U of x such that f (U ) ⊆ V .

Proof. (⇒) Suppose f is continuous. Let x ∈ X and V 3 f (x) is open in Y . We take U = f −1 (V ).

Since f is open and U is preimage of open set, so U is open. Also,


f (x) ∈ V =⇒ x ∈ f −1 (V ) = U and f (U ) = f f −1 (V ) ⊆ V

(⇐) Let V ⊆ Y be open. We need to show that f −1 (V ) is open. Take x ∈ f −1 (V ). Then f (x) ∈ V ,
so V is a neighborhood of f (x). By assumption, there exists open U 3 x such that

f (U ) ⊆ V =⇒ U ⊆ f −1 (V )

So for every x ∈ f −1 (V ), there exists a neighborhood of x that is contained in f −1 (V ). So f −1 (V ) is

open, and hence f is continuous. ■

§0.6 Quotient Topology

Quotient topology is defined using an equivalence relation. An equivalence relation is a binary relation
on a set that has some properties.

Definition 0.6.1 (Equivalence Relation and Equivalence Class). An equivalence relation ∼ on a set
S is a binary relation which is reflexive, symmetric and transitive. That is

(i) a ∼ a for every a ∈ S

(ii) a ∼ b =⇒ b ∼ a

(iii) a ∼ b , b ∼ c =⇒ a ∼ c

The equivalence class [x], if x ∈ S, is the set of all elements in S equivalent to x.

11
0 Topology Review 12

An equivalence relation on S partitions S into disjoint equivalence classes. We denote the set of all
equivalence classes with S/∼ and call this the quotient of S by the equivalence relation ∼. There is a
natural projection map π : S → S/∼ which projects x ∈ S to its own equivalence class [x] ∈ S/∼.
Abuse of Notation. Ideally [x] denotes a point in S/∼. But we will use the same notation [x] to
identify a set in S whose elements are all equivalent to each other under the given equivalence relation.

Definition 0.6.2 (Quotient Topology). Let S be a topological space. We define a topology called
quotient topology on S/ ∼ by declaring a set U in S/ ∼ to be open if and only if π −1 (U ) is
open in S.

It’s not hard to see that quotient topology is a well defined topology. Note that π −1 (∅) = ∅ and
π −1 (S/∼) = S and hence ∅ and S/∼ are both open  sets in quotient topology. Now let {Uα }α∈A be
an arbitrary collection of open sets in S/∼. Then π −1 (Uα ) α∈A is an an arbitrary collection of open
sets in S. So,
!
[ [ [
−1 −1
π (Uα ) = π Uα is open in S =⇒ Uα open in S/∼
α∈A α∈A α∈A

So arbitrary
 −1 union of open sets is open in S/ ∼. Now for a finite collection of open sets {Ui }ni=1 in
n
S/∼, π (Ui ) i=1 is a finite collection of open sets in S. So,
!
\n \n \
n
−1 −1
π (Ui ) = π Uα is open in S =⇒ Uα open in S/∼
i=1 i=1 i=1

So finite intersection of open sets is open in S/∼. Therefore, we’ve verified that the open sets defined
on S/∼ indeed form a topology.

Continuity on Quotient Topology

Let ∼ be a equivalence relation on the topological space S and give S/ ∼ the quotient topology.
Suppose that the function f : S → Y is continuous from S to another topological space Y . Further
assume that f is constant on each equivalence class. Then f induces a map

f¯ : S/∼ → Y ; f¯ ([p]) = f (p) ∀p ∈ S

Note that this latter function f¯ wouldn’t be well-defined had we not assumed f to be constant on
each equivalence class in S/∼.

f
S Y
f =f ◦π
π f (p) = f (π(p)) = f ([p])
f

S/∼

Proposition 0.6.1
The induced map f¯ : S/∼→ Y is continuous if and only if the map f : S → Y is continuous.


Proof. (⇒). Suppose f : S → Y is continuous. Let V be open in Y . Then f −1 (V ) = π −1 f¯−1 (V )
is open in S. Therefore, by the definition of quotient topology, then f¯−1 (V ) is open in S/∼. Hence,
we’ve shown that for a given open set V in Y , f¯−1 (V ) is open in S/∼. So, f¯ : S/∼→ Y is continuous.
(⇐). That the continuity of f¯ : S/∼→ Y implies the continuity of f : S → Y is easy to see using
the equality f = f¯ ◦ π and the fact that π : S → S/∼ is continuous. ■

12
0 Topology Review 13

Identification of a subset to a point

If A is a subspace of a topological space S, we can define a relation ∼ on S by declaring

x∼x, ∀x∈S and x ∼ y , ∀ x, y ∈ A

It is immediate that ∼ is an equivalence relation. We say that the quotient space S/ ∼ is obtained
from S by identifying A to a point.

§0.7 Compactness

Definition 0.7.1 (Open Cover). Let S be a topological space. A collection {Uα }α∈I of open subsets
of S is said to be an open cover of S if
[
S⊆ Uα
α∈I

Since the open sets are in the topology of S and consequently Uα ⊆ S for S
S every α ∈ I, one has
α∈I Uα ⊆ S. Therefore, the open cover condition in this case reduces to S = α∈I Uα .
With the subspace topology, a subset A of a topological space S is a topological space by its own
right. The subspace A can be covered by open sets in A or by open sets in S.

S cover of A in S is a collectionS{Uα }α of open sets in S that covers A. In other words,

• An open
A ⊆ α Uα (Note that in this case A = α Uα might not hold in general).

• An open cover of A in A is S a collection {Uα }α of openSsets in A in subsapce topology that

covers A. In other words, A ⊆ α Uα (Here, in fact, A = α Uα as each Uα ⊆ A).

Definition 0.7.2 (Compact Set). Let S be a topological space and A ⊆ S. A is compact if and
only if every open cover of A in A has finite subcover.

Proposition 0.7.1
A subspace A of a topological space S is compact if and only if every open cover of A in S
has a finite subcover.
S
Proof. (⇒) Assume A is compact and let {Uα } be an open cover of A in S. This means that A ⊆ Uα .
α
Hence, !
[ \ [ \ 
A⊆ Uα A= Uα A
α α

Now, {Uα ∩ A}α is an open cover of A in A. Since A is compact, every open cover of A in A has a
finite subcover. Let the finite sub-cover be {Uαi ∩ A}ni=1 . Thus,

[
n
A⊆ U αi
i=1

which means that {Uαi }ni=1 is a finite sub-cover of the open cover {Uα }α of A in S.
(⇐) Suppose every open cover of A in S has a finite subcover, and let {Vα }α be an open cover of A
in A. Then each Vα is an open set of A in subspace topology. According to the definition of subspace
topology, there is an open set Uα in S such that Vα = Uα ∩ A. Now,
!
[ [ [ [
A⊆ Vα = (Uα ∩ A) = Uα ∩ A ⊆ Uα
α α α α

13
0 Topology Review 14

Therefore, {Uα }α is an open cover of A in S. By hypothesis, there are finitely many sets {Uαi }ni=1
Sn
such that A ⊆ Uαi . Hence,
i=1
!
[
n [
n [
n
A⊆ U αi ∩A= ( Uαi ∩ A) = Vαi
i=1 i=1 i=1

So {Vαi }ni=1 is a finite subcover of {Vα } that covers A in A. Therefore, A is compact. ■

Proposition 0.7.2
Every compact subset of K of a Hausdorff space S is closed.

Proof. We shall prove that S \ K is open. Let’s take x ∈ S \ K. We claim that there is a neighborhood
Ux of x that is disjoint from K.
Since S is hausdorff, for each y ∈ K, we can choose disjoint open sets Uy and Vy such that Uy 3 x
and Vy 3 y. The collection {Vy : y ∈ K}Sis an open cover of K in S. Since K is compact, there exists
a finite subcover {Vyi }ni=1 . That is K ⊆ ni=1 Vyi . Since Uyi ∩ Vyi for every i, we have
! !
\n [n \n
U yi ∩ Vyi = ∅ =⇒ Ux ∩ K = ∅ where Ux = U yi
i=1 i=1 i=1

Ux is the finite intersection of open sets, hence open. Also, every Uyi contains x, hence their intersection
Ux also contains x. So Ux is the desired open set that is disjoint from K, in other words x ∈ Ux ⊆ S \K.
As a result, [ [
S\K ⊆ Ux ⊆ S \ K =⇒ S \ K =
x∈S\K x∈S\K

S \ K is the union of open sets, hence open. Therefore K is closed. ■

Proposition 0.7.3
The image of a compact set under a continuous map is compact.

Proof. Let f : X → Y be a continuous and K a compact subset of X. Suppose {Uα } is an open cover
of f (K) by open subsets of Y . Since, f is continuous, the inverse images of f −1 (Uα ) are all open in
X. Moreover, !
[ [
K ⊆ f −1 (f (K)) ⊆ f −1 Uα = f −1 (Uα )
α α
 −1
So
 −1f (U )nα is an open cover of K in X. By Proposition 0.7.1, there is a finite sub-collection
f (Uαi ) i=1 such that
! !!
[
n [
n [
n [
n
−1 −1 −1
K⊆ f (Uαi ) = f U αi =⇒ f (K) ⊆ f f U αi ⊆ U αi
i=1 i=1 i=1 i=1

Thus f (K) is compact. ■

Lemma 0.7.4
A closed subset F of a compact topological space S is compact.

14
0 Topology Review 15

Proof. Let {Uα }α be an open cover of F in S. The collection {Uα , S \ F } is an open cover of S itself.
By compactness of S, there is a finite sub-cover {Uαi , S \ F }ni=1 of S, that is,
!
[n [n
F ⊆S⊆ Uαi ∪ (S \ F ) =⇒ F ⊆ U αi
i=1 i=1

Therefore, {Uαi }ni=1 is a finite subcover of the open cover {Uα } of F in S. Hence, F is also compact. ■

Proposition 0.7.5
A continuous map f : X → Y form a compact space X to a Hausdorff space Y is a closed map
(a map that takes closed sets to closed sets).

Proof. Let F ⊆ X be closed. Then F is compact by Lemma 0.7.4. Since f : X → Y is a continuous

map, by Proposition 0.7.3, f (F ) is compact in Y . Since Y is Hausdorff, by Proposition 0.7.2, f (F ) is
closed in Y . Hence, f is a closed map. ■

Corollary 0.7.6
A continuous bijection f : X → Y from a compact space X to a Hausdorff space is a homeomor-
phism.

Proof. We want to show thaat f −1 : Y
→ X is continuous. And in order to that it suffices to show
−1
that for every closed set F in X, f −1 (F ) = f (F ) is closed in Y . In other words, it suffices to
show that f is a closed map. The corollary then follows from Proposition 0.7.5. ■

Definition 0.7.3 (Bounded Set). A subset A of Rn is said to be bounded if it is contained in some

open ball B(p, r). otherwise, it is unbounded.

Theorem 0.7.7 (Heine-Borel Theorem)

A subset of Rn is compact if and only if it is closed and bounded.

Definition 0.7.4 (Diameter of Set). Let A ⊆ X be a bounded subset of a metric space (X, d). The
diameter of A is defined by

diam(A) := sup {d (a1 , a2 ) : a1 , a2 ∈ A}

Lemma 0.7.8 (Lebesgue Number Lemma)

Let (X, d) be a compact metric space. Given an open cover U = {Uα }α∈J of X, there exists a
number δ > 0 — called the Lebesgue number associated with the cover — such that for a given
A ⊆ X with diam(A) < δ, one must have A ⊆ Uα for some α ∈ J.

Proof. Take x ∈ X. As U covers X, we can find Uα ∈ U such that x ∈ Uα . Since Uα is open and
x ∈ Uα , there exists rx > 0 such that
B (x, rx ) ⊆ Uα
We do this for every x ∈ X. So we get an open cover of X
[  rx 
X= B x,
2
x∈X

15
0 Topology Review 16

Since X is compact, there exists a finite subcover of this open cover. So

[
n  r 
x
X= B xi , i
2
i=1

We define δ > 0 in the following way:

nr o
xi
δ = min : i = 1, 2, . . . , n
2
We claim that this δ is our desired Lebesgue number of the open cover U . Let A ⊆ X with diam(A) < δ.
Fix a ∈ A. Then there exists j ∈ {1, 2, . . . , n} such that
 rx  rx
a ∈ B xj , j =⇒ d (xj , a) < j
2 2


By the construction of rxj , there exists Uβ ∈ U such that B xj , rxj ⊆ Uβ . We claim that A ⊆ Uβ .
Take any b ∈ A.
rx rx
d (a, b) ≤ diam(A) < δ ≤ j =⇒ d (a, b) < j
2 2
rx rx

d (xj , b) ≤ d (xj , a) + d (a, b) < j + j = rxj =⇒ b ∈ B xj , rxj
2 2



For every b ∈ A, we have b ∈ B xj , rxj . Therefore, A ⊆ B xj , rxj ⊆ Uβ . ■

§0.8 Quotient Topology Continued

Let I be the closed interval [0, 1] in the standard topology of Rn and I/ ∼ be the quotient space
obtained from I by identifying the 2 points {0, 1} to a point. Denote by S 1 the unit circle in the
complex plane. Define f by f (x) = e2πix .
Now the function f : I → S 1 defined above assumes the same value at 0 and 1 and based on the
discussion prior to Proposition 0.6.1, f induces the map f¯ : I/∼→ S 1 .

Proposition 0.8.1
The function f¯ : I/∼ → S 1 is a homeomorphism.

Proof. The function f : I → S 1 defined by f (x) = e2πix is continuous (check!). Therefore, by

Proposition 0.6.1, f¯ : I → S 1 is also continuous.
Note that I = [0, 1] in R is closed and bounded and hence by Heine-Borel Theorem, I is compact.
Since the projection π : I → I/ ∼ is continuous, by Proposition 0.7.3, the image of I under π, i.e.,
I/∼ is compact.
It should also be obvious that f¯ : I/ ∼ → S 1 is a bijection. Since S 1 is a of the Hausdorff space
R2 , by Lemma 0.4.2, S 1 is also Hausdorff. Hence, f¯ is a continuous bijection from the compact
space I/ ∼ to the Hausdorff topological space S 1 . Therefore, by Corollary 0.7.6, f¯ : I/ ∼→ S 1 is a
homeomorphism. ■

Necessary Condition for a Hausdorff quotient

Even if S is a Hausdorff space, the quotient space S/∼ may fail to be Hausdorff.

Proposition 0.8.2
If the quotient space S/∼ is Hausdorff, then the equivalence class [p] of any point p in S is closed
in S.

16
0 Topology Review 17

Proof. By Proposition 0.4.1, every singleton set is closed in a Hausdorff topological space. Now,
consider the canonical projection map π : S → S/∼. For a point p ∈ S, {π(p)} is a singleton set in
S/∼.
Since, by hypothesis S/ ∼ is Hausdorff, {π(p)} must be closed in S/ ∼ with respect to quotient
topology. By continuity of π, π −1 ({π(p)}) is closed in S. But π −1 ({π(p)}) = [p]. Hence, [p] is a
closed set in S. ■

Remark. In order to prove that a quotient space S/∼ is not Hausdorff it is sufficient to prove that
the equivalence class [p] of some point p ∈ S is not closed in S. We have the following example to
elucidate this remark.

Example 0.8.1
Define an equivalence relation ∼ on R by identifying the open interval (0, ∞) to a point. The
resulting quotient space R/∼ is not Hausdorff since the equivalence class (0, ∞) is not a closed
subset of R.

§0.9 Open Equivalence Relations

Definition 0.9.1. An equivalence relation ∼ on a topological space S is said to be open if the

underlying projection map π : S → S/∼ is open (maps open sets to open sets).

S
π −1 (π(U )) = [b] ∪ [c] ∪ [d] = [x]
x∈U
e

a e a
π
b U π(U )
d
c c
d b f
f
S/∼
S
S
Figure 3: Indeed π −1 (π(U )) = [x]
x∈U

In other words, the equivalence relation ∼ on S is open if and only if for every open set U ∈ S, the
set π(U ) ∈ S/∼ is open. Or equivalently, by definition of quotient topology,
[
π −1 (π(U )) = [x] is open in S
x∈U
S
[x] denotes all points equivalent to some point of U (shaded region in Figure 3).
x∈U

Example 0.9.1
The projection map onto a quotient space is, in general, not open. For example, let ∼ be the
equivalence relation on the real line R that identifies the two points 1 and −1, and π : R → R/∼
the projection map.
The map π is open if and only if for every open set V in R, its image π(V ) is open in R/∼, or

17
0 Topology Review 18

equivalently π −1 (π(V )) is open in R. Let V be the open interval (−2, 0) in R. Then,

π −1 (π(V )) = (−2, 0) ∪ {1} , [Since π(1) ∈ π(V )]

which is not open in R and hence π is not an open map. In other words, the equivalence relation
∼ is not open.

Definition 0.9.2 (Graph of Equivalence Relation). Given an equivalence relation∼ on S, let R be

the subset of S × S that defines the relation R = {(x, y) ∈ S × S | x ∼ y}. We call R the graph
of the equivalence relation ∼.

We have a necessary and sufficient condition for a quotient space to be Hausdorff if the underlying
equivalence relation is an open equivalence relation. We state this condition by means of a theorem
(the proof is omitted).

Theorem 0.9.1
Suppose ∼ is an open equivalence relation on a topological space S. Then the quotient space
S/∼ is Hausdorff if and only if the graph R of ∼ is closed in S × S.

If the equivalence relation ∼ is equality, i.e., x ∼ y iff x = y, then the quotient space S/∼ is S itself
and the graph R of ∼ is simply the diagonal ∆ = {(x, x) ∈ S × S}.

Corollary 0.9.2
A topological space is Hausdorff if and only if the diagonal ∆ is closed in S × S.

Theorem 0.9.3
Let ∼ be an open equivalence relation on a topological space S with projection π : S → S/∼. If
B = {Bα } is a basis for S, then its image {π(Bα )} under π is a basis for S/∼.

Proof. Since π is open, {π(Bα )} is a collection of open sets in S/∼. Let W be an open set in S/∼
and [x] ∈ W with x ∈ S. So π(x) ∈ W , i.e., x ∈ π −1 (W ). Since π −1 (W ) is open in S, there is a basic
open set B ∈ B such that, x ∈ B ⊆ π −1 (W ). Hence

[x] = π(x) ∈ π(B) ⊆ π(π −1 (W )) ⊆ W

Now, we have seen that given W open in S/ ∼ and [x] ∈ W , there exists an open set π(B) in the
collection {π(Bα )} such that [x] ∈ π(B) ⊆ W . This proves that {π(Bα )} is a basis for S/∼. ■

Corollary 0.9.4
If ∼ is an open equivalence relation on a second-countable topological space, then the quotient
space S/∼ is second countable.

§0.10 An Alternate Approach to Quotient Topology

Instead of considering the projection map π : X → X/∼ under some equivalence relation ∼ defined on
the given topological space X, we shall consider a surjective map p : X → Y between two topological
spaces X and Y .

18
0 Topology Review 19

Definition 0.10.1 (Quotient Map). Let X and Y be topological spaces, and p : X → Y be

surjective. Then p is said to be a quotient map provided U ⊆ Y is open if and only if p−1 (U ) ⊆
X is open.

Another equivalent way to define quotient map would be using closed sets instead of open sets. In
other words, F ⊆ Y is closed if and only if p−1 (F ) ⊆ X is closed. The equivalence follows directly
from the following equation:
p−1 (Y \ F ) = X \ p−1 (F )

Lemma 0.10.1
If p : X → Y is a surjective continuous map, and p is either open or closed, then p is a quotient
map.

Proof. Suppose p is an open map. Continuity of p gives us p−1 (U ) is open in X, if U is open
in Y .

Now for the other direction, suppose −1 −1

p (U ) is open in X. p is an open map, so p p (U ) is open
in Y . As p is surjective, p p−1 (U ) = U . So U is open in Y .
Now suppose p is a closed map. Similar as before, continuity of p gives one direction. For the other
direction, suppose p−1 (U ) is open in X. Then X \ p−1 (U ) is closed in X. As p is a closed map, the
image of closed sets are closed. Therfore,



p X \ p−1 (U ) = Y \ p p−1 (U ) = Y \ U

is closed in Y . So U is open in Y .
Henceforth, p is a quotient map. ■

However, the converse is not necessarily true. There are quotient maps that are neither open nor
closed.

Example 0.10.1
Let A be a subset of R2 defined by

A = (x, y) ∈ R2 : x ≥ 0 or y = 0 (or both)

Let π1 : R2 → R be the projection on the first coordinate. Let q : A → R be the restriction of π1 ,

i.e. q = π1 A . Then q is a quotient map. But it is neither open, nor closed.

Definition 0.10.2 (Quotient Topology). Let X be a topological space, and A be a set. If p : X → A

is a surjective map, then there exists exactly one topology τ on A, relative to which p is a quotient
map. It is called the quotient topology induced by p.

τ consists of the subsets U of A such that p−1 (U ) is open in X. It’s easy to check that τ is a topology.
τ contains ∅ and A, because p−1 (∅) = ∅ and p−1 (A) = X. The other two conditions are also
satisfied, because
! !
[ [ \n \n
−1 −1 −1
p Uα = p (Uα ) and p Ui = p−1 (Ui )
α∈J α∈J i=1 i=1

Definition 0.10.3 (Quotient Space). Let X be a topological space and X∗ be a partition of X

into disjoint subsets. Let p : X → X∗ be the surjective map, that carries each point of X to
the element of X∗ containing it. In the quotient topology induced by p, X∗ is called a quotient

19
0 Topology Review 20

space of X.

It is often said that there is an equivalence relation ∼ defined on X, by means of which elements of
X∗ are said to be the equivalence classes. That is the reason we denoted the quotient space by X/∼
previously.
This broader definition quotient map p is a surjective map betweem topologiacal spaces. This is not
necessairly the canonical projection map π : X → X/∼. But essentially it turns our that if p : X → Y
is a quotient map, then Y is homeomorphic to the quotient space X/∼, provided that p is constant
on each equivalence classes of ∼.

Lemma 0.10.2
Composition of quotient maps is also a quotient map.

Proof. Let p : X → Y and q : Y → Z be quotient maps. We need to show that q ◦ p : X → Z is also

a quotient map.
Let U ⊆ Z. As q is a quotient map, U is open if and only
if q −1 (U ) is open in Y . Since p is a
quotient map, q −1 (U ) is open in Y if and only if p−1 q −1 (U ) is open in X. Therefore, U is open in


Z if and only if p−1 q −1 (U ) = (q ◦ p)−1 (U ) is open in X. Hence q ◦ p is a quotient map. ■

Theorem 0.10.3
Let p : X → Y be a quotient map. Let Z be a topological space, and g : X → Z be a map that
is constant on each set p−1 ({y}) for every y ∈ Y . Then g induces a map f : Y → Z such that
f ◦ p = g. The induced map f is continuous if and only if g is continuous; f is a quotient map if
and only if g is a quotient map.


Proof. For each y ∈ Y , the set g p−1 ({y}) is a singleton. If we let f (y) denote this point, then we
have defined a map f : Y → Z; and for every x ∈ X, f (p (x)) = g(x), so f ◦ p = g.

X
g
p

Y f
Z

If f is continuous, then g is a composition of continuous map, hence continuous. Now suppose
g is

continuous. So for V open in Z, g −1 (V ) is open in X. As f ◦ p = g, g −1 (V ) = p−1 f −1 (V ) is open
in X. Since p is a quotient map, p−1 (U ) open in X implies U is open in Y . Therefore, f −1 (V ) is
open in Y . Hence f is continuous.
If f is a quotient map, then g is a composition of quotient map. So g is also a quotient map by
Lemma 0.10.2.
Now suppose g is a quotient map. Then g is surjective. g = f ◦ p gives us f is also surjective. Since
g is continuous (because it’s a quotient map), f is also continuous. So V is open in Z gives us f −1 (V )
is open in Y . For the other direction,

assume f −1 (V ) is open in Y . As p is a quotient map, f −1 (V )
is open in Y implies p −1 f (V ) is open in X. This is the same as g −1 (V ). g is a quotient map, so
−1

g −1 (V ) is open in X gives us V is open in Z.

Assuming f −1 (V ) is open in Y we proved that V is open in Z. So we’ve proven that V is open in
Z if and only if f −1 (V ) is open in Y . Henceforth, f is a quotient map. ■

Proposition 0.10.4
Let p : X → Y be a continuous map. If p has a continuous right inverse, i.e. there exists a
continuous map f : Y → X such that p ◦ f = idY , then p is a quotient map.

20
0 Topology Review 21

Proof. p (f (y)) = y for every y ∈ Y . So p is a surjective map from X to Y . By the continuity of p,

p−1 (U ) is open in X if U is open in Y .

For the converse, suppose p−1 (U ) is open in X. By the continuity of f , f −1 p−1 (U ) is open in Y .


As p ◦ f = idY , f −1 ◦ p−1 = (p ◦ f )−1 = id−1
Y = idY . So f
−1 p−1 (U ) = id (U ) = U is open in Y .
Y
So we’ve proven that U is open in Y if and only if p−1 (U ) is open in X. Henceforth, p is a quotient
map. ■

Lemma 0.10.5
The map q : A → R in Example 0.10.1 is a quotient map, which is neither open nor closed.

Proof. Consider the map f : R → R2 defined by f (x) = (x, 0). We claim that this is continuous.
Consider a basis element U × V of R2 , where both U and V are open in R. Any open set in R2 is
composed of such bases. So it’s enough for us to prove that f −1 (U × V ) is open in R for such U and
V.
If 0 6∈ V , then f −1 (U × V ) = ∅ is open in R. If 0 ∈ V , f −1 (U × V ) = U is open in R. So f is
continuous.
The range of f is a subset of A, so we can think of f as a map from R to A. The topology on A is
the subspace topology inherited from R2 , so f : R → A is continuous.
π1 is continuous. Because for open W ⊆ R, π1−1 (W ) = W × R, which is open in R2 . As a restriction
of a continuous map, q is also continuous.
Now, f is the right inverse of q. Because, for x ∈ R, q (f (x)) = q ((x, 0)) = x = idR (x). We’ve
shown that both q and f are continuous, so by Proposition 0.10.4, q is a quotient map. Now we need
to show that it’s neither open nor closed.
U = (−1, 1) × (0, 1) is open in R2 . Therefore, V = A ∩ U = [0, 1) × (0, 1) is open in A. But
q (V ) = [0, 1) is not open in R. Because, no matter how small a neighborhood of 0 we choose, it will
always contain some negative part. So the neighborhood will not be contained in [0, 1). Hence, q is
not an open map.  1

n , n : n ∈ N . X is closed in R , so
To show that q is not a closed map, take the set X = 2
1
X ∩ A = X is closed in A. Then q (X) = n : n ∈ N = Y is not closed. Because R \ Y contains 0,
but no neighborhood of 0 is contained in R \ Y . No matter how small a neighborhood we choose, it
will always contain some n1 . Therefore, R \ Y is not open, consequently Y is not closed, and thus q is
not a closed map. ■

Proposition 0.10.6
Let f : X → Y be continuous surjective, where X is compact and Y is hausdorff. Then f is a
quotient map.

Proof. f is a continuous map from a compact space to a Hausdorff space. Therefore, by Proposi-
tion 0.7.5, f is a closed map. f is a continuous surjective closed map. By Lemma 0.10.1, f is a
quotient map. ■

21
1 Lecture 1
§1.1 Homotopy

Definition 1.1.1 (Homotopy). If f and f ′ are continuous maps of the space X into the space Y ,
we say that f is homotopic to f ′ if there is a continuous map F : X × I → Y such that

F (x, 0) = f (x) and F (x, 1) = f ′ (x)

for each x ∈ X. (Here I = [0, 1]) The map F is called a homotopy between f and f ′ . If
f is homotopic to f ′ , we write f ' f ′ . If f ' f ′ and f ′ is a constant map, we say that f is
nulhomotopic.

We considered a continuous function f : X → Y . Now we consider a special case where f : [0, 1] → X

is a continuous map such that f (0) = x0 ∈ X and f (1) = x1 ∈ X. We say that f is a path from x0
(the initial point) to x1 (the final point) in X.
If f and f ′ are 2 paths in X, there is a stronger relation between them than hometopy It is defined as
follows:

Definition 1.1.2 (Path Homotopy). Two paths f, f ′ : I → X, are said to be path homotopic if
they have the same initial and final points, dented by x0 and x1 respectively, and if there is a
continuous map F : I × I → X such that

F (s, 0) = f (s) and F (s, 1) = f ′ (s),

F (0, t) = x0 and F (1, t) = x1 ,

for each s ∈ I and each t ∈ I. We call F a path homotopy between f and f ′ . If f is path
homotopic to f ′ , we write f 'p f ′ .

f x1
F
x0
f′

X
I × I

The first condition simply says that F is a homotopy between f and f ′ , and the second says that for
each t, the path ft defined by the equation ft (s) = F (s, t) is a path from x0 to x1 , such that f0 = f
and f1 = f ′ .

Lemma 1.1.1
The relations ' and 'p are equivalence relations.

22
1 Lecture 1 23

Proof. We need to verify the three properties of an equivalence relation. Firstly, we need to check
that f ' f for every f . That is, we need to find a homotopy F : X × I → Y such that

F (x, 0) = f (x) = F (x, 1) ∀x ∈ X

We choose F (x, t) = f (x) for every t ∈ I. Continuity of F follows directly from continuity of f .
For the symmetry part, let f ' f ′ . We need to show that f ′ ' f . Let F : X × I → Y be a homotopy
between f and f ′ . That is,

F (x, 0) = f (x) and F (x, 1) = f ′ (x) ∀x ∈ X

We need to find a homotopy G between f ′ and f . We choose G : X × I → Y such that G(x, t) =

F (x, 1 − t). Then it’s immediate that

G(x, 0) = F (x, 1) = f ′ (x) and G(x, 1) = F (x, 0) = f (x) ∀x ∈ X

Continuity of G follows directly from continuity of F . Therefore, G is a homotopy between f ′ and f .

In other words, f ′ ' f . The same construction of G works for path homotopy too.
Now for the transitivity part, let f ' f ′ and f ′ ' f ′′ . We need to show that f ' f ′′ . Let F be a
homotopy between f and f ′ ; and F ′ be a homotopy between f ′ and f ′′ . That is,

F (x, 0) = f (x) and F (x, 1) = f ′ (x) ∀x ∈ X

′ ′ ′ ′′
F (x, 0) = f (x) and F (x, 1) = f (x) ∀x ∈ X

Now we define G : X × I → Y as follows:

(  
F (x, 2t) for t ∈ 0, 12
G(x, t) =  
F ′ (x, 2t − 1) for t ∈ 12 , 1

Claim — G is a homotopy between f and f ′′ .

   
Proof. Let A = X × 0, 12 and B = X × 21 , 1 . So A ∪ B = X × I. X is both open and closed
in X. The closed intervals 0, 12 and 12 , 1 are also closed. Therefore, A and B are both closed
in X × I. Here the underlying topology is product topology. A ∩ B = X × 21 . The functions
indeed agree on A ∩ B, because
   
1 ′ ′ ′ 1
F x, 2 · = F (x, 1) = f (x) = F (x, 0) = F x, 2 · − 1
2 2

Therefore, by Pasting Lemma, G is continuous on X × I. Now,

G(x, 0) = F (x, 0) = f (x) and G(x, 1) = F ′ (x, 1) = f ′′ (x) ∀x ∈ X

Therefore, G is a homotopy between f and f ′′ . □

Therefore, ' is an equivalence relation. For path homotopy 'p , transitivity can be checked in a similar
way.
Let f 'p f ′ and F be a path homotopy between Let f ′ 'p f ′′ and F ′ be a path homotopy between
f and f ′ . That is, f ′ and f ′′ . That is,

F (s, 0) = f (s) and F (s, 1) = f ′ (s) , F ′ (s, 0) = f ′ (s) and F ′ (s, 1) = f ′′ (s) ,
F (0, t) = x0 and F (1, t) = x1 ∀ s, t ∈ I F ′ (0, t) = x0 and F ′ (1, t) = x1 ∀ s, t ∈ I
We construct G : I × I → X similar as before.
(  
F (s, 2t) for t ∈ 0, 21
G(s, t) =  
F ′ (s, 2t − 1) for t ∈ 12 , 1

23
1 Lecture 1 24

Now we want to show that G preserves the endpoints x0 and x1 .

(  
F (0, 2t) for t ∈ 0, 21
G(0, t) =   = x0 [F (0, t) = F ′ (0, t) = x0 for all t]
F ′ (0, 2t − 1) for t ∈ 12 , 1
(  
F (1, 2t) for t ∈ 0, 12
G(1, t) =   = x1 [F (1, t) = F ′ (1, t) = x1 for all t]
F ′ (1, 2t − 1) for t ∈ 12 , 1

Therefore, 'p is also an equivalence relation. ■

Example 1.1.1
Let f and g be any two continuous maps from a space X to R2 . It’s easy to see that f ' g. The
map
F (x, t) = (1 − t)f (x) + tg(x)
is a homotopy between them. It is called a straight-line homotopy because it moves the point
f (x) to the point g(x) along the straight-line segment joining them.

If f is a path, we shall denote its path-homotopy equivalence class by [f ].

Abuse of Notation. When we say f is a path in X from x0 to x1 , we generally mean that f : I → X

is the path and f (0) = x0 , f (1) = x1 .

Definition 1.1.3 (Product of Path). If f is a path in X from x0 to x1 , and if g is a path in X from

x1 to x2 , we define the product f ∗ g of f and g to be the path h given by the equations
(  
f (2s) for s ∈ 0, 21
h(s) =  
g(2s − 1) for s ∈ 12 , 1

The function h is well-defined and continuous, by Pasting Lemma. It is a path in X from x0 to x2 ,

since h(0) = f (0) = x0 and h(1) = g(1) = x2 . We think of h as the path whose first half is the path f
and whose second half is the path g. The product operation on paths induces a well-defined operation
on path-homotopy classes, defined by the equation [f ] ∗ [g] = [f ∗ g].

Definition 1.1.4. Let [f ] and [g] be path classes such that f (1) = g(0). Then we define the
operation ∗ between path classes as follows:

[f ] ∗ [g] = [f ∗ g]

Proposition 1.1.2
The operation [f ] ∗ [g] = [f ∗ g] is well defined.

Proof. For the operation to make sense, let f and f ′ be homotopic paths in X from x0 to x1 ; g and
g ′ be homotopic paths in X from x1 to x2 . Let F be a path homotopy between f and f ′ and let G be
a path homotopy between g and g ′ . We need to prove that f ∗ g 'p f ′ ∗ g ′ . Define H : I × I → X by
(  
F (2s, t) for s ∈ 0, 12
H(s, t) =  
G(2s − 1, t) for s ∈ 21 , 1

24
1 Lecture 1 25

Claim — H is a path homotopy between f ∗ g and f ′ ∗ g ′ .

Proof. F and G are path homotopies.

f 'p f ′ and F is a path homotopy between f g 'p g ′ and G be a path homotopy between g
and f ′ . That is, and g ′ . That is,

F (s, 0) = f (s) and F (s, 1) = f ′ (s) , G(s, 0) = g(s) and G(s, 1) = g ′ (s) ,
F (0, t) = x0 and F (1, t) = x1 ∀ s, t ∈ I G(0, t) = x1 and G(1, t) = x2 ∀ s, t ∈ I
Now, from the definition of product,
(  
f (2s) for s ∈ 0, 21
(f ∗ g) (s) =  
g(2s − 1) for s ∈ 12 , 1
(  
F (2s, 0) for s ∈ 0, 12
=   = H(s, 0)
G(2s − 1, 0) for s ∈ 12 , 1
(  

f ′ (2s) for s ∈ 0, 12
f ′ ∗ g ′ (s) =  
g ′ (2s − 1) for s ∈ 12 , 1
(  
F (2s, 1) for s ∈ 0, 12
=   = H(s, 1)
G(2s − 1, 1) for s ∈ 12 , 1

So the first condition of path homotopy is satisfied. The second condition is also satisfied, since

H(0, t) = F (0, t) = x0 and H(1, t) = G(1, t) = x2

Hence H is a path homotopy between f ∗ g and f ′ ∗ g ′ . □

Therefore, f ∗ g 'p f ′ ∗ g ′ and the product operation is well defined. ■

Theorem 1.1.3
The operation ∗ has the following properties:

(i) (Associativity) If [f ] ∗ ([g] ∗ [h]) is defined, so is ([f ] ∗ [g]) ∗ [h], and they are equal.

(ii) (Right and left identities) Given x ∈ X, let ex denote the constant path ex : I → X carrying
all of I to the point x. If f is a path in X from x0 to x1 , then

[f ] ∗ [ex1 ] = [f ] and [ex0 ] ∗ [f ] = [f ]

(iii) (Inverse) Given the path f in X from x0 to x1 , let f¯ be the path defined by f¯(s) = f (1 − s).
It is called the reverse of f . Then

[f ] ∗ [f¯] = [ex0 ] and [f¯] ∗ [f ] = [ex1 ]

Proof. The proof is based on two key facts.

1. If k : X → Y is a continuous map, and if F is a path homotopy in X between the paths

f, f ′ : I → X, then k ◦ F is a path homotopy in Y between the paths k ◦ f, k ◦ f ′ : I → Y . See
the figure below:

25
1 Lecture 1 26

k◦F

f k◦f
F k
f′ k ◦ f′

I ×I X Y
Given F is a path homotopy between f and f ′ , we have

F (s, 0) = f (s) and F (s, 1) = f ′ (s),

F (0, t) = x0 and F (1, t) = x1 ,

Indeed we have
(k ◦ F ) (s, 0) = (k ◦ f ) (s) and (k ◦ F ) (s, 1) = (k ◦ f ′ ) (s),
(k ◦ F ) (0, t) = k (x0 ) and (k ◦ F ) (1, t) = k (x1 )

Both k : X → Y and F : I × I → X are continuous, so is k ◦ F : I × I → Y . Therefore, k ◦ F is

a path homotopy between k ◦ f and k ◦ f ′ .

2. The second is the fact that if k : X → Y is a continuous map, and if f and g are paths in X
with f (1) = g(0), then
k ◦ (f ∗ g) = (k ◦ f ) ∗ (k ◦ g)
Proving this fact is not hard. Indeed for continuous f, g : I → X and x0 , x1 , x2 ∈ X with
f (0) = x0 , f (1) = x1 = g(0), g(1) = x2 one finds f ∗ g making sense as a path in X with
(f ∗ g)(0) = x0 and (f ∗ g)(1) = x2 defined by
(  
f (2s) for s ∈ 0, 21
(f ∗ g)(s) =  
g(2s − 1) for s ∈ 12 , 1

Hence, k ◦ (f ∗ g) : I → Y is a path with (k ◦ (f ∗ g)) (0) = x0 and (k ◦ (f ∗ g)) (1) = x2 is given

by (  
(k ◦ f )(2s) for s ∈ 0, 21
(k ◦ (f ∗ g)) (s) =  
(k ◦ g)(2s − 1) for s ∈ 12 , 1
Since k is continuous, we have k ◦ f, k ◦ g : I → Y continuous. Also (k ◦ f )(1) = k(x1 ) = (k ◦ g)(0).
So it makes sense to talk about the product of k ◦ f and k ◦ g.
(  
(k ◦ f )(2s) for s ∈ 0, 21
((k ◦ f ) ∗ (k ◦ g)) (s) =  
(k ◦ g)(2s − 1) for s ∈ 12 , 1

So indeed we have k ◦ (f ∗ g) = (k ◦ f ) ∗ (k ◦ g).

Now we proceed to prove the theorem. We shall prove the existence of identities and inverses first,
then we will go on to prove associativity.
Let e0 : I → I be the constant path in I that maps all of I to 0; in other words, e(t) = 0 for every
t ∈ I. Also we denote by i : I → I, the identity map on I; in other words, i(s) = s for every s ∈ I.
Note that, i is also a path on I from 0 to 1. Now, using the definition of product of paths,
(   (  
e0 (2s) for s ∈ 0, 12 0 for s ∈ 0, 21
(e0 ∗ i) (s) =   =  
i(2s − 1) for s ∈ 12 , 1 2s − 1 for s ∈ 12 , 1

We draw the graphs of the maps i and e0 ∗ i below:

26
1 Lecture 1 27

i(s)

(e0 ∗ i)(s)

Claim — i 'p e0 ∗ i

Proof. We construct a path homotopy G : I × I → I between them as follows:

G(s, t) = (1 − t)i(s) + t (e0 ∗ i) (s)

Indeed G(s, 0) = i(s) and G(s, 1) = (e0 ∗ i) (s). Also,

G(0, t) = (1 − t)i(0) + t (e0 ∗ i) (0) = 0 = i(0) = (e0 ∗ i) (0) and

G(1, t) = (1 − t)i(1) + t (e0 ∗ i) (1) = 1 − t + t = 1 = i(1) = (e0 ∗ i) (1)

Both i and e0 ∗ i are continuous, therefore G is also continuous. Hence G is a path homotopy
between i and e0 ∗ i. □

Now, let f be a path in X from x0 to x1 . That is, f : I → X is a continuous map with f (0) = x0
and f (1) = x1 . Since i 'p e0 ∗ i with G being a path homotopy between them, by the key facts
stated above, f ◦ G is a path homotopy between f ◦ i and f ◦ (e0 ∗ i) = (f ◦ e0 ) ∗ (f ◦ i). Now,
(f ◦ i) (s) = f (i(s)) = f (s), so f ◦ i = f . Also,

(f ◦ e0 ) (s) = f (e0 (s)) = f (0) = x0 ∀s ∈ I =⇒ f ◦ e0 = ex0

Therefore, f 'p ex0 ∗ f . It makes sense to talk about ex0 ∗ f because ex0 (1) = x0 = f (0). So, by the
well definedness of ∗,
[ex0 ] ∗ [f ] = [ex0 ∗ f ] = [f ]
In a similar way, one can prove that [f ] ∗ [ex1 ] = [f ]. For that we need i 'p i ∗ e1 , where e1 is the
constant path that sends all of I to 1.
Now we shall prove the existence of inverses. Firstly, we define the reverse path of i, ī : I → I by
ī(s) = 1 − s for every t ∈ I. Since i is a path in I from 0 to 1, ī is a path in I from 1 to 0. So it makes
sense to talk about i ∗ ī. Using the definition of product of paths,
(   (  
i(2s) for s ∈ 0, 12 2s for s ∈ 0, 12
(i ∗ ī) (s) =   =  
ī(2s − 1) for s ∈ 12 , 1 2 − 2s for s ∈ 12 , 1

We draw the graphs of the maps e0 and i ∗ ī below:

27
1 Lecture 1 28

(i ∗ i)(s)

e0 (s)

Claim — i ∗ ī 'p e0

Proof. We construct a path homotopy H : I × I → I between them as follows:

H(s, t) = (1 − t) (i ∗ ī) (s) + t e0 (s)

Indeed H(s, 0) = (i ∗ ī) (s) and H(s, 1) = e0 (s). Also,

H(0, t) = (1 − t) (i ∗ ī) (0) + t e0 (0) = 0 = (i ∗ ī) (0) = e0 (0) and

H(1, t) = (1 − t) (i ∗ ī) (1) + t e0 (1) = 0 = (i ∗ ī) (1) = e0 (1)

Both i ∗ ī and e0 are continuous, therefore H is also continuous. Hence H is a path homotopy
between i ∗ ī and e0 . □

Now, let f be a path in X from x0 to x1 . That is, f : I → X is a continuous map with f (0) = x0
and f (1) = x1 . Since i ∗ ī 'p e0 with H being a path homotopy between them, by the key facts
stated above, f ◦ H is a path homotopy between f ◦ e0 = ex0 and f ◦ (i ∗ ī) = (f ◦ i) ∗ (f ◦ ī). Now,
(f ◦ i) (s) = f (i(s)) = f (s), so f ◦ i = f . Also,

(f ◦ ī) (s) = f (ī(s)) = f (1 − s) = f¯(s) =⇒ f ◦ ī = f¯

Therefore, ex0 'p f ∗ f¯. It makes sense to talk about f ∗ f¯ because f (1) = f¯(0). So, by the well
definedness of ∗,    
[f ] ∗ f¯ = f ∗ f¯ = [ex0 ]
In a similar way, one can prove that [f¯] ∗ [f ] = [ex1 ]. For that we need ī ∗ i 'p e1 , where e1 is the
constant path that sends all of I to 1.
Now we shall move on to proving associativity. Let f, g, h be three paths in X such that f (1) = g(0)
and g(1) = h(0), so that the products (f ∗ g) ∗ h and f ∗ (g ∗ h) are defined.
Now we define the positive linear mapping pem,n : [m, n] → [0, 1] such that pem,n (m) = 0 and pem,n (n) = 1
and the map is linear, where 0 ≤ m ≤ n ≤ 1. Using this map, we will define the “triple product” of
three paths f , g and h.

28
1 Lecture 1 29

pea,b (s) = s
b−a − a
b−a

pe0,a (s) = a1 s
peb,1 (s) = s
1−b − b
1−b

0 a b 1

We choose a and b such that 0 < a < b < 1. Then we define a path Ka,b : I → X (depending on the
choice of a and b) in the following way:


(f ◦ pe0,a ) (s) if s ∈ [0, a]
Ka,b (s) = (g ◦ pea,b ) (s) if s ∈ [a, b]


(h ◦ peb,1 ) (s) if s ∈ [b, 1]

We wish to prove that, no matter what a, b we choose, the paths will belong to the same path homotopy
class. That is Ka,b 'p Kc,d for all 0 < a < b < 1 and 0 < c < d < 1. For this, we need to define a new
map P : I → I in the following way:


p0,a (s) if s ∈ [0, a]
P (s) = pa,b (s) if s ∈ [a, b]


pb,1 (s) if s ∈ [b, 1]

where p0,a : [0, a] → [0, c] is a positive linear map with p0,a (0) = 0 and p0,a (a) = c; pa,b : [a, b] → [c, d]
is a positive linear map with pa,b (a) = c and pa,b (b) = d; pb,1 : [b, 1] → [d, 1] is a positive linear map
with pb,1 (b) = d and pb,1 (1) = 1.

1
pb,1 (s) = 1−d
1−b (s − b) + d
d

c pa,b (s) = d−c

b−a (s − a) + c

p0,a (s) = ac s

0 a b 1

Claim — pe0,c ◦ p0,a = pe0,a ; pec,d ◦ pa,b = pea,b ; ped,1 ◦ pb,1 = peb,1 .

29
1 Lecture 1 30

Proof. The proofs follow immediately from the equations shown in the figures.
1 1 c
p0,c ◦ p0,a ) (s) = pe0,c (p0,a (s)) =
(e p0,a (s) = s
c c a
1
= s = pe0,a (s)
a
1 c
pc,d ◦ pa,b ) (s) = pec,d (pa,b (s)) =
(e pa,b (s) −
 d−c  d−c
1 d−c c
= (s − a) + c −
d−c b−a d−c
s−a
= = pea,b (s)
b−a
1 d
pd,1 ◦ pb,1 ) (s) = ped,1 (pb,1 (s)) =
(e pb,1 (s) −
 1−d  1−d
1 1−d d
= (s − b) + d −
1−d 1−b 1−d
s−b
= = peb,1 (s)
1−b
So the claim is proved. □

Claim — P 'p i, where i : I → I is the identity map.

Proof. We construct a path homotopy F : I × I → I between them as follows:

F (s, t) = (1 − t)P (s) + t i(s)

Indeed F (s, 0) = P (s) and F (s, 1) = i(s). Also,

F (0, t) = (1 − t)P (0) + t i(0) = 0 = P (0) = i(0) and

F (1, t) = (1 − t)P (1) + t i(1) = 1 = P (1) = i(1)

Both P and i are continuous, therefore F is also continuous. Hence F is a path homotopy between
P and i. □

Now we wanna show that Ka,b = Kc,d ◦ P . For this, let’s recall the definition of Kc,d once again.


(f ◦ pe0,c ) (s) if s ∈ [0, c]
Kc,d (s) = (g ◦ pec,d ) (s) if s ∈ [c, d]


(h ◦ ped,1 ) (s) if s ∈ [d, 1]


Now we shall compose it with P . If s ∈ [0, a], P (s) = p0,a (s) ∈ [0, c]. That’s why P −1 [0, c] = [a, b].

30
1 Lecture 1 31



Similarly, P −1 [c, d] = [a, b] and P −1 [d, 1] = [b, 1]. Now,



(f ◦ pe0,c ◦ P ) (s) if s ∈ P −1 [0, c]


(Kc,d ◦ P ) (s) = (g ◦ pec,d ◦ P ) (s) if s ∈ P −1 [c, d]



(h ◦ ped,1 ◦ P ) (s) if s ∈ P −1 [d, 1]


(f ◦ pe0,c ◦ P ) (s) if s ∈ [0, a]
= (g ◦ pec,d ◦ P ) (s) if s ∈ [a, b]


(h ◦ ped,1 ◦ P ) (s) if s ∈ [b, 1]


(f ◦ pe0,c ◦ p0,a ) (s) if s ∈ [0, a]
= (g ◦ pec,d ◦ pa,b ) (s) if s ∈ [a, b]


(h ◦ ped,1 ◦ pb,1 ) (s) if s ∈ [b, 1]


(f ◦ pe0,a ) (s) if s ∈ [0, a]
= (g ◦ pea,b ) (s) if s ∈ [a, b]


(h ◦ peb,1 ) (s) if s ∈ [b, 1]
= Ka,b (s)

We have shown that F is a path homotopy between P and i. So, by the first key fact stated at the
beginning of the proof, Kc,d ◦ F is a path homotopy between Kc,d ◦ P and Kc,d ◦ i. As a result,

Kc,d ◦ P 'p Kc,d ◦ i =⇒ Ka,b 'p Kc,d

Now we can go back to our original problem. The main idea is to write f ∗ (g ∗ h) and (f ∗ g) ∗ h as
Ka,b and Kc,d for some 0 < a < b < 1 and 0 < c < d < 1.
(  
f (2s) for s ∈ 0, 12
(f ∗ (g ∗ h)) (s) =  
(g ∗ h) (2s − 1) for s ∈ 12 , 1
  

f (2s) for s ∈ 0, 21
 
= g (2(2s − 1)) for 2s − 1 ∈ 0, 12

  
h (2(2s − 1) − 1) for 2s − 1 ∈ 12 , 1
  

f (2s) for s ∈ 0, 12
 
= g (4s − 2) for s ∈ 12 , 34

  
h (4s − 3) for s ∈ 34 , 1

From the equations of pea,b maps,

1
pe0, 1 (s) = 1 (s) = 2s
2
2
1 1
s s
pe1 , 3 (s) = − 2
= − 2
= 4s − 2
2 4 3
4 − 1
2
3
4 − 1
2
1
4
1
4
3 3
s s
pe3 ,1 (s) = − 4
= − 4
= 4s − 3
4 1− 3
4 1− 3
4
1
4
1
4

31
1 Lecture 1 32

As a result,
    

 f e
p 1 (s) for s ∈ 0, 12

  0, 2   
(f ∗ (g ∗ h)) (s) = g pe1 , 3 (s) for s ∈ 12 , 34

  2 4   

h pe3 (s)
,1 for s ∈ 34 , 1
 4   

 f ◦ e
p for s ∈ 0, 21

 0, 2 (s)
1
  
= g ◦ pe1 , 3 (s) for s ∈ 12 , 34

  2 4
 

 h ◦ pe3
,1 (s) for s ∈ 34 , 1
4

= K 1 , 3 (s)
2 4

So f ∗ (g ∗ h) = K 1 , 3 . In a similar manner, one can show that (f ∗ g) ∗ h = K 1 , 1 . Since Ka,b 'p Kc,d ,
2 4 4 2
we get

K 1 , 3 'p K 1 , 1 =⇒ f ∗ (g ∗ h) 'p (f ∗ g) ∗ h
2 4 4 2

=⇒ [f ] ∗ [g ∗ h] = [f ∗ g] ∗ [h]
=⇒ [f ] ∗ ([g] ∗ [h]) = ([f ] ∗ [g]) ∗ [h]

Therefore, the operation ∗ on path homotopy classes is associative. ■

32
2 Lecture 2
§2.1 The Fundamental Group
The set of path-homotopy classes of paths in a space X does not form a group under the operation
∗ because the product of two path-homotopy classes is not always defined. But suppose we pick out
a point (to be called “base point”) x0 ∈ X and restrict ourselves to all the paths that begin and
end at x0 . The set of these path-homotopy classes does form a group under ∗. It will be called the
fundamental group of X relative to the base point x0 .

Definition 2.1.1 (Loop). Let X be a space and x0 ∈ X. A path in X that begins and ends at x0
is called a loop based at x0 .

A path f begins and ends at x0 means f : I → X is a continuous map with f (0) = f (1) = x0 .

Definition 2.1.2 (The Fundamental Group). The set of path homotopy classes of loops based at x0 ,
equipped with the operation ∗ between any two path homotopy classes, is called the fundamental
group of X relative to the base point x0 . It is denoted by π1 (X, x0 ).

Exercise 2.1. Convince yourself that π1 (X, x0 ) is indeed a group by using Theorem 1.1.3 from lecture
1.

Example 2.1.1
For the Euclidean space Rn , the fundamental group π1 (Rn , x0 ) is the trivial group (the group
consisting of the identity alone). Recall that the identity element of π1 (X, x0 ) is [ex0 ], the constant
path in X beginning and ending at x0 . So in this example, π1 (Rn , x0 ) = {[ex0 ]}.

In fact, one can prove that if X is a convex subset of Rh , then π1 (X, x0 ) = {[ex0 ]}.

Lemma 2.1.1
If X is a convex subset of Rn , then π1 (X, x0 ) = {[ex0 ]}.

Proof. Let f be a loop in X based at x0 . That is, f (0) = f (1) = x0 . We need to show that f 'p ex0 .
We define the continuous map H : I × I → X as

H(s, t) = (1 − t)f (s) + t ex0 (s) = (1 − t)f (t) + tx0

It can be done because X ⊆ Rn is convex. That’s why the straight line, given by the above equation,
connecting f (s) ∈ X and x0 ∈ X will lie in X. Here we have,
H(s, 0) = f (s) and H(s, 1) = ex0 (s)
H(0, t) = x0 = f (0) = ex0 (0) and H(1, t) = x0 = f (1) = ex0 (1)
Also, H is a conntinuous map because both f and ex0 are continuous. Hence H is a path homotopy
between f and ex0 . Therefore, for any loop f based at x0 , we have f ∈ [ex0 ]. So π1 (X, x0 ) =
{[ex0 ]}. ■

Question. To what extent the fundamental group depends on the base point?

§2.2 Path Connectedness and Simply Connectedness

33
2 Lecture 2 34

Definition 2.2.1 (Path Connected Space). A topological space X is path connected if for any
x0 , x1 ∈ X, there is a path from x0 to x1 .

If X is a path connected space, then X is connected. But the converse is not true, a connected space
may fail to be path connected.

Example 2.2.1
Consider “Topologist’s sine curve”
  
1
A = (x, y) ∈ R2 : x > 0 and y = sin
x


and it’s closure A = A∪ {0}×[−1, 1] . Both A and A are connected, but A is not path connected.

Consider the relation ∼ on X by x ∼ y if there exists a path between x and y. Then ∼ is an equivalence
relation and the equivalence classes are called the path components of X. If X is path connected,
then there is only one equivalence class.

Definition 2.2.2. Let α be a path in X from x0 to x1 . We define a map α̂ : π1 (X, x0 ) −→ π1 (X, x1 )

by the equation
α̂([f ]) = [ᾱ] ∗ [f ] ∗ [α]
Here ᾱ is the reverse of α.

Exercise 2.2. If f is a loop based at x0 , then check that ᾱ ∗ f ∗ α is a loop based at x1 .

x1
α
ᾱ
x0

Figure 2.1: ᾱ ∗ f ∗ α is a loop based at x1

The map α̂ is well-defined because the operation ∗ is a well defined product between path homotopy
classes.

Theorem 2.2.1
The map α̂ is a group isomorphism.

Proof. Let us first show that α̂ is a group homomorphism.

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

= [ᾱ] ∗ [f ] ∗ ([α] ∗ [ᾱ]) ∗ [g] ∗ [α]
= [ᾱ] ∗ [f ] ∗ [ex0 ] ∗ [g] ∗ [α]
= [ᾱ] ∗ [f ] ∗ [g] ∗ [α]
= [ᾱ] ∗ ([f ] ∗ [g]) ∗ [α] = α̂ ([f ] ∗ [g])

34
2 Lecture 2 35

Now we shall prove that α̂ is invertible. It would prove that α̂ is bijective. Let β = ᾱ be the reverse
of α. Consequently β̄ = α. We claim that β̂ is the inverse of α̂.
Let [h] ∈ π1 (X, x1 ). Then we have
 
β̂ ([h]) = β̄ ∗ [h] ∗ [β] = [α] ∗ [h] ∗ [ᾱ]

As a result,
   
α̂ β̂ ([h]) = [ᾱ] ∗ β̂ ([h]) ∗ [α]
= [ᾱ] ∗ [α] ∗ [h] ∗ [ᾱ] ∗ [α]
= [ex1 ] ∗ [h] ∗ [ex1 ]
= [h]

Now let [f ] ∈ π1 (X, x0 ). Then,

β̂ α̂ ([f ]) = [α] ∗ α̂ ([f ]) ∗ [ᾱ]
= [α] ∗ [ᾱ] ∗ [f ] ∗ [α] ∗ [ᾱ]
= [ex0 ] ∗ [f ] ∗ [ex0 ]
= [f ]

So β̂ = α̂−1 . Therefore α̂ is invertible, and so α̂ is a bijective homomorphism. Hence α̂ is a group

isomorphism. ■

Corollary 2.2.2
If X is path connected and x0 and x1 are two points of X then π1 (X, x0 ) is isomorphic to
π1 (X, x1 ).

Let X be a topological space that is not path connected and x0 ∈ X. Also let C be the path component
of X that contains the point x0 . All the path homotopy classes of loops based at x0 belong in the
same equivalence class or path component of X containing x0 . So we have

π1 (X, x0 ) = π1 (C, x0 )

Remark. If X is path connected, then all the groups π1 (X, x) are isomorphic. In particular, π1 (X, x0 ) ∼
=
π1 (X, x1 ) for x0 , x1 ∈ X. But the isomorphism between π1 (X, x0 ) and π1 (X, x1 ) depends on the
choice of path from x0 to x1 . However, the isomorphism will be independent of the chosen path
between x0 and x1 if and only if the fundamental group is abelian (problem ??).

Definition 2.2.3 (Simply Connected Space). A space X is said to be simply connected if it is

a path-connected space and if π1 (X, x0 ) is the trivial (one-element) group for some x0 ∈ X, and
hence for every x0 ∈ X. In other words,

π1 (X, x0 ) = {[ex0 ]} ∀ x0 ∈ X

Abuse of Notation. We often express the fact that π1 (X, x0 ) is the trivial group by writing
π1 (X, x0 ) = 0.

35
2 Lecture 2 36

Lemma 2.2.3
In a simply connected space X, any two paths having the same initial and final points are path
homotopic.

Proof. Let α and β be two paths in X, both from x0 to x1 . Then α ∗ β̄ is a loop in X based at x0 .
Since X is a simply connected space, we have α ∗ β̄ = [ex0 ].
 
α ∗ β̄ ∗ [β] = [ex0 ] ∗ [β]
 
=⇒ [α] ∗ β̄ ∗ [β] = [β]
=⇒ [α] ∗ [ex1 ] = [β]
=⇒ [α] = [β]
Therefore, α 'p β. ■

§2.3 Induced Homomorphism and Its Properties

Let X be a topological space with a distinguished base point x0 ∈ X. Such a topological space is
called a pointed topological space and is denoted by (X, x0 ).
Suppose hx0 : X → Y is a continuous map that carries the point x0 ∈ X to the point y0 ∈ Y ; in
other words, hx0 (x0 ) = y0 . We often write the map as a map between two pointed topological spaces
(X, x0 ) and (Y, y0 ).
hx0 : (X, x0 ) → (Y, y0 )
Now, if f is a loop in X based at x0 , then hx0 ◦ f : I → Y with
(hx0 ◦ f ) (0) = hx0 (f (0)) = hx0 (x0 ) = y0
and (hx0 ◦ f ) (1) = hx0 (f (1)) = hx0 (x0 ) = y0
In other words, hx0 ◦ f is a loop in Y based at y0 . The correspondence f → hx0 ◦ f gives rise to a map
from π1 (X, x0 ) to π1 (Y, y0 ). We define it formally as follows:

Definition 2.3.1. Let hx0 : (X, x0 ) → (Y, y0 ) be a continuous map. We define (hx0 )∗ : π1 (X, x0 ) →
π1 (Y, y0 ) by the following equation:

(hx0 )∗ ([f ]) = [hx0 ◦ f ]

The map (hx0 )∗ is called the homomorphism induced by hx0 , relative to the base point x0 in X.

The map (hx0 )∗ is well-defined. To check the well-definedness, let f 'p f ′ and F is a path homotopy
between them. Then by the key facts used in proving Theorem 1.1.3 in last lecture, hx0 ◦ F is a path
homotopy between hx0 ◦ f and hx0 ◦ f ′ . So, for [f ] = [f ′ ], we have [hx0 ◦ f ] = [hx0 ◦ f ′ ]. As a result,
(hx0 )∗ is a well-defined map.

Proposition 2.3.1
(hx0 )∗ is a gruop homomorphism.

Proof. Let [f ] , [g] ∈ π1 (X, x0 ).

(hx0 )∗ [f ] ∗ [g] = (hx0 )∗ ([f ∗ g])
= [hx0 ◦ (f ∗ g)]
= [(hx0 ◦ f ) ∗ (hx0 ◦ g)]
= [hx0 ◦ f ] ∗ [hx0 ◦ g]



= (hx0 )∗ [f ] ∗ (hx0 )∗ [g]
So (hx0 )∗ is indeed a group homomorphism. ■

36
2 Lecture 2 37

As the notation suggests, the induced homomorphism depends not only on the continuous map hx0 ,
but also on the choice of the base point x0 .

Theorem 2.3.2
If hx0 : (X, x0 ) → (Y, y0 ) and ky0 : (Y, y0 ) → (Z, z0 ) are continuous maps between pointed
topological spaces, then
(ky0 ◦ hx0 )∗ = (ky0 )∗ ◦ (hx0 )∗
Furthermore, if ix0 : (X, x0 ) → (X, x0 ) is
the identity map, then (ix0 )∗ : π1 (X, x0 ) → π1 (X, x0 )
is the identity homomorphism (ix0 )∗ [f ] = [f ].

Proof. Note that, the composition map ky0 ◦hx0 is continuous from (X, x0 ) to (Z, z0 ). Indeed ky0 ◦hx0 :
(X, x0 ) → (Z, z0 ) as a map between pointed topological spaces, as z0 = ky0 (y0 ) = ky0 (hx0 (x0 )) =
(ky0 ◦ hx0 ) (x0 ). Therefore,


(ky0 ◦ hx0 )∗ : π1 (X, x0 ) → π1 (Z, z0 ) with (ky0 ◦ hx0 )∗ [f ] = [(ky0 ◦ hx0 ) ◦ f ]

for a given [f ] ∈ π1 (X, x0 ).

On the other hand, since (hx0 )∗ : π1 (X, x0 ) → π1 (Y, y0 ) and (ky0 )∗ : π1 (Y, y0 ) → π1 (Z, z0 ), one has
(ky0 )∗ ◦ (hx0 )∗ : π1 (X, x0 ) → π1 (Z, z0 ) with





(ky0 )∗ ◦ (hx0 )∗ [f ] = (ky0 )∗ (hx0 )∗ [f ]


= (ky0 )∗ [hx0 ◦ f ]
= [ky0 ◦ (hx0 ◦ f )]

We know that composition of maps is associative. Therefore,

(ky0 ◦ hx0 )∗ = (ky0 )∗ ◦ (hx0 )∗

Again for the identity map ix0 : (X, x0 ) → (X, x0 ) we have ix0 (x) = x for every x ∈ X. So the induced
homomorphism (ix0 )∗ : π1 (X, x0 ) → π1 (X, x0 ) gives us


(ix0 )∗ [f ] = [ix0 ◦ f ] = [f ] ∀ [f ] ∈ π1 (X, x0 )

So (ix0 )∗ is indeed the identity group homomorphism. ■

Theorem 2.3.2 is often known as functorial properties1 of induced homomorphism.

Corollary 2.3.3
If hx0 : (X, x0 ) → (Y, y0 ) is a homeomorphism from X to Y , then (hx0 )∗ : π1 (X, x0 ) → π1 (Y, y0 )
is a group isomorphism.

Proof. By Proposition 2.3.1, (hx0 )∗ is a group homomorphism. So we only need to prove that the
inverse of (hx0 )∗ exists.
hx0 : (X, x0 ) → (Y, y0 ) is continuous with continuous inverse. Let ky0 : (Y, y0 ) → (X, x0 ) be the
inverse of hx0 . Hence ky0 ◦ hx0 : (X, x0 ) → (X, x0 ) is the identity map ix0 on (X, x0 ). Similarly,
hx0 ◦ kx0 : (Y, y0 ) → (Y, y0 ) is the identity map jy0 on (Y, y0 ). That is

ky0 ◦ hx0 = ix0 and hx0 ◦ kx0 = jy0

Applying Theorem 2.3.2,

(ky0 ◦ hx0 )∗ = (ky0 )∗ ◦ (hx0 )∗ and (hx0 ◦ ky0 )∗ = (hx0 )∗ ◦ (ky0 )∗

=⇒ (ky0 )∗ ◦ (hx0 )∗ = (ix0 )∗ =⇒ (hx0 )∗ ◦ (ky0 )∗ = (jy0 )∗
1
However, this naming is not random. See Appendix A for more details.

37
2 Lecture 2 38

(ix0 )∗ and (jy0 )∗ are identity group homomorphism on π1 (X, x0 ) and π1 (Y, y0 ) respectively. Therefore,
(hx0 )∗ and (ky0 )∗ are inverses of one another. It means the inverse of (hx0 )∗ exists, hence it’s a group
isomorphism. ■

38
3 Lecture 3
Any convex subspace of the Euclidean space Rn has trivial fundamental group. We now need to
compute fundamental groups of topological spaces that are not necessarily trivial. One useful technique
to achieve this is to use the tool of covering spaces.

§3.1 Covering Space

Definition 3.1.1. Let p : E → B be a continuous surjective map. The open set U of B is said to
be evenly covered by p if the inverse image p−1 (U ) can be written as the union of disjoint open
sets Vα in E such that for each α, the restriction of p to Vα is a homeomorphism of Vα onto U .
The collection {Vα } will be called a partition of p−1 (U ) into slices.

We often picture p−1 (U ) as a some copies of U ; the map p maps them all down onto U (Firgure 3.1).

p−1 (U )

Figure 3.1: Even covering of U ⊆ B by p : E → B.

Definition 3.1.2. Let p : E → B be continuous and surjective. If every point b ∈ B has a

neighborhood U that is evenly covered by p, then p is called a covering map; and E is called a
covering space of B.

There are a few properties of p : E → B that ensue from its definition.

Lemma 3.1.1
If p : E → B is a covering map, then for every b ∈ B the subspace p−1 (b), equipped with the
subspace topology inherited from E, has the discrete topology.

Proof. We claim that p−1
(b)∩Vα is a singleton set for each α ∈ J. Assume for the sake of contradiction
that, c, d ∈ p−1 (b) ∩ Vα with c 6= d for a given α. Then we would have that, for c, d ∈ Vα , p(c) =

39
3 Lecture 3 40

p(d) = b. As a result, p Vα
: Vα → U is no longer injective, and hence not a homeomorphism.
Contradiction! Also,
!
G G G

p−1 (b) ⊆ p−1 (U ) = Vα =⇒ p−1 (b) = p−1 (b) ∩ Vα = p−1 (b) ∩ Vα
α α α

Therefore, −1
 −1each element of p (b) is an element belonging to a singleton set of the mutually disjoint
family p (b) ∩ Vα α . Each Vα is open in E. According to the definition of Subspace topology,
p−1 (b) ∩ Vα is open in p−1 (b). As a result, every singleton set is open in p−1 (b). So p−1 (b) has the
discrete topology. ■

Lemma 3.1.2
If p : E → B is a covering map, then p is an open map.

Proof. Let A ⊆ E be open. We need to show that p(A) is open in B. Take x ∈ p(A). We need to find
a neighborhood of x that is contained inside p(A).
Since p is a covering map of B, one can choose a neighborhoodFU of x that is evenly covered by p.
Let {Vα }α∈J be a partition of p−1 (U ) into slices, i.e., p−1 (U ) = Vα .
α∈J
−1
FSince x ∈ p(A), there exists some y ∈ A such that x = p(y). Now p(y) = x ∈ U =⇒ y ∈ p (U ) =
Vα . Let Vβ be the slice containing y.
α∈J
Vβ is a subset of E, and A is open in E. Thus Vβ ∩ A is open in Vβ with respect to the subspace
topology. p maps Vβ homeomorphically onto U , so it maps open subsets of Vβ to open subsets of U .
Hence p (Vβ ∩ A) is open in U . Since U is open in B, by Lemma 0.2.2, p (Vβ ∩ A) is open in B.
We know that y ∈ Vβ and y ∈ A, so y ∈ Vβ ∩ A =⇒ x = p(y) ∈ p (Vβ ∩ A) ⊆ p(A), in particular
x ∈ p (Vβ ∩ A) ⊆ p(A). So we have found our desired open neighborhood of x, contained in p(A).
Hence p(A) is open. ■

Example 3.1.1
Let X be an arbitrary topological space. Set E = X × {1, 2, . . . , N }, with the latter set given the
discrete topology. The the projection p : E → X given by p(x, i) = x is a covering map.
−1
FNNote that the whole set X, being open in X, is evenly covered by p. Indeed p (X) =
i=1 X × {i}, where each X × {i} is open in E. It’s immediate that p X×{i} : X × {i} → X
is a homeomorphism. So p is indeed a covering map.

Theorem 3.1.3
Define p : R → S 1 by p(x) = (cos 2πx, sin 2πx). Then p is a covering map.

Proof. Clearly p is Surjective. Because every point y on S 1 has norm 1, so they can be expressed as
(cos θ, sin θ) for some θ ∈ R. So one can always find x ∈ R such that p(x) = y.
About the continuity, p can be expressed as follows:

p:R→R×R , x 7→ (p1 (x), p2 (x))

where p1 , p2 : R → R are given by p1 (x) = cos 2πx and p2 (x) = sin 2πx. Since both p1 and p2 are
continuous, p is continuous (by Theorem 18.4 from Munkres).

Look at the point (1, 0) in S 1 . It’s clear that p−1 (1, 0) = Z. Also, using the formula for p, one
obtains easily that
   
−1

1 −1

1
p (0, 1) = n + | n ∈ Z and p (0, −1) = n − | n ∈ Z
4 4

40
3 Lecture 3 41
F

As a result, the open arc U1 (colored blue in the figure below), has the preimage p−1 (U1 ) = n − 14 , n + 1
4 .
n∈Z
Similarly we find the preimage of other Ui ’s.
F
F

p−1 (U1 ) = n − 14 , n + 14 p−1 (U2 ) = n, n + 1
2
n∈Z n∈Z
U2
U1

U3

U4
F
F

p−1 (U3 ) = n + 14 , n + 3
4 p−1 (U4 ) = n + 12 , n + 1
n∈Z n∈Z

Now, each point of S 1 belong to at least one of the Ui ’s. So our job is done if we can show that each
Ui is evenly covered by p. Here we will only show that U1 is evenly covered by p. The proofs for U2 ,
U3 , U4 are similar. F

We saw that p−1 (U1 ) = Vn , where Vn = n − 41 , n + 14 . It’s immediate that the open sets Vn are
disjoint. So we now need to show that the p Vn : Vn → U1 is a homeomorphism. But first we will show
that p Vn : Vn → U1 is a homeomorphism, where X stands for the closure of X.
 
Clearly Vn = n − 41 , n + 14 . Firstly we need to check that p Vn is injective. Suppose we have
x1 , x2 ∈ Vn with p Vn (x1 ) = p Vn (x2 ). That is, (cos 2πx1 , sin 2πx1 ) = (cos 2πx2 , sin 2πx2 ). In other
words,
cos 2πx1 = cos 2πx2 and sin 2πx1 = sin 2πx2
 
But sin θ is a monotonically increasing continuous function for θ ∈ 2πn − π2 , 2πn + π2 . In this case
sin 2πx is monotonically increasing continuous function for x ∈ Vn . So, to acquire sin 2πx1 = sin 2πx2 ,
we must have x1 = x2 . Thus p Vn is injective.
Now we will prove the surjectivity of p Vn . Observe that, p Vn maps the endpoints of Vn to the
endpoints of U 1 , with he endpoints of U being (0, 1) and (0, −1).
      
1 1 1
p Vn n − = cos 2π n − , sin 2π n −
4 4 4
  π  π 
= cos − , sin − = (0, −1)
   2
  2  
1 1 1
p Vn n + = cos 2π n + , sin 2π n +
4 4 4
 π   π 
= cos , sin = (0, 1)
2 2

41
3 Lecture 3 42

p Vn is continuous, Vn is connected, and U1 is ordered. So by Intermediate value theorem, for any

c ∈ U1 there exists x ∈ Vn such that p Vn (x) = c. Hence p Vn is surjective.
We’ve proved that p Vn is a bijective continuous map from Vn to U1 . Vn is a bounded closed subset
of R, hence compact (by Heine-Borel Theorem). U1 is a subset of the Hausdorff space R2 , so it is
also Hausdorff (by Lemma 0.4.2). So p Vn being a continuous bijective map from a compact set to a
Hausdorff space, by Corollary 0.7.6, p Vn is a homeomorphism.
Restriction of a homeomorphism is again a homeomorphism. Therefore, p Vn : Vn → U1 is a
homeomorphism for every n ∈ Z. Hence U1 is evenly covered by p. In a similar manner, one can
prove that U2 , U3 , U4 are also evenly covered by p. So p : R → S 1 is indeed a covering map, and R is
a covering space of S 1 . ■

Definition 3.1.3 (Local Homeomorphism). The map p : E → B is called a local homeomorphism

if each point e ∈ E has a neighborhood U which is mapped homeomorphically by p onto an open
subset of B.

Let p : E → B be a covering map and e ∈ E. Also, let p(e) = x ∈ B. Since p is a covering F

map, there
−1
exists open U ⊆ B containing x, that is evenly covered by p. In other words, p (U ) = Vα with
α∈J
each Vα being open in E. F
Now, p(e) = x ∈ U =⇒ e ∈ p−1 (U ). Therefore, e ∈ α∈J Vα . So e ∈ Vα for some α ∈ J. From the
definition of covering map, p Vα : Vα → U is a homeomorphism. Therefore, for a given e ∈ E, we’ve
found an open set Vα ⊆ E that contains e; and Vα gets homeomorphically mapped to an open subset
U of B. Therefore, p is a local homeomorphism.
We’ve just seen that, if p : E → B is a covering map, then p is also a local homeomorphism.
However, the converse is not true in general. The following example illustrates a counterexample to
the converse.

Example 3.1.2
The map pe : R+ → S 1 given by pe(x) = (cos 2πx, sin 2πx) is surjective, and it’s a local homeomor-
phism. We have seen that there is an open cover {Ui }4i=1 of S 1 with
G 1 1
 G 1

−1 −1
p (U1 ) = n − ,n + , p (U2 ) = n, n +
4 4 2
n∈Z n∈Z
G   G  
−1 1 3 −1 1
p (U3 ) = n + ,n + , p (U4 ) = n + ,n + 1
4 4 2
n∈Z n∈Z

for the covering map p : R → S 1 given by p(x) = (cos 2πx, sin 2πx). Now, for this map pe, we can
calculate pe−1 (Ui ) in using p−1 (Ui ).
  G ! G
1 1 1
pe−1 (U1 ) = p−1 (U1 ) ∩ R+ = 0, t n − ,n + = Ven
4 4 4
n∈N n∈N∪{0}
G  1
 G
pe−1 (U2 ) = p−1 (U2 ) ∩ R+ = n, n + = Ven′
2
n∈N∪{0} n∈N∪{0}
G  1 3
 G
pe−1 (U3 ) = p−1 (U3 ) ∩ R+ = n + ,n + = Ven′′
4 4
n∈N∪{0} n∈N∪{0}
G   G
1
pe−1 (U4 ) = p−1 (U4 ) ∩ R+ = n + ,n + 1 = Ven′′′
2
n∈N∪{0} n∈N∪{0}

The arc U1 , an open subset of S 1 containing (1, 0), is not evenly covered by pe. Because, for

42
3 Lecture 3 43


the slice Ve0 = 0, 14 , the map pe Ve0 : Ve0 → U1 is not a homeomorphism, it’s just a topological
imbedding. So pe is not a covering map.
For any point x in R+ , some neighborhood of x will be contained in at least one of the open
sets Ven , Ven′ , Ven′′ , Ven′′′ for some n ∈ N ∪ {0}. Let W 3 x and W ⊆ Ven′′ (for example). Then one
can easily show that the map pe W : W → pe(W ) ⊆ U3 is a homeomorphism. Despite not being a
covering map, pe is a local homeomorphism.

In Example 3.1.2, we’ve seen that the restriction of a covering map, in general, is not a covering map.
The following theorem tells us when the restriction of a covering map will be a covering map.

Theorem 3.1.4
Let p : E → B be a covering map. If B0 is a subspace of B, and E0 = p−1 (B0 ); then the map
p0 : E0 → B0 obtained by restricting p is a covering map.

Proof. Given b0 ∈ B0 , let U be an open set in B containing b0 , this is F evenly covered by p. Let
{Vα }α∈J be a partition of p−1 (U ) into slices. In other words, p−1 (U ) = Vα .
α∈J
Each Vα is open in E. U ∩ B0 is open in B0 , with respect to the subspace topology, that contains
b0 . The sets {Vα ∩ E0 }α∈J are disjoint open sets in E0 , because for distinct α, β ∈ J

(Vα ∩ E0 ) ∩ (Vβ ∩ E0 ) = (Vα ∩ Vβ ) ∩ E0 = ∅ ∩ E0 = ∅

This disjoint collection of open sets cover p−1 (U ∩ B0 ).

!
G G
−1 −1 −1
p (U ∩ B0 ) = p (U ) ∩ p (B0 ) = Vα ∩ E0 = (Vα ∩ E0 )
α∈J α∈J

In addition, since p Vα : Vα → U is a homeomorphism, p Vα ∩E0

: Vα ∩ E0 → U ∩ B0 is also a
homeomorphism. Hence, p0 : E0 → B0 is a covering map. ■

43
4 Lecture 4
§4.1 Lifting Map

Definition 4.1.1 (Lifting). Let p : E → B be a map. If f is a continuous map from some topological
space X to B, a lifting of f is a map f¯ : X → E such that p ◦ f¯ = f .

f¯
p

X B
f

In this lecture we will be concerned with liftings when p is a covering map.

Example 4.1.1
Consider the covering map p : R → S 1 given by p(x) = (cos 2πx, sin 2πx). The path f : [0, 1] → S 1
beginning at (1, 0) ∈ S 1 given by f (s) = (cos πs, sin πs) lifts to the path f¯(s) = 2s beginning at 0
and ending at 21 .


s
(cos πs, sin πs) = f (s) = p f¯(s) = cos 2π f¯(s), sin 2π f¯(s) =⇒ f¯(s) =
2
The path g : [0, 1] → S 1 given by g(s) = (cos πs, − sin πs) lifts to the path ḡ(s) = − 2s beginning
at 0 and ending at − 21 .
s
(cos πs, − sin πs) = g(s) = p (ḡ(s)) = (cos 2πḡ(s), sin 2πḡ(s)) =⇒ ḡ(s) = −
2
The path h : [0, 1] → S 1 given by f (s) = (cos 4πs, sin 4πs) lifts to the path h̄(s) = 2s beginning
at 0 and ending at 2.



(cos 4πs, sin 4πs) = h(s) = p h̄(s) = cos 2π h̄(s), sin 2π h̄(s) =⇒ h̄(s) = 2s

0 1
2 1 − 12 0 1 0 2

p p p
f¯ ḡ h̄

I I I
f
g h

Lemma 4.1.1
Let p : E → B be a covering map, and p (e0 ) = b0 . Any path f : I → B beginning at b0 has a
unique lifting to a path f¯ in E beginning at e0 .

Proof. Since p : E → B is a covering map, for every x ∈ B, there exists a neighborhood Ux that is
evenly covered by p. Ux is open in B and it contains x, so the collection {Ux }x∈B is an open cover

44
4 Lecture 4 45

of B. f : I → B is a continuous map, so the preimages of open sets are open in I. Therefore, the
collection f −1 (Ux ) x∈B is an open cover of I.
I is bounded and closed subset of R. By Heine-Borel Theorem,  −1I is a compact metric space. Now
we shall use Lebesgue Number Lemma. We have an open cover f (Ux ) x∈B of the compact metric
space I. Let δ be the Lebesgue Number of this open cover. Now we are gonna partition [0, 1] into
subintervals [s0 , s1 ] , [s1 , s2 ] , . . . , [sn−1 , sn ], with s0 = 0 and sn = 1, such that

diam ([si , si+1 ]) = si − si−1 < δ , ∀ i ∈ {0, 1, 2, . . . , n − 1}

Each [si , si+1 ] has diameter less than δ. By Lebesgue Number Lemma,

[si , si+1 ] ⊆ f −1 (Ux ) for some x ∈ B =⇒ f ([si , si+1 ]) ⊆ Ux for some x ∈ B

Now we will inductively define the lift f¯ : I → E with f (0) = e0 . Note that, by the definition of lift,
one must have p ◦ f¯ = f . The base case of the inductive definition is

f¯ (0) = f¯ ([0, s0 ]) = 0

Then we assume that f¯ ([0, si ]) is defined. We need to show that f¯ ([0, si+1 ]) can be defined; in other
words f¯ ([si , si+1 ]) can be defined. Because if we can prove that f¯ can be defined on [si , si+1 ], we can
conclude by pasting lemma that f¯ can be defined on [0, si+1 ]. By inductive hypothesis, f¯ ([si , si+1 ]) is
defined. In particular, f (si ) is defined. Since p ◦ f¯ = f , we have

G
p f¯ (si ) = f (si ) ∈ Ux =⇒ f¯ (si ) ∈ p−1 (Ux ) = Vα =⇒ f¯ (si ) ∈ Vα0
α∈J

for a unique α0 ∈ J. For s ∈ [si , si+1 ], we define f¯(s) as

 
−1
f¯(s) = p Vα ◦ f (s)
0

−1
p V α0
is a homeomorphism, so p is continuous. Hence f¯ is composition of continuous maps, thus
V α0
continuous on [si , si+1 ]. This definition satisfies p ◦ f¯ = f , because

 
−1
p ◦ f¯ (s) = p ◦ p Vα ◦ f (s) = f (s)
0

So f¯ can be defined on [si , si+1 ] with the required properties. f¯ defined on the subintervals [si , si+1 ]
can be continuously extended to all of [0, 1] by pasting lemma, since f¯ is defined on the closed intervals
[s0 , s1 ] , [s1 , s2 ] , . . . , [sn−1 , sn ], with s0 = 0 and sn = 1 and f¯ agrees on the pointwise overlaps.
We’ve shown that the lifting f¯ can be defined on I with the desired properties. Now we need to
show that, f¯ is unique. The uniqueness can be proved inductively as well.
Suppose fe is another lifting of f beginning at e0 , i.e, fe(0) = e0 = f¯(0). So the base case of the
induction is f¯(s) = fe(s) for every s ∈ [0, s0 ]. Assume that f¯(s) = fe(s) for every s ∈ [0, si ]; this is the
inductive hypothesis. We want to show that f¯(s) = fe(s) for every s ∈ [si , si+1 ]. That will establish
f¯(s) = fe(s) for every s ∈ [0, si+1 ].
We’ve seen earlier from the inductive construction of f¯ that f¯ (si ) ∈ Vα0 for a unique α0 ∈ J. Also,
we defined f¯ on [si , si+1 ] by  
−1 −1
f¯(s) = p Vα ◦ f (s) = p Vα (f (s))
0 0

Since fe is a lift of f , we must have p ◦ fe = f . Now,

  G
f ([si , si+1 ]) ⊆ Ux =⇒ p ◦ fe ([si , si+1 ]) ⊆ Ux =⇒ fe([si , si+1 ]) ⊆ p−1 (Ux ) = Vα
α∈J

[si , si+1 ] is an interval, thus connected. By Theorem 23.5 from Munkres, the image of a connected
space under a continuous map is connected. So fe([si , si+1 ]) is connected. The slices Vα are open and

45
4 Lecture 4 46

disjoint. If fe([si , si+1 ]) is distributed among the Vα ’s, connectedness of fe([si , si+1 ]) is contradicted.
So fe([si , si+1 ]) must lie in a single slice, namely Vα1 .
But fe(si ) = f¯ (si ) ∈ Vα0 , so α1 = α0 . Hence fe([si , si+1 ]) ⊆ Vα0 . Let s ∈ [si , si+1 ] and y0 = fe(s) ∈
Vα0 . p Vα is a homeomorphism, so it’s invertible.
0
 
−1
f (s) = p fe(s) = p V α0
(y0 ) =⇒ y0 = p V α0
(f (s)) = f¯(s)

So fe and f¯ agrees on [si , si+1 ] and we are done. ■

In fact, the uniqueness of lifting has a generalization.

Lemma 4.1.2
Let p : E → B be a covering map and X be a connected topological space. Given any two
continuous maps fe0 , fe1 : X → E such that p ◦ fe0 = p ◦ fe1 =: f (in other words, both fe0 and fe1
are lifts of f ), consider the set
n o
A = x ∈ X : fe0 (x) = fe1 (x)

Then A = ∅ or A = X.

Proof. The only sets that are both open and closed in a connected space are the empty set and the
whole set itself. So this lemma is equivalent to proving A is both open and closed.
Firstly we will show that A is closed. It is enough to show that A = A. Let y ∈ A, we nned to show
that y ∈ A, i.e. fe0 (y) = fe1 (y). Assume for the sake of contradiction that fe0 (y) 6= fe1 (y).
   
p fe0 (y) = p fe1 (y) = f (y) =: x

Consider a neighborhood Ux of x that is evenly covered by p. Now,

    G
p fe0 (y) = p fe1 (y) ∈ Ux =⇒ fe0 (y), fe1 (y) ∈ p−1 (Ux ) = Vα
α∈J

Let V0 and V1 be the disjoint open slices contaiing fe0 (y) and fe1 (y) respectively. fe0 :: X → E, V0 is
open in E, so V0 is a neighborhood of fe0 (y) in E. By Lemma 0.5.3, there exists a neighborhood W0
of y in X such that fe0 (W0 ) ⊆ V0 . Similarly, there exists a neighborhood W1 of y in X such that
fe1 (W1 ) ⊆ V1 .
Now W = W0 ∩ W1 is a neighborhood of y in X.

fe0 (W ) ⊆ V0 , fe1 (W ) ⊆ V1 =⇒ fe0 (W ) ∩ fe1 (W ) = ∅

But since y ∈ A and W is a neighborhood of y, W ∩A 6= ∅. This contradicts with fe0 (W )∩ fe1 (W ) = ∅.

Therefore, y ∈ A. As a result A ⊆ A.By the definition of closure, A ⊆ A. Hence A = A and A is
closed.
To prove that A is open, take any y ∈ A. So we have fe0 (y) = fe1 (y).
   
p fe0 (y) = p fe1 (y) = f (y) =: x

Consider a neighborhood Ux of x that is evenly covered by p. Now,

    G
p fe0 (y) = p fe1 (y) ∈ Ux =⇒ fe0 (y), fe1 (y) ∈ p−1 (Ux ) = Vα
α∈J

Let V0 be the open slice contaiing fe0 (y) = fe1 (y). fe0 :: X → E, V0 is open in E, so V0 is a neighborhood
of fe0 (y) in E. By Lemma 0.5.3, there exists a neighborhood W0 of y in X such that fe0 (W0 ) ⊆ V0 .
Similarly, there exists a neighborhood W1 of y in X such that fe1 (W1 ) ⊆ V0 .

46
4 Lecture 4 47

Now W = W0 ∩ W1 is a neighborhood of y in X. fe0 (W ) ⊆ V0 , fe1 (W ) ⊆ V0 Let w ∈ W , so

f0 (w), fe1 (w) ∈ V0 . p0 = p V0 is a homeomorphism, so it’s injective.
e
   
f (w) = p0 fe0 (w) = p0 fe1 (w) =⇒ fe0 (w) = fe1 (w) =⇒ w ∈ A =⇒ W ⊆ A

W is a neighborhood of y that is fully contained in A. So every y ∈ A is an interior point, hence A is

open.
Therefore, A is both open and closed. Hence A = ∅ or A = X. ■

Lemma 4.1.3
Let p : E → B be a covering map; let p (e0 ) = b0 . Let F : I × I → B be continuous with
F (0, 0) = b0 . Then there is a unique lifting of F to a continuous map Fe : I × I → E such that
Fe (0, 0) = e0 . Furthermore, if F is a path homotopy, so is Fe .

Proof. Given F , we first define F (0, 0) = e0 . Then we can extend F to the left hand edge 0 × I and
bottom edge I × 0 of the square I × I using Lemma 4.1.1. Because F restricted on 0 × I is basically
a path, so F 0×I ≡ f : I → B. We can definitely find a unique lifting f¯ of f that starts at b0 . We can
then define Fe (0, t) = f¯(t). In a similar manner, we can extend Fe on I × 0 too. Now we will extend it
to whole I × I.
Since p : E → B is a covering map, for every x ∈ B, there exists a neighborhood Ux that is evenly
covered by p. Ux is open in B and it contains x, so the collection {Ux }x∈B is an open cover of B.
F : I ×I → B is a continuous map, so the preimages of open sets are open in I × I. Therefore, the
−1
collection F (Ux ) x∈B is an open cover of I.
I × I is bounded and closed subset of R2 . By Heine-Borel Theorem, I is a compact metric space.
Let δ be the Lebesgue Number of this open cover. Now we are gonna choose subdivisions

0 = s0 < s1 < s2 · · · < sm = 1 and 0 = t0 < t1 < t2 · · · < tn = 1

such that diam ([si−1 , si ] × [tj−1 , tj ]) < δ for every i ∈ {1, 2, . . . , m} , j ∈ {1, 2, . . . , n}. Let Ii =
[si−1 , si ] and Jj = [tj−1 , tj ]. By Lebesgue Number Lemma,

Ii × Jj ⊆ F −1 (Ux ) for some x ∈ B =⇒ F (Ii × Jj ) ⊆ Ux for some x ∈ B

Now we shall define F : I × I → E inductively beginning with I1 × J1 , then continuing with the other
rectangles Ii × J1 of the “bottom row”, and then with the rectangles Ii × J2 and so on. Now, fix
1 ≤ i0 ≤ m and 1 ≤ j0 ≤ n. Denote by A the union of 0 × I and I × 0 and all the rectangles “previous”
to Ii0 × Jj0 . In other words,
  !
[ [
A = (0 × I) ∪ (I × 0) ∪  (Ii × Jj ) ∪ (Ii × Jj0 )
j<j0 i<i0

47
4 Lecture 4 48

Ii0 × Jj0

V0

Fe

p0

I ×I Ux

A is colored in grey in the figure above, and the edges of A are the thick black edges. For our inductive
construction, the base case is Fe can be constructed on I1 × J1 . By the inductive hypothesis, the lifting
Fe of F A can be constructed. Now we will define Fe on Ii0 × Jj0 in the following way.
One can find Ux being open in B that is evenly covered by p such that F (Ii0 × Jj0 ) ⊆ Ux . Let
C = A ∩ (Ii0 × Jj0 ) be the union of left and bottom edge of Ii0 × Jj0 . According to the inductive
hypothesis, Fe is defined on C ⊂ A.
  G
p Fe (Ii0 × Jj0 ) = F (Ii0 × Jj0 ) ⊆ Ux =⇒ Fe (Ii0 × Jj0 ) ⊆ p−1 (Ux ) = Vα
α∈J
G
=⇒ Fe (C) ⊆ Fe (Ii0 × Jj0 ) ⊆ Vα
α∈J

C is connected, Fe is continuous, so Fe (C) is connected. The slices Vα are open and disjoint. If Fe (C)
is distributed among the Vα ’s, connectedness of Fe (C) is contradicted. So Fe (C) must lie in a single
slice Vα . Let us denote the Vα entirely containing Fe (C) by V0 . Let p0 : V0 → Ux be the restriction of
p on V0 . p0 is a homeomorphism. We extend Fe from C to Ii0 × Jj0 by


Fe (x) = p−1 −1
0 ◦ F (x) = p0 (F (x))

Now we are only left with the base case of I1 × J1 . We showed in the beginning that Fe can be
constructed on (0 × I)∪(I × 0). And the left and bottom edge of I1 ×J1 is a subset of (0 × I)∪(I × 0).
Using this, we can construct Fe on I1 × J1 using a similar construction as we’ve just done for Ii0 × Jj0 .
Thus the base case is proved.
So we have proved that Fe can be constructed on each of the rectangles Ii × Ji . The rectangles
are closed, and Fe agrees on the boundaries of adjecent rectangles. Therefore, by pasting lemma, the
extended map Fe on I × I is continuous. By construction, it satisfies p ◦ Fe = F . So Fe is indeed a
lifting of F .
About the uniqueness issue, assume the contrary. If F̄ : I ×I → E is another lifting of F : I ×I → B
with F̄ (0, 0) = e0 , then the set
n o
(x, y) ∈ I × I : Fe (x, y) = F̄ (x, y

contains (0, 0), so it is not empty. Hence, by Lemma 4.1.2, this set equal to I × I. Therefore, Fe is
unique.
Now, let g and h be two paths in B and F be a path homotopy between them. F (0, 0) = b0 , so
both g and h start at b0 . By Lemma 4.1.1, there exists unique lifts of g and h, namely ge and e
h, such

48
4 Lecture 4 49

that ge(0) = e
h(0) = e0 . So g = p ◦ ge and h = p ◦ e h. We shall prove that Fe is a path homotopy between
e
ge and h in E. Suppose they both end at b1 .
 
F (s, 0) = (p ◦ ge) (s) and F (s, 1) = p ◦ e
h (s)
   
F (0, t) = (p ◦ ge) (0) = p ◦ e h (0) = b0 and F (1, t) = (p ◦ ge) (1) = p ◦ eh (1) = b1

−1
Recall that, we defined the unique map Fe as Fe = p V0
◦ F . Hence
−1 −1 −1
Fe (s, 0) = p V0
(F (s, 0)) = p V0
((p ◦ ge) (s)) = p V0 (p (e
g (s))) = ge(s)
     
−1 −1 e −1 e
Fe (s, 1) = p V0
(F (s, 1)) = p V0
p ◦ h (s) = p V0
p h(s) =e h(s)
 
We’ve seen before that p Fe (0 × I) = F (0 × I) = {b0 }, so Fe (0 × I) ⊆ p−1 (b0 ). 0 × I is connected,
so is Fe (0 × I). By Lemma 3.1.1, p−1 (b0 ) has the discrete topology; so every singleton set is open in
p−1 (b0 ). If Fe (0 × I) has more that one element, it can be splitted into two disjoint nonempty open
sets, contradicting its connectedness. Hence Fe (0 × I) must be a singleton set. Similarly Fe (1 × I) is
also a singleton set. In fact,
−1 −1 −1
Fe (0, t) = p V0
(F (0, t)) = p V0
((p ◦ ge) (0)) = p V0 (p (e
g (0))) = ge(0)
     
−1 e −1 e
=p V0
p ◦ h (0) = p V0
p h(0) =e h(0)
−1 −1 −1
Fe (1, t) = p V0
(F (1, t)) = p V0
((p ◦ ge) (1)) = p V0 (p (e
g (1))) = ge(1)
     
−1 e −1 e
=p V0
p ◦ h (1) = p V0
p h(1) =e h(1)

Therefore, Fe is a path homotopy between ge and e

h. ■

Using Lemma 4.1.1 and Lemma 4.1.3, we can have the following theorem.

Theorem 4.1.4
Let p : E → B be a covering map and p (e0 ) = b0 . Let f and g be two paths in B from b0 to b1 ;
let f¯ and ḡ be their respective unique liftings to paths in E beginning at e0 . If f 'p g, then f¯
and ḡ end at the same point and f¯ 'p ḡ.

§4.2 Lifting Correspondence

Definition 4.2.1 (Lifting Correspondence). Let p : E → B be a covering map and b0 ∈ B. Choose

e0 ∈ E such that p (e0 ) = b0 . Given an element [f ] ∈ π1 (B, b0 ), let f¯ be the unique lifting of f to
a path in E such that f¯(0) = e0 . Then we define a set map ϕ

ϕ : π1 (B, b0 ) → p−1 (b0 ) , ϕ ([f ]) = f¯(1)

We call ϕ the lifting correspondence derived from the covering map p. It depends on the choice
of the point e0 .

ϕ is indeed a well-defined set map. Since [f ] ∈ π1 (B, b0 ), f is a loop in B based at b0 .

b0 = f (0) = f (1) = p f¯(1) =⇒ f¯(1) ∈ p−1 (b0 )

Now let [f ] = [g] for two loops f and g. So f and g are path homotopic. Then by Theorem 4.1.4, f¯
and ḡ ends at the same point. So f¯(1) = ḡ(1). Therefore, ϕ ([f ]) = ϕ ([g]). Hence ϕ is a well-defined
set map.

49
4 Lecture 4 50

Theorem 4.2.1
Let p : E → B be a covering map and p (e0 ) = b0 . If E is path connected, then the lifting
correspondece
ϕ : π1 (B, b0 ) → p−1 (b0 )
is surjective. If E is simply connected, it is bijective.

Proof. Suppose E is path connected. p (e0 ) = b0 , so e0 ∈ p−1 (b0 ) ⊆ E. Now, take any e1 ∈ p−1 (b0 ).
e0 and e1 are elements of E, so there exists a path f¯ in E from e0 to e1 . In other words, f¯ : I → E
such that f¯(0) = e0 and f¯(1) = e1 . Consider the map f = p ◦ f¯. It is a map from I to B. In fact, it
is a loop based at b0 . Because



f (0) = p f¯(0) = p (e0 ) = b0 and f (1) = p f¯(1) = p (e1 ) = b0

So [f ] ∈ π1 (B, b0 ). It has a unique lifting in E that starts at e0 . f¯ is one such lift. Therefore,

ϕ ([f ]) = f¯(1) = e1

In other words, for every e1 ∈ p−1 (b0 ), we can find [f ] ∈ π1 (B, b0 ) such that ϕ ([f ]) = e1 . Hence ϕ is
surjective.
Now suppose E is simply connected. Then it is also path connected. As a result, ϕ is surjective. So
we only need to prove that ϕ is injective. Let [f ] , [g] ∈ π1 (B, b0 ) such that ϕ ([f ]) = ϕ ([g]). We need
to show that [f ] = [g].
Let f¯ and ḡ be unique lifts of f and g, respectively, such that both of them start at e0 . In other
words, f¯, ḡ : I → E with f¯(0) = ḡ(0) = e0 and p ◦ f¯ = f, p ◦ ḡ = g. Since ϕ ([f ]) = ϕ ([g]), we have
f¯(1) = ḡ(1). Thus, f¯ and ḡ have the same initial and final points in a simply connected space E.
Therefore, by Lemma 2.2.3, f¯ ' ḡ.
Let F̄ be the path homotopy between f¯ and ḡ. By the 1st key fact used in proving Theorem 1.1.3,
p ◦ F̄ is a path homotopy between p ◦ f¯ = f and p ◦ ḡ = g. So f ' g, and thus we have [f ] = [g].
Hence ϕ is injective. ■

Now the main theorem of this lecture.

Theorem 4.2.2
The fundamental group of S 1 is isomorphic to the additive group of integers.

Proof. Let p : R → S 1 be the covering map of Theorem 3.1.3, defined by p(x) = (cos 2πx, sin 2πx).
Let e0 = 0, and b0 = p (e0 ) = p(0) = (1, 0). We’ve seen earlier that p−1 (b0 ) = Z. Since R is simply
connected, by Theorem 4.2.1, the lifting correspondence


ϕ : π1 S 1 , b0 → Z

is bijective. So we only need
to show that ϕ is a group homomorphism.

Given [f ] , [g] ∈ π1 S 1 , b0 , let f¯ and ḡ be their respective unique liftings on R both starting at
0 = e0 . Let ϕ ([f ]) = f¯(1) = n and ϕ ([g]) = ḡ(1) = m, where m and n are integers. Now, we define a
new path ge on R by
ge : I → R , ge(s) = n + ḡ(s)
From the definition of p, one immediately finds that

p(n + x) = (cos (2πn + 2πx) , sin (2πn + 2πx)) = (cos 2πx, sin 2πx) = p(x)

(p ◦ ge) (s) = p (e
g (s)) = p (n + ḡ(s)) = p (ḡ(s)) = g(s)
Also, ge(0) = n + ḡ(0) = n. So ge is the unique lifting of g thet begins at n. Since f¯(1) = n = ge(0), we
can form the product of paths f¯ ∗ ge. We claim that f¯ ∗ ge is the lift of f ∗ g that begins at 0.

50
4 Lecture 4 51

Using the 2nd key fact used in proving Theorem 1.1.3, we obtain that



p ◦ f¯ ∗ ge = p ◦ f¯ ∗ (p ◦ ge) = f ∗ g

Also, f¯ ∗ ge begins at where f¯ begins, i.e. it begins at f¯(0) = 0. So it is indeed the unique lift of f ∗ g
that begins at 0. Now, the endpoint of f¯∗e g is the endpoint of ge, i.e. it ends at ge(1) = n+ ḡ(1) = n+m.
Then using the definition of lifting correspondence, we get


ϕ ([f ] ∗ [g]) = ϕ ([f ∗ g]) = f¯ ∗ ge (1) = n + m = ϕ ([f ]) + ϕ ([g])


So ϕ is indeed a homomorphism from π1
S 1 , b0 to the additive group of integers. We’ve seen earlier
that ϕ is bijective. Therefore, π1 S 1 , b0 is isomorphic to Z. ■

51
5 Lecture 5
§5.1 Retraction

Definition 5.1.1 (Retraction). If A ⊂ X, a retraction of X onto A is a continuou map r : X → A

such that r A = idA . If such r exists, we say that A is a retract of X.

Lemma 5.1.1
If A is a retract of X, then the homomorphism of fundamental groups induced by the inclusion
map j : A → X is injective.

Proof. If r : X → A is a retraction, then r ◦ j = ia is the identity map of the pointed topological space
(A, a). By Theorem 2.3.2, (r ◦ j)∗ = r∗ ◦ j∗ = (ia )∗ is the identity homomorphism of the fundamental
group π1 (A, a). Therefore, j∗ has a left inverse.

Claim — Let f : X → Y and g : Y → X such that g ◦ f = idX (in other words, f has a left
inverse). Then f is injective.

Proof. Suppose f (x1 ) = f (x2 ), we need to show that x1 = x2 .

f (x1 ) = f (x2 ) =⇒ g (f (x1 )) = g (f (x2 )) =⇒ idX (x1 ) = idX (x2 ) =⇒ x1 = x2

So f is injective. □

By the claim, j∗ is injective. ■

Theorem 5.1.2 (No-retraction theorem)

There is no retraction of B 2 onto S 1 .

Proof. Assume the contrary. By Lemma 5.1.1, the homomorphism

induced by the inclusion map
j : S 1 → B 2 is injective. In other words, j∗ : π1 S 1 , b0
→ π1 B 2 , b0 is injective.
As B 2 is convex, the fundamental
group π1 B 2 , b0 is trivial (contains only the identity element).
But the fundamental group π1 S 1 , b0 is non-trivial (we know from Theorem
4.2.2 that

this is isomor-
phic to Z). So it’s not possible to have an injective map from π1 S 1 , b0 to π1 B 2 , b0 . Contradiction!
Hence S 1 is not a retract of B 2 . ■

Lemma 5.1.3
Let h : S 1 → X be continuous. Then the following are equivalent:
(1) h is nulhomotopic.
(2) h extends to a continuos map k : B 2 → X.
(3) h∗ is the trivial homomorphism of fundamental groups.

Proof. (1)⇒(2). Let H : S 1 × I → X be a path homotopy between h and a constant map. We define
π : S 1 × I → B 2 by π (x, t) = (1 − t)x.

Claim — π is a quotient map.

Proof. Consider a map π1 : R3 → R2 defined by π1 (x1 , x2 , t) = ((1 − t) x1 , (1 − t) x2 ). The map

52
5 Lecture 5 53

(x1 , x2 , t) 7→ 1 − t is continuous, so is (x1 , x2 , t) 7→ x1 . Then as a product of continuous real

values functions, (x1 , x2 , t) 7→ (1 − t) x1 is continuous. Similarly, the second coordinate of π1 is
also continuous. Therefore, π1 is continuous. Since π is the restriction of π1 on S 1 × I, π is also
continuous.
If y ∈ B 2 , then kyk ≤ 1. If kyk = 0, y = 0. Then taking t = 1 gives us π (x, t) = 0. If kyk 6= 0,
y
Then ∥y∥ has unit norm, so
 
y y
π , 1 − kyk = kyk =y
kyk kyk

So π is surjective.
S 1 × I is a bounded closed subset of R3 , so it’s compact by Heine-Borel Theorem. B 2 is a
subspace of R2 , by Lemma 0.4.2, B 2 is hausdorff. π is a continuous map from a compact space
to a hausdorff space. Therefore, by Proposition 0.7.5, π is a closed map.
π is continuous, surjective, closed. Lemma 0.10.1 gives us π is a quotient map. □

We want this following diagram to commute. In other words, we want H = k ◦ π.

S1 × I

π H

B2 k
X

Note that, the preimage of 0 under π is S 1 × {1}. As H is a homotopy between h and a constant
map,


H (x, 1) = ex0 (x) = x0 =⇒ H π −1 ({0}) = H S 1 × {1} = {x0 }
where ex0 is a constant map that maps all of S 1 to x0 ∈ X. So H is constant on π −1 ({0}).
For b 6= 0, π −1 ({b}) is a singleton set. Because,
b
(1 − t)x = b =⇒ |1 − t| kxk = kbk =⇒ 1 − t = kbk =⇒ t = 1 − kbk and x =
kbk

So H is trivially constant on π −1 ({b}). Therefore, by Theorem 0.10.3, H induces a map k : B 2 → X

such that k ◦ π = H. Since H is continuous, by the same theorem, k is continuous.
Now we need to show that k is an extension of h. For x ∈ S 1 , H (x, 0) = h(x) since H is a homotopy.
As a result,
h(x) = H (x, 0) = k (π (x, 0)) = k (x) =⇒ k S 1 = h
So h extends to a continuous map k : B 2 → X.
(2)⇒(3). If j : S 1 → B 2 is the inclusion map, then h = k ◦ j.
j k
S1 B2 X

h=k ◦ j


Hence h∗ =
(k ◦ j)∗ = k∗
◦ j∗ . As B 2 is convex, the fundamental group π1 B 2 , b0 is trivial. So
j∗ : π1 S
1 , b0 → π1 B 2 , b0 must be the trivial homomorphism, a homomorphism

that maps all of
1
π1 S , b0 to the identity element 2
of the one-element group π1 B , b0 .



We take any g ∈ π1 S 1 , b0 . Let e1 and e2 respectively denote the identity elements of π1 B 2 , b0
and π1 (X, x0 ), where x0 = h(b0 ). We’ve just shown that j∗ (g) = e1 . Group homomorphisms map
identity elements to identity elements. So we have

h∗ (g) = k∗ (j∗ (g)) = k∗ (e1 ) = e2

So h∗ is the trivial homomorphism.

53
5 Lecture 5 54

(3)⇒(1). Let p : R → S 1 be the standard covering map given by p (x) = (cos 2πx, sin 2πx). Also,
let p0 be the restriction of p to I, i.e. p0 = p I .


h∗ : π1 S 1 , b0 → π1 (X, x0 ) with h (b0 ) = x0 . Since S 1 is path connected, π1 (S 1 , b0 ) is isomorphic
to π1 (S 1 , b1 ) for any b0 , b1 ∈ S 1 . So we can assume without loss of generality that b0 = (1, 0).


Claim — [p0 ] generates π1 S 1 , b0 .


Proof. We proved in Theorem 4.2.2
that π1 S 1 , b0 is isomorphic to Z. In context

of the proof of
the same theorem, ϕ : π1 S , b0 → Z is a group isomorphism. So π1 S , b0 is also an infinite
1 1

cyclic group generated by ϕ−1 (1). Now we need to show that ϕ−1 (1) = [p0 ], i.e. ϕ ([p0 ]) = 1.
p0 is a path in S 1 beginning at (1, 0), and p : R → S 1 is a covering map. So p0 has a unique
lifting pe0 : I → R such that beginning at 0, and pe0 (1) = ϕ ([p0 ]).

(cos 2πs, sin 2πs) = p0 (s) = p (pe0 (s)) = (cos 2π pe0 (s) , sin 2π pe0 (s))

This gives us
2π pe0 (s) = 2πs + 2πn , for some n ∈ Z =⇒ pe0 (s) = s + n


Since pe0 begins at 0, pe0 (s) = s. So ϕ ([p0 ]) = pe0 (1) = 1, and π1 S 1 , b0 is generated by [p0 ]. □

The inverse
 −1  element of [p0 ] is [p0 ], where p0
(s) = p0 (1 − s) is the reverse loop of p0 . We denote
1
[p0 ] = p0 . So all the elements of π1 S , b0 can be expressed as
   −2   −1   0     
. . . , p−3
0 , p0 , p0 , p0 , [p0 ] , p20 , p30 , . . .
 
where p00 is the identity element, the constant loop based at b0 . And

pn0 = p0 ∗ p0 ∗ · · · ∗ p0 and p−n −1 −1 −1

0 = p0 ∗ p0 ∗ · · · ∗ p0 , for n ∈ N
| {z } | {z }
n times n times

Let f = h ◦ p0 . h∗ is the trivial homomorphism, so [h ◦ p0 ] = h∗ ([p0 ]) = [ex0 ], where ex0 is the constant
loop based at x0 . So f 'p ex0 and let F : I × I → X be a path homotopy between them.
Consider the map p0 × id : I × I → S 1 × I defined by (p0 × id) (s, t) = (p0 (s) , t).

Claim — p0 × id is a quotient map.

Proof. (s, t) 7→ s 7→ p0 (s) is continuous, (s, t) 7→ t is continuous. Both coordinates of p0 × id are

continuous, so p0 × id is continuous.
p0 is a surjective map. So for any x ∈ S 1 , we can find s0 ∈ I such that p0 (s0 ) = x. Now for
any (x, t) ∈ S 1 × I, (p0 × id) (s0 , t) = (x, t), so p0 × id is surjective.
I × I is a bounded closed subset of R2 , so it’s compact by Heine-Borel Theorem. S 1 × I is a
subspace of R3 , thus hausdorff. p0 × id is a continuous map from a compact space to a hausdorff
space. Therefore, by Proposition 0.7.5, p0 × id is a closed map.
p0 × id is continuous, surjective, closed. By Lemma 0.10.1, it is a quotient map. □

p0 (0) = b0 = p0 (1), othewise p0 is injective on (0, 1). So p0 × id is injective on (0, 1) × I. Each point of
the form (b0 , t) has two preimages under p0 × id, namely (0, t) and (1, t). Since F is a path homotopy
between f and ex0 ,
F (0, t) = x0 = F (1, t)
So F is constant on the preimage of (b0 , t). For all other points (x, t), the preimage of (x, t) is a
singleton set. So F is trivially constant on the preimage.
Therefore, by Theorem 0.10.3, F induces a map H : S 1 × I → X such that H ◦ (p0 × id) = F . Since
F is continuous, by the same theorem, H is continuous.

54
5 Lecture 5 55

I ×I

p0 ×id F

S1 × I H
X

Since F is a path homotopy between f and ex0 ,

F (s, 0) = f (s) = h (p0 (s)) and F (s, 1) = ex0 (s) = x0 ∀s ∈ I

If x ∈ S 1 , there exists some s0 ∈ I such that p0 (s0 ) = x.

H (x, 0) = H (p0 (s0 ) , 0) = H ((p0 × id) (s0 , 0)) = F (s0 , 0) = h (p0 (s0 )) = h(x)

This is true for every x ∈ S 1 . On the other hand, let ef

x0 : S → X be the continuous map that maps
1

all of S 1 to x0 ∈ X.

H (x, 1) = H (p0 (s0 ) , 1) = H ((p0 × id) (s0 , 1)) = F (s0 , 1) = x0 = ef

x0 (x)

Therefore, H is a homotopy between h and ef

x0 . This proves that h is nulhomotopic.

We’ve proved that (1) ⇒ (2) ⇒ (3) ⇒ (1). So the three statements are equivalent. ■

Corollary 5.1.4
The inclusion map j : S 1 → R2 \ {0} is not nulhomotopic. Also, the identity map i : S 1 → S 1 is
not nulhomotopic.

Proof. There is a retraction of R2 \ {0} onto S 1 given by r (x) = ∥x∥

x
(check that this is continuous).



By Lemma 5.1.1, the induced homomorphism j∗ : π1 S , b0 → π1 R \ {0} , b0 is injective.
1 2

Assume for the sake of contradiction that j is nulhomotopic. Then by
Lemma 5.1.3, j∗ is a trivial
homomorphism. But we’ve just shown that j∗ is injective.
As π1 S 1 , b0 is nontrivial, the image of
j∗ can’t be just the identity element of π1 R \ {0} , b0 . So we arrive at a contradiction! Therefore,
2

j is not nulhomotopic.



Similarly, the induced homomorphism i∗ of the identity
map i : S 1 , b0 → S 1 , b0 is the identity
group homomorphism. i∗ can’t be trivial since π1 S 1 , b0 is nontrivial. Hence, using Lemma 5.1.3
again, i is not nulhomotopic. ■

Theorem 5.1.5
Given a nonvanishing vector field on B 2 , there exists a point of S 1 where the vector field points
directly inward, and a point of S 1 where it points directly outward.

Proof. A vector field on B 2 is an ordered pair (x, v(x)), where x ∈ B 2 and v : B 2 → R2 is a

continuous map. To say that a vector field is nonvanishing means that v(x) 6= 0 for every x ∈ B 2 .
In such a case, v actually maps B 2 to R2 \ {0}.
Assume for the sake of contradiction that v(x) doesn’t directly point inward at any point x ∈ S 1 .
Let w be the restriction of v on S 1 , i.e. w = v S 1 . In other words, w : S 1 → R2 \ {0} has a continuous
extension given by v : B 2 → R2 \ {0}. Then by Lemma 5.1.3, w is nulhomomtopic.
Now observe that w is homotopic to the inclusion map j : S 1 → R2 \{0}; the homotopy F : S 1 ×I →
R2 \ {0} is given by

F (x, t) = t j(x) + (1 − t) w(x) = tx + (1 − t) w(x) , for x ∈ S 1 , t ∈ I

55
5 Lecture 5 56

ty + (1 − t)w(y)
tx + (1 − t)w(x)
(y, j(y)) (y, w(y)) (x, j(x))

(x, w(x))

(z, w(z))

tz + (1 − t)w(z) (z, j(z))

But we are yet to show that F (x, t) ∈ R2 \ {0}. From the fact that w(x) 6= 0 and j(x) 6= 0, it
immediately follows that F (x, 1) 6= 0 and F (x, 0) 6= 0. If for some t ∈ (0, 1), F (x, t) = 0 then
−t
tx + (1 − t) w(x) = 0 =⇒ w(x) = x
1−t
This means w(x) is a negative scalar multiple of x. Geometrically this interprets as w(x) pointing
directly inwards, which is a contradiction! So F (x, t) ∈ R2 \ {0} for every x ∈ S 1 , t ∈ I.
So we have proved that F : S 1 × I → R2 \ {0} is a homotopy between w and j. In other words,
w ' j. We’ve also proved that w is nulhomotopic. Since ' is an equivalence relation, j : S 1 → R2 \{0}
is nulhomotopic. But this contradicts with Corollary 5.1.4. Hence, there must exist a point x ∈ S 1
such that v(x) points directly inward.
If we had started the proof with the vector field (x, −v(x)) and carried out the same arguments, we
would have reached the conclusion that −v(x) points directly inward for some x ∈ S 1 . In other words,
v(x) points directly outward for some x ∈ S 1 . ■

§5.2 Fixed Point Theorems

Lemma 5.2.1
Let f : X → X be continuous. If X = [0, 1] then there exists x ∈ X such that f (x) = x. This x
is called a fixed point of f .

Proof. Let g : [0, 1] → [−1, 1] be defined by g (x) = f (x) − x. Then g is continuous. f (x) ∈ [0, 1], so
f (0) ≥ 0. In other words,
g (0) = f (0) − 0 ≥ 0
Similarly, f (x) ∈ [0, 1], gives us f (1) ≤ 1. In other words,
g (1) = f (1) − 1 ≤ 0
Combining these two inequalities, we get g (1) ≤ 0 ≤ g (0). By Intermediate Value Theorem, there
exists y ∈ [0, 1] such that g (y) = 0. For that y, 0 = g (y) = f (y) − y. Hence f (y) = y. ■

56
5 Lecture 5 57

Theorem 5.2.2 (Brouwer fixed point theorem)

If f : B 2 → B 2 is continuous, then there exists x ∈ B 2 such that f (x) = x.

Proof. We shall proceed by contradiction. Suppose f (x) 6= x for every x ∈ B 2 . We define v (x) =
f (x) − x. Clearly v is continuous. So we obtain a nonvanishing vector field (x, v (x)) on B 2 .
Using Theorem 5.1.5, we get that there is a point x ∈ S 1 at which v (x) points directly outward. In
other words, v (x) = ax for some positive real number a.

f (x) − x = v (x) = ax =⇒ f (x) = (1 + a) x =⇒ kf (x)k = |1 + a| kxk = 1 + a > 1

But f is a map from B 2 to itself. So kf (x)k ≤ 1 for every x ∈ B 2 . Contradiction! Therefore, there
must exist some x ∈ B 2 such that f (x) = x. ■

Now we shall prove a linear algebra result using topological techniques.

Corollary 5.2.3
Let A be a 3 × 3 matrix with positive real entries. Then A has a positive real eigenvalue.

Proof. Let T : R3 → R3 be the linear transformation whose matrix representation (with reference to
the standard basis for R3 ) is A. Let F denote the first octant of R3 . In other words,

F = (x1 , x2 , x3 ) ∈ R3 : x1 ≥ 0 , x2 ≥ 0 , x3 ≥ 0

Let B be the intersection of F with S 2 , i.e. B = F ∩ S 2 . It’s easy to see that B is homeomorphic to
the unit ball B 2 (left as an exercise for the reader). Now we claim that Brouwer fixed point theorem
holds for continuous maps from B to itself.

Claim — Let f : B → B be continuous, then there exists y ∈ B such that f (y) = y.

Proof. Let g : B → B 2 be a homeomorphism. Then g ◦ f ◦ g −1 : B 2 → B 2 is a continuous map

from B 2 to itself. According to Brouwer fixed point theorem, it has a fixed point x ∈ B 2 .




g f g −1 (x) = x =⇒ f g −1 (x) = g −1 (x) =⇒ f (y) = y

where y = g −1 (x) ∈ B. □

Take x = (x1 , x2 , x3 ) ∈ B. Then all the components of x are nonnegative. Since kxk = 1, at least one
of the components must be positive. A is a 3 × 3 matrix with positive real entries. So T (x) = Ax has
all components positive. As a result, ∥TT (x)
(x)∥ ∈ B.
Consider that map f : B → B defined by f (x) = ∥TT (x)
(x)∥ . T is a linear map from a finite dimensional
vector space to itself, so it’s continuous. Furthermore, the map x 7→ ∥x∥ x
is also continuous. Therefore,
their composition f must also be continuous.
f is a continuous map from B to itself. By the claim stated above, there exists x0 ∈ B such that
f (x0 ) = x0 .
T (x0 )
= f (x0 ) = x0 =⇒ Ax0 = T (x0 ) = kT (x0 )k x0
kT (x0 )k
Therefore, x0 is an eigenvector of the linear transformation T , with eigenvalue kT (x0 )k, which is
necessarily positive. Thus we’ve proved that A has a positive real eigenvalue. ■

§5.3 Fundamental Theorem of Algebra

In this section we shall prove the Fundamental Theorem of Algebra (FTA) using topological
techniques.

57
5 Lecture 5 58

Theorem 5.3.1 (Fundamental theorem of algebra)

A polynomial equation xn + an−1 xn−1 + · · · + a1 x + a0 = 0 with complex coefficients has at least
one complex root.

Proof. Step 1. Consider thet map f : S 1 → S 1 given by f (z) = z n , where z is a complex number
with unit modulus. In this step we shall prove that the induced homomorphism f∗ of fundamental
groups is injective.
Let p0 : I → S 1 be the standard loop in S 1 . Here p I
= p0 , with the familiar covering map
p : R → S 1 given by p (x) = (cos 2πx, sin 2πx). So we have

p0 (s) = e2πis ≡ (cos 2πs, sin 2πs)

Let g = f ◦ p0 . Then f∗ ([p0 ]) = [g].

n
g (s) = f (p0 (s)) = f e2πis = e2πis = e2πins ≡ (cos 2πns, sin 2πns)

We’ve proved before that the unique lifting of p0 starting at 0 is given by pe0 (s) = s. g is also a loop
in S 1 based at b0 = (1, 0). Using Lemma 4.1.1, g has a unique lifting ge : I → R beginning at 0.

R
ge
p

[0, 1] g S1

g (s) = p (e
g (s)) =⇒ (cos 2πns, sin 2πns) = (cos 2πe
g (s) , sin 2πe
g (s))
g (s) = 2πns + 2πk , for some k ∈ Z
=⇒ 2πe
=⇒ ge (s) = ns + k


Since ge begins at 0, ge (s) = ns. If we consider the isomorphism ϕ : π1 S 1 , b0 → Z given in the proof
of Theorem 4.2.2, then
n = ge (1) = ϕ ([g]) = ϕ ([f ◦ p0 ]) = ϕ (f∗ ([p0 ]))
while ϕ ([p0 ]) = 1.

We’ve seen before that all elements of π1 S 1 , b0 can be expressed as [pm 0 ], where m ∈ Z. To show
that f∗ is injective, suppose f∗ ([px0 ]) = f∗ ([py0 ]) for x, y ∈ Z. We need to show that x = y.

f∗ ([px0 ]) = f∗ ([py0 ]) =⇒ f∗ ([p0 ])x = f∗ ([p0 ])y =⇒ ϕ (f∗ ([p0 ])x ) = ϕ (f∗ ([p0 ])y )
=⇒ x ϕ (f∗ ([p0 ])) = y ϕ (f∗ ([p0 ]))
=⇒ xn = yn =⇒ x = y

So f∗ is injective.

Step 2. In this step, we shall prove that if g : S 1 → R2 \ {0} is the map given by g (z) = z n , then
g is not nulhomotopic.
In Step 1, we had the map f : S 1 → S 1 defined by f (z) = z n . If j : S 1 → R2 \ {0} is the inclusion
map, then we have
j ◦ f = g =⇒ j∗ ◦ f∗ = g∗
We’ve shown in the proof of Corollary 5.1.4 that S 1 is a retract of R2 \ {0}. As a result, j∗ is injective
by Lemma 5.1.1. By step 1, f∗ is injective. As a composition

of two injective

maps, g∗ is injective.

We, therefore, have proved that g∗ : π1 S 1 , b0 → π1 R2 \ {0} , b0 is injective.

As π1 S 1 , b0 is
nontrivial, the image of g∗ can’t be just the identity element of π1 R \ {0} , b0 . So g∗ is not trivial.
2

Hence, by Lemma 5.1.3, g is not nulhomotopic.

58
5 Lecture 5 59

Step 3. In this step, we shall prove a special case of FTA. We assume that |an−1 | + |an−2 | + · · · +
|a1 | + |a0 | < 1. Then we shall prove that the polynomial equation xn + an−1 xn−1 + · · · + a1 x + a0 = 0
has a root lying in B 2 .
Assume the contrary. Then we can define k : B 2 → R2 \ {0} by

k (z) = z n + an−1 z n−1 + · · · + a1 z + a0 = 0

Let h be the restriction of k to S 1 , i.e. k S 1 = h. In other words, h : S 1 → R2 \ {0} extends to a

continuous map k : B 2 → R2 \ {0}. By Lemma 5.1.3, h is nulhomotopic.
We shall now define a homotopy F between h and g (as defined in step 2). Consider F : S 1 × I →
R \ {0} defined by
2


F (z, t) = z n + t an−1 z n−1 + · · · + a1 z + a0
However, we are yet to show that F (z, t) 6= 0 for any z ∈ S 1 , t ∈ I. For z ∈ S 1 , |z| = 1. Using triangle
inequality,



t an−1 z n−1 + · · · + a1 z + a0 ≤ t an−1 z n−1 + |a1 z| + |a0 |
= t (|an−1 | + · · · + |a1 | + |a0 |)


1 − t an−1 z n−1 + · · · + a1 z + a0 ≥ 1 − t (|an−1 | + · · · + |a1 | + |a0 |)
≥ 1 − (|an−1 | + · · · + |a1 | + |a0 |) > 0


z n = F (z, t) − t an−1 z n−1 + · · · + a1 z + a0


|z n | = F (z, t) − t an−1 z n−1 + · · · + a1 z + a0


≤ |F (z, t)| + t an−1 z n−1 + · · · + a1 z + a0


|F (z, t)| ≥ |z n | − t an−1 z n−1 + · · · + a1 z + a0


= 1 − t an−1 z n−1 + · · · + a1 z + a0 > 0
So F (z, t) 6= 0, F is a map from S 1 × I to R2 \ {0}. So F is indeed a homotopy between h and g,
i.e. h ' g. We’ve seen that h is nulhomotopic. By the equivalence of ', g is nulhomotopic. But we
proved in step 2 that g is not nulhomotopic. Contradiction! So the polynomial equation must have a
root in B 2 .

Step 4. In this step we shall prove the general result.

Consider the polynomial equation xn + an−1 xn−1 + · · · + a1 x + a0 = 0. We choose c ∈ R with c > 0,
and substitute x = cy.

(cy)n + an−1 (cy)n−1 + · · · + a1 (cy) + a0 = 0

=⇒ cn y n + cn−1 an−1 y n−1 + · · · + ca1 y + a0 = 0
an−1 n−1 a1 a0
=⇒ y n + y + · · · + n−1 y + n = 0
c c c
If this equation can be shown to admit a root y0 , then the original equation is easily seen to have a
root x0 = cy0 . Now we choose large enough c such that
an−1 an−2 a1 a0
+ 2
+ · · · + n−1 + n < 1
c c c c
Then using step 3, y n + an−1
c y
n−1 + · · · + a1 y + a0 = 0 has a root y ∈ B 2 . Therefore, xn + a
cn−1 cn 0 n−1 x
n−1 +

· · · + a1 x + a0 = 0 has a root x0 = cy0 . ■

59
6 Lecture 6
§6.1 Simplices

Definition 6.1.1 (Geometric independence). Given a set {a0 , a1 , a2 , . . . , an } of points of RN , i.e.

each ai is an N -tuple, this set is said to be geometrically independent if for any scalars ti ∈ R,
the equations
Xn Xn
ti = 0 and ti ai = 0
i=0 i=0

imply that t0 = t1 = · · · = tn = 0.

It is immediate that a one-point set {a0 } is always geometrically independent.

Lemma 6.1.1
{a0 , a1 , a2 , . . . , an } is geometrically independent if and only if the vectors

a1 − a0 , a2 − a0 , . . . , an − a0 are linearly independent.

Proof. (⇒) Suppose {a0 , a1 , a2 , . . . , an } is geometrically independent. Consider ti ∈ R for i =

1, 2, . . . , n such that
Xn
ti (ai − a0 ) = 0
i=1
P
n
Take t0 ∈ R such that ti = 0. Now we have
i=0

X
n X
n X
n X
n
t i ai = ti ai + t0 a0 = ti (ai − a0 ) + ti a0 + t 0 a0
i=0 i=1 i=1 i=1
Xn X
n
= ti (ai − a0 ) + a0 ti = 0
i=1 i=0

P
n Pn
Since {a0 , a1 , a2 , . . . , an } is geometrically independent, ti = 0 and i=0 ti ai together gives us ti = 0
i=0
for every i = 0, 1, 2, . . . , n. Therefore, the vectors a1 − a0 , a2 − a0 , . . . , an − a0 are linearly independent.
(⇐) Conversely, suppose the vectors a1 − a0 , a2 − a0 , . . . , an − a0 are linearly independent. Consider
ti ∈ R for i = 0, 1, 2, . . . , n such that
X
n X
n
ti = 0 and ti ai = 0
i=0 i=0

Using these, we get

X
n X
n X
n X
n X
n
0= ti ai = ti (ai − a0 ) + a0 ti = ti (ai − a0 ) =⇒ ti (ai − a0 ) = 0
i=0 i=1 i=0 i=1 i=1

Since a1 − a0 , a2 − a0 , . . . , an − a0 are linearly independent, all the ti ’s must be 0, for i = 1, 2, . . . , n. As

Pn
ti = 0, t0 is also 0. Therefore, t0 = t1 = · · · = tn = 0, and thus {a0 , a1 , a2 , . . . , an } is geometrically
i=0
independent. ■

60
6 Lecture 6 61

Definition 6.1.2 (Simplex). Let {a0 , a1 , . . . , an } be a geometrically independent set in RN . We

define the n-simplex σ spanned by a0 , a1 , . . . , an to be the set of all points x of RN such that

X
n X
n
x= ti ai where ti = 1 and ti ≥ 0 ∀ i
i=0 i=0

The numbers ti are uniquely determined by x; they arre called the barycentric coordinates of
the point x of σ with respect to a0 , a1 , . . . , an .

Example 6.1.1 (Simplices in low dimensions)

A 0-simplex is a point. The 1-simplex spanned by a0 and a1 consists of all points x of the form

x = ta0 + (1 − t) a1 where t ∈ [0, 1]

This is just the line segment joining a0 and a1 .

The 2-simplex spanned by a0 , a1 , a2 equals the triangle (geometric independece guarantees non-
collinearity of the points) having these three points as vertices. This can be seen in the following
way:

X
2    
t1 t2 t1 t2
x= ti ai = t0 a0 + (1 − t0 ) a1 + a2 = t0 a0 + (1 − t0 ) a1 + a2
1 − t0 1 − t0 λ λ
i=0

where λ = 1 − t0 . Note that, t0 + t1 + t2 = 1 gives us t1

λ + t2
λ = t1 +t2
1−t0 = 1. So p is a point on the
line joining a1 and a1 given by
t1 t2 t1 t2 ti
p= a1 + a2 because + = 1 and ≥0
λ λ λ λ λ
Any point lying on the line segment joining a0 and p is given by
 
t1 t2
t0 a0 + (1 − t0 ) p = t0 a0 + (1 − t0 ) a1 + a2
λ λ

Therefore, x represents a point in the triangular region formed by joining the points a0 , a1 , a2 .
See the figure below:
a0

x
a2
p
a1

Definition 6.1.3. A subset A ⊆ RN is said to be convex if for each pair x, y of points of A, the
line segment joining them lies in A.

Definition 6.1.4. The points a0 , a1 , . . . , an that span a are called the vertices of σ; the number
n is called the dimension of σ. Any simplex spanned by a subset of {a0 , a1 , . . . , an } is called
a face of σ. In particular, the face spanned by {a1 , . . . , an } is called the face opposite to a0 .
The faces of σ different from σ itself are called the proper faces of σ; their union is called the

61
6 Lecture 6 62

boundary of σ and denoted Bd σ. The interior of σ is defined by the equation Int σ = σ \ Bd σ.

The set Int σ is sometimes called an open simplex.

Let us list some basic properties of simplices. Throughout, let P be the n-plane determined by the
points of the geometrically independent set {a0 , a1 , . . . , an }; and let σ be the n-simplex spanned by
these points. If x ∈ σ, let {ti (x)} be the barycentric coordinates of x; they are determined uniquely
by the conditions
Xn Xn
x= ti ai and ti = 1
i=0 i=0

Then the following properties hold:

1. The barycentric coordinates ti (x) with respect to a0 , a1 , . . . , an are continuous functions of x.

2. σ equals the union of all line segments joining a0 to points of the simplex s spanned by a1 , . . . , an .
Two such line segments intersect only in the point a0 .

3. σ is a compact, convex set in RN , which equals the intersection of all convex sets in RN containing
a0 , a1 , . . . , an .

4. Given a simplex σ, there is one and only one geometrically independent set of points spanning
σ.

5. Int σ is convex and is open in the plane P ; its closure is σ. Furthermore, Int σ equals the union
of all open line segments joining a0 to points of Int s, where s is the face of σ opposite to a0 .

6. There is a homeomorphism of σ with the unit ball B n that carries Bd σ onto the unit sphere
S n−1 .

§6.2 Simplicial Complexes

Definition 6.2.1. A simplicial complex K in RN is a collection of simplices in RN such that:

i. Every face of a simplex of K is in K.

ii. The intersection of any two simplices of K is a face of each of them.

NOT
Simplicial Simplicial Simplicial
Simplicial
Complex Complex Complex
Complex

Lemma 6.2.1
A collection K of simplices is a simplicial complex if and only if the following hold:

i. Every face of a simplex of K is in K.

ii. Every pair of distinct simplices of K have disjoint interiors.

62
6 Lecture 6 63

Definition 6.2.2. If L is a subcollection of K that contains all faces of its elements, then L is a
simplicial complex in its own right; it is called a subcomplex of K. One subcomplex of K is the
collection of all simplices of K of dimension at most p; it is called the p-skeleton of K and is
denoted K (p) . The points of the collection K (0) are called the vertices of K.

Definition 6.2.3. Let |K| be the subset of RN that is the union of the simplices of K. Giving each
simplex its natural topology as a subspace of RN , we then topologize |K| by declaring a subset A
of |K| to be closed in |K| if and only if A ∩ σ is closed in σ, for each σ in K. The space |K| is
called the underlying space of K, or the polytope of K.

In general, the topology of |K| is finer (more open sets) than the topology |K| inherits as a subspace
of RN . If A is closed in the subspace topology, then according to the definition of subspace topology,
A = B ∩ |K| for some closed set B in RN . Since B is closed in RN , B ∩ σ is closed in σ in the subspace
topology σ inherits from RN .
Therefore, for each σ in K, B ∩ σ is closed in σ. σ is a subset of |K|, so |K| ∩ σ = σ.

∴ B ∩ σ = B ∩ (|K| ∩ σ) = (B ∩ |K|) ∩ σ = A ∩ σ

As a result, A ∩ σ is closed in σ for every σ in K. Hence, by the definition of topology on |K|, A is

closed in |K|.
For each closed set A in the subspace topology on |K|, A is also closed in |K| with respect to the
topology defined earlier. So, the subspace topology is contained in the topology of |K|.
The two topologies on |K| are, in general, different as we will see using some examples. In fact, if
K is finite, then they are the same. We will prove the containment in the other direction for the case
of K being finite.
Suppose K is finite and A is closed in |K|. Then A ∩ σi is closed in σi for each i ∈ {1, 2, . . . , n} with
Sn
|K| = σi .
i=1
Now, A ∩ σi is closed in σi for every i, and σi is closed in RN . Hence, by Lemma 0.2.2 (the lemma
is true if you replace “open” by “closed”), each A ∩ σi is closed in RN .

[
n [
n
A⊆K= σi =⇒ A = (A ∩ σi )
i=1 i=1

Finite union of closed sets is closed. Since each A ∩ σi is closed in RN , A is also closed in RN . As
A ⊆ |K|, A = A ∩ |K|. Therefore, A is closed in |K| in the subspace topology inherited by |K| from
RN .

Example 6.2.1
Let K be the collection of all 1-simplices
h ini R of the form [m, m + 1], where m ∈ Z \ {0}; along
with all the 1-simplices of the form n+1 , n1 where n ∈ N.
1

We claim that the set X = n1 : n ∈ N is closed in |K|. X ∩ [m, m + 1] is {1} when m = 1,
and ∅ when m 6= 1. Empty set is obviously closed. R is hausdorff, so is its subspace [1, 2].
Singleton sets are closed in hausdorffh space i (Proposition
h 0.4.1). n {1} isoclosed
i So h in [1,
i 2].
If we take a simplex of the form n+1 , n , then X ∩ n+1 , n = n+1 , n . n+1 , n is hausdorff
1 1 1 1 1 1 1 1
n o  h i
as a subspace of R, so both n+1 1
and n1 are closed in n+1 1
, n1 . Therefore, their union
n o h i
1 1 1 1
,
n+1 n is closed in n+1 n .
,
For each simplex σ, X ∩ σ is closed in σ. Therefore, X is closed in |K|.
But X has an accumulation point 0. But 0 6∈ X, so X is not closed in R. Therefore, X is
not closed in the subspace topology of |K|. Hence, the topology of |K| is strictly finer than the

63
6 Lecture 6 64

subspace topology.

Example 6.2.2
Let K be the
collection of simplices σ1 , σ2 , . . ., where σi is the 1-simplex in R2 having vertices
(0, 0) and 1, 1i . Then K is a simplicial complex.

(1, 1)

( 1)
1,
( 21 )
1, 3



We take X = |K| ∩ x, x2 : x > 0 . The  1-simplex

σi is given by y = 1i x for
x ∈ [0, 1].
Its
 1intersection

with the open parabolic arc x, x 2 : x > 0 is the singleton 1i , i12 . So X =
1
i , i2 : i∈N .
X ∩ σi is singleton for every i. σi is a subspace of Hausdorff space R2 , so singleton sets are
closed. Therefore, X ∩ σi is closed in σi for every i. Hence, X is closed in the topology on |K|.
X has an accumulation point (0, 0). But (0, 0) 6∈ X, so X is not closed in R2 . Therefore, X is
not closed in the subspace topology of |K|. Hence, the topology of |K| is strictly finer than the
subspace topology.

§6.3 Abstract Simplicial Complex

Definition 6.3.1 (Abstract Simplicial Complex). An abstract simplicial complex is a collection

S of finite nonempty sets, such that if A is an element of S, so is every nonempty subset of A.

Definition 6.3.2. Suppose S is an abstract simplicial complex. The element A of S is called a

simplex of S; its dimension is one less than the number of its elements. Each nonempty subset
of A is called a face of A.
The dimension of S is the largest dimension of one of its simplices, or is infinite if there is no
such largest dimension. The vertex set V of S is the union of the one-point elements of S. An
element v ∈ V is called a vertex and it will be considered the same as the 0-simplex {v} ∈ S. A
subcollection of S that is itself a complex is called a subcomplex of S.

Definition 6.3.3 (Isomorphism). Two abstract complexes S and T are said to be isomorphic if
there is a bijective correspondence f mapping the vertex set of S to the vertex set of T such that

{a0 , a1 , . . . , an } ∈ S ⇐⇒ {f (a0 ) , f (a1 ) , . . . , f (an )} ∈ T

Example 6.3.1
Consider the following collection of finite nonempty sets

S = {a0 , a1 , a2 } , {a0 , a1 } , {a1 , a2 } , {a0 , a2 } , {a2 , a3 } , {a0 } , {a1 } , {a2 } , {a3 } , {a4 }

Then S is an abstract simplicial complex. It can be visualized as

64
6 Lecture 6 65

a1

a0 a2

a4 a3

V = {a0 , a1 , a2 , a3 , a4 } and V 6∈ S. In general, the vertex set is not a simplex. The nonempty
set {a0 , a1 , a2 } is a simplex of dimension 3 − 1 = 2. The dimension of S is also 2. Here {a0 , a1 , a2 }
is a subcomplex of S.

Definition 6.3.4 (Vertex Scheme). If K is a geometric simplicial complex, let V be the vertex set of
K. Let K be the collection of all subsets {a0 , a1 , . . . , an } of K such that the vertices a0 , a1 , . . . , an
span a simplex of K. The collection K is called the vertex scheme of K.

The collection K is an example of an abstract simplicial complex.

Remark. Whenever we refer to a simplicial complex, we mean a geometric simplicial complex, which
is a collection of simplices in RN . We can take the vertex set V of the geometric simplicial complex,
and we can construct an abstract simplicial complex K out of it, which we call the vertex scheme.

Theorem 6.3.1
Every abstract simplicial complex S is isomorphic to the vertex scheme of some geometric sim-
plicial complex.

Lemma 6.3.2
Let K and L be simplicial complexes, and let f : K (0) → L(0) be a map. Suppose that whenever
the vertices v0 , v1 , . . . , vn of K span a simplex of K, the points f (v0 ) , f (v1 ) , . . . , f (vn ) are
vertices of a simplex of L. Then f can be extended to a continuous map g : |K| → |L| such that

X
n X
n
x= ti vi =⇒ g (x) = ti f (vi )
i=0 i=0

We call g the simplicial map induced by the vertex map f .

Lemma 6.3.3
Let K and L be simplicial complexed. Suppose f : K (0) → L(0) is a bijective correspondence such
that the vertices v0 , v1 , . . . , vn of K span a simplex of K if and only if f (v0 ) , f (v1 ) , . . . , f (vn )
span a simplex of L. Then the induced simplicial map g : |K| → |L| is a homeomorphism.

Definition 6.3.5 (Geometric Realization). It is guaranteed by Theorem 6.3.1 that every abstract
simplicial complex S is isomorphic to the vertex scheme of some simplicial complex K. We call
K a geometric realization of S.

Let’s see an example that illustrates these concepts. Suppose we wish to identify a simplicial complex
K whose underlying space |K| is homeomorphic to the cylinder S 1 × I. Let us draw the simplicial
complex K consisting of six 2-simplices, twelve 1-simplices and six 0-simplices.

65
6 Lecture 6 66

e d f d e f
f

c b
a b c a
a
K L

The abstract simplicial complex S has the vertex set V = {a, b, c, d, e, f } while

S = {a, f, d} , {a, b, d} , {b, c, d} , {c, d, e} , {a, c, e} , {a, e, f } ,
{a, f } , {f, d} , {a, d} , {a, b} , {b, d} , {b, c} ,
{c, d} , {d, e} , {c, e} , {a, c} , {a, e} , {e, f } ,
{a} , {b} , {c} , {d} , {e} , {f }

Of course this abstract simplicial complex S is isomorphic to the vertex scheme of the simplicial
complex K pictured above.
Let f : L(0) → K (0) be the map that assigns each vertex of L to the correspondingly labeled vertex
of K. It is immediate that f is surjective. Then f extends to a simplicial map g : |L| → |K|. This
map is continuous surjective. Since |L| , |K| are compact hausdorff, the map g is a quotient map
(Proposition 0.10.6). This quotient map g is also called “pasting map”. It identifies the right edge of
|L| linearly with the left edge of |L|.

§6.4 Abelian Group Essentials

Here we shall write abelian groups additively. Then 0 denotes the neutral element, and −g denotes
the inverse of g. If n is a positive integer, then ng denotes the n-fold sum g + · · · + g, and (−n)g
denotes n(−g).

Definition 6.4.1 (Free Abelian Group). An abelian group G is free if it has a basis – that is, if
there is a family {gα }α∈J of elements of G such that each g ∈ G can be written uniquely as a
finite sum X
g= nα g α , (there are finitely many summands here)
α

Uniqueness implies that each element gα has infinite order; i.e. gα generates an infinite cyclic subgroup
of G. Suppose the contrary, i.e. ∃ n ∈ N such that gα has order n. This means ngα = 0. But then

gα = 1gα and gα = 0 + gα = ngα + gα = (n + 1)gα

Since n ∈ N, n + 1 = 6 1, so the representation of gα is not unique. Contradiction! Hence, there is no

n ∈ N for which gα is annihilated, meaning that gα has infinite order.
P
More generally, if each g ∈ G can be written as the finite sum α nα gα , but not necessarily uniquely,
then we say that the family {gα }α∈J generates G. In particular, if {gα }α∈J is finite, we say that G
is finitely generated.
If G is free and has a basis consisting of n elements, say g1 , g2 , . . . , gn , then every other basis for G
consists of precisely n elements.
Now 2G = {2g : g ∈ G} is easily seen to be a subgroup of G. An arbitrary element 2g ∈ 2G can
be expressed uniquely as
Xn
2g = 2 mi gi , with mi ∈ Z
i=1

66
6 Lecture 6 67

Now, the group G/2G consists of cosets, the elements of which are of the form
X
n
(2mi + εi ) gi , with mi ∈ Z and εi ∈ {0, 1}
i=1

If all the εi ’s are 0, then the element belongs to the subgroup 2G. In other words, when all the εi ’s are
0, one gets the representative of the coset 0 + 2G. Since εi can be either 0 or 1, and i ∈ {1, 2, . . . , n},
there are 2n distinct cosets. Hence |G/2G| = 2n .
Choosing a different basis, with m elements g1′ , g2′ , . . . , gm
′ , we get that |G/2G| = 2m . Therefore,
n m
2 = 2 , so m = n.

Definition 6.4.2 (Rank). The number of elements in a basis for G is called the rank of G.

There is a specific way of constructing free abelian groups. Given a set S, we define the free abelian
group G generated by S to be the set of all functions ϕ : S → Z such that

ϕ (x) = 0 for all but finitely many values of x ∈ S

The addition of two such functions ϕ and ψ is defined as

(ϕ + ψ) (x) = ϕ (x) + ψ (x)

Given x ∈ S, the characteristic function ϕx for x is defined as

(
0 6 x
if y =
ϕx (y) =
1 if y = x

The functions {ϕx : x ∈ S} form a basis for G. Because, each ϕ ∈ G can be written as
X
ϕ= nx ϕ x , where nx = ϕ (x)
x∈S ,ϕ(x)̸=0

Since ϕ (x) 6= 0 for only finitely many x, the sum is a finite sum.

Definition 6.4.3 (Torsion Subgroup). If G is an abelian group, an element g ∈ G has finite order
if ng = 0 for some n ∈ N. The set of all elements of finite order in G is a subgroup T of G, called
the torsion subgroup. If T vanishes, we say G is torsion-free.

We’ve seen earlier that in a free abelian group, no nonzero element has finite order. In other words, a
free abelian group is necessarily torsion-free. But the converse is not true in general.

Definition 6.4.4 (Direct Sum). Suppose G is an abelian group and suppose {Gα }α∈J isPa collection
of subgroups of G. Suppose each g ∈ G can be written uniquely as a finite sum g = gα , where
gα ∈ Gα for every α ∈ J. Then G is called the direct sum of the groups Gα . It is written as
M
G= Gα
α∈J

If the collection {Gα } is finite, say {Gα } = {G1 , G2 , . . . , Gn }, we write the direct sums as

G = G 1 ⊕ G2 ⊕ · · · ⊕ Gn
P
More generally, if each g ∈ G can be written as a finite sum g = gα , where gα ∈ Gα for every α ∈ J.
But the finite sum expression is not necessarily unique. Then G is called the sum of the groups Gα .
It is written as X
G= Gα
α∈J

67
6 Lecture 6 68

If the collection {Gα } is finite, say {Gα } = {G1 , G2 , . . . , Gn }, we write the direct sums as

G = G1 + G2 + · · · + G n
P
If G = α∈J Gα , then this sum will be direct if and only if for every fixed index α0 ∈ J,
 
X
Gα0 ∩  Gα  = {0}
α∈J,α̸=α0

If G is free, then it has a basis, say {gα }α∈J . Then all these basis elements are of infinite order. The
subgroup Gα that gα generates is infinite cyclic, and G is the direct sum of these subgroups {Gα }α∈J .
Conversely, if G is a direct sum of infinite cyclic groups, then G is a free abelian group.

Theorem 6.4.1 (Fundamental Theorem of Finitely Generated Abelian Groups)

Let G be a finitely generated abelian group. Let T be its torsion subgroup.
(a) There is a free abelian subgroup H of G having finite rank β such that G = H ⊕ T .
(b) There are finite cyclic groups T1 , T2 , . . . , Tk where Ti has order ti > 1, such that ti |ti+1 , i.e.
ti+1 is divisible by ti for each i ∈ {1, 2, . . . , k − 1} and

T = T1 ⊕ T2 ⊕ · · · ⊕ Tk

(c) The numbers β and t1 , t2 , . . . , tk are uniquely determined by G.

The number β is called the Betti number of G; the numbers t1 , t2 , . . . , tk are called the torsion
coefficients of G.

The group Z/mZ is often written as Zm . It’s a well-known result from group theory that if m and n
are coprime positive integers, then
Zm ⊕ Zn ∼ = Zmn
Using this, one can refine the statement of Fundamental Theorem of Finitely Generated Abelian
Groups. H is a free abelian group having rank β, so

H∼
= Z ⊕ Z ⊕ ··· ⊕ Z

which is the β-fold direct sum of Z. Also, each Ti is finite cyclic group of order ti . So Ti ∼
= Zti .

T = T1 ⊕ T2 ⊕ · · · ⊕ Tk ∼
= Zt1 ⊕ Z t2 ⊕ · · · Z tk

Each ti can be prime power factorized, thus we get

T ∼
= Z a1 ⊕ Z a2 ⊕ · · · Z as

where each ai is power of some prime. Combining the results,

G=H ⊕T ∼
= (Z ⊕ Z ⊕ · · · ⊕ Z) ⊕ (Za1 ⊕ Za2 ⊕ · · · Zas )

These ai ’s are called invariant factors of G.

68
7 Lecture 7
§7.1 Notion of Orientation

Definition 7.1.1. Let σ be a simplex. Define two orderings of its vertex set to be equivalent if
they differ from one another by an even permutation. If dim σ > 0, the orderings of the vertices
can be classified in two ways; in other words, there are two equivalence classes. Each of these
classes is called an orientation of σ. (If σ is a 0-simplex, then there is only one class and hence
only one orientation of σ.) An oriented simplex is a simplex σ together with an orientation of
σ.

If the points v0 , v1 , . . . , vp are geometrically independent, we shall use the symbol v0 v1 . . . vp (without
comma) to denote the simplex they span; and we use the symbol

[v0 , v1 , . . . , vp ]

to denote the oriented simplex consisting of the simplex v0 v1 . . . vp with the given ordering (v0 , v1 , . . . , vp )
.

Example 7.1.1
We attach an arrow to a given 1-simplex to prescribe an orientation. The oriented 1-simplex
[v0 , v1 ] is pictured with the arrow directed from v0 to v1 .

v0 v1

An orientation of a 2-simplex is depicted by a circular arrow. The oriented 2-simplex [v0 , v1 , v2 ]

is pictured as
v2

v0 v1

The 2-simplex v0 v1 v2 can be given a clockwise orientation to obtain the oriented 2-simplex
[v0 , v2 , v1 ].

v2

v0 v1

[v0 , v2 , v1 ], [v1 , v0 , v2 ] and [v2 , v1 , v0 ] all represent the same oriented 2-simplex; because they
differ from each other by an even permutation.

Example 7.1.2
The oriented 3-simplex [v0 , v1 , v2 , v3 ] is drawn by attching a spiral arrow to the simplex v0 v1 v2 v3 .

69
7 Lecture 7 70

v3

v2
v0

v1

The orientation thus prescribed obeys the “right hand screw” rule. If one curls their right hand
fingers in the direction from v0 to v1 to v2 , then the thumb will be pointed towards the direction
of v3 .
It can also be verified that the oriented 3-simplex [v0 , v2 , v3 , v1 ] obeys the right hand screw rule.
Indeed, if one curls their right hand fingers in the direction from v0 to v2 to v3 , then the thumb
will be pointed towards the direction of v1 .
The ordering of the vertex set of the oriented 3-simplex [v0 , v1 , v2 , v3 ] differs from that of
[v0 , v2 , v3 , v1 ] by an even permutation. Indeed, by permuting the vertices of [v0 , v2 , v3 , v1 ] twice
(v2 ↔ v3 followed by v3 ↔ v1 ) one arrives at [v0 , v1 , v2 , v3 ].
Consider the 3-simplex [v0 , v2 , v1 , v3 ] which differs from [v0 , v1 , v2 , v3 ] by a single permutation,
i.e. an odd permutation. Hence, [v0 , v2 , v1 , v3 ] is oppositely oriented. It is represeted with a
downwards spiral arrow:
v3

v2
v0

v1

Definition 7.1.2 (p-chain). Let K be a simplicial complex. A p-chain on K is a function c from

the set of oriented p-simplices of K to the integers, such that:
1. c (σ) = −c (σ ′ ), if σ and σ ′ are opposite orientations of the same simplex.
2. c (σ) = 0 for all but finitely many oriented p-simplices σ.

We add two p-chains c and c′ pointwise

c + c′ (σ) := c (σ) + c′ (σ) , where σ is an oriented p-simplex of K

With this addition operation between p-chains, the set of all p-chains on the simpicial complex K,
becomes a group. We call this the group of oriented p-chains of K, and denote it by Cp (K).
If p < 0 or p > dim K, we let Cp (K) denote the trivial group, the group consisting of the identity
element only.

Definition 7.1.3 (Elementary chain). If σ is a oriented simplex, then the elementary chain c
corresponding to σ is the function defined as:
1. c (σ) = 1.
2. c (σ ′ ) = −1; where σ ′ is the same simplex as σ, but with opposite orientation.

70
7 Lecture 7 71

3. c (τ ) = 0 for every other oriented simplex τ .

Abuse of Notation. We often use the symbol σ to denote not only a simplex, or an oriented
simplex, but also to denote the elementary p-chain c corresponding to the oriented simplex σ. With
this convention, if σ and σ are opposite orientations of the same simplex, then we can write σ ′ = −σ
instead of writing c (σ) = −c (σ ′ ).

Lemma 7.1.1
Cp (K) is free abelian; a basis for Cp (K) can be obtained by orienting each p-simplex and using
the corresponding elementary chains as a basis.

Proof. The proof is straightforward. First, orient the p simplices of K arbitrarily. Then each p-chain
on K can be written as a finite linear combination
X
c= ni σ i

of the corresponding elementary chains σi . The chain c takes the integer value ni on the oriented
p-simplex σi , the value −ni on the same p-simplex as σi but with opposite orientation, and 0 on all
oriented p-simplices not appearing in the summation. ■

Note that, the free abelian group C0 (K) has a natural basis, since a 0-simplex has only one orientation.
But Cp (K) has no “natural” basis if p > 0. Because one must orient the p-simplices arbitrarily in
order to obtain a basis.

Corollary 7.1.2
Any function f from the oriented p-simplices of K to an abelian group G extends uniquely to a
homomorphism Cp (K) → G provided that f (−σ) = −f (σ) for all oriented p-simplices σ.

Corollary 7.1.2 is a consequence of Lemma 7.1.1 and the universal property of free abelian groups,
which states that:

Theorem 7.1.3 (Universal Property of Free Abelian Groups)

If S is a set and Z(S) is the free abelian group on S, and given an abelian group G, f : S → G is
a mapping; then there exists a unique group homomorphism g : Z(S) → G with g ◦ ι = f . In
other words, the following diagram commutes:

f
S G

ι
∃! g

Z(S)

In Theorem 7.1.3, Z(S) refers to the group of all functions ϕ : S → Z such that ϕ (s) = 0 for all but
finitely many s ∈ S, and ι is the map that maps x ∈ S to ϕx ∈ Z(S) .
In the present context,
P S is the set of all oriented p-simplices of K and Z(S) is the free abelian group
Cp (K). Given c = ni σi ∈ Cp (K), define g : Cp (K) → G by
X  X
g ni σ i = ni f (σi )

Here ni ∈ Z for each i, and f (σi ) ∈ G. This g is our desired unique homomorphism Cp (K) → G, that
is an extension of f .

71
7 Lecture 7 72

§7.2 Homology Groups

Definition 7.2.1. We define a homomorphism

∂p : Cp (K) → Cp−1 (K)

called the boundary operator. If σ = [v0 , v1 , . . . , vp ] is an oriented p-simplex (p > 0), we define

X
p
∂p σ = ∂p [v0 , v1 , . . . , vp ] = (−1)i [v0 , . . . , vbi , . . . , vp ]
i=0

vbi means that the vertex vi is removed from the given array. Since Cp (K) is the trivial group for
p < 0, the operator ∂p is the trivial group homomorphism for p ≤ 0.

We need to check the well-definedness of ∂p and ∂p (−σ) = −∂p (σ). It suffices to check that in case of
exchange of two adjacent vertices of [v0 , v1 , . . . , vp ], ∂p changes sign. In other words, we need to show
that
∂p [v0 , . . . , vj , vj+1 , . . . , vp ] = −∂p [v0 , . . . , vj+1 , vj , . . . , vp ]
For i 6= j, j +1, the i-th terms in these two expressions differ precisely by a sign; the terms are identical
except that vj and vj+1 , have been interchanged. Now we need to consider the terms corresponding
to i = j and i = j + 1 separately.
For the first case ∂p [v0 , . . . , vj , vj+1 , . . . , vp ], the i = j and i = j + 1 terms are:

(−1)j [v0 , . . . , vj−1 , vbj , vj+1 , . . . , vp ] + (−1)j+1 [v0 , . . . , vj−1 , vj , vd

j+1 , vj+2 , . . . , vp ]

On the other hand, the i = j and i = j + 1 terms for ∂p [v0 , . . . , vj , vj+1 , . . . , vp ] are:

(−1)j [v0 , . . . , vj−1 , vd

j+1 , vj , . . . , vp ] + (−1)
j+1
[v0 , . . . , vj−1 , vj+1 , vbj , vj+2 , . . . , vp ]

It’s straightforward to see that

− (−1)j [v0 , . . . , vj−1 , vbj , vj+1 , . . . , vp ] = (−1)j+1 [v0 , . . . , vj−1 , vj+1 , vbj , vj+2 , . . . , vp ] and

− (−1)j [v0 , . . . , vj−1 , vd

j+1 , vj , . . . , vp ] = (−1)
j+1
[v0 , . . . , vj−1 , vj , vd
j+1 , vj+2 , . . . , vp ]

So ∂p [v0 , . . . , vj , vj+1 , . . . , vp ] and ∂p [v0 , . . . , vj+1 , vj , . . . , vp ] have opposite signs.

Example 7.2.1
For an oriented 1-simplex [v0 , v1 ],
∂1 [v0 , v1 ] = v1 − v0

∂1 =
v0 v1 v0 v1

For an oriented 2-simplex [v0 , v1 , v2 ],

∂2 [v0 , v1 , v2 ] = [v1 , v2 ] − [v0 , v2 ] + [v0 , v1 ]

v2
v2

=
∂2

v1
v1 v0
v0

72
7 Lecture 7 73

For an oriented 3-simplex [v0 , v1 , v2 , v3 ],

∂3 [v0 , v1 , v2 , v3 ] = [v1 , v2 , v3 ] − [v0 , v2 , v3 ] + [v0 , v1 , v3 ] − [v0 , v1 , v2 ]

v3
v3

∂3 =

v2
v2 v0
v0
v1
v1

If we apply the ∂1 operator on ∂2 [v0 , v1 , v2 ], we get

∂1 ∂2 [v0 , v1 , v2 ] = ∂1 [v1 , v2 ] − ∂1 [v0 , v2 ] + ∂1 [v0 , v1 ]

= (v2 − v1 ) − (v2 − v0 ) + (v1 − v0 ) = 0

If we apply the ∂2 operator on ∂3 [v0 , v1 , v2 v3 ], we get

∂2 ∂3 [v0 , v1 , v2 , v3 ] = ∂2 [v1 , v2 , v3 ] − ∂2 [v0 , v2 , v3 ] + ∂2 [v0 , v1 , v3 ] − ∂2 [v0 , v1 , v2 ]

= [v2 , v3 ] − [v1 , v3 ] + [v1 , v2 ] − [v2 , v3 ] + [v0 , v3 ] − [v0 , v2 ] + [v1 , v3 ]
− [v0 , v3 ] + [v0 , v1 ] − [v1 , v2 ] + [v0 , v2 ] − [v0 , v1 ]
=0
These computations illustrate a general fact.

Lemma 7.2.1
∂p−1 ◦ ∂p = 0.

Proof. Let σ = [v0 , . . . , vp ].

X
p
∂p−1 ∂p σ = (−1)i ∂p−1 [v0 , . . . , vbi , . . . , vp ]
i=0
 
X
p X X
= (−1)i  (−1)j [. . . , vbj , . . . , vbi , . . .] + (−1)j−1 [. . . , vbi , . . . , vbj , . . .]
i=0 j<i j>i
p X
X p X
X
= (−1)i (−1)j [. . . , vbj , . . . , vbi , . . .] + (−1)i (−1)j−1 [. . . , vbi , . . . , vbj , . . .]
i=0 j<i i=0 j>i
Xp X XX
p
= (−1)i+j [. . . , vbj , . . . , vbi , . . .] − (−1)i+j [. . . , vbi , . . . , vbj , . . .]
i=0 j<i i=0 j>i

Now let’s fix two positive integers i0 and j0 with j0 < i0 . Now, the term corresponding to i = i0 ≤ p
and j = j0 < i0 in the first summand is:

(−1)i0 +j0 [. . . , vc c
j0 , . . . , v i0 , . . .]

The term corresponding to i = j0 ≤ p and j = i0 > j0 in the second summand is:

(−1)i0 +j0 [. . . , vc c
j0 , . . . , v i0 , . . .]

These two terms cancel each other. This way, all the terms get canceled. So ∂p−1 ∂p σ = 0. ■

73
7 Lecture 7 74

Definition 7.2.2. The kernel of the homomorphism ∂p : Cp (K) → Cp−1 (K) is a subgroup of
Cp (K). It is called the group of p-cycles and is denoted by Zp (K).
The image of the homomorphism ∂p+1 : Cp+1 (K) → Cp (K) is a subgroup of Cp (K). It is
called the group of p-cycles and is denoted by Bp (K).

By Lemma 7.2.1, ∂p ◦∂p+1 = 0. As a result, the image of the boundary operator ∂p+1 is contained in the
kernel of the boundary operator ∂p . In other words, each boundary of a (p + 1)-chain is automatically
a p-chain. So Bp (K) ⊆ Zp (K). Zp (K) is a abelian group, so all its subgroups are normal. Thus
Bp (K) is a normal subgroup of Zp (K), and so the quotient group Zp (K) /Bp (K) is defined.

Definition 7.2.3 (Homology Group). The p-th homology group Hp (K) of K is defined as

Hp (K) = Zp (K) /Bp (K)

§7.3 Homology Group Computation

Let’s now do a few computation of homology groups.

Example 7.3.1. Consider the following 1-dimensional simplicial complex K. The underlying space
|K| of this complex K is the boundary of a square with edges e1 , e2 , e3 , e4 .

v3 e3 v2
K

e4
e2

v0 v1
e1

The group C1 (K) is free abelian of rank 4. It is generated by {e1 , e2 , e3 , e4 }. To put it more con-
cretely, it is generated by the elementary 1-chains corresponding to the oriented 1-simplices e1 , e2 , e3 , e4 .
A generic element of C1 (K) is a Z-linear combination of e1 , e2 , e3 , e4 . So, if c is a general 1-chain,

X
4
c= ni e i = n1 e 1 + n2 e 2 + n3 e 3 + n4 e 4 ; ni ∈ Z
i=1
∂1 c = n1 (v1 − v0 ) + n2 (v2 − v1 ) + n3 (v3 − v2 ) + n4 (v0 − v3 )
= (n4 − n1 ) v0 + (n1 − n2 ) v1 + (n2 − n3 ) v2 + (n3 − n4 ) v3

Therefore, c is a 1-cycle, i.e. ∂1 c = 0 if and only if n1 = n2 = n3 = n4 . Hence, a generic 1-cycle can

be written as
c = n1 (e1 + e2 + e3 + e4 ) with n1 ∈ Z
Therefore, Z1 (K) is infinite cyclic and generated by the 1-chain e1 + e2 + e3 + e4 . So Z1 (K) ∼
= Z.
Since there are no 2-simplex in K, C2 (K) is, by definition, trivial. And hence ∂2 is the trivial
homomorphism. As a result, B1 (K) is trivial. Therefore,

H1 (K) = Z1 (K) /B1 (K) = Z1 (K) ∼

=Z

Example 7.3.2. Now consider the following complex L.

74
7 Lecture 7 75

v3 e3 v2
L
σ2
e4
e5 e2
σ1

v0 v1
e1

The underlying space |L| is a square. So, if c is a general 1 chain,

X
4
c= ni e i = n1 e 1 + n2 e 2 + n3 e 3 + n4 e 4 + n5 e 5 ; ni ∈ Z
i=1
∂1 c = n1 (v1 − v0 ) + n2 (v2 − v1 ) + n3 (v3 − v2 ) + n4 (v0 − v3 ) + n5 (v2 − v0 )
= (n4 − n1 − n5 ) v0 + (n1 − n2 ) v1 + (n2 − n3 + n5 ) v2 + (n3 − n4 ) v3

Hence, c is a 1-cycle if the values of ∂1 c on the vertices are all 0. In other words,

n4 − n1 − n5 = 0 , n 1 − n2 = 0 , n 2 − n3 + n5 = 0 , n 3 − n4 = 0

These equations give us n1 = n2 ; n3 = n4 ; n5 = n3 − n2 . Here we have 2 degrees of freedom. Because

we can assign the values of n2 and n3 arbitrarily to obtain the values of n1 , n4 and n5 . Thus a generic
1-cycle c can be written as

c = n2 e1 + n2 e2 + n3 e3 + n3 e4 + (n3 − n2 ) e5 = n2 (e1 + e2 − e5 ) + n3 (e3 + e4 + e5 )

So Z1 (L) is free abelian with rank 2. A basis for Z1 (L) consists of the 1-chains e1 + e2 − e5 and
e3 + e4 + e5 . Hence, Z1 (L) ∼
= Z2 .
On the other hand,
∂2 σ1 = e1 + e2 − e5 and ∂2 σ2 = e3 + e4 + e5
So, a generic 1-boundary can be written as a Z-linear combination of e1 + e2 − e5 and e3 + e4 + e5 .
Hence, B1 (L) ∼= Z2 .
H1 (L) = Z1 (L) /B1 (L) ∼= Z2 /Z2 =⇒ H1 (L) = 0
Since there are no 3-simplex in L, C3 (L) is, by definition, trivial. And hence ∂3 is the trivial homo-
morphism. As a result, B2 (L) is trivial. Now, if c2 is a generic 2-chain,

c2 = m1 σ1 + m2 σ2 =⇒ ∂2 c2 = m1 (e1 + e2 − e5 ) + m2 (e3 + e4 + e5 )

If c2 is a 2-cycle, we must have m1 = m2 = 0. Hence Z2 (L) is trivial. Therefore,

H2 (L) = Z2 (L) /B2 (L) = 0

At this stage, we need some terminologies.

Definition 7.3.1. We shall say that a chain c is carried by a subsimplex L of K, if c has

value 0 on every simplex that is not in L. And we say two p-chains c and c′ are homologous if
c − c′ = ∂p+1 d for some (p + 1)-chain d. In particular, if c = ∂p+1 d, we say that c is homologous
to 0, or simply c bounds.

Example 7.3.3. Consider the following complex M whose underlying space is a square.

75
7 Lecture 7 76

v3 e7 v2
M
e4 σ3 e3

e8 v
σ4 σ2 e6

e1 σ1 e2

v0 v1
e5

Instead of computing the group of 1-cycles, we reason as follows: Given a 1-chain c, let a be its
value on e1 . Then we claim that the chain

c1 = c + ∂2 (aσ1 )

has value 0 on e1 . To prove it, we write c as

X
c = ae1 + ni ei , with ni ∈ Z
i̸=1

Since ∂2 (σ1 ) = e5 + e2 − e1 , we get

X X
c1 = ae1 + ni ei + a (e5 + e2 − e1 ) = n′i ei
i̸=1 i̸=1

So, the value of c1 on e1 is indeed 0.

This can be interpreted as follows: after modifying the 1-chain c by a boundary, one “pushes it off
e1 ”. Our next goal is to “push c1 off e2 ”. Let b be the value of c1 on e2 . So we have
X
c1 = be2 + ni ei , with ni ∈ Z
i̸=2

We take the chain c2 = c1 +∂2 (bσ2 ), and we claim that its value on e2 is 0. Using ∂2 (σ2 ) = e6 +e3 −e2 ,
we get X X
c2 = be2 + ni ei + b (e6 + e3 − e2 ) = n′i ei
i̸=2 i̸=2

So, the value of c2 on e2 is indeed 0. In other words, by modifying c1 by a boundary, we “pushed c1

off e2 ”.
Now our goal is to “push c2 off e3 ”. Let d be the value of c2 on e3 . So we have
X
c2 = de3 + ni ei , with ni ∈ Z
i̸=3

We take the chain c3 = c2 +∂2 (dσ3 ), and we claim that its value on e3 is 0. Using ∂2 (σ3 ) = e7 +e4 −e3 ,
we get X X
c3 = de3 + ni ei + d (e7 + e4 − e3 ) = n′i ei
i̸=3 i̸=3

So, the value of c3 on e3 is indeed 0. In other words, by modifying c1 by a boundary, we “pushed c2

off e3 ”.

c3 = c2 + ∂2 (dσ3 ) = c1 + ∂2 (bσ2 ) + ∂2 (dσ3 ) = c + ∂2 (aσ1 ) + ∂2 (bσ2 ) + ∂2 (dσ3 )

= c + ∂2 (aσ1 + bσ2 + dσ3 )

Therefore, given a 1-chain c, we can find a chain c3 that is homologous to c, and carried by the
following subcomplex:

76
7 Lecture 7 77

v3 e7 v2

e4

e8 v
e6

v0 v1
e5

If c happens to be a cycle, then c3 is also a cycle. Let c3 = n4 e4 + n5 e5 + n6 e6 + n7 e7 + n8 e8 , with

each ni ∈ Z.
∂2 c3 = n4 (v − v3 ) + n5 (v1 − v0 ) + n6 (v2 − v1 ) + n7 (v3 − v2 ) + n8 (v0 − v3 )
= n4 v + (n8 − n5 ) v0 + (n5 − n6 ) v1 + (n6 − n7 ) v2 + (n7 − n8 − n4 ) v3

The only way ∂2 c3 can be 0 is n4 = 0 and n5 = n6 = n7 = n8 . Thus, for the 1-chain c3 to be a cycle,
it must vanish on e4 . In other words, every 1-cycle of M is homologous to a 1-cycle carried by the
boundary of the square.
∂2 σ1 + ∂2 σ2 + ∂2 σ3 + ∂2 σ4 = (e5 + e2 − e1 ) + (e6 + e3 − e2 ) + (e7 + e4 − e3 ) + (e8 + e1 − e4 )
= e5 + e6 + e7 + e8
=⇒ c3 = n5 (e5 + e6 + e7 + e8 ) = ∂2 (n5 σ1 + n5 σ2 + n5 σ3 + n5 σ4 )

Therefore, the 1-cycle c3 is homologous to 0, i.e. c3 bounds. Since the given 1-cycle c is homologous
to c3 , c also bounds. Hence, every 1-cycle of M bounds; so H1 (M ) = 0.
A generic 2-chain on M is given by

c = m1 σ1 + m2 σ2 + m3 σ3 + m4 σ4 ; with mi ∈ Z
∂2 c = m1 (e5 + e2 − e1 ) + m2 (e6 + e3 − e2 ) + m3 (e7 + e4 − e3 ) + m4 (e8 + e1 − e4 )
= (m4 − m1 ) e1 + (m1 − m2 ) e2 + (m2 − m3 ) e3 + (m3 − m4 ) e4
+ m1 e5 + m2 e6 + m3 e7 + m4 e8
∂2 c = 0 if and only if m1 = m2 = m3 = m4 = 0. This means that the group Z2 (M ) of 2-cycles on M
is trivial. Hence, H2 (M ) = 0.
Note that, the underlying spaces for the complex L and M , dealt in Example 7.3.2 and Example 7.3.3,
respectively, are both square. Using the result of the respective examples, we find that the homology
groups of L and M are exactly the same. This validates the fact that the homology groups of a
complex depend only on the underlying space, which will be proved in Algebraic Topology II.
Example 7.3.4. In this example, we are going to compute the first homology groups of the following
complexes and interpret the results geometrically:

v3 e3 v2 e3 v2
v3
K
L
e8 e7

e4 v4 e4
e2 e5 e2

e5 e6

v0 v1 v0 v1
e1 e1

77
7 Lecture 7 78

First let’s focus on H1 (K). There are no 2-simplices in K. So C2 (K) = 0, and hence ∂2 is trivial.
Therefore,
H1 (K) = Ker ∂1 / im ∂2 = Ker ∂1
P8
A general 1-chain of K is of the form c = ni ei , where ni ∈ Z.
i=1

∂1 c = n1 (v1 − v0 ) + n2 (v2 − v1 ) + n3 (v3 − v2 ) + n4 (v0 − v3 ) + n5 (v4 − v0 ) + n6 (v4 − v1 )

+ n7 (v4 − v2 ) + n8 (v4 − v3 )
= v0 (−n1 + n4 − n5 ) + v1 (n1 − n2 − n6 ) + v2 (n2 − n3 − n7 ) + v3 (n3 − n4 − n8 )
+ v4 (n5 + n6 + n7 + n8 )
If c is a 1-cycle, then ∂1 c = 0, so we must have
−n1 + n4 − n5 = n1 − n2 − n6 = n2 − n3 − n7 = n3 − n4 − n8 = n5 + n6 + n7 + n8 = 0
So n5 = n4 − n1 , n6 = n1 − n2 , n7 = n2 − n3 and n8 = n3 − n4 .
n5 + n6 + n7 + n8 = (n4 − n1 ) + (n1 − n2 ) + (n2 − n3 ) + (n3 − n4 ) = 0
Therefore,
c = n1 e1 + n2 e2 + n3 e3 + n4 e4 + (n4 − n1 ) e5 + (n1 − n2 ) e6 + (n2 − n3 ) e7 + (n3 − n4 ) e8
= n1 (e1 − e5 + e6 ) + n2 (e2 − e6 + e7 ) + n3 (e3 − e7 + e8 ) + n4 (e4 + e5 − e8 )
So, if c ∈ Ker ∂1 , it can be expressed as an integral linear combination of e1 − e5 + e6 , e2 − e6 + e7 ,
e3 − e7 + e8 and e4 + e5 − e8 . So Ker ∂1 is free abelian of rank 4. Hence,
H1 (K) = Ker ∂1 ∼ = Z4
Now we are going to compute H1 (L). As before, there are no 2-simplices in L. So C2 (L) = 0, and
hence ∂2 is trivial. Therefore,
H1 (L) = Ker ∂1 / im ∂2 = Ker ∂1
P
5
A general 1-chain of L is of the form c = ni ei , where ni ∈ Z.
i=1

∂1 c = n1 (v1 − v0 ) + n2 (v2 − v1 ) + n3 (v3 − v2 ) + n4 (v0 − v3 ) + n5 (v2 − v0 )

= v0 (−n1 + n4 − n5 ) + v1 (n1 − n2 ) + v2 (n2 − n3 + n5 ) + v3 (n3 − n4 )
If c is a 1-cycle, then ∂1 c = 0, so we must have
−n1 + n4 − n5 = n1 − n2 = n2 − n3 + n5 = n3 − n4 = 0
So, n1 = n2 , n3 = n4 , and n5 = −n1 + n4 = n3 − n1 . Therefore,
c = n1 e1 + n1 e2 + n3 e3 + n3 e4 + (n3 − n1 ) e5
= n1 (e1 + e2 − e5 ) + n3 (e3 + e4 + e5 )
So, if c ∈ Ker ∂1 , it can be expressed as an integral linear combination of e1 + e2 − e5 and e3 + e4 + e5 .
So Ker ∂1 is free abelian of rank 2. Hence,
H1 (L) = Ker ∂1 ∼ = Z2
Now, how can we interpret these results geometrically? Notice that K has 4 “holes”. A n-
dimensional hole can be mathematically defined as an n-cycle which is not a boundary of any
(n + 1)-dimensional object, in our case (n + 1)-chain. By this definition, e1 + e6 − e5 is a 1-dimensional
hole of K. Similarly, e2 + e7 − e6 , e3 + e8 − e7 and e4 + e5 − e8 are also 1-dimensional holes of K. And
all of them generate H1 (K). So, we can say that the Betti number of H1 (K) correspond to the
number of 1-dimensional holes of K. In general, it is true that the Betti number of Hp (K) denotes the
number of p-dimensional holes of K. It is not so hard to prove, and interested readers are encouraged
to prove it.
Similarly, for L, since H1 (L) ∼ = Z2 , the Betti number of H1 (L) is 2. And L has exactly two 1-
dimensional holes: they are e3 + e4 + e5 and e1 + e2 − e5 . It is of no surprise that they are the generator
of H1 (L). So the Betti number of H1 (L) is equal to the number of 1-dimensional holes.

78
A Category Theory Basics
Lecture videos and lecture notes of Category Theory course.

§A.1 What is a Category?

Definition A.1.1 (Category). A category C consists of

• A collection C0 whose elements are called the objects of C. Elements of C0 are denoted by
uppercase letters X, Y, Z, . . .

• A collection C1 whose elements are called the morphisms or arrows of C. Elements of C1 are
denoted by lowercase letters f, g, h, . . .

such that the following hold:

(i) Each morphism assigns two objects called source (or domain) and target (or codomain).
We denote them by s(f ) and t(f ), respectively for a given arrow f . If s(f ) = X ∈ C0 and
t(f ) = Y ∈ C0 for a given f ∈ C1 , we write
f
f : X → Y , or X −→ Y

(ii) Each object X ∈ C0 has a distinguished morphism idX : X → X.

(iii) For each pair of morphisms f, g ∈ C1 such that t(f ) = s(g), there exist specified morphisms
g ◦ f called composite morphisms such that

s (g ◦ f ) = s(f ) and t (g ◦ f ) = t(g)

f g
In other words, X −→ Y −→ Z implying g ◦ f : X → Z.

These structures need to satisfy the following axioms:

(a) (Unitality) For each morphism f : X → Y ,

f ◦ idX = idY ◦f = f

Warning: idX ◦f doesn’t make sense, so doesn’t f ◦ idY .

(b) (Associativity) For X, Y, Z, W ∈ C0 and f, g, h ∈ C1 satisfying

f g h
X −→ Y −→ Z −→ W

one must have

h ◦ (g ◦ f ) = (h ◦ g) ◦ f

The two morphisms that are set equal using associativity axiom can be understood more clearly using
the following diagrams:
f g h f g h
X Y Z W X Y Z W
g◦f h◦g

h ◦ (g ◦ f ) (h ◦ g) ◦ f

79
A Category Theory Basics 80

Example A.1.1
We can take a collection of groups in C0 and the group homomorphisms between them in C1 . In
other words, the objects are groups and the morphisms are group homomorphisms. This forms a
category.
Similarly, one can form a category of topological spaces too. In that case, the morphisms will
be continuous functions between the spaces.

§A.2 Functor and Functorial Properties

Definition A.2.1 (Functor). Let C = (C0 , C1 ) and D = (D0 , D1 ) be two categories. A functor F
from C to D, denoted by F : C → D, is a map that has the following properties:

i. F maps objects of C0 to objects of D0 .

ii. F maps morphisms of C1 to morphisms of D1 , such that for f ∈ C1 and X, Y ∈ C0

f : X → Y =⇒ F (f ) : F (X) → F (Y )

iii. For every X ∈ C0 ,

F (idX ) = idF (X)

iv. For f, g ∈ C1 with t(f ) = s(g) (in other words, g ◦ f makes sense), condition ii guarantees
that t (F (f )) = s (F (g)) (so F (g) ◦ F (f ) makes sense). Then we must have

F (g ◦ f ) = F (g) ◦ F (f )

The definition of functor can be visualized using the following diagram:

Category C
idX

f g
··· X Y Z ···

idF (X)
Category D

F (f ) F (g)
··· F(X) F(Y ) F(Z) ···

Question. Why is Theorem 2.3.2 called functorial properties of induced homomorphism? Is it some-
how related to functors?

The answer is, yes. It is indeed related to functors. We shall use the same notations as Theorem 2.3.2
here.
Recall from Example A.1.1 that, we can make a category of topological spaces. Pointed topological
spaces are not an exception; we can make a category of pointed topological spaces with the spaces
(X, x0 ), (Y, y0 ) and (Z, z0 ). These three spaces are gonna be the objects of the category.
But what are the morphisms of this category? We’ve seen in the aforementioned example that the
morphisms are continuous functions. Here we already have two continuous maps,

hx0 : (X, x0 ) → (Y, y0 ) and ky0 : (Y, y0 ) → (Z, z0 )

80
A Category Theory Basics 81

Also, composition of continuous maps is indeed continuous, so ky0 ◦ hx0 : (X, x0 ) → (Z, z0 ) is also a
morphism of this category.
The identity maps are also continuous, so we can take the identity maps ix0 : (X, x0 ) → (X, x0 ),
iy0 : (Y, y0 ) → (Y, y0 ) and iz0 : (Z, z0 ) → (Z, z0 ) as morphisms. Thus this category of pointed
topological spaces is formed.
ix0 i y0 i z0

hx 0 k y0
(X, x0 ) (Y, y0 ) (Z, z0 )

We can also make a category of fundamental groups. For that, the fundamental groups π1 (X, x0 ),
π1 (Y, y0 ) and π1 (Z, z0 ) are gonna be the objects of the category. And the morphisms are groups homo-
morphisms between them. In this case, we can take the induced homomorphisms between fundamental
groups. Also, we take the identity homomorphisms to make it a category.
idπ1 (X,x0 ) idπ1 (Y,y0 ) idπ1 (Z,z0 )

( hx 0 ) ∗ ( k y0 ) ∗
π1 (X, x0 ) π1 (Y, y0 ) π1 (Z, z0 )

Now, if there is a functor F between these two categories, it must take each topological spaces to
the respective fundamental groups.




F (X, x0 ) = π1 (X, x0 ) , F (Y, y0 ) = π1 (Y, y0 ) , F (Z, z0 ) = π1 (Z, z0 )

To satisfy the second criterion of functor, as hx0 : (X, x0 ) → (Y, y0 ), we must have



F (hx0 ) : F (X, x0 ) → F (Y, y0 ) =⇒ F (hx0 ) : π1 (X, x0 ) → π1 (Y, y0 )
=⇒ F (hx0 ) = (hx0 )∗

Similarly for other continuous functions, F maps them to their respective induced homomorphisms.
For F to satisfy the 3rd property of functor, we must have


F id(X,x0 ) = id
=⇒ F (ix ) = id
F (X,x0 ) 0 π1 (X,x0 ) =⇒ (ix0 )∗ = idπ1 (X,x0 )

So, for F to be a functor, (ix0 )∗ must be the identity homomorphism of π1 (X, x0 ). This is guaranteed
by Theorem 2.3.2.
Now, For F to satisfy the 4th property of functor, we must have

F (ky0 ◦ hx0 ) = F (ky0 ) ◦ F (hx0 ) =⇒ (ky0 ◦ hx0 )∗ = (ky0 )∗ ◦ (hx0 )∗

which is, again, guaranteed by Theorem 2.3.2.

81
A Category Theory Basics 82

Category of Pointed Topological Spaces

i x0 i y0 i z0

hx 0 k y0
(X, x0 ) (Y, y0 ) (Z, z0 )

k y0 ◦ h x 0

Category of Groups
( ix0 ) ∗ ( i y0 ) ∗ ( i z0 ) ∗

( hx 0 ) ∗ ( k y0 ) ∗
π1 (X, x0 ) π1 (Y, y0 ) π1 (Z, z0 )

( k y0 ) ∗ ◦ ( h x 0 ) ∗

To summarize, Theorem 2.3.2 ensures that F is a functor between the category of pointed topological
spaces and the category groups. That’s why this theorem is called functorial properties of induced
homomorphism.

82