Basic Algebraic & Differential

Topology

Arjun Paul
Department of Mathematics
Jadavpur University,
Kolkata - 700032, India.
Email: [email protected].

Version: December 12, 2021 at 12:38am (IST).

Note: This note will be updated from time to time.

If there are any mistakes, please bring it to my notice.
 iii

Contents

List of Symbols v

1 Review of Point Set Topology 1

1.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Quotient space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.3 Projective space and Grassmannian . . . . . . . . . . . . . . . . . . 20

1.3.1 Real and complex projective spaces . . . . . . . . . . . . . 20

1.3.2 Grassmannian Gr(k, Rn ) . . . . . . . . . . . . . . . . . . . . 23

1.4 Topological group . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2 Homotopy Theory 31

2.1 Homotopy of maps . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

2.2 Fundamental group . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.2.1 Construction . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.2.2 Functoriality . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.2.3 Dependency on base point . . . . . . . . . . . . . . . . . . 45

2.2.4 Fundamental group of some spaces . . . . . . . . . . . . . 50

2.3 Covering Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

2.3.1 Covering map . . . . . . . . . . . . . . . . . . . . . . . . . . 51

2.3.2 Fundamental group of S1 . . . . . . . . . . . . . . . . . . . 65

iv

2.3.3 Fundamental group of Sn , for n ≥ 2 . . . . . . . . . . . . . 67

2.3.4 Some applications . . . . . . . . . . . . . . . . . . . . . . . 69

2.4 Galois theory for covering spaces . . . . . . . . . . . . . . . . . . . 76

2.4.1 Universal cover . . . . . . . . . . . . . . . . . . . . . . . . . 76

2.4.2 Construction of universal cover . . . . . . . . . . . . . . . . 78

2.4.3 Group action and covering map . . . . . . . . . . . . . . . 82

2.4.4 Group of Deck transformations . . . . . . . . . . . . . . . . 84

2.4.5 Galois covers . . . . . . . . . . . . . . . . . . . . . . . . . . 88

2.4.6 Galois correspondence for covering spaces . . . . . . . . . 90

3 Differential Topology 91

3.1 Review of Rn Calculus . . . . . . . . . . . . . . . . . . . . . . . . . 91

3.1.1 Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . 93

3.2 Differentiable Manifolds . . . . . . . . . . . . . . . . . . . . . . . . 100

4 Appendix 103

4.1 Category and Functor . . . . . . . . . . . . . . . . . . . . . . . . . 103

 v

List of Symbols

∅ Empty set
Z The set of all integers
Z≥0 The set of all non-negative integers
N The set of all natural numbers (i.e., positive integers)
Q The set of all rational numbers
R The set of all real numbers
C The set of all complex numbers
< Less than
≤ Less than or equal to
> Greater than
≥ Greater than or equal to
⊂ Proper subset
⊆ Subset or equal to
( Subset but not equal to (c.f. proper subset)
∃ There exists
@ Does not exists
∀ For all
∈ Belongs to
∈
/ Does not belong to
∑ Sum
∏ Product
± Plus and/or minus
∞
√ Infinity
a Square root of a
∪
F
Union
Disjoint union
∩ Intersection
A → B A mapping into B
a 7→ b a maps to b
,→ Inclusion map
A\B A setminus B
∼
= Isomorphic to
A := ... A is defined to be ...
 End of a proof
vi

Symbol Name Symbol Name

α alpha β beta
γ gamma δ delta
π pi φ phi
ϕ var-phi ψ psi
e epsilon ε var-epsilon
ζ zeta η eta
θ theta ι iota
κ kappa λ lambda
µ mu ν nu
υ upsilon ρ rho
$ var-rho ξ xi
σ sigma τ tau
χ chi ω omega
Ω Capital omega Γ Capital gamma
Θ Capital theta ∆ Capital delta
Λ Capital lambda Ξ Capital xi
Σ Capital sigma Π Capital pi
Φ Capital phi Ψ Capital psi

Some of the useful Greek letters

 1

Chapter 1

Review of Point Set Topology

In this chapter, we recall some topological preliminaries from basic topology

course. Star marked exercises are difficult one and are not required for this
course.

1.1 Preliminaries

A topology on a set X is given by specifying which subsets of X are ‘open’.

Naturally those subsets should satisfies certain properties as we are familiar
from basic analysis and metric space courses.

Definition 1.1.1. A topology is on a set X is a collection τ of subsets of X satis-

fying the following properties:

(i) ∅ and X are in τ,

(ii) for any collection {Uα }α∈Λ of objects of τ, their union Uα ∈ τ,

S
α∈Λ

n
(iii) for a finite collection of objects U1 , . . ., Un ∈ τ, their intersection Ui ∈ τ.
T
i =1

The pair ( X, τ ) is called a topological space, and the objects of τ are called open
subsets of ( X, τ ). For notational simplicity, we suppress τ and denote a topo-
logical space ( X, τ ) simply by X.

Joke: An empty set may contain some air since it is open!

2 Chapter 1. Review of Point Set Topology

Remark 1.1.2. One can also define a topology on a set X by considering a col-
lection τc of subsets of X such that

(a) both ∅ and X are in τc ,

(b) τc is closed under arbitrary intersections, and

(c) τc is closed under finite unions.

This is known as the closed set axioms for a topology. In this settings, the objects of
τc are called closed subsets of X. It is easy to switch between these two definitions
by taking complements of objects of τ and τc in X. However, unless explicitly
mentioned, we usually work with open set axioms for topology.

Example 1.1.3. (i) If X = ∅ then τ = {∅} is the only topology on ∅.

(ii) Let X be non-empty set X. Then ( X, P ( X )) is a topological space. Such

a topology, where all subsets of X are open, is called the discrete topology
on X. Note that ( X, τ ) with τ = {∅, X } is a topological space; such a
topology on X is called the indiscrete topology on X.

(iii) Let X 6= ∅, and let τ = { A ∈ P ( X ) : X \ A is finite}. Then ( X, τ ) is a

topological space; such a topology is called the cofinite topology on X.

(iv) The set Rn together with the collection of all usual open subsets of Rn is a
topological space, known as the usual/Euclidean topology on Rn .

(v) Any metric space ( X, d) is a topological space where the topology on X is

given by the collection of all open subsets of ( X, d).

Exercise∗ 1.1.4. This exercise is for readers interested in commutative algebra.

Let A be a commutative ring with identity. Let Spec( A) be the set of all prime
ideals of A, known as the spectrum of A. For each subset E ⊆ A, let

V ( E) := {p ∈ Spec( A) : E ⊆ p}.

Prove the following.

(i) V ( A) = ∅ and V (0) = Spec( A).

(ii) V ( E) = V (a), where a ⊆ A is the ideal generated by E ⊆ A.

1.1. Preliminaries 3

(iii) V a ∪ b = V (ab), for all ideals a, b of A.

V (ai ) = V ( ∑ ai ), for any collection of ideals {ai : i ∈ I } of A.

T
(iv)
i∈ I i∈ I

(v) Conclude that the collection {V (a) : a is an ideal of A} satisfies axioms

for closed subsets of a topological space. The resulting topology on Spec( A)
is called the Zariski topology on Spec( A).

(vi) For any ideal a of A, show that Spec( A/a) is homeomorphic to the closed
subspace V (a) of Spec( A).

(vii) Let X be a topological space. A point ξ ∈ X is said to be a closed point

(resp., a generic point) of X if {ξ } = {ξ } (resp., {ξ } = X). If A is an inte-
gral domain, show that Spec( A) contains a unique generic point, which is
precisely the zero ideal of A.

(viii) Show that the Zariski topology on Spec( A) is not even T1 let alone be it
Hausdorff.

(ix) Show that, a point m ∈ Spec( A) is closed if and only if m is a maximal

ideal of A.

(x) Let k be an algebraically closed field, and let A = k [ x1 , . . . , xn ]. Use

Hilbert’s Nullstellensatz to show that the set of all closed points of Spec( A)
is in bijection with the set kn := ( a1 , . . . , an ) : ai ∈ k, ∀ i ∈ {1, . . . , n} .


(xi) Let k be an algebraically closed field. Fix a subset S ⊆ A := k [ x1 , . . . , xn ],

and let aS ⊆ A be the ideal generated by S. Show that the set

Z (S) := {( a1 , . . . , an ) ∈ kn : f ( a1 , . . . , an ) = 0, ∀ f ∈ S}

is in bijection with the set of all closed points of V (aS ) ⊆ Spec( A).

The space Spec( A) carries rich algebro-geometric structure. They are called
affine scheme, and are building block of all schemes in the sense that any scheme
is build by suitably gluing affine schemes.

Definition 1.1.5. Let X and Y be topological spaces. A map f : X → Y is said

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

As an immediate consequence of the above definition, we get the following.

4 Chapter 1. Review of Point Set Topology

Corollary 1.1.6. Let X be a non-empty set together with two topologies τ1 and τ2 . For
each j = 1, 2, let X j = ( X, τj ) be the topological space whose underlying set is X and
the topology is τj . Then τ2 is finer than τ1 (i.e., τ1 ( τ2 ) if and only if the identity map
IdX : X2 → X1 is continuous.

Definition 1.1.7. A topological space X is said to be Hausdorff or, T2 or, separated

if each pair of distinct points of X can be separated by a pair of disjoint open
neighbourhoods of them. In other words, give x1 , x2 ∈ X with x1 6= x2 , there
exist open subsets V1 , V2 of X with x1 ∈ V, x2 ∈ V2 and V1 ∩ V2 = ∅.

Exercise 1.1.8. Show that a topological space is Hausdorff if and only if the
image of the diagonal map

∆ X : X → X × X, x 7→ ( x, x )

is closed in X × X. (Hint: Look at the complement of ∆ X ( X ) in X × X.)

Exercise 1.1.9. Let f , g : X → Y be continuous maps of topological spaces. If Y

is Hausdorff, show that { x ∈ X : f ( x ) = g( x )} is closed in X. (Hint: Look at
the inverse image of ∆Y (Y ) ⊂ Y × Y under the map ( f , g) : X → Y × Y given
by x 7→ ( f ( x ), g( x )).)

Exercise∗ 1.1.10. Given continuous maps f 1 : X1 → Y and f 2 : X2 → Y of

topological spaces, we define their fiber product to be a topological space X × Z Y
together with continuous maps p1 : X1 ×Y X2 → X1 and p2 : X1 ×Y X2 → X2
such that

(i) f 1 ◦ p1 = f 2 ◦ p2 , and

(ii) given any topological space Z, and continuous maps g1 : Z → X1 and

g2 : Z → X2 satisfying f 1 ◦ g1 = f 2 ◦ g2 , there is a unique continuous map
g : Z → X1 ×Y X2 such that p1 ◦ g = g1 and p2 ◦ g = g2 .

Z g2
g

$ p2 &
X1 × Y X2 / X2
g1
p1 f2
%  f1 
X1 /Y

Prove the following.

1.2. Quotient space 5

(a) The fiber product X1 ×Y X2 exists and is unique up to a unique homeomor-

Take X1 ×Y X2 := {( x1 , x2 ) ∈ X1 × X2 : f 1 ( x1 ) = f 2 ( x2 )}
phism. (Hint: .)

(b) If X2 ⊆ Y and f 2 : X2 ,→ Y is the inclusion map, show that f 1−1 ( X2 ) is

homeomorphic to the fiber product X1 ×Y X2 . (Hint: Verify properties (i)
and (ii) for f 1−1 ( X2 ).)

1.2 Quotient space

Let’s recall the notion of quotients from algebra course. Let G be a group
and H a normal subgroup of G. Then we have a relation ∼ on G defined by

g1 ∼ g2 if g1−1 g2 ∈ H.

Clearly this is an equivalence relation on G, and we have a partition of G into a

disjoint union of its subsets (equivalence classes)
[
G= gH,
g∈ G

where gH = { g0 ∈ G : g ∼ g0 } is the equivalence class of g in G, for all g ∈ G.

Now question is does there exists a pair ( Q, q) consisting of a group Q and

a map q : G → Q such that

(QG1) q : G → Q is a surjective group homomorphism, and

(QG2) given any group G 0 and a group homomorphism f : G → G 0 with H ⊆

Ker( f ), there should exists a unique group homomorphism fe : Q → G 0
such that fe ◦ q = f ?
f
G / G0
8
q (1.2.1)
 fe
Q

Interesting point is that, without knowing existence of such a pair ( Q, q), it fol-
lows immediately from the properties (QG1) and (QG2) that such a pair ( Q, q),
if it exists, must be unique up to a unique isomorphism of groups in the sense
6 Chapter 1. Review of Point Set Topology

that, given another such pair ( Q0 , q0 ) satisfying the above two conditions, there
is a unique group isomorphism φ : Q → Q0 such that φ ◦ q = q0 ).

Exercise 1.2.2. Prove the above mentioned uniqueness statement.

Now question is about its existence. The condition (QG1) suggests that the
elements of Q should be the fibers of the map q, which are nothing but the
∼-equivalence classes

[ g]∼ = { g0 ∈ G : g0 ∼ g} = gH, ∀ g ∈ G.

This suggests us to consider { gH : g ∈ G } as a possible candidate for the set Q.

Now question is what should be the appropriate group structure on it? Take
any group homomorphism f : G → G 0 such that H ⊆ Ker( f ). This says that
f ( g1 ) = f ( g2 ) if g1 ∼ g2 (equivalently, g1−1 g2 ∈ H). The commutativity of the
diagram (1.2.1) tells us to send gH ∈ Q to f ( g) ∈ G 0 to define the map fe : Q →
G 0 (note that this is well-defined!), and since we want fe : Q → G 0 to be a group
homomorphism, we should define a binary operation on Q = { gH : g ∈ G } in
fe
such a way that ( g1 H ) ∗ ( g2 H ) −→ f ( g1 ) f ( g2 ) = f ( g1 g2 ), for all g1 , g2 ∈ G. So
the obvious choice is to define

( g1 H ) ∗ ( g2 H ) := ( g1 g2 ) H, ∀ g1 , g2 ∈ G. (1.2.3)

Clearly this is a well-defined binary operation on Q = { gH : g ∈ G }.

Exercise 1.2.4. Verify that (1.2.3) makes Q a group such that the pair ( Q, q)
satisfies the condition (QG1) and (QG2).

Exercise 1.2.5. Verify analogous stories for the cases rings and vector spaces.

Now we are going to witness the same story in topology! Let X be a topo-
logical space.

Definition 1.2.6. Given an equivalence relation ∼ on X, the associated quotient

topological space (or, identification space) X/∼ is a pair ( Q, q) consisting of a topo-
logical space Q and a continuous map q : X → Q such that

(QT1) q is surjective and satisfies q( x ) = q( x 0 ) whenever x ∼ x 0 in ( X, ∼); and

1.2. Quotient space 7

(QT2) given any topological space Y and a continuous map f : X → Y satisfy-

ing f ( x ) = f ( x 0 ) whenever x ∼ x 0 in ( X, ∼), there is a unique continu-
ous map fe : Q → Y such that fe ◦ q = f .

f
X 8/ Y
q
 fe
Q

The map q is called the quotient map (or, identification map) for ( X, ∼).

As an immediate corollary to the Definition 1.2.6, we have the following.

Corollary 1.2.7. If ( Q, q) is a quotient space for ( X, ∼), then the topology on Q is

the largest topology on the set Q such that the map q : X → Q is continuous.

Proof. By a topology on a set S we mean a collection τ of subsets of S that sat-

isfies axioms for open subsets in S. Suppose on the contrary that the statement
in Corollary 1.2.7 is false. Then there is a topology τ 0 on the set Q finer than
the quotient topology on Q such that the set map q : X → Q0 := ( Q, τ 0 ) is con-
tinuous. Then by property (QT2) of ( Q, q) in Definition 1.2.6, there is a unique
(continuous) map f : Q → Q0 such that f ◦ q = q. Since q : X → Q is surjective,
it admits a right inverse (set theoretically). This forces f : Q → Q0 to be the
identity map. This is not possible because f is continuous and the topology on
Q0 is finer than that of Q by our assumption (see Corollary 1.1.6). Hence the
result follows.

Remark 1.2.8. In Definition 1.2.6, the first condition suggests what should be
the underlying set of points of Q and the map q : X → Q, and the second
condition suggests what should be the topology on the set Q.

Theorem 1.2.9. Given a topological space X and an equivalence relation ∼ on X,

the associated quotient space ( Q, q) for ( X, ∼) exists, and is unique up to a unique
homeomorphism (i.e., if ( Q, q) and ( Q0 , q0 ) are two quotient spaces for ( X, ∼), then
there is a unique homeomorphism ϕ : Q → Q0 such that ϕ ◦ q = q0 ).

Proof. We first prove uniqueness of the pair ( Q, q), up to a unique homeomor-

phism. Let ( Q0 , q0 ) be another quotient space for the pair ( X, ∼). Since q0 is
8 Chapter 1. Review of Point Set Topology

continuous and q0 ( x ) = q0 (y) whenever x ∼ y in ( X, ∼), we have a unique

continuous map qe : Q0 → Q such that qe ◦ q0 = q.

X
q0 q0
q

(1.2.10)
x &
Q0 / Q / Q0
qe qe0

Similarly, interchanging the role of ( Q, q) and ( Q0 , q0 ) we get a unique contin-

uous map qe0 : Q → Q0 such that qe0 ◦ q = q0 . Then we have (qe0 ◦ qe) ◦ q0 = q0 .
Since the identity map IdQ0 : Q0 → Q0 is continuous and satisfies IdQ0 ◦q0 = q0 ,
we must have qe0 ◦ qe = IdQ0 . Similarly, we have qe ◦ qe0 = IdQ . Therefore, both
qe and qe0 are homeomorphisms. Thus the pair ( Q, q) is unique, up to a unique
homeomorphism.

Now (following Remark 1.2.8) we give an explicit construction of ( Q, ∼).

For each x ∈ X, the equivalence class of x in ( X, ∼) is the subset

[ x ] := { x 0 ∈ X : x ∼ x 0 } ⊆ X.

Let Q be the set of all distinct equivalence classes of elements of X. Consider

the map q : X → Q defined by sending each point x ∈ X to its equivalence class
[ x ] ∈ Q. Note that, the map q is surjective. As suggested in Corollary 1.2.7, we
define a topology on Q by declaring a subset U ⊆ Q to be open if its inverse
image q−1 (U ) ⊆ X is open in X. Clearly this makes q : X → Q continuous. It
remains to check property (QT2) as in Definition 1.2.6. Let Y be any topological
space and f : X → Y any continuous map satisfying f ( x ) = f ( x 0 ) for x ∼ x 0 in
( X, ∼). Define a map fe : Q → Y by fe([ x ]) = f ( x ), for all [ x ] ∈ Q. Clearly fe is
well-defined, and by its construction it satisfies

fe ◦ q = f . (1.2.11)

Since f is continuous, for any open subset V ⊆ Y, the subset

q−1 fe−1 (V ) = ( fe ◦ q)−1 (V ) = f −1 (V )

is open in X, and hence fe−1 (V ) is open in Q by definition of the topology on

Q. Therefore, fe is continuous. If g : Q → Y is any continuous map satisfying
g ◦ q = f , then g([ x ]) = ( g ◦ q)( x ) = f ( x ), for all [ x ] ∈ Q, and hence g = fe.
1.2. Quotient space 9

Theorem 1.2.12. Let X and Y be topological spaces, and let p : X → Y be a surjective

continuous map. Then the following are equivalent.

(i) The pair (Y, p) is a quotient space for some equivalence relation ∼ on X.

(ii) A subset U ⊆ Y is open in Y if and only if p−1 (U ) is open in X.

(iii) A subset Z ⊆ Y is closed in Y if and only if p−1 ( Z ) is closed in X.

Proof. (i) ⇒ (ii) and (i) ⇒ (iii): Follow from Corollary 1.2.7.

We show (ii) ⇒ (i) and (iii) ⇒ (i) together. Consider the equivalence relation
∼ on X defined by

x1 ∼ x2 in X, if p( x1 ) = p( x2 ).

Let ( Q, q) be the associated quotient space for the pair ( X, ∼). Note that,

Q = { p −1 ( y ) ⊆ X : y ∈ Y } ,

and the quotient map q : X → Q is given by

q( x ) = p−1 ( p( x )), ∀ x ∈ X.

Since p( x1 ) = p( x2 ) whenever x1 ∼ x2 in ( X, ∼), the universal property of the

quotient space ( Q, q) gives a unique continuous map

pe : Q → Y

such that pe ◦ q = p (see Theorem 1.2.9). In fact, in this case, the map pe can be
explicitly given as follow. Since p is surjective, given any element p−1 (y) ∈ Q,
there exists an element x ∈ p−1 (y) ⊆ X so that p( x ) = y. Then pe( p−1 (y)) =
pe(q( x )) = p( x ) = y. Then the map

φ:Y→Q
10 Chapter 1. Review of Point Set Topology

defined by φ(y) := p−1 (y), for all y ∈ Y, is the set theoretic inverse of the map
pe : Q → Y.
p
X /8 Y
pe
q

Qp φ

Therefore, to show pe a homeomorphism, it is enough to show that φ is continu-

ous. Let V ⊆ Q be open (resp., closed). Since φ−1 (V ) = {y ∈ Y : p−1 (y) ∈ V },
the subset

p−1 φ−1 (V ) = { x ∈ X : p( x ) ∈ φ−1 (V )}

= { x ∈ X : p−1 ( p( x )) ∈ V }
= { x ∈ X : q( x ) ∈ V }
= q −1 ( V )

is open (resp., closed) in X, because the quotient map q is continuous. Then

φ−1 (V ) is open (resp., closed) in Y by assumption (ii ) (reps., (iii )). Therefore,
φ is continuous as required.

Remark 1.2.13. It follows from construction of quotient space in Theorem 1.2.9,

and the proof of Theorem 1.2.12 that if f : X → Y is a quotient map then the
set of all fibers { f −1 (y) : y ∈ Y } of f gives a partition of X, and hence defines
an equivalence relation ∼ on X such that the associated quotient space X/ ∼ is
homeomorphic to (Y, f ).

Exercise 1.2.14. Let q : X → Q be a quotient map. Show that for any subset
Z ⊆ Q, the restriction map qq−1 (Z) : q−1 ( Z ) → Z is a quotient map. (Hint:

Use Theorem 1.2.12
).

Definition 1.2.15. A map f : X → Y is said to be open (resp., closed) if f (U ) is

open (resp., closed) in Y for any open (resp., closed) subset U of X.

Corollary 1.2.16. A surjective continuous open (or, closed) map is a quotient map.

Proof. Let f : X → Y be a surjective map. Then for any V ⊆ Y we have

f ( f −1 (V )) = V. Suppose that f is also continuous and open (resp., closed).
Then for any V ⊆ Y with f −1 (V ) open (resp., closed) in X, V = f ( f −1 (V )) is
open (resp., closed) in Y. Hence the result follows from Theorem 1.2.12.
1.2. Quotient space 11

Remark 1.2.17. Corollary 1.2.16 fails without continuity assumption on f . For

example, take a set X with at least two elements. Let τ0 and τ1 be the trivial
topology and the discrete topology on X, respectively. Then the identity map
IdX : ( X, τ0 ) → ( X, τ1 ) is a surjective open map, which is not continuous let
alone be a quotient map.
Corollary 1.2.18. Let f : X → Y be a continuous surjective map. If X is compact and
Y is Hausdorff, then f is a quotient map.

Proof. Let Z be a closed subset of X. Since X is compact, Z is compact. Since

f is continuous, f ( Z ) is a compact subset of Y. Since Y is Hausdorff, f ( Z ) is
closed in Y. Hence the result follows from Corollary 1.2.16.
Proposition 1.2.19. Let ∼ be an equivalence relation on a topological space X, and let
( Q, q) be the associated quotient space. Given a topological space Y, a map φ : Q → Y
is continuous if and only if the composite map φ ◦ q : X → Y is continuous.

φ◦q
X 8/ Y
q
 φ
Q

Proof. Since the quotient map q is continuous, the composite map φ ◦ q is con-
tinuous whenever φ is continuous. Conversely, let φ ◦ q be continuous. Since
for any open subset V ⊆ Y, we have q−1 (φ−1 (V )) = (φ ◦ q)−1 (V ) is open in
X, by construction of topology of Q, the subset φ−1 (V ) is open in Q. Thus φ is
continuous.
Example 1.2.20. (i) Cylinder: Let I = [0, 1] ⊂ R. Consider the unit square

I × I = {( x, y) ∈ R2 : 0 ≤ x, y ≤ 1}

in R2 . Define an equivalence relation ∼1 on I × I by setting

( x, y) ∼1 ( x 0 , y0 ), if x 0 = x + 1 = 1 and y = y0 .

This identifies points of two vertical sides of I × I (see Figure 1.1 below),
and the associated quotient space ( I × I )/ ∼1 is homeomorphic to the
cylinder

S1 × [0, 1] = {( x, y, z) ∈ R3 : x2 + y2 = 1, 0 ≤ z ≤ 1}.
12 Chapter 1. Review of Point Set Topology

F IGURE 1.1

Indeed, we can define a (continuous) map

φ : I × I → S1 × [0, 1]

by φ(s, t) = (exp(2πis), t), for all (s, t) ∈ I × I. Then the set {φ−1 (z, t) :
(z, t) ∈ S1 × [0, 1]} of all fibers of φ is precisely the partition of I × I given
by the equivalence relation ∼1 on I × I. It follows from Corollary 1.2.18
that φ is a quotient map, and by Remark 1.2.13 the associated quotient
space ( I × I )/ ∼ is homeomorphic to S1 × I.

(ii) Torus: Consider an equivalence relation ∼2 on the cylinder S1 × I defined

by
(z, t) ∼2 (z0 , t0 ) if z = z0 , and t0 = t + 1 = 1.

This identifies each point of the bottom circle of S1 × I with the corre-
sponding point of the top circle on S1 × I (see Figure 1.2 below).
Then the associated quotient space (S1 × I )/ ∼2 is homeomorphic to the
torus T := S1 × S1 in R3 . Indeed, we can define a (continuous) map

ψ : S1 × I → S1 × S1

by ψ(z, t) = (z, exp(2πit)), for all (z, t) ∈ S1 × I. As before, it is easy

to see that the set of all fibers of the map ψ is precisely the partition of
S1 × I defined by the equivalence relation ∼2 on the cylinder S1 × I. As
before, it follows from Corollary 1.2.18 that ψ is a quotient map, and by
1.2. Quotient space 13

F IGURE 1.2

Remark 1.2.13 the associated quotient space (S1 × I )/ ∼ is homeomorphic

to S1 × S1 .

(iii) Cone: Let I = [0, 1] ⊆ R. The cone of a topological space X is the quotient
space CX := ( X × I )/ ∼ of X × I for the equivalence relation ∼ on X × I
defined by
( x, t) ∼ ( x 0 , t0 ), if t = t0 = 1. (1.2.21)

The associated set of all partitions of X × I is the set


X × {1}, {( x, t)} : x ∈ X, 0 ≤ t < 1 .

Thus we identify all points of X × {1} ⊆ X × I into a single point, called

F IGURE 1.3

the vertex of the cone CX, and the remaining points of X × [0, 1) remains
as they are.
14 Chapter 1. Review of Point Set Topology

If X is a compact subset of an Euclidean space Rn , then we can con-

struct CX more geometrically as follow. Embed Rn into Rn+1 by the map
( x1 , . . . , xn ) 7→ ( x1 , . . . , xn , 0), and fix a point v ∈ Rn+1 which lies outside
the image of this embedding; for example take v = (0, . . . , 0, 1) ∈ Rn+1 .
Note that `[v, x] := {tv + (1 − t) x : 0 ≤ t ≤ 1} ⊆ Rn+1 is the straight line
segment in Rn+1 joining v and x ∈ X. The subset

`[v, x] ⊆ Rn+1
[

x∈X

with the subspace topology induced from Rn+1 is called the geometric cone
of X. We show that the geometric cone of X is homeomorphic to the cone
of X, i.e.,
CX ∼
[
= `[v, x] .
x∈X

Define a map
[
f : X×I → `[v, x]
x∈X

by f ( x, t) = tv + (1 − t) x, for all ( x, t) ∈ X × I. Clearly f is a surjective

continuous map, and f ( x, t) = f ( x, t0 ) if and only if either x = x 0 and
t = t0 , or t = t0 = 1. Since X is compact and its image is Hausdorff (being
a subspace of Rn+1 ), it follows from Corollary 1.2.18 that f is a quotient
map. Since the fibers of the map f are precisely the equivalence classes
for the equivalence relation on X × I defined in (1.2.21), it follows from
Remark 1.2.13 that CX = ( X × I )/ ∼ is homeomorphic to the geometric
cone of X.

(iv) The space X/A: Let A be a subset of a topological space X. Define an

equivalence relation ∼ on X by

x ∼ x 0 if both x and x 0 are in A.

We denote by X/A the associated quotient space X/∼. Here we collapse

the subspace A into a single point, and the remaining points of X \ A
remains as they were. For example, CX = ( X × I )/( X × {1}).
1.2. Quotient space 15

(v) The space Bn /Sn−1 : Consider the closed unit ball

n
n
B := {( x1 , . . . , xn ) ∈ R : n
∑ x2j ≤ 1}
j =1

in Rn , and its boundary

n
∂Bn = {( x1 , . . . , xn ) ∈ Bn : ∑ x2j = 1} = Sn−1.
j =1

Then the associated quotient space is denoted by Bn /Sn−1 is homeomor-

phic to Sn . This is quite easy to visualize for n = 1 and 2. For n = 1,
B1 = [−1, 1] ⊆ R, and S0 = {−1, 1} is its boundary. If we identify all
points of S0 = {−1, 1} into a single point and keep all other points of B1
as they were, we get a circle S1 in R2 ; see Figure 1.4 below.

F IGURE 1.4

The case n = 2 is explained in the Figure 1.5 below.

F IGURE 1.5

In general, it suffices to construct a surjective continuous map

f : Bn → Sn
16 Chapter 1. Review of Point Set Topology

such that f  Bn \Sn−1 is injective and f (Sn−1 ) is a singleton subset of Sn .



Then by Corollary 1.2.18, f become a quotient map producing a home-

omorphism of Bn /Sn−1 onto Sn . To construct such a map f , note that Rn
is homeomorphic to Bn \ Sn−1 and Sn \ { p}, for any p ∈ Sn . Fix two home-
omorphisms h1 : Bn \ Sn−1 → Rn and h2 : Rn → Sn \ { p}, and define
(
h2 (h1 ( x )), if x ∈ Bn \ Sn−1 ,
f ( x ) := (1.2.22)
p, if x ∈ Sn−1 .

It is easy to check that f has desired properties (verify).

Example 1.2.23 (Attaching spaces along a map). Let X and Y be two topological
spaces. Suppose we wish to attach X by identifying points of a subspace A ⊆ X
with points of Y in a continuous way. This can be done by using a continuous
map f : A → Y. Indeed, we identify x ∈ A with its image f ( x ) ∈ Y. This
defines an equivalence relation on X t Y, and we denote the associated quotient
space by X t f Y, and call it the space Y with X attached along A via f . Let us
discuss some examples.

(i) Let I = [0, 1] ⊂ R. Given a space X, we call X × I the cylinder over X. Let
f : X → Y be a continuous map of topological spaces. Then f induces a
continuous map fe : X × {0} → Y given by

fe( x, 0) = f ( x ), ∀ ( x, 0) ∈ X × {0}.

If we attach Y with the cylinder X × I of X along its base X × {0} ⊂ X × I

via the map fe, by identifying ( x, 0) ∼ f ( x ), then the associated quotient
space M f = ( X × I ) t fe Y is called the mapping cylinder of f (see Figure
1.6).

F IGURE 1.6: Mapping Cylinder

(ii) Let CX = ( X × I )/( X × {1}) be the cone over X obtained by collapsing

the subspace X × {1} of the cylinder X × I over X to a single point. Let
1.2. Quotient space 17

f : X → Y be a continuous map. If we attach this cone CX with Y along its

base X × {0} ⊂ CX by identifying ( x, 0) ∈ X × {0} with f ( x ) ∈ Y, then
the resulting quotient space C f = Y t f CX is called the mapping cone of f
(see Figure 1.7). Note that, the mapping cone C f can also be obtained as a

F IGURE 1.7: Mapping Cone

quotient space of the mapping cylinder M f by collapsing X × {1} ⊂ M f

to a point.

(iii) Let X be a topological space. The suspension SX of X is the quotient space

of X × I obtained by collapsing X × {0} to one point and X × {1} to an-
other point.

F IGURE 1.8: Suspension

For example, if we take X to be the unit circle S1 = {( x, y) ∈ R2 : x2 +

y2 = 1}, then X × I is a cylinder C = {( x, y, z) ∈ R3 : x2 + y2 = 1, 0 ≤
z ≤ 1}, and then we collapse two circular edges of C to two points to
get SX, which is homeomorphic to the 2-sphere S2 = {( x, y, z) ∈ R3 :
x 2 + y2 + z2 = 1}.
We can think of SX as a double cone on X: take disjoint union of two cones
C1 := ( X × I )/( X × {1}) and C2 := ( X × I )/( X × {0}), and then attach
C1 with C2 via the continuous map f : X × {0} → C2 given by f ( x, 0) =
( x, 0) ∈ C2 , ∀ ( x, 0) ∈ X × {0} ⊂ C1 .

The following exercise shows that a quotient of a Hausdorff space need not
be Hausdorff in general.
18 Chapter 1. Review of Point Set Topology

Exercise 1.2.24. Consider the double real line X = R × {0, 1} ⊂ R2 , and the
equivalence relation ∼ on X defined by (t, 0) ∼ (t, 1), for all t ∈ R \ {0}. The
associated quotient space X/ ∼ is called the real line with double origin. Show
that X/∼ is not Hausdorff.

Proposition 1.2.25. Let ρ ⊆ X × X be an equivalence relation on a topological space

X, and let q : X → Q := X/ρ be the quotient map. Then we have the following.

(i) Q is a T1 space if and only if every ρ-equivalence class is closed in X.

(ii) If Q is Hausdorff then ρ is a closed subspace of the product space X × X. The

converse holds if q : X → Q is an open map.

Proof. (i) Let Q be T1. Let x ∈ X. Choose a y ∈ X \ [ x ]. Then [ x ] 6= [y] in Q.

Since Q is T1, there is an open subset Vy ⊆ Q such that [y] ∈ Vy and [ x ] ∈ /
Vy . Then q (Vy ) is an open neighbourhood of y with q (Vy ) ∩ [ x ] = ∅.
− 1 − 1

Therefore, y is an interior point of X \ [ x ]. Therefore, X \ [ x ] is open, and

hence [ x ] ⊆ X is closed.
Conversely, suppose that [ x ] ⊆ X is closed, for all x ∈ X. To show Q
is T1, we need to show that {[ x ]} is closed in Q, for all x ∈ X. Since
q−1 ( Q \ {[ x ]}) = {y ∈ X : q(y) 6= [ x ]} = X \ [ x ] is open, Q \ {[ x ]} is open
in Q, for all [ x ] ∈ Q. Therefore, Q is a T1 space.

(ii) Consider the commutative diagram of continuous maps

q
X / Q
∆X ∆Q
 q×q 
X×X / Q × Q,

where q × q : X × X → Q × Q is the product map given by

(q × q)( x, y) = (q( x ), q(y)), ∀ ( x, y) ∈ X × X.

Note that, (q × q)−1 (∆Q ( Q)) = {( x, y) ∈ X × X : q( x ) = q(y)} = ρ.

If Q is Hausdorff, then ∆Q ( Q) is closed by Exercise 1.1.8. Since q × q is
continuous, ρ is closed in X × X.
Now we assume that q is an open map, and that ρ is closed in X × X.
Since q × q is a continuous surjective open map (verify!), it is a quotient
1.2. Quotient space 19

map by Corollary 1.2.16. Since (q × q)−1 (∆Q ( Q)) = ρ is closed in X × X,

the diagonal ∆Q ( Q) is closed in Q × Q by Theorem 1.2.12 (iii). Therefore,
Q is Hausdorff by Exercise 1.1.8.

Now we give an example to show that even if X is Hausdorff and the equiv-
alence relation ρ is closed in X × X, the associated quotient space Q = X/ρ need
not be Hausdorff without the assumption that q is an open map. For this, we
first recall the following.

Definition 1.2.26. A topological space X is said to be normal if any two disjoint

closed subsets can be separated by a pair of disjoint open subsets containing
them. In other words, given two closed subsets A, B ⊂ X with A ∩ B = ∅,
there are open subsets U, V ⊂ X such that A ⊂ U, B ⊂ V and U ∩ V = ∅.

Most of the familiar examples of topological spaces are generally normal

(e.g., Rn ), and a closed subspace of a normal space is normal. The following
example shows that a Hausdorff space need not be normal.

Example 1.2.27. Let K = { n1 : n ∈ Z+ }. Consider the topology τK on R whose

basis for open subsets is given by the collection

B = {( a, b) : a, b ∈ R with a < b} ∪ {( a, b) \ K : a, b ∈ R with a < b}.

Clearly this topology on R is strictly finer than the Euclidean topology on R,

and hence (R, τK ) is a Hausdorff space. Note that in this topology, K and {0}
are disjoint closed subsets that cannot be separated by a pair of disjoint open
subsets containing them. Therefore, (R, τK ) is not normal.

Example 1.2.28. Start with a Hausdorff space X that is not normal. Choose two
disjoint closed subsets A, B ⊂ X that cannot be separated by two disjoint open
subsets containing them. Take ρ = ∆ X ( X ) ∪ ( A × A) ∪ ( B × B). Note that ρ
is an equivalence relation on X, and is closed in X × X (why?). Show that the
associated quotient space X/ρ is not Hausdorff. In this example, can you think
of an open subset of X whose image under the quotient map is not open?)
20 Chapter 1. Review of Point Set Topology

1.3 Projective space and Grassmannian

1.3.1 Real and complex projective spaces

Fix an integer n ≥ 0. Define an equivalence relation ∼ on Rn+1 \ {0} by

v ∼ v0 if v0 = λ · v, for some λ ∈ R \ {0}. (1.3.1)

In other words, identify all points lying on the same straight-line in Rn+1 pass-
ing through the origin 0 ∈ Rn+1 . Then the associated quotient space

RPn := (Rn+1 \ {0})/ ∼

is called the real projective n-space. As a set, RPn consists of all straight-lines in
Rn+1 passing through the origin 0 ∈ Rn+1 . So an element of RPn is of the form

[ a 0 : · · · : a n ] : = λ · ( a 0 , . . . , a n ) ∈ Rn +1 \ { 0 } : λ ∈ R \ { 0 } .

(1.3.2)

Let q : Rn+1 \ {0} → RPn be the quotient map for the projective n-space. Note
n
that the unit n-sphere Sn = {( a0 , . . . , an ) ∈ Rn+1 : ∑ a2j = 1} is a compact
j =0
connected subspace of R n + 1 \ {0}. Since the restriction map qSn : Sn −→ RPn


is continuous and surjective, RPn is compact and connected.

Exercise 1.3.3. Prove the following.

(i) Show that the map f := qSn : Sn → RPn is a quotient map.



(ii) For each ` ∈ RPn , show that f −1 (`) = {v, −v}, for some v ∈ Sn .

(iii) Show that RPn is Hausdorff.

Outline of solution. Note that, the quotient map q : Rn+1 \ {0} → RPn is given
by sending ( a0 , . . . , an ) ∈ Rn+1 \ {0} to the straight line

[ a0 : · · · : an ] := {λ( a0 , . . . , an ) : λ ∈ R} ∈ RPn .

Since RPn consists of all straight lines in Rn+1 passing through the origin, given
a straight-line ` ∈ RPn , choosing any non-zero point v := ( a0 , . . . , an ) ∈ `, we
1.3. Projective space and Grassmannian 21

n
1/2
find an element v/||v|| ∈ Sn with f (v/||v||) = `, where ||v|| := ∑ a2j . Thus,
j =1
f is surjective.
ι q
f : Sn ,→ Rn+1 \ {0} −→ RPn .

Note that given any subset V ⊆ RPn , we have f −1 (V ) = q−1 (V ) ∩ Sn . So con-

tinuity of f follows from that of q. To see that f is a quotient map, suppose that
f −1 (V ) is open in Sn . To show that V is open in RPn , fix a point ` ∈ V. Its fiber
f −1 (`) = {v, −v} consists of the two antipodal points of Sn obtained by inter-
secting the line ` with Sn . Since the points v and −v lies on two hemispheres
separated by a great circle on Sn , we can find a small enough (connected) open
neighbourhood U ⊂ f −1 (V ) of v such that −U := {−u : u ∈ U } ⊂ Sn is
an open neighbourhood of −v in Sn , and U ∩ (−U ) = ∅. Note that, −U ⊆
f −1 (V ). Then f U is a homeomorphism of U onto the open neighbourhood


f (U ) ⊆ V of ` in RPn . Thus f is a quotient map.

Next we show that RPn can be covered by n + 1 open subsets each homeo-
morphic to Rn . Let p j : Rn+1 → R be the j-th projection map defined by

p j ( x 0 , . . . , x n ) = x j , ∀ ( x 0 , . . . , x n ) ∈ Rn +1 .

For each j ∈ {0, 1, . . . , n}, consider the hyperplane

Hj := {[ a0 : · · · : an ] ∈ RPn : a j = 0} ⊂ RPn .

Since q is a quotient map and

q−1 ( Hj ) = {( a0 , . . . , an ) ∈ Rn+1 \ {0} : a j = 0}

= p−
j (0) ∩ (R
1 n +1
\ {0}),

we conclude that Hj is a closed subset of RPn . Let Uj := Pn \ Hj , ∀ j =

0, 1, . . . , n. Since any point of RPn is of the form

[ a 0 : · · · : a n ] : = { λ ( a 0 , . . . , a n ) ∈ Rn +1 \ { 0 } : λ ∈ R } ,

with a j 6= 0, for some j, we see that {U0 , U1 , . . . , Un } is an open cover of RPn .

Proposition 1.3.4. The open subset Uj ⊂ RPn is homeomorphic to Rn , for all j.

22 Chapter 1. Review of Point Set Topology

Proof. Consider the map φj : Uj −→ Rn given by

!
φj a1 a j −1 a j +1 an
[ a0 : · · · : an ] 7−→ ,..., , ,..., .
aj aj aj aj

Note that φj is a well-defined bijective map with its inverse ψj : Rn → Uj given

by
 
(b0 , . . . , bn−1 ) 7→ b0 : · · · : b j−1 : 1 : b j : · · · : bn .

Note that Vj = q−1 (Uj ) = {( a0 , . . . , an ) ∈ Rn+1 : a j 6= 0} is open in Rn+1 \ {0},

fj  a a j −1 a j +1

an
and the map f j : Vj → R given by ( a0 , . . . , an ) 7−→ a , . . . , a , a , . . . , a is
n 0
j j j j

continuous. Since q −1 ( φ − 1
j (V ))
= f j−1 (V ),
∀V⊆ Rn ,
and q is a quotient map,
we conclude that φj is continuous, for all j (c.f. Proposition 1.2.19).

fj
Rn +1 \ { 0 } o ? _V /
9R
n
j
φj
q qj
 
RPn o ? _U q ψj
j

Since f j is a quotient map by Corollary 1.2.16, as before we see that ψj = φ− 1

j is
also continuous. This completes the proof.

Corollary 1.3.5. RPn is a compact connected Hausdorff space.

Exercise 1.3.6. Define an equivalence relation ∼ on Sn by

v ∼ v0 if v0 = −v.

Show that the associated quotient space Sn / ∼ is homeomorphic to RPn . Con-

clude that RPn is a compact connected Hausdorff space.

The complex projective n-space CPn is the quotient space of Cn+1 \ {0} under
the equivalence relation ∼ defined by

v ∼ v0 if v0 = λv, for some λ ∈ C.

So the points of CPn are precisely one dimensional C-linear subspaces of Cn+1
(i.e., complex lines in Cn+1 passing through the origin 0 ∈ Cn+1 ).

Exercise 1.3.7. Show that CPn is a compact connected Hausdorff space.

1.3. Projective space and Grassmannian 23

Remark on notations: The real projective n-space RPn is also denoted by PR

n and

Pn (R). Similar notations PC n and Pn (C) are also used for complex projective

n-space CPn .

1.3.2 Grassmannian Gr(k, Rn )

Fix two positive integers k and n, with k < n. Let

(Rn ) k : = R
|
n
· · × Rn}
× ·{z
k-times

be k-fold product of Rn together with the product topology. A typical element

of (Rn )k is of the form (v1 , . . . , vk ), where v j = ( a j1 , . . . , a jn ) ∈ Rn , for all j =
1, . . . , k. Note that, we can identify (Rn )k with Mk, n (R) using the bijective map
 
a11 . . . a1n
 . .
. . ...  .

(v1 , . . . , vk ) 7−→  .
 . 
ak1 . . . akn

Consider the subset

X := {(v1 , . . . , vk ) ∈ (Rn )k  {v1 , . . . , vk } is R-linearly independent}



with the subspace topology induced from (Rn )k . Given A := (v1 , . . . , vk ) and
A0 := (v10 , . . . , v0k ) in X, we define A ∼ A0 if

SpanR {v1 , . . . , vk } = SpanR {v10 , . . . , v0k }.

Clearly ∼ is an equivalence relation on X. The associated quotient topological

space X/ ∼ is known as the Grassmannian of k-dimensional R-linear subspaces of
Rn , and is denoted by Gr(k, Rn ). As a set, Gr(k, Rn ) consists of all R-linear
subspaces of Rn of dimension k.

Corollary 1.3.8. Gr(1, Rn ) is homeomorphic to RPn−1 .

Remark 1.3.9 (Plücker embedding). Given a n-dimensional R-vector space V,

its k-th exterior power k V is a R-vector space of dimension (nk). Sending
V

W ∈ Gr(k, Rn ) to its k-th exterior power k W ⊂ k Rn , we get a continuous

V V
24 Chapter 1. Review of Point Set Topology

map
Φ : Gr(k, Rn ) −→ RP N ,

where N = (nk) − 1. It turns out that Φ is a closed embedding (homeomorphism

onto a closed subspace of RP N ). From this, one can conclude that Gr(k, Rn ) is
a compact Hausdorff space. We shall not go into detailed proofs of the above
statements.

1.4 Topological group

Definition 1.4.1. A topological group is a topological space G which is also a

group G such that the binary map (group operation)

m : G × G → G, ( x, y) 7→ xy,

and the inversion map

inv : G → G, x 7→ x −1 ,

involved in its group structure, are continuous. Here we consider G × G as the

product topological space.

We recast the above definition of topological group in more formal lan-

guage, without using points of G. This formalism, with appropriate type of
spaces and maps between them, defines Lie group, algebraic group, group-scheme
and more generally, a group object in a category (for curious readers!). Denote
by ∗ the topological space whose underlying set is singleton. This space is
unique up to a unique homeomorphism. Given any topological space X, any
map ∗ → X is continuous, and they are in bijection with the underlying set of
points of X. On the other hand, the space ∗ is the final object in the category
of topological spaces in the sense that, given any topological space X, there is
a unique continuous map X → ∗. Clearly the product space X × ∗ is home-
omorphic to X, and the set of all such homeomorphisms are in bijection with
the set of all automorphisms of X (i.e., homeomorphisms of X onto itself). Unless
explicitly specified, we consider the homeomorphism X × ∗ → X given by the
identity map IdX : X → X of X.
1.4. Topological group 25

Now the above Definition 1.4.1 essentially says that, a topological group is a
pair ( G, m), where G is a topological space and m : G × G → G is a continuous
map such that the following axioms holds.

(TG1) Associativity: The following diagram is commutative.

m×IdG
G×G×G / G×G
IdG ×m m
 
m /
G×G G

(TG2) Existence of neutral element: There is a continuous map e : ∗ → G such

that the following diagram is commutative.

e×IdG IdG ×e
∗×G / G×Go G×∗
∼ m ∼
=  =
( v
G

(TG3) Existence of inverse: There is a continuous map inv : G → G such that

the following diagram is commutative.

(IdG , inv) (inv, IdG )

G / G×Go G
m
  
e / e
∗ Go ∗
Example 1.4.2. (i) Any abstract group is a topological group with respect to
the discrete topology on it.

(ii) (R, +), the real line with usual addition of real numbers, is a topological
group.

(iii) (R∗ , ·), the subspace of non-zero real numbers with usual multiplication
is a topological group.

(iv) (Z, +) is a topological group, where the topology on Z is discrete.

(v) For any integer n ≥ 1, the Euclidean space Rn with the component wise
addition, i.e.,

( a1 , . . . , an ) + (b1 , . . . , bn ) := ( a1 + b1 , . . . , an + bn ), ∀ ai , bi ∈ R,
26 Chapter 1. Review of Point Set Topology

is a topological group.

(vi) Given integers m, n ≥ 1, the set of all (m × n)-matrices with real entries
Mm,n (R), considered as the Euclidean topological space Rmn , is a topolog-
ical group with respect to the usual matrix addition.

(vii) GLn (R), the subspace of all invertible (n × n)-matrices with real entries,
is a topological group with respect to multiplication of matrices.

(viii) Circle group: The space S1 = {z ∈ C : |z| = 1} ⊂ C, together with the

multiplication of complex numbers, is a topological group.

(ix) Any abstract subgroup of a topological group is a topological group with

respect to the subspace topology.

(x) Product of two topological groups is a topological group.

Exercise 1.4.3. Let G be a topological group. If U ⊆ G is an open neighbour-

hood of identity e ∈ G, show that there is an open neighbourhood V ⊂ G of
identity such that V 2 := { ab : a, b ∈ V } ⊆ U. (Hint: Use continuity of the
multiplication map m.)

Exercise 1.4.4. Show that for any a ∈ G, the right translation by a map

R a : G → G, g 7→ ga,

is a homeomorphism. Prove the same statement for the left translation by a map
given by L a ( g) = ga, for all g ∈ G. (Hint: Note that R a is the composite map
m
g 7→ ( g, a) 7→ ga with inverse R a−1 .)

Exercise 1.4.5. Show that a topological group G is Hausdorff if and only if it is

a T1 space. (Hint: ∆G ( G ) is precisely the inverse image of {e} ⊆ G under the
map ( x, y) 7→ x −1 y.)

Lemma 1.4.6. Let G be a topological group. Let H be the connected component of G

containing the neutral element e ∈ G. Then H is a closed normal subgroup of G.

Proof. Since connected components are closed, H is closed. Since for any a ∈ H,
the set Ha−1 = {ha−1 : h ∈ H } = R a−1 ( H ) contains e, and is homeomorphic
to H, we must have Ha−1 ⊆ H. Since this holds for all a ∈ H, we see that
H is a subgroup of G. To see that H is normal, note that, for any g ∈ G, the
1.4. Topological group 27

set gHg−1 = L g ( R g−1 ( H )) is a connected subset of G containing e, and hence

gHg−1 ⊆ H. This completes the proof.
Definition 1.4.7. A right action of a topological group G on a topological space X
is a continuous map σ : X × G → X such that σ( x, e) = x, and σ(σ( x, g1 ), g2 ) =
σ ( x, m( g1 , g2 )), for all x ∈ X and g1 , g2 ∈ G, where m : G × G → G is the
product operation (multiplication map) on G. Similarly, one can left action of
G on X.

Without using points, a right G-action σ on X can be defined by commuta-

tivity of the following diagrams.

(i)
IdX ×e
X×∗ / X×G
σ
∼
= 
(
X

(ii)
σ×IdG
X×G×G / X×G
IdX ×m σ
 
X×G
σ / X

A right G-action σ on X induces an equivalence relation on X, which gives a

partition of X as a disjoint union of equivalence classes. A typical equivalence
class is of the form

orbG ( x ) := { x 0 ∈ X : x 0 = xg, for some g ∈ G } = xG,

and is called the G-orbit of x ∈ X. The associated quotient space, denoted by

X/σ or X/G, consists of all G-orbits of elements of X as its points. For this
reason, X/G is also called orbit space. If the G-action on X is transitive (i.e.,
given any x, x 0 ∈ X, there exists g ∈ G such that x 0 = xg), then X is called a
homogeneous space. In this case, the associated quotient space X/G is singleton.
Exercise 1.4.8. Given a subgroup H of a topological group G, the H-action on
G defined by
G × H 7→ G, ( g, h) 7→ gh

gives a partition of G into all right cosets of H in G. Show that the orbit space
G/H is a homogeneous space.
28 Chapter 1. Review of Point Set Topology

Exercise 1.4.9. Let I = [0, 1] ⊂ R. Define the Z2 -action on I × I which gives

identifications (0, t) ∼ (1, 1 − t), for each t ∈ I. Convince yourself that the
associated quotient space is homeomorphic to the Möbius strip (see Figure 1.9).
Note that, Möbius strip has only one side!

F IGURE 1.9: Möbius strip

Exercise 1.4.10. Define a Z2 -action on the n-sphere Sn ⊂ Rn+1 which identifies

v ∈ Sn with its antipodal point −v ∈ Sn . Show that the associated quotient space
Sn /Z2 is homeomorphic to RPn .

Exercise 1.4.11. Show that R/Q is a non-Hausdorff topological group. Hint:

Note that Q is dense in R, and right cosets are just translates of Q.

Exercise 1.4.12. Let σ : X × G → X be a right action of a topological group on

a space X. For each g ∈ G, show that the induced map

σg : X → X, x 7→ xg := σ ( x, g)

is a homeomorphism.

Exercise 1.4.13. Let X be a topological space together with an action of a topo-

logical group G. Show that the quotient map q : X → X/G is open. (Hint: For
V ⊆ X open, show that q−1 (q(V )) =
S
Vg is open by Exercise 1.4.12, where
g∈ G
Vg = {vg : v ∈ V }, ∀ g ∈ G.)

Proposition 1.4.14. Let H be a subgroup of a topological group G. Then the orbit

space G/H is Hausdorff if and only if H is closed in G. (Here G/H is not necessarily
a group because H need not be a normal subgroup of G.)

Proof. If G/H is Hausdorff, then it is a T1 space so that H = orb H (e) ∈ G/H

is a closed point. Since H is the inverse image of this point under the quotient
map q : G → G/H (continuous), H is closed in G. Conversely, suppose that
1.4. Topological group 29

H is closed in G. Since the equivalence relation given by the H-action on G is

precisely the inverse image of H under the continuous map

G × G −→ G, ( g1 , g2 ) 7→ g1−1 g2 ,

and the quotient map q : G → G/H is open by Exercise 1.4.13, the converse
part follows from Proposition 1.2.25 because H is closed in G.

Corollary 1.4.15. The topological group R/Q is not Hausdorff.

Definition 1.4.16. A homomorphism of topological groups is a continuous group

homomorphism. An isomorphism of topological groups is a bijective bi-continuous
homomorphism of topological groups.

Exercise 1.4.17. If f : G → H is a homomorphism of topological groups with

H Hausdorff, show that Ker( f ) := { g ∈ G : f ( g) = e H } is a closed normal
subgroup of G.

Exercise 1.4.18. If f : G → H is a homomorphism of topological groups, show

that the induced map G/Ker( f ) → Im( f ) is an isomorphism of topological
groups.

Exercise 1.4.19. Show that f : R → S1 defined by f (t) = e2πit , for all t ∈ R, is

a surjective homomorphism of topological groups. Use Exercise 1.4.18 to show
that R/Z ∼ = S1 as topological groups.

Exercise 1.4.20. Let f : G → H be a continuous bijective homomorphism of

topological groups. Show that f −1 : H → G is continuous (Hint: Use Exercise
1.4.13).

Corollary 1.4.21. A bijective homomorphism f : G → H of topological groups is an

isomorphism.

Exercise 1.4.22. Consider the Z-action on R given by σ (t, n) = t + n, for all t ∈

R and n ∈ Z. Show that the associated quotient space R/σ is homeomorphic
to S1 .

Exercise 1.4.23. Show that GLn (R)/ SLn (R) ∼

= R∗ as topological groups.

Exercise 1.4.24. Show that GLn (R) is disconnected, and has precisely two con-
nected components, whereas GLn (C) is path-connected. (Hint: For the first
part, use determinant map. For the second part, given A ∈ GLn (C) use left
30 Chapter 1. Review of Point Set Topology

and right translation homeomorphisms to move it to an upper triangular ma-

trix, and then use convex combination map for its entries to move it to the
identity matrix in GLn (C).)

Exercise∗ 1.4.25. Show that the group

SOn = { A ∈ GLn (R) : AAt = At A = In and det( A) = 1}

is compact and connected.

 31

Chapter 2

Homotopy Theory

2.1 Homotopy of maps

Let I be the closed interval [0, 1] ⊂ R. Let X and Y be topological spaces.

Definition 2.1.1. Let f 0 , f 1 : X → Y be continuous maps. We say that f 0 is

homotopic to f 1 , written as f 0 ' f 1 , if there is a continuous map

F : X × I −→ Y

such that F ( x, 0) = f 0 ( x ), ∀ x ∈ X, and F ( x, 1) = f 1 ( x ), ∀ x ∈ X. In this case,

the continuous map F is called the homotopy from f 0 to f 1 . A continuous map
f : X → Y is said to be null homotopic if f is homotopic to a constant map from
X into Y.

f1

f0 Y

F IGURE 2.1: Homotopy

Example 2.1.2. 1. Let X be a space. Then any two continuous maps f , g :

X → R2 are homotopic. To see this, note that the map F : X × I → R2
defined by

F ( x, t) = (1 − t) f ( x ) + tg( x ), ∀ ( x, t) ∈ X × I,
32 Chapter 2. Homotopy Theory

is continuous and satisfies F ( x, 0) = f ( x ) and F ( x, 1) = g( x ), for all x ∈

X. Thus F is a homotopy from f to g; such a homotopy is called a straight-
line homotopy, because for each x ∈ X, it movies f ( x ) to g( x ) along the
straight-line segment joining them.

x
f0

f1

t
0 1
F IGURE 2.2: Example of a straight-line homotopy

Before proceeding further, let us recall the following useful result from basic
topology course, that we need frequently in this course.
Lemma 2.1.3 (Joining continuous maps). Let A and B be two closed subsets of
topological space X such that X = A ∪ B. Let Y be any topological space. Let f : A →
Y and g : B → Y be continuous maps such that f ( x ) = g( x ), for all x ∈ A ∩ B. Then
the function h : X → Y defined by
(
f ( x ), if x ∈ A,
h( x ) =
g( x ), if x ∈ B,

is continuous.

Proof. Let Z ⊆ Y be a closed subset. It is enough to check that h−1 ( Z ) is closed

in X. Note that

h −1 ( Z ) = h −1 ( Z ) ∩ A h −1 ( Z ) ∩ B

[

= f −1 ( Z ) g −1 ( Z ).
[

Since f and g are continuous, f −1 ( Z ) is closed in A and g−1 ( Z ) is closed in B.

Since A and B are closed in X, both f −1 ( Z ) and g−1 ( Z ) are closed in X, and so
is their union h−1 ( Z ). This completes the proof.
2.1. Homotopy of maps 33

Lemma 2.1.4. The relation “being homotopic maps” is an equivalence relation on the
set C ( X, Y ) of all continuous maps from X into Y.

Proof. For any f ∈ C ( X, Y ), taking

F : X × I → Y, ( x, t) 7→ f ( x )

we see that f is homotopic to itself, and hence “being homotopic maps” is a

reflexive relation. Let f 0 , f 1 ∈ C ( X, Y ) be such that f 0 is homotopic to f 1 with
homotopy F. Then the continuous map

G : X × I → Y, ( x, t) 7→ F ( x, 1 − t)

is a homotopy from f 1 to f 0 . So “being homotopic maps” is a symmetric re-

lation. Let f 0 , f 1 , f 2 ∈ C ( X, Y ) be such that f 0 ' f 1 with a homotopy F, and
f 1 ' f 2 with a homotopy G. Consider the map H : X × I → Y defined by
(
F ( x, 2t), if t ∈ [0, 12 ],
H ( x, t) :=
G ( x, 2t − 1), if t ∈ [ 12 , 1].

Since at t = 21 , we have F ( x, 2t) = F ( x, 1) = f 1 ( x ) = G ( x, 0) = G ( x, 2t − 1),

for all x ∈ X, we see that H is a well-defined continuous map (c.f., Lemma
2.1.3). Clearly H satisfies H ( x, 0) = f 0 ( x ) and H ( x, 1) = f 2 ( x ), for all x ∈ X.
Therefore, H is a homotopy from f 0 to f 2 . Thus “being homotopic maps” is a
transitive relation, and hence is an equivalence relation on C ( X, Y ).

Exercise 2.1.5. Let f , g ∈ C ( X, Y ), and F : X × I → Y be a homotopy from f to g.

Use F to construct a homotopy G from f to g with G 6= F. Therefore, homotopy
between two maps need not be unique. (Hint: take G ( x, t) = F ( x, t2 )).

Lemma 2.1.6. Let f 0 , f 1 : X → Y be two continuous maps such that f 0 is homotopic

to f 1 . Then for any spaces Z and W, and continuous maps g : Z → X and h : Y → W,
we have f 0 ◦ g ' f 1 ◦ g and h ◦ f 0 ' h ◦ f 1 .

f0
g
/ & h /
Z X 8Y W.
f1
34 Chapter 2. Homotopy Theory

Proof. Let F : X × I → Y be a continuous map such that F ( x, 0) = f 0 ( x ) and

F ( x, 1) = f 1 ( x ), for all x ∈ X. Define G : Z × I → Y by setting

G (z, t) = F ( g(z), t), ∀ (z, t) ∈ Z × I.

Clearly G is a continuous function with G (z, 0) = F ( g(z), 0) = ( f 0 ◦ g)(z), and

G (z, 1) = F ( g(z), 1) = ( f 1 ◦ g)(z), for all z ∈ Z. Therefore, G gives a homotopy
f 0 ◦ g ' f 1 ◦ g. Similarly, taking

H : X × I → W, ( x, t) 7→ h( F ( x, t)),

we see that H is a continuous map satisfying H ( x, 0) = h( F ( x, 0)) = (h ◦ f 0 )( x )

and H ( x, 1) = h( F ( x, 0)) = (h ◦ f 1 )( x ), for all x ∈ X. Therefore, H gives a
homotopy h ◦ f 0 ' h ◦ f 1 .

Definition 2.1.7. Let f 0 , f 1 : ( X, x0 ) → (Y, y0 ) be continuous maps of pointed

topological spaces. A homotopy from f 0 to f 1 is a continuous map F : X × I → Y
such that

(i) F ( x, 0) = f 0 ( x ), ∀ x ∈ X,

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

(iii) F ( x0 , t) = y0 , ∀ t ∈ [0, 1].

When we talk about homotopy of continuous maps of pointed topological

spaces, we always mean that the homotopy preserve the marked points in the
sense of (iii) mentioned above.

Exercise 2.1.8. Let f 0 , f 1 : ( X, x0 ) → (Y, y0 ) be two continuous maps of pointed

topological spaces. If f 0 is homotopic to f 1 in the sense of Definition 2.1.7, show
that for any spaces Z and W, and continuous maps g : ( Z, z0 ) → ( X, x0 ) and
h : (Y, y0 ) → (W, w0 ), we have

(i) f 0 ◦ g is homotopic to f 1 ◦ g in the sense of Definition 2.1.7, and

(ii) h ◦ f 0 is homotopic to h ◦ f 1 in the sense of Definition 2.1.7.

Definition 2.1.9. Let X and Y be topological spaces. A continuous map f : X →

Y is said to be a homotopy equivalence if there exist a continuous map g : Y → X
2.1. Homotopy of maps 35

such that f ◦ g ' IdY and g ◦ f ' IdX . In this case, we say that X is homotopy
equivalent to Y (or, X and Y have the same homotopy type), and write it as X ' Y.

Lemma 2.1.10. Being homotopy equivalent spaces is an equivalence relation.

Proof. For any space X, we can take f = g = IdX to get f ◦ g = IdX = g ◦ f so

that X is homotopy equivalent to itself (verify!). It follows from the Definition
2.1.9 that the relation “being homotopy equivalent spaces” is symmetric. Let
X, Y and Z be topological spaces such that X ' Y and Y ' Z. Let f 1 : X → Y
and f 2 : Y → Z be homotopy equivalences. Then there are continuous maps
g1 : Y → X and g2 : Z → Y such that g1 ◦ f 1 ' IdX , f 1 ◦ g1 ' IdY , g2 ◦ f 2 ' IdY
and f 2 ◦ g2 ' IdZ . Now using Lemma 2.1.6 we have

( f 2 ◦ f 1 ) ◦ ( g1 ◦ g2 ) = f 2 ◦ ( f 1 ◦ g1 ) ◦ g2
' f 2 ◦ IdY ◦ g2
= f 2 ◦ g2 ' IdZ .

Similarly, we have ( g1 ◦ g2 ) ◦ ( f 2 ◦ f 1 ) ' IdX . Therefore, f 2 ◦ f 1 : X → Z is a

homotopy equivalence, and hence X ' Z. Thus “being homotopy equivalent
spaces” is a transitive relation, and hence is an equivalence relation.

Definition 2.1.11. A space X is said to be contractible if the identity map IdX :

X → X is null homotopic.

Exercise 2.1.12. Show that a contractible space is path-connected.

Corollary 2.1.13. A space X is contractible if and only if given any topological space
T, any two continuous maps f , g : T → X are homotopic.

Proof. Suppose that X is contractible. Let T be any topological space, and let
f , g : T → X be any two continuous maps. Since X is contractible, the identity
map IdX : X → X of X is homotopic to a constant map c x0 : X → X given
by c x0 ( x ) = x0 , ∀ x ∈ X. Then f = IdX ◦ f is homotopic to the constant map
c x0 ◦ f : T → X. Similarly, g is homotopic to the constant map c x0 ◦ g : T → X.
Since c x0 ◦ f = c x0 ◦ g, and being homotopic maps is an equivalence relation by
Lemma 2.1.4, we see that f is homotopic to g. Converse part is obvious.
36 Chapter 2. Homotopy Theory

2.2 Fundamental group

2.2.1 Construction

A path in X is a continuous map γ : I → X; the point γ(0) ∈ X is called the

initial point of γ, and γ(1) ∈ X is called the terminal point or the final point of γ.

Definition 2.2.1. Fix two points x0 , x1 ∈ X. Two paths f , g : I → X with the

same initial point x0 and terminal point x1 are said to be path homotopic if

(i) f (0) = g(0) = x0 and f (1) = g(1) = x1 , and

(ii) there is a continuous map F : I × I → X such that for each t ∈ I, the map

γt : I → X, s 7→ F (s, t)

is a path in X from x0 to x1 , and that γ0 = f and γ1 = g.

γ1 = g

x0 x1

X γ0 = f

F IGURE 2.3: Path homotopy

Exercise 2.2.2. Given x0 , x1 ∈ X, let

Path( X; x0 , x1 ) := { f : I → X | f (0) = x0 , f (1) = x1 }

be the set of all paths in X starting at x0 and ending at x1 . Show that being path
homotopic is an equivalence relation on Path( X; x0 , x1 ). (Hint: Follow the proof
of Lemma 2.1.4).

Remark 2.2.3. If γ, δ : I → X are two paths in X, we use the symbol γ ' δ to

mean γ and δ are path-homotopic in X in the sense of Definition 2.2.1. Unless
explicitly mentioned, by a homotopy between two paths we always mean a path-
homotopy between them.
2.2. Fundamental group 37

A loop in X is a path γ : I → X with the same initial and terminal point: i.e.,
γ(0) = γ(1) = x0 ∈ X; the point x0 is called the base point of the loop γ. For a
loop γ : I → X based at x0 ∈ X, let

[γ] := {δ : I → X | δ(0) = δ(1) = x0 and δ ' γ},

the homotopy equivalence class of γ. Fix a base point x0 ∈ X, and let


π1 ( X, x0 ) := {[γ]  γ : I → X with γ(0) = γ(1) = x0 }

be the set of all equivalence classes of loops in X based at x0 . Next we define a

binary operation on π1 ( X, x0 ) and show that it is a group.

Given any two loops γ1 , γ2 : I → X in X with the base point x0 ∈ X, we

define the product of γ1 with γ2 to be the map γ1 ? γ2 : I → X defined by
(
γ1 (2t), if t ∈ [0, 21 ],
(γ1 ? γ2 )(t) := (2.2.4)
γ2 (2t − 1), if t ∈ [ 12 , 1].

That is, we first travel along γ1 with double speed from t = 0 to t = 12 , and
then along γ2 from t = 21 to t = 1. Clearly γ1 ? γ2 is a continuous map with
(γ1 ◦ γ2 )(0) = (γ1 ◦ γ2 )(1) = x0 , and hence γ1 ◦ γ2 is a loop in X with the base
point x0 . Note that γ1 ? γ2 6= γ2 ? γ1 , in general (Find such an example).

Remark 2.2.5. In fact, we shall see later examples of topological spaces X ad-
mitting loops γ1 , γ2 : I → X with the same base point x0 ∈ X such that γ1 ? γ2
is not homotopic to γ2 ? γ1 .

Proposition 2.2.6. Let γ1 , γ2 , δ1 , δ2 be loops in X based at x0 . If γ1 ' δ1 and

γ2 ' δ2 , then (γ1 ? γ2 ) ' (δ1 ? δ2 ). Consequently, the map

π1 ( X, x0 ) × π1 ( X, x0 ) → π1 ( X, x0 ), ([γ1 ], [γ2 ]) 7→ [γ1 ? γ2 ] (2.2.7)

is well-defined, and hence is a binary operation on the set π1 ( X, x0 ).

Proof. Let F : I × I → X be a homotopy from F (−, 0) = γ1 to F (−, 1) = δ1 , and

let G : I × I → X be a homotopy from G (−, 0) = γ2 to G (−, 1) = δ2 . Define a
38 Chapter 2. Homotopy Theory

map F ? G : I × I → X by sending (s, t) ∈ I × I to

(
F (2s, t), if 0 ≤ s ≤ 1/2,
( F ? G )(s, t) :=
G (2s − 1, t), if 1/2 ≤ s ≤ 1.

Clearly F ? G is a continuous map with ( F ? G )(−, 0) = γ1 ? γ2 and ( F ? G )(−, 1) =

δ1 ? δ2 . Therefore, γ1 ? γ2 ' δ1 ? δ2 .

Theorem 2.2.8. The set π1 ( X, x0 ) together with the binary operation (2.2.7) defined
in Proposition 2.2.6 is a group, known as the fundamental group of X with base point
x0 ∈ X.

To prove this theorem, we use the following technical tool (Lemma 2.2.10).

Definition 2.2.9. A reparametrization of a path γ : I → X is defined to be a

composition γ ◦ ϕ, where ϕ : I → I is a continuous map with ϕ(0) = 0 and
ϕ(1) = 1.

Lemma 2.2.10. A reparametrization of a path preserves its homotopy class.

Proof. Let γ : I → X be a path in X. Let

ϕ γ
γ ◦ ϕ : I −→ I −→ X

be a reparametrization of γ in X, for some continuous map ϕ : I → I with

ϕ(0) = 0 and ϕ(1) = 1. Consider the straight-line homotopy from ϕ to the
identity map of I given by

ϕt (s) := (1 − t) ϕ(s) + ts, ∀ s, t ∈ I.

Now it is easy to check that the map

F : I × I → X, (s, t) 7→ γ( ϕt (s)),

is continuous and satisfies F (s, 0) = (γ ◦ ϕ)(s) and F (s, 1) = γ(s), for all s ∈ I.
Therefore, γ ◦ ϕ ' γ via the homotopy F.

Proof of Theorem 2.2.8. We need to verify group axioms.

Associativity: Given any three loops γ1 , γ2 , γ3 : I → X based at x0 , it is enough

to show that (γ1 ? γ2 ) ? γ3 ' γ1 ? (γ2 ? γ3 ).
2.2. Fundamental group 39

t γ1 γ2 γ3 γ1 ? (γ2 ? γ3 )
t=1

(γ1 ? γ2 ) ? γ3
γ1 γ2 γ3
s
0 1 1 3 1
4 2 4

F IGURE 2.4: Homotopy for associativity

Note that

γ1 (4t),

 if 0 ≤ t ≤ 1/4,
((γ1 ? γ2 ) ? γ3 )(t) = γ2 (4t − 1), if 1/4 ≤ t ≤ 1/2,

γ3 (2t − 1), if 1/2 ≤ t ≤ 1,


and 
γ1 (2t),

 if 0 ≤ t ≤ 1/2,
(γ1 ? (γ2 ? γ3 ))(t) = γ2 (4t − 2), if 1/2 ≤ t ≤ 3/4,

γ3 (4t − 3), if 3/4 ≤ t ≤ 1.


It’s an easy exercise to check that γ1 ? (γ2 ? γ3 ) is a reparametrization of (γ1 ?

γ2 ) ? γ3 by a piece-wise linear function (hence, continuous) ϕ : I → I defined
by

 t/2, if 0 ≤ t ≤ 1/2,

ϕ(t) = t − 14 , if 1/2 ≤ t ≤ 3/4,

2t − 1, if 3/4 ≤ t ≤ 1,


(see Figure 2.5). Then using Lemma 2.2.10 we conclude that γ1 ? (γ2 ? γ3 ) '
(γ1 ? γ2 ) ? γ3 .

F IGURE 2.5: Graph of ϕ

Existence of identity: Let e ∈ π1 ( X, x0 ) be the homotopy class of constant loop,

c x0 : I → X, t 7→ x0 ,
40 Chapter 2. Homotopy Theory

at x0 . Let γ : I → X be any loop in X based at x0 . Since γ ? c x0 is a reparametriza-

tion of γ via the function
(
2t, if 0 ≤ t ≤ 1/2,
ψ(t) :=
1, if 1/2 ≤ t ≤ 1,

by Lemma 2.2.10 we have γ ? c x0 ' γ. Similarly, c x0 ? γ is a reparametrization

of γ by via the function
(
0, if 0 ≤ t ≤ 1/2,
η (t) :=
2t − 1, if 1/2 ≤ t ≤ 1,

by Lemma 2.2.10 we have c x0 ? γ ' γ.

Existence of inverse: Given any loop γ in X based at x0 , we can define its inverse
loop or opposite loop γ : I → X by setting γ(t) = γ(1 − t), for all t ∈ I. We need
to show that γ ? γ ' c x0 and γ ? γ ' c x0 . To show γ ? γ ' c x0 , consider the map
H : I × I → X given by

H (s, t) := f t (s) ? gt (s), ∀ (s, t) ∈ I × I,

where f t : I → X is the path defined by

(
γ ( s ), for 0 ≤ s ≤ 1 − t,
f t (s) =
γ(1 − t), for 1 − t ≤ s ≤ 1,

and gt : I → X is the inverse path of f t , i.e., gt (s) = f t (1 − s), ∀ s ∈ I. It is an

easy exercise to check that H is a continuous map satisfying

H (s, 0) = γ ? γ, and H (s, 1) = c x0 , ∀ s ∈ I.

The homotopy H can be understood using the Figure 2.6. In the bottom line
t = 0, we have γ ? γ while on the top line t = 1 we have the constant loop
c x0 . And below the ‘V’ shape we let H (s, t) be independent of t while above
the ‘V’ shape we let H (s, t) be independent of s. Therefore, we have γ ? γ '
c x0 . Interchanging the roles of γ and γ in the above construction, we see that
γ ? γ ' c x0 . Therefore, π1 ( X, x0 ) is a group.
2.2. Fundamental group 41

t
c x0
1

s
0 γ 1 γ 1
2

F IGURE 2.6: Homotopy H

2.2.2 Functoriality

By a pointed topological space we mean a pair ( X, x0 ) consisting of a topolog-

ical space X and a point x0 ∈ X. In the above construction, given a pointed
topological space ( X, x0 ) we attached a group π1 ( X, x0 ), known as the funda-
mental group of X with the base point at x0 . Next we see how fundamental group
of a pointed space behaves under continuous maps and their compositions.

Let f : ( X, x0 ) → (Y, y0 ) be a continuous map of pointed spaces (this means,

f : X → Y is a continuous map with f ( x0 ) = y0 ). Let γ : I → X be a loop in X
based at x0 . Then the composition f ◦ γ,

γ f
I −→ X −→ Y,

is a loop in Y based at f ( x0 ) = y0 . Let γ, δ : I → X be loops in X based at x0 . If

F : I × I → X is a homotopy from γ to δ, then f ◦ F : I × I → Y is a homotopy
from f ◦ γ to f ◦ δ (see Lemma 2.1.6). Thus we have a well-defined map

f ∗ : π1 ( X, x0 ) → π1 (Y, y0 ), [γ] 7→ [ f ◦ γ]. (2.2.11)

Proposition 2.2.12. The map f ∗ : π1 ( X, x0 ) → π1 (Y, y0 ) induced by f is a group

homomorphism.
42 Chapter 2. Homotopy Theory

Proof. Note that for any two loops γ, δ : I → X based at x0 , we have

f ∗ ([γ ? δ]) = [ f ◦ (γ ? δ)]

= [( f ◦ γ) ? ( f ◦ δ)]
= [ f ◦ γ] · [ f ◦ δ]
= f ∗ ([γ]) · f ∗ ([δ]).

Remark 2.2.13. If f ∗ : π1 ( X, x0 ) → π1 ( X, x0 ) is the homomorphism of fun-

damental group of a pointed topological space ( X, x0 ) induced by the identity
map of ( X, x0 ) onto itself, then it follows from the construction of the map f ∗
given in (2.2.11) that f ∗ = Idπ1 (X, x0 ) , the identity map of π1 ( X, x0 ) onto itself.

Proposition 2.2.14. Let f : ( X, x0 ) → (Y, y0 ) and g : (Y, y0 ) → ( Z, z0 ) be

continuous maps of pointed spaces. Then g∗ ◦ f ∗ = ( g ◦ f )∗ . In other words, the
following diagram commutes.

f∗
π1 ( X, x0 ) / π1 (Y, y0 )
g∗
( g◦ f )∗ ) 
π1 ( Z, z0 )

Proof. Left as an exercise.

Corollary 2.2.15. If f : ( X, x0 ) → (Y, y0 ) is a homeomorphism of pointed spaces

with its inverse g : (Y, y0 ) → ( X, x0 ), then f ∗ : π1 ( X, x0 ) → π1 (Y, y0 ) is an
isomorphism of groups with its inverse g∗ : π1 (Y, y0 ) → π1 ( X, x0 ).

Proof. Since g ◦ f = Id(X, x0 ) and f ◦ g = Id(Y, y0 ) , applying Proposition 2.2.14

we have g∗ ◦ f ∗ = Idπ1 (X, x0 ) and f ∗ ◦ g∗ = Idπ1 (Y, y0 ) .

Lemma 2.2.16. Homotopic continuous maps of pointed topological spaces induces the
same homomorphism of fundamental groups.

Proof. Let f , g : ( X, x0 ) → (Y, y0 ) be two continuous maps of pointed space. If

f is homotopic to g in the sense of Definition 2.1.7, then for any loop γ : I → X
based at x0 , using Exercise 2.1.8 (c.f. Lemma 2.1.6) we have f ◦ γ is homotopic
to g ◦ γ in the sense of Definition 2.1.7, and hence f ∗ ([γ]) = [ f ◦ γ] = [ g ◦ γ] =
g∗ ([γ]). Hence the result follows.
2.2. Fundamental group 43

Definition 2.2.17. A category C consists of the following data:

(i) a collection of objects ob(C ),

(ii) for each ordered pair of objects ( X, Y ) of ob(C ), there is a collection MorC ( X, Y ),
whose members are called arrows or morphisms from X to Y in C ; an object
ϕ ∈ MorC ( X, Y ) is usually denoted by an arrow ϕ : X → Y.

(iii) for each ordered triple ( X, Y, Z ) of objects of C , there is a map (called

composition map)

◦ : MorC ( X, Y ) × MorC (Y, Z ) → MorC ( X, Z ), ( f , g) 7→ g ◦ f ,

such that the following conditions hold.

(a) Associativity: Given X, Y, Z, W ∈ ob(C ), and f ∈ MorC ( X, Y ), g ∈

MorC (Y, Z ) and h ∈ MorC ( Z, W ), we have h ◦ ( g ◦ f ) = (h ◦ g) ◦ f .
(b) Existence of identity: For each X ∈ ob(C ), there exists a morphism
IdX ∈ MorC ( X, X ) such that given any objects Y, Z ∈ ob(C ) and
morphism f : Y → Z we have f ◦ IdY = f and IdZ ◦ f = f .

Example 2.2.18. (i) Let (Set) be the category of sets; its objects are sets and
arrows are map of sets.

(ii) Let (G rp) be the category of groups; its objects are groups and arrows are
group homomorphisms.

(iii) Let (T op) be the category of topological spaces; its objects are topological
spaces and arrows are continuous maps.

(iv) Let (Ring) be the category of rings; its objects are rings and morphisms
are ring homomorphisms.

Definition 2.2.19. Let C be a category. A morphism f : X → Y in C is said to

be an isomorphism if there is a morphism g : Y → X in C such that g ◦ f = IdX
and f ◦ g = IdY .

Definition 2.2.20. Let C and D be two categories. A covariant functor (resp., a

contravariant functor) from C to D is a rule

F :C →D
44 Chapter 2. Homotopy Theory

which associate to each object X ∈ C an object F ( X ) ∈ D, and to each mor-

phism f ∈ MorC ( X, Y ) a morphism F ( f ) ∈ MorD (F ( X ), F (Y )) (resp., a mor-
phism F ( f ) ∈ MorD (F (Y ), F ( X ))) such that

(i) F (IdX ) = IdF (X ) , for all X ∈ C , and

(ii) given any objects X, Y, Z ∈ C and morphisms f ∈ MorC ( X, Y ) and g ∈

MorC (Y, Z ), we have F ( g ◦ f ) = F ( g) ◦ F ( f ) (resp., F ( g ◦ f ) = F ( f ) ◦
F ( g)).

If F : C → D is a functor, for each ordered pair of objects X, Y ∈ C we

denote by F X,Y the induced map

F X,Y : MorC ( X, Y ) → MorD (F ( X ), F (Y )),

defined by F X,Y ( f ) := F ( f ). The same notation is used for contravariant func-

tor.

Definition 2.2.21. A functor F : C → D is said to be

(i) faithful if F X,Y is injective, ∀ X, Y ∈ C .

(ii) full if F X,Y is surjective, ∀ X, Y ∈ C .

(iii) fully faithful if F X,Y is bijective, ∀ X, Y ∈ C .

(iv) essentially surjective if given any object Y ∈ D, there is an object X ∈ C

'
and an isomorphism ϕ : F ( X ) → Y in D.

(v) equivalence of categories if there is a functor G : D → C such that G ◦ F ∼=

∼
IdC and F ◦ G = IdD . This is equivalent to say that F is fully faithful and
essentially surjective.

Remark 2.2.22. Let T op0 be the category of pointed topological spaces; its ob-
jects are pointed topological space, and given any two pointed topological
spaces ( X, x0 ) and (Y, y0 ), a morphism f : ( X, x0 ) → (Y, y0 ) in T op0 is a contin-
uous map f : X → Y such that f ( x0 ) = y0 . Then it follows from Propositions
2.2. Fundamental group 45

2.2.12 and 2.2.14 and the Remark 2.2.13 that

π1 : T op0 −→ (G rp)
( X, x0 ) 7→ π1 ( X, x0 )
f 7→ f ∗

is a covariant functor from the category of pointed topological spaces to the cat-
egory of groups. It follows from Lemma 2.2.16 that the functor π1 is not faith-
ful. It is a non-trivial fact that π1 is not full. (i.e., there exist pointed topolog-
ical spaces ( X, x0 ) and (Y, y0 ), and a group homomorphism ϕ : π1 ( X, x0 ) →
π1 (Y, y0 ) such that ϕ 6= f ∗ , for all continuous map f : ( X, x0 ) → (Y, y0 ).)
However, π1 is essentially surjective (i.e., given any group G there is a pointed
topological space ( X, x0 ) such that π1 ( X, x0 ) ∼
= G).

2.2.3 Dependency on base point

Now we investigate relation between fundamental groups of X for different

choices of base point. Let x0 , x1 ∈ X. Let f : I → X be a path in X joining x0
to x1 , i.e., f is a continuous map satisfying f (0) = x0 and f (1) = x1 . We define
the opposite path of f to be the map

f : I → X, t 7→ f (1 − t); (2.2.23)

note that f (0) = x1 and f (1) = x0 , hence f is a path from x1 to x0 .

Exercise 2.2.24. Show that f ? f ' c x0 and f ? f ' c x1 , where ' stands for
path-homotopy relation (see Definition 2.2.1).

Given a loop γ in X based at x1 , we can define γ e := f ? γ ? f . Note that

e : I → X is a continuous map satisfying γ
γ e( 0 ) = f ( 0 ) = x 0 = f ( 1 ) = γ
e( 1 ) ,
and hence is a loop in X based at x1 . Strictly speaking, we have two choices to
define this product γe, namely ( f ? γ) ? f or f ? (γ ? f ), but we are interested in
only homotopy classes of paths, and following the proof of associativity as in
Theorem 2.2.8 one can easily verify that ( f ? γ) ? f ' f ? (γ ? f ), therefore, we
just fix one ordering of taking products to define γ e.
46 Chapter 2. Homotopy Theory

If γ and γ0 are two loops in X based at x1 with γ ' γ0 via a homotopy

e to γe0 (Exercise: Write down the
{ht }t∈ I , then { f ? ht ? f }t∈ I is a homotopy from γ
homotopy explicitly and check details). Thus, we have a well-defined map

β f : π1 ( X, x1 ) → π1 ( X, x0 ), [γ] 7→ [( f ? γ) ? f ]. (2.2.25)

Proposition 2.2.26. The map β f defined in (2.2.25) is a group isomorphism.

Proof. Since f ? f ' c x0 for any two loops γ and δ in X based at x1 , using
Exercise 2.2.24, we have

f ? ( γ ? δ ) ? f ' f ? γ ? c x0 ? δ ? f
' ( f ? γ ? f ) ? ( f ? δ ? f ).

Therefore, β f ([γ ? δ]) = [ f ? (γ ? δ) ? f ] = [ f ? γ ? f ][ f ? δ ? f ] = β f ([γ]) β f ([δ]),

and hence β f is a group homomorphism. To show β f an isomorphism of
groups, it is enough to show that the group homomorphism

β f : π1 ( X, x0 ) → π1 ( X, x1 ), [γ] 7→ [ f ? γ ? f ]

is the inverse of β f . Indeed, for any γ ∈ π1 ( X, x0 ) we have

β f ( β f ([γ])) = β f ([ f ? γ ? f ])
= [f ? f ? γ ? f ? f]
= [ c x0 ? γ ? c x0 ] = [ γ ],

and similarly, for any δ ∈ π1 ( X, x1 ) we have

β f ( β f ([δ])) = β f ([ f ? δ ? f ])
= [f ? f ? δ ? f ? f]
= [ c x1 ? δ ? c x1 ] = [ δ ].

Therefore, β f is the inverse homomorphism of β f , and hence both of them are

isomorphisms.

Remark 2.2.27. Thus if X is a path connected space, up to isomorphism its fun-

damental group is independent of choice of base point, and so we may denote
it by π1 ( X ) without specifying its base point.
2.2. Fundamental group 47

Proposition 2.2.28. Let f , g : X → Y be two continuous maps of topological spaces.

Fix a point x0 ∈ X, and let y0 = f ( x0 ) and y1 = g( x0 ). Let F : X × I → Y be a
continuous map such that F (−, 0) = f and F (−, 1) = g. Then for any loop γ in X
based at x0 , the loop f ◦ γ is path-homotopic to the loop F0 ? ( g ◦ γ) ? F0 in Y, where
F0 : I → Y is the path in Y defined by F0 (t) = F ( x0 , t), ∀ t ∈ I.

g∗
π1 ( X, x0 ) / π1 (Y, y1 )

f∗ & x β F0
π1 (Y, y0 )

Proof. Left as an exercise.

Corollary 2.2.29. Let f , g : X → Y be two homotopic continuous maps of topological

spaces. Let x0 ∈ X be such that f ( x0 ) = g( x0 ) = y0 ∈ Y. Then the homomorphisms
of fundamental groups f ∗ and g∗ , induced by f and g, respectively, are conjugate by
an element of π1 (Y, y0 ). In other words, there exists an element [η ] ∈ π1 (Y, y0 ) such
that g∗ ([γ]) = [η ] f ∗ ([γ])[η ]−1 , for all [γ] ∈ π1 ( X, x0 ).

Proof. Let F : X × I → X be a continuous map such that

(
f ( x ), if t = 0,
F ( x, t) =
g( x ), if t = 1.

Then by Proposition 2.2.28 we have g∗ ([γ]) = [η ] f ∗ ([γ])[η ]−1 , for all [γ] ∈
π1 ( X, x0 ), where η : I → Y is the loop defined by η (t) := F ( x0 , t), ∀ t ∈ I.

Corollary 2.2.30. If f , g : ( X, x0 ) → (Y, y0 ) are two homotopic continuous maps of

pointed topological spaces (see Definition 2.1.7), then f ∗ = g∗ .

Proof. Follows from Corollary 2.2.29.

Definition 2.2.31. A space X is said to be simply connected if X is path connected

and π1 ( X ) is trivial.

Corollary 2.2.32. A contractible space (see Definition 2.1.11) is simply connected.

Proof. Let X be a contractible space. Then X is path-connected by Exercise

2.1.12. Fix a point x0 ∈ X, and let c x0 : X → X be the constant map send-
ing all points to x0 . Since X is contractible, the identity map IdX : X → X is
48 Chapter 2. Homotopy Theory

homotopic to the constant map c x0 in X. Then by Corollary 2.2.29 the identity

homomorphism Idπ1 (X,x0 ) : π1 ( X, x0 ) → π1 ( X, x0 ) is conjugate to the trivial
homomorphism (c x0 )∗ : π1 ( X, x0 ) → π1 ( X, x0 ) by an element of π1 ( X, x0 ).
Therefore, the image of the identity homomorphism Idπ1 (X, x0 ) is trivial, and
hence π1 ( X, x0 ) is trivial.

Corollary 2.2.33. A space X is simply connected if and only if there is a unique path-
homotopy class of paths connecting any two points of X.

Proof. Suppose that X is simply connected. Fix x0 , x1 ∈ X. Since X is path-

connected, there is a path in X joining x0 to x1 . Let f , g : I → X be any two paths
in X from x0 to x1 . Let f and g be the opposite paths of f and g, respectively.
Since f ? g is a loop in X based at x0 and π1 ( X, x0 ) is trivial, we have f ? g is
path-homotopic to the constant loop c x0 in X based at x0 . Since g ? g is path-
homotopic to the constant loop c x1 by Exercise 2.2.24, we have

f ' f ? c x1 ' f ? g ? g ' c x0 ? g ' g.

To see the converse part, note that path connectedness of X means any two
points of X can be joined by a path in X. Since there is a unique homotopy class
of paths connecting any two points of X, path connectedness of X is automatic,
and any loop in X based at a given point x0 ∈ X is homotopically trivial. Thus,
X is path connected with π1 ( X, x0 ) trivial, and hence is simply connected.

Proposition 2.2.34. π1 ( X × Y, ( x0 , y0 )) ∼
= π1 ( X, x0 ) × π1 (Y, y0 ).

Proof. Note that X × Y naturally acquires product topology induced from X

and Y, and the projection maps p1 : X × Y → X and p2 : X × Y → Y defined by
p1 ( x, y) = x and p2 ( x, y) = y, for all ( x, y) ∈ X × Y, are continuous. Moreover,
given any space Z and a map f : Z → X × Y, we have f = ( p1 ◦ f , p2 ◦ f ). From
this, it follows that f is continuous if and only if its components p1 ◦ f : Z → X
and p2 ◦ f : Z → Y are continuous. Therefore, to give a loop γ : I → X × Y
based at ( x0 , y0 ) ∈ X × Y is equivalent to give a pair of loops ( p1 ◦ γ, p2 ◦
γ) in the pointed spaces ( X, x0 ) and (Y, y0 ), respectively. Similarly, to give a
homotopy F : I × I → X × Y of loops γ, δ : I → X × Y based at ( x0 , y0 )) is
equivalent to give a pair of homotopies ( p1 ◦ F, p2 ◦ F ) of the corresponding
2.2. Fundamental group 49

loops p j ◦ γ with p j ◦ δ, where j ∈ {1, 2}. Thus we have a bijection

φ : π1 ( X × Y, ( x0 , y0 )) → π1 ( X, x0 ) × π1 (Y, y0 ), [γ] 7→ ([ p1 ◦ γ], [ p2 ◦ γ]).

To see φ is an isomorphism, note that for any two loops γ and δ in X × Y based
at ( x0 , y0 ) ∈ X × Y, we have

φ([γ] · [δ]) = φ([γ ? δ])

= ([ p1 ◦ (γ ? δ)], [ p2 ◦ (γ ? δ)])
= ([( p1 ◦ γ) ? ( p1 ◦ δ)], [( p2 ◦ γ) ? ( p2 ◦ δ)])
= ([( p1 ◦ γ)] · [( p1 ◦ δ)], [( p2 ◦ γ)] · [( p2 ◦ δ)])
= ([ p1 ◦ γ], [ p2 ◦ γ]) · ([ p1 ◦ δ], [ p2 ◦ δ])
= φ([γ]) · φ([δ]).

This completes the proof.

Example 2.2.35. As an immediate application of Proposition 2.2.34 we see that

the fundamental group of the 1-torus S1 × S1 is isomorphic to Z × Z.

We end this subsection with the following useful remark.

Remark 2.2.36. A loop in X based at x0 can equivalently be defined as a contin-

uous map of pointed spaces γ : (S1 , 1) → ( X, x0 ). Indeed, since a loops in X
based at x0 ∈ X is a continuous map γ : I = [0, 1] → X with γ(0) = γ(1) = x0 ,
and since S1 is homeomorphic to the quotient space [0, 1]/ ∼, where only the
end points 0 and 1 of the interval I are identified, γ : I → X uniquely factors as

8/
γ
I X
q
 γ0
S1

where q : I → S1 is the quotient map given by q(t) = e2πit , for all t ∈ I.

Therefore, π1 ( X, x0 ) is the group of all homotopy classes of continuous maps
(S1 , 1) → ( X, x0 ).
50 Chapter 2. Homotopy Theory

2.2.4 Fundamental group of some spaces

Proposition 2.2.37. π1 (R, 0) = {1}.

Proof. Consider the continuous map F : R × I → R defined by

F ( x, t) = (1 − t) x, ∀ ( x, t) ∈ R × I.

Note that, for all x ∈ R and t ∈ I we have

• F ( x, 0) = x,

• F ( x, 1) = 0, and

• F (0, t) = 0.

Therefore, F “contracts” whole R to the point 0 leaving the point 0 intact at all
times. Let γ : I → R be a loop based at 0. Then the composite map

γ×Id I F
F ◦ (γ × Id I ) : I × I −→ R × I −→ R

is a homotopy from γ to the constant loop 0 : I → R which sends all points of

S1 to 0 ∈ R. This completes the proof.

Proposition 2.2.38. Let

D2 := {z ∈ C : |z| ≤ 1} = {( x, y) ∈ R2 : x2 + y2 ≤ 1}

be the closed unit disk in the plane. Then π1 ( D2 , 1) = {1}.

Proof. Consider the map F : D2 × I → D2 defined by

F (z, t) = (1 − t)z + t, ∀ (z, t) ∈ D2 × I.

Note that F is continuous and for all z ∈ D2 and t ∈ I we have

• F (z, 0) = z,

• F (z, 1) = 1, and
2.3. Covering Space 51

• F (1, t) = 1.

Therefore, F contracts D2 to the point 1 leaving 1 intact at all times. Let γ : I →

D2 be a loop based at 1. Then the composite map

γ×Id I F
F ◦ (γ × Id I ) : I × I −→ D2 × I −→ D

is a homotopy from γ to the constant loop 1 : I → D2 which sends all points of

I to 1 ∈ D2 . This completes the proof.

2.3 Covering Space

2.3.1 Covering map

We begin this section with an aim to compute fundamental group of the

unit circle in plane

S1 = {z ∈ C : |z| = 1} = {( x, y) ∈ R2 : x2 + y2 = 1},

and we’ll see how the idea of a ‘covering map’ could help us.
√
Let ω : I → S1 be the map defined by ω (t) = e2πit , ∀ t ∈ I, where i = −1.
Then ω is a loop in S1 based at x0 := 1 ∈ S1 . For each integer n, let ωn : I → S1
be the loop based at x0 defined by ωn (t) = e2πint , ∀ t ∈ I. So ωn winds around
the circle |n|-times in the anti-clockwise direction if n > 0, and in the clockwise
direction if n < 0. We shall see later that [ω ]n = [ωn ] in π1 (S1 , 1), for all n ∈ Z.
The following is the main theorem of this section.

Theorem 2.3.1. π1 (S1 , x0 ) is the infinite cyclic group Z generated by the loop ω.

To prove this theorem, we compare paths in S1 with paths in

R via the map

p : R → S1 given by p(s) = (cos 2πs, sin 2πs), ∀ s ∈ R.

F IGURE 2.7
52 Chapter 2. Homotopy Theory

We can visualize this map geometrically by embedding R in-

side R3 as the helix parametrized as

s 7→ (cos 2πs, sin 2πs, s),

and then p is the restriction of the projection map

R3 → R2 , ( x, y, z) 7→ ( x, y)

from this helix onto S1 ⊂ R2 , as shown in the Figure 2.7.

In this setup, the loop

ωn : I → S1 , s 7→ (cos 2nπs, sin 2nπs)

is the composition p ◦ ω
e n , where

e n : I → R, s 7→ ns
ω

is the path in R starting at 0 and ending at n, winding around the helix |n|-
times, upward direction if n > 0, and downward direction if n < 0.

Before proceeding further, we introduce notion of a covering map, and dis-

cuss some of its useful properties.

Definition 2.3.2. Let f : X → Y be a continuous map. An open subset V ⊆ Y is

said to be evenly covered by f if f −1 (V ) is a union of pairwise disjoint open sub-
sets of X each of which are homeomorphic to V by f (meaning that, f −1 (V ) =
Ui , where Ui ⊆ X is an open subset of X with Ui ∩ Uj = ∅, for all i 6= j in I,
S
i∈ I 
and f U : Ui → V is a homeomorphism, for all i ∈ I).
i

Example 2.3.3. (i) Let f : R → S1 := {z ∈ C : |z| = 1} be the map defined

by f (t) = e2πit = (cos 2πt, sin 2πt), for all t ∈ R. For a, b ∈ R with a < b,
we define an open subset

Va, b := { f (t) : a ≤ t ≤ b} ⊆ S1 .

If b − a < 1, then Va,b is evenly covered by f . In fact, in this case, we

'
have f −1 (Va,b ) = ( a + n, b + n), and f : ( a + n, b + n) −→ Va,b is a
F
n ∈Z
homeomorphism, ∀ n ∈ Z. See Figure 2.8.
2.3. Covering Space 53

F IGURE 2.8

If b − a ≥ 1, then Va, b = S1 , and hence f −1 (Va, b ) = R. In this case, Va, b

is not evenly covered by f , for otherwise we would have R =
F
Ui with
i∈ I
each Ui open subset of R and f U : Ui → S1 is a homeomorphism, which

i
is not possible because S1 is compact, whereas an open subset of R cannot
be compact.

(ii) Let R>0 := {t ∈ R : t > 0} be the positive part of the real line. Let

f : R>0 → S1 , t 7→ e2πit . (2.3.4)

For any point x ∈ S1 with x 6= 1 := (1, 0) ∈ S1 , we can choose a small

enough open neigbourhood V of x in S1 with 1 ∈ / V. Then it is easy to
see that V is evenly covered by f . However, there is no evenly covered
neighbourhood of 1 ∈ S1 . To see this, note that if U ⊆ V is an open subset
of an evenly covered neighbourhood V, then U is also evenly covered.
Thus, if there is a neighbourhood V of 1 which is evenly covered, then
we may find e ∈ (0, 1/2) small enough such that V−e, e ⊆ V, and hence
V−e, e is evenly covered. Then we must have f −1 (V−e, e ) =
F
Ui , with
 i ∈ I
f U : Ui → V−e, e homeomorphism, for all i ∈ I. In particular, each Ui
i
is connected and are path components of f −1 (V−e, e ). Let U0 be the path
54 Chapter 2. Homotopy Theory

F IGURE 2.9

component of e/2 ∈ R>0 . Since

!
−1
[ [
f (V−e, e ) = (0, e) (n − e, n + e) ,
n ≥1


we must have U0 = (0, e). But f (0, e) : (0, e) → V−e, e cannot be surjec-
tive because only possible preimage of 1 ∈ V−e, e in R+ could be positive

integers, and none of which are in the domain of f (0, e) . Thus we get a
contradiction. See Figure 2.9. Therefore, there is no evenly covered neigh-
bourhood of 1 ∈ S1 for the map f in (2.3.4).

Definition 2.3.5. A continuous map f : X → Y is called a covering map if each

point y ∈ Y has an open neighbourhood Vy ⊆ Y that is evenly covered by f .

Note that, a covering map is always surjective. This follows immediately

from the Definition 2.3.5.

Example 2.3.6. (i) Let F be a non-empty discrete topological space, and let
X be any topological space. Give X × F the product topology. Then the
projection map pr1 : X × F → X defined by pr1 ( x, v) = x, ∀ ( x, v) ∈
X × F, is a covering map. Such a covering map is called a trivial cover of
X.

(ii) The continuous map

f : R → S1 , t 7→ e2πit ,
2.3. Covering Space 55

as discussed in Example 2.3.3 (i), is a covering map, while its restriction

f R+ : R+ → S1 , in Example 2.3.3 (ii), is not a covering map.


(iii) The map f : C → C∗ := C \ {0} defined by f (z) = ez , for all z ∈ C, is a

covering map.

(iv) Fix an integer n ≥ 1. Then the map f : C∗ → C∗ defined by f (z) = zn , for

all z ∈ C, is a covering map, known as the n-sheeted covering map of C∗ .

Exercise 2.3.7. If f i : Xi → Yi is a covering map, for i = 1, 2, show that the

map f 1 × f 2 : X1 × X2 → Y1 × Y2 defined by sending ( x1 , x2 ) ∈ X1 × X2 to
( f 1 ( x1 ), f 2 ( x2 )) ∈ Y1 × Y2 , is a covering map.

Exercise 2.3.8. If f : X → Y is a covering map, for any subspace Z ⊆ Y, the

restriction of f on f −1 ( Z ) ⊆ X is a covering map.

Definition 2.3.9. Let p1 : Y1 → X and p2 : Y2 → X be two covering maps. A

morphism of covering maps from p1 to p2 is a continuous map φ : Y1 → Y2 such
that p2 ◦ φ = p1 . In other words, the following diagram commutes.

φ
Y1 /Y
2

p1 p2
 
X

A morphism of covering maps φ : Y1 → Y2 is said to be an isomorphism of

covering maps if there is a covering map ψ : Y2 → Y1 such that φ ◦ ψ = IdY2
and ψ ◦ φ = IdY1 . In other words, an isomorphism of covering spaces is a
homeomorphism of the covers compatible with the base. An isomorphism of
a covering map p : Y → X to itself is called a Deck transformation or a covering
transformation.

Exercise 2.3.10. Show that any covering map p : Y → X is locally trivial (i.e.,
each point x ∈ X has an open neighbourhood Ux ⊆ X such that the restriction
map p : p−1 (Ux ) → Ux is isomorphic to a trivial covering map over Ux ).

A continuous map f : X → Y is said to be an open map if for any open subset

U of X, f (U ) is open in Y.

Proposition 2.3.11. If f : X → Y is a covering map, then f is an open map.

56 Chapter 2. Homotopy Theory

Proof. Let U ⊆ X be an open subset of X, and let y ∈ f (U ). Then there is x0 ∈ U

such that f ( x0 ) = y. Since f is a covering map, there is an open neighbourhood
V ⊆ Y of y such that f −1 (V ) =
S
Wi is a union of pairwise disjoint open
j∈ J

subsets Wj ⊆ X, and that f W : Wj → V is a homeomorphism, for all j ∈ J.
j 
Then x0 ∈ U ∩ Wj , for some unique i0 ∈ I. Since f  is a homeomorphism,
0 Wi
f (U ∩ Wj0 ) ⊆ V is an open neighbourhood of f ( x0 ) = y. Since V is open in
Y, f (U ∩ Wj0 ) is open in Y. Thus f (U ) is open in Y, and hence f is an open
map.

Theorem 2.3.12 (Lifting path to a cover). Let f : X → Y be a covering map. Let

γ : [0, 1] → Y be a path in Y. Fix a point x0 ∈ X such that f ( x0 ) = y0 := γ(0).
e : [0, 1] → X with γ
Then there is a unique path γ e(0) = x0 and f ◦ γ
e = γ.

7 X
γ
e
f

[0, 1] /Y
γ

The path γ
e is called a lift of γ in X starting at x0 .

Proof. We first prove uniqueness of lift of γ, if it exists. Let η1 , η2 : [0, 1] → X be

any two continuous maps such that η1 (0) = x0 = η2 (0) and f ◦ η1 = γ = f ◦ η2 .
We need to show that η1 = η2 on [0, 1]. Let

S = {t ∈ [0, 1] : η1 (t) = η2 (t)}.

Since both η1 and η2 are continuous, S is a closed subset of [0, 1]. Note that
S 6= ∅ since 0 ∈ S. Since [0, 1] is connected, it is enough to show that S is both
open and closed in [0, 1], so that S is a connected component of [0, 1], and hence
S = [0, 1].

Fix a t ∈ S, and let V ⊆ Y be an open neighbourhood of y := γ(t) that is

evenly covered by f . So f −1 (V ) = Uj , where {Uj } j∈ J is a collection of pair-
S
j∈ J
wise disjoint open subsets of X each of which gets mapped homeomorphically
onto V by f . Then there are j1 , j2 ∈ J such that η1 (t) ∈ Uj1 and η2 (t) ∈ Uj2 .
Since η1 and η2 are continuous at t ∈ [0, 1], there is an open neighbourhood
W ⊆ [0, 1] of t such that η1 (W ) ⊆ Uj1 and η2 (W ) ⊆ Uj2 . Since Uj1 ∩ Uj2 = ∅
for j1 6= j2 , and since η1 (t) = η2 (t) by assumption, we must have j1 = j2 and
2.3. Covering Space 57

F IGURE 2.10


Uj1 = Uj2 . Since f U : Uj → V is injective (in fact, homeomorphism), for all
j  
j ∈ J, and f ◦ η1 = f ◦ η2 , we must have η1 W = η2 W . Therefore, W ⊆ S. Thus
S is both open and closed in [0, 1], and hence is the connected component of
[0, 1]. Therefore, S = [0, 1], and hence η1 = η2 on [0, 1].
Remark 2.3.13. Note that, by replacing [0, 1] with any connected topological
space T in the above proof of uniqueness of lift of γ, we get the following result:

Lemma 2.3.13. Let f : X → Y be a covering map. Let η1 , η2 : T → X be any

continuous maps such that f ◦ η1 = f ◦ η2 . If T is connected and η1 (t) = η2 (t), for
some t ∈ T, then η1 = η2 on whole T.

To complete the proof of Theorem 2.3.12, it remains to construct an explicit

lift of γ to the cover f : X → Y starting at x0 . For this we use a result from basic
topology course, called Lebesgue number lemma (to keep the note self-contained,
we include its proof later).
Lemma 2.3.14 (Lebesgue number lemma). Let {Uj } j∈ J be an open cover of a com-
pact metric space ( X, d). Then there is a δ > 0 such that for each x0 ∈ X, the open ball
Bδ ( x0 ) is contained in Uj0 , for some j0 ∈ J.

Since f : X → Y is a covering map, we can write Y = Vy , where Vy ⊆

S
y ∈Y
Y is an open neighbourhood of y that is evenly covered by f , for all y ∈ Y.
S −1
Since [0, 1] = γ (Vy ), by Lebesgue covering lemma (c.f. Lemma 2.3.14)
y ∈Y
we can find a δ > 0 such that for each t ∈ (0, 1) there is a yt ∈ Y such that
58 Chapter 2. Homotopy Theory

1


γ [t − 2δ , t + 2δ ] ∩ [0, 1] ⊆ Vyt . Choose n  0 such that n < δ, and write

−1 
n[ 
k k+1
[0, 1] = , .
k =0
n n

Now γ([0, 1/n]) ⊆ V0 , for some open subset V0 ⊂ Y evenly covered by f , and
y0 = γ(0) ∈ V0 . Write
f −1 (V0 ) =
G
U0, j ,
j∈ J

where {U0, j } j∈ J is a collection of pair-wise disjoint open subsets of X each of

which are homeomorphic to V0 via the restriction of f onto them. Since x0 ∈
f −1 (V0 ), there is a unique j0 ∈ J such that x0 ∈ U0, j0 . Let s0 : V0 → U0, j0 be the

inverse of the homeomorphism f U . Clearly s0 (y0 ) = x0 . Consider the map
0, j0
1
e0 : [0,
γ n] → U0, j0 defined by

e0 (t) := s0 (γ(t)), ∀ t ∈ [0, 1/n] .

γ

Then γ
e0 satisfies γ e0 = γ on [0, n1 ].
e0 (0) = x0 and f ◦ γ

e0 ( n1 ) and y1 = γ( n1 ) = ( f ◦ γ
Let x1 = γ e0 )( n1 ). Then there is an open subset
V1 ⊆ Y which is evenly covered by f and γ([ n1 , n2 ]) ⊆ V1 . Proceeding in the
same way as above, we can write

f −1 (V1 ) =
G
U1, j ,
j∈ J

where U1, j are pairwise disjoint open subsets of X each of which are homeo-
morphic to V1 by the restriction of f onto them. Since x1 = γ e0 ( n1 ) ∈ f −1 (V1 ),
there is a j1 ∈ J such that x1 ∈ U1,j1 . Let s1 : V1 → U1, j1 be the inverse of the
homeomorphism f : U1, j1 → V1 . Clearly s1 (y1 ) = x1 . Then the continuous
map γe1 : [ n1 , n2 ] → U1, j1 defined by

e1 (t) = s1 (γ(t)), ∀ t ∈ [1/n, 2/n]

γ

e1 ( n1 ) = x1 and f ◦ γ
satisfies γ e1 = γ on [ n1 , n2 ]. Since the maps γ e0 and γe1 agrees
1 1 2 1
on [0, n ] ∩ [ n , n ] = { n }, by Lemma 2.1.3 we can join them to get a continuous
map γ e : [0, n2 ] → X such that γ e(0) = x0 and f ◦ γ e = γ on [0, n2 ]. Proceeding in
this way we can construct a lift γ e of γ to the whole [0, 1] as required.
2.3. Covering Space 59

We now give a proof of Lemma 2.3.14. If you have already learned it from
your basic topology course, you may skip it. Let’s begin with the following
observation from metric space.

Proposition 2.3.15. Let X be a metric space with the metric d, and let Y be a subspace
of X. Then the function φY : X → R defined by

φY ( x ) = inf d( x, y), ∀ x ∈ X,
y ∈Y

is continuous.

Proof. Let x0 ∈ X. Fix a δ > 0, and let z ∈ Bδ ( x0 ) := { x ∈ X : d( x0 , x ) < δ}.

Then for any y ∈ X, using the triangle inequality we have

φY (z) = inf d(z, y0 ) ≤ d(z, y) ≤ d(z, x0 ) + d( x0 , y).

y 0 ∈Y

Now taking infimum over y ∈ Y and subtracting φY ( x0 ), we have

φY (z) − φY ( x0 ) ≤ d(z, x0 ).

Since z ∈ Bδ ( x0 ) implies x0 ∈ Bδ (z), repeating the above construction we have

φY ( x0 ) − φY (z) ≤ d(z, x0 ), and hence

|φY ( x0 ) − φY (z)| ≤ d(z, x0 ).

Therefore, given any e > 0, taking δ = e, we see that

|φY ( x0 ) − φY (z)| < e, whenever d( x0 , z) < δ.

Since x0 ∈ X is arbitrary, φY is continuous on X.

Now we give a proof of Lebesgue number lemma.

Proof of Lemma 2.3.14. Since X is compact, we can cover it by finitely many Uj ’s,
n
Uj . Consider the function φ : X → R defined by
S
say X =
j =1

n
φ( x ) = ∑ φYj (x),
j =1
60 Chapter 2. Homotopy Theory

where φYj : X → R is the map as defined in Proposition 2.3.15 with Yj = X \ Uj ,

for j = 1, . . . , n. Since each φYj is continuous, φ is continuous. Since ( X, d) is
compact, its image φ( X ) is a compact subset of R. Since each φYj takes non-
negative values, for x ∈ X with φ( x ) = 0 we must have φYj ( x ) = 0, for all
n
j = 1, . . . , n, and hence x ∈
T
Yj , since each Yj is closed in X. But this is not
j =1
n n
Uj = ∅. Therefore, 0 ∈


Yj = X \
T S
possible because / φ( X ). Since φ is
j =1 j =1
a non-negative valued function, there is a δ > 0 such that nδ < φ( x ), for all
x ∈ X. Then for each x ∈ X, there is at least one jx ∈ {1, . . . , n} such that
φYjx ( x ) > nδ
n = δ, and hence Bδ ( x ) ⊆ X \ Yjx = U jx .

Next we lift homotopy from a base to its cover.

Lemma 2.3.16 (Glueing continuous maps). Let X and Y be two topological spaces.
Let {Uj } j∈ J be an open covering of X. Then given a family of continuous maps
 
{ f j : Uj → Y } j∈ J satisfying f j U ∩U = f k U ∩U , for all j, k ∈ J, there is a unique
j k  j k
continuous map f : X → Y such that f U = f j , for all j ∈ J.

j

Proof. Left as an exercise.

Theorem 2.3.17 (Lifting homotopy to covers). Let I := [0, 1] ⊂ R. Let f : X → Y

be a covering map. Let F : I × I → Y be a continuous map. Let y0 := F (0, 0) and fix
a point x0 ∈ f −1 (y0 ). Then there is a unique continuous map Fe : I × I → X such
that Fe(0, 0) = x0 and f ◦ Fe = F.

Proof. Since I × I is connected, uniqueness of F,

e if it exists, follows from Remark
2.3.13. We only show a construction of such a lift F.e

It is enough to show that, for each s ∈ I there is a connected open neigh-

bourhood Us ⊆ I of s ∈ I such that Fe can be constructed on Us × I. Indeed,
since {Us × I : s ∈ I } is a connected open covering of I × I and those F’s e agree
on their intersections (Us × I ) ∩ (Us0 × I ) = (Us ∩ Us0 ) × I, which are connected
(because Us0 are open intervals), uniqueness of liftings F’se defined on connected
domains ensures that they can be glued together to get a well-defined continu-
ous map Fe : I × I → X such that Fe(0, 0) = x0 and f ◦ Fe = F on I × I.

Now we construct such a lift Fe : U × I → X, for some open neighbourhood

U ⊆ I of a given point s0 ∈ I. Since F is continuous, each point (s0 , t) ∈ I × I
2.3. Covering Space 61

has an open neighbourhood Ut × ( at , bt ) ⊂ I × I such that F (Ut × ( at , bt )) is

contained in some open neighbourhood of F ((s0 , t)) ∈ Y that is evenly covered
by f . Since {s0 } × I is compact, finitely many such open subsets Ut × ( at , bt )
cover {s0 } × I. Taking intersection of those finitely many open subsets Ut ⊆ I,
we can find a single open neighbourhood U ⊂ I of s0 and a partition 0 =
t0 < t1 < · · · < tm = 1 of I = [0, 1] such that for each i ∈ {0, 1, . . . , m},
F (U × [ti , ti+1 ]) ⊆ Vi , for some open subset Vi ⊂ Y that is evenly covered by f .

By Theorem 2.3.12 (Lifting paths to a cover), we can find a unique continu-


ous function Fe : I × {0} → X with Fe(0, 0) = x0 and f ◦ Fe = F  I ×{0} . Assume
inductively that Fe has been constructed on U × [0, ti ], starting with the given Fe
on U × {0} ⊆ I × {0}. Since F (U × [ti , ti+1 ]) ⊆ Vi , and Vi is evenly covered by

f , there is an open subset Wi ⊆ X such that Fe(s0 , ti ) ∈ Wi and f W : Wi → Vi
i
is a homeomorphism. Replacing U by a smaller open neighbourhood of s0 ∈ I,
if required, we may assume that Fe(U × {ti }) ⊆ Wi ; for instance, it is enough

T 
−1
to replace U × {ti } with U × {ti } FeU ×{t } (Wi ). Then we can define Fe
i
on U × [ti , ti+1 ] to be the composition ϕ ◦ F, where ϕ : Vi → Wi is the inverse

of the homeomorphism f W : Wi → Vi . Continuing in this way, after a finite
i
number of steps, we get a continuous map Fe : U × I → X with Fe(0, 0) = x0

and f ◦ Fe = F 
U× I
, as required.

Lemma 2.3.18. Let f : X → Y be a covering map, and let γ : I → X be a continuous

map. If f ◦ γ is a constant map, so is γ.

Proof. Suppose that ( f ◦ γ)(t) = y0 , for all t ∈ I. Let V ⊆ Y be an open

neighbourhood of y0 that is evenly covered by f . Then f −1 (V ) =
F
Uα , where
α∈Λ
{Uα }α∈Λ is a family of pairwise disjoint open subsets of X with f U : Uα → V
α
a homeomorphism, for all α ∈ Λ. Since γ(t) ∈ f −1 (V ), for all t ∈ I, and
I is connected, there is a unique α0 ∈ Λ such that γ(t) ∈ Uα0 , for all t ∈ I.

Since f U is a homeomorphism, its restriction on the image of γ must be a
α0
homeomorphism; this is not possible since f ◦ γ is a constant map.

Corollary 2.3.19 (Lifting of path-homotopy). Let f : X → Y be a covering map.

Let γ0 , γ1 : I → Y be two paths in Y with γ0 (0) = γ1 (0) = y0 and γ0 (1) = γ1 (1) =
y1 . Let F : I × I → Y be a path-homotopy from γ0 to γ1 in X. If Fe : I × I → X is a
lifting of F on X, then Fe is a path-homotopy.
62 Chapter 2. Homotopy Theory

Proof. Fix a point x0 ∈ f −1 (y0 ), and let Fe : I × I → X be the lifting of F on

X with Fe(0, 0) = x0 . Then by Theorem 2.3.17, Fe is a homotopy of maps from
e0 := Fe(−, 0) to γ
γ e1 := Fe(−, 1). Let x1 := γ e0 (1) = Fe(1, 0). To show Fe is a
path-homotopy, we need to ensure that Fe(0, t) = x0 and Fe(1, t) = x1 , for all
t ∈ I. This follows from the Lemma 2.3.18 applied to the paths t 7→ Fe(0, t) and
t 7→ Fe(1, t).

Corollary 2.3.20. Let f : X → Y be a covering map. Let y0 ∈ Y and fix a point

x0 ∈ f −1 (y0 ). Then the group homomorphism f ∗ : π1 ( X, x0 ) → π1 (Y, y0 ) induced


by f is injective. The image subgroup f ∗ π1 ( X, x0 ) in π1 (Y, y0 ) consists of the
homotopy classes of loops in Y based at y0 whose lifts to X starting at x0 are loops.

Proof. Let [γ], [δ] ∈ π1 ( X, x0 ) be such that f ∗ ([γ]) = f ∗ ([δ]). Then f ◦ γ is

homotopic to f ◦ δ. Let F : I × I → Y be a path homotopy from f ◦ γ to
f ◦ δ. Then by Theorem 2.3.17 and its Corollary 2.3.19, we can lift F to a path-
homotopy Fe : I × I → X with Fe(0, 0) = x0 . By uniqueness of path-lifting (see
Theorem 2.3.12), Fe must be a path-homotopy from γ to δ (verify!). Therefore,
f ∗ is injective.


To see the second part, note that any element of f ∗ π1 ( X, x0 ) is of the form
[ f ◦ γ], for some loop γ : I → X in X based at x0 . By Theorem 2.3.12 (Path-
lifting) we can lift f ◦ γ to a path ]
f ◦ γ starting at γ(0). Then by uniqueness of
path-lifting, we have ] f ◦ γ = γ. Conversely, if δ is a loop in Y based at x0 such
that its lift δ in X is a loop in X based at x0 , then f ∗ ([δe]) = [ f ◦ δe] = [δ]. This
e
completes the proof.

Exercise 2.3.21 (Lifting of opposite path). Let f : X → Y be a covering map.

Let γ : I → Y be a path in Y from y0 to y1 . Fix a point x0 ∈ f −1 (y0 ), and let γ e
be the lift of γ in X starting at x0 . Let γ be the opposite path of γ. If γ
e is the lift
of γ in X starting at γe(1), then show that γ e = γ.
e

Exercise 2.3.22 (Lifting of product of paths). Let f : X → Y be a covering

map. Let γ, δ : I → Y be two paths in Y such that γ(1) = δ(0). Fix a point
x0 ∈ f −1 (γ(0)), and let γ e and γ ] ? δ be the liftings of the paths γ and γ ? δ,
respectively, in X starting at x0 . If δe is the lifting of δ in X starting at x1 := γ
e( 1 ) ,
show that γe ? δe = γ
] ? δ.

Lemma 2.3.23. Let f : X → Y be a covering space. If both X and Y are path-

connected, then the cardinality of the fiber f −1 (y) is independent of y ∈ Y.
2.3. Covering Space 63

Proof. Fix a point y0 ∈ Y, and a point x0 ∈ f −1 (y0 ) ⊆ X. Let G = π1 (Y, y0 )

and H = f ∗ π1 ( X, x0 ) . Let H \G := { Hg : g ∈ G } be the set of all right cosets
of H in G. Since both X and Y are path-connected, the cardinality of the set
H \G is independent of choices of y0 ∈ Y and x0 ∈ f −1 (y0 ). Therefore, to show
the cardinality of the fibers f −1 (y) is independent of y ∈ Y, it is enough to
construct a bijective map

Φ : H \G −→ f −1 (y0 ). (2.3.24)

Given a loop γ in Y based at y0 , let γ

e be the lifting of γ in X starting at x0 . Note
− 1
e(1) ∈ f (y0 ). Then we define
that, x1 := γ

Φ( H [γ]) := γ
e( 1 ) . (2.3.25)

We need to show that x1 is independent of choice of γ. Let δ be a loop in Y

based at y0 with H [γ] = H [δ]. Then [γ ? δ] = [γ][δ]−1 ∈ H = f ∗ π1 ( X, x0 ) ,

where δ is the opposite path of δ. Then by Corollary 2.3.20 the loop γ ? δ lifts to
a unique loop (^γ ? δ) in X based at x . Let e
0 δ be the lifting of δ in X starting at
x1 : = γ
e(1). Then by Exercises 2.3.22 we have γ ] ?δ = γ
e?e ]
δ. Since γ ? δ is a loop
in X based at x0 , we have δ(1) = x0 . Let η be the opposite path of δ in X. Since
e e

( f ◦ η )(t) = f (η (t)) = f eδ(1 − t)
= δ(1 − t) = δ(t), ∀ t ∈ I,

η is a lift of δ in X starting at η (0) = eδ(1) = x0 . Then by uniqueness of path-

lifting (Theorem 2.3.12) we have η = δ. e Then δe(1) = η (1) = e δ (0) = x1 .
Therefore, the map Φ in (2.3.25) is well-defined. Since X is path connected,
given any x1 ∈ f −1 (y0 ), there is a path ϕ in X from x0 to x1 . Then f ◦ ϕ is a
loop in Y based at y0 whose lift ] f ◦ ϕ starting at x0 is the unique path ϕ end-
ing at x1 = ϕ(1). Therefore, Φ is surjective. Let [γ], [δ] ∈ π1 (Y, y0 ) = G
be such that Φ( H [γ]) = Φ( H [δ]). Let γ e and δe be the lifts of γ and δ, re-
spectively, in X starting at x0 . Let δe be the opposite path of δe in X. Since
Φ( H [γ]) = Φ( H [δ]), we have γ e(1) = δe(1), and hence γ e ? δe is a loop in X based


at x0 . Since f ◦ γ e ? δe = γ ◦ δ, by uniqueness of path-lifting and Corollary


2.3.20, we conclude that [γ ? δ] ∈ f ∗ π1 ( X, x0 ) = H. Therefore, H [γ] = H [δ],
and hence Φ is injective. Therefore, Φ : H \G → f −1 (y0 ) is a bijection.
64 Chapter 2. Homotopy Theory

Exercise 2.3.26. Give an example to show that the Lemma 2.3.23 fails if X and
Y are not path-connected.

Theorem 2.3.27 (General Lifting Criterion). Let f : ( X, x0 ) → (Y, y0 ) be a cover-

ing map. Let T be a path-connected and locally path-connected space. A continuous
map g : ( T, t0 ) → (Y, y0 ) lifts to a continuous map ge : ( T, t0 ) → ( X, x0 ) if and only



if g∗ π1 ( T, t0 ) ⊆ f ∗ π1 ( X, x0 ) . Note that, such a lift ge of g, if it exists, is unique
by Lemma 2.3.13.

Proof. If g lifts to a continuous map ge : ( T, t0 ) → ( X, x0 ) such that f ◦ ge = g,

then g∗ π1 ( T, t0 ) = f ∗ ge∗ (π1 ( T, t0 )) ⊆ f ∗ π1 ( X, x0 ) .



To see the converse, suppose that g∗ π1 ( T, t0 ) ⊆ f ∗ π1 ( X, x0 ) . Since T is
path-connected, given a point t1 ∈ T, there is a path γ : I → T with γ(0) = t0
and γ(1) = t1 . Then g ◦ γ : I → Y is a path in Y from g(t0 ) = y0 to y1 :=
g(t1 ) = ( g ◦ γ)(1). Since f : ( X, x0 ) → (Y, y0 ) is a covering map, by Theorem
2.3.12 (Path-lifting) the path g ◦ γ lifts to a unique path g] ◦ γ in X starting at x0 .
Define a map
ge : T → X (2.3.28)

by sending t1 to x1 := g] ◦ γ(1) ∈ X. To show the map ge is independent of

choice of a path γ in T from t0 to t1 , note that given any path δ : I → T from


t0 to t1 , the product path γ ? δ is a loop in T based at t0 . Since g∗ π1 ( T, t0 ) ⊆


f ∗ π1 ( X, x0 ) , by the second part of the Corollary 2.3.20 the loop g ◦ (γ ? δ) =
( g ◦ γ) ? ( g ◦ δ) lifts to a unique loop, say ϕ, in X based at x . Let (^
0 g ◦ δ) be the
lifting of g ◦ δ in X starting at x1 := g] ◦ γ(1). Then by Exercises 2.3.22 we have
ϕ = (^ g ◦ γ) ? ( g ◦ δ). Since ϕ is a loop in X based at x0 , we have (^
^ g ◦ δ)(1) = x0 .
Let η be the opposite path of (^
g ◦ δ). Since

( f ◦ η )(t) = f ((^
g ◦ δ)(1 − t))
= ( g ◦ δ)(1 − t)
= ( g ◦ δ)(t), ∀ t ∈ I,

and η (0) = (^ g ◦ δ)(1) = x0 , by uniqueness of path-lifting, we have η = (^g ◦ δ ).

Then (^ g ◦ δ)(1) = η (1) = (^ g ◦ δ)(0) = (^
g ◦ γ)(1) = x1 . Therefore, the map
ge in (2.3.28) is well-defined. It follows from the construction of ge that f ◦ ge =
2.3. Covering Space 65

g. It remains to show that ge is continuous. Here we need to use local path-

connectedness of T.

Fix a point t1 ∈ T and let y1 = g(t1 ) ∈ Y and x1 := ge(t1 ) ∈ f −1 (y1 ). Since

f is a covering map, there is an open neighbourhood U ⊆ Y of y1 and an open
neighbourhood U e ⊆ X of x1 such that


e →U
f Ue : U (2.3.29)

is a homeomorphism. Since T is locally path-connected and g is continuous,

there is a path-connected neighbourhood V ⊆ T of t1 such that g(V ) ⊆ U. To
show ge : T → X continuous, it is enough to show that ge(V ) ⊆ U. e Given t0 ∈ V,
choose a path α inside V joining t1 to t0 . Then γ ? α is a path in T joining t0
to t0 , and its image g ◦ (γ ? α) has a lifting, say β, in X starting at x0 . Let e α :=

s ◦ ( g ◦ α), where s : U → U e is the inverse of the homeomorphism f e given in

U
(2.3.29). Since γ e(1) = (s ◦ g ◦ α)(0), by uniqueness of path-lifting, β coincides
with γ e ? (s ◦ g ◦ α). Then ge(t0 ) = β(1) = (γ
e ? (s ◦ g ◦ α))(1) = (s ◦ g ◦ α)(1) ∈ U.
e
Therefore, ge(V ) ⊆ U, e and hence ge is continuous.

2.3.2 Fundamental group of S1

Now we are in a position to compute fundamental group of the unit circle

S1 := {z ∈ C : |z| = 1} = {( x, y) ∈ R2 : x2 + y2 = 1}.

Assuming that the reader has forgotten the statement of Theorem 2.3.1 by now,
let’s recall it once again.

Theorem 2.3.1. The fundamental group π1 (S1 , 1) of the unit circle S1 with the base
point 1 ∈ S1 is isomorphic to the infinite cyclic group Z generated by the loop ω : I →
S1 defined by ω (t) = e2πit , for all t ∈ I = [0, 1].

Proof. Let γ : I → S1 be a loop based at x0 = 1 ∈ S1 . Since

p : R → S1 , t 7→ e2πit

is a covering map (c.f. Example 2.3.6 (i)), there is a unique continuous map
e : I → R such that γ
γ e = γ. Since p−1 (γ(1)) = Z, the path γ
e(0) = 0 and p ◦ γ e
66 Chapter 2. Homotopy Theory

ends at some integer, say n. Note that, we have a path

e n : I → R, s 7→ ns,
ω

starting at 0 and ending at n. Clearly the path γ

e is homotopic to ω
e n by the
linear homotopy

F : I × I → R, (s, t) 7→ (1 − t)γ
e( s ) + t ω
e n ( s ).

Then the composition p ◦ F : I × I → S1 is a homotopy from γ to ωn , where

ωn : I → S1 is the loop based at 1 ∈ S1 defined by

ωn (s) = e2πins , ∀ s ∈ I.

Therefore, [γ] = [ωn ] in π1 (S1 , 1).

Define a map
ϕ : Z −→ π1 (S1 , 1), n 7→ [ωn ].

It follows from the above construction that ϕ is surjective. To show that ϕ is a

group homomorphism, we need to show that ωm ? ωn ' ωm+n , for all m, n ∈ Z.
To see this, consider the “translation by m” map

τm : R → R, x 7→ x + m.

Note that τm ◦ ω e n is a path in R starting at m and ending at m + n, and hence

the path ω e m ? (τm ◦ ωe n ) in R starts at 0 and ends at m + n. Then it follows
from the first paragraph that p ◦ (ω e m ? (τm ◦ ω
e n )) is homotopic to ωm+n . Since
p ◦ (ω
e m ? (τm ◦ ω
e n )) = ωm ? ωn , we conclude that ϕ is a group homomorphism.

To show that ϕ is injective, it is enough to show if a loop γ : I → S1 based

at 1 is homotopic to both ωn and ωm , for some m, n ∈ Z, then m = n. Indeed,
if γ ' ωm and γ ' ωn , then ωm ' ωn by Lemma 2.1.4. Let G : I × I → S1 be a
homotopy from ωm to ωn in S1 . By Theorem 2.3.17 there is a unique continuous
map G e : I × I → R such that p ◦ G e = G and G e(0, 0) = 0. Then by uniqueness
 
of path lifting (c.f. Theorem 2.3.12) we have G e
{0}× I
= ω
e n and G
e
{1}×I
=ω em.
: I → R t∈ I is a homotopy of paths, the end points G
 
Since G e
{t}× I
e
{t}× I
(1)
 
are independent of t. Thus, m = G e
{0}× I
(1) = G e
{1}× I
(1) = n, and hence ϕ is
injective. This completes the proof.
2.3. Covering Space 67

2.3.3 Fundamental group of Sn , for n ≥ 2

In this subsection we show that Sn is simply connected, for n ≥ 2. First we

need the following.

Lemma 2.3.30. Let ( X, x0 ) be a pointed topological space. Let {Uα }α∈Λ be an open
cover of X such that

1. each Uα is path-connected,

2. x0 ∈ Uα , for all α ∈ Λ,

3. Uα ∩ Uβ is path-connected, for all α, β ∈ Λ.

Then any loop in X based at x0 is homotopic to a finite product of loops each of which
is contained in a single Uα , for finitely many α’s.

Proof. Let γ : I → X be a loop based at x0 . Since γ is continuous, each s ∈ I

is contained in an open neighbourhood Vs := (s − δs , s + δs ) ⊆ I of s such
that γ(V s ) ⊆ Uαs , for some αs ∈ Λ. Since I is compact, we can choose finitely
many such open neighbourhoods Vs ’s to cover I. Thus we get a finite partition
0 = s0 < s1 < · · · < sm = 1 of I = [0, 1] such that γ([s j−1 , s j ]) ⊆ Uα j , for some
α j ∈ Λ, for all j = 1, . . . , m. Therefore, the restriction

γ j : = γ [s : [s j−1 , s j ] → Uα j ⊆ X
j −1 , s j ]

is a path in Uα j , for each j = 1, . . . , m, and that γ = γ1 ? · · · ? γm . Since Uj ∩ Uj+1

F IGURE 2.11

is path-connected, we may choose a path ηi in Uα j ∩ Uα j+1 from the base point

68 Chapter 2. Homotopy Theory

x0 to the point γ(s j ) ∈ Uα j ∩ Uα j+1 , for all j (see Figure 2.11). Denote by η j the
opposite path of η j , for all j (see definition (2.2.23) in §2.2.3). Then the product
loop

(γ1 ? η1 ) ? (η1 ? γ2 ? η2 ) ? (η2 ? γ3 ? η3 ) ? · · · ? (ηm−1 ? γm ) (2.3.31)

is homotopic to γ (see Exercise 2.2.24). Clearly this loop is a composition of

the loops γ1 ? η1 , η1 ? γ2 ? η2 , η2 ? γ3 ? η3 , · · · , ηm−1 ? γm based at x0 , each lying
inside a single Uα j , for all j = 1, . . . , m. This completes the proof.

Exercise 2.3.32. Fix an integer n ≥ 1.

(i) For any x0 ∈ Sn , show that Sn \ { x0 } is homeomorphic to Rn .

(ii) For a pair of antipodal points x1 , x2 ∈ Sn , let Uj := Sn \ { x j }, for j = 1, 2.

Show that U1 ∩ U2 is homeomorphic to Sn−1 × R.

Proposition 2.3.33. For an integer n ≥ 2, we have π1 (Sn ) = {1}.

Proof. Fix a pair of antipodal points x1 , x2 in Sn . Then we have two open subsets
U1 = Sn \ { x1 } and U2 = Sn \ { x2 } each homeomorphic to Rn . Clearly Sn =
U1 ∪ U2 and U1 ∩ U2 is homeomorphic to Sn−1 × R. Then by Exercise 2.3.32 we
have U1 ∩ U2 is homeomorphic to Sn−1 × R, which is path-connected because
n ≥ 2. Fix a base point x0 ∈ U1 ∩ U2 . Let γ be a loop in Sn based at x0 . Then by
Lemma 2.3.30 γ is homotopic to a product of finitely many loops in Sn based
at x0 each of which are contained in either U1 or U2 . Since both U1 and U2 are
homeomorphic to Rn by Exercise 2.3.32, we have π1 (Uj ) = π1 (Rn ) = {1}, for
j = 1, 2. Therefore, γ is homotopic to a finite product of loops based at x0 each
of which are null-homotopic, and hence γ is null-homotopic.

Corollary 2.3.34. Sn is simply connected, for n ≥ 2.

Exercise 2.3.35. For a point x0 ∈ Rn , show that the space Rn \ { x0 } is homeo-

morphic to Sn−1 × R.

Corollary 2.3.36. R2 is not homeomorphic to Rn , for n 6= 2.

Proof. If possible let f : R2 → Rn be a homeomorphism. For n = 1, since

R2 \ {0} is path-connected while R \ { f (0)} is disconnected, there is no such
homeomorphism in this case. Suppose that n > 2. In this case, we cannot
2.3. Covering Space 69

distinguish R2 \ {0} with Rn \ { f (0)} in terms of number of path-components;

but we can distinguish them by their fundamental groups.

Since for any point x ∈ Rn the space Rn \ { x } is homeomorphic to Sn−1 × R

by Exercise 2.3.35, we have

π1 (Rn \ { x }) ∼
= π 1 ( S n −1 × R )
∼
= π 1 ( S n −1 ) × π 1 (R )
∼
= π 1 ( S n −1 ),

because π1 (R) is trivial. Since π1 (S1 ) ∼

= Z by Theorem 2.3.1 while π1 (Sn−1 ) ∼
=
{1}, for n > 2, by Proposition 2.3.33, such a homeomorphism cannot exists.

Remark 2.3.37. A more general result that Rm is homeomorphic to Rn if and

only if m = n can be proved in a similar fashion using higher homotopy groups
or homology groups. In fact, using homology groups one can show that non-
empty open subsets of Rm and Rn can be homeomorphic if and only if m = n.

2.3.4 Some applications

Theorem 2.3.38 (Fundamental theorem of algebra). Every non-constant polyno-

mial with coefficients from C has a root in C.

Proof. Take a non-constant polynomial p(z) ∈ C[z]. Diving p(z) by its leading
coefficient, if required, we may assume that

p ( z ) = z n + a 1 z n −1 + · · · + a n ∈ C [ z ] .

If p(z) has no roots in C, then for each real number r ≥ 0, the map γr : I →
S1 ⊂ C defined by

p(re2πis )/p(r )
γr (s) := , ∀ s ∈ I, (2.3.39)
| p(re2πis )/p(r )|

is a loop in the unit circle S1 := {z ∈ C : |z| = 1} with the base point 1 ∈ C. As r

varies, the collection {γr }r≥0 defines a homotopy of loops in S1 based at 1. Since
γ0 is the constant loop 1 in S1 , we see that the homotopy class [γr ] ∈ π1 (S1 , 1)
is trivial, for all r ≥ 0.
70 Chapter 2. Homotopy Theory

Choose any r ∈ R with r > max{1, | a1 | + · · · + | an |}. Then for |z| = r we

have

| z n | = r n = r · r n −1 > | a 1 | + · · · + | a n | | z n −1 |

≥ | a 1 z n −1 + · · · + a n |

From this inequality, it follows that for each t ∈ [0, 1], the polynomial

p t ( z ) : = z n + t ( a 1 z n −1 + · · · + a n )

has no roots on the circle |z| = r. Replacing p(z) with pt (z) in the expression
of γr in (2.3.39) and letting t vary from 1 to 0, we get a homotopy from the loop
γr to the loop
ωn : I → S1 , s 7→ e2πins .

Since the loop ωn represents n times a generator of the infinite cyclic group
π1 ( S1 , 1) ∼
= Z, and that [ωn ] = [γt ] = 0, we must have n = 0. Thus the only
polynomials without roots in C are constants.

Definition 2.3.40. A deformation retraction of X onto its subspace A is a contin-

uous map F : X × I → X such that the associated family of continuous maps
n  o
f t := F  X ×{t} : X → X
t∈ I

obtained by restricting F on the slices X × {t} ,→ X × I, for each t ∈ I, satisfies


f 0 = IdX , f 1 ( X ) = A, and f t  A = Id A , ∀ t ∈ I. In this case, we say that A is a
deformation retract of X.

Example 2.3.41. (i) Let D = {reiθ ∈ C : 0 < r ≤ 1, 0 ≤ θ < 2π } be the

punctured disk of radius 1 in the plane C, and let S1 = {z ∈ C : |z| =
1} ⊆ X be the unit circle. For each t ∈ I = [0, 1], we define a map

f t : D −→ D

by sending reiθ ∈ D to t + (1 − t)r eiθ ∈ D. It is easy to verify that { f t }t∈ I

is a family of continuous maps from D into itself, and satisfies f 0 = IdD ,

f 1 ( D ) = S1 and f t S1 = IdS1 . Therefore, { f t }t∈ I is a deformation retraction


of D onto S1 .
2.3. Covering Space 71

(ii) Let X be the Möbius strip (see Figure 2.12) and A ⊂ X be the central
simple loop of X. Then there is a deformation retraction of X onto A.

F IGURE 2.12: Möbius strip

Definition 2.3.42. A retraction of X onto a subspace A ⊂ X is a continuous map


f : X → X such that f ( X ) = A and f  A = Id A . A subspace A ⊆ X is said to be
a retract of X if there is a retraction of X onto A.

Note that a retraction f : X → X of X onto a subspace A ⊆ X can be charac-

terized by its property f ◦ f = f , and hence we can think of it as a topological
analogue of a projection operator in algebra.

Lemma 2.3.43. If A ⊆ X is a retract of X, for any a0 ∈ A the homomorphism of

fundamental groups
ι ∗ : π1 ( A, a0 ) → π1 ( X, a0 ),

induced by the inclusion map ι : A ,→ X, is injective.

Proof. Let f : X → X be a retraction of X onto A. Then f ◦ ι = Id A , the

identity map of A. Then by Proposition 2.2.12 and Remark 2.2.13 we have
f ∗ ◦ ι ∗ = Idπ1 ( A,a0 ) . Thus ι ∗ admits a left inverse, and hence is injective.

Proposition 2.3.44. If A ⊆ X is a deformation retract of X, then X is homotopically

equivalent to A (see Definition 2.1.9).

Proof. Let F : X × I → X be a deformation retract of X onto its subspace A.

Since
f 0 : X → X, x 7→ F ( x, 0)

is the identity map IdX : X → X, and

f 1 : X → X, x 7→ F ( x, 1)
72 Chapter 2. Homotopy Theory

is a retraction of X onto A, we conclude that F is a homotopy from IdX to a

retraction of X onto A. Since f 1 ◦ ι = Id A and ι ◦ f 1 is homotopic to the identity
map of X, we conclude that X and A are homotopically equivalent.

Corollary 2.3.45. If A ⊆ X is a deformation retract of X, then for any a0 ∈ A we

have an isomorphism of fundamental groups π1 ( A, a0 ) ∼
= π1 ( X, a0 ).

Proof. Follows from Lemma 2.2.16.

Remark 2.3.46. Note that the constant map X → { x0 } ⊆ X being continuous,

every space X admits a retraction onto a point of it. However, the next Propo-
sition 2.3.47 and Lemma 2.3.48 produce examples of topological spaces that do
not admit any deformation retract onto a point of it.

Proposition 2.3.47. If there is a deformation retract of X onto a point x0 ∈ X, then X

is path connected.

Proof. Let F : X × I → X be a deformation retract of X onto a point x0 ∈ X.

Since for any point x ∈ X, the continuous map

φx : I → X, t 7→ F ( x, t)

is a path joining F ( x, 0) = x and F ( x, 1) = x0 , X is path connected.

Lemma 2.3.48. If A ⊆ X is a deformation retract of X, then for any a0 ∈ A, the

homomorphism of fundamental groups ι ∗ : π1 ( A, a0 ) → π1 ( X, a0 ) induced by the
inclusion map ι : A ,→ X is an isomorphism.

Proof. Let F : X × I → X be a deformation retraction of X onto X. Then


f 1 := F  X ×{1} : X → X is a retraction of X onto A. Then by Lemma 2.3.43
the homomorphism ι ∗ : π1 ( A, a0 ) → π1 ( X, a0 ) is injective. To show ι ∗ is an
isomorphism, it enough to show that it is surjective. Note that, given any loop
γ : I → X in X based at a0 , the composite map

γ×Id I F
G : I × I −→ X × I −→ X
 
is a path-homotopy from G  I ×{0} = γ to a loop g := G  I ×{1} : I → A based at
a0 . Thus, ι ∗ ([ g]) = [ g] = [γ], and hence ι ∗ is surjective.
2.3. Covering Space 73

Remark 2.3.49. The notion of deformation retraction of a space X onto a sub-

space A ⊆ X is a way to continuously deform X onto A in a very strong sense,
while the notion of homotopy equivalence seems to be a weaker notion of be-
ing able to deform a space into another space. However, if two spaces X and
Y are homotopically equivalent, then there is a space Z such that both X and Y
are deformation retracts of Z. Such a space Z can be constructed as a mapping
cylinder


M f := ( X × I ) t Y ( x, 1) ∼ f ( x )

of a homotopy equivalence f : X → Y. We shall not go into details for its proof

in this course.

Exercise 2.3.50. Show that the unit circle S1 = {z ∈ C : |z| = 1} do not admit
any deformation retraction onto a point of it.

Exercise 2.3.51. Show that π1 (R2 \ {(0, 0)}) ∼

= Z.

For an integer n ≥ 1, let

n
n n
D := {( x1 , . . . , xn ) ∈ R : ∑ x2j ≤ 1}
j =1

be the closed unit disk in Rn . Its boundary ∂D n is the unit sphere in Rn given by

n
Sn−1 := {( x1 , . . . , xn ) ∈ Rn : ∑ x2j = 1}.
j =1

Theorem 2.3.52 (Brouwer’s fixed point theorem). Every continuous map f : D2 →

D2 has a fixed point.

Proof. Suppose on the contrary that f : D2 → D2 has no fixed point, i.e., f ( x ) 6=

x, ∀ x ∈ D2 . Then for each x ∈ D2 , the ray in R2 starting at f ( x ) and passing
through x hits a unique point, say r ( x ) ∈ S1 . This defines a map r : D2 → S1 .
74 Chapter 2. Homotopy Theory

Since f is continuous, small perturbations of x produce small perturbations of

f ( x ), and hence small perturbations of the ray starting from f ( x ) and passing
through x, it follows that the function x 7→ r ( x ) is continuous. Explicit proof
of continuity could be given by writing down the explicit expression for r ( x )
in terms of f ( x ). Note that r ( x ) = x, for all x ∈ S1 . Therefore, r : D2 → S1
is a retraction of D2 onto its subspace S1 = ∂D2 . Then by Lemma 2.3.43 the
homomorphism of fundamental groups

ι ∗ : π1 (S1 , (1, 0)) −→ π1 ( D2 , (1, 0))

induced by the inclusion map ι : S1 ,→ D2 , is injective. Since π1 (S1 , (1, 0)) ∼

=Z
and π1 ( D2 , (1, 0)) is trivial, we get a contradiction.

Remark 2.3.53. The corresponding statement for Brouwer’s fixed point theo-
rem holds, more generally, for a closed unit disk D n ⊂ Rn , for all n ≥ 2. If
time permits, we shall give a proof of it using homology. However, the original
proof of it, due to Brouwer, neither uses homology nor uses homotopy groups,
which was not invented at that time. Instead, Brouwer’s proof uses the notion
of degree of maps Sn → Sn , which could be defined later using homology, but
Brouwer defined it more directly in a geometric way.

Definition 2.3.54. For x = ( x1 , . . . , xn+1 ) ∈ Sn , we define its antipodal point to

be the point − x := (− x1 , . . . , − xn+1 ) ∈ Sn .

Theorem 2.3.55 (Borsuk-Ulam). Let n ∈ {1, 2}. Then for every continuous map
f : Sn → Rn , there is a pair of antipodal points x and − x in Sn with f ( x ) = f (− x ).

Proof. The case n = 1 is easy. Indeed, since the function

g : S1 → R, x 7→ f ( x ) − f (− x )

changes its sign after the point x ∈ S1 moves half way along the circle S1 , there
must be a pint x ∈ S1 such that f ( x ) = f (− x ).

Assume that n = 2. We use the same technique used to compute the fun-
damental group of S1 . Suppose on the contrary that there is a continuous map
f : S2 → R2 such that f ( x ) 6= f (− x ), for all x ∈ S2 . Then we can define a map
g : S2 → R2 by
f ( x ) − f (− x )
g( x ) := , ∀ x ∈ S2 ,
|| f ( x ) − f (− x )||
2.3. Covering Space 75

q
where ||(y1 , y2 )|| := y21 + y22 is the norm of (y1 , y2 ) ∈ R2 . Since || g( x )|| = 1,
the image of the map g lands inside S1 ⊂ R2 . Note that the map g : S2 → S1 is
continuous. Define a loop η : I = [0, 1] → S2 by

η (s) = (cos 2πs, sin 2πs, 0), ∀ s ∈ I. (2.3.56)

Then η circles around the equator of the sphere S2 ⊂ R3 . Let h : I → S1 be the

composite map h := g ◦ η.

η g
h : I −→ S2 −→ S1 .

Since g( x ) = − g(− x ), we have

1
h(s + ) = −h(s), ∀ s ∈ [0, 1/2]. (2.3.57)
2

Now consider the covering map

p : R → S1 , s 7→ e2πis = (cos 2πs, sin 2πs).

h : I → R starting at
Lift the loop h : I → S1 to this cover to get a unique path e
0 ∈ R (see Theorem 2.3.12). Then it follows from the relation (2.3.57) that

1 q(s)
h(s + ) = e
e h(s) + , (2.3.58)
2 2

for some odd integer q(s) depending on s ∈ [0, 12 ]. Since e

h is continuous, it
follows from the equation (2.3.58) that the map

I → R, s 7→ q(s),

is continuous on [0, 12 ]. Since q is a discrete function taking values in odd in-

tegers, we must have q(s) = q, for some odd integer q, for all s ∈ [0, 12 ]. In
particular, putting s = 1/2 and 0 in (2.3.58) we have

q
h (1) = e
e h(1/2) + = e
h(0) + q.
2

This means that the loop h represents q times a generator of π1 (S1 ). Since q is
an odd integer, h cannot be null homotopic. But this cannot happen because
the loop η : I → S2 being null-homotopic, the loop h := g ◦ η : I → S2 → S1
76 Chapter 2. Homotopy Theory

should be null-homotopic. Thus we get a contradiction. This completes the

proof.

Remark 2.3.59. (i) Borsuk-Ulam theorem (Theorem 2.3.55) holds for all inte-
ger n ≥ 1. A general proof could be given using homology theory later.

(ii) Theorem 2.3.55 says that there is no one-to-one continuous map from Sn
into Rn . As a result, Sn cannot be homeomorphic to a subspace of Rn .

2.4 Galois theory for covering spaces

2.4.1 Universal cover

Since we shall work with paths in X, and a locally path-connected space is

connected if and only if it is path-connected, and path-connected components
of X are the same as connected components of X, there is no harm in assum-
ing that X is connected or equivalently path-connected. Unless explicitly men-
tioned, in this section, we always assume that X is path-connected and locally
path-connected.

Proposition 2.4.1. Let X be a connected and locally path-connected topological space.

Fix a point x0 ∈ X. Let p : ( X,e xe0 ) → ( X, x0 ) be a simply connected covering. Then
for any connected covering f : (Y, y0 ) → ( X, x0 ), there is a unique continuous map
e xe0 ) → (Y, y0 ) such that p ◦ F = f .
F : ( X,

( X,
e xe0 )
F
y
(Y, y0 ) p

f % 
( X, x0 )

Proof. Since X is locally path-connected and Xe is a simply connected covering

of X, Xe is path-connected and locally path-connected. Since π1 ( X,
e xe0 ) is trivial
and f : (Y, y0 ) → ( X, x0 ) is a covering map, by general lifting criterion (see
Theorem 2.3.27) there is a unique continuous map F : ( X, e xe0 ) → (Y, y0 ) such
that f ◦ F = p.
2.4. Galois theory for covering spaces 77

Proposition 2.4.2. Let ( X, x0 ) be a locally path-connected and path-connected topo-

logical space. Let p1 : ( Xf1 , xe1 ) → ( X, x0 ) and p2 : ( X f2 , xe2 ) → ( X, x0 ) be two
simply connected covering spaces of ( X, x0 ). Then there is a unique homeomorphism
of pointed topological spaces F : ( Xf1 , xe1 ) → ( X
f2 , xe2 ) such that p2 ◦ F = p1 .

F /
(X
f1 , xe1 ) (X
f2 , xe2 )

p1 p2
% y
( X, x0 )

Proof. Follows from Proposition 2.4.1.

Definition 2.4.3. A simply connected covering space of a path-connected lo-

cally path-connected topological space ( X, x0 ) is called the universal cover of
( X, x0 ). This name is due to its universal property (c.f. Proposition 2.4.1) and
uniqueness upto a unique homeomorphism (c.f. Proposition 2.4.2).

It is not yet clear if universal cover of a path-connected locally path-connected

topological space exists or not, however if it exists, it is unique up to a unique
homeomorphism of pointed topological space by Proposition 2.4.2. The fol-
lowing Lemma 2.4.4 gives a necessary condition on ( X, x0 ) for existence of a
universal covering space.

e xe0 ) → ( X, x0 ) be the universal cover of ( X, x0 ). Then

Lemma 2.4.4. Let p : ( X,
each point x ∈ X has a path-connected open neighbourhood U ⊆ X such that the
homomorphism of fundamental groups ι ∗ : π1 (U, x ) → π1 ( X, x ), induced by the
inclusion map ι : U ,→ X, is trivial.

Proof. Fix x ∈ X. Then there is a path-connected open neighbourhood U ⊆

X which is evenly covered by the covering map p. Let U e ⊆ Xe be the path-

connected open subset such that pUe : U e → U is a homeomorphism. Let γ

be a loop in U based at x. Using the homeomorphism pUe , we can lift it to a
loop γ e based at the point xe ∈ U
e in X e ∩ p−1 ( x ). Since Xe is simply-connected,
we have a path-homotopy F : I × I → X e from γ e to the constant loop c xe at xe in
e Composing F with p we get a path-homotopy p ◦ F from γ to the constant
X.
loop c x at x in X. This shows that the homomorphism ι ∗ : π1 (U, x ) → π1 ( X, x )
induced by the inclusion map ι : U ,→ X is trivial.
78 Chapter 2. Homotopy Theory

Definition 2.4.5. A path-connected and locally path-connected topological space

X is said to be semi-locally simply connected if each point x ∈ X has a path-
connected open neighbourhood U ⊆ X such that the homomorphism of fun-
damental groups ι ∗ : π1 (U, x ) → π1 ( X, x ), induced by the inclusion map
ι : U ,→ X, is trivial.

2.4.2 Construction of universal cover

The following theorem shows that the condition on ( X, x0 ) for existence of

its universal covering space given in Lemma 2.4.4 is, in fact, sufficient.

Theorem 2.4.6. Let X be a path-connected, locally path-connected topological space.

Fix a point x0 ∈ X. Then a simply connected covering space of ( X, x0 ) exists if and
only if X is semi-locally simply connected.

Proof. If a simply connected covering space for X exists, then X is semi-locally

simply connected by Lemma 2.4.4.

Suppose that X is semi-locally simply connected. We give an explicit con-

struction of a simply connected covering space of X. Note that, if p : ( X, e xe0 ) →
( X, x0 ) is a simply connected covering space for ( X, x0 ), then for each xe ∈ X, e
there is a unique path-homotopy class of paths in X e from xe0 to xe (see Corollary
2.2.33). Thus, points of X e can be thought of as homotopy classes of paths in X e
starting at xe0 , and hence can be thought of as the homotopy classes of paths in
X starting at x0 thanks to the homotopy lifting property. This motivates us to
construct the underlined set of points of Xe as

e := {[γ] : γ is a path in X starting at x0 },

X

where [γ] denotes the path-homotopy class of a path γ in X. Define

e→X
p:X (2.4.7)

by sending a [γ] ∈ X
e to the end point γ(1) ∈ X of γ; this map is well-defined
because of the definition of path-homotopy (see Definition 2.2.1). Since X is
path-connected, given any x1 ∈ X there is a path γ in X with γ(0) = x0 and
γ(1) = x1 . Then [γ] ∈ X e with p([γ]) = x1 . Thus, p is surjective. If we set
2.4. Galois theory for covering spaces 79

xe0 ∈ X e to be the path-homotopy class of the constant path c x : I → X given by

0

c x0 (t) = x0 , ∀ t ∈ I, then p( xe0 ) = x0 .

It remains to give a suitable topology on X e xe0 ) → ( X, x0 ) a

e to make p : ( X,
simply connected covering space of ( X, x0 ). Let

ι
U := {V ,→ X | V is a path-connected open subset of X such that
the homomorphism ι ∗ : π1 (V ) → π1 ( X ) is trivial }.

Note that, if the homomorphism ι ∗ : π1 (V, x ) → π1 ( X, x ), induced by the

inclusion map ι : V ,→ X, is trivial for some x ∈ V, then it is trivial for all
points of V, whenever V is path-connected. Moreover, if U and V are two
path-connected open subsets of X with V ⊆ U and U ∈ U , then it follows
from the following commutative diagram

ιV,U ∗
π1 (V ) / π 1 (U )

ιV ∗
$ z ιU ∗
π1 ( X )

that V ∈ U , where ιU : U ,→ X, ιV : V ,→ X and ιV,U : V ,→ U are inclusion

maps. Since X is locally-path-connected, path-connected and semi-locally sim-
ply connected, now it follows that U is a basis for the topology on X (verify!).

We now use the collection U to construct a collection B of subsets of X e

which forms a basis for the desired topology on X.e Given U ∈ U and a path γ
in X starting at x0 and ending at a point in U, consider the subset

U[γ] := {[γ ? η ] : η is a path in U starting at γ(1)} ⊆ X.

e

Note that, if γ is path-homotopic to γ0 in X, then γ(1) = γ0 (1), and hence for

any path η in U starting at γ(1) = γ0 (1), we have [γ ? η ] = [γ0 ? η ]. Therefore,
the subset U[γ] ⊆ X e depends only on U and the path-homotopy class of γ in X.

Observation 1: The restriction map


p U : U[γ] → U (2.4.8)
[γ]

is bijective. Indeed, it is surjective because U is path-connected. To see it is

injective, note that if p([γ ? η ]) = p([γ ? η 0 ]), then η (1) = η 0 (1) and so the
80 Chapter 2. Homotopy Theory

loop η ? η 0 is path-homotopic to the constant path cη (0) inside X, because the

homomorphism ι ∗ : π1 (U ) → π1 ( X ) is trivial. Then it follows that [γ ? η ] =
[γ ? η 0 ]. Therefore, the restriction of p on U[γ] (see (2.4.8)) is injective, and hence
is bijective.

Observation 2: Given U ∈ U and any two paths γ and δ in X with γ(0) =

δ(0) = x0 and γ(1), δ(1) ∈ U, if [δ] ∈ U[γ] , then we must have U[γ] = U[δ] .
Indeed, if [δ] ∈ U[γ] , then [δ] = [γ ? η ], for some path η in U with η (0) = γ(1).
Then for any path α in U with α(0) = δ(1), we have [δ ? α] = [(γ ? η ) ? α] =
[γ ? (η ? α)] ∈ U[γ] . Thus U[δ] ⊆ U[γ] . Conversely, given any [γ ? α] ∈ U[γ] we
have [γ ? α] = [γ ? η ? η ? α] = [δ ? (η ? α)] ∈ U[δ] , which shows that U[γ] ⊆ U[δ] .
Therefore, we conclude that U[γ] = U[δ] if [δ] ∈ U[γ] .

Now we use the above two observations to show that the collection

B := {U[γ] : U ∈ U and γ is a path in X with γ(0) = x0 and γ(1) ∈ U }

forms a basis for a topology on X.e Note that, X being path-connected, we have
U[γ] . To check the second property for B to be a basis for a topology
S
e=
X
U[γ] ∈B
e suppose that we are given two objects U[γ] , V[δ] ∈ B and an element
on X,

[α] ∈ U[γ] ∩ V[δ] . (2.4.9)

Now U, V ∈ U , and γ and δ are paths in X with γ(0) = δ(0) = x0 and γ(1) ∈
U, δ(1) ∈ V. We claim that

U[γ] = U[α] and V[δ] = V[α] . (2.4.10)

Since [α] ∈ U[γ] ∩ V[δ] , we have [α] = [γ ? η ] = [δ ? η 0 ], for some paths η and
η 0 in U and V respectively, with η (0) = γ(1) and η 0 (0) = δ(1). Since γ ? η
is path-homotopic to δ ? η 0 , both of them have the same end point, and hence
α(1) = η (1) = η 0 (1) ∈ U ∩ V. Then the claim in (2.4.10) follows from the
Observation 2. Since U is a basis for the topology on X, and α(1) ∈ U ∩ V, there
is an object W ∈ U such that α(1) ∈ W and W ⊆ U ∩ V. Since [α] ∈ U[γ] ∩ V[δ] ,
the argument given in Observation 2 shows that

W[α] ⊆ U[α] ∩ V[α] = U[γ] ∩ V[δ] ,

2.4. Galois theory for covering spaces 81

where the equality of sets on the right side is by (2.4.10). Clearly [α] ∈ W[α] .
Therefore, B is a basis for a topology on X.
e Give Xe the topology generated by
this basis B.

e → X in (2.4.7) is a covering map and that

Now it remains to show that p : X
e is simply connected. We first show that, for each U[γ] ∈ B, the restriction
X
map

pU : U[γ] → U
[γ]

is a homeomorphism. We already have shown that pU is bijective. Note that,
[γ]
for any V[0δ] ∈ B with V[0δ] ⊆ U[γ] we have p(V[0δ] ) = V0 ⊆ U. Since both U
and B are basis for the topologies of X and X,
e respectively, this shows that the
 
restriction map pU is open. To show that pU is continuous, it suffices to
[γ] [γ]
show that for any V ∈ U with V ⊆ U, we have p−1 (V ) ∩ U[γ] = V[γ] . Indeed,
for any [α] ∈ p−1 (V ) ∩ U[γ] , we have α(1) ∈ V ∩ U, and so V[α] ⊆ U[α] = U[γ]
by Observation 2. Since p(V[α] ) = V, it follows that p−1 (V ) ∩ U[γ] = V[α] .

Since B is a basis for the topologies on X,

e it follows that p−1 (V ) is open
e for all V ∈ U . Since U is a basis for the topology on X, it follows that
in X,
p:X e → X is continuous. Given a point x ∈ X, choose an object U ∈ U with
x ∈ U. We claim that the collection

CU := {U[γ] : γ is a path in X with γ(0) = x0 and γ(1) ∈ U }

is a partition of p−1 (U ). Since p−1 (U ) =

S
U[γ] , it suffices to show that
U[γ] ∈CU
objects of the collection CU are either disjoint or identical. If [α] ∈ U[γ] ∩ U[δ] ,
then α is a path in X with α(0) = x0 and α(1) ∈ U, and hence by Observation
2 we have U[γ] = U[α] = U[δ] . Since the restriction of p on each of U[γ] is a
homeomorphism, p : X e → X is a covering map.

It remains to show that X e is simply connected. Given a point [γ] ∈ X

e and
t ∈ I, consider the map γt : I → X defined by
(
γ(s), if 0 ≤ s ≤ t, and
γt ( s ) : = (2.4.11)
γ(t), if t ≤ s ≤ 1.
82 Chapter 2. Homotopy Theory

Note that, each γt is a path in X starting at x0 , and hence its path-homotopy

e Then the map φ[γ] : I → X
class is an element of X. e defined by

φ[γ] (t) = [γt ], ∀ t ∈ I,

is a path (why it is continuous?) in X e starting at xe0 = [c x ] ∈ X e and ending

0

at [γ] ∈ X.e Therefore, X e is path-connected. Since p : ( X, e xe0 ) → ( X, x0 ) is a

covering map, the homomorphism p∗ : π1 ( X, e xe0 ) → π1 ( X, x0 ) induced by the
map p is injective by Corollary 2.3.20. Therefore, to show π1 ( X, e xe0 ) is trivial


it suffices to show that p∗ π1 ( X,
e xe0 ) is the trivial subgroup of π1 ( X, x0 ). By


Corollary 2.3.20 elements of p∗ π1 ( X, e xe0 ) ⊆ π1 ( X, x0 ) are given by loops γ in
X based at x0 whose lift to the cover p : X e → X starting at xe0 is a loop in X e
based at xe0 . Since φ[γ] is a path in Xe starting at xe0 and p ◦ φ[γ] = γ, we must
have [γ] = φ[γ] (1) = xe0 = [c x0 ]. In other words, γ is path-homotopic to the
constant loop c x0 in X. This completes the proof.

We now go towards establishing Galois correspondence for covering spaces.

Whenever we talk about simply connected covering space of X, we assume
that X is semi-locally simply connected in addition to be it path-connected and
locally path-connected.

2.4.3 Group action and covering map

Before proceeding further, let’s recall some standard terminologies related

to group action. Let G be a group, and let σ : G × X → X be a left G-action on
X. For notational simplicity, we denote by g · x the element σ( g, x ) ∈ X, for all
( g, x ) ∈ G × X. Given x ∈ X, the subset

StabG ( x ) := { g ∈ G : g · x = x } ⊆ G

is a subgroup of G, known as the stabilizer of x or the isotropy subgroup for x. The

G-action σ is said to be free if StabG ( x ) = {e}, for all x ∈ X. This means that,
for each x ∈ X, given g1 , g2 ∈ G, we have g1 · x = g2 · x if and only if g1 = g2 .
Note that the G-action σ on X defines an equivalence relation on X; for x ∈ X,
its equivalence class is the subset

OrbG ( x ) := { g · x : g ∈ G } ⊆ X,
2.4. Galois theory for covering spaces 83

called the G-orbit of x in X. The G-action σ is said to be transitive if there is

exactly one G-orbit in X. In other words, given any two points x1 , x2 ∈ X, there
exists g ∈ G such that x2 = g · x1 .

Definition 2.4.12. Let G be a group. A G-action σ : G × X → X on X is said to

be even (or, properly discontinuous according to old texts) if the G-action map σ
is continuous, and each point x0 ∈ X has an open neighbourhood V ⊆ X such
that ( g · V ) ∩ V = ∅, for all g 6= e in G, where g · V := { g · x : x ∈ V } ⊆ X.

Remark on old notation: Most of the old texts uses the term properly discontinuous
G-action to mean an even G-action. This terminology is awkward because the
G-action on X itself is a continuous map.

Proposition 2.4.13. If a group G is acting evenly on a path-connected and locally

path-connected topological space Y, then the associated quotient map q : Y → Y/G is
a covering map.

Proof. Clearly the quotient map q : Y → Y/G is continuous. Note that, for any
subset V ⊆ Y we have
q −1 q ( V ) =

[
g · V, (2.4.14)
g∈ G

where g · V = { g · v : v ∈ V } ⊆ X, for all g ∈ G. Since the left translation map

L g : Y → Y given by

L g (y) = g · y := σ( g, y), ∀ y ∈ Y

is a homeomorphism, V is open in Y if and only if g · V = L g (V ) is open in Y,

for all g ∈ G. Since q is a quotient map, it follows that q(V ) is open in Y/G if V
is open in Y. Therefore, q is an open map.

To see q : Y → Y/G is a covering map, let’s fix a point v ∈ Y/G, and a point
y ∈ q−1 (v). Since the G-action on Y is even, y has an open neighbourhood
Uy ⊆ Y such that ( g · Uy ) ∩ Uy = ∅, for all g 6= e in G. Take Vy := q(Uy ). Then
it follows that
q−1 (Vy ) =
G
g · Uy .
g∈ G

It remains to show that the restriction map


q g·U : g · Uy → Vy = q(Uy )
y
84 Chapter 2. Homotopy Theory

is a homeomorphism, for all g ∈ G. Since q is continuous and open, it suffices


to show that q g·U is bijective, for all g ∈ G.
y

If q g·U were not injective, then there exist y1 , y2 ∈ g · Uy with y1 6= y2
y
such that q(y1 ) = q(y2 ). Then there exists h ∈ G such that y2 = h · y1 . Then
y2 = h · y1 ∈ Uy ∩ (h · U1 ) implies h = e because the G-action on Y is even.

This contradicts our assumption that y1 6= y2 = h · y1 . Therefore, q g·U must be
 y
injective. To show q is surjective, note that a typical element of Vy = q(Uy )
g·Uy
is of the form q(y1 ), for some y1 ∈ Uy . Since q(y1 ) = OrbG (y1 ) = { a · y1 : a ∈
 
G }, we see that g · y1 ∈ g · Uy satisfies q g·U ( g · y1 ) = q(y1 ). Therefore, q g·U is
y y
surjective.

Proposition 2.4.13 allow us to construct a lot of examples of covering maps.

2.4.4 Group of Deck transformations

Let f : Y → X be a covering map. An automorphism of f : Y → X is a

homeomorphisms φ : Y → Y satisfying f ◦ φ = f . The set

Aut(Y/X ) := {φ : Y → Y | φ is a homeomorphism satisfying f ◦ φ = f }

of all automorphisms of f : Y → X forms a group with respect to the binary

operation on Aut(Y/X ) given by composition of homeomorphisms. The group
Aut(Y/X ) is also known as the group of Deck transformations or covering trans-
formations of f : Y → X. Note that, Aut(Y/X ) acts on Y from the left by
automorphisms:

a : Aut(Y/X ) × Y → Y, (φ, y) 7→ φ(y). (2.4.15)

We shall show in Proposition 2.4.19 that if we equip Aut(Y/X ) with discrete

topology, then the action map in (2.4.15) become continuous.

Proposition 2.4.16. Fix a point x0 ∈ X, and a path-connected covering space f : Y →

X of X. Then the natural Aut(Y/X )-action on Y restricts to give a free Aut(Y/X )-
action on the fiber f −1 ( x0 ). If Y is simply connected, then the Aut(Y/X )-action on
the fiber f −1 ( x0 ) is transitive.
2.4. Galois theory for covering spaces 85

Proof. Let y0 ∈ f −1 ( x0 ) be given. Since φ ∈ Aut(Y/X ) satisfies f ◦ φ = f , we

have f (φ(y0 )) = f (y0 ) = x0 , and hence φ(y0 ) ∈ f −1 ( x0 ). Therefore, the natural
Aut(Y/X )-action on Y restricts to an Aut(Y/X )-action on the fiber f −1 ( x0 ). If
φ(y0 ) = y0 , for some φ ∈ Aut(Y/X ), then by uniqueness of lifting of maps
(see Theorem 2.3.27 or Lemma 2.3.13) we must have φ = IdY . Therefore, the
Aut(Y/X )-action on the fiber f −1 ( x0 ) is free.

Now assume that Y is simply connected. To show that Aut(Y/X )-action

on the fiber f −1 ( x0 ) is transitive, choose two points y0 , y1 ∈ f −1 ( x0 ). Since
X is locally path-connected and f : Y → X is a covering map, Y is locally
path-connected. Since by assumption Y is path-connected and locally path-
connected (since X is so) with π1 (Y ) trivial, by general lifting criterion (Theo-
rem 2.3.27) there is a unique continuous map φ : (Y, y0 ) → (Y, y1 ) such that
f ◦ φ = f . Similarly, there is a unique continuous map ψ : (Y, y1 ) → (Y, y0 )
such that f ◦ ψ = f . Then by uniqueness of lifting (see Theorem 2.3.27), we
must have φ ◦ ψ = Id(Y, y1 ) and ψ ◦ φ = Id(Y, y0 ) . Therefore, both φ and ψ are
homeomorphisms, and that φ(y0 ) = y1 . Thus, the Aut(Y/X )-action on f −1 ( x0 )
is transitive.

Let f : Y → X be a covering map. Fix a point x0 ∈ X. Since X is locally

path-connected, there is a path-connected open neighbourhood U ⊂ X of x0
which is evenly covered by f . Then we can write

f −1 (U ) =
G
Vy , (2.4.17)
y ∈ f −1 ( x 0)

where Vy ⊂ Y is the path-connected open neighbourhood of y ∈ f −1 (y0 ) such

that f V : Vy → U is a homeomorphism. Note that, {Vy : y ∈ f −1 ( x0 )} is

y
precisely the set of all path-components of f −1 (U ).

Proposition 2.4.18. With the above notations, Aut(Y/X ) acts freely on the set of all
path-components {Vy : y ∈ f −1 ( x0 )} of f −1 (U ). Moreover, this action is transitive
when Y is simply connected.

Proof. Since f : Y → X is a covering map, the restricted map

f U := f  f −1 (U ) : f −1 (U ) −→ U

86 Chapter 2. Homotopy Theory

is a covering map. Since for any φ ∈ Aut(Y/X ) we have f ◦ φ = f , image

of the restriction map φ f −1 (U ) : f −1 (U ) → Y lands inside f −1 (U ), and hence


gives rise to an automorphism of the covering space f U : f −1 (U ) → U, i.e.,

φ f −1 (U ) ∈ Aut( f −1 (U )/U ). Clearly φ ∈ Aut(Y/X ) takes path-components of


f −1 (U ) to path-components of f −1 (U ). In particular, for each y ∈ f −1 ( x0 ), the

induced map
φ : Vy → Vφ(y)

is a homeomorphism. Since Aut(Y/X )-action on f −1 ( x0 ) is free by Proposition

2.4.16, if for some y ∈ f −1 ( x0 ), the automorphism φ ∈ Aut(Y/X ) takes Vy to
itself, then we must have Φ = IdY .

Now assume that Y is simply connected. Since the Aut(Y/X )-action on

f −1 ( x
0 ) is transitive by Proposition 2.4.16, and the path-components of f
−1 (U )

are uniquely determined by the conditions that Vy ∩ f −1 ( x0 ) = {y} and Vy1 ∩

Vy2 = ∅ for y1 6= y2 in f −1 ( x0 ), given any two path-components Vy1 , Vy2 of
f −1 (U ), there exists φ ∈ Aut(Y/X ) such that φ(y1 ) = y2 , and hence φ(Vy1 ) =
Vy2 . Thus, the Aut(Y/X )-action on the set of all path-components of f −1 (U ) is
transitive.

Proposition 2.4.19. Let f : Y → X be a path-connected covering space of X. Equip

Aut(Y/X ) with discrete topology. Then there is a continuous map (action map)

a : Aut(Y/X ) × Y → Y, (2.4.20)

such that the following diagram commutes:

a /Y
Aut(Y/X ) × Y
pr2 f (2.4.21)
 f 
Y / X,

where pr2 : Aut(Y/X ) × Y → Y is the projection map onto the second factor.

Proof. Clearly a : Aut(Y/X ) × Y → X is defined by

a(φ, y) = φ(y), ∀ (φ, y) ∈ Aut(Y/X ) × Y,

makes the above diagram commutative. We only need to show that the action
map a is continuous.
2.4. Galois theory for covering spaces 87

Let

B := {V ⊆ Y : V is path-connected, open and

f (V ) is evenly covered by f }.

Since Y is path-connected and locally path-connected covering space for X, it

is easy to check that B is a basis for the topology on Y. Therefore, to show
the action map a is continuous, it is enough to show that a−1 (V ) is open in
Aut(Y/X ) × Y, for all V ∈ B . Fix V ∈ B . Since V is path-connected and
f : Y → X is a covering map, U := f (V ) is path-connected and open in X. Fix
a point x0 ∈ U. Since U = f (V ) is evenly covered by f , we can write

f −1 (U ) =
G
Vy ,
y ∈ f −1 ( x 0 )

where Vy ⊆ Y is an open neighbourhood of y ∈ f −1 ( x0 ) such that f V : Vy →


y
U is a homeomorphism. Since V is path-connected and p(V ) = U, we have
V ⊆ Vy0 , for some y0 ∈ f −1 ( x0 ). Since f V : Vy0 → U is a homeomorphism,

y0

we must have V = Vy0 , for some y0 ∈ f −1 ( x

0 ). Therefore, it is enough to show
that a (Vy ) is open in Aut(Y/X ) × Y, for all y ∈ f −1 ( x0 ).
− 1

Let (φ, y) ∈ a−1 (Vy0 ) = {(ψ, y0 ) ∈ Aut(Y/X ) × Y : ψ(y0 ) ∈ Vy0 } be arbi-

trary. Then φ(y) ∈ Vy0 . Since φ is an automorphism of Y, there is a unique
y1 ∈ Y such that φ(y1 ) = y0 . Then φ : Vy1 → Vy0 is a homeomorphism. Since
φ(y) ∈ Vy0 , we must have y ∈ Vy1 . Then {φ} × Vy1 is an open neighbourhood
of (φ, y) in Aut(Y/X ) × Y such that a {φ} × Vy1 ⊆ Vy0 . Therefore, a−1 (Vy0 ) is

open in Aut(Y/X ) × Y. This completes the proof.

Corollary 2.4.22. If f : Y → X is a connected cover of X, the action of Aut(Y/X )

on Y is even (see Definition 2.4.12).

Proof. Follows from Proposition 2.4.19 and 2.4.18.

Proposition 2.4.23. If a group G acts evenly on a connected topological space Y, then

the automorphism group Aut(Y/X ) of the covering map q : Y → X := Y/G is
naturally isomorphic to G.
88 Chapter 2. Homotopy Theory

Proof. Let σ : G × Y → Y be the left G-action which is even. Since σ is continu-

ous, for each g ∈ G, the induced map

σg : Y → Y, y 7→ g · y := σ( g, y)

is a homeomorphism of Y onto itself. Since the quotient map q : Y → X := Y/G

sends a point y ∈ Y to its G-orbit OrbG (y) ∈ Y/G, it follows that q(σg (y)) =
q(y), for all g ∈ G. Therefore, σg ∈ Aut(Y/X ). Thus we have a natural map

Φ : G −→ Aut(Y/X ), g 7−→ σg . (2.4.24)

Note that, for any g, h ∈ G we have

σgh (y) = ( gh) · y = g · (h · y) = σg (σh (y)), ∀ y ∈ Y.

Therefore, Φ is a group homomorphism. Since the G-action on Y is even (see

Definition 2.4.12), it follows that Ker(Φ) is trivial, and hence Φ is injective. Let
ϕ ∈ Aut(Y/X ) be arbitrary. Fix a pint y ∈ Y, and let x := q(y) ∈ X. Since
ϕ(y) ∈ q−1 ( x ) = OrbG (y), we have ϕ(y) = g · y = σg (y), for some g ∈ G.
Since both ϕ, σg ∈ Aut(Y/X ) and they agree at a point of Y and Y is connected,
by uniqueness of lifting (see Lemma 2.3.13) we have ϕ = σg . Therefore, Φ is
surjective, and hence is an isomorphism.

2.4.5 Galois covers

Let f : Y → X be a path-connected covering space of X. Then the natural

left Aut(Y/X )-action on Y gives rise to an equivalence relation on Y, where
the equivalence classes are Aut(Y/X )-orbits of points of Y. Given y ∈ Y, its
Aut(Y/X )-orbit is the subset

OrbAut(Y/X ) (y) = {φ(y) : φ ∈ Aut(Y/X )} ⊆ Y.

Fix y0 ∈ Y, and let x0 = f (y0 ). Clearly, OrbAut(Y/X ) (y0 ) ⊆ f −1 ( x0 ), and equal-

ity holds if and only if the Aut(Y/X )-action on the fiber f −1 ( x0 ) is transitive.
By the universal property of quotient space, there is a unique continuous map

fe : Y/ Aut(Y/X ) → X (2.4.25)
2.4. Galois theory for covering spaces 89

such that the following diagram commutes:

f
Y /8 X
q
(2.4.26)
& ∃ ! fe
Y/ Aut(Y/X )

where q : Y → Y/ Aut(Y/X ) is the quotient map.

Definition 2.4.27 (Galois cover). A covering map f : Y → X is said to be a Galois

cover of X if Y is path-connected and the continuous map fe : Y/ Aut(Y/X ) →
X in (2.4.25), induced by f , is a homeomorphism (see the diagram (2.4.26)).

Proposition 2.4.28. A connected covering map p : Y → X is Galois if and only if

Aut(Y/X ) acts transitively on each fiber of the covering map p.

Proof. Suppose that p : Y → X is Galois cover. Consider the commutative

diagram.
p
Y 8/ X
q pe
&
Y/ Aut(Y/X )

Since the induced map pe : Y/ Aut(Y/X ) → X is a homeomorphism (by defini-

tion), for each x ∈ X, the fiber p−1 ( x ) coincides with the Aut(Y/X )-orbit of a
point of the fiber p−1 ( x ). In other words, the Aut(Y/X )-action on each of the
fibers of p is transitive.

Conversely, if the Aut(Y/X )-action on each of the fibers of p is transitive,

then the induced continuous map pe : Y/ Aut(Y/X ) → X is bijective. There-
fore, to show that p : Y → X a Galois cover, it suffices to show that pe is an
open map. Let U ⊆ Y/ Aut(Y/X ) be an open subset. Since the quotient map
q : Y → Y/ Aut(Y/X ) is continuous, q−1 (U ) is open in Y. Since the covering
map p : Y → X is an open map, p q−1 (U ) is open in X. Since q is surjective,

we have q(q−1 (U )) = U. Since p = pe ◦ q, we have

pe(U ) = pe q(q−1 (U )) = p(q−1 (U )).

Therefore, pe(U ) is open in X. This completes the proof.

90 Chapter 2. Homotopy Theory

If Y is simply connected, as remarked above, the Aut(Y/X )-orbit of y0 is

precisely the fiber f −1 ( x0 ), for all y0 ∈ f −1 ( x0 ). Therefore, in that case, the map
fe is bijective. This leads to the following.

e → X is Galois cover.
Corollary 2.4.29. A simply-connected covering map p : X

e being simply connected, Aut( X/X

Proof. X e ) acts transitively on each fiber of p
by Proposition 2.4.18. Therefore, the result follows from Proposition 2.4.28.

Remark 2.4.30. If p : Y → X is a covering map with Y connected, then to show

p : Y → X is a Galois cover it suffices to show that Aut(Y/X ) acts transitively
on one fibre. Indeed, since in this case Y/ Aut(Y/X ) is a connected cover of X
where one of the fibres is singleton, it follows that pe : Y/ Aut(Y/X ) → X is a
homeomorphism.

2.4.6 Galois correspondence for covering spaces

Theorem 2.4.31. Let p : Y → X be a Galois cover. For each subgroup H of the

Galois group G := Aut(Y/X ), the projection map p induces a natural continuous
map peH : Y/H → X which is a covering map. Conversely, if f : Z → X is a
connected cover of X fitting into a commutative diagram

φ
Y / Z
p   f
X

then φ : Y → Z is a Galois cover and Z is homeomorphic to Y/H. The maps H 7→

Y/H and Z 7→ Aut(Y/Z ) induces a natural one-to-one correspondence between the
collection of subgroups of G and the intermediate covers of p : Y → X as above.
Moreover, the cover f : Z := Y/H → X is Galois if and only if H is a normal
subgroup of G; and in this case we have Aut( Z/X ) ∼
= G/H.
 91

Chapter 3

Differential Topology

3.1 Review of Rn Calculus

Recall that the set Rn together with component-wise addition and scalar
multiplication

( x1 , . . . , x n ) + ( y1 , . . . , y n ) : = ( x1 + y1 , . . . , x n + y n )
λ · ( x1 , . . . , xn ) := (λx1 , . . . , λxn )

is a R-vector space of dimension n. The Euclidean norm of a vector a = ( a1 , . . . , an ) ∈

Rn is the non-negative real number || a|| defined by
q
|| a|| := a21 + · · · + a2n . (3.1.1)

The Euclidean norm induces a metric on Rn , known as the Euclidean metric,

which is defined as follow: given two points (vectors) x, y ∈ Rn , the Euclidean
distance between them is the non-negative real number

d( x, y) := || x − y||,

the Euclidean norm of the vector x − y ∈ Rn . More precisely, if x := ( x1 , . . . , xn )

and y := (y1 , . . . , yn ) in Rn , then d( x, y) is the non-negative real number
q
||( x1 , . . . , xn ) − (y1 , . . . , yn )|| = ( x1 − y1 )2 + · · · + ( x n − y n )2 . (3.1.2)

It is easy to check that the Euclidean distance between points, as defined above,
is a metric on Rn . Thus, Rn has a structure of a normed linear space and hence
92 Chapter 3. Differential Topology

of a metric space. Given a point a ∈ Rn and a real number r > 0, we define the
open ball in Rn with centre at a and radius r to be the subset

B( a, r ) := { x ∈ Rn : || x − a|| < r }.

The collection { B( a, r ) : a ∈ Rn , r > 0} forms a basis for a topology on Rn ,

known as the Euclidean topology on Rn . One can easily check that the Euclidean
topology on Rn coincides with the product topology on it induced from the
Euclidean topology on R.

Recall that the closure of a subset A in a topological space X is the smallest

closed subset A containing A; this can be constructed as the intersection of all
closed subsets of X containing the given subset A, i.e.,
\
A= Z,
Z ∈C A

where C A is the collection of all closed subsets of X that contains A. With this
definition, check that the closure of B( a, r ) in Rn is the subset

B( a, r ) = { x ∈ Rn : || x − a|| ≤ r },

known as the closed ball in Rn with centre at a and radius r. Let A be a non-empty
subset of Rn .
Definition 3.1.3. A function f : A → Rm is said to be continuous at a ∈ A if for
given a real number e > 0, there exists a real number δ > 0 (depending on both
a and e) such that for any x ∈ A with || x − a|| < δ, we have || f ( x ) − f ( a)|| < e.
f is said to be continuous on A if f is continuous at each point of A.

Let A be a non-empty subset of Rn , and let a ∈ A. Let f : A → Rm be a

map. A vector v ∈ Rm is said to be a limit of f at a, denoted by the symbol
lim f ( x ), if for each real number e > 0 there exists a real number δ > 0 such
x→a
that



f B( a, δ) \ { a} ∩ A ⊆ B(v, e).
Exercise 3.1.4. If a ∈ A, show that f is continuous at a if and only if lim f ( x ) =
x→a
f ( a ).
Proposition 3.1.5. A function f : A → Rm is continuous if and only if the inverse
image f −1 (U ) of any open subset U ⊆ Rn is open in A.
3.1. Review of Rn Calculus 93

Proof. Suppose that f is continuous on A. Let U ⊆ Rn be an open subset. To

show f −1 (U ) open in A, let a ∈ f −1 (U ) be arbitrary but fixed after choice. Since
U is open subset of Rm containing f ( a), there is a real number e > 0 such that
the open ball B( f ( a), e) ⊆ U. Since f is continuous at a, there is a real number
δ > 0 such that for any point x ∈ B( a, δ) ∩ A, we have f ( x ) ∈ B( f ( a), e). In
other words, f ( B( a, δ) ∩ A) ⊆ B( f ( a), e). Thus B( a, δ) ⊆ f −1 (U ) is an open
subset containing a. Since a ∈ f −1 (U ) is arbitrary, f −1 (U ) is open in A.

Conversely, suppose that f −1 (U ) is open in A for any open subset U ⊆ Rn .

Fix a ∈ A. Let e > 0 be given. Since U := B( f ( a), e) is open subset of Rm
containing f ( a), its inverse image f −1 (U ) is an open subset of A containing
a. Then there is a real number δ > 0 such that B( a, δ) ∩ A ⊆ f −1 (U ). Con-
sequently, f ( B( a, δ) ∩ A) ⊆ B( f ( a), e). Therefore, f a is continuous at a. Since
a ∈ A is arbitrary, f is continuous on A.

Fix two positive integers m, n. For each j ∈ {1, . . . , n}, the projection map
onto the j-th factor is the map p j : Rn → R defined by p j ( x1 , . . . , xn ) := x j , for all
( x1 , . . . , xn ) ∈ Rn . Let A ⊆ Rm . Given a map f : A → Rn , its j-th component
map is the composition f j := p j ◦ f ,

f pj
f j : A −→ Rn −→ R, ∀ j = 1, . . . , n.

Exercise 3.1.6. Show that f : A → Rm is continuous if and only if its j-th

component function f j := p j ◦ f : A → R is continuous, for all j = 1, . . . , m.

3.1.1 Differentiation

Let A ⊆ R, and a ∈ A be an interior point of A. A map f : A → R is said to

be differentiable at a ∈ A if there is a real number D f ( a) such that

f ( a + h) − f ( a)
D f ( a) = lim ; (3.1.7)
h →0 h

in this case D f ( a) is called the derivative of f at a. Note that, the above equation
is equivalent to the following one.

f ( a + h) − [ f ( a) + D f ( a)h]
lim = 0. (3.1.8)
h →0 h
94 Chapter 3. Differential Topology

Note that, h 7→ D f ( a)h is a R-linear map from R to R, whose matrix represen-

tation is the 1 × 1 matrix ( D f ( a)). Therefore, the above equation says that in
a small enough neighbourhood of a in A, the map f : A → R can be approxi-
mated by the unique affine transformation∗ defined by

h 7→ f ( a) + D f ( a)(h),

where D f ( a) : R → R is the R-linear map defined by

h 7→ D f ( a) · h.

This gives a way to extend the notion of differentiability of a function of several

variables. Here is a formal definition.

Definition 3.1.9. Let A ⊆ Rn and a ∈ A an interior point of A. A map f : A →

Rm is said to be differentiable at a if there is a R-linear map D f ( a) : Rn → Rm
such that
|| f ( a + h) − f ( a) − D f ( a)(h)||
lim = 0, (3.1.10)
h →0 ||h||
where the norms in the numerator and the denominator are the Euclidean
norms in Rm and Rn , respectively.

The next Proposition 3.1.12 says that the R-linear map D f ( a) : Rn → Rm in

Definition 3.1.9 is unique. We use the following.

Exercise 3.1.11. For any real number δ > 0, show that B(0, δ) ⊂ Rn contains a
basis for the R-vector space Rn .

Proposition 3.1.12. If f : Rn → Rm is differentiable at a ∈ Rn , then there is a

unique R-linear map D f ( a) : Rn → Rm such that

|| f ( a + h) − f ( a) − D f ( a)(h)||
lim = 0.
h →0 ||h||

Proof. Let T : Rn → Rm be any R-linear map such that

|| f ( a + h) − f ( a) − T (h)||
lim = 0.
h →0 ||h||
∗ Given two vector spaces V
and W, an affine transformation from V to W is a map of the form
w0 + T, where T : V → W is a linear map and w0 ∈ W.
3.1. Review of Rn Calculus 95

Since a ∈ A is an interior point, there is a real number δ > 0 such that B( a, δ) ⊆

A. Given a real number e > 0, ∃ positive real numbers δ1 , δ2 > 0 such that

|| f ( a + h) − f ( a) − D f ( a)(h)||
< e/2, (3.1.13)
||h||
|| f ( a + h) − f ( a) − T (h)||
and < e/2. (3.1.14)
||h||

Take δ0 := min{δ1 , δ2 , δ}. Then for any h ∈ B(0, δ0 ) \ {0} ⊂ Rn we have

|| T (h) − D f ( a)(h)||
0≤
||h||
|| [ f ( a + h) − f ( a) − D f ( a)(h)] − [ f ( a + h) − f ( a) − T (h)] ||
=
||h||
|| f ( a + h) − f ( a) − D f ( a)(h)|| || f ( a + h) − f ( a) − T (h)||
≤ +
||h|| ||h||
e e
< + = e.
2 2

Since e > 0 is arbitrary, we have

|| T (h) − D f ( a)(h)||
lim = 0.
h →0 ||h||

Fix a non-zero element h0 ∈ B(0, δ0 ) \ {0}. Then we have

|| T (th0 ) − D f ( a)(th0 )||

0 = lim
t →0 ||th0 ||
|t| · || T (h0 ) − D f ( a)(h0 )||
= lim
t →0 |t| · ||h0 ||
|| T (h0 ) − D f ( a)(h0 )||
= lim
t →0 ||h0 ||
|| T (h0 ) − D f ( a)(h0 )||
= .
||h0 ||

Since h0 6= 0, we conclude that T (h) = D f ( a)(h), ∀ h ∈ B(0, δ0 ) \ {0}. Then by

Exercise 3.1.11 we have

T (h) = D f ( a)(h), ∀ h ∈ Rn .

This completes the proof.

96 Chapter 3. Differential Topology

Note that, the unique R-linear map D f ( a) : Rn → Rm in Definition 3.1.9 is

the best R-linear approximation of f at a ∈ A, and is called the derivative of f at a.
The matrix representation of D f ( a) with respect to the standard ordered bases
for Rn and Rm is a m × n matrix with coefficients from R, called the Jacobian
of f at a. If A ⊆ Rn is an open subset of Rn , a map f : A → Rm is said to
be differentiable on A if it is differentiable at each point of A. When A ⊆ Rn is
not necessarily open, f : A → Rm is said to be differentiable on A if it can be
extended to a differentiable function on an open neighbourhood of A in Rn .

Proposition 3.1.15. Let T : Rn → Rm be a R-linear map. Then there exists a positive

real number M ∈ R such that || T ( x )|| ≤ M · || x ||, ∀ x ∈ Rn .

Proof. Let {e1 , . . . , en } be the standard ordered basis for Rn , and for each i ∈
{1, . . . , n}, let wi := T (ei ) ∈ Rm . Let x = ( x1 , . . . , xn ) ∈ Rn with || x || = 1. Since
T ( x ) = x1 w1 + · · · + xn wn , by Cauchy-Schwarz inequality we have
!1/2 !1/2
n 1
n n
|| T ( x )|| ≤ ∑ |x j | · ||w j || ≤ ||x|| 2 · ∑ ||w j || = ∑ ||w j || .
j =1 j =1 j =1

Therefore, the subset S := {|| T ( x )|| : x ∈ Rn with || x || = 1} ⊆ R is bounded

above, and hence || T || := sup(S) exists in R. Since for any non-zero x ∈ Rn we
have ||( x/|| x ||)|| = 1, it follows that || T ( x )|| ≤ || T || · || x ||, for all x ∈ Rn .

Exercise 3.1.16. Let A ⊆ Rn , and let f : A → Rm be differentiable at an interior

point a ∈ A. Show that f is continuous at a.

Theorem 3.1.17 (Chain rule). Let A ⊆ Rn and B ⊆ Rm . If f : A → Rn is

differentiable at a ∈ A, with f ( A) ⊆ B, and if g : B → R p is differentiable at f ( a),
then g ◦ f : A → R p is differentiable at a, and D ( g ◦ f )( a) = Dg( f ( a)) ◦ D f ( a).

Proof. For notational simplicity, let’s introduce the following.

ϕ( x ) = f ( x ) − f ( a) − D f ( a)( x − a), (3.1.18)

ψ(y) = g(y) − g(b) − Dg(b)(y − b), (3.1.19)


and ρ( x ) = ( g ◦ f )( x ) − ( g ◦ f )( a) − Dg(b) ◦ D f ( a) ( x − a). (3.1.20)
3.1. Review of Rn Calculus 97

Since f and g are differentiable at a and b = f ( a), respectively, we have

|| ϕ( x )||
lim = 0, (3.1.21)
x → a || x − a ||
||ψ(y)||
and lim = 0. (3.1.22)
y→b || y − b ||

Since


ρ( x ) =( g ◦ f )( x ) − ( g ◦ f )( a) − Dg(b) ◦ D f ( a) ( x − a)


= [ g(y) − g(b) − Dg(b)(y − b)] + Dg(b) y − b − D f ( a)( x − a)
=ψ(y) + Dg(b)( f ( x ) − f ( a) − D f ( a)( x − a))
=ψ( f ( x )) + Dg(b)( ϕ( x )),

it suffices to show that

||ψ( f ( x ))||
lim = 0, (3.1.23)
x → a || x − a |
|| Dg(b)( ϕ( x ))||
and lim = 0. (3.1.24)
x→a || x − a|

Now it follows from (3.1.22) that given a real number e > 0, there exists a real
number δ > 0 such that if || f ( x ) − b|| < δ, we have

||ψ( f ( x ))||e ≤ || f ( x ) − f ( a)||

= e|| ϕ( x ) + D f ( a)( x − a)||, by equation 3.1.21.
≤ e|| ϕ( x )|| + e|| D f ( a)|| · || x − a||, by Proposition 3.1.15.
≤ e2 || x − a|| + e|| D f ( a)|| · || x − a||


= e e + || D f ( a)|| || x − a||.

From this, the equation (3.1.23) follows. And the equation (3.1.24) follows from
the inequality
|| Dg(b)( ϕ( x ))|| ≤ || Dg(b)|| · | ϕ( x )||, ∀ x,

and the equation (3.1.21).

Theorem 3.1.25. Let f : Rn → Rm .

(i) If f is a constant map, then it is differentiable on Rn with D f ( a) = 0, for all

a ∈ Rn .
98 Chapter 3. Differential Topology

(ii) If f is a R-linear map, then f is differentiable on Rn with D f ( a) = f , for all

a ∈ Rn .

(iii) For each j ∈ {1, . . . , m}, let f j := p j ◦ f : Rn → R be the j-th component map
of f . Then f is differentiable at a ∈ Rn if and only if f j is differentiable at a, for
all j ∈ {1, . . . , m}. And in this case, we have D f ( a) = ( D f 1 ( a), . . . , D f m ( a)),
for all a ∈ Rn .
n
(iv) If f : Rn → R is defined by f ( x1 , . . . , xn ) = ∑ x j , for all ( x1 , . . . , xn ) ∈ Rn ,
j =1
then f is differentiable on Rn with D f ( a) = f , for all a ∈ Rn .

(v) If f : R2 → R is defined by f ( x, y) = xy, for all ( x, y) ∈ R2 , then f is

differentiable on R2 , and its derivative D f ( a, b) : R2 → R at ( a, b) ∈ R2 is
given by
D f ( a, b)( x, y) = bx + ay, ∀ ( x, y) ∈ R2 .

Proof. (i) Note that, if f is a constant function, then

|| f ( a + h) − f ( a) − 0||
lim = 0.
h →0 ||h||

(ii) If f : Rn → Rm is a R-linear map, then

|| f ( a + h) − f ( a) − f (h)|| || f ( a) + f (h) − f ( a) − f (h)||

lim = lim = 0.
h →0 ||h|| h →0 ||h||

(iii) Since the projection maps pi : Rm → R are R-linear, they are differentiable
with derivative Dpi (b) = pi , for all b ∈ Rm . If f is differentiable at a ∈ Rn ,
by chain rule (Theorem 3.1.17) the component functions f i := pi ◦ f are
differentiable at a with

D f i ( a) = Dpi ( f ( a)) ◦ D f ( a) = pi ◦ D f ( a), ∀ i ∈ {1, . . . , m}.

Conversely suppose that the component maps f i : Rn → R are differen-

tiable at a ∈ Rn . Let Ta : Rn → Rm be the R-linear map defined by

Ta ( x ) = ( D f 1 ( a)( x ), . . . , D f m ( a)( x )), ∀ x ∈ Rn .

3.1. Review of Rn Calculus 99

Since f ( x ) = ( f 1 ( x ), . . . , f m ( x )), ∀ x ∈ Rn , and

f ( a + h) − f ( a) − Ta (h)
= ( f 1 ( a + h) − f 1 (h) − Ta (h), . . . , f m ( a + h) − f m ( a) − Ta (h)), (3.1.26)

we have

|| f ( a + h) − f ( a) − Ta (h)||
lim
h →0 ||h||
m
| f ( a + h) − f i ( a) − D f i ( a)(h)|
≤ lim ∑ i = 0.
h →0 i =1 ||h||

Therefore, f is differentiable at a with D f ( a) = ( D f 1 ( a), . . . , D f m ( a)).

n
(iv) Since f ( x1 , . . . , xn ) := ∑ xi is a R-linear map (verify!), it follows that f is
i =1
differentiable on Rn with D f ( a) = f , for all a ∈ Rn .

(v) Let f : R2 → R be defined by f ( x, y) = xy, for all ( x, y) ∈ R2 . Fix

( a, b) ∈ R2 , and consider the map Ta,b : R2 → R defined by

Ta,b ( x, y) = bx + ay, ∀ ( x, y) ∈ R2 .

Then we have,

|| f ( a + h, b + k) − f ( a, b) − Ta,b (h, k)||

lim
(h, k)→(0, 0) ||(h, k)||
||( a + h)(b + k) − ab − bh − ak||
= lim √
(h, k)→(0, 0) h2 + k 2
|hk|
= lim √ .
(h, k)→(0, 0) h2 + k 2

Since (
|h2 |, if |k| ≤ |h|,
|hk| ≤
|k2 |, if |h| ≤ |k|,
|hk| h2 + k 2
we have lim √ ≤ lim √ = 0. Hence the result
(h, k)→(0, 0) h2 + k 2 (h, k)→(0, 0) h2 + k 2
follows.
100 Chapter 3. Differential Topology

Given f , g : Rn → R, we define f + g, f · g : Rn → R by

( f + g)( x ) := f ( x ) + g( x ), ∀ x ∈ Rn ,
( f · g)( x ) := f ( x ) · g( x ), ∀ x ∈ Rn .

If g is continuous at a ∈ Rn with g( a) 6= 0, then there is an open neighbourhood

Va ⊆ Rn of a with g( x ) 6= 0, for all x ∈ Va (verify!). Therefore, we can define a
f (x)
function f /g : Va → R by ( f /g)( x ) = g( x) , for all x ∈ Va . Then we have the
following.

Corollary 3.1.27. If f , g : Rn → R are differentiable at a ∈ Rn , so are f + g and

f · g; and in this case

D ( f + g)( a) = D f ( a) + Dg( a),

D ( f · g)( a) = g( a) D f ( a) + f ( a) Dg( a).

Moreover, if g( a) 6= 0, then f /g is differentiable at a with

g( a) D f ( a) − f ( a) Dg( a)
D ( f /g)( a) = .
[ g( a)]2

3.2 Differentiable Manifolds

A topological manifold is a Hausdorff second countable topological space X

such that each point x ∈ X has an open neighbourhood Ux ⊆ X which admits
a homeomorphism ϕUx : Ux → Rn .

Let X be a topological manifold. A C ∞ -atlas on X is an indexed family of

pairs C = {(Uα , ϕα )}α∈Λ , where Uα is an open subset of X such that
S
Uα =
α∈Λ
X and ϕα : Uα → Vα ⊆ Rn is a homeomorphism of Uα onto an open subset Vα
of Rn such that given any α, β ∈ Λ, the induced map

ϕαβ := ϕ β ◦ ϕ− 1
α : ϕα (Uα ∩ Uβ ) → ϕ β (Uα ∩ Uβ ) (3.2.1)

is a C ∞ map. Two C ∞ atlas C = {(Uα , ϕα )}α∈Λ and C 0 = {(Wβ , ψβ )} β∈Λ0 are

said to be equivalent if for any α ∈ Λ and β0 ∈ Λ0 the induced map

ψβ ◦ ϕ − 1
α : ϕα (Uα ∩ Wβ ) → ψβ (Uα ∩ Uβ ) (3.2.2)
3.2. Differentiable Manifolds 101

is a C ∞ map. A C ∞ manifold is a topological manifold X together with an equiv-

alence class of a C ∞ atlas on X.
 103

Chapter 4

Appendix

4.1 Category and Functor

Definition 4.1.1. ∗ A category C consists of the following data:

(i) a class of objects, denoted ob(C ),

(ii) for X, Y ∈ ob(C ), a class of morphisms from X into Y, denoted MorC ( X, Y ),

(iii) for each X, Y, Z ∈ ob(C ), a composition map

MorC ( X, Y ) × MorC (Y, Z ) → MorC ( X, Z ), ( f , g) 7→ g ◦ f ,

which satisfies associative property: h ◦ ( g ◦ f ) = (h ◦ g) ◦ f , for all f ∈

MorC ( X, Y ), g ∈ MorC (Y, Z ) and h ∈ MorC ( Z, W ), for all X, Y, Z, W ∈
ob(C ).

A category C is said to be locally small if MorC ( X, Y ) is a set, for all X, Y ∈

ob(C ). A category C is said to be small if it is locally small and the class of
objects ob(C ) is a set.

Example 4.1.2. The category (Set), whose objects are sets and morphisms are
given by set maps, is a locally small, but not small. However, the category
(FinSet), whose objects are finite sets and morphisms are given by set maps, is
a small category.
∗ Joke: Category theory is like Ramayana & Mahabharata — there are lots of arrows!
104 Chapter 4. Appendix

Two objects A1 , A2 ∈ C are said to be isomorphic if there are morphisms

(arrows) f : A1 → A2 and g : A2 → A1 in C such that g ◦ f = Id A1 and
f ◦ g = Id A2 .

Let A and B be two categories. A functor F : A → B is given by the

following data:

(i) for each X ∈ A there is an object F ( X ) ∈ B,

(ii) for X, Y ∈ A and f ∈ HomA ( X, Y ), there is F ( f ) ∈ MorB (F ( X ), F (Y )),

which are compatible with the composition maps.

A functor F : A → B is said to be faithful (resp., full) if for any two objects

A1 , A2 ∈ A , the induced map

F : MorA ( A1 , A2 ) −→ MorB ( F ( A1 ), F ( A2 ))

is injective (resp., surjective). We say that F is fully faithful if it is both full and
faithful.

Let F , G : A → B be two functors. A morphism of functors ϕ : F → G is

given by the following data: for each object A ∈ A , a map ϕ A : F ( A) → G( A)
which is functorial; that means, for any arrow f : A → A0 in A , the following
diagram commutes.
F(f)
F ( A) / F ( A0 )
ϕA ϕ A0 (4.1.3)
 G( f ) 
G( A) / G( A0 )
Definition 4.1.4. A morphism f ∈ MorA ( A, B) is said to be a monomorphism
if for any object T ∈ A and two morphisms g, h ∈ HomA ( T, A) with f ◦ g =
f ◦ h, we have g = h.

A morphism f ∈ MorA ( A, B) is said to be a epimorphism if for any object

T ∈ A and two morphisms g, h ∈ MorA ( B, T ) with g ◦ f = h ◦ f , we have
g = h.

Given any two categories A and B, we can define a category Func(A , B ),

whose objects are functors F : A → B, and for any two such objects F , G ∈
Func(A , B ), there is a morphism set Mor(F , G) consisting of all morphisms of
functors ϕ A : F → G , as defined above.
4.1. Category and Functor 105

Proposition 4.1.5. Let A and B be two small categories. Then F , G ∈ Func(A , B )

are isomorphic if there exists a morphism of functors ϕ : F → G such that for any
object A ∈ A , the induced morphism ϕ A : F ( A) → G( A) is an isomorphism in B.

Definition 4.1.6. A category A is said to be pre-additive if for any two objects

X, Y ∈ A , the set MorA ( X, Y ) has a structure of an abelian group such that the
composition map

MorA ( X, Y ) × MorA (Y, Z ) −→ MorA ( X, Z ),

written as ( f , g) 7→ g ◦ f , is Z-bilinear, for all X, Y, Z ∈ A .

Notation. For any pre-additive category A , we denote by HomA ( X, Y ) the

abelian group MorA ( X, Y ), for all X, Y ∈ ob(A ).

Let A and B be pre-additive categories. A functor F : A −→ B is said to

be additive if for all objects X, Y ∈ A , the induced map

F X,Y : HomA ( X, Y ) −→ HomB (F ( X ), F (Y ))

is a group homomorphism.

Definition 4.1.7 (Additive category). A category A is said to be additive if for

any two objects A, B ∈ A , the set HomA ( A, B) has a structure of an abelian
group such that the following conditions holds.

(i) The composition map HomA ( A, B) × HomA ( B, C ) −→ HomA ( A, C ),

written as ( f , g) 7→ g ◦ f , is Z-bilinear, for all A, B, C ∈ A .

(ii) There is a zero object 0 in A , i.e., HomA (0, 0) is the trivial group with one
element.

(iii) For any two objects A1 , A2 ∈ A , there is an object B ∈ A together with

morphisms ji : Ai → B and pi : B → Ai , for i = 1, 2, which makes B the
direct sum and the direct product of A1 and A2 in A .

Definition 4.1.8. Let k be a field. A k-linear category is an additive category A

such that for any A, B ∈ A , the abelian groups HomA ( A, B) are k-vector spaces
such that the composition morphisms

HomA ( A, B) × HomA ( B, C ) −→ HomA ( A, C ) , ( f , g) 7→ g ◦ f

106 Chapter 4. Appendix

are k-bilinear, for all A, B, C ∈ A .

Remark 4.1.9. Additive functors F : A −→ B between two k-linear addi-

tive categories A and B over the same base field k are assumed to be k-linear,
i.e., for any two objects A1 , A2 ∈ A , the map F A1 ,A2 : HomA ( A1 , A2 ) −→
HomB (F ( A1 ), F ( A2 )) is k-linear.

Let A be an additive category. Then there is a unique object 0 ∈ A , called

the zero object such that for any object A ∈ A , there are unique morphisms
0 → A and A → 0 in A . For any two objects A, B ∈ A , the zero morphism
0 ∈ HomA ( A, B) is defined to be the composite morphism

A −→ 0 −→ B .

In particular, taking A = 0, we see that, the set HomA (0, B) is the trivial group
consisting of one element, which is, in fact, the zero morphism of 0 into B in A .

Definition 4.1.10. Let f : A → B be a morphism in A . Then kernel of f is a pair

(ι, Ker( f )), where Ker( f ) ∈ A and ι ∈ HomA (Ker( f ), A) such that

(i) f ◦ ι = 0 in HomA (Ker( f ), B), and

(ii) given any object C ∈ A and a morphism g : C → A with f ◦ g = 0, there

is a unique morphism ge : C → Ker( f ) such that ι ◦ ge = g.

C
∃ ! ge 0
g
w 
ι /
f &/
Ker( f ) A B

The cokernel of f ∈ HomA ( A, B) is defined by reversing the arrows of the

above diagram.

Definition 4.1.11. The cokernel of f : A → B is a pair (π, Coker( f )), where

Coker( f ) is an object of A together with a morphism π : B → Coker( f ) in A
such that

(i) π ◦ f = 0 in HomA ( A, Coker( f )), and

4.1. Category and Functor 107

(ii) given any object C ∈ A and a morphism g : B → C with g ◦ f = 0 in

HomA ( A, C ), there is a unique morphism ge : Coker( f ) → C such that
ge ◦ π = g.
f
A / B
π / Coker( f )

g
&  v
0 ∃ ! ge
C
Definition 4.1.12. The coimage of f ∈ HomA ( A, B), denoted by Coim( f ), is the
cokernel of ι : Ker( f ) −→ A of f , and the image of f , denoted Im( f ), is the
kernel of the cokernel π : B −→ Coker( f ) of f .

Lemma 4.1.13. Let C be a preadditive category, and f : X → Y a morphism in C .

(i) If a kernel of f exists, then it is a monomorphism.

(ii) If a cokernel of f exists, then it is an epimorphism.

(iii) If a kernel and coimage of f exist, then the coimage is an epimorphism.

(iv) If a cokernel and image of f exist, then the image is a monomorphism.

Proof. Assume that a kernel ι : Ker( f ) → X of f exists. Let α, β ∈ HomC ( Z, Ker( f ))

be such that ι ◦ α = ι ◦ β. Since f ◦ (ι ◦ α) = f ◦ (ι ◦ β) = 0, by definition
ι
of Ker( f ) −→ X there is a unique morphism g ∈ Hom( Z, Ker( f )) such that
ι ◦ α = ι ◦ g = ι ◦ β. Therefore, α = g = β.

The proof of (ii ) is dual.

(iii ) follows from (ii ), since the coimage is a cokernel. Similarly, (iv) follows
from (i ).

Exercise 4.1.14. Let A be an additive category. Let f ∈ HomA ( X, Y ) be such

that Ker( f ) → X exists in A . Then the kernel of ι : Ker( f ) → X is the unique
ι

morphism 0 → Ker( f ) in A .

Lemma 4.1.15. Let f : X → Y be a morphism in a preadditive category C such that

the kernel, cokernel, image and coimage all exist in C . Then f uniquely factors as
X → Coim( f ) → Im( f ) → Y in C .

Proof. Since Ker( f ) → X → Y is zero, there is a canonical morphism Coim( f ) →

Y such that the composite morphism X → Coim( f ) → Y is f . The composi-
tion Coim( f ) → Y → Coker( f ) is zero, because it is the unique morphism
108 Chapter 4. Appendix

which gives rise to the morphism X → Y → Coker( f ), which is zero. Hence by

Lemma 4.1.13 (iii ), Coim( f ) → Y factors uniquely through Im( f ) = Ker(π f ).
This completes the proof.

ι f πf
Ker( f ) X Y Coker( f )

πι j
Coim( f ) Im( f )
(4.1.16)

Definition 4.1.17. An abelian category A is an additive category such that for

any morphism f : A → B in A , its kernel ι : Ker( f ) → A and cokernel p :
B → Coker( f ) exists in A , and the natural morphism Coim( f ) → Im( f ) is an
isomorphism in A (c.f. Definition 4.1.12).

Example 4.1.18. For any commutative ring A with identity, the category Mod A
of A-modules is an abelian category.