You are on page 1of 40

ALGEBRAIC TOPOLOGY I

Dr. A. Deb Ray


Department of Pure Mathematics
University of Calcutta
e-mail : debrayatasi@gmail.com
Contents

1 Introduction to Algebraic Topology 2

2 Homotopy 7
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2 Homotopy and relative homotopy - some examples and basic properties . . . 9
2.3 Contractible spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.4 Retracts and Deformation retracts . . . . . . . . . . . . . . . . . . . . . . . . 17
2.5 Extension of a continuous function . . . . . . . . . . . . . . . . . . . . . . . 19

3 Fundamental Group and its basic properties 21


3.1 Path Homotopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2 Construction of Fundamental group and Induced homomorphism . . . . . . . 25
3.3 Fundamental groups - Homeomorphic and Contractible spaces . . . . . . . . 28

4 Covering spaces and Covering maps 32


4.1 Covering spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2 Properties of Covering maps . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

1
Chapter 1

Introduction to Algebraic Topology

Sometimes it is difficult to answer several natural questions regarding properties of topologi-


cal spaces by using the techniques of Point-set topology. For example, our intuition suggests
that the Euclidean m-space is topologically different from Euclidean n-space, when m ̸= n.
Unfortunately, for large values of m and n there is no topological method by which this result
can be established in general. (We shall see shortly that some special cases of the problems
of this type can be solved by means of point-set topology.)
We begin our discussion by picking up just one important situation, known as ‘Homeo-
morphism Problem’, from a wide range of problems where sole topological methods are
either inadequate or it is difficult to use such methods for solving them.
Let X and Y be two topological spaces. The problem, ‘Is X homeomorphic to Y ’ is in
general referred to as Homeomorphism problem. This is a very important and a fundamental
question in general topology. Intuitively, two topological spaces are considered same, if they
are homeomorphic.
We now observe a few situations where topological techniques are useful for determining the
sameness of two topological spaces.

Example 1.0.1. An open interval (i.e., a line segment without its end points) and a semi-
circular arc (without its end points) are homeomorphic.
An open interval is nothing but a line segment without its end points, whereas a semi-circular
arc (without end-points) is a bended line segment. So the obvious way of getting the home-
omorphism is just flattening the circular arc into a straight line. The actual correspondence
is obtained trivially as follows :

ˆ Draw a line through the centre of the semi-circular arc that intersect both the objects.

ˆ This line cuts the semicircle at x and the open interval at y.


2
A. Deb Ray, C.U. 3

ˆ The correspondence is x ↔ y.

Another well-known technique in point set topology, called puncturing objects, is also used
to establish that two spaces are not topologically same (i.e., not homeomorphic). In the next
example, we explain this technique by showing that a circle is not homeomorphic to an open
interval, though ‘compactness’ is enough to show this.

Example 1.0.2. A circle is not homeomorphic to an open interval.


Removing a point from the circle (this removal is called puncturing) transforms it to a single
line segment. On the other hand, removing a point from the open interval breaks the open
interval into two pieces of line segments. A single line segment (i.e., a space with only one
component) can’t be homeomorphic to two pieces of lines (i.e., a space with two components),
and consequently, the original figures are not homeomorphic. (Here, the argument depends
on the fact that connectedness is a topological property and so, a connected space can not
be homeomorphic to a disconnected space.)

Example 1.0.3. The real line R is not homeomorphic to R2 .


Removal of one point from R breaks the real line into two components (hence, punctured R is
disconnected), but R2 is still connected when a single point is removed from it. A connected
space can not be homeomorphic to a disconnected space (as connectedness is a topological
property) and so, the original spaces are not homeomorphic to each other.

In fact, in the last example, there is nothing special about R2 and so, the same arguments
establish that Rn is not homeomorphic to R, for any n > 1. But, it is not possible to argue
in the same line to obtain in general that Rm is not homeomorphic to Rn , for any m ̸= n.

Example 1.0.4. In R3 , a sphere (S 2 ) is not homoemorphic to a torus (S 1 × S 1 ).


If we remove a circle from the torus T (as indicated in the following figure), we get a pipe-
like shape. This new shape remains connected. But removing a circle from a sphere S 2 (as
shown in the next diagram) breaks the sphere into two parts (i.e., the result is a disconnected
space). Hence, a torus is not homeomorphic to a sphere in Euclidean 3-space.
A. Deb Ray, C.U. 4

Unlike the case of a 2-sphere S2 and a torus (i.e., S 1 × S 1 ), it is highly difficult even to prove
that S 3 is not homeomorphic to S 1 × S 1 × S 1 .
Algebraic topology partially solves ‘Homeomorphism problems’. The general scheme is as
follows :

ˆ To each topological space X, associate an algebraic object H(X)

ˆ If there is a continuous function X → Y then it induces a homomorphism H(X) →


H(Y )

ˆ If the induced homomorphism is not an isomorphism then the given topological spaces
are not homeomorphic.

The algebraic object H(X) as mentioned above, may be a group, module, a sequence of
modules, etc. There are several methods of construction of such algebraic objects. In this
two semester course on Algebraic Topology, we explore two constructions of H(X):

1. Fundamental group / Homotopy group

2. Homology group

One should note carefully that by using the techniques of algebraic topology, ‘homeomor-
phism problems’ can be solved partially - not completely. In fact, we prove that
homeo iso
X −→ Y =⇒ H(X) −→ H(Y )

But the converse of this result is not in general true. So, this result helps to distinguish
topological spaces than establishing homeomorphisms.

However, there are situations where this implication is both-sided.


In a nutshell, Algebraic topology aims at creating and defining various algebraic objects
A. Deb Ray, C.U. 5

associated to topological spaces and homomorphisms induced by continuous maps. We shall


learn in the subsequent chapters how algebraic objects (i.e., fundamental groups and homol-
ogy groups) are constructed from topological spaces. We observe that the algebraic objects
so associated depend on the topology of X in such a manner that if f : X → Y is a con-
tinuous map from X to Y . then it naturally induces a homomorphism H(f ) between the
groups H(X) and H(Y ). Furthermore, the induced homomorphisms have the following two
properties:

1. If f : X → Y , g : Y → Z are continuous maps, then the induced homomorphisms H(f ),


H(g) and H(g ◦ f ) are related by H(g ◦ f ) = H(g) ◦ H(f ) (or, H(g ◦ f ) = H(f ) ◦ H(g)).

2. If IX : X → X is the identity map of the space X, then H(IX ) = IH(X) is the identity
map of the group H(X)

It is evident that these properties establish that H is a functor (either covariant or con-
travariant) from the Category of topological spaces to the category of groups. These two
properties are thus referred to as functorial properties.

Syllabus
Algebraic Topology I
PM3/11, Group B (Full marks 25)

ˆ Homotopy between continuous maps, homotopy relative to a subset, homotopy class,


null homotopy, contractibility of spaces, homotopy equivalent spaces, homotopy prop-
erties.

ˆ Deformability, deformation retracts, strong deformation retracts, homotopy between


paths, product of paths, fundamental group π(X, x) of a space X based at the point
x ∈ X, induced homomorphism and related properties. simply connected space, special
Van Kampan theorem and fundamental group of S n (n ≥ 2).

ˆ Fundamental Group of S 1 , fundamental group of the product and of torus, R2 and Rn


(n > 2) are not homeomorphic.

ˆ Fundamental theorem of algebra and Brouwer fixed point theorem. Covering projec-
tion, covering spaces, lifting of paths and homotopies, the fundamental group of a cov-
ering space. The Monodromy theorem, The Borsuk-Ulam theorem and ham-sandwich
theorem.

References

1. A. Dold , Lectures on Algebraic Topology, Springer-Verlag (1972).

2. W. Fulton , Algebraic Topology : A First Course, Springer-Verlag (1995).

3. M. Greenberg , Lectures on Algebraic Topology, W.D.Benjamin, N.Y. (1967).

4. Allen Hatcher , Algebraic Topology, Cambridge Univ. Press (2002).


A. Deb Ray, C.U. 6

5. C. Kosniowski, A First Course in Algebraic Topology, Cambridge University Press


(1980).

6. W. S. Massey, Algebraic Topology : An Introduction, Springer-Verlag, N.Y. (1990).

7. James R. Munkres : Topology (2nd Edit.), Pearson Education Inc. (2004).

8. E. H. Spanier, Algebraic Topology, McGraw Hill Book Co. N.Y. (1966).

9. C.T.C.Wall, A Geometric Introduction to Topology, Addison-Wesley Publ. Co. Inc(1972).


Chapter 2

Homotopy

Throughout this course, whenever we write ‘a function’ or ‘a map’ we always mean a con-
tinuous function, unless stated otherwise.

2.1 Introduction
In this chapter we learn a concept called ‘equivalence’ of two spaces, which is much broader
than homeomorphism. We introduce here, mainly four interrelated terms - Homotopy (and
homotopy relative to a subspace), deformation retract, homotopy equivalence and con-
tractible spaces. Let us first try to understand the concepts intuitively. Suppose, we start
with an annulus and draw a middle circle, as shown in the following figure:

The annulus is X and the middle circle is A. Certainly, A is a subspace of X. If we contin-


uously shrink the annulus X from both sides towards the middle circle A and assume that
this shrinking process is taking place during a time interval 0 ≤ t ≤ 1, then it defines a
family of functions ϕt : X → A parametrized by t ∈ [0, 1], where ϕt (x) is the point to which
a given point x ∈ X has moved in time t. This ϕt (x) must depend continuously on both of t
7
A. Deb Ray, C.U. 8

and x. We further put a restriction on this process by saying that the points on the middle
circle A remains stationary throughout 0 ≤ t ≤ 1.

This is an example of what we call a ‘deformation retraction’ of a space X onto a sub-


space A of X. The family of functions ϕt : X → A are called deformation retraction and
the subspace A is called a deformation retract of X. If we drop the restriction of A to
remain stationary during this shrinkingg process then we may call it a weak deformation
retraction. Some authors prefer to call deformation retraction as ‘strong’ deformation re-
traction and therefore, our weak deformation retraction becomes deformation retraction in
their terminology. We now observe that (weak) deformation retraction is a special case of a
more general notion called ‘homotopy’ . Suppose f : X → Y and g : X → Y are continuous
functions from a topological space X to a topological space Y . If a family of continuous
maps {ϕt : X → Y : t ∈ [0, 1]} exists, such that

ˆ ϕ0 = f

ˆ ϕ1 = g

ˆ ϕt changes continuously as t varies from 0 to 1.

then {ϕt : X → Y : t ∈ [0, 1]} is called a homotopy between f and g. Further, to formalize
‘ϕt changes continuously with t’, we make use of the product space X ×I, where I stands
for the closed interval [0, 1] and get the following:

Definition 2.1.1. Let f, g : X → Y be continuous functions. f is said to be homotopic to


F
g, denoted by f ∼= g (or, by f ∼
= g), if there exists a continuous function F : X × I → Y such
that F (x, 0) = f (x) and F (x, 1) = g(x), for all x ∈ X. The map F is called a homotopy.
If f is homotopic to a constant map then f is called null homotopic, denoted by f ∼ = 0.
Further, if F : X × I → Y is a continuous function such that F (x, 0) = f (x), F (x, 1) = g(x),
for all x ∈ X and F (a, t) = f (a) = g(a) for all a ∈ A, where A is some subspace of X,
then F is called a relative homotopy and f is said to be homotopic to g relative to A,
denoted by ∼A .

If in particular, one chooses A = ∅, then the definition of relative homotopy reduces to that
of a homotopy. Hence, homotopy is a special case of relative homotopy.
Before exploring homotopy and relative homotopy further, we formalize the idea of defor-
mation retraction, using Definition 2.1.1.Described in terms of functions, the initial state of
X (before deformation) is realised as the identity map IX : X → X (i.e., IX (x) = x for all
x ∈ X) and the final state (after deformation) by a surjective map f : X → A such that
f |A = IA ( f |A = IA means that in the process of deformation A remains unchanged or
stationary), then considering the inclusion map i : A → X, we feel that ϕt : X → A is a
homotopy from IX to i ◦ f relative to A. Equivalently,

Definition 2.1.2. A subspace A of X is called a deformation retract of X if there is


a continuous map H : X × I → X such that H(x, 0) = x, H(x, 1) ∈ A for all x ∈ X
and H(a, t) = a for all t ∈ I, for all a ∈ A. This homotopy H is called a deformation
retraction.
A. Deb Ray, C.U. 9

If there exists a singleton subspace A of X such that A is a deformation retract of X then


X is called a contractible space w.r.t. the singleton. So, in otherwords, if there exists
x0 ∈ X such that IX : X → X is homotopic to Cx0 : X → X given by Cx0 (x) = x0 , for all
x ∈ X, relative to the point x0 , then X is called contractible relative to x0 .

Exercise 2.1.1. Show that the definition of deformation retraction given by the family of
functions ϕt , t ∈ [0, 1] is equivalent to Definition 2.1.2.

The name ‘contractiblity’ of a space relative to a point seems quite understandable. For
contracting a space to a point there is no need to put the restriction of keeping the point x0
stationary. This leads to the definition of a contractible space. i.e.,

Definition 2.1.3. A space X is called contractible, if there exists x0 ∈ X such that


IX : X → X is homotopic to Cx0 : X → X given by Cx0 (x) = x0 , for all x ∈ X.

In Section 3 of this chapter we discuss contractible spaces in more details and will see that
a contractible space need not be contractible w.r.t a point, though the other implication is
immediate.
It is already mentioned before, if A is a subspace of X and there exists a continuous function
f : X → A such that f |A = IA and i ◦ f ∼ =A IX then A is a deformation retract of X. It is
not hard to observe that in such case, f ◦ i = IA ∼ = IA holds. Generalizing this idea we get
the definition of homotopy equivalence, as given below:

Definition 2.1.4. A continuous function f : X → Y is said to be a homotopy equivalence


if there exists a continuous function g : Y → X such that g ◦ f ∼
= IX and f ◦ g ∼
= IY . In such
case, f and g are called homotopy inverse of each other and X and Y are said to be of the
same homotopy type or homotopically equivalent spaces.

A most natural example of homotopy equivalence is a homeomorphism. Because, if X and


Y are homeomorphic there exist continuous functions f, g : X → Y such that g ◦ f = IX
and f ◦ g = IY . As ‘homotopy’ is a reflexive relation, g ◦ f = IX ∼
= IX and f ◦ g = IY ∼
= IY
and hence, f is a homotopy equivalence.

Exercise 2.1.2. Show that the relation X ∼ Y if and only if X and Y are of the same
homotopy type is an equivalence relation.

In the next two sections we discuss these concepts in details.

2.2 Homotopy and relative homotopy - some examples


and basic properties
We recall the definitions of homotopy and relative homotopy from Definition 2.1.1.

Example 2.2.1. Let f, g : Rn → Rn are given by f (x) = x and g(x) = 0, for all x ∈ Rn .
F
Define F : Rn × I → Rn by F (x, t) = (1 − t)x then f ∼
= g.
A. Deb Ray, C.U. 10

In Example 2.2.1, we have constructed one F such that F is continuous, F (x, 0) = f (x)
and F (x, 1) = g(x). But one may note that F is not the only function satisfying those
conditions. In fact, H : Rn × I → Rn given by H(x, t) = (1 − t2 )x also serves the purpose.
So, there are several ways a function can homotope to another function.

The following example explains that there is nothing special about the chosen functions
f and g on Rn in Example 2.2.1:
Example 2.2.2. Let f, g : Rn → Rn are any two continuous functions. Define F : Rn × I →
Rn by F (x, t) = tg(x)+(1−t)f (x). Then F is a continuous function such that F (x, 0) = f (x)
F
and F (x, 1) = g(x). i.e., f ∼
= g.
Even further, the domain of f and g in Example 2.2.2 need not be Rn and can be replaced
by any topological space X. What is so special about Rn in the co-domain? The only thing
we need is that the convex combination of two points of Rn should be inside Rn . Therefore,
we get a more general example :
Example 2.2.3. Let X be a topological space and Y a convex subset of Rn (or, Y be
a topological vector space) and f, g : X → Y be any two continuous functions. Define
F : X × I → Y by F (x, t) = tg(x) + (1 − t)f (x). Then F is a continuous function such that
F
F (x, 0) = f (x) and F (x, 1) = g(x). i.e., f ∼
= g.
The homotopy described in Example 2.2.1 and 2.2.2 is called straight line homotopy.
However, ‘convexity’ of the co-domain space is not mandatory for any two continuous func-
tions to be homotopic.
Example 2.2.4. Let X be a topological space and S n−1 be the n − 1-sphere of unit radius.
Then of course, S n−1 ⊆ Rn \ {0}. If f, g : X → S n−1 are any two continuous functions such
that f (x) and g(x) are not antipodal (i.e., f (x) ̸= −g(x)) then F : X × I → S n−1 given by
tg(x)+(1−t)f (x)
F (x, t) = ||tg(x)+(1−t)f (x)||
is a homotopy. It is easy to see that if f (x) ̸= −g(x), each point on
tg(x)+(1−t)f (x)
the line segment tg(x) + (1 − t)f (x) has norm > 0. So, ||tg(x)+(1−t)f (x)||
has unit length and
n−1
therefore, ∈ S for each x ∈ X and t ∈ I. The continuity of F follows from the facts that
the function ϕ : X × I → Rn \ {0} given by ϕ(x, t) = tg(x) + (1 − t)f (x) is continuous and
x
ψ : Rn \ {0} → S n−1 given by ψ(x) = ||x|| is continuous. (check!)
This example leads to a result that follows next:
Theorem 2.2.1. Any non-surjective continuous function f : X → S n−1 is null homotopic.
Proof. Since f is non-surjective, we can choose s ∈ S n−1 \f (X) and define a constant function
g : X → S n−1 by g(x) = −s. Clearly, f (x) ̸= −g(x) holds for all x ∈ X. Proceeding as
Example 2.2.4, f ∼ = g.
Theorem 2.2.2. Consider the collection C(X, Y ) of all continuous functions from X to Y .
Define a relation on C(X, Y ) by
f∼
= g ⇔ f is homotopic to g
Then ∼
= is an equivalence relation.
A. Deb Ray, C.U. 11

Proof. For any f ∈ C(X, Y ) set H(x, t) = f (x) for all x ∈ X and t ∈ I. So, ∼ = is reflexive.

If f, g ∈ C(X, Y ) are such that f = g then there exists a homotopy H : X × I → Y
such that H(x, 0) = f (x) and H(x, 1) = g(x). Consider the map F : X × I → Y defined
by F (x, t) = H(x, 1 − t). Then F is also a continuous map such that F (x, 0) = g(x) and
F (x, 1) = f (x) and hence, g ∼
= f . i.e., ∼
= is symmetric.
H1 H2
Let f, g, h ∈ C(X, Y ) are such that f ∼ = g and g ∼ = h. Define H : X × I → Y by
(
H1 (x, 2t) if 0 ≤ t ≤ 12
H(x, t) =
H2 (x, 2t − 1) if 21 ≤ t ≤ 1
for all x ∈ X. By pasting lemma, H is continuous. Further, H(x, 0) = f (x) and H(x, 1) =
h(x) for all x ∈ X. Therefore, f ∼
= h and consequently, ∼
= is transitive.
Exercise 2.2.1. Let X and Y be topological spaces and A ⊆ X. Show that the ‘homotopy
relative to A’ is an equivalence relation on C(X, Y ) of all continuous functions from X to Y .
That ∼ = is an equivalence relation is efficiently utilized to establish a very useful result
regarding composition of homotopic maps.
Theorem 2.2.3. If f1 , g1 : X → Y and f2 , g2 : Y → Z are continuous functions such that
f1 ∼
= g1 and f2 ∼
= g2 then f2 ◦ f1 ∼
= g2 ◦ g1 .
H1 2H
Proof. Let f1 ∼= g1 and f2 ∼
= g2 . H1 : X × I → Y and f2 : Y → Z are continuous imply that
f2 ◦ H1 : X × I → Z is a continuous function. Again (f2 ◦ H1 )(x, 0) = f2 (f1 (x)) = (f2 ◦ f1 )(x)
f2 ◦H1
and (f2 ◦ H1 )(x, 1) = f2 (g1 (x)) = (f2 ◦ g1 )(x), for all x ∈ X. Hence f2 ◦ f1 ∼
= f 2 ◦ g1 .
Define ψ : X × I → Y × I by ψ(x, t) = (g1 (x), t) and H : X × I → Z by H = H2 ◦ ψ. Then H
is a continuous function such that H(x, 0) = (H2 ◦ ψ)(x, 0) = H2 (g1 (x), 0) = (f2 ◦ g1 )(x) and
H
H(x, 1) = (H2 ◦ ψ)(x, 1) = H2 (g1 (x), 1) = (g2 ◦ g1 )(x), for all x ∈ X. Hence f2 ◦ g1 ∼
= g2 ◦ g1 .
Since ‘being homotopic to’ is a transitive relation, f2 ◦ f1 ∼ = g2 ◦ g1 .
Recall that for any two continuous functions f : X1 → Y1 and g : X2 → Y2 , f ×g : X1 ×X2 →
Y1 ×Y2 given by (f ×g)(x, y) = (f (x), g(y)) is continuous. Similarly, the continuous functions
f : X → Y and g : X → Z also define a continuous function h : X → Y × Z described by
h(x) = (f (x), g(x)).
Theorem 2.2.4. (1) If f1 , f2 : X1 → X2 and g1 , g2 : Y1 → Y2 are such that f1 ∼ = f2 and
∼ ∼
g1 = g2 , then f1 × g1 = f2 × g2 .
(2) If f1 , f2 : X → Y and g1 , g2 : X → Z are such that f1 ∼ = f2 and g1 ∼ = g2 then h1 ∼
= h2 ,
where hi (x) = (fi (x), gi (x)), for i = 1, 2.
H 1 H 2
Proof. (1) Let f1 ∼ = f2 and g1 ∼
= g2 . Define H : (X1 × Y1 ) × I → X2 × Y2 by H((x, y), t) =
(H1 (x, t), H2 (y, t)). Then H is a continuous function such that H((x, y), 0) = (f1 (x), g1 (y)) =
H
(f1 × g1 )(x, y) and H((x, y), 1) = (f2 (x), g2 (y)) = (f2 × g2 )(x, y). So, f1 × g1 ∼
= f2 × g2 .
H 1 H 2
(2) Let f1 ∼ = f2 and g1 ∼= g2 . Define H : X × I → Y × Z by H(x, t) = (H1 (x, t), H2 (x, t)).
Then H is a continuous function with H(x, 0) = (f1 (x), g1 (x)) = h1 (x) and H(x, 1) =
H
(f2 (x), g2 (x)) = h2 (x). Therefore, h1 ∼
= h2 .
A. Deb Ray, C.U. 12

More generally,
Q
Theorem 2.2.5. Let X and Yα (α ∈ Λ) be topological spaces and f, g : X → Y = Yα
α∈Λ
be continuous. f ∼
= g if and only if pα ◦ f ∼
= pα ◦ g, for all α ∈ Λ. (Here pα denotes the α-th
projection map)

Proof. If f ∼= g then pα ◦ f ∼= pα ◦ g, for all α ∈ Λ, follows from Theorem 2.2.3 and that ∼=
is a reflexive relation.
Conversely, let pα ◦ f ∼ = pα ◦ g, for all α ∈ Λ. For each α ∈ Λ, there is a continuous
function Hα : X × I → Yα such that Hα (x, 0) = (pα ◦ f )(x), Hα (x, 1) = (pα ◦ g)(x). Define
H : X ×I → Y by H(x, t) = (Hα (x, t)α . Then H is a continuous function such that H(x, 0) =
(Hα (x, 0))α = ((pα ◦ f )(x))α = f (x) and H(x, 1) = (Hα (x, 1))α = ((pα ◦ g)(x))α = g(x), for
H
all x ∈ X. Hence, f ∼ = g.

Exercise 2.2.2. If X is a discrete space and f : X → X is homotopic to the identity map


IX : X → X, then f = IX .
Since f is homotopic to IX , there exists a continuous function F : X × I → X such that
F (x, 0) = f (x) and F (x, 1) = x, for all x ∈ X. F is continuous and {x} × I is connected.
So, F ({x} × I) is a connected subset of the discrete space X and hence a singleton. i.e.,
f (x) = F (x, 0) = F (x, 1) = x. It proves that f = IX .

We have already observed that the definition of homotopy can be derived from the definition
of relative homotopy, by choosing in particular the set A as an empty set. But a ‘homotopy’
is not always a ‘homotopy relative to a given set A’, where A is a nonempty set. Before
giving an example, we prove an important characteristic to homotopy relative to a set.

Theorem 2.2.6. If X is a topological space such that the identity map IX is homotopic to
a constant map Ca relative to {a} then for each neighbourhood U of a in X, there exists a
neighbourhood V of a contained in U such that any point of V can be joined to a by a path
lying completely inside U .

Proof. Done in Class.

Example 2.2.5. (Comb space) Consider the subspace C = {(x, 0) ∈ R2 : 0 ≤ x ≤


1} ∪ {(1/n, y) : 0 ≤ y ≤ 1, n ∈ N} ∪ {(0, y) : 0 ≤ y ≤ 1} of the Euclidean plane R2 . Let
IC : C → C be the identity map and C(0,1) : C → C, the constant map that takes all points
of C to the point (0, 1). Then IC and C(0,1) are homotopic to each other but they are not
homotopic relative to {(0, 1)}. (apply the above theorem to check this).

Exercise 2.2.3. If f and g are continuous functions from a topological space X to Rn such
that f (x) = g(x) for all x ∈ A ⊆ X, then show that f ∼A g.

Exercise 2.2.4. If f, g : X → Y are homotopic and A ⊆ X then show that f |A ∼


= g|A .

Exercise 2.2.5. Find an example to establish that any two constant functions from any
space X to any space Y need not be homotopic. (Solution : Choose X to be a connected
space and Y a disconnected space (hence, not path connected). Choose y, z from two distinct
A. Deb Ray, C.U. 13

components of Y and consider the constant functions on X as f (x) = y and g(x) = z. Then
f is not homotopic to g. For if there is a homotopy H : X × I → Y , then H(X × I) (being
a continuous image of a connected space) is a connected subset of Y and so can not contain
both y and z, a contradiction to the fact that H(x, 0) = y and H(x, 1) = z.)

Exercise 2.2.6. Show that any two constant functions from any space X to a path connected
space Y are homotopic.

One may feel that ‘homotopy’ does not only depend on the nature of the continuous functions
but also on the characteristic of the space Y . We may even generalize the last exercise as
follows:

Exercise 2.2.7. Any two constant functions from a topological space X to a topological
space Y are homotopic if and only if their images lie in one path component of Y .

2.3 Contractible spaces


We have already defined contractible spaces in Definition 2.1.3. It is not hard to guess that
the spaces Rn , the n-ball Dn , more generally convex subsets of Rn are contractible. It is
evident from Example 2.2.3, that our intuitive understanding of the term ‘contractibility’
matches the formal definition. We generalize that example a bit more as follows:

Example 2.3.1. A subspace X of Rn is said to be star-shaped, if there exists some x0 ∈ X


such that tx + (1 − t)x0 ∈ X, for each x ∈ X and 0 ≤ t ≤ 1. A star-shaped subspace of Rn
is contractible.
Define H : X × I → X by H(x, t) = tx + (1 − t)x0 . Then H(x, 0) = x0 = Cx0 (x) and
H
H(x, 1) = x = IX (x), proving Cx ∼
0= IX .

Note that a star-shaped subspace of Rn need not be a convex subspace, though the re-
verse implication certainly holds. For example, if we take two lines L1 and L2 in Rn with
L1 ∩ L2 = {a} for some a ∈ Rn , then L1 ∪ L2 is a star-shaped non-convex subspace of Rn .
In view of Example 2.2.5, Comb space is contractible. A rigorous proof of this statement is
given later.

Example 2.2.2 shows that a discrete space (with at least two points) is not contractible.
For if any constant function Cx is homotopic to IX then Cx = IX , proving that X is a
singleton.
In the next theorem we show that path connectedness is a necessary condition for a space
to be contractible. That the condition is not sufficient is well understood by visualizing the
unit circle S 1 . For a rigorous proof, we have to wait till the next chapter where we discuss
the fundamental group of S 1 . It is quite expected that if a space X is contractible to {x0 }
then it is surely contractible to any point x ∈ X.

Theorem 2.3.1. Any two continuous functions from a topological space to a contractible
space are homotopic.
A. Deb Ray, C.U. 14

Proof. Let f, g : X → Y be continuous functions from a space X to a contractible space


Y . By definition of contractible space, there exists some y0 ∈ Y such that IY ∼
= Cy0 . Since
∼ ∼
homotopy is a reflexive relation, f = f and hence by Theorem 2.2.2 IY ◦ f = Cy0 ◦ f . Now,
C
by
(Cy0 ◦ f )(x) = y0 shows that Cy0 ◦ f : X → Y is a constant function x 7→0 y0 . Similarly,
Cy0
IY ◦ g ∼
b
= Cy0 ◦ g and Cy0 ◦ g : X → Y is the same constant function x 7→ y0 . Hence,
f = IY ◦ f ∼
=Cby0 ∼
= IY ◦ g = g. By transitivity of homotopy it follows that f ∼
= g.
Corollary 2.3.1. A space X is contractible if and only if the identity map IX is homotopic
to a constant map Cx , for any x ∈ X.
Proof. If the space X is contractible, by Theorem 2.3.1, IX ∼
= Cx , for any x ∈ X. Converse
is obvious.
Theorem 2.3.2. A contractible space is path connected.
Proof. Let X be a contractible space and y, z ∈ X. By Theorem 2.3.1, Cy ∼ = Cz . Using
Exercise 2.2.7, y and z are in the same path component of X. By arbitrariness of the points
y, z ∈ X, it follows that X is path connected.
Theorem 2.3.3. Product of contractible spaces is contractible.
Proof. Suppose {Xα : α ∈ Λ} is a collection of contractible spaces and X = Πα∈Λ Xα .
By Theorem 2.3.1, every projection map pα : X → Xα given by pα ((xα )) = xα is null
homotopic. i.e., there exists some zα ∈ Xα such that Czα : X → Xα is homotopic to pα .
Using Theorem 2.2.5, p : X → X given by p(x) = (pα (x)), where x = (xα ), is homotopic to
C : X → X given by C(x) = (zα ). But such p is the identity function on X and hence IX is
null homotopic, proving X is contractible.
We recall from Definition 2.1.4 that two spaces are of the same homotopy type if there is a
homotopy equivalence between them. The following theorem meets our natural expectation.
Theorem 2.3.4. A space is contractible if and only if it is the same homotopy type with a
singleton topological space P = {p}.
H
Proof. Let X be a contractible space and IX ∼
= Cx0 , for some x0 ∈ X. Define maps C : X →
H
P by C(x) = p for all x ∈ X and f : P → X by f (p) = x0 . Then C ◦ f = IP and f ◦ C ∼= IX .
Hence X and P are of the same homotopy type.
Conversely, suppose there are continuous functions g : X → P and h : P → X such that
g◦h ∼ = IP and h ◦ g ∼ = IX . Let h : P → X be h(p) = q, for some q ∈ X. Since for all
x ∈ X, (h ◦ g)(x) = h(g(x)) = h(p) = q, h ◦ g = Cq , showing that Cq ∼
= IX . Therefore, X is
contractible.
Corollary 2.3.2. Any two contractible spaces are of the same homotopy type.
Example 2.3.2. Comb space C is contractible : Let L be the line segment joining (0, 0)
and (1, 0) and p : C → L is the projection map given by p(x, y) = (x, 0). If i : L → C is the
inclusion map (i.e., i(x, 0) = (x, 0)) then p ◦ i = IL and the map H : C × [0, 1] → C given by
A. Deb Ray, C.U. 15

H((x, y), t) = (x, y)(1−t)+(x, 0)t = (x, (1−t)y) is a continuous map such that H((x, y), 0) =
IC (x, y) and H((x, y), 1) = i◦p(x, y), for all (x, y) ∈ C. Hence i◦p ∼
= IC . So, p is a homotopy
equivalence. i.e., C and L are of the same homotopy type. Since L is contractible by the
homotopy F : L × [0, 1] → L given by F ((x, 0), t) = tIL (x, 0) + (1 − t)C(0,0) (x, 0), we get that
C is contractible. (why?)

By Example 2.2.5, Comb space is not contractible relative to (0, 1).


As remarked earlier, if a space A is a deformation retract of X then A and X are of the same
homotopy type, because, i ◦ f is a homotopy equivalence where i : A → X is the inclusion
map and f : X → A is a surjective map such that f |A = IA . So, there are several examples
of non-contractible spaces which are of the same homotopy type. One such example is :

Example 2.3.3. Consider the annulus X = {z ∈ C : 1 ≤ |z| ≤ 2} and the circle


z
A = S 1 = {z ∈ C : |z| = 1}. If f : X → A is defined as f (z) = |z| , and i : A → X is the
inclusion map z ,→ z, then it is easy to see that f ◦ i = IA . On the otherhand i ◦ f ∼= IX .
Hence X and A are of the same homotopy type, but intuition ensures that none of them are
contractible. (We shall formally prove their non-contractibility in a subsequent chapter.)

Exercise 2.3.1. Show that the annulus A = {z ∈ C : 2 ≤ |z| ≤ 3} is of the same homotopy
type with S 1 .

Though it is generally hard to prove that two spaces are not of the same homotopy type,
the following theorem can be used to distinguish homotopy nonequivalent spaces.

Theorem 2.3.5. A 2-point discrete space is not of the same homotopy type with a point
space.

Proof. Let X = {a, b} be a 2-point discrete space and Y = {p} is a point space. If possible, let
X and Y are of the same homotopy type. Then there exist continuous functions f : X → Y
and g : Y → X such that g ◦ f ∼ = IX and f ◦ g ∼= IY . Clearly, f is the constant map Cp and
thus, f ◦ g = IY . Without loss of generality, we assume that g(p) = a. Let H : X × I → X
be a homotopy between g ◦ f and IX . Define w : I → X by w(t) = H(b, t). Then

w(0) = H(b, 0) = g(f (b)) = g(p) = a

and
w(1) = H(b, 1) = IX (b) = b.
Since I is connected and w is continuous, w(I) is also connected. But. X being a discrete
space, it is disconnected. So, w(I) can not be the whole of X. This contradiction proves
that X and Y are not of the same homotopy type.

This idea can further be generalized :

Theorem 2.3.6. If X is connected and Y is disconnected then X and Y are not of the same
homotopy type.
A. Deb Ray, C.U. 16

Proof. If possible let X and Y are of the same homotopy type. Then there are continuous
H F
maps f : X → Y and g : Y → X such that g ◦ f ∼ = IX and f ◦ g ∼= IY . So, F (y, 0) = f (g(y))
and F (y, 1) = y, for all y ∈ Y`. Y is disconnected implies that there are nonempty open sets
U , V in Y such that Y = U V . There exists a continuous map p : Y → {a, b}, such that
(
a, y ∈ U
p(y) =
b, y ∈ V

On the other hand, X being connected, f (X) ⊆ U or f (X) ⊆ V . Without loss of generality
assume, f (X) ⊆ U . Choose v ∈ V (such a choice is possible as V ̸= ∅). Define h : I → {a, b}
by h(t) = p(F (v, t)). Then
h(0) = p(F (v, 0)) = p(f (g(v))) = a (since f (g(v)) ∈ f (X) ⊆ U ) and h(1) = p(F (v, 1)) =
p(v) = b. i.e., h is a continuous surjection from a connected set I to a 2-point discrete space
{a, b}. This is a contradiction.’

Remark 2.3.1. Stating contrapositively, Theorem 2.3.6 reads as : If two spaces are of the
same homotopy type then either both of them are connected or both are disconnected. So,
‘connectedness’ can be used to distinguish spaces which are not of the same homotopy type.
But, ‘compactness’ can not be used for a similar purpose. There are examples of non-compact
spaces (e.g., R) that are of the same homotopy type with compact spaces (e.g., point space).

Definition 2.3.1. A property of topological space is called a homotopy property if it is


preserved by every homotopy equivalence. i.e., A homotopy property is one which possessed
by a space X is also possessed by every space which is of the same homotopy type with X.

Since homeomorphism is a homotopy equivalence, homotopy property is always a topological


property. But the converse is not true. e.g., ‘compactness’ is a topological property but not
a homotopy property. Note that, Corollary 2.3.2 shows that ‘contractiblity’ is a homotopy
property.
In the next theorem we prove that the ‘number of components’ of a space is a homotopy
property. To prove this theorem we need a lemma :

Lemma 2.3.1. Let X, Y be arbitrary topological spaces and Γ(X), Γ(Y ) denote the set of
all components of X and Y respectively. For any continuous function f : X → Y , define
a function Γ(f ) : Γ(X) → Γ(Y ) by Γ(f )(α) = β whenever f (α) ⊆ β. If f, g : X → Y are
homotopic then Γ(f ) = Γ(g).

Proof. Let α ∈ Γ(X). Since f ∼ = g, there exists a continuous function H : X × I → Y


such that H(x, 0) = f (x) and H(x, 1) = g(x) for all x ∈ X. Since α × I is connected and
continuous image of a connected set is connected, H(α × I) is connected and therefore there
exists a unique β ∈ Γ(Y ), such that H(α×I) ⊆ β. For any x ∈ α, f (x), g(x) ∈ H(α×I) ⊆ β.
i.e., f (α) ⊆ β and g(α) ⊆ β. So, Γ(f )(α) = Γ(g)(α). The arbitrariness of α ∈ Γ(X) proves
the claim.

Theorem 2.3.7. If f : X → Y is a homotopy equivalence, then Γ(f ) : Γ(X) → Γ(Y ) is


bijective.
A. Deb Ray, C.U. 17

Proof. Since f is a homotopy equivalence, there exists g : Y → X such that g ◦ f ∼ = IX



and f ◦ g = IY . Then by Lemma 2.3.1, Γ(g) ◦ Γ(f ) = Γ(g ◦ f ) = Γ(IX ) = IΓ(X) . Similarly,
Γ(f ) ◦ Γ(g) = Γ(f ◦ g) = Γ(IY ) = IΓ(Y ) . i.e., Γ(f ) and Γ(g) are inverse of each other.
Corollary 2.3.3. Connectedness is a homotopy property.
Proof. Follows from Theorem 2.3.7.

2.4 Retracts and Deformation retracts


We begin with the definition of a topological analogue of the projection operator, called
retraction.
Definition 2.4.1. A subspace A of a topological space X is called a retract of X if there
exists a continuous function r : X → A such that r|A = IA . i.e., r(a) = a for all a ∈ A. Such
a map r is known as a retraction.
A retraction defined in Definition 2.4.1 is always a surjective map. However, one may
equivalently say that a retraction is a map r : X → X such that r2 = r. In such case,
A = Im(r) is the retract of X and the map r is certainly not surjective, except for the trivial
case when r = IX .
The following is a trivial example of a retraction:
Example 2.4.1. Let X be any topological space and x0 ∈ X. Consider the subspace
A = {x0 } and the constant map Cx0 : X → A given by Cx0 (x) = x0 . Then clearly A is a
retract of X and Cx0 is a retraction.
From this example it is easy to see that a space may have several (may be infinitely many)
retracts. In fact, Example 2.4.1 suggests that any singleton of a space X can be viewed
trivially as a retract of X. We are actually interested in a special type of retract, called
deformation retract as defined in the introduction of this chapter. We recall the definition
here :
Definition 2.4.2. Let A be a subspace of a topological space X. A is called a deformation
retract of X if there is a continuous map H : X ×I → X such that H(x, 0) = x, H(x, 1) ∈ A
for all x ∈ X and H(a, t) = a for all t ∈ I, for all a ∈ A. This homotopy H is called a
deformation retraction.
Example 2.4.2. S 1 is a deformation retract of R2 \ {0}. (i.e., A deformation retract of an
annulus is a circle.)
z
Define f : R2 \ {(0, 0)} → S 1 by f (z) = |z| . Then f is a continuous function. If i : S 1 → R2 \
{(0, 0)} is the inclusion map z ,→ z, then f ◦i = IS 1 . Define H : R2 \{(0, 0)×I → R2 \{(0, 0)
by
z
H(z, t) = (1 − t)z + t
|z|
z
Then H is certainly a continuous map such that H(z, 0) = z = IS 1 (z), H(z, 1) = |z| =
1
(i ◦ f )(z) and H(a, t) = a, for all a ∈ S and for all t ∈ I. Hence, H is a deformation
retraction.
A. Deb Ray, C.U. 18

Since no special property of two dimensional space is utilized here, one may generalize
Example 2.4.2 for any n ∈ N :

Example 2.4.3. S n−1 is a deformation retract of Rn \ {0}.

Example 2.4.4. Circle is a deformation retract of a cylinder : Consider C =


{(x, y, 0) ∈ R3 : x2 + y 2 = 1}. Then to show that C is a deformation retract of
Y = {(x, y, z) ∈ R3 : x2 + y 2 = 1, −k ≤ z ≤ k}, k ∈ R. Define f : Y → C by
f (x, y, z) = (x, y, 0) and a map H : Y × I → Y by H((x, y, z), t) = (x, y, (1 − t)z). Then it
is easy to check that such an H is continuous. Now, H((x, y, z), 0) = (x, y, z) = IY (x, y, z);
H((x, y, z), 1) = (x, y, 0) = i ◦ f (x, y, z) and H((x, y, 0), t) = (x, y, 0), for all t ∈ I. So, H is
a homotopy between IY and i ◦ f relative to C. Hence C is a deformation retract of Y .

Example 2.4.5. Circle is a deformation retract of a möbius band : The möbius


band M is described as the set {(x, y) : 0 ≤ x ≤ a, −1 ≤ y ≤ 1} with identification
(0, y) ∼ (a, −y). Then the möbius band contains a circle C = {(x, 0) : 0 ≤ x ≤ a}. We shall
show that C is a deformation retract of M . Define a retract f : M → C by f (x, y) = (x, 0)
that preserves the gluing and a map H : M × I → M by H((x, y), t) = (x, (1 − t)y). Then
it is easy to check that such an H is continuous. Now, H((x, y), 0) = (x, y) = IM (x, y);
H((x, y), 1) = (x, 0) = i ◦ f (x, y) and H((x, 0), t) = (x, 0), for all t ∈ I. So, H is a homotopy
between IM and i ◦ f relative to C which also preserves the gluing. Hence C is a deformation
retract of M .

In the following example we observe that a space X may have more than one deformation
retracts, say, Y1 and Y2 , but none of Y1 and Y2 will deformation retract to the other.

Example 2.4.6. Consider a doubly punctured plane, i.e., a space obtained by removing
two points p, q from R2 . θ-space and figure-8 space are deformation retracts of a doubly
punctured plane R2 \ {p, q}.

The following figure shows how R2 \ {p, q} deformation retracts to figure-8 space :

On the otherhand, the following diagram shows how R2 \ {p, q} deformation retracts to a
θ-space:
A. Deb Ray, C.U. 19

But neither θ-space deformation retracts to figure-8 space nor it is a deformation retract
of a figure-8 space. However, Y1 and Y2 are of the same homotopy type with the doubly
punctured disc and hence, they are the same homotopy type with each other.

Exercise 2.4.1. If C is a deformation retract of B and B is a deformation retract of A then


C is a deformation retract of A.

Solution : Let iB : B → A and iC : C → B be the respective inclusion maps. Then it


is clear that IB (b) = iB (b), for all b ∈ B. By definition of deformation retracts, there are
continuous functions f : B → C and g : A → B such that iC ◦ f ∼ =C IB and iB ◦ g ∼=B IA .
So, (iC ◦ f ) ◦ g ∼C IB ◦ g = iB ◦ g ∼C IA . Hence, iC ◦ (f ◦ g) ∼C IA , which shows that C is
a deformation retract of A.

2.5 Extension of a continuous function


In this section we observe how existence of the extension of a continuous function relates to
null homotopy.

Theorem 2.5.1. Let Y be any topological space. A continuous function f : S n → Y is null


homotopic if and only if f can be extended to a continuous function F : Dn+1 → Y .

Proof. Suppose f is null homotopic. Then there exist y0 ∈ Y and a continuous function
H : S n × I → Y such that H(x, 0) = f (x), H(x, 1) = y0 , for all x ∈ S n . Define a function
F : Dn+1 → Y as follows:
(
y0 , if 0 ≤ ||x|| ≤ 21
F (x) = x
H( ||x|| , 2 − 2||x||), if 12 ≤ ||x|| ≤ 1

It is easy to see that 2 − 2||x|| ∈ I whenever 12 ≤ ||x|| ≤ 1. So, the map F is well-
defined. By pasting lemma, F is continuous. (Because, F is continuous on the closed sets
{x ∈ Dn+1 : 0 ≤ ||x|| ≤ 12 } and {x ∈ Dn+1 : 12 ≤ ||x|| ≤ 1} and F (x) = y0 = H(x, 1) for
||x|| = 12 .) Also, F (x) = f (x) for all x ∈ S n . Therefore, F is a continuous extension of f .
A. Deb Ray, C.U. 20

Let f : S n → Y be extended to F : Dn+1 → Y . Choose any p ∈ S n . For each x ∈ Dn+1 ,


the line segment (1 − t)x + tp ∈ Dn+1 , t ∈ I (since Dn+1 is a convex subset of Rn+1 ). Define
H : S n × I → Y by H(x, t) = F ((1 − t)x + tp). Then clearly H is a continuous function and
H
H(x, 0) = F (x) = f (x), for all x ∈ S n and H(x, 1) = F (p) = f (p). Hence, f ∼ Cf (p) .

Definition 2.5.1. let X be a topological space. A cone over X, denoted by T X, is the


quotient space X × I/ρ, where ρ is the equivalence relation on X × I given by (x, 1) ρ (z, 1)
for every x, z ∈ X.
Intuitively, take X × I and then paste all the points on X × {1} to a single point on X × {1}
(i.e., T X is obtained by pinching X × {1} to a single point).

The following theorem shows that ‘null homotopy’ provides a necessary and sufficient con-
dition for a continuous function on X to be extended to T X.

Theorem 2.5.2. Let X be any topological space and T X the cone over X. A continuous
function f : X → Y is null homotopic if and only if f has a continuous extension F : T X → Y
(where Y is any topological space).
H
Proof. Let f : X → Y be null homotopic. Then there exists some y0 ∈ Y such that f ∼ = C y0 ,
where Cy0 : X → Y is given by Cy0 (x) = y0 . If p : X ×I → T X is the identification map, then
p|X×{0} : X × {0} → p(X × {0}) is a homeomorphism (check!) and so, we identify X with
p(X × {0}) (as X and X × {0} are homeomorphic). Now, H(x, 0) = f (x) and H(x, 1) = y0 ,
for all x ∈ X. Then (H ◦ p−1 ) (z) is a single valued continuous function T X → Y such that
H ◦ p−1 (p(a, 0)) = f (a) for all a ∈ X. Hence, H ◦ p−1 : T X → Y is a continuous extension
of f .
Conversely, let f : X → Y has a continuous extension F : T X → Y . Consider F ◦ p :
X × I → Y . Then (F ◦ p)(x, 0) = f (x) for all x ∈ X. Also, (F ◦ p)(x, 1) is a single point,
say y0 ∈ Y . Hence, we get a homotopy H : X × I → Y given by H(x, t) = F ◦ p between f
and the constant map C(x) = y0 , for all x ∈ X. i.e., f ∼= 0.
The following example shows that f ∼
= g and f is extendible may not imply that g is
extendible.

Example 2.5.1. Consider the comb space C and A = {(a, 0) ; a = 0, 1, 12 , 13 , . . .}. If


f : A → C is the inclusion map f (a, 0) = (a, 0) and g : A → C given by g(a, 0) = (a, 1) then
clearly f and g are continuous functions. Also, f has the continuous extension Id : C → C.
But g can not be extended to a continuous function. However, H : A × I → C defined by
H((a, 0), t) = (1 − t)f (a) + tg(a) = (a, t) is a continuous function such that H((a, 0), 0) =
f (a, 0) and H((a, 0), 1) = g(a). i.e., f ∼
= g.
Exercise 2.5.1. Show that a function f : S n → Y is null homotopic if and only if it is null
homotopic relative to a point x0 of S n .
Chapter 3

Fundamental Group and its basic


properties

3.1 Path Homotopy


In what follows, X, Y denote topological spaces and I stands for the closed interval [0, 1] of
the real line R.
Definition 3.1.1. Let x0 , x1 ∈ X. Any continuous function α : I → X is called a path
from x0 to x1 , if α(0) = x0 and α(1) = x1 . The points x0 and x1 are respectively called the
initial and terminal points of the path α. In case the initial and terminal points coincide,
i.e., x0 = x1 , then the path α is called a loop based at x0 .
Definition 3.1.2. If α is a path from x0 to x1 and β is a path from x1 to x2 , then α ∗ β
defined as follows, is a path joining x0 and x2 :
(
α(2t), 0 ≤ t ≤ 12
(α ∗ β)(t) =
β(2t − 1), 21 ≤ t ≤ 1
α ∗ β is known as the product of the paths α and β.
Remark 3.1.1. It may be observed that
ˆ For any two paths α and β α ∗ β is defined if the terminal point of α coincides with
the initial point of β.
ˆ Clearly, for α ∗ β and β ∗ α both to be defined one needs x0 = α(0) = β(1) and
x1 = α(1) = β(0). Even if both the products are defined, the product of the paths is
generally not commutative.
ˆ The product of paths is not associative in general.
Definition 3.1.3. Let α and β be two paths in a topological space X joining x0 and x1 . If
there exists a continuous function H : I × I → X such that
H(s, 0) = α(s), H(s, 1) = β(s), H(0, t) = α(0) = β(0) = x0 and H(1, t) = α(1) = β(1) = x1 ,
then α is said to be homotopic to β, denoted by α ∼{x0 ,x1 } β or simply by α ∼ β. H is
called a path homotopy from α to β.
21
A. Deb Ray, C.U. 22

Note that H of the above definition is a special case of Homotopy relative to {0, 1}, as
discussed before. The following theorem guarantees that for three compatible paths α, β
and γ, (α ∗ β) ∗ γ and α ∗ (β ∗ γ) are path homotopic.

Theorem 3.1.1. If α, β and γ are paths in a topological space from x0 to x1 , from x1 to


x2 and from x2 to x3 respectively, then (α ∗ β) ∗ γ is path homotopic to α ∗ (β ∗ γ); i.e.,

(α ∗ β) ∗ γ ∼{x0 ,x3 } α ∗ (β ∗ γ)

Proof. The descriptions of the paths (α ∗ β) ∗ γ and α ∗ (β ∗ γ) are as follows:



α(4s),
 if 0 ≤ s ≤ 14
((α ∗ β) ∗ γ)(s) = β(4s − 1), if 14 ≤ s ≤ 12

γ(2s − 1), if 12 ≤ s ≤ 1

Schematically,

and

α(2s),
 if 0 ≤ s ≤ 12
(α ∗ (β ∗ γ))(s) = β(4s − 2), if 12 ≤ s ≤ 34

γ(4s − 3), if 34 ≤ s ≤ 1

So,

For constructing a path homotopy, consider the following diagram :


A. Deb Ray, C.U. 23

The equations of the lines l1 , l2 , l3 and l4 are given by


l1 : s = 0
l2 : t = 4s − 1
l3 : t = 4s − 2
l4 : s = 1
We define H : I × I → X by pasting three continuous functions, defined on
ˆ the region bounded by the lines t = 0, t = 1, l1 and l2 ;
i.e., {(s, t) : 0 ≤ s ≤ (t+1)
4
, 0 ≤ t ≤ 1};

ˆ the region bounded by the lines t = 0, t = 1, l2 and l3 ;


i.e., {(s, t) : (t+1)
4
≤ s ≤ (t+2)
4
, 0 ≤ t ≤ 1}; and

ˆ the region bounded by the lines t = 0, t = 1, l3 and l4 ;


i.e., {(s, t) : (t+2)
4
≤ s ≤ 1, 0 ≤ t ≤ 1}.

  
4s (t+1)
 (1+t) ,

 α if 0 ≤ s ≤ 4
, 0≤ t≤1
(t+1)
H(s, t) =
β (4s − 1 − t) , if 4
≤ s ≤ (t+2)
4
,0≤t≤1
γ 4s−2−t  ,
 (t+2)
2−t
if 4
≤ s ≤ 1, 0 ≤ t ≤ 1.

It is left as an exercise to the reader that H, as constructed above, is continuous (by pasting
lemma) and satisfies the desired conditions for being a path homotopy between the paths
(α ∗ β) ∗ γ and α ∗ (β ∗ γ).
As a special case of Theorem 3.1.1, we get
Theorem 3.1.2. If α, β, γ are loops based at some x0 ∈ X, then (α ∗ β) ∗ γ ∼{x0 } α ∗ (β ∗ γ).
Some important results are proved next, that we require in our subsequent discussion.
Theorem 3.1.3. Let α, α1 be two homotopic paths joining x0 , x1 and β, β1 be two homotopic
paths joining x1 , x2 . Then α ∗ β is path homotopic to α1 ∗ β1 . i.e.,

α ∼{x0 ,x1 } α1 and β ∼{x1 ,x2 } β1 ⇒ α ∗ β ∼{x0 ,x2 } α1 ∗ β1

Proof. By definition there exist continuous H1 : I × I → X and H2 : I × I → X such that


H1 (s, 0) = α(s), H1 (s, 1) = α1 (s), H1 (0, t) = x0 , H1 (1, t) = x1
and
H2 (s, 0) = β(s), H2 (s, 1) = β1 (s), H2 (0, t) = x1 , H2 (1, t) = x2
Define H : I × I → X by
(
H1 (2s, t), if 0 ≤ s ≤ 12 , 0 ≤ t ≤ 1
H(s, t) =
H2 (2s − 1, t), if 21 ≤ s ≤ 1, 0 ≤ t ≤ 1

Then H is continuous and H(s, 0) = (α ∗ β)(s), H(s, 1) = (α1 ∗ β1 )(s), H(0, t) = x0 and
H(1, t) = 1. This implies H is a path homotopy between α ∗ β and α1 ∗ β1 .
A. Deb Ray, C.U. 24

Corollary 3.1.1. Let α, α1 be two homotopic paths from x0 to x1 and β be any path from
x1 to x2 . Then α ∗ β is path homotopic to α1 ∗ β.

Proof. It is immediate as β ∼ β (since ∼ is a reflexive relation) and by applying Theo-


rem 3.1.3 we get α ∗ β is path homotopic to α1 ∗ β.

As a special case of Theorem 3.1.3, we obtain

Theorem 3.1.4. Let α, α1 , β, β1 be loops at x0 ∈ X such that α is path homotopic to α1


and β is path homotopic to β1 . Then α ∗ β is path homotopic to α1 ∗ β1 . i.e.,

α ∼{x0 } α1 and β ∼{x0 } β1 ⇒ α ∗ β ∼{x0 } α1 ∗ β1

Throughout our discussion, we use the following definitions / notations, without any further
clarification.

Definition 3.1.4. (i) For any path α : I → X from x0 to x1 , define α−1 : I → X by


α−1 (t) = α(1 − t), which is a path from x1 to x0 , called the inverse of α.
(ii) For any x ∈ X, Cx : I → X defined by Cx (t) = x is called a constant loop at x.

Theorem 3.1.5. If α is a path from x0 to x1 in a space X, then


(i) α ∗ Cx1 ∼ α
(ii) Cx0 ∗ α ∼ α
(iii) α ∗ α−1 ∼ Cx0
(iv) α−1 ∗ α ∼ Cx1

Proof. (i) Define H : I × I → X by


(
2s
, if 0 ≤ s ≤ t+1

α 1+t 2
, 0≤t≤1
H(s, t) =
x1 , if t+1
2
≤ s ≤ 1, 0 ≤ t ≤ 1

H is a path homotopy between α ∗ Cx1 and α.


(ii) Define H : I × I → X by
(
x0 , if 0 ≤ s ≤ 1−t
2
, 0≤t≤1
H(s, t) =
α 2s−1+t , if 1−t

1+t 2
≤ s ≤ 1, 0 ≤ t ≤ 1

H is a path homotopy between Cx0 ∗ α and α.


(iii) Define H : I × I → X by
(
α (2ts) , if 0 ≤ s ≤ 21 , 0 ≤ t ≤ 1
H(s, t) =
α (2t(1 − s)) , if 21 ≤ s ≤ 1, 0 ≤ t ≤ 1

H is a path homotopy between α ∗ α−1 and Cx0 .


(iv) Similar to (iii).

In particular, if α is a loop based at x0 ∈ X, then the following hold.


A. Deb Ray, C.U. 25

Theorem 3.1.6. If α is a loop based at x0 ∈ X, then


(i) α ∗ Cx0 ∼ Cx0 ∗ α ∼ α
(ii) α ∗ α−1 ∼ α−1 ∗ α ∼ Cx0

Exercise 3.1.1. If α and β are paths from x0 to x1 ∈ X, then show that α ∼ β if and only
if α ∗ β −1 ∼ Cx0 .

Remark 3.1.2. It is important to note that ‘Path homotopy’ is an equivalence relation on


the collection of all loops based at some x0 ∈ X. This is a consequence of the facts that the
relation of ‘being homotopic to’ is an equivalence relation and path homotopy is a special
case of homotopy relative to a set.

In view of the above Remark 3.1.2, for each x0 ∈ X, if P = {α : I → X : α(0) = α(1) = x0 },


then P is partitioned by the equivalence classes determined by the relation α ∼{x0 } β. We
consider the set π(X, x0 ) consisting of the equivalence classes [α], corresponding to each
α ∈ P.

3.2 Construction of Fundamental group and Induced


homomorphism
Now we define a binary operation on the collection π(X, x0 ).

Theorem 3.2.1. Let (X, x0 ) be a pointed topological space and π(X, x0 ), the collection
of all equivalence classes of loops in X based at x0 , arising from the equivalence relation
α ∼ β iff α is path homotopic to β. Then ([α], [β]) 7→ [α] ◦ [β] = [α ∗ β] is a well defined
binary operation on π(X, x0 ).

Proof. For any α, β ∈ P, α ∗ β ∈ P so that [α ∗ β] ∈ π(X, x0 ). Let [α] = [α1 ] and [β] = [β1 ],
where α, β, α1 and β1 ∈ P. Then α ∼ α1 and β ∼ β1 . Now, α ∗ β ∼ α1 ∗ β1 . Hence,
[α] ◦ [β] = [α1 ] ◦ [β1 ], which shows that the operation is well defined.

Theorem 3.2.2. For any pointed topological space (X, x0 ), (π(X, x0 ), ◦) is a group; where
[α] ◦ [β] = [α ∗ β].

Proof. Follows directly from Theorems on path homotopy.

Definition 3.2.1. The group π(X, x0 ), obtained in Theorem 3.2.2 is called the first fun-
damental group or Poincaré group of X based at x0 .

The definition of fundamental group depends on the base point x0 . So, fundamental groups
are associated to pointed topological spaces rather than topological spaces, in general. Later,
we may convince ourselves by constructing examples of fundamental groups of topological
spaces that depend on the base point. However, we shall see next that if x0 and x1 are two
points lying in the same path component then their fundamental groups are isomorphic.

Theorem 3.2.3. Let X be a topological space and x0 , x1 ∈ X. If there is a path connecting


x0 and x1 then π(X, x0 ) is isomorphic to π(X, x1 ).
A. Deb Ray, C.U. 26

Proof. Let ω : [0, 1] → X be a path connecting x0 and x1 . Then ω(0) = x0 and ω(1) = x1 .
Define fω : π(X, x0 ) → π(X, x1 ) by fω ([α]) = [ω −1 ∗ α ∗ ω]. We first see that this map is
well defined. Let [α] = [β]. Then α ∼x0 β and therefore, ω −1 ∗ α ∗ ω ∼x1 ω −1 ∗ β ∗ ω. So,
fω ([α]) = fω ([β]).
If α ∼x0 α1 and β ∼x0 β1 then α ∗ α1 ∼x0 β ∗ β1 . Using this and a Theorem that we have
obtained in Chapter 5, we get

fω ([α] ◦ [β]) = fω ([α ∗ β])


= [ω −1 ∗ (α ∗ β) ∗ ω]
= [ω −1 ∗ (α ∗ Cx0 ∗ β) ∗ ω]
= [ω −1 ∗ (α ∗ ω ∗ ω −1 ∗ β) ∗ ω]
= [ω −1 ∗ α ∗ ω] ◦ [ω −1 ∗ β ∗ ω]
= fω ([α]) ◦ fω ([β]).

Hence, fω is a homomorphism. If fω−1 : π(X, x1 ) → π(X, x0 ) is defined by fω−1 ([γ]) =


[ω ∗ γ ∗ ω −1 ], then it is also a group homomorphism. We see that fω−1 and fω are inverse of
each other :

(fω ◦ fω−1 )([α]) = fω−1 ([ω −1 ∗ α ∗ ω])


= [ω ∗ (ω −1 ∗ α ∗ ω) ∗ ω −1 ]
= [Cx0 ∗ α ∗ Cx0 ]
= [α]

for all [α] ∈ π(X, x0 ). This establishes, fω is an isomorphism. Consequently, π(X, x0 ) ∼


=
π(X, x1 ).
Corollary 3.2.1. If X is path connected then for any two points x0 , x1 ∈ X, π(X, x0 ) ∼
=
π(X, x1 ).
Henceforth, in view of the last result, we denote the fundamental group of a path connected
space X by π(X), without mentioning the base point explicitly.
Theorem 3.2.4. If X is a topological space and ω, η are homotopic paths joining x0 , x1
∈ X, then fω = fη .
Proof. Since ω ∼ η ⇒ ω −1 ∼ η −1 , we get ω ∗ α ∗ ω −1 ∼ η ∗ α ∗ η −1 , for any [α] ∈ π(X, x0 ).
Hence, fη ([α]) = [η ∗ α ∗ η −1 ] = [ω ∗ α ∗ ω −1 ] = fω ([α]).
The following result gives a necessary and sufficient condition for π(X, x0 ) to be abelian.
Theorem 3.2.5. Let X be a path connected space and x0 , x1 ∈ X. π(X, x0 ) is abelian if
and only if for each pair of paths ω and η from x0 to x1 , fω = fη .
Proof. Suppose π(X, x0 ) is abelian. Since ω ∗ η −1 is a loop based at x0 , [ω ∗ η −1 ] ∈ π(X, x0 ).
Hence, [ω ∗ η −1 ] ◦ [α] = [α] ◦ [ω ∗ η −1 ], for all [α] ∈ π(X, x0 ). Then ω ∗ η −1 ∗ α ∼ α ∗ ω ∗ η −1
which implies that η −1 ∗ α ∗ η ∼ ω −1 ∗ α ∗ ω. i.e., [η −1 ∗ α ∗ η] = [ω −1 ∗ α ∗ ω] and so,
fη ([α]) = fω ([α]) for all [α] ∈ π(X, x0 ).
A. Deb Ray, C.U. 27

Conversely, let the condition hold and [α], [β] ∈ π(X, x0 ). Since X is path connected, there is
a path ω (say) from x0 to x1 . Then β ∗ω is also a path from x0 to x1 . By the hypothesis, fω =
fβ∗ω . i.e., for all [α] ∈ π(X, x0 ), fω ([α]) = fβ∗ω ([α]). So, [ω −1 ∗ α ∗ ω] = [(β ∗ ω)−1 ∗ α ∗ (β ∗ ω)]
= [(ω −1 ∗ β −1 ) ∗ α ∗ (β ∗ ω)]. So, ω −1 ∗ α ∗ ω ∼ ω −1 ∗ (β −1 ∗ α ∗ β) ∗ ω which gives α ∼ β −1 ∗ α ∗ β,
i.e., β ∗ α ∼ α ∗ β. Hence, [β ∗ α] = [α ∗ β].

Definition 3.2.2. A topological space X is called simply connected if it is path connected


and its fundamental group is trivial.

In what follows, we shall mean a trivial fundamental group when we write π(X, x0 ) = 0.

We observe next that a continuous function between pointed topological spaces induces
a homomorphism between the corresponding fundamental groups.

Theorem 3.2.6. If f : (X, x0 ) → (Y, y0 ) is a continuous function between the pointed


topological spaces (X, x0 ) and (Y, y0 ), then there exists a homomorphism f# : π(X, x0 ) →
π(Y, y0 ).

Proof. By definition of continuous function between pointed spaces, we have f (x0 ) = y0 .


Define f# : π(X, x0 ) → π(Y, f (x0 )) by f# ([α]) = [f ◦ α]. We show that such f# is well
defined. Suppose [α] = [β] ∈ π(X, x0 ). Then α ∼x0 β. Consequently, f ◦ α ∼f (x0 ) f ◦ β. i.e.,
f# ([α]) = [f ◦ α] = [f ◦ β] = f# ([β]). For any [α], [β] ∈ π(X, x0 ) and t ∈ [0, 1],

f ◦ (α ∗ β)(t) = f ((α ∗ β)(t))


(
f (α(2t)), 0 ≤ t ≤ 21
= 1
f (β(2t − 1)), 2
≤t≤1
= ((f ◦ α) ∗ (f ◦ β))(t)

Hence, [f ◦ (α ∗ β)] = [(f ◦ α) ∗ (f ◦ β)]. So,

f# ([α] ◦ [β]) = f# ([α ∗ β])


= [f ◦ (α ∗ β)]
= [(f ◦ α) ∗ (f ◦ β)]
= [f ◦ α] ◦ [f ◦ β]
= f# ([α]) ◦ f# ([β]).

Therefore, f# is a homomorphism.

The function f# , defined in Theorem 3.2.6 is called the homomorphism induced by f .


The following are two important properties of the induced homomorphism.

Theorem 3.2.7. (i) If I : (X, x0 ) → (X, x0 ) is the identity map then I# : π(X, x0 ) →
π(X, x0 ) is the identity homomorphism.
(ii) If f : (X, x0 ) → (Y, y0 ) and g : (Y, y0 ) → (Z, z0 ) are continuous maps then (g ◦ f )# =
g# ◦ f# .
A. Deb Ray, C.U. 28

Proof. (i) For each [α] ∈ π(X, x0 ), (I ◦ α)(t) = α(t), for all t ∈ [0, 1] and hence, we have
I# ([α]) = [I ◦ α] = [α]. So, the induced homomorphism I# : π(X, x0 ) → π(X, x0 ) is the
identity map.
(ii) For all [α] ∈ π(X, x0 ),

(g ◦ f )# ([α]) = [(g ◦ f ) ◦ α]
= [g ◦ (f ◦ α)]
= g# ([f ◦ α])
= g# (f# ([α]))
= (g# ◦ f# )([α])

Hence, (g ◦ f )# = g# ◦ f# .

In view of Theorem 3.2.7, we conclude that π is a covariant functor π : T op∗ ⇒ Grp described
as follows:
(X, x0 ) 7−→ π(X, x0 )
f f#
 
(Y, y0 ) 7−→ π(Y, y0 )
The properties mentioned in Theorem 3.2.7 are referred to as functorial properties of π.

Exercise 3.2.1. Let X be a topological space and a ∈ A ⊆ X. Prove that


(i) the induced homomorphism r# : π(X, a) → π(A, a) is surjective if r : X → A is a
retraction;
(ii) the induced homomorphism i# : π(A, a) → π(X, a) is injective if i : A → X is the
inclusion map.
Solution. (i) By definition of induced homomorphism r# ([α]) = [r ◦ α]. Let [β] ∈ π(A, a).
Then [i ◦ β] ∈ π(X, a), where i : A → X is the inclusion map. Clearly, r# ([i ◦ β]) = [β],
proving the map to be surjective.
(ii) Follows immediately from injectiveness of the inclusion map.

3.3 Fundamental groups - Homeomorphic and Con-


tractible spaces
In this section, our aim is to prove two very basic but important results. The first one among
the two is that the fundamental groups of homeomorphic spaces are isomorphic. To prove
this result we use the functorial properties as discussed in the last section and observe how
crucial those properties are.

Theorem 3.3.1. If X and Y are two path connected homeomorphic topological spaces,
then π(X) is isomorphic to π(Y ).

Proof. Suppose f : X → Y is a homeomorphism. Since X and Y are path connected, their


fundamental groups are independent of the choice of base points. So choose any x0 ∈ X
A. Deb Ray, C.U. 29

such that y0 = f (x0 ) ∈ Y . We shall show that π(X, x0 ) ∼ = π(Y, y0 ). Since f : X → Y


−1
is a homeomorphism, f : (X, x0 ) → (Y, y0 ) and f : (Y, y0 ) → (X, x0 ) are continuous
maps between respective pointed topological spaces. Then f# : π(X, x0 ) → π(Y, y0 ) and
f#−1 : π(Y, y0 ) → π(X, x0 ) are the induced homomorphisms, such that

(f −1 ◦ f )# = (IX )# ⇒ (f −1 )# ◦ f# = (IX )# = Iπ(X,x0 ) .

and
(f ◦ f −1 )# = (IY )# ⇒ f# ◦ (f −1 )# = (IY )# = Iπ(Y,y0 ) .
Hence f# becomes a group isomorphism, as desired.

Theorem 3.3.2. If f, g : (X, x0 ) → (Y, y0 ) are two continuous functions between two pointed
topological spaces such that f ∼x0 g, then f# = g# : π(X, x0 ) → π(Y, y0 ).

Proof. Let [α] ∈ π(X, x0 ). Then f ∼x0 g ⇒ f ◦ α ∼x0 g ◦ α. Hence, [f ◦ α] = [g ◦ α] so that


f# ([α]) = g# ([α]) and arbitrariness of [α] proves the result.

A continuous function f : (X, x0 ) → (Y, y0 ) is called a homotopy equivalence between


pointed spaces if there exists a continuous function g : (Y, y0 ) → (X, x0 ) such that g ◦f ∼x0
IX and f ◦ g ∼y0 IY .

Theorem 3.3.3. If f : (X, x0 ) → (Y, y0 ) is a homotopy equivalence between pointed spaces,


then f# : π(X, x0 ) → π(Y, y0 ) is an isomorphism.

Proof. By definition of homotopy equivalence, for f : (X, x0 ) → (Y, y0 ) there exists a g :


(Y, y0 ) → (X, x0 ) such that g ◦f ∼x0 IX and f ◦g ∼y0 IY .By Theorem 3.3.2, (g ◦f )# = (IX )#
and (f ◦ g)# = (IY )# . Using functorial properties, (as stated before, we get g# ◦ f# = Iπ(X,x0 )
and f# ◦ g# = Iπ(Y,y0 ) , proving that f# is an isomorphism.

Corollary 3.3.1. If A is a deformation retract of X and a ∈ A then π(A, a) ∼


= π(X, a).

Proof. By definition of a deformation retract, the inclusion map i : (A, a) → (X, a) is a


homotopy equivalence. So, by Theorem 3.3.3, i# : π(A, a) → π(X, a) is an isomorphism.

Remark 3.3.1. As discussed before, the figure-8 space and θ-space are the deformation
retracts of the same space (a disc with two holes), and so, their fundamental groups are
isomorphic. But figure-8 space and θ-space are not deformation retracts of each other.
Hence, this cites an example of spaces whose fundamental groups are same though one is
not a deformation retract of the other.

Suppose, X and Y are two topological spaces and x0 ∈ X. In Theorem 3.3.3 we have shown
that if f : X → Y is a homotopy equivalence such that the base points x0 and f (x0 ) remain
unaltered during homotopy, then the induced homomorphism f# is an isomorphism. We see
next that even if the base points are not kept fixed during homotopy, the homomorphism
induced by a homotopy equivalence is an isomorphism.
A. Deb Ray, C.U. 30

Theorem 3.3.4. Let f, g : X → Y be continuous maps such that f (x0 ) = y0 and g(x0 ) = y1 .
If f is homotopic to g then there exists a path η in Y connecting y0 and y1 , determined by
the homotopy, such that the following diagram commutes:
π(X, x0 ) / π(Y, y0 )
f#
ηb
g#
& 
π(Y, y1 )
where ηb : π(Y, y0 ) → π(Y, y1 ) is given by ηb([β]) = [η −1 ∗ β ∗ η].
Proof. Suppose H : X × I → Y is the homotopy between f and g. i.e., H : X × I → Y is
a continuous function such that H(x, 0) = f (x) and H(x, 1) = g(x). Define η : [0, 1] → Y
by η(t) = H(x0 , t). Then η is certainly a continuous map with η(0) = y0 and η(1) = y1 . So,
η is a path in Y determined by the homotopy H. We claim that ηb ◦ f# = g# . In order to
establish our claim, we need to show [η −1 ∗ (f ◦ α) ∗ η] = [g ◦ α], for all [α] ∈ π(X, x0 ); i.e.,
to show η −1 ∗ (f ◦ α) ∗ η ∼ (g ◦ α).
We prove this result in two steps:
Step 1. Define a continuous map G : I × I → Y by G(s, t) = H(α(s), t). Then for each
s, t ∈ I
G(s, 0) = H(α(s), 0) = (f ◦ α)(s), G(s, 1) = F (α(s), 1) = (g ◦ α)(s)
and
G(0, t) = H(α(0), t) = H(x0 , t) = η(t), G(1, t) = H(α(1), t) = H(x0 , t) = η(t)
Step 2. Define a map F : I × I → Y as follows:

−1
η (2s),
 if 0 ≤ s ≤ 1−t
2
(1−t)
F (s, t) = G (4s + 2t − 2)/(3t + 1), t) , if 2 ≤ s ≤ t+3 4

η(4s − 3), if t+3 ≤ s ≤ 1

4

Using pasting lemma it is easy to see that F is a continuous function such that,
F (s, 0) = η −1 ∗ (f ◦ α) ∗ η
F (s, 1) = g ◦ α
F (0, t) = F (1, t) = g(x0 ).
Hence F is path homotopy between η −1 ∗ (f ◦ α) ∗ η and g ∗ α, as desired.
Corollary 3.3.2. If f : X → Y is homotopic to a constant map C : X → Y and x0 ∈ X
then the induced homomorphism f# : π(X, x0 ) → π(Y, f (x0 )) is the zero map.
Proof. Choose any [α] ∈ Π(X, x0 ). To show that f# ([α]) = [Cf (x0 ) ]. Let C : X → Y
be the constant map described by x 7→ y0 . Then by Theorem 3.3.4, we get the following
commutative diagram:
π(X, x0 ) f / π(Y, f (x0 ))
#

ηb
C# ' 
π(Y, y0 )
Here C# ([α]) = [C ◦ α] = [Cy0 ] and ηb ◦ f# = C# implies [η −1 ∗ (f ◦ α) ∗ η] = [Cy0 ]. So,
f ◦ α ∼ η ∗ Cy0 ∗ η −1 ∼ Cf (x0 ) . Hence, f# ([α]) = [f ◦ α] = [Cf (x0 ) ].
A. Deb Ray, C.U. 31

Corollary 3.3.3. The fundamental group of a contractible space is trivial.


Proof. Let X be a contractible space and x0 ∈ X. As every contractible space is path
connected, it is enough to show that π(X, x0 ) = {[Cx0 ]}. Since X is contractible, the
identity map I is homotopic to the constant map Cx0 (= C), by the homotopy H. Then
using Theorem 3.3.4, we get the following commutative diagram:
π(X, x0 ) / π(X, x0 )
I#
ηb
C# & 
π(X, x0 )

where ηb is described as in Theorem 3.3.4, induced from η(t) = H(x0 , t). By commutativity
of the diagram, we get, for all [α] ∈ π(X, x0 ),
η ◦ I# )([α]) = [η −1 ∗ α ∗ η]
[Cx0 ] = C# ([α]) = (b
Consequently, Cx0 ∼ η −1 ∗ α ∗ η, i.e., η −1 ∗ Cx0 ∗ η ∼ α and hence, Cx0 ∼ α. Therefore, by
arbitrariness of [α], we get π(X, x0 ) = {[Cx0 ]}.
Exercise 3.3.1. Find π(Rn ).
Solution. Let I : Rn → Rn be the identity map and C0 : Rn → Rn be the constant map
x 7→ 0. Define H : Rn × I → Rn by H(x, t) = tx which is a homotopy between I and
C0 . Therefore, Rn is contractible. Again Rn being contractible, it is path connected and
therefore the fundamental group does not depend on the base point. Hence, π(Rn ) is trivial
(by Corollary 3.3.3).
Theorem 3.3.5. If X and Y are two path connected spaces which are of the same homotopy
type then their fundamental groups are isomorphic.
Proof. By the definition of homotopy type, there are continuous functions f : X → Y and
H G
g : Y → X such that g ◦ f ∼= IX and f ◦ g ∼= IY . Let x0 ∈ X. Suppose f (x0 ) = y0 , g(y0 ) = x1
and f (x1 ) = y1 . Then (g ◦ f )(x0 ) = x1 and (f ◦ g)(y0 ) = y1 . Since X is path connected,
define a path from x0 to x1 by η(t) = H(x0 , t) and hence, applying Theorem 3.3.4, we get
the following commutative diagram:
π(X, x0 ) / π(X, x0 )
I#
ηb
(g◦f )# & 
π(X, x1 )

i.e., g# ◦ f# = (g ◦ f )# = ηb ◦ I# = ηb. Defining the path η −1 from x1 to x0 by η −1 (t) =


−1 are inverse of each other. Hence, η
H(x0 , 1 − t), we get that ηb and ηd b is an isomorphism.
Consequently, from g# ◦ f# = ηb it follows that f# is an injective function and g# is a
surjective function.
Proceeding similarly, from f ◦ g ∼ = IY , we get, f# is a surjective function and g# is an
inective function. Hence, both f# and g# are isomorphisms, proving π(X, x0 ) ∼ = π(Y, y0 )
(i.e., π(X) ∼
= π(Y )).
Chapter 4

Covering spaces and Covering maps

4.1 Covering spaces


In Chapter 3, it has been observed that the fundamental group of a circle is non-trivial.
The continuous function exp : R → S 1 described by x 7→ (cos 2πx, sin 2πx) plays crucial
role in the construction of the fundamental group of S 1 . We have established two major
properties of exp, known as Path Lifting property and Homotopy Lifting property.
In this chapter, we see that exp is a typical example of a covering map which serves as an
important tool for computing fundamental groups.

In this chapter, we explore in details certain general properties of covering maps and
covering spaces and prove Path Lifting property and Homotopy Lifting property
of any covering map. It is to be noted that ‘covering spaces’ are important in their own
right and have a wide range of applications in various branches of mathematics, but for our
purpose, we only discuss their use as a tool to compute fundamental groups.

e → X be a continuous surjection.
Definition 4.1.1. Let p : X

ˆ An open set U of X is said to be evenly covered by p in X, e if p−1 (U ) is expressed


e (in notation, p−1 (U ) = ` Vα ), such that each
as disjoint union of open sets Vα of X
α∈I
p|Vα : Vα → U is a homeomorphism. The collection {Vα } is called a partition of p−1 (U )
into slices and each Vα is called a slice over U .

ˆ If each x ∈ X has an open neighbourhood U in X such that U is evenly covered by


p in X,
e then p is called a covering map (or, covering projection) and X
e is called
a covering space of X. Such an open neighbourhood U of x is referred to as an
admissible open set.
e → X is a covering map if there is a cover for X by
Therefore, it is to note that p : X
admissible open sets. We begin with trivial examples of covering maps :

Example 4.1.1. (i) Let X be a topological space and IX : X → X be the identity map.
−1
For each x ∈ X, consider the open set X containing x. Then IX (X) = X is certainly evenly
32
A. Deb Ray, C.U. 33

covered by IX and the partition is {X}. So, IX is a covering map and X is a covering space
of itself.

(ii) Let X be a topological space. Consider the space X e = X × {1, 2, . . . , n}. Then of

course, Xe is a disjoint union of copies of X, i.e., X
e= (X × {i}). Defining p : X e → X by
i=1
p(x, i) = x, we observe that p is a covering map and hence X
e is a covering space of X.

e → X is trivially a covering map.


(iii) Any homeomorphism p : X

(iv) Let Y be a discrete space and X any topological space. Then the projection map
p`: X × Y → X given by (x, y) 7→ x is a covering map. Because, p−1 (X) = X × Y =
X × {y} and each X × {y} is homeomorphic to X.
y∈Y

Example 4.1.2. Consider exp : R → S 1 given by exp(t) = (cos 2πt, sin 2πt). Clearly exp is a
continuous surjection. For any point x ∈ S 1 \ {(1, 0)}, consider the open neighbourhood U =

S 1 \{(1, 0)}. Then exp−1 (U ) =
`
(n, n+1) such that each (n, n+1) is homeomorphic to U ,
n=−∞
i.e., U is evenly covered by exp. For the point (1, 0), consider the open set V = S 1 \{(−1, 0)}.

exp−1 (V ) = (n − 12 , n + 12 ) and each (n − 12 , n + 12 ) is homeomorphic to V imply that V
`
n=−∞
is also evenly covered by exp. Hence, exp is a covering map and R is a covering space of S 1 .
Example 4.1.2 differs from Example 4.1.1(ii) in the sense that the space X (and hence X)
e
1
of Example 4.1.1(ii) is not path connected while in Example 4.1.2, S (and also R) is path
connected. In this course, we are mainly concerned with path connected spaces and hence
in what follows, we discard the covering spaces which are not path connected.
Next example shows that a (path connected) topological space may have more than one path
connected covering spaces.
Example 4.1.3. Let p1 : S 1 → S 1 be given by p1 (z) = z 2 . Then p1 is a covering map and
hence, S 1 is a covering space of itself.
The above example can further be generalized as follows:
Example 4.1.4. Let p : S 1 → S 1 be given by p(z) = z n , for any nonzero integer n. Then p
is a covering map and so, S 1 is a covering space of itself via the map p.
Remark 4.1.1. We observe that for a covering map p : X e → X and each b ∈ X, the subspace
p−1 (b) of X
e is a discrete space. We explain it as follows: If U is an open neighbourhood
of b evenly covered by p then p−1 (U ) =
`
Vα and each Vα is homeomorphic to U by the
α
y ∈ p−1 (b). We intend to show that {y} is open in p−1 (b).
restriction of p on Vα . Choose any `
−1 −1
Since b ∈ U , p (b) ⊆ p (U ) = Vα and therefore, y ∈ Vα for some Vα . We claim that
α
there is no z ̸= y such that z ∈ Vα ∩ p−1 (b). For if such z exists, then p(y) = p(z) = b which
contradicts the fact that p|Vα : Vα → U is a homeomorphism. Hence Vα ∩ p−1 (b) = {y}.
Since Vα is open in X,e it follows that {y} is open in p−1 (b).
A. Deb Ray, C.U. 34

e → X is an open map.
Theorem 4.1.1. A covering map p : X

Proof. Let A be an open subset of X. e Take any x ∈ p(A). Then there exists y ∈ A such
that p(y) =`x. Since p is a covering map there exists an open neighourhood U of x such that
p−1 (U ) = Vα . Let y ∈ Vβ for some β ∈ Λ. Therefore, y ∈ A ∩ Vβ and A ∩ Vβ is open in
α∈Λ
e So, A ∩ Vβ is open in Vβ . p|V : Vβ → U being a homeomorphism, p(A ∩ Vβ ) is open in
X. β
U and hence in X. So, there exists an open set p(A ∩ Vβ ), such that x ∈ p(A ∩ Vβ ) ⊆ p(A).
Arbitrariness of x proves that p(A) is a neighbourhood of each of its points and hence it is
open in X.

Corollary 4.1.1. A covering map is a quotient map.

Proof. Follows from the fact that a continuous open surjection is a quotient map.
e → X and q : Ye → Y be covering maps. Then p × q : X
Theorem 4.1.2. Let p : X e × Ye →
X × Y given by
(p × q)(x, y) = (p(x), q(y))
is a covering map.

Proof. Let (b, c) ∈ X × Y . Then b ∈ X and c ∈ Y . By definition of covering maps,


there are open neighbbourhoods U and V of ` b and c respectively in the respective spaces,
such that p−1 (U ) = Sα and q −1 (V ) =
`
Tβ , where Sα ’s are open in Xe and Tβ ’s are
α β
open in Ye . Then U × V is
` an open neighbourhood of (b, c) in the product space X × Y
and (p × q)−1 (U × V ) = (Sα × Tβ ) where each p × q|Sα ×Tβ : Sα × Tβ −→ U × V is a
α,β
homeomorphism.

As a consequence of Theorem 4.1.2 we get the following example:

Example 4.1.5. The plane R2 is a covering space of a Torus S 1 × S 1 . In Example 4.1.2, we


have seen that exp : R → S 1 given by exp(t) = (cos 2πt, sin 2πt) is a covering map. Using
Theorem 4.1.2, exp × exp : R2 → S 1 × S 1 is a covering map.

However, restriction of a covering map need not be a covering map.

Example 4.1.6. Consider exp : R → S 1 given by exp(t) = (cos 2πt, sin 2πt) and take the
restriction exp |R+ : R+ → S 1 . Then exp |R+ : R+ → S 1 is a continuous surjection but not
a covering map because, (1, 0) has no evenly covered open neighbourhood by the restriction
map.

Though it is in general false, the next result shows one situation when restriction of a covering
map becomes a covering map.

Theorem 4.1.3. Let p : X e → X be a covering map. If Y is a subspace of X and if


−1
Ye = p (Y ), then the map p|Ye : Ye → Y is a covering map.
A. Deb Ray, C.U. 35

Proof. Let b ∈ Y . Since b ∈ X and `


p is a covering map, there exists an open neighbourhood
−1 e U ∩ Y is an open
U of b in X such that p (U ) = Vα , where each Vα is open in X.
α∈Λ
neighbourhood of b in Y such that
a
p|Y−1 −1
e (U ∩ Y ) = p (U ) ∩ Y =
e (Vα ∩ Ye ).
α∈Λ

Hence, the open neighbourhood U ∩ Y of b is evenly covered by p|Ye .


We recall that a function f : X → Y is called a local homeomorphism, if each point x ∈ X
has an open neighbourhood U such that f (U ) is open in Y and the map f |U : U → f (U ) is
a homeomorphism.
Theorem 4.1.4. A covering map p : X e → X is a local homeomorphism.
e ∈ X.
Proof. Let x e Since p is a covering map, p(e x) has an open neighbourhood U , evenly
−1
covered by p. Suppose Vα ’s are slices of p (U ) in X e ∈ Vβ . Then Vβ is the desired
e and let x
open neighbourhood of xe, which is homeomorphic to U by the restriction of p on Vβ .
But the converse of Theorem 4.1.4 is not in general true, follows from Example 4.1.6.
Example 4.1.7. The quotient map q : S 2 → RP 2 is a covering map. More generally, the
quotient map q : S n → RP n (n ≥ 2) is a covering map.
Solution : If α : S 2 → S 2 is given by z → −z, then α is a homeomorphism (check).
Now, the quotient map q : S 2 → RP 2 is given by z 7→ [z] where [z] = {z, −z}. Hence,
q(z) = q(−z) = [z]. So, if we choose any open nbd U of [z] in RP 2 such that d(x, y) < 2 for
any x, y ∈ U , then we get q −1 (U ) = U ∪ α(U ) with U ∩ α(U ) = ∅ and U is homeomorphic
to both U and α(U ) via the respective restrictions of q. Hence, each [z] has admissible open
nbd. i.e., S 2 is a covering space of RP 2 . (The proof is exactly same for any n ≥ 2).
Definition 4.1.2. Let p : E → X and q : F → X be two covering maps. The covering
spaces E and F of X are said to be equivalent if there exists a homeomorphism α : E → F
such that q ◦ α = p.
It can be proved that if p : E → X is a covering map such that E is simply connected then
for any covering map q : F → X with F simply connected, there exists a homeomorphism
α : E → F such that q ◦ α = p. In such case, p : E → X is called a universal covering of X.
So, exp : R → S 1 , exp : R2 → S 1 × S 1 , q : S 2 → RP 2 are all universal coverings.

4.2 Properties of Covering maps


In this section, we mainly discuss two most important properties of covering maps :
(i) Path Lifting property and
(ii) Homotopy Lifting property.
In a previous Chapter, we have established these properties for exp : R → S 1 . Now we see
that no special feature of exp is necessary for path lifing and homotopy lifting. The only
information we need is that exp is a covering map.

The following couple of results are required to establish the aforementioned properties.
A. Deb Ray, C.U. 36

Lemma 4.2.1. Let Y be any topological space and p : X e → X a covering map. If f, g :


Y → X are any two continuous functions such that p ◦ f = p ◦ g then the set S = {a ∈ Y :
e
f (a) = g(a)} is a clopen (closed as well as open) subset of Y .

Proof. To prove S is open in Y : Choose any x ∈ S. Then f (x) = g(x) in X. e Then


p(f (x)) = p(g(x)), i.e., (p ◦ f )(x) = (p ◦ g)(x). By definition of a covering map, there
exists a neighbourhood U of p(f (x)) such that p−1 (U ) =
`
Vα where each Vα is open in
α
Xe and p|Vα : Vα → U is a homeomorphism. Then f (x) = g(x) ∈ Vβ for some β. i.e.,
x ∈ f −1 (Vβ ) ∩ g −1 (Vβ ). Since f and g are continuous, f −1 (Vβ ) and g −1 (Vβ ) are open in Y
and consequently, we get an open neighbourhood f −1 (Vβ ) ∩ g −1 (Vβ ) of x in Y . We claim
that f −1 (Vβ ) ∩ g −1 (Vβ ) ⊆ S. If y ∈ f −1 (Vβ ) ∩ g −1 (Vβ ) then y ∈ f −1 (Vβ ) and y ∈ g −1 (Vβ ).
So, f (y), g(y) ∈ Vβ . Since p|Vβ is a homeomorphism and p(f (y)) = p(g(y)) holds, we get
f (y) = g(y). Therefore, y ∈ S. Hence, x ∈ f −1 (Vβ ) ∩ g −1 (Vβ ) ⊆ S, i.e., S is a neighbourhood
of each of its points.
To prove S is closed in Y : If possible, let b ∈ S \ S. Then f (b) ̸= g(b). But p(f (b)) =
p(g(b)). Let W be an admissible open neighbourhood of p(f (b)) and {Vα } form a partition
of p−1 (W ). We claim that ∃α ̸= β, such that f(b) ∈ Vα and g(b) ∈ Vβ .. If it is not true
then both f (b), g(b) ∈ Vγ with p(f (b)) = p(g(b)) implies that f (b) = g(b), because, p|Vγ is
one-one. This proves our claim.
So, b ∈ f −1 (Vα ) ∩ g −1 (Vβ ). By continuity of f and g, we get that f −1 (Vα ) ∩ g −1 (Vβ ) is
an open neighbourhood of b in Y . As b ∈ S \ S, (f −1 (Vα ) ∩ g −1 (Vβ )) ∩ S ̸= ∅. Choose
a ∈ (f −1 (Vα ) ∩ g −1 (Vβ )) ∩ S. Then f (a) ∈ Vα , g(a) ∈ Vβ and f (a) = g(a). Hence Vα ∩ Vβ ̸= ∅.
This contradiction proves S = S.

Lemma 4.2.2. Let p : X e → X be a covering map and Y a connected space. If g1 , g2 : Y → X


e
and f : Y → X are continuous functions such that p ◦ g1 = f = p ◦ g2 and there exists a
point y0 ∈ Y such that g1 (y0 ) = g2 (y0 ), then g1 = g2 .

Proof. Consider S = {a ∈ Y : g1 (a) = g2 (a)}. By Lemma 4.2.1, S is a clopen (i.e., closed


as well as open) set in Y . Since g1 (y0 ) = g2 (y0 ), for some y0 ∈ Y , we have y0 ∈ S. i.e., S ̸= ∅.
Since Y is connected and S is a nonempty clopen subset of Y , S = Y . Hence g1 = g2 .

Theorem 4.2.1. (Path Lifting Property) Let p : X e → X be a covering map and


α : I → X be a path beginning at some x0 ∈ X. Given any point x e0 ∈ Xe with p(e
x0 ) = x0 ,
there exists a unique path α e : I → X starting at x
e e0 such that p ◦ α
e = α.
(This αe is called a lift of α at xe0 )
i.e, the following diagram commutes:

@X
e
∃ unique α
e
p

I
α / X

Proof. Since α is a continuous function and I is connected, α(I) is a connected subset of X.


Choose an open cover {Vβ } of X by admissible open neighbourhoods of X. Then {α−1 (Vβ )}
is an open cover of I. Since I is compact, this open cover has a Lebesgue number ϵ. Choose
A. Deb Ray, C.U. 37

a partition 0 = t0 < t1 < . . . < tn = 1 of I such that |ti+1 − ti | ≤ ϵ, for all i = 0, 1, . . . , n − 1.


Then α[ti , ti+1 ] is contained in some admissible open neighbourhood Vi+1 , i = 0, 1, . . . n − 1.
−1
Let U1 be the path  component of p (V1 ) such that xe0 ∈ U1 . Define α e1 : [t0 , t1 ] → Xe by
−1
e1 (t) = p|U1 ◦ α (t). Then α
α e1 is a continuous function such that (p ◦ α e1 ) (t) = α(t) for all
e1 (0) = p|−1
t ∈ [t0 , t1 ] and α U1 ◦ α (0) = p|−1
U1 (x 0 ) = x
e 0 . So, α
e 1 is a lift of α|[t0 ,t1 ] . Assume that
α
fk is already defined on [t0 , tk ], for k > 1 such that α fk is a lift of α|[t0 ,tk ] . Then α([tk , tk+1 ]) ⊆
−1
Vk+1 . Choose a path component Uk+1 of p (Vk+1 ) such that Uk+1 contains  α
fk . Since
 p|Uk+1
−1
is a homeomorphism, we define an extension αg k+1 of α
fk by αg k+1 (t) = p|Uk+1 ◦ α (t) for all

t ∈ [tk , tk+1 ]. Then αg


k+1 is a path in X defined on [tk , tk+1 ] which lifts α|[t0 ,tk+1 ] . Taking
e
k = n − 1, we have a path α e in X
e defined on I which lifts the given path α in X such that
α
e(0) = x e0 .
The uniqueness of the lift follows from Lemma 4.2.2.
Remark 4.2.1. It is to be noted that even if α is a loop, α e may not be a loop. For example,
1
consider the covering map exp : R → S given by exp (t) = (cos 2πt, sin 2πt). For a loop
α(t) = (cos 2πt, sin 2πt), we get α
e(t) as line segment joining 0 and 1.
Theorem 4.2.2. (Homotopy Lifting Property) Let p : X e → X be a covering map and
Y be a topological space. If fe : Y → X e is a lift of f : Y → X (i.e., p ◦ fe = f ) and
H : Y × I → X is a continuous function with H(y, 0) = f (y), for all y ∈ Y , then there exists
e : Y × I → X,
a unique continuous function H e 0) = fe(y), for all y ∈ Y .
e such that H(y,

Proof. Left as self-reading exercise.


As a consequence of Homotopy lifting theorem, we get the following:
Theorem 4.2.3. (Monodromy Theorem) Let p : X e → X be a covering projection. Let
σ, τ : I → X be paths from x0 to x1 such that σ ∼ τ (i.e., σ is path homotopic to τ ). Let σ e
and τe be lifts of σ and τ respectively, both beginning at x e ∼ τe and σ
e0 . Then σ e(1) = τe(1).
Proof. Let H : σ ∼ τ be the path homotopy. By Theorem 4.2.2, there exists a lift H
e :
I ×I → X e of H, such that H(s,
e 0) = σ e:I →X
e(s). Consider ω e given by ω
e (t) = H(0,
e t).
Then ωe (0) = H(0,
e 0) = σ e0 and for all t ∈ I
e(0) = x
(p ◦ w)(t)
e = p(H(0,
e t)) = H(0, t) = σ(0) = x0 = Cx0 (t).

So, ωe lifts Cx0 and begins from x e0 . By uniqueness of the lift we have, ω e = Cxe0 . Similarly,
e : I → X given by γ
γ e e(t) = H(1, t) is a lift of Cx1 , beginning at x1 and so, by uniqueness of
e
the lift, γ
e = Cxe1 , where x
e1 = σe(1).
Define δ : I → X by δ(t) = H(t, 1). Then
e e e e

(p ◦ δ)(t)
e = p(H(t,
e 1)) = H(t, 1) = τ (t).

Also, δ(0)
e = H(0,
e 1) = w(1) e = Cxe0 (1) = x
e0 . Hence, δe is a lift of τ beginning at x e0 . By
uniqueness of the lift, δe = τe. Therefore, combining all, H e : I×I → X e is a continuous
function such that H(s, 0) = σ
e e(s), H(s, 1) = τe(s), H(0, t) = x
e e e0 and H(1, t) = x
e e1 . i.e., σ
e is
path homotopic to τe by H. e
Moreover, σe(1) = H(1,
e 0) = H(1, e 1) = τe(1).
A. Deb Ray, C.U. 38

Corollary 4.2.1. Let p : X e → X be a covering projection. Then p# : π(X, e xe0 ) → π(X, x0 )


is a monomorphism.
Proof. Let [ω] ∈ π(X, e0 ) be such that p# ([ω]) = [Cx0 ]. Then p ◦ ω ∼ Cx0 . By Theorem 4.2.3
e x
(Monodromy Theorem), p] ◦ω ∼ C g x0 , where p]◦ ω and C g x0 are the lifts of p ◦ ω and Cx0
respectively. But ω is a lift of p ◦ ω and Cxe0 is a lift of Cx0 both beginning at x e0 . Hence, by
uniqueness of lifts, ω = p]◦ ω and Cxe0 = Cx0 . i.e., [ω] = [Cxe0 ].
g

We conclude this section by proving a very important theorem, called Borsuk-Ulam Theorem
(for n = 2).
Definition 4.2.1. A point −x ∈ S n is said to be an antipodal point (or an antipode) of
x ∈ S n . A map h : S n → S m is said to be an antipode preserving map if h(−x) = −h(x)
for all x ∈ S n .
Example 4.2.1. Let Rϕ : S 1 → S 1 be a rotation by an angle ϕ (0 ≤ ϕ ≤ π). i.e.,
Rϕ (cos θ, sin θ) = (cos (θ + ϕ), sin (θ + ϕ)). Then Rϕ is antipode preserving. Because, for
any point z = (cos θ, sin θ), −z = (cos (π + θ), sin (π + θ)) and so,
Rϕ (cos (π + θ), sin (π + θ)) = (cos (π + θ + ϕ), sin (π + θ + ϕ))
= (− cos (θ + ϕ), − sin (θ + ϕ))
= −Rϕ (cos θ, sin θ)
The following are very useful observations about antipode preserving maps:
ˆ Composition of two antipode preserving maps is antipode preserving.
ˆ If antipodal maps g, h : S n → S n are such that h is null homotopic then g ◦ h is an
antipodal map which is also null homotopic.
Theorem 4.2.4. If h : S 1 → S 1 is a continuous and antipode preserving function then h is
not nullhomotopic.
Proof. Suppose b0 = (1, 0) ∈ S 1 and h(b0 ) = a0 ∈ S 1 . If b0 = a0 then we proceed with the
function h. If not, then choose a rotation Rϕ : S 1 → S 1 such that Rϕ (h(b0 )) = b0 . Then
consider the map k = Rϕ ◦h. In view of the previous observations, it follows that k : S 1 → S 1
is a continuous and antipode preserving map such that k(b0 ) = b0 . For proving that h is not
null homotopic, it is enough to show that k is not nullhomotopic.
Consider a map q : S 1 → S 1 given by q(cos θ, sin θ) = (cos 2θ, sin 2θ). Then q is a closed,
continuous surjection and hence a quotient map. The inverse image of any point w on S 1
under the map q consists of the antipodal points z and −z of S 1 . i.e., w = q(z) = q(−z).
Therefore,
q(k(−z)) = q(−k(z)) = q(k(z)).
Therefore, q ◦ k is constant on each fiber f −1 (w), w ∈ S 1 . So, by the property of quotient
maps, q ◦ k induces a continuous map f : S 1 → S 1 such that f ◦ q = q ◦ k.
S1 / S1
k
q q
 
S1 / S1
f
A. Deb Ray, C.U. 39

It is to be noted that b0 = q(b0 ) = q(k(b0 )) = f (q(b0 )) = f (b0 ). Also, k(−b0 ) = −b0 .


From the above commutative diagram we obtain a commutative diagram among the corre-
sponding fundamental groups and induced homomorphisms as follows:

Z∼
= π(S 1 , b0 ) / π(S 1 , b0 ) ∼
=Z
k#
q# q#
 
Z∼
= π(S 1 , b0 ) / π(S 1 , b0 ) ∼
=Z
f#

Also, q is a covering map:


S 1 is covered by four open sets U1 = S 1 ∩ {(x, y) ∈ R2 : y > 0}, U2 = S 1 ∩ {(x, y) ∈ R2 :
y < 0}, U3 = S 1 ∩ {(x, y) ∈ R2 : x > 0} and U4 = S 1 ∩ {(x, y) ∈ R2 : x < 0}. q −1 (U1 ) is the
disjoint union of the points on S 1 belonging to the first and third quadrant of R2 , each of
which is homeomorphic to U1 . Similarly, for U2 , U3 and U4 .
We observe that if α is any path from b0 to −b0 in S 1 , then the loop β = q ◦ α represents a
nontrivial element of π(S 1 , b0 ). For if β is trivial, i.e., β ∼ Cb0 , applying Monodromy theorem,
the lift of β (i.e., α) and lift of Cb0 (i.e., Cb0 ) are also equivalent such that α(1) = Cb0 (1) = b0
- a contradiction to the construction of α.
Now, f# ([β]) = [f ◦ β] = [f ◦ (q ◦ α)] = [q ◦ (k ◦ α)]. Since (k ◦ α) is a path in S 1 that starts
from b0 and ends at −b0 , [q ◦ (k ◦ α)] is non-trivial. Hence, f# is a nontrivial homomorphism
from an infinite cyclic group Z to itself. So, it is injective. Similarly, q# is injective so that
f# ◦ q# is injective. Using functorial property, q# ◦ k# = f# ◦ q# and therefore, k# is injective
and hence nontrivial. Therefore, k is not null homotopic.

Theorem 4.2.5. There is no continuous antipode preserving map g : S 2 → S 1 .

Proof. If possible let there be a continuous antipode preserving function g : S 2 → S 1 . Let


S 1 be the equator of S 2 . Then g|S 1 : S 1 → S 1 is a continuous antipode preserving map. By
Theorem 4.2.4, g|S 1 is not null homotopic. It is known that the upper closed hemisphere E
of S 2 is homeomorphic to the 2-disc D2 . Consider the continuous extension g|E : E → S 1
of g|S 1 . Since g|S 1 has a continuous extension to g|E , by a previous theorem, g|S 1 is null
homotopic. This contradiction proves that there is no continuous antipode preserving map
from S 2 → S 1 .

Theorem 4.2.6. (Borsuk Ulam Theorem) If f : S 2 → R2 is a continuous function then


there exists a point x of S 2 such that f (−x) = f (x).

Proof. Suppose for all x ∈ S 2 , f (x) ̸= f (−x). Define g : S 2 → S 1 by g(x) = ||ff (x)−f
(x)−f (−x)
(−x)||
.
2
Then g is a continuous function such that g(−x) = −g(x), for all x ∈ S . By Theorem 4.2.5,
it is not possible. Hence, there exists some x ∈ S 2 such that f (x) = f (−x).

Exercise 4.2.1. Find π(RP n ).

You might also like