The Little Big Book of Topology Notes

The Students of Math 147

Spring 2010

Compiled by Daniel Moore

 CHAPTER 1

Definitions and Examples

1. Introduction

What is geometry? Geometry is the study of rigid shapes that can be distinguished
with measurements (length, angle, area, . . . ).
What is topology? Topology is the study of shapes which are equivalent via deforma-
tions.
Topology versus Geometry: Objects that have the same topology do not necessarily
have the same geometry. For instance, a square and a triangle have different geometries
but the same topology.

Motivation: Our goal is to understand the shape of our universe. Consider the following
examples of two-dimensional universes: a plane, a sphere, a torus, and planes connected
by tubes.

These are topologically distinct universes. Intuitively, we can see that their “holes” distin-
guish them. Hence we seek a mathematical way to describe the holes; one familiar concept
we will use is continuity, which is related to a lack of holes.
3
4 1. DEFINITIONS AND EXAMPLES

2. Some Review from Analysis

Definition. Let f : R → R and a ∈ R. We say f is continuous at a if ∀ > 0 ∃ δ > 0

such that if |x − a| < δ, then |f (x) − f (a)| < .
Definition. Let M be a set and d : M × M → R be a function such that
(1) d(a, b) = 0 iff a = b (nondegeneracy)
(2) ∀a, b, c ∈ M , d(b, c) ≤ d(a, b) + d(a, c) (triangle inequality).
Then we say d is a metric or distance and (M, d) denotes a metric space.

Compare this to the usual definition of a metric space. The above definition is equivalent
to the usual definition of a metric space, but does not explicitly state the properties of
positivity or symmetry:
(1) d(a, b) ≥ 0 ∀a, b ∈ M (positivity)
(2) d(a, b) = d(b, a) ∀a, b ∈ M (symmetry).
Definition. The usual metric on Rn is
v
u n
uX
d((x1 , . . . , xn ), (y1 , . . . , yn )) = t (xi − yi )2 .
i=1

Definition. Let M be any set. Then the discrete metric is defined as


1 if a 6= b
d(a, b) =
0 if a = b

The discrete metric can be useful for testing conjectures as it does not rely on Rn .
Example (The Comb Metric for R2 .). Let X0 = {0} × [0, 1], Y0 = [0, 1] × {0}; and ∀n ∈ N,
let Xn = { n1 } × [0, 1]. Let M = (∪∞
n=0 Xn ) ∪ Y0 . The distance is the distance measured
along the comb in R2 .

...

0 1
 3. CONTINUITY, METRIC SPACES, AND OPEN BALLS 5

Using the comb metric,

• Does the sequence {( n1 , 0)} converge? Yes, to the origin.
• Does the sequence {( n1 , a)} converge when a ∈ (0, 1]? No, since d(( n1 , a), (0, a)) >
2a ∀n.
Example (A Non-Example of a Metric Space.). Consider the line of real numbers with
d(a, b) = a − b. Clearly, d(a, b) = 0 only when a = b, satisfying the first property (non-
degeneracy). However, the second property (triangle inequality) is not satisfied: if a = 0,
b = 1, and c = −1, then
2 = d(b, c)  d(a, b) + d(a, c) = −1 + 1 = 0.
Notice that this non-example also fails to satisfy the other two properties listed in the usual
definition of a metric, positivity and symmetry.
Example. Let M be a set and d : M × M → R a function which satisfies property 2 but
not property 1 in the definition of a metric space. Then
d(a, b) = 0 ∀a, b ∈ M
and M has at least 2 points.

3. Continuity, metric spaces, and open balls

We’d like to be able to talk about continuity in metric spaces (not just the reals).
Definition. Let (M1 , d1 ) and (M2 , d2 ) be metric spaces and a ∈ M1 . Let f : M1 → M2 .
We say f is continuous at a if ∀ > 0 there exists a δ > 0 such that ∀x ∈ M1 with
d1 (x, a) < δ then d2 (f (x), f (a)) < .

a
f(x)
δ
x ε
f(a)

Definition. Let (M,d) be a metric space and a ∈ M . The open ball of radius  > 0
about a is defined to be
B (a) = {x ∈ M | d(x, a) < }
6 1. DEFINITIONS AND EXAMPLES

We can redefine ‘continuous’ in terms of open balls:

Definition. A function f : X → Y is continuous at a ∈ X iff ∀ > 0 there exists a δ > 0
such that if x ∈ Bδ (a) then f (x) ∈ B (f (a)).
In other words, f is continuous at a iff for every  > 0 there is some δ > 0 such that
f (Bδ (a)) ⊆ B (f (a)).

f(Bδ)
δ
ε
Bδ

It should be noted that open balls don’t always look like open balls...
Example. Let M = upper half of plane with usual distance

-1 1

Because the dark-blue semicircle is the intersection of the open set

{(x, y) ∈ R2 | x2 + y 2 < 1}
with M , it is open in M .
Example. In R with the discrete metric, B1 (47) = 47 and B2 (47) = R.
 4. OPEN SETS 7

Example (Comb metric). Under the comb metric, B 1 ((0, 1)) is just an interval along the
2
y-axis starting at (0, 1).
B2 ((0, 1) is everything on the comb below the line of slope −1 that connects points (0, 1)
and (1, 0). Since it’s an open ball, the line is not contained in the ball.
B½((0,1))

...

B2((0,1))

0 1

- in R2 , B1 ((0, 0)) ∩ B1/2 ((1, 0)) and B1 ((0, 0)) ∪

S T
Note: Balls are not closed under and
B1/2 ((1, 0)) are not balls.

4. Open sets

Definition. Let (M, d) be a metric space. A set U ⊆ M is said to be open if ∀a ∈ U

there exists an  > 0 such that B (a) ⊆ U .
Example. Every subset of a set M under the discrete metric is open.

Small Fact. In a metric space, open balls are open sets.

Proof. Let a ∈ M and  > 0. We want to show that B (a) is open. That is, we want
to show that ∀x ∈ B (a) there exists an r > 0 such that Br (x) ⊆ B (a).
Let x ∈ B (a) and let r <  − d(x, a) and r > 0. We claim that Br (x) ⊆ B (a).
Let y ∈ Br (x). Then d(y, a) ≤ d(y, x) + d(x, a) <  − d(x, a) + d(x, a) = .
Therefore, Br (x) ⊆ B (a) 
8 1. DEFINITIONS AND EXAMPLES

Definition. Let
[
Ui = {x ∈ Ui | i ∈ I}
i∈I

Theorem 1 (the “open sets are nice” theorem). Let F be the family of open sets in a
metric space (M, d). Then:

(1) M , ∅ ∈ F

(2) If U, V ∈ F then U ∩ V ∈ F
S
(3) If ∀i ∈ I, Ui ∈ F then i∈I Ui ∈ F

Proof.

(1) Let x ∈ M , and  = 47. By definition,

B47 (x) = {y ∈ M | d(x, y) < 47} ⊆ M

Also, trivially ∀x ∈ ∅ there exists an  > 0 such that B (x) ⊆ ∅. Therefore, M

and ∅ are open, and so M, ∅ ∈ F .

(2) Let x ∈ U ∪ V , let r1 > 0 such that Br1 (x) ⊆ U , and let r2 > 0 such that
Br2 (x) ⊆ V . Let r = min{r1 , r2 }. Then clearly Br (x) ⊆ Br1 (x) ⊆ U and
Br (x) ⊆ Br2 (x) ⊆ V . Thus Br (x) ⊆ U ∩ V as desired.
S
Let x ∈ i∈I Ui . We want to show that there exists an r > 0 such that Br (x) ⊆
(3) S
i∈I Ui .
S
Because x ∈ i∈I Ui , there exists an i0 ∈ I such that x ∈ Ui0 . Then there must
exist an r > 0 such that Br (x) ⊆ Ui0 . Then
[
Br (x) ⊆ Ui0 ⊆ Ui
i∈I
S
and hence i∈I Ui ∈ F .

Definition. If f : X → Y and U ⊆ Y then f −1 (U ) = {x ∈ X | f (x) ∈ U }.

Theorem 2. Let (M1 , d1 ) and (M2 , d2 ) be metric spaces and f : M1 → M2 . Then f is

continuous if and only if for every U ⊆ M2 that is open in (M2 , d2 ), f −1 (U ) is open in
(M1 , d1 ).
 5. TOPOLOGIES 9

Proof.

(⇒) Suppose f is continuous. Let U ⊆ M2 be open. We want to show that ∀x ∈ f −1 (U )

there exists an r > 0 such that Br (x) ⊆ f −1 (U ).
Let x ∈ f −1 (U ). Since f (x) ∈ U and U is open, there exists  > 0 such that
B (f (x)) ⊆ U . Since f is continuous there exists an r > 0 such that
f (Br (x)) ⊆ B (f (x)) ⊆ U

Thus, f −1 (f (Br (x))) ⊆ f −1 (U ). Therefore Br (x) ⊆ f −1 (f (Br (x))) ⊆ f −1 (U ) as

desired.
(⇐) Suppose that for all open U ⊆ M2 , f −1 (U ) is open in M1 . Let p ∈ M1 . We want
to show that f is continuous at p.
Let  > 0. B (f (p)) is open, so f −1 (B (f (p))) is open. p ∈ f −1 (B (f (p))), so ∃
δ > 0 such that Bδ (p) ⊆ f −1 (B (f (p))).
f (Bδ (p)) ⊆ f (f −1 (B (f (p)))) = B (f (p)), so f is continuous at p, and thus every-
where.


We’ve proven a small fact:

Small Fact. If f : M1 → M2 , then f is continuous if for all p ∈ M1 , and for all  > 0
f −1 (B (f (p))) is open.

5. Topologies

We haven’t actually defined what a topology is yet. Let’s fix that.

Definition. Let X be a set and F be some collection of subsets of X such that
1) X, ∅ ∈ F .
2) If U, V ∈ F then U ∩ V ∈ F .
S
3) If for all i ∈ I, Ui ∈ F , then Ui ∈ F .
Then we say that (X, F ) is a topological space with open sets the elements of F . We
also say F is the topology on X.
Example. Let (M, d) be a metric space and F be the set of open sets in M . Then (M, F )
is a topological space.
Definition. Let M be a set and d be the discrete metric, then we say (M, F ) is the
discrete topology.
10 1. DEFINITIONS AND EXAMPLES

Definition. Let X be a set with at least 2 points. Let F = {X, ∅}. Then we say (X, F )
is the indiscrete, or concrete topology.
Definition. If F1 and F2 are topologies on X and F1 ⊆ F2 then we say that F1 is weaker
than F2 , or F2 is stronger than F1 .

Two notes:
• weaker = fewer = coarser and stronger = more = finer.
• The discrete topology is the strongest topology on M and the indiscrete topology
is the weakest topology on X.
Example. Let X = R and U ∈ F iff U is the union of sets of the form [a, b) such that
a, b ∈ R. (This is called the half-open interval topology).

Is this weaker, stronger, or neither compared to the usual topology?

If the interval (a, b) ∈ F , then F is stronger. Let a, b ∈ R, and a < b. Then,
[ 1
(a, b) = [a + , b),
n
n∈N

so F is stronger.
Example. Let X = R2 have the “dictionary order”. This means that (a, b) < (c, d) if
either a < c or a = c and b < d.

Note that U ∈ F iff U is a union of “open intervals”, i.e.

{(x, y) | (a, b) < (x, y) < (c, d)}.
• Is a vertical line open? Yes.
• Is a horizontal line open? No.
• Is this topology finer or coarser than the usual topology on R2 ?
Well, any point in a ball in the usual topology can be found in a ball of the
dictionary topology, which is contained in the usual ball. Open balls are open in
the dictionary topology, so the dictionary topology is finer than the usual topology.
S
Note: In R, n∈Z (n, n + 1) is open.

Question: Are the topologies on a set linearly ordered? No!

Example. Consider R with the usual topology and (R, F ) with F = {R, φ, 47}. We say
these two topologies are incomparable.
 5. TOPOLOGIES 11

Example. Consider (R, F ) Define U ∈ F if and only if either U = φ or U = R, or R − U is

finite. We see that if U is open in (R, F ), then U is open in the usual topology. Therefore
the usual topology is finer. This topology is called the finite complement topology on
R.
Definition. A set C in a topological space (X, F ) is closed iff its complement X \ C is
open.

Note that a set can be both open and closed as well as neither open nor closed. (It is not
a door!)
S
Example. Let us consider R with the half-open interval topology. ThenS[0, ∞) = n∈N [0, n)
is open. On the other hand, the complement of this set is (−∞, 0) = n∈N [−n, 0), which
is also open. Thus this is a clopen set.
Example. A “vertical line” in R2 with the dictionary order is both open and closed. The
proof is left as exercise.
Lemma. Let (X, F ) be a topological space and A be the set of all closed sets in X. Then:
(1) X, φ ∈ A
T
(2) If C, D ∈ A, then C D∈A
T
(3) If Ci ∈ A for every i ∈ I, then i∈I Ci ∈ A

The proof is basically the same as that for open sets back in section 4.
Definition. Let (X, F ) be a topological space and A ⊆ X. Let {Uj | j ∈ J} be the set of
◦ S ◦
all open sets contained in A. Then we define A = Int(A) = j∈J Uj , and we say A is the
interior of A.
Small Fact. Let (X, F ) be a topological space and A ⊆ X. Then
◦
(1) A ⊆ A
◦
(2) A is open
◦
(3) If U ⊆ A is open, then U ⊆ A
◦
(4) A is open iff A = A.

◦ S ◦
Proof. (1) Since A = j∈J Uj and Uj ⊆ A for every j ∈ J, A ⊆ A.
◦ ◦
(2) By definition, A is a union of open sets, so A is open.
12 1. DEFINITIONS AND EXAMPLES

S ◦
(3) Since U is open in A, U ∈ {Uj | j ∈ J}. Therefore U ⊆ j∈J Uj = A.
◦
Note that this means that A is the “largest” open set in A.
◦ ◦
(4) Suppose that A = A. Then A is open by part (2), hence A is open.
◦
Conversely, suppose that A is open. Then A ∈ {Uj | j ∈ J}. Hence A ⊆ A. From
◦
(1), A = A.

Example. In the half-open interval topology on R, Int((0, 1]) = (0, 1). To prove this,
assume that 1 ∈ Int((0, 1]) and show that it leads to a contradiction.
Example. In the finite-complement topology on R, Int((0, 1]) = ∅. This follows from the
fact that R is infinite and all subsets of (0, 1] are finite.
Example. In the dictionary-order topology on R2 , Int([0, 1] × [0, 1]) = [0, 1] × (0, 1).

1 1

R Int(R)

1 1

Definition. Let A be a subset of a topological space (X, F ), and let {Fj | j ∈ J} be the set
T
of all closed sets containing A. Then the closure of A is defined as A = cl(A) = j∈J Fj .
Small Fact. Small facts about closures:
(1) A ⊆ A
(2) A is closed
(3) If A ⊆ C and C is closed, then A ⊆ C
(4) A = A if and only if A is closed.

The proofs are left as exercises.

1
1
Example. In R with the usual topology, cl n |n ∈N = n |n ∈ N ∪ {0}
Example. In R with the half-open interval topology, cl((0, 1]) = [0, 1]
 6. CONTINUITY IN TOPOLOGICAL SPACES 13

Lemma (an Important Lemma). Let (X, F ) be a topological space and Y ⊆ X. Then p ∈ Y
if and only if for every open set U ⊆ X containing p, U ∩ Y 6= ∅.

Proof. Let p ∈ Y and U ⊆ X be open with p ∈ U . Suppose U ∩ Y = ∅ and let

C = X \ U . Then p ∈/ C because p ∈ U . It follows that Y ⊆ C because Y ⊆ C and C is
closed. We obtain that p ∈ C, which is a contradiction.
Conversely, suppose that for every open set U ⊆ X such that p ∈ U , U ∩ Y 6= ∅. Let
C ⊆ X be closed such that Y ⊆ C. Suppose p ∈ / C. Let U = X \ C, which is open with
p ∈ U . Now U ∩ Y 6= ∅, so there exists some x ∈ U ∩ Y . This means that x ∈ X \ C, i.e.
x∈/ C. But since Y ⊆ C, we have x ∈ / Y , a contradiction. Therefore we conclude that
p∈Y. 
Corollary.
T T that U is an open set in a topological space (X, F ) and Y ⊆ X. If
Suppose
U Y 6= ∅, then U Y 6= ∅.

Proof. The proof is immediate from the important lemma.

Now we can use the important lemma to prove that in R in the usual topology, cl({1/n | n ∈
T =1 {1/n | n ∈ N} ∪ {0}. Here, we can show that for all open sets U containing 0,
N})
U n | n ∈ N 6= ∅.

6. Continuity in Topological Spaces

Definition. Let (X1 , F1 ) and (X2 , F2 ) be topological spaces and f : X1 → X2 . We say f

is continuous if and only if for every U ∈ F2 , f −1 (U ) ∈ F1 .
Small Fact. Let (X1 , F1 ) and (X2 , F2 ) and (X3 , F3 ) be topological spaces, f : X1 → X2
and f : X2 → X3 be continuous functions. Then g ◦ f : X1 → X3 is continuous.

Proof. Let U ∈ F3 . Then g −1 (U ) ∈ F2 because g is continuous, and f −1 (g −1 (U )) ∈ F1

because f is continuous. Therefore (g ◦ f )−1 (U ) ∈ F1 . Thus g ◦ f is continuous. 
Theorem 3. Let X, Y be topological spaces and f : X → Y . Then f is continuous if and
only if for every closed set C in Y, f −1 (C) is closed in X.

Proof. (⇒) Suppose C is closed in Y . Then Y \ C is open, implying that f −1 (Y \

C) is open in X. Since f −1 (Y \C) = {x ∈ X | f (x) ∈ Y \C} = {x ∈ X | f (x) ∈
/ C}
and X \ f −1 (C) = {x ∈ X | f (x) ∈/ C}, we have the desired results.
(⇐) This is exactly like the analogous proof for open sets.


This is nice, because sometimes it’s easier to work with closed sets than with open sets.
Definition. Let X and Y be topological spaces, and let f : X → Y .
14 1. DEFINITIONS AND EXAMPLES

(1) If for every open set U ⊆ X, f (U ) is open in Y , then f is open.

(2) If for every closed set U ⊆ X, f (U ) is closed in Y , then f is closed.

This is sort of like continuity, except that we care about the image of sets instead of their
preimages.
Example time!
Example. Let F be the half-open topology on R, and define a function
f : (R, F ) → (R, usual) by f (x) = x
• Is f continuous? Yes! An open
set in (R, usual) is a union of intervals of the form
(a, b). We know that f −1 (a, b) = (a, b), which is open in F .
• Is f open? No. Take any U = [a, b) ∈ F . Then f (U ) = [a, b), which is not open
in (R, usual).


• Is f closed? Also no, since f [a, b) = [a, b) is not closed in (R, usual).

7. Homeomorphisms

Let’s define a notion of equivalency for topological spaces.

Definition. Let (X1 , F1 ) and (X2 , F2 ) be topological spaces, and let f : X1 → X2 be
continuous, bijective, and open. Then f is a homeomorphism, and X1 and X2 are
homeomorphic (denoted by X1 ∼ = X2 ).
Example. Let I = [0, 1] and X = I × I ⊂ R2 , under the usual metric for R2 . Let
Y = {(x, y) ∈ R2 |x2 + y 2 ≤ 1} = D2 be the unit disk in R2 . Question: Are X and Y
homeomorphic?
Answer: Yes! Let’s see how.

Define the centers of X and Y to be x and y, respectively. Fix some point a on the
boundary of X, and some b on the boundary of Y (that is, a ∈ ∂X and b ∈ ∂Y ).
Define a function f as follows:
• f (x) = y
• f (a) = b
• For t ∈ ∂X, look at the distance along the boundary from a to t. Then f (t) is a
point proportionally far along the boundary of Y .
• For s ∈ int(X), draw the ray connecting x and s. Let t be the point at which
this ray intersects ∂X. Now, in Y , draw the ray connecting y and f (s). Then s
is mapped to a point on this ray that is proportionally far from y.
 7. HOMEOMORPHISMS 15

The last two, in pictures:

f(t)
t

a
x y

b
X Y

t f(t)
s
a f(s)

x y

b
X Y

• Is this well-defined?
– Yes! There is always exactly one point in Y that can be mapped to be a
point in X 1.
• Is this a bijection?
– Yes! The inverse is defined identically, so it would make sense for this to be
a bijection. Also, consider the images of concentric squares centered on x
under f : they are mapped to disjoint concentric circles centered on y.
• Is f continuous and open?
1This works because both a square and a circle are convex shapes - for example, for all points p, q ∈ X,
the line pq that connects p and q lies entirely within X. This also implies that x and y didn’t actually have
to be the exact centers of X and Y respectively.
16 1. DEFINITIONS AND EXAMPLES

– Yes! Intuitively, it’s easy to see that an open set in X is mapped to an open
set in Y , and that the preimage of an open set in Y is open.
Question: Can we extend this to (some) non-convex regions?
Answer: Sure. Just divide up the non-convex region into smaller regions. Actually, a
subregion will work as long as there is some point in its interior such that any ray from
that point intersects the subregion’s boundary exactly once.
However, there are limits to this: for example, a donut is not homeomorphic to a circle.
This is hard to show; we’ll see a proof later.
There are also some homeomorphisms that we might find unsettling. For example, a
knot in R3 and the unit circle S 1 = {(x, y) ∈ R2 |x2 + y 2 = 1} are homeomorphic; the
argument works exactly like the one used for ∂X when showing that a square and circle
are homeomorphic (above). It turns out that, while the knot and circle are homeomorphic,
their complements are not.
Example. Define a function f : [0, 1) → S 1 by


f (t) = cos(2πt), sin(2πt)

This takes the interval and wraps it counterclockwise around the circle.
• Is this a bijection?
– Yes! The interval wraps once around the circle; one end is open, so there is
no overlap.
• Is f continuous?
– Yes! First we define an open ball in S 1 as B,S 1 (a) = {x ∈ S 1 | d(x, a) < }.
Then the preimage of every open ball in S 1 is an open interval, so f is
continuous.
• Is f open?
– No. Let U = [0, 1/2), which is open in [0, 1). Sadly, its image is not open in
S1.
So this f is not a homeomorphism.

We will show that R2 6∼

= R3 near the end of Chapter 3. This is difficult.
Example. Some non-examples of homeomorphisms:
• Q under the usual topology is not homeomorphic to R under the usual topology,
since there is no bijection between Q and R.
 8. SUBSPACES 17

• R with the finite-complement topology (F CT ) is not homeomorphic to R with

the usual topology:
– Suppose that there is some homeomorphism f : (R, F CT ) → (R, usual). Let
U = (0, 1).
– Look at f −1 (U ) - it must be open, since f is a homeomorphism.
– It is definitely not equal to ∅, since U 6= ∅.
– It is also not R, since f is a bijection.
– So R \ f −1 (U ) must be finite. But R \ U is not finite.
– Because f is a bijection, this is a contradiction.
– Therefore, (R, F CT ) ∼
6 (R, usual).
=

8. Subspaces

Definition. Let (X, FX ) be a topological space, and let Y ⊆ X. Let FY = {U ∩ Y |U ∈

FX }. Then FY is the subspace topology or induced topology on Y .
Example. Consider the topological space R2 with the dictionary order.

Q: What are the open sets in the subspace Z × Z?

A: All sets are open. We can draw a small open interval about any point, so the set of any
one point is open, and any set is the union of such sets. Because of all sets are open, we
can consider this topology the discrete topology on Z × Z.
Example. Consider Z × Z as a subspace of R2 , but this time with the usual topology.
Again, all sets are open, as we can construct an open ball of radius 21 about any point that
doesn’t intersect any others in the same way that we can construct an open interval.

Small facts about subspaces:

Let (X, FX ) and (Y, FY ) be topological subspaces with (S, FS ) a subspace of X and (T, FT )
a subspace of Y . Then
(1) If S ∈ FX , then FS ⊆ FX .
(2) C ≤ S iff ∃ a closed set A in X such that C = A ∩ S.
(3) Suppose f : X → Y is continuous. Then f | S : S → Y is continuous.
(4) Suppose f : X → Y is continuous and f (X) ⊆ T . Let g : X → T be such that
g(x) = f (x) for every x ∈ X. Then g is continuous.

Proof. (1) If S ∈ FX , then S is open in X. By definition FS = {U ∩S : U ∈ FX }

18 1. DEFINITIONS AND EXAMPLES

(2) (⇒) Let C ⊆ S be closed. Because S \ C ∈ FS , there is some U ∈ FS such that

U ∩ S = S \ C. Since U is open in X, we know that X \ U is closed in X.
We claim that the closed set we desire is X \ U . Note that
X \ (U ∩ S) = (X ∩ S) \ (U ∩ S) = S \ (U ∩ S) = S \ (S \ C) = C
and so we are done
(⇐) Suppose that there is a closed set A ⊆ X such that C = A ∩ S. Then
X \ A ∈ FX , so (X \ A) ∩ S ∈ FS . But X \ (A ∩ S) = (X ∩ S) \ (A ∩ S) = S \ C
is open in SS, so C is closed in S.
(3) Let U ∈ FY . Since f is continuous, f −1 (U ) ∈ FX . Because (f |S)−1 (U ) =
f −1 (U ) ∩ S, (f |S)−1 (U ) is the intersection of an open set with S and so is open.
Therefore, f |S is continuous.
(4) Let U ∈ FT . Then there is some V ∈ FY such that U = V ∩ T . Because
f : X → Y , f −1 (V ) ∈ FX . Since f (X) ⊆ T ,
f −1 (U ) = f −1 (V ) ∩ T = f −1 (V ∩ T ) = f −1 (V )
Therefore, f −1 (U ) is open in X. But f −1 (U ) = g −1 (U ), so g −1 (U ) is open and so
g is continuous.


9. Bases

You may have noticed that, in metric spaces, the idea of open sets was quite useful. We’d
like to extend this idea to more general topological spaces.
Definition. Let (X, F ) be a topological space and β ⊆ F such that for every U ∈ F , U
is a union of elements in β. Then we say that β is a basis for F .

Note that this basis is not necessarily minimal (like the basis of a vector space), but it is
more useful the more specific it is.

Example. R with the half-open interval topology was defined by the basis [a, b) | a < b .
Example. R2 with the dictionary topology was similarly defined by a basis of open inter-
vals.
Theorem 4. Let X be a set and β be a collection of subsets such that
S
(1) X = B∈β B
(2) ∀B1 , B2 , if x ∈ B1 ∩ B2 , then there is some B3 ∈ β such that x ∈ B3 ⊆ B1 ∩ B2 .
Let F = collection of elements of β. Then F is a topology on X with basis β.

The proof is left as an exercise.

 9. BASES 19

Theorem 5. Let X be a set and β a collection of subsets of X such that

(1) X = ∪B∈β B
(2) For all B1 , B2 ∈ β and x ∈ B1 ∩ B2 , ∃ B3 ∈ β such that x ∈ B3 ⊆ B1 ∩ B2 .
Let F = {unions of elements of β}. Then F is a topology for X with basis β.

Proof. First we prove that F is a topology for X.

(1) Since X = ∪B∈β B, by definition X ∈ F . Since ∅ is the union of zero elements of
the β, we also have ∅ ∈ F .
(2) Let U, V ∈ F . Hence we know that there exist index sets I and J such that
U = ∪i∈I Bi and V = ∪j∈J Bj . Consider U ∩ V = (∪i∈I Bi ) ∩ (∪j∈J Bj ). Let
x ∈ U ∩ V . Then there exists ix ∈ I and jx ∈ J such that x ∈ Bix ∩ Bjx . From our
second assumption we know there exists a Bx ∈ β such that x ∈ Bx ⊆ Bix ∩ Bjx .
Let W = ∪x∈U ∩V Bx . Since W is a union of elements of β it is clearly in F .
WTS: W = U ∩ V
(⊇) For all x ∈ U ∩ V , we know that x ∈ Bx ⊆ ∪x∈U ∩V Bx = W . Therefore
U ∩ V ⊆ W.
(⊆) For all x ∈ U ∩ V , we have a Bx ⊆ Bix ∩ Bjx ⊆ U ∩ V . Therefore, W =
∪x∈U ∪V Bx ⊆ U ∩ V .
We have containment in both directions, so W = U ∩ V .
(3) Suppose ∀k ∈ K, Uk ∈ F .
WTS: ∪k∈K Uk ∈ F .
For all k ∈ K, Uk = ∪i∈Ik Bi . Hence ∪k∈K Uk = ∪k∈K (∪i∈Ik Bi ) ∈ F , since it is a
union of elements of β.
Therefore F is a topology. By definition, β is also a basis of F . 
Small Fact. Let (X, FX ) and (Y, FY ) be topological spaces with bases of βX and βY
respectively, and f : X → Y .
(1) f is continuous iff ∀ B ∈ βY , f −1 (B) ∈ FX .
(2) f is open iff ∀ B ∈ βX , f (B) ∈ FY .

Proof. (of (1) only. (2) is virtually identical.)

(⇒) Suppose f is continuous. Then ∀ U ∈ FY , f −1 (U ) ∈ FX by the definition of continuity.
In particular, if B ∈ βy , then B ∈ FY and f −1 (B) ∈ FX .
(⇐) Suppose B ∈ βY implies f −1 (B) ∈ FX . Let U ∈ FY . Hence U = ∪i∈I Bi for some
index set I. Therefore,
f −1 (U ) = f −1 (∪i∈I Bi ) = ∪i∈I f −1 (Bi ) ∈ FX ,
since we know f −1 (Bi ) ∈ FX for all i and unions of elements of FX are in FX . 
 CHAPTER 2

Making new spaces from old

1. Quotient Spaces

First we will consider quotients of sets.

Definition. Let X be a set and ∼ a relation on X. We say ∼ is an equivalence relation
if
(1) ∀ x ∈ X, x ∼ x (reflexivity)
(2) ∀ x, y ∈ X if x ∼ y then y ∼ x (symmetry)
(3) If x, y, z ∈ X and x ∼ y and y ∼ z, then x ∼ z (transitivity)
Definition. Let X be a set with an equivalence relation ∼. Then ∀ a ∈ X define the
equivalence class of a as
[a] = {x ∈ X : x ∼ a}.
Definition. Let X be a set with an equivalence relation ∼. Then the quotient of X by
∼ is
X/ ∼= {[p] : p ∈ X}.

NOTE: The equivalence classes partition X (i.e. ∀ x ∈ X, x ∈ exactly one equivalence

class)
Example. Similarity (i.e. A ∼ B iff there exists a P such that A = P −1 BP ) is an
equivalence relationship on the set of n × n matrices. However, this is not the “kind” of
equivalence relationships that we will be studying.
Example. Let X = [0, 1] and x ∼ y iff x = y or x, y ∈ {0, 1}. This “glues” the interval
[0, 1] into a circle.
Definition. Let X be a set and ∼ an equivalence relation. We define
π : X → X/ ∼
such that π(x) = [x]∀ x ∈ X. We say π is the projection map.
Tiny Fact. Let X be a set with equivalence relation ∼. Then
(1) π is onto
21
22 2. MAKING NEW SPACES FROM OLD

(2) π is one-to-one iff “∼” is “=”.

Proof. (1) Let [x] ∈ X/ ∼. Then π(x) = [x]. Therefore π is onto.

(2) (⇒) Suppose π is one-to-one. Let x, y ∈ X and x ∼ y. Then [x] = [y] and
π(x) = π(y), implying x = y since π is one-to-one.
(⇐) Suppose “∼” is “=”. Let x, y ∈ X such that π(x) = π(y). Then {x} = [x] =
[y] = {y}, and x = y.

Now we would like to define a topology such that π is continuous.

Definition. Let (X, FX ) be a topological space and ∼ be an equivalence relation on X.

We define
F∼ = {U ⊆ X/ ∼: π −1 (U ) ∈ FX }

and call (X/ ∼, F∼ ) the quotient space of X with respect to ∼.

Small Fact. Let (X, FX ) be a topological space and ∼ be an equivalence relation on X.

(X/ ∼, F∼ ) is a topological space and π is continuous.

Proof. First let us prove that F∼ is a topology on X/ ∼.

(1) Since π −1 (X/ ∼) = X ∈ FX , clearly (X/ ∼) ∈ F∼ . Since π −1 (∅) = ∅ ∈ FX ,

∅ ∈ F∼ .

(2) Let U, V ∈ F∼ . Now

π −1 (U ∩ V ) = π −1 (U ) ∩ π −1 (V ) ∈ FX

since π −1 (U ) ∈ FX and π −1 (V ) ∈ FX . Therefore U ∩ V ∈ F∼ .

(3) Consider π −1 (∪i∈I Ui ) = ∪i∈I π −1 (Ui ).

Since π −1 (Ui ) ∈ F∼ ∀i, the arbitrary union of such sets must also be open. Thus
by the above equality, π −1 (∪i∈I Ui ) ∈ F∼ , completing the proof.

Note that the continuity of π follows directly from the quotient topology. 

Question: Is π necessarily open? Answer: No

 1. QUOTIENT SPACES 23

Consider the interval [0,1] mapped to a circle under π. Then the interval [0,1/2), which is
open in [0,1] is mapped to a half circle which is not open in the whole circle.
π

Example. Let X = I × I, where I is the unit interval. Define an equivalence relation on X

as follows: (x, y) ∼ (x0 , y 0 ) if and only if either (x, y) = (x0 , y 0 ) or x = x0 and y, y 0 ∈ {0, 1}.
This equivalence “glues together” the top and bottom edges of the unit square. This
basically rolls up the unit square, so topologically X/ ∼ gives us a cylinder.
(0,1) (1,1)

(1,0)~(1,1)

(0,0)~(0,1)

(0,0) (1,0)
X X/~
Example. Let X = R2 and suppose (x, y) ∼ (x0 , y 0 ) if and only if ∃n, m ∈ Z such that
x = x0 + n, y = y 0 + m.

Note that X may be divided into integer side length squares, such that under ∼ all of the
given squares are equivalent. Thus we need only consider one such square, noting that the
opposite sides are equivalent, yielding a torus. This is pretty much the same as the above
example, except that we roll up the cylinder too. In the picture, we are identifying all lines
of the same color.

We may also generalize this idea to higher dimensions, yielding the analogous torus for
that dimension (i.e. X = R3 yields a 3-torus and so on).
24 2. MAKING NEW SPACES FROM OLD

Example. Let S n = {x ∈ Rn+1 | kxk = 1}. Define x ∼ y ⇔ x = ±y.

Note that for n = 1 we obtain the unit circle such that points connected by a diameter are
equivalent. Thus any semi-circle forms a fundamental domain. Such a semicircle has it’s
endpoints as equivalent, and is thus topologically equivalent to the original circle.
We denote the quotient space S n / ∼ by RPn , or the real projective space. The above
argument concludes that RP1 ∼
= S1.
Example. From the above, RP2 = S 2 / ∼. We simply note that the fundamental domain is
a hemisphere whose boundary takes on the same topology as RP1 since it is a circle under
∼.
0 0
Example. Let X = R2 and define (x, y) ∼ (x0 , y 0 ) ⇔ x2 + y 2 = x 2 + y 2 .
Under ∼, any points on a circle centered at the origin are equivalent. Collapsing all such
circles to points, we find that X/ ∼ is a ray emanating from the origin (it’s pink in the
picture).

2
2
/~

Definition. Let X, Y be sets and f : X → Y be a function. We define the relation ∼

induced by f as follows:

∀p, q ∈ X, p ∼ q ⇔ f (p) = f (q)

It is clear that ∼ is an equivalence relation on X because = is an equivalence relation on
Y.
Definition. Suppose (X, FX ) is a topological space and Y is a set. Let f : X → Y be
onto. Then the quotient topology on Y with respect to f is given by:

Ff = {U ⊆ Y |f −1 (U ) ∈ FX }

If Y has this topology we say that f is a quotient map.

 1. QUOTIENT SPACES 25

Tiny Fact. (1) (Y, Ff ) is a topological space.

(2) f : (X, FX ) → (Y, Ff ) is continuous.

Proof. Left as an exercise.

Example. Let f map the closed segment [0, 1] to a figure-eight, where f (0) = f (1) = f ( 12 )
at the intersection of the two sides of the figure-eight. What is open in the quotient topology
(Y, Ff )?

• Any open-looking interval on the eight that does not contain the intersection point
is open since its preimage is clearly open on the segment; for the same reason, any
appropriate union or intersection of these open intervals is also open.

• Furthermore, any open set in the figure-eight that includes the point at the cross
must also contain an open interval of nonzero length extending along each of the
four legs of the figure-eight. This is necessary because the definition of Ff requires
that the preimage of anything open must be open itself; hence the only way for the
preimage of a set containing the intersection point to be open is if the preimage
is an open set in [0, 1] containing the three preimages of the intersection point (0,
1
2 , and 1).

-1
f

f(0) = f(1/2) = f(1)

0 ½ 1

Theorem 6. Let (X, FX ) be a topological space, f : X  Y onto, and ∼ induced by f .

Then (Y, Ff ) ∼
= (X/ ∼, F∼ ).

If a diagram commutes, then the path taken does not affect the result. Note that the arrow
representing g is dashed because we have not yet shown that the diagram commutes. In
the proof, we will define g so that the diagram does commute.
26 2. MAKING NEW SPACES FROM OLD

f
X5 /Y
55 D
55
55
55π g
55
55
5
X/ ∼

Proof. Define g : X/ ∼→ Y by g([π]) = f (x) where x is a representative of equivalence

class [π]. We want to show that f is well-defined, one-to-one, onto, continuous, and open.
• Well-defined: WTS if [x] = [y], then g([x]) = g([y]). That is, any representative of
a given equivalence class maps to the same value. Suppose [x] = [y]. Then x ∼ y;
since ∼ is the relation induced by f , f (x) = f (y). ∴ g([x]) = g([y]).
• One-to-one: Suppose g([x]) = g([y]). Then f (x) = f (y) ⇒ x ∼ y ⇒ [x] = [y].
• Onto: Suppose y ∈ Y . f is onto, so ∃x ∈ X such that f (x) = y. So g([x]) =
f (x) = y.
• Continuous: WTS U ∈ Ff ⇒ g −1 (U ) ∈ F∼ . Suppose U ∈ Ff . Recalling that
F∼ = {O ⊆ X/ ∼ st. π −1 (O) ∈ FX }, we equivalently WTS that π −1 (g −1 (U )) ∈
FX . Since U ∈ Ff ⇔ f −1 (U ) ∈ FX , we WTS f −1 (U ) = π −1 (g −1 (U )).
(⊆) Let x ∈ f −1 (U ). Hence f (x) = g([x]) ∈ U . Since π −1 (g −1 (U )) = {p st. g ◦
π(p) ∈ U }, g ◦ π(x) = g([x]) ∈ U ⇒ x ∈ π −1 (g −1 (U )).
(⊇) Let x ∈ π −1 (g −1 (U )). Hence g(π(x)) ∈ U ⇒ g([x]) ∈ U ⇒ f (x) ∈ U . So
x ∈ f −1 (U ).
∴ f −1 (U ) = π −1 (g −1 (U )). Since f −1 (U ) ∈ FX , π −1 (g −1 (U )) ∈ FX . Thus
g −1 (U ) ∈ F∼ .
• Open: WTS U ∈ F∼ ⇒ g(U ) ∈ Ff . Suppose U ∈ F∼ . Recalling that Ff = {O ⊆
Y st. f −1 (O) ∈ FX } and that U ∈ F∼ ⇔ π −1 (U ) ∈ FX , we equivalently WTS
f −1 (g(U )) = π −1 (U ).
(⊆) Let x ∈ f −1 (g(U )). So f (x) = g([x]) ∈ g(U ). Since g is bijective (from the
earlier parts of this proof), [x] ∈ U ⇒ π(x) ∈ U ⇒ x ∈ π −1 (U ).
(⊇) Let x ∈ π −1 (U ). So π(x) ∈ U , implying that g(π(x)) ∈ g(U ). Then g(π(x)) =
g([x]) = f (x) ⇒ x ∈ f −1 (g(U )).
Therefore g is a homeomorphism, and (Y, Ff ) ∼
= (X/ ∼, F∼ ). 
Theorem 7. Let (X, FX ) be a topological space and ∼ an equivalence relation. Let Y =
X/ ∼ and let f : X → Y be π. Then F∼ = Ff and f is a quotient map.
 1. QUOTIENT SPACES 27

Proof. Recall that:

F∼ = {U ⊆ X/ ∼ st. π −1 (U ) ∈ Fx }
Ff = {U ⊆ Y st. f −1 (U ) ∈ Fx }.
We know that X/ ∼= Y and f = π. ∴ F∼ = Ff . Also, f = π is onto since it is a projection,
and by definition F∼ = Ff is a quotient topology. ∴ f is a quotient map. 
Lemma. (Important Lemma about Quotients)
Let (X, FX ), (Z, FZ ) be topological spaces and let Y be a set. Suppose f : X  Y is onto
and let g : (Y, Ff ) → (Z, FZ ). Then g is continuous if and only if g ◦ f is continuous.

Proof. (⇒) : Suppose that g is continuous. Since f is a quotient map, it is

continuous. Therefore, g ◦ f is a composition of continuous functions, and so g ◦ f
is continuous.
(⇐) : Suppose, now, that g ◦ f is continuous. Let U ∈ FZ . Then (g ◦ f )−1 (U ) ∈ FX .
Therefore, (g ◦ f )−1 (U ) = f −1 (g −1 (U )) ∈ FX .
Recall that Ff = {V ⊆ Y : f −1 (V ) ∈ FX }. So, g −1 (U ) ∈ Ff , and g is continuous.

Example (A Non-Example). Let f : R − {1} → R by
(
x2 x ∈ (−∞, 1)
f (x) = −1
x−1 x ∈ (1, ∞)
f is continuous. Let g : R → R by
(
1 x≥0
g(x) =
−1 x<0
g is not continuous. But g ◦ f : R − {1} → R is defined by
(
−1 x > 1
(g ◦ f )(x) =
1 x<1
g ◦ f is continuous.

Question: Does this contradict the Important Lemma?

Answer: No! The solution is, of course, that g actually is continuous - the topology of its
domain isn’t the usual topology. Instead, it has the topology induced by f . That is,
Ff = {U ⊆ R : f −1 (U ) open in R \ {1}}
More precisely, we have g −1 ({1}) = [0, ∞) and f −1 ([0, ∞)) = (−∞, 1), which is open.
Also, g −1 ({−1}) = (−∞, 0) and f −1 ((−∞, 0)) = (1, ∞), which again is open.
Therefore, g is continuous (as we only need to look at the range of g, which is {−1, 1}).
28 2. MAKING NEW SPACES FROM OLD

Theorem 8. Let (A, FA ) and (B, FB ) be topological spaces and f : A → B be a homeo-

morphism. Let ∼A and ∼B be equivalence relations on A and B, respectively, such that
x ∼A x0 if and only if f (x) ∼B f (x0 ).
Then A/ ∼A ≡ B/ ∼B .

Proof. We want to show that there is some function g such that the following diagram
commutes:
f
A /B

πA πB

 
C _ _ _ g _ _ _/ D

Define g : A/ ∼A → B/ ∼B by g ([x]A ) = [f (x)]B . To show it’s a homeomorphism:

• Well-defined: Suppose [y]A = [z]A . Then
y ∼A z ⇒ f (y) ∼B f (z) ⇒ [f (y)]B = [f (z)]B ]
and so g is well-defined.
• 1-to-1: Suppose that [x]A and [y]A are such that g([x]A ) = g([y]A ). Then
[f (x)]B = [f (y)]B ⇒ f (x) ∼B f (y) ⇒ x ∼A y ⇒ [x]A = [y]A
and so g is 1-to-1.
• Onto: Let [y]B ∈ B/ ∼B , and let x = f −1 (y). Then
g([x]A ) = [f (x)]B = [y]B
and g is onto.
• Continuous: We want to show that g ◦ πA is continuous, so we can use the
Important Lemma. By the definition of g, g ◦πA = πB ◦f . Since f is a homeomor-
phism, it is continuous; πB is a quotient map and is thus continuous. So πB ◦ f is
continuous. Since g ◦ πA = πB ◦ f , we know that g ◦ πA is continuous. As πA is a
quotient map, it is continuous, so from the Important Lemma, g is continuous.
• Open: We can show equivalently that g −1 is continuous instead, since g is a
bijection.
g ◦ πA = πB ◦ f
g −1 ◦ (g ◦ πA ) ◦ f −1 = g −1 ◦ (πB ◦ f ) ◦ f −1
πA ◦ f −1 = g −1 ◦ πB
From here, the argument is eerily similar to the argument for continuity above.
 2. THE PRODUCT TOPOLOGY 29

Note that πA is a quotient map and f −1 s a homeomorphism, so both are contin-

uous. As such, πA ◦ f −1 is continuous, and therefore so is g −1 ◦ πB . Since πB is
a quotient map, it is continuous; from the Important Lemma, g −1 is continuous.
Therefore, g is open.
Therefore, g is a homeomorphism. Hurrah! 

2. The Product Topology

For the past several pages, we’ve been building new topological spaces from old by using
equivalence relations to form quotient spaces. Here, we will change directions and build
new topological spaces from old by taking products of spaces. To that end, we wish to
define the product of two sets:
Definition. Let X, Y be sets. The product of X and Y is given by:
X × Y ≡ {(x, y) : x ∈ X, y ∈ Y }

This is a topology class, so our intrinsic urge is to find a natural topology for the product
of two topological spaces. While our first instinct might be to take products of open sets
in our original spaces, this approach will give unsatisfactory results:
Example. Consider the sets A = (0, 1) × (0, 1) and B = (1/2, 3/2) × (1/2, 3/2) as subsets
of R × R = R2 with the usual topology. Then A ∪ B in R2 is NOT a product of an open
set in R with an open set in R as we would like. To intuitively see that this is not the case,
see the picture! On the other hand, A ∪ B is open in the usual topology on R2 .
30 2. MAKING NEW SPACES FROM OLD

Although products of open sets will not work because they are not closed under unions,
we can use products of open sets to construct our topology. Define the following:
Definition. Let (X, Fx ) and (Y, Fy ) be topological spaces. Then the set βX×Y is defined
by:
βX×Y = {A × B : A ∈ Fx , B ∈ Fy }
and define:
FX×Y = {∪i∈I Ui : I is some index set, and Ui ∈ βX×Y }

In other words, βX×Y is the set of products of an open set in X and an open set in Y , and
FX×Y is the set of unions of elements in βX×Y . The motivation for defining our sets this
way is that we want FX×Y to be a topology on X × Y and for β to be its basis. We will
now verify this claim with the following small fact:
Theorem 9. If (X, Fx ) and (Y, Fy ) are topological spaces, the space (X × Y, FX×Y ) is a
topological space with basis βX×Y .

Proof. We will apply our basis theorem; i.e. we want to show that
S
(1) U ∈βX×Y U = X × Y
(2) Given B1 , B2 ∈ βX×Y , for each x ∈ B1 ∩ B2 , there exists B3 ∈ BX×Y such that
x ∈ B3 ⊆ B1 ∩ B2 .
For the first statement, we know that X ∈ Fx and Y ∈ Fy by definition of a topology so
that X × Y ∈ βX×Y . It follows then that because each U ∈ βX×Y is subset of X × Y :
[
X ×Y ⊆ U ⊆X ×Y
U ∈βX×Y
S
so that X × Y = U ∈βX×Y U , as desired.
For the second statement, let B1 , B2 ∈ βX×Y and let (x, y) ∈ B1 ∩ B2 . Then by definition
of βX×Y , there exist U1 , U2 ∈ Fx and V1 , V2 ∈ Fy such that B1 ∩ B2 = (U1 × V1 ) ∩ (U2 × V2 ).
Our aim is to find B3 ∈ βX×Y such that B3 contains (x, y) and B3 ⊆ B1 ∩ B2 , so define:
B3 = (U1 ∩ U2 ) × (V1 ∩ V2 ).
Thus U1 , U2 ∈ Fx ⇒ U1 ∩ U2 ∈ Fx by the closure of topologies under finite intersections
and similarly, V1 ∩ V2 ∈ Fy , so B3 ∈ βX×Y . Since (x, y) ∈ (U1 × V1 ) ∩ (U2 × V2 ), then
(x, y) ∈ (U1 × V1 ) and (x, y) ∈ (U2 × V2 ). Therefore, x ∈ U1 , U2 and y ∈ V1 , V2 , so
x ∈ U1 ∩ U2 , and y ∈ V1 ∩ V2 . It immediately follows by the definition of the intersection
of sets that:
(x, y) ∈ (U1 ∩ U2 ) × (V1 ∩ V2 ) = B3
The only thing left to show is that B3 ⊆ B1 ∩ B2 . To that end, let (a, b) ∈ B3 . Then
a ∈ U1 ∩ U2 and b ∈ V1 ∩ V2 by definition of B3 . It follows that:
a ∈ U1 , U2 and b ∈ V1 , V2 ⇒ (a, b) ∈ U1 ×V1 and (a, b) ∈ U2 ×V2 ⇒ a ∈ (U1 ×V1 )∩(U2 ×V2 ).
 2. THE PRODUCT TOPOLOGY 31

Therefore, (a, b) ∈ B1 ∩ B2 so that B3 ⊆ B1 ∩ B2 because (a, b) is an arbitrary element of

B3 .
Therefore, βX×Y satisfies the hypotheses of our basis theorem, so the set of unions of
elements of βX×Y , FX×Y , is a topology for X × Y and βX×Y is a basis for the topology
FX×Y . 

We have successfully devised a topology for product spaces as unions of products of open
sets.

2.1. Examples.
Example. The set S 1 × [0, 1] looks like a cylinder! What kinds of sets are open in the
cylinder? A particular example is an open disc projected on the face of the cylinder and
unions thereof.

[0,1]

S1

Example. The set S 1 × S 1 looks like a torus! What kinds of sets are open in the torus?
Similar to the previous example, open discs projected onto the torus surface are examples
of open sets in S 1 × S 1 . In the picture, one copy of S 1 is red and one is green; they
determine a torus.

Example. The set S 1 × S 1 × S 1 is actually a 3-torus!

32 2. MAKING NEW SPACES FROM OLD

Not every space is a product of more than one space. For example the sphere S 2 is not
a product of more than one space. An intuitive (non-rigorous) justification is that the
natural axes of a sphere are the great circles; however, every pair of distinct great circles
intersect twice which makes it hard to define a coordinate system on the sphere.

2.2. The Product Projection Map. Now that we have a product topology to im-
pose on product spaces, we want a way to relate the product topology to the topologies of
the constituent spaces. In order to do so, we will define the projection maps as follows:
Definition. Let (X, Fx ) and (Y, Fy ) be topological spaces and create X × Y endowed
with the product topology FX×Y . Define πX : (X × Y, FX×Y ) → (X, Fx ) and πY : (X ×
Y, FX×Y ) → (Y, Fy ) by:
πX ((x, y)) = x πY ((x, y)) = y.
The map πX is the projection onto X and πY is the projection onto Y .

A small fact which we will derive now is that the product projection maps are continuous:
Theorem 10. Let (X, FX ) and (Y, FY ) be topological spaces and (X × Y, FX×Y ) be their
product with the induced product topology. Then the projection maps πX and πY onto X
and Y are continuous.

Proof. Suppose O ⊆ X and O ∈ Fx . We see that π −1 (O) = O × Y ∈ FX×Y .

Therefore, the preimage of any open set in X under πX is open in X × Y with the product
topology. A similar argument shows that πY is continuous. 

Recall that the quotient projection map is not necessarily an open map. It turns out that
the product projection map is an open map. Accidentally assuming that the quotient map
is open is a very common mistake that one should be aware of! We will now prove that
the product projection map is open:
Theorem 11. Let (X, FX ) and (Y, FY ) be topological spaces and (X × Y, FX×Y ) be their
product with the induced product topology. Then the projection maps πX and πY onto X
and Y are open.

Proof. Let O = U × V such that U ∈ Fx and V ∈ Fy and O ∈ βX×Y . Therefore:

πX (O) = U ∈ Fx
So if C is some collection of sets in βX×Y , then:
[
V
V ∈C
is an arbitrary element of FX×Y by the definition of a basis and:
!
[ [
πX V = πX (V )
V ∈C V ∈C
 2. THE PRODUCT TOPOLOGY 33

which is open because the preceding statments indicate that V ∈ βX×Y ⇒ πX (V ) ∈ FX .

Consequently, the image of the arbitrary open set in (X × Y, FX×Y ) is a union of open sets
in (X, Fx ) and is thus open.
The proof is similar for πY . 

2.3. Using Bases More Effectively. In metric spaces, open balls are bases for the
metric topology. By proving properties about open balls, we were able to say they apply
to the entire set. We would like to prove the following important yet tiny lemma so that
we can use bases in topological spaces like open balls in metric spaces:
Lemma. Let (X, Fx ) be a topological space with basis β. If W ⊆ X then W ∈ Fx if and
only if for all p ∈ W, there exists Bp ∈ β such that p ∈ Bp ⊆ W .

Proof. (⇒) Suppose W ∈ Fx . Let p ∈ W . Since W ∈ Fx , for some index set I:

[
W = Bi ∀i ∈ I, Bi ∈ β
i∈I
because every element of a topology can be written as a union of basis elements.
Therefore, by definition of the union, p ∈ W implies that there exists ip ∈ I such
that p ∈ Bip , and Bip ⊆ W . If we let Bp = Bip , we are finished.
(⇐) Suppose that for all p ∈ W , there exists Bp ∈ β such that p ∈ Bp ⊆ W . We want
to show that W ∈ Fx . We see that:
[
V = Bp ⊆ W
p∈W
because each of the constituent Bp ⊆ W , and every p ∈ W is an element of Bp ,
so the union of Bp over p ∈ W , contains every p ∈ W . Consequently, V ⊆ W ⇒
V = W . Since β is a basis and V is a union of basis elements, V ∈ Fx ⇒ W ∈ Fx .


In other words, if we have a topological space with a basis, then every point in an open
set U is an element of a basis element contained in U , giving us a structure very similar to
the metric topology.

2.4. Finding and Constructing Continuous Maps to Product Spaces. Now

that we have product spaces and have addressed their basic topological properties, we
would like a way to easily find and construct continuous maps to the product space. To
that end we introduce the following important lemma:
Lemma. Let (X, FX ), (Y, FY ) and (A, FA ) be topological space and let (X × Y, FX×Y ) be
the product space of X, Y with the induced product topology. Suppose f : A → X and
g : A → Y and define
h : A → (X × Y ) by h(a) = (f (a), g(a)),
34 2. MAKING NEW SPACES FROM OLD

then h is continuous if and only if f, g are continuous.

Proof. (⇒) Suppose h is continuous: then we see that:

f = πX ◦ h g = πY ◦ h
so f, g are continuous because they are compositions of continuous functions.
(⇐) Suppose that f and g are continuous. We prove that the preimage of every basis
element under h is open. From a theorem we proved some time ago, this shows
that h is continuous.
Let U × V ∈ βX×Y . Then U ∈ FX and V ∈ FY . Since f and g are continuous,
f −1 (U ) ∈ FA and g −1 (V ) ∈ FA . Then
h−1 (U × V ) = {a ∈ A | h(a) ∈ U × V }
= {a ∈ A | (f (a), g(a)) ∈ U × V }
= {a ∈ A | f (a) ∈ U, g(a) ∈ V }
= f −1 (U ) ∩ g −1 (V )
Then h−1 (U × V ) is the intersection of two open sets, hence it is open. Therefore,
h is continuous.


The following example illustrates how this lemma makes it very easy to define continuous
functions to the product space:
Example. Suppose f : R → R and g : R → R with f (x) = x2 + 3x and g(x) = sin(x).
Then the map h : R → R2 defined by h(x) = (x2 + 3x, sin(x)) is continuous because f, g
are.

2.5. Relating Products to Subspaces.

Small Fact. Let (X, FX ) and (Y, FY ) be topological spaces with subspaces (A, FA ) and
(B, FB ) respectively. Let FS denote the subspace topology on A × B ⊆ X × Y and let
FA×B denote the product topology on A × B. Then FS = FA×B .

This means that we can either look at A × B as a subspace of X × Y , then induce the
subspace topology, or we can induce the subspace topologies on A and B, then take the
cross product. Essentially, taking the cross product and creating subspaces “commute.”

Proof. We start by reviewing which sets are open under each topology:
FS = {(A × B) ∩ U | U ∈ FX×Y }
FX×Y = unions of sets of the form U × V , with U ∈ FA , V ∈ FB
 2. THE PRODUCT TOPOLOGY 35

Let U ∩ (A × B) ∈ FS . Then
[
U ∩ (A × B) = (Ui × Vi ) ∩ (A × B) (where ∀i ∈ I, Ui ∈ FX , Vi ∈ FY )
i∈I
Then
[
U ∩ (A × B) = (Ui × Vi ) ∩ (A × B)
i∈I
= {(x, y) ∈ A × B | (x, y) ∈ Ui0 × Vi0 for some i0 ∈ I}
= {(x, y) ∈ X × Y | x ∈ Ui0 ∩ A, y ∈ Vi0 ∩ B for some i0 ∈ I}
[
= (Uj ∩ A) × (Vj ∩ B) for some index setJ}
j∈J
∈ FX×Y
The same argument run backwards shows that FX×Y ⊆ FS .
Therefore, FS = FX×Y . 

2.6. Products and Quotient Spaces. Taking products and quotients does not
“commute” in the same way that taking products and subspaces does. We demonstrate a
(lengthy) counterexample to this idea.
Consider R with the usual topology, with x ∼ Y iff x = y or x, y ∈ N. This looks something
like a “sideways infinite flower.”

If a set open in R/ ∼ contains a (“the”) natural number, then its preimage must contain
every natural number, so it must contain an open interval about “the” natural number in
every direction (all infinity of them).
36 2. MAKING NEW SPACES FROM OLD

More formally, let π : R → R/ ∼ be the projection map. Let idQ : Q → Q be the identity
map, where each copy of Q has the usual topology. Note that idQ is a quotient map (where
∼ is ‘=’).
Now, consider π × idQ : R → Q → (R/ ∼) × Q by (π × idQ )(x, y) = (π(x), i(y)). We claim
that π × idQ is not a quotient map. To prove this, we show that the product topology on
(R/ ∼) × Q is not the same as the quotient topology
{U ∈ (R/ ∼) × Q | (π × idQ )(U ) ∈ FR×Q }
To do this, let’s find some U 0 ∈ Fπ×i such that U 0 ∈
/ F(R/∼)×Q . We want to construct U 0
as a union.
For all n ∈ N define Un to be the interior of the region of R × Q bounded
√ by the vertical
lines n − 41 and n + 14 and above and below by the lines through (n, n2 ) with slope ±1.
Each one of these sets has a vertical line through the natural number.
So ∀n ∈ N, {n} × Q ⊆ Un . Observe that ∀n ∈ N, Un ∈ FR×Q (it is an interior, so it is open).
S
Let n∈N Un ∈ FR×Q (also open in R × Q because it is a union).
Let U 0 = (π × i)(U ). This will glue together the strips along the vertical lines with R
coordinates in N. This can be pictured as a fan.
.

..
U1 U2 U3 U4 ...

U’

Claim: U 0 ∈ Fπ×i

Proof. : (π × i)−1 (U 0 ) = U and U is a union of Un0 s

 2. THE PRODUCT TOPOLOGY 37

Therefore, by definition, U 0 ∈ Fπ×i .

Claim: U 0 6∈ F(R/∼)×Q .

Proof. : Let ([n], 0) ∈ U 0 . Suppose ∃W ∈ FR/∼ and V ∈ FQ such that ([n], 0) ∈

W × V ⊆ U 0.
Then π −1 (W ) ∈ FR and of course i−1 (V ) = V .
So π −1 (W ) × V is open in R × Q such that π −1 (W ) × V ⊆ (π × i)−1 (U 0 ) = U .
Also, π −1 (W ) is open in R and contains N and V is open in Q and contains {0}.
√
So ∃δ > 0 such that (−δ, δ) ∩ Q ⊆ V . There exists an n ∈ N such that δ > 2
n . π −1 (W ) is
open in R, so it’s on the x-axis.
So ∃ > 0 such that (n − , n + ) ⊆ π −1 (W ).
So the interval (n − , n + ) × (−δ, δ) ⊆ π −1 (W ) × V . So  < 1
4 from our previous definition
of Un .
Let x = n + 2 . Question: Where does x = n + 
2 meet the boundaries of Un ? It meets at
√
2 
y= n ± 2.
√ √
2  2
∃y ∈ Q such that n − 2 <y< n + 2 .
So (x, y) ∈ / Um since (x, y) ∈ (n − , n + ) ⊆ (n − 14 , n + 41 .
/ Un and ∀m 6= n, (x, y) ∈
But (x, y) ∈ (n − , n + ) × (−δ, δ) ⊆ π −1 (W ) × V ⊆ U ⇒ (x, y) ∈
/ U . Therefore we have
a contradiction.
∴ U0 ∈
/ F(R/∼)×Q
So we have shown that Fπ×i 6= F(R/∼)×Q .

2.7. Infinite Products. We would of course like to be able to generalize this to deal
with infinite products. Initially, we may simply want to define such products as follows:
X1 xX2 x...= {(x1 , x2 , ...)|xi ∈ Xi ∀i ∈ N}. However, such a definition limits us to countable
products, so we look more generally.
Definition. j ∈ J, let Xj be a set. Define the product Πj∈J Xj = {f : J → ∪j∈J Xj |f (j) ∈
Xj }. We refer to f (j) as the j th coordinate of the point f .
Example. Suppose J = {1, 2}. Then Πj∈{1,2} Xj = {f : {1, 2} → X1 ∪ X2 |f (j) ∈ Xj } =
{(f (1), f (2))|f (j) ∈ Xj } = {(x1 , x2 )|xj ∈ Xj } = X1 x X2 . Thus we see that this definition
agrees with our previous definition for finite products.
38 2. MAKING NEW SPACES FROM OLD

Example. Consider Πj∈R {1, 2}. By definition this is equivalent to {f : R → {1, 2}|f (j) ∈
{1, 2}, j ∈ R}, which precisely correspond to subsets of R if we simply think of the preimage
of 1 under f as the elements in the set and the preimage of 2 under f as those elements
outside the set. Commonly this product is then denoted by {1, 2}R .

Now we wish to define a topology on these products which, as in example 1, agrees with
our prior definition for a product of two sets if the indexing set J = {1, 2}. Perhaps the
most natural way of doing this is to define the following basis:

β = {Πj∈J Uj |Uj ∈ Fj }, where Xj has topology Fj .

F = {Unions of elements of β }.

This topology is not what we desire, but is a topology, aptly named the box topology on
the product.
Example. The product topology on Πj∈J Xj is given by the basis βΠ = {Πj∈J Uj |Uj = Xj
for all but finitely many j and ∀j ∈ J, Uj ∈ Fj }.

Remarks:
(1) βΠ ⊆ β
(2) Both are bases for topologies on the product.
Definition. Define the projection map πj : Πj∈J Xj → Xj by π(f ) = f (j).
Lemma. For all j ∈ J, Suppose (Xj , Fj ) is a topological space. Then πj as defined above
is continuous for all j ∈ J.

Proof. Let j0 ∈ J and consider U ∈ Fj0 . We wish to show that πj−1 0

(U ) ∈ FΠ .
−1
Note that πj0 (U ) = {f ∈ Πj∈J Xj |πj0 (f ) ∈ U } = {f ∈ Πj∈J Xj |f (j0 ) ∈ U } = {f ∈
Πj∈J Xj |f (j0 ) ∈ U, ∀j 6= j0 f (j) ∈ Xj } = Πj∈J Uj such that Uj0 = U and ∀j 6= j0 , Uj = Xj .
It then follows from definitions that πj−1 0
(U ) ∈ βΠ ⊆ FΠ so the projection map is continuous.


We now prove an Important Lemma for Infinite Products.

Lemma. Let (Xj , Fj ) and (Y, FY ) be topological spaces for all j ∈ J. Moreover, for each
such j let gj : Y → Xj be a function. Define h : Y → Πj∈J Xj by: h(y) = f such that
∀j ∈ J, f (j) = gj (y). Then h is continuous if and only if gj is continuous for all j ∈ J.

Proof. (⇒) Suppose h is continuous and let j ∈ J. Then πj ◦ h = gj . Thus gj is

the composition of continuous functions and must itself be continuous.
 2. THE PRODUCT TOPOLOGY 39

(⇐) Suppose that gj is continuous for all j ∈ J. Let U ∈ βΠ . We wish to show that
h−1 (U ) ∈ FY .
Note that h−1 (U ) = {y ∈ Y |h(x) ∈ U } As an open set in the product topology,
U = Πj∈J Uj where Uj ∈ Fj . Thus, h−1 (U ) = {y ∈ Y |h(x) ∈ Πj∈J Uj } = {f ∈
Πj∈J Uj such that ∀j ∈ J, f (j) = gj (y)} = {y ∈ Y |gj (y) ∈ Uj } = {y ∈ Y |y ∈
gj−1 (Uj )∀j ∈ J} = ∩j∈J Uj .

Since gj is continuous for all j, gj−1 (Uj ) is open in Y for all j. Also, by definition
of the product topology, Uj = Xj for all but at most finitely many j. Thus,
gj−1 (Uj ) = Y for all but at most finitely many j. It follows that the intersection
∩j∈J Uj is a finite intersection of open sets since removing all trivial indices will
not change the intersection. Thus, h−1 (U ) ∈ FY so h is continuous, completing
the proof.

 CHAPTER 3

Distinguishing Spaces

Definition. A topological property (or ‘top. prop.’) is a property of a topological

spaces that is preserved by homeomorphisms.

List of Topological Properties thus Far

(1) Cardinality of X
(2) Cardinality of Fx
(3) Metrizability (we proved this in the homework)
(4) Discreteness
(5) Indiscreteness
Lemma (Rachel’s Lemma). Let (X, FX ) and (Y, FY ) be topological spaces and f : X → Y
be an open bijection. If FX is the discrete topology, then FY is the discrete topology.

Proof. Let y ∈ Y . Since f is surjective, there exists x ∈ X such that f (x) = y. Since
{x} ∈ FX and f open, f ({x}) ∈ FY . Thus all singletons in are elements of FY and all sets
in Y are unions of singletons and hence elements of FY , so every set is an open set and FY
is the discrete topology. 
Lemma (Daniel’s Lemma). Let (X, FX ) and (Y, FY ) be topological spaces and f : X → Y be
a continuous bijection. If FX is the indiscrete topology, then FY is the indiscrete topology.

Proof. Let U ∈ FY . WTS U = Y or U = ∅. Since f is continuous, f −1 (U ) ∈ FX .

Therefore f −1 (U ) = X or ∅.
Suppose f −1 (U ) = X. Then U = f (f −1 (U )) = f (X) = Y since f surjective.
Suppose f −1 (U ) = ∅. Then U = ∅. Therefore FY is the indiscrete topology. 

Corollaries: If (X, FX ) and (Y, FY ) topological spaces and f : X → Y a homeomorphism,

then
(1) FX the discrete topology =⇒ FY the discrete topology
(2) FX the indiscrete topology =⇒ FY the indiscrete topology
41
42 3. DISTINGUISHING SPACES

Unfortunately, even with these five lovely Topological Properties, we can’t yet distinguish
a circle from a line. Clearly there is more work to do.
Example (a non-example). Distance is not a topological property. A big circle is homeo-
morphic to a little circle.

1. Compactness

Definition. Let (X, FX ) be a topological space and S ⊆ X. S

We say {Uj : j ∈ J} is an open cover of S if for all j ∈ J, Uj ∈ FX and S ⊆ j∈J Uj .
S
We say {Uj : j ∈ K} is a subcover if K ⊆ J and S ⊆ j∈K Uj .
We say S is compact if every open cover of S has a finite subcover.

IMPORTANT WARNING!
We learned in Math 131 that in Rn , a set if compact ⇐⇒ it is closed and bounded. THIS
IS NOT TRUE IN TOPOLOGICAL SPACES. DO NOT TRY TO USE IT.
Example. Is the set (0, 1) compact in
(1) R with the discrete topology?
(2) R with the half-open topology?
(3) R with the finite complement topology?

Answers:
(1) No. Take the open covering {B1/2 (x) : x ∈ (0, 1)}. If we remove any of these
balls, the corresponding x will no longer be covered. However, there are clearly
uncountably infinitely many balls.

(2) No. Take the open covering {[ n1 , 1) : n ∈ N} so that (0, 1) = 1

S
n∈N ([ n , 1). There
is not finite subcover.

(3) Yes. Suppose we have a cover of (0, 1). Take any element of the cover, say
U47 . U47 is missing at most finitely may elements of (0, 1) since its complement
is finite. For each element x ∈ R \ U47 , select one Ujx containing x. The set
{U47 } ∪ {Ujx : x ∈ R \ U47 } is a finite open subcover.
Theorem 12 (analogous to a theorem from 131). Let (X, FX ) and (Y, FY ) be topological
spaces, S ⊆ X, and f : X → Y continuous. If S is compact then f (S) is compact.

Proof. (1) Take a covering of f (S).

 1. COMPACTNESS 43

(2) Applying f −1 to these sets, we pull back to an open covering of S.

(3) Take a finite subcover.
(4) Push these forward again. We have a finite cover of f (S).


As a corollary we have Topological Property 6: Compactness.

Theorem 13. Any closed subset of a compact space is compact.

Proof. (1) Take an open cover of S

(2) Add X \ S. Now we have an open cover of X.
(3) Take a finite subcover.
(4) Remove X \ S. Not we have a finite subcover of S.

Theorem 14 (also analagous to a 131 theorem). Let (X, FX ) be compact and f : X → R
be continuous. Then f has a max value and a min value.

Proof. (1) f (X) is compact by Analogous Theorem 1

(2) Therefore f (X) is closed and bounded
(3) Therefore f (X) has a lub and a glb (ie a max and a min)

Small Fact. Let (X, FX ) be a topological space. Suppose every open cover of S ⊆ X
made up of basis elements has a finite subcover. Then X is compact.

S {Uj : j ∈ J} be an open cover of S and β be a basis of S. For every

Proof. Let
j ∈ J, Uj = i∈Ij Bi where for every i ∈ Ij Bi ∈ β. Hence {Bi : i ∈ Ij and j ∈ J} is an
open cover of S. By hypothesis, we have a finite subcover, {Bi : i ∈ Kj and j ∈ K} where
K ⊆ J and K is finite, and for all j ∈ K, Kj ⊆ Ij and Kj is finite. Now note that given
any j ∈ K, for every i ∈ Kj Bi ⊆ Uj . Hence
 
[ [ [
X=  Bi  ⊆ Uj ⊆ X
j∈K i∈Kj j∈K

and {Uj : j ∈ K} is a finite subcover. 

44 3. DISTINGUISHING SPACES

Theorem 15 (Bolzano-Weierstrass Theorem). Let (X, FX ) be a compact topological space.

Then for all infinite S ⊆ X, ∃ p ∈ X such that ∀ U ∈ FX with p ∈ U , U contains infinitely
many points of S.

Proof. Suppose that for some such S ⊆ X, no such p ∈ X exists. Therefore, [ for all
p ∈ X there exists some Up ∈ FX such that p ∈ Up and |S ∩Up | is finite. So, as Up = X,
p∈X
{Up | p ∈ X} is an open cover of X. As X is compact, there exists some finite subcover
n
[
{Up1 , ..., Upn }. Hence, from the definition of a subcover, X = Upi . Intersecting both
i=1
sides with S gives that:
n n
!
[ [
S= Upi ∩S = (Upi ∩ S)
i=1 i=1

However, as |Up ∩ S| is finite for all pS∈ X, |Upi ∩ S| is finite for 1 ≤ i ≤ n. The union
of finite sets is finite, and therefore | ni=1 (Upi ∩ S)| is finite. However, S is infinite by
assumption. So this is a contradiction, and no such S exists. 
Theorem 16 (Finite Tychonoff Theorem). Let (X, FX ) and (Y, FY ) be compact topological
spaces. Then (X × Y, FX×Y ) is also a compact topological space.

Proof. (Comic book style)

(a) Consider some open cover. (b) Restrict to some {x} × Y , which is compact,
so has a finite subcover.
 1. COMPACTNESS 45

Ux Ux Ux
1 2 3

(c) So there is some open Ux with x ∈ Ux s.t. (d) So consider just the Ux , which are a cover of
Ux × Y is contained in this subcover. X, so have a finite subcover.

So the original sets that covered {x} × Y corresponding to the chosen Ux for the desired
finite cover of X × Y . 

Now let’s do a real proof.

Proof. (Real) Consider an open cover of X × Y consisting of basis elements, {Ui × Vi |

i ∈ I}, where Ui ∈ FX and Vi ∈ FY for all i ∈ I.
Let x ∈ X. It is clear that {x} × Y ∼ = Y , so {x} × Y is compact. So the original open
cover is a cover of {x} × Y , and there exists some finite subcover {Ui(x) × Vi(x) | i(x) ∈ Kx }
where Kx ⊆ I and Kx is finite. If x ∈ / Ui(x) , removing the set Ui(x) × Vi(x) does not change
the fact that the remaining set is a finite subcover of {x} × Y . Hence, without loss of
generality, we assume that x ∈ Ui(x) for all i(x) ∈ Kx .
\
Next, define Ux = Ui(x) . As Ui(x) ∈ FX for all i(x) ∈ Kx , Ux is the intersection of
i(x)∈Kx
finitely many open sets, and so is open. Additionally, as x ∈ Ui(x) for all i(x) ∈ Kx , x ∈ Ux .
Claim 1: {Ux × Vi(x) | i(x) ∈ Kx } is an open cover of {x} × Y .
Proof: Let (x, y) ∈ {x} × Y . Then for some i(x) ∈ Kx , (x, y) ∈ Ui(x) × Vi(x) . So y ∈ Vi(x) .
Then, as x ∈ Ux , (x, y) ∈ Ux × Vi(x) . As such an i(x) ∈ Kx exists for all (x, y) ∈ {x} × Y ,
{Ux × Vi(x) | i(x) ∈ Kx } is an open cover of {x} × Y .
Now, as x ∈ Ux and Ux ∈ FX for all x ∈ X, {Ux | x ∈ X} is an open cover of X. As X is
compact, this must have a finite subcover, {Ux | x ∈ L}, where L ⊆ X and L is finite.
46 3. DISTINGUISHING SPACES

Claim 2: {Ux × Vi(x) | x ∈ L, i(x) ∈ Kx } is a cover of X × Y .

Proof: Let (x0 , y0 ) ∈ X × Y . As {Ux | x ∈ L} covers X, there exists some x ∈ L such that
x0 ∈ Ux . As {Ux × Vi(x) | i(x) ∈ Kx } covers {x} × Y , there exists some i(x) ∈ Kx such
that (x, y0 ) ∈ Ux × Vi(x) , so y0 ∈ Vi(x) . Hence (x0 , y0 ) ∈ Ux × Vi(x) , and (x0 , y0 ) is in the
cover. As this holds for all such (x0 , y0 ) ∈ X × Y , {Ux × Vi(x) | x ∈ L, i(x) ∈ Kx } is a cover
of X × Y .
Claim 3: {Ui(x) × Vi(x) | x ∈ L, i(x) ∈ Kx } is a cover of X × Y .
Proof: Let (x0 , y0 ) ∈ X × Y . As {Ux × Vi(x) | x ∈ L, i(x) ∈ Kx } is a cover of X × Y , there
exists some x ∈ L, i(x) ∈ Kx such that (x0 , y0 ) ∈ Ux × Vi(x) . From the definition of Ux ,
Ux ⊆ Ui(x) . Therefore, as x0 ∈ Ux , x0 ∈ Ui(x) , and (x0 , y0 ) ∈ Ui(x) × Vi(x) . As such a x ∈ L,
i(x) ∈ Kx exists for all (x0 , y0 ) ∈ X × Y , this is a cover of X × Y .
S
Finally, let J = x∈L Kx . As L is finite, and Kx is finite for all x ∈ X, J is finite.
Additionally, as Kx ⊆ I for all x ∈ X, J ⊆ I. So {Ui × Vi | i ∈ J} is a finite subcover of
{Ui × Vi | i ∈ I}. Also, as this has {Ui × Vi | i ∈ J} = {Ui(x) × Vi(x) | x ∈ L, i(x) ∈ Kx },
this is a cover of X × Y .
So {Ui × Vi | i ∈ J} is a finite subcover of X × Y . As such a subcover exists for any cover
of basis elements, X × Y is compact. 

2. Hausdorffness

Definition. Let (X, FX ) be a topological space. We say that X is Hausdorff if ∀ p, q ∈ X

such that p 6= q, ∃ disjoint sets U, V ∈ FX such that p ∈ U , q ∈ V .
Example. Metric spaces are Hausdorff.
Example (a non-example). Any space containing at least two points under the indiscrete
topology is not Hausdorff.

Hausdorff is important because any “reasonable” space/any space that we want to be in

will be Hausdorff.
Note: Continuous functions may not preserve Hausdorff-ness.
Example. Define f : (R, discrete) → (R, indiscrete) by the identity map.

This is continuous because the domain has the discrete topology, but the domain is Haus-
dorff while the range clearly is not.
Small Fact. Suppose f : (X, FX ) → (Y, FY ) is a homeomorphism and (X, FX ) is Haus-
dorff. Then (Y, FY ) is also Hausdorff.
 2. HAUSDORFFNESS 47

Proof. Let p 6= q ∈ Y . Because f is a bijection, f −1 (p) and f −1 (q) are distinct points
in X.
Since X is Hausdorff, ∃ U, V ∈ FX such that U ∩ V = ∅, f −1 (p) ∈ U , f −1 (q) ∈ V .
Consider f (U ), f (V ). Since f is a homeomorphism and, thus, open, f (U ) and f (V ) are
open. Then, because f is a bijection, we have p ∈ f (U ), q ∈ f (V ), and f (U ) ∩ f (V ) =
∅ 

We are now interested in how Hausdorff-ness and compactness are related.

Flapan Says... “ Hausdorff-ness and compactness are like two people who love each other,
because when two people love each other, they can do so much more together than they
can alone.”

Theorem 17. Let (X, FX ) be Hausdorff. Let A ⊆ X be compact. Then A is closed.

Proof. (Compare the following to the proof that all compact subsets of metric spaces
are closed. That is, we are going to show that we don’t need a metric, just Hausdorff-ness,
for closed subsets to be compact.)
Rather than show A is closed, we will show X − A is open.
Let p ∈ X − A. We want to show ∃ U ∈ FX such that p ∈ U ⊆ X \ A. Let a ∈ A. Because
X is Hausdorff, ∃ Ua , Va ∈ FX such that p ∈ Ua , a ∈ Va , Ua ∩ Va = ∅.
Consider {Va |a ∈ A}. This is an open cover of A because Va open and a ∈ Va ∀a ∈ A. So,
because A is compact, ∃ finite subcover {Va1 , Va2 , ..., Van }
Let U = ni=1 Uai . U ∈ FX since it is a finite intersection of elements of FX and p ∈ U
T
since p ∈ Uai ∀ i = 1, 2, ..., n.
Claim: U ⊆ X − A
Proof. ∀ i = 1, 2, ... n, we have Uai ∩ Vai = ∅. Also, A ⊆ ni=1 Vai and U ∩ A ⊆ U ∩ ni=1 Vai .
S S

However, U ∩ ni=1 Vai = ∅ because if ∃ x ∈ U ∩ ni=1 Vai , then ∃ io such that x ∈ Vaio ∩Uaio .
S S
But, by definition, Vaio ∩ Uaio = ∅, so no such x exists. Thus,

n
[
U ∩A⊆U ∩ Vai = ∅,
i=1
meaning U ∩ A = ∅. So, because U ⊆ X but U ∩ A = ∅, U ⊆ X − A
Thus, we have found U ∈ FX such that p ∈ U , U ⊆ X − A. Thus, X − A is open. Thus,
A is closed. 
Corollary. In any Hausdorff space, (finite sets of ) points are closed sets.
48 3. DISTINGUISHING SPACES

Proof. Let p ∈ (X, FX ) and let (X, FX ) be Hausdorff. Let {Ui |i ∈ I} be an open
cover of {p}. Then ∃ io ∈ I such that p ∈ Uio . Thus {Uio } is a finite subcover. Hence, {p}
is compact.
Thus, by Theorem above, {p} is closed. 
Lemma (it’s important!). Let f : (X, FX ) → (Y, FY ) be continuous. Let X be compact,
and Y be Hausdorff. Then f is a homeomorphism if and only if it is a bijection.

Proof. (⇒) Since f is a homeomorphism, f is a bijection.

(⇐) Suppose f is a continuous bijection. We want to show that f is open. Since f is
a bijection, this is equivalent to showing that f is closed.
Let A ⊆ X be closed. Since X is compact, by the second analogous theorem
above, A is compact. By the first analogous theorem, f (A) is compact. By the
above theorem, f (A) is closed. Thus, f is closed, and, thus, open. Thus, f is an
open, continuous bijection.
Thus, f is a homeomorphism.
Thus, f is a homeomorphism if and only if it is a bijection.

Example. Let X = [0, 1] × [0, 1], and let (a, b) ∼ (c, d) if either (a, b) = (c, d) or b = d 6= 0.

X X/~

Points of the same color are identified. Points on the x-axis are not separable, so this is a
Hausdorff space X and an equivalence relation ∼ such that X/ ∼ is not Hausdorff.
Example. Let X = R, and set x ∼ y if and only if x = y or x, y > 0.

Let p > 0. Then [p], [0] ∈ X/ ∼. Let U ⊆ X/ ∼ be an open set containing [0]. We claim
that [p] ∈ U . To see why, note that π −1 (U ) is an open subset of R containing 0. Hence,
 3. NORMALNESS 49

there exists q ∈ π −1 (U ) such that q > 0. Therefore, q ∼ p ⇒ [q] = [p] ∈ U , and X/ ∼ is

not Hausdorff.

Note: this proves that the continuous images of a Hausdorff space need not be Hausdorff!

3. Normalness

Definition. Let (X, Fx ) be a topological space. We say X is normal if for every pair of
disjoint closed sets A, B ⊆ X, there exist disjoint open sets U, V ⊆ X such that A ⊆ U
and B ⊆ V .

On Homework 2, we proved that metric spaces are normal.

Example. Consider R with a topology F defined by a basis of all sets of the form (a, b)
plus all sets of the form (a, b) ∩ Q. This is known as the rational topology, and it is finer
than the usual topology. As an exercise, you can prove that that this collection of sets
really form the basis of a topology. (Use the basis lemma.)
We want to show that (R, F ) is Hausdorff. This is easy, because F contains the usual
topology on on R, which is Hausdorff.
Next, we will show that (R, F ) is not normal. To see why, let A = R \ Q. Then A is closed,
because Q is contained in F . Let B = {47}. Suppose there exist U, V ∈ F such that A ⊆ U,
and B ⊆ V . Then there exists  > 0 such that (47 − , 47 + ) ∩ Q ⊆ V .
Let p ∈ (47 − , 47 + ) such that p 6∈ Q. Then p ∈ U because p 6∈ Q. So, there exists δ > 0
such that (p − δ, p + δ) ⊆ U .
Now, there must exist a point in x ∈ (p−δ, p+δ)∩((47−, 47+)∩Q). But then x ∈ U ∩V ,
and so U ∩ V 6= ∅. This proves that (R, F ) is not a normal space.

In particular, we have shown that a space can be Hausdorff but not normal. Under what
conditions must a Hausdorff space be normal?
Lemma. If (X, Fx ) is Hausdorff and compact, then X is normal.

Proof. Comic-book style

50 3. DISTINGUISHING SPACES

A
a a

(a) Let A and B be closed and disjoint sets. (b) Each point in B can be separated from a by an
Let a ∈ A. open set. This forms an open cover. Take a finite
subcover.

(c) Repeat for each a ∈ A. This makes an open (d) Take the intersection of all of the covers of B.
cover of A and a lot of finite open covers of B. It’s disjoint from the cover of A, so A and B are
separated by open sets, so X is normal.


 3. NORMALNESS 51

Now for the real proof.

Proof. Let A and B be disjoint closed subsets of X. Then A and B are compact
because X is compact. Let a ∈ A. For every b ∈ B, there exist open sets Ub and Vb such
that a ∈ Ub and b ∈ Vb , and Ub ∩ Vb = ∅. Then {Vb | b ∈ B} is an open cover of B. Since
B is compact, we can choose a finite subcover {Vb1 , . . . , Vbn }. Let Ua = ∩ni=1 Ubi . For every
a ∈ A, Ua ∈ Fx and a ∈ Ua . Let Va = ∪ni=1 Vbi . Then B ⊆ Va ∈ Fx .
We claim that Ua ∩ Va = ∅ for all a ∈ A. To see why, note that (∩ni=1 Ubi ) ∩ (∪ni=1 Vbi ) = ∅
because for all i, Ubi ∩ Vbi = ∅.
Now, {Ua | a ∈ A} is an open cover of A. Since A is compact, this cover has a finite
subcover {Ua1 , . . . , Uam }. Now, V = ∩m i=1 Vai is open and B ⊆ V . For all i = 1, . . . , m,
Uai ∩ (∩m V
i=1 ai ) = ∅ because for all i = 1, . . . , m, Uai ∩ Vai = ∅.
Let U = ∪m
i=1 Uai ∈ Fx . Then A ⊆ U and U ∩ V = ∅ because Uai ∩ Vai = ∅ for all i.

Therefore, X is normal. 
Theorem 18. Let (X, FX ) be a compact and Hausdorff topological space. If (Y, FY ) is a
topological space and f : X → Y is continuous, onto, and closed, then (Y, FY ) is compact
and Hausdorff.

Proof. From previous notes, since f is continuous and X is compact, Y is compact.

So now consider Hausdorff:
Let p, q ∈ Y such that p 6= q. As f is onto, ∃ a, b ∈ X such that f (a) = p and f (b) = q. As
X is Hausdorff, {a} and {b} are closed. As f is closed, f ({a}) = {p} and f ({b}) = {q} are
closed. Therefore, as f is continuous, f −1 ({p}) and f −1 ({q}) are closed, and are clearly
disjoint.
As X is both compact and Hausdorff, it is also normal. So ∃ U, V ∈ FX such that
f −1 ({p}) ⊆ U , f −1 ({q}) ⊆ V , and U ∩ V = ∅. As these sets are open, A = U c and B = V c
are closed. As f is closed, f (A) and f (B) are also closed, and hence f (A)c and f (B)c are
open.
Suppose that p ∈ f (A). Then for some x ∈ A, f (x) = p. But then x ∈ f −1 ({p}) ⊆ U = Ac .
So x ∈ A∩Ac , which is a contradiction. So p ∈ f (A)c , and by a similar argument, q ∈ f (B)c .
Now suppose that c ∈ f (A)c ∩ f (B)c . So, as f is onto, there exists some d ∈ X such
that f (d) = c. As c ∈
/ f (A), d ∈
/ A, so d ∈ U . Similarly, as c ∈
/ f (B), d ∈ V . But then
d ∈ U ∩ V = ∅. This is a contradiction, so no such c exists, and f (A)c ∩ f (B)c = ∅.
So f (A)c and f (B)c are disjoint open sets with p ∈ f (A)c and q ∈ f (B)c . As such open
sets exist for all p, q ∈ Y , Y is Hausdorff.
Therefore Y is compact and Hausdorff, as desired. 
52 3. DISTINGUISHING SPACES

4. Connectedness

4.1. Connected Sets and Spaces.

Definition. (X, FX ) is a topological space. Then X is disconnected if X has a proper,
nonempty, clopen subset. Otherwise, X is connected.
Equivalently, X is disconnected if there exist nonempty, disjoint A, B ∈ FX such that
A ∪ B = X. A and B are called a separation of X.

Recall from Math 131 that A ⊆ R is connected iff A is an interval.

Theorem 19. Let f : (X, FX ) → (Y, FY ) be continuous. Then X is connected only if f (X)
is connected.

Proof. Suppose that U, V ∈ FY are separating sets of f (X). Then, as f is continuous,

f −1 (U ) and f −1 (V ) are open. As U and V cover f (X), f −1 (U ) and f −1 (V ) cover X.
Furthermore, suppose that x ∈ f −1 (U ) ∩ f −1 (V ). Then f (x) ∈ U ∩ V , and f (x) ∈ f (X).
However, this contradicts the assumption that U and V are separating sets. So f −1 (U )
and f −1 (V ) are disjoint open sets that cover X, so they are a separating set. But X is
connected, so this is a contradiction. Hence, no such U and V exist, and f (X) is also
connected. 

Are the following connected?

• R with the finite complement topology - Yes. Let U be some proper, nonempty
open subset of R. Then it is infinite, so its complement does not have finite
complement. Therefore, its complement is not open. So there are no proper,
nonempty clopen sets, and the topology is connected.
• R with the half-open interval
S S Let U = [0, ∞) and V = (−∞, 0).
topology - No.
These are open as U = n∈N [0, n) and V = n∈N [−n, 0), and the arbitrary union
of basis elements is open. Furthermore, they are clearly disjoint, and cover R.
We can also define (dis)connectedness for subsets of a topological space.
Definition. (X, FX ) is a topological space. S ⊆ X is disconnected if and only if there
exist U, V ∈ FX such that (U ∩ S) ∩ (V ∩ S) = ∅, U ∩ S, V ∩ S 6= ∅ and S ⊆ U ∪ V . Note
that this does not require U ∩ V = ∅.
Small Fact. (X, FX ) is connected if and only if ∀ f : X → Y where Y has the discrete
topology and f is continuous, then f constant.

Proof. (⇒) Suppose (X, Fx ) is connected and that ∃ f : X → Y where Y has

the discrete topology, such that f is continuous and not constant. So there exist
p, q ∈ Y such that p 6= q and f −1 ({p}), f −1 ({q}) 6= ∅. Therefore, U = f −1 ({p})
and V = f −1 ({p}c ) ⊇ f −1 ({q}) have U, V 6= ∅. As every set is open in the
 4. CONNECTEDNESS 53

discrete topology and f is continuous, U, V ∈ FX . Furthermore, as {p}∩{p}c = ∅,

U ∩ V = ∅, and U ∪ V = f −1 (Y ) = X. So U and V separate X. However, X is
connected, so this is a contradiction, and no such f exists.
(⇐) Suppose that f : X → Y is constant for all continuous f where Y has the discrete
topology and that X is disconnected. Then there exist separating sets A, B ∈ FX .
Define Y = {0, 1} with the discrete topology and f : X → Y as

0 x∈A
f (x) =
1 x∈B
Then, as f −1 ({0}) = A ∈ FX , f −1 ({1}) = B ∈ FX , f −1 (∅) = ∅ ∈ FX and
f −1 (Y ) = X ∈ FX , f −1 (U ) ∈ FX for all U ⊆ Y such that U is open. So f is
continuous. Furthermore, as A, B 6= ∅, there exist a ∈ A and b ∈ B, so f (a) = 0
and f (b) = 1. Hence f is not constant. Therefore f : X → Y is a continuous
function from X to a space with the discrete topology that is not constant. This
contradicts the initial assumption, so X is connected.


We now have an Important Lemma about Flowers.

[
Lemma. Suppose {yj | j ∈ J} is a collection of connected subspaces of X = yj , and
j∈J
\
yj 6= ∅. Then X is connected.
j∈J

Proof. Suppose that X is disconnected. Then X has a separation, U, V .

Let j0 ∈ J. We know that

• (U ∩ yj0 ) ∪ (V ∩ yj0 ) = yj0 ,
54 3. DISTINGUISHING SPACES

• (U ∩ yj0 ) ∩ (V ∩ yj0 ) = ∅,
• (U ∩ yj0 ) and (V ∩ yj0 ) are open in yj0 ,
• yj0 is connected, so either U ∩ yj0 = ∅ or V ∩ yj0 = ∅.
Without loss of generality, say U ∩ yj0 = ∅. Then V ∩ yj0 = yj0 . By the same argument,
∀j ∈ J, yj ⊆ U or yj ⊆ V .
\ \
But yj 6= ∅, so there exists p ∈ yj such that p ∈ yj0 ⊆ V . Thus ∀j ∈ J, yj ⊆ V .
j∈J j∈J
[
Since X = yj ⊆ V , U, V is not a separation of X. Thus, X is connected. 
j∈J

Theorem 20. Let (X, FX ) and (Y, FY ) be topological spaces. Then (X × Y, FX×Y ) is
connected if and only if both X and Y are connected.

Proof. (⇒) πX : X × Y −→ X and πY : X × Y −→ Y are continuous surjections.

Continuous functions preserve connectedness, so X and Y are connected.
(⇐) First, a picture proof.

The axes are connected, because X and Y are connected. If we make a copy of
the axes, shifted over a little bit, that copy is connected too. The union of these
shifted axis crosses will be the entirety of X × Y , and their intersection will be
nonempty, so X × Y is connected.
Now for the actual proof. Let (x0 , y0 ) ∈ X ×Y. Since {x0 }×Y ∼
= Y , it is connected
(as connectedness is a top. prop.). Similarly, {y0 } × X is connected.
Observe that (x0 , y0 ) ∈ ({x0 }×Y )∩({y0 }×X). That is, ({x0 }×Y )∩({y0 }×X) 6= ∅.
Then (x0 , y0 ) ∈ ({x0 } × Y ) ∪ ({y0 } × X) is connected by the Important Flower
Lemma. By the same process, we see that ∀x ∈ X, (X × {y0 }) ∪ ({x} × Y ) is
connected.
\
Note that (X × {y0 }) ⊆ (X × {y0 }) ∪ ({x} × Y ).
x∈X
 4. CONNECTEDNESS 55
[
Claim: (X × {y0 }) ∪ ({x} × Y ) = X × Y .
x∈X

Proof of claim:
[
(⊆) Clearly (X × {y0 }) ∪ ({x} × Y ) ⊆ X × Y because X × Y is the entire
x∈X
space.
(⊇) Let (x1 , y1 ) ∈ X × Y . [
Then (x1 , y1 ) ∈ ({x1 } × Y ). Thus, ∀x ∈ X and y ∈ Y ,
(x, y) ∈ ({x} × Y ) ⊆ (X × {y0 }) ∪ ({x} × Y ).
x∈X
\
Since (X×{y0 })∪({x}×Y ) is connected ∀x ∈ X, and (X × {y0 }) ∪ ({x} × Y ) 6=
x∈X
∅, X × Y is connected by the Important Flower Lemma.


4.2. Connected Components.

Definition. Let (X, FX ) be a topological space and let p ∈ X. Let {Cj | j ∈ J} be the set
of all connected subspaces of X containing p. Then ∪j∈J Cj is said to be the connected
component, Cp , of p.

And now, a series of tiny facts about connected components.

Small Fact. Let (X, FX ) be a topological space. Then,
(1) ∀p ∈ X, Cp is connected.
(2) If Cp and Cq are connected components, then either Cp ∩ Cq = ∅ or Cp = Cq .
(That is, connected components partition a space.)

Proof. (1) p ∈ ∩j∈J Cj , so Cp = ∪j∈J Cj is connected by the lemma above.

(2) Suppose Cp ∩ Cq 6= ∅ and let x ∈ Cp ∩ Cq . Then Cp ∪ Cq is connected by the
lemma above.
p ∈ Cp ∪ Cq , so Cp ∪ Cq ∈ {Cj | j ∈ J}. Also, Cp ∪ Cq ⊆ Cp , since Cp = ∪j∈J Cj .
So, Cq ⊆ Cp . Similarly, Cp ⊆ Cq .
Thus Cp = Cq , so we can say X is partitioned by its connected components. 
Example. Let X = R2 with the dictionary topology. Connected components are vertical
lines. (Proof is an exercise.) Define x ∼ y (x, y ∈ R2 ) if x and y are in the same connected
component. Then X/ ∼ = R with the discrete topology.
56 3. DISTINGUISHING SPACES

X X/~

Connectedness is not always intuitive, though. Here’s a large example showing this.
Define the Comb, the Flea, and the space X as follows.
∞
[
Comb: yn in R2 where y0 = [0, 1] × {0} and ∀n ∈ N, yn = {1/n} × [0, 1].
n=0

Flea: {(0, 1)}.

X: Flea ∪ Comb with the subspace topology.
Example. X is connected.

Proof. First, we claim that the Comb is connected.

Proof:
∀n ∈ {0} ∪ N, yn is connected because yn ∼ = [0, 1]. Also, ∀n ∈ N, y0 ∩ yn 6= ∅ (because
(1/n, 0) is in both), so y0 ∪ yn is connected.
\ ∞
[ ∞
[
Thus y0 ∪ yn 6= ∅, so y0 ∪ yn = yn and so the Comb is connected by the Impor-
n∈N n=0 n=0
tant Flower Lemma.

Now we want to show that X is connected. Suppose U , V are a separation of X. Since

the Comb is connected, we can say, WLOG, that Comb ⊆ U . Then it must be that Flea
⊆ V because U , V are a separation. V is open in X, so ∃ > 0 s.t. B ((0, 1); X) ⊆ V . Let
n ∈ N s.t. n > 1 . Then d(( n1 , 1), (0, 1)) = n1 <  ⇒ (1/n, 1) ∈ B ((0, 1); X) ⊆ V . Observe
that since Yn = {1/n} × [0, 1], (1/n, 1) ∈ Yn . Thus, (1/n, 1) ∈ V ∩ Yn ⊆ V ∩ Y ⊆ V ∩ U ⇒
V ∩ U 6= ∅. But this is a contradiction, since we assumed that U , V was a separation of
X. Thus, X is connected. 
 5. PATH-CONNECTEDNESS 57

5. Path-Connectedness

Definition. Let (X, FX ) be a topological space and let f : [0, 1] → X be continuous.

Then we say that f is a path from f (0) to f (1) (Note: it is handy to think of t ∈ [0, 1] as
time).
Definition. Let (X, FX ) be a topological space. If ∀p, q ∈ X, there exists a path in X
from p to q then we say that X is path-connected.
Example. Rn is path-connected:
Let a, b ∈ Rn be given. Let f : I → Rn be f (t) = (1 − t)a + bt. Then f (0) = a, and
f (1) = b. f can be shown to be continuous by performing an involved -δ proof.

Some remarks:
(1) Connected is a negative definition and path-connected is a positive definition (for
connectedness, we are trying to show that a separation doesn’t exist, so it is
easiest to do connectedness proofs by contradiction. For path-connectedness, we
are trying to show that a path exists, so path-connectedness proofs are more easily
done constructively).
(2) In general, it is easier to prove that a space is disconnected than to prove that it
is not path-connected.
(3) In general, it is easier to prove that a space is path-connected than to prove that
it is connected.
Theorem 21. If (X, FX ) is path-connected, then (X, FX ) is connected.

Proof. Suppose there exists a separation U , V of X. Since U , V is a separation of

X, U 6= ∅, V 6= ∅. Let p ∈ U , q ∈ V be given. Since X is path-connected, ∃ a path f from
p to q. Since paths are continuous by definition, f is continuous, so f −1 (U ), f −1 (V ) are
open in [0, 1].
Claim 1: f −1 (U ) ∪ f −1 (V ) = [0, 1].
Proof of Claim 1: Let x ∈ [0, 1] be given. Then f (x) ∈ X = U ∪ V ⇒ f (x) ∈ U
or f (x) ∈ V ⇒ x ∈ f −1 (U ) or x ∈ f −1 (V ) ⇒ x ∈ f −1 (U ) ∪ f −1 (V ). Thus, [0, 1] ⊆
f −1 (U ) ∪ f −1 (V ) ⇒ [0, 1] = f −1 (U ) ∪ f −1 (V ) (since f −1 (U ), f −1 (V ) ⊆ [0, 1]).
Claim 2: f −1 (U ) ∩ f −1 (V ) = ∅.
proof of Claim 2: Suppose ∃x ∈ f −1 (U ) ∩ f −1 (V ). Then x ∈ f −1 (U ) ⇒ f (x) ∈ U , and
x ∈ f −1 (V ) ⇒ f (x) ∈ V , so f (x) ∈ U ∩ V , which is impossible since we are assuming that
U and V are a separation of X. Thus, f −1 (U ) ∩ f −1 (V ) = ∅.
Observe that since f is a path from p to q, f (0) = p and f (1) = q, so 0 ∈ f −1 (U ), 1 ∈
f −1 (V ) so f −1 (U ) and f −1 (V ) are non-empty and proper. Thus, since f −1 (U )∩f −1 (V ) = ∅
and f −1 (U ) ∪ f −1 (V ) = [0, 1], f −1 (U ) and f −1 (V ) form a separation of [0, 1]. But this is a
58 3. DISTINGUISHING SPACES

contradiction, since we know from Math 131 that [0, 1] is connected. Thus, (X, FX ) must
be connected. 

Recall the example of the flea and the comb. We showed that the flea and comb is con-
nected.
Theorem 22. The flea and comb is not path-connected.

Proof. We want to show that @ a path from the flea to (0, 0).
Suppose ∃ a path f from the flea to (0, 0). Let p = flea. Then f −1 ({p}) is not empty
because f (0) = p by the definition of f . Similarly, by the definition of f , f (1) = (0, 0) 6= p,
so f −1 ({p}) is proper as well. Observe that since {p} is closed in X and f is continuous,
f −1 ({p}) is closed.
We’d like to show that f −1 ({p}) is open, since then it would be a nontrivial clopen subset
of the flea and the comb. This implies that the set is not connected, a contradiction.
Let y ∈ f −1 ({p}) be given.
Observe that B 1 (p; X) is open in X. Since f is a path, f is continuous, so f −1 (B 1 (p; X))
2 2
is open in [0, 1]. Since y ∈ f −1 ({p}), f (y) = p ∈ B 1 (p; X) ⇒ y ∈ f −1 (B 1 (p; X)). Thus,
2 2
since f −1 (B 1 (p; X)) is open in [0, 1], ∃ > 0 s.t. B (y; [0, 1]) ⊆ f −1 (B 1 (p; X)).
2 2

Let z ∈ B (y; [0, 1]). If f (z) = p, then B (y; [0, 1]) ⊆ f −1 (p) and so f −1 (p) is open,
completing the proof.
Suppose f (z) 6= p. Since z ∈ B (y; [0, 1]), z ∈ f −1 (B 1 (p; X)), so f (z) ∈ B 1 (p; X) (so
2 2
d(f (z), p) < 12 ). Since Y0 = [0, 1] × {0}, for each q ∈ Y0 , d(p, q) ≥ 1 ⇒ q 6∈ B 1 (p; X). Thus,
2
f (z) 6∈ Y0 . Thus since f (z) 6= p, ∃n ∈ N s.t. f (z) ∈ Yn .
There exists r ∈ R \ Q such that 0 < r < n1 . Define sets A, B ⊆ f (B (y; [0, 1])), as
A = {(x, y) ∈ f (B (y; [0, 1])) |x < r} and B = {(x, y) ∈ f (B (y; [0, 1])) |x > r}
We next show that A and B are a separation for f (B (y; [0, 1])). To this end, we’ll need
to show that A and B are disjoint, proper and non-empty, that their union is the entire
set, and that they’re are both clopen.
We note that A ∩ B = ∅ by definition. We know that f (z) ∈ B, and p ∈ A, so neither set
will be empty and both will be proper.
Next, we want to show that A ∪ B = f (B (y; [0, 1])). Certainly A ∪ B ⊆ f (B (y; [0, 1])).
1
Now let (x, f (x)) ∈ f (B (y; [0, 1])). We know that f (x) 6= 0, and in fact x = m for some
m ∈ N. Then x 6= r, so x ∈ A ∪ B and we’re done.
Next we want to show that A is open in f (B (y; [0, 1])). We know that {(x, y)|x < r} is
open in R2 , the half plane to the left of r. It follows using the subspace topology that
 5. PATH-CONNECTEDNESS 59

A = {(x, y)|x < r} ∩ f (B (y; [0, 1])) is open in f (B (y; [0, 1])). Similarly, B is open in
f (B (y; [0, 1])). Since each is the other’s complement in f (B (y; [0, 1])), then both are also
closed.
We’ve hence shown that A and B are a separation of f (B (y; [0, 1])). This is a contradiction,
since B (y; [0, 1]) is connected and f is continuous, and thus we’ve disconnected B (y; [0, 1]).
We conclude that f (z) = p, for all such z ∈ B (y; [0, 1]).
Therefore B (y; [0, 1]) ⊆ f −1 ({p}), so f −1 ({p}) is open. So f −1 ({p}) is a clopen, non-
empty, proper subset of [0, 1]. This is a contradiction, so we conclude that there does not
exist a path in X from p to (0, 0) (or anywhere). 

The point of the whole example is to show that path-connected is stronger than connected.
We knew already that path-connected implies connected, and now we see that connected
doesn’t necessarily imply path-connected.

5.1. Combining paths. It turns out that it’s useful to have a way to combine paths.
Definition. Let f, g be paths in a topological space (X, FX ) such that f (1) = g(0). Then
we define f ∗ g : I → X by
(
f (2t) 0 ≤ t ≤ 21
(f ∗ g)(t) =
g(2t − 1) 12 ≤ t ≤ 1

Intuitively what’s happening is that we’re connecting two paths while speeding things
up, creating a new single path parametrized from 0 to 1 from two paths which were
parametrized from 0 to 1.
Small Fact. f ∗ g is a path from f (0) to g(1).

Proof. By the Pasting Lemma1, since [0, 12 ] and [ 12 , 1] are closed subsets under [0, 1],
and f (2( 12 )) = f (1) = g(0) = g(2( 12 ) − 1), f ∗ g is continuous. So f ∗ g is a path. Since
(f ∗ g)(0) = f (0) and (f ∗ g)(1) = g(1), then f ∗ g is a path from f (0) to f (1). 

This definition also allows us to create an analogue to the flower lemma.

Theorem 23 (Flower Lemma for Path-connected). Let X = ∪i∈I Yi such that ∀i ∈ I, Yi is
path-connected, and ∩i∈I Yi 6= ∅. Then X is path-connected.

Proof. Let a, b ∈ X. If ∃n ∈ I such that a, b ∈ Yn , then there exists a path from a to

b in Yn ⊆ X. So without loss of generality, suppose that a ∈ Yn , b ∈ Ym , and m 6= n. Let
x ∈ ∩i∈I Yi . Then there exists a path f from a to x in Yn , and a path g from x to b in Ym .
So f ∗ g is a path in X from a → b, and we’re done. 
1HW3 #1, which states that if we have two continuous functions with closed sets as domains, and they
agree over the intersection of these domains, then the combined function is continuous
60 3. DISTINGUISHING SPACES

Corollary. The product of path-connected spaces is path-connected.

Proof. The proof is identical to the connected one; the key is the flower lemma. 
Definition. Let (X, FX ) be a topological space, and p ∈ X. Let {Cj |j ∈ J} be the
set of all path-connected subspaces of X containing p. Then ∪j∈J Cj is said to be the
path-connected component Cp .

Some tiny facts about path-connected components:

Let (X, FX ) be a topological space. Then
(1) ∀p ∈ X, Cp is path-connected
(2) If Cp , Cq are path-connected components, then either Cp ∩ Cq = ∅, or Cp = Cq .
In other words, path-connected components partition the set.
We’ll omit this proof, as it is identical to the one presented on connectedness.

6. Homotopies

We now have enough background to start talking about geometric (or algebraic) topology.

Definition. Let f be a path in (X, FX ), and define f : I → X by f (t) = f (1 − t).

A few remarks:
(1) f is a path, because it is a composition of continuous functions.
(2) f is a path from f (1) to f (0).
(3) f ∗ f 6= the “identity”. We haven’t created an inverse.
Definition. Let (X, FX ) be a topological space, and a ∈ X. We define ea : I → X by
ea (t) = a for all t ∈ I.

However, note that f ∗ ea 6= f . This isn’t an identity.

Definition. Let (X, FX ) and (Y, FY ) be top spaces, and let f0 : X → Y and f1 : X → Y
be continuous. Then we say that f0 is homotopic to f1 if there exists a function F :
X × I → Y , continuous, such that F (x, 0) = f0 (x), and F (x, 1) = f1 (x). We say that F is
a homotopy from f0 to f1 , and we write f0 ' f1 .

This is probably the most important concept in the rest of these notes.
 6. HOMOTOPIES 61

To illustrate homotopies, a couple of pictures.

f0 f1

It is visually apparent that one image can be warped to make the other, so f0 and f1 are
homotopic. However, homotopic functions need not be homeomorphic.

These are homotopic, for example:

f0 f2

Example. Let X = I and Y = R2 . Define f0 : I → R2 by f0 (x) = (x, 0), f1 : I → R2 by

f1 (x) = (x, x2 ), and F : I × I → R2 by F (x, t) = t(x, x2 ) + (1 − t)(x, 0).

Observe that F is continuous. Note also that F (x, 0) = (x, 0) = f0 (x) and F (x, 1) =
(x, x2 ) = f1 (x).

Thus F is a homotopy.
62 3. DISTINGUISHING SPACES

6.1. Drawing Homotopies. How to draw a homotopy:

f1(1)
f1(0) f1(s) f1(1)
t=1
f1(s)

f1(0)
F(s0,t)
f0(1)
F(s0,t)

s0
f0(s)

t=0 s0 f0(1)
f0(0) f0(s) f0(0)

The image of a vertical segment in the square is the path taken by the x coordinate of that
segment during the homotopy.

A few remarks.

(1) Suppose Y is not path connected and f0 (x) and f1 (x) are in different path com-
ponents. Then there does not exist a homotopy from f0 to f1 .

f1 f1

F(0,t)
f0 f0

If f0 and f1 were homotopic, then F (0, t) would be a path from f0 (0) to f1 (0).
But f0 and f1 are in different path components.

(2) If Y is path connected and f0 and f1 are paths in Y , then f0 ' f1 . Homotope (the
verb!) f0 to its initial point, move it to the initial point of f1 and then stretch it
 6. HOMOTOPIES 63

back out into f1 .)

f0

f1 f1 f1 f1

Example. Let Y = R2 − {( 12 , 31 )}. Define f0 : I → Y by f0 (s) = (s, s2 ), and f1 : I → Y

by f1 (s) = (s, s). Because there’s a hole in Y , we won’t be able to just bend f0 over to f1 .
Define G : I × I → Y by G(s, t) = f0 ((1 − t)s). Observe that G is continuous, and
G(s, 0) = f0 (s) and G(s, 1) = f0 (0).
Now define H : I × I → Y by H(s, t) = f1 (ts). Observe that H is continuous and
H(s, 0) = f1 (0) and H(s, 1) = f1 (s).
Finally, define F : I × I → Y by
: t ∈ [0, 21 ]

G(s, 2t)
F (s, t) = .
H(s, 2t − 1) : t ∈ [ 12 , 1]

F is continuous: Let A = I × [0, 21 ] and B = I × [ 21 , 1]. Both are closed in I × I.

A ∩ B = I × { 12 }. G(s, 2( 12 )) = G(s, 1) = f0 (0) = (0, 0) and H(s, 2( 21 ) − 1) =
H(s, 0) = f1 (0) = (0, 0).
Since G and H are continuous and agree at A ∩ B, by the Pasting Lemma, F is
continuous.
F is a homotopy :
F (s, 0) = G(s, 0) = f0 (s) and F (s, 1) = H(s, 1) = f1 (s)

Thus F is indeed an homotopy, and f0 is homotopic to f1 !

What path does (1, 1) take during the aforementioned homotopy?

- Informally put, it moves down along f0 and climbs back up f1 to its old position.
More formally,
( (
G(1, 2t) t ∈ [0, 21 ] f0 (1 − 2t) t ∈ [0, 12 ]
F (1, t) = 1 = = f0 ∗ f1 .
H(1, 2t − 1) t ∈ [ 2 , 1] f1 (2t − 1) t ∈ [ 21 , 1]
64 3. DISTINGUISHING SPACES

6.2. Path Homotopies.

Definition. Let f0 and f1 be paths in (X, FX ) from a to b. We say f0 is path-homotopic
if there exists a homotopy F from f0 to f1 s.t. ∀t ∈ I, F (0, t) = a and F (1, t) = b. We say
F is a path homotopy and write f0 ∼ f1 .

F(s,0) b

a F(s,1)

Example. Let X be a convex region of Rn , let a, b ∈ X, and let f0 and f1 be paths in X

from a to b.

Claim: f0 ∼ f1 .

Proof. Let F (s, t) = (1 − t)f0 (s) + tf1 (s) (Note: we call this the straight line
homotopy). Observe that ∀s ∈ I, F (s, 0) = f0 (s) and F (s, 1) = f1 (s), and that F is
continuous, so F is a homotopy from f0 to f1 . Now let t ∈ I be given. Observe that
F (0, t) = (1 − t)f0 (0) + tf1 (0) = f0 (0) = a, and that F (1, t) = f0 (1) = b. Thus ∀t ∈ I,
F (0, t) = a and F (1, t) = b, so F is a path homotopy, and thus f0 ∼ f1 . 
Example. Let X ∼ 2
= D , let a, b ∈ X, and let f0 and f1 be paths in X from a to b. Then
f0 ∼ f1 .

Proof. Let g : X → D2 be a homeomorphism. Let F : (I × I) → D2 be the straight

line homotopy in D2 from g ◦ f0 to g ◦ f1 (Note: D2 is a convex region of R2 , so by the last
example we can use the straight line homotopy here).
First, we need to show that g −1 ◦ F is continuous. Note that F is continuous since F is a
homotopy. Note also that since g is a homeomorphism, g −1 is continuous. Thus g −1 ◦ F is
the composition of continuous functions, and hence g −1 ◦ F is continuous.
 6. HOMOTOPIES 65

Now we need to show that g −1 ◦ F is a homotopy from f0 to f1 .

First, observe that ∀s ∈ I, (g −1 ◦F )(s, 0) = g −1 ((1−0)g(f0 (s))+(0)g(f1 (s))) = g −1 (g(f0 (s))) =
f0 (s) since g is a bijection, and similarly (g −1 ◦ F )(s, 1) = f1 (s). Thus, since g −1 ◦ F is
continuous, g −1 ◦ F is a homotopy from f0 to f1 .
Lastly, we need to show that g −1 ◦ F is a path homotopy.
Note that ∀t ∈ I, (g −1 ◦ F )(0, t) = g −1 ((1 − t)g(f0 (0)) + tg(f1 (0))) = g −1 ((1 − t)g(a) +
tg(a)) = g −1 (g(a)) = a since g is a bijection. Similarly, ∀t ∈ I, (g −1 ◦ F )(1, t) = b. Thus,
since g −1 ◦F is a homotopy from f0 to f1 , g −1 ◦F is a path homotopy, and thus f0 ∼ f1 . 
Theorem 24. Let (X, FX ) be a topological space and let a, b ∈ X be given. Then ∼ is an
equivalence relation of paths in X from a to b.

Proof. In order to show ∼ is an equivalence relation, we need to show that ∼ is

reflexive, symmetric, and transitive.
Reflexive: If f is a path in X from a to b, let F : (I ×I) → X be given by F (s, t) = f (s). Note
that F is a homotopy from f to f since, ∀s ∈ I, F (s, 0) = f (s) and F (s, 1) = f (s)
and F is continuous since f is continuous. Observe that ∀t ∈ I, F (0, t) = f (0) = a
and F (1, t) = f (1) = b, so F is a path homotopy and f ∼ f . Thus, ∼ is reflexive.
Symmetric: Suppose that f1 ∼ f2 . Then there exists a path homotopy F from f1 to f2 . Define
F 0 : (I × I) → X given by F 0 (s, t) = F (s, 1 − t), ∀(s, t) ∈ (I × I). Note that F 0 is
continuous since F is continuous so F 0 is a composition of continuous functions.
Recall that F is a homotopy from f1 to f2 , and thus F 0 is a homotopy from f2 to
f1 since, ∀s ∈ I, F 0 (s, 0) = F (s, 1) = f2 (s) and F 0 (s, 1) = F (s, 0) = f1 (s). Now
observe that ∀t ∈ I, F 0 (0, t) = F (0, 1 − t) = a and F 0 (1, t) = F (1, 1 − t) = b, since
F is a path homotopy. Thus, F 0 is a path homotopy, and thus f2 ∼ f1 , so ∼ is
symmetric.
Transitive: Suppose that f1 , f2 , and f3 are paths in X from a to b s.t. f1 ∼ f2 and f2 ∼ f3 .
Since f1 ∼ f2 , there exists a path homotopy F1 from f1 to f2 , and since f2 ∼ f3 ,
there exists a path homotopy F2 from f2 to f3 . Define F3 : (I × I) → X by
(
F1 (s, 2t) t ∈ [0, 12 ]
F3 (s, t) =
F2 (s, 2t − 1) t ∈ [ 21 , 1]

Now we must show that F3 is a path homotopy from f1 to f3 . Observe that

A = I × [0, 12 ] and B = I × [ 21 , 1] are closed in I × I, and F1 and F2 are continuous,
so if F1 (s, t) = F2 (s, t) ∀(s, t) ∈ A∩B, then by the pasting lemma F3 is continuous.
Since A∩B = I ×{ 21 }, and, ∀s ∈ I, F1 (s, 2( 12 )) = F1 (s, 1) = f2 (s) and F2 (s, 2( 21 )−
1) = F2 (s, 0) = f2 (s), F3 is continuous by the pasting lemma. Now, observe that
∀s ∈ I, F3 (s, 0) = F1 (s, 2(0)) = F1 (s, 0) = f1 (s) and F3 (s, 1) = F2 (s, 2(1) − 1) =
66 3. DISTINGUISHING SPACES

F2 (s, 1) = f3 (s), so F3 is a homotopy from f1 to f3 . Now, observe that ∀t ∈ I,

(
F1 (0, 2t) if t ∈ [0, 21 ]
F3 (0, t) =
F2 (0, 2t − 1) if t ∈ [ 21 , 1]
(
a t ∈ [0, 12 ]
=
a t ∈ [ 21 , 1]

(since F1 and F2 are path homotopies), and thus F3 (0, t) = a, ∀t ∈ I. Similarly,

∀t ∈ I, F3 (1, t) = b. Thus, F3 is a path homotopy from f1 to f3 , so f1 ∼ f3 , and
thus ∼ is transitive.

Thus, ∼ is an equivalence relation of paths in X from a to b. 

Note that the same proof (using only the parts related to continuity and homotopy) works
for '.

Definition. Let (X, FX ) be a topological space and let a, b ∈ X be given. For each path
f from a to b in X, define [f ] to be the path homotopy class of f.

Definition. Let f be a path in X from a to b and g be a path in X from b to c. Define

an invisible symbol by [f ][g] = [f ∗ g].

Some remarks:

(1) [f ], [g], and [f ∗ g] are not elements of the same quotient ‘world’ unless a = b = c.

(2) We have to prove that invisible multiplication is well-defined, i.e. if f ∼ f 0 and

g ∼ g 0 , then we want [f ∗ g] = [f 0 ∗ g 0 ] because [f ] = [f 0 ] and [g] = [g 0 ].

Lemma (Important). The product of path homotopy classes is well-defined. Formally, let
(X, Fx ) be a topological space. Let f, f 0 be paths in X from a to b and let g, g 0 be paths in
X from b to c. Then:

f ∗ g ∼ f 0 ∗ g 0 ⇒ [f ][g] = [f 0 ][g 0 ]

Proof. We know f ∗ g and f 0 ∗ g 0 are paths in X from a to c by a previous result. Let

F be the path homotopy from f to f 0 , and let G be the path homotopy from g to g 0 .
 6. HOMOTOPIES 67

At this point, it may be wise to draw a homotopy diagram:

a f’ b g’ c

F G

a b g c
f

Define H : I × I → X by:
(
F (2s, t) s ∈ [0, 1/2]
H(s, t) =
G((2s − 1), t) s ∈ [1/2, 1]
We claim H is a homotopy from f ∗ g to f 0 ∗ g 0 . We will now verify the claim.
We see that:
F (2 · (1/2), t) = F (1, t) = b G(2 · (1/2) − 1, t) = G(0, t) = b
so by the pasting lemma, it follows that H is continuous.
We now show H gives the desired paths by calculating:
H(0, t) = F (0, t) = a H(1, t) = G(1, t) = c
so H(s, [0, 1]) gives a set of paths from a to c.
Finally, we need to prove that H gives a homotopy between the intended paths:
( (
F (2s, 0) s ∈ [0, 1/2] f (2s) s ∈ [0, 1/2]
H(s, 0) = = =f ∗g
G(2s − 1, 0) s ∈ [1/2, 1] g(2s − 1) s ∈ [1/2, 1]
Similarly:
( (
F (2s, 1) s ∈ [0, 1/2] f 0 (2s) s ∈ [0, 1/2]
H(s, 1) = = 0
= f 0 ∗ g0
G(2s − 1, 1) s ∈ [1/2, 1] g (2s − 1) s ∈ [1/2, 1]

Consequently, H satisfies the definition of a path homotopy from f ∗ g to f 0 ∗ g 0 . We then

have that f ∗ g ∼ f 0 ∗ g 0 so that by definition of the product of paths:
[f ][g] = [f ∗ g] = [f 0 ∗ g 0 ] = [f 0 ][g 0 ]
68 3. DISTINGUISHING SPACES

as desired. 

7. Loops and the Fundamental Group

We have shown that the product of two path homotopy classes is well defined. For the
purposes of defining a group of path homotopy classes, we would like a set of paths whose
products have the same endpoints as the original paths. This simplification motivates the
two definitions which follow:
Definition. Let f be a path in X such that f (0) = f (1) = x0 ∈ X. Then f is a loop in
X based at x0 .

Note that if f, g are loops in X based at some point x0 ∈ X, then their product f ∗ g is
also a loop based at x0 . In particular, we then have that [f ], [g] and [f ][g] = [f ∗ g] are all
path-homotopic classes of loops based at x0 .
Definition. Let (X, Fx ) be a topological space, and let x0 ∈ X.
Define π1 (X, x0 ) as the set of path homotopy classes of loops based at x0 endowed with
the product of path homotopy classes. We call π1 (X, x0 ) the fundamental group of X
based at x0 .

7.1. The Fundamental Group is a Group. Our ultimate goal is to harness the
power of group theory from abstract algebra to study topological spaces. We defined the
fundamental group of X based at x0 with the path homotopy class product operation.
Naturally, we would like to prove that π1 (X, x0 ) endowed with the product operation is
actually a group. In other words, if (X, Fx ) is a topological space with x0 ∈ X, we must
prove the following:
(1) π1 (X, x0 ) is closed under the operation [][]. In other words, for [f ], [g] ∈ π1 (X, x0 ),
[f ][g] ∈ π1 (X, x0 ).
(2) The operation [][] is associative. In other words, given [f ], [g], [h] ∈ π1 (X, x0 ):
([f ][g])[h] = [f ]([g][h])

(3) π1 (X, x0 ) contains an identity element. In other words, there exists [e] ∈ π1 (X, x0 )
such that for all [f ] ∈ π1 (X, x0 ):
[e][f ] = [f ][e] = [f ]

(4) Every element of π1 (X, x0 ) has an inverse. In other words, given f ∈ π1 (X, x0 ),
there exists g ∈ π1 (X, x0 ) such that:
[f ][g] = [g][f ] = [e]

If π1 (X, x0 ) satisfies all of these requirements, π1 (X, x0 ) is a group.

 7. LOOPS AND THE FUNDAMENTAL GROUP 69

Lemma (Closure). Let (X, Fx ) be a topological space, and let x0 ∈ X. Let [f ], [g] ∈
π1 (X, x0 ). Then:
[f ][g] ∈ π1 (X, x0 )

Proof. We see that:

[f ][g] = [f ∗ g]
We know by a previous result that f ∗ g is a path in X from x0 to x0 since x0 is both the
starting point of [f ] and the endpoint of [g]. Consequently, f ∗ g is a loop in X based at
x0 , so [f ∗ g] ∈ π1 (X, x0 ). 

Next, we will show that path homotopy path products of loops based at a point are
associative.
Lemma (Associativity). Let (X, Fx ) be a topological space, and let x0 ∈ X. Let [f ], [g], [h] ∈
π1 (X, x0 ). Then:
([f ][g])[h] = [f ]([g][h])

Proof. Before actually proving the result, we will explicitly write the formulas for
(f ∗ g) ∗ h and f ∗ (g ∗ h):
 
f (4s)
 s ∈ [0, 41 ] f (2s)
 s ∈ [0, 21 ]
(f ∗ g) ∗ h = g(4s − 1) s ∈ [ 14 , 12 ] f ∗ (g ∗ h) = g(4s − 2) s ∈ [ 21 , 34 ]
1
h(4s − 3) s ∈ [ 43 , 1]
 
h(2s − 1) s ∈ [ 2 , 1]
 

We need to construct a homotopy from (f ∗ g) ∗ h to f ∗ (h ∗ g).

f ½ g ¾ h

f ¼ g ½ h
70 3. DISTINGUISHING SPACES

Based on our formulas for (f ∗g)∗h and f ∗(g∗h), we parameterize the function F : I×I → X
by:
  
4s
s ∈ 0, 1+t
 
f 1+t

 4
 1+t 2+t 
F (s, t) = g(4s − 1 − t) s ∈ 4 , 4
  
h 4s − 2+t

s∈
 2+t
,1

2−t 2−t 4

We claim and will verify that F is a homotopy from (f ∗ g) ∗ h to f ∗ (g ∗ h). We first prove
continuity using the pasting lemma as usual. The proof that each of the piecewise parts of
F is defined on a closed set is left as a (relatively trivial but tedious) exercise.
We will, however, show that on the boundaries between these regions, the piecewise parts
agree. We see:
  
4 1+t
f · = f (1) = x0
1+t 4
   
1+t
g 4· −1−t = g(0) = x0
4
   
2+t
g 4· −1−t = g(1) = x0
4
   
4 2+t 2+t
h · − = h(0) = x0
2−t 4 2−t

so continuity follows by
 the pasting
 lemma. Note that we assume f, g, h are continuous
4s 2+t
so that things like h 2−t − 2−t are because compositions of continuous functions are
4s
continuous and over t ∈ I, 2−t − 2+t
2−t is continuous. The justification is similar for the other
piecewise parts in the definition of F .
We now verify that F gives a path from x0 to x0 for every fixed t ∈ I. We see that:

F (0, t) = f (0) = x0
F (1, t) = h(1) = x0

as desired.
Finally, we need to show that F provides a homotopy from (f ∗ g) ∗ h to f ∗ (g ∗ h). We
have that:

s ∈ 0, 1+t
 
f (4s)
 4
F (s, 0) = g(4s − 1) s ∈ 1+t 2+t
 
4 , 4
= (f ∗ g) ∗ h
2+t
 
h (2s − 1) s ∈ 4 , 1

 7. LOOPS AND THE FUNDAMENTAL GROUP 71

s ∈ 0, 1+t
 
f (2s)
 4
F (s, 1) = g(4s − 2) s ∈ 1+t , 2+t
 
4 4
= f ∗ (g ∗ h)
h (4s − 3) s ∈ 2+t
  
4 ,1


as desired, so F satisfies the three properties of a homotopy from (f ∗ g) ∗ h to f ∗ (g ∗ h).

We have constructed F (s, t) such that F is a homotopy from (f ∗ g) ∗ h to f ∗ (g ∗ h) which
implies that (f ∗ g) ∗ h ∼ f ∗ (g ∗ h) ⇒ [(f ∗ g) ∗ h] = [f ∗ (g ∗ h)]. It immediately follows
that
([f ][g])[h] = [f ]([g][h])

Lemma (Identity). Let f be a path in X which begins at x0 and ends at x1 . Then f ∗ex1 ∼ f
and ex0 ∗ f ∼ f . 2

Proof. We prove that f ∗ ex1 ∼ f . The other case is similar.

In order to define a homotopy, at t, we will “do” f for s ∈ [0, t(1) + (1 − t) 21 ] or, by
simplification, s ∈ [0, t+1 t+1
2 ] and ex1 (t) for s ∈ [ 2 , 1].

Define F : I × I → X by
(  
2s
s ∈ 0, t+1
 
f t+1 2
F (s, t) =
s ∈ t+1
 
ex1 2 ,1

This is:
Continuous: By the Pasting lemma, as f, ex1 are continuous on closed  domains it suffices to
t+1 
2
check that they agree for s = t+1
2 . This follows easily, as f
2
t+1 = f (1) = x1 ,
and ex1 t+1


2 = x1 .
A homotopy: (
f (2s) s ∈ 0, 12
 
F (s, 0) =  = f ∗ ex1
ex1 (t) s ∈ 21 , 1


and (
f (s) s ∈ [0, 1]
F (s, 1) =
e x1 s ∈ [1, 1] = f.

A path: F (0, t) = f (0) = x0 , and F (1, t) = ex1 = x1 , hence F is a path.

Thus f ∗ ex1 ∼ f . 

2Recall that ∼ denotes “path homotopy.”

 72 3. DISTINGUISHING SPACES

We’ve now shown that π1 (X, x0 ) is closed, has an identity, and the operation is associative,
so just showing that inverses exist proves that it’s a group.
Lemma (Inverses). Let f be a path in X from x0 to x1 . Then f ∗ f ∼ ex0 and f ∗ f ∼ ex0 .

The “wrong” approach: increase the speed over f and f and wait at x1 . The “right”
approach: travel successively smaller distances along f .

Proof. We show only that f ∗ f ∼ ex0 , and the other case follows similarly.
Define F : I × I → X by

s ∈ 0, 1−t
 
f (2s)

 1−t 2 1+t 
F (s, t) = f (1 − t) s∈ 2 , 2
f (2s − 1) s ∈ 1+t
  
, 1

2
This is:
Continuous: Using the pasting lemma, it suffices to check that these functions agree on their
(shared) endpoints. As f (2( 1−t 1+t
2 )) = f (1 − t), and f (2( 2 ) − 1) = f (t) = f (1 − t),
we conclude that F is continuous.
A homotopy: Let t = 0,

f (2s)
 s ∈ [0, 1/2]
F (s, 0) = f (1) s ∈ [1/2, 1/2] = f ∗ f (s)

f (2s − 1) s ∈ [1/2, 1]


For t = 1, 
f (2s)
 s ∈ [0, 0]
F (s, 1) = f (0) s ∈ [0, 1] = ex0 .

f (2s − 1) s ∈ [1, 1]


A path: For s = 0, F (0, t) = f (0) = x0 . And for s = 1, F (1, t) = f (1) = x0 .

For each loop f based at x0 , f is a loop based at x0 . Thus we have shown closure under
inverses. 

It follows that the action ∗ defines a group on the set of loops based at x0 .
Observe that the set of paths does not have a group structure, as there is no definition of
multiplication between arbitrary paths.
Example. Let X = Rn and x0 be a point in X. Then π1 (X, x0 ) = h[ex0 ]i.

The following theorem relates the fundamental group at a point within a path component
to the fundamental group of that point in the ambient space:
 7. LOOPS AND THE FUNDAMENTAL GROUP 73

Theorem 25. Let A be a path component of a topological space X, and let x0 ∈ A. Then:
π1 (A, x0 ) ∼
= π1 (X, x0 )
(Note that ∼
= denotes a group isomorphism and not a homeomorphism)

Before proving the theorem, we will cover two quick non-examples of cases where the
theorem could break down.
Recall the circle S 1 which we can consider as a subspace of R2 . Since R2 is a simply
connected space, the fundamental group at every point is trivial. On the other hand,
picking some point x0 ∈ S 1 , the loop around the circle cannot be deformed in S 1 to the
point x0 , so π1 (S 1 , x0 ) is non-trivial and hence not isomorphic to π1 (R2 , x0 ) if we embed
S 1 in R2 . The following picture illustrates S 1 with the point x0 :

r0
'$
x

&%

This may seem like a counterexample to the theorem, but in R2 , S 1 is not a distinct path
component of R2 , so the theorem does not apply. Intuitively, by embedding S 1 in R2 , the
interior of the circle is part of R2 , so we can deform a loop around S 1 based at x0 to the
trivial loop by “pulling” the loop through the middle of the circle, which we could not do
when S 1 was considered as a space in its own right.
Consider the following diagram:

r0
'$
x
By A
y

&%

where we have the space R2 with A, B removed, so loops based at x0 that run around A, B
cannot be deformed into the trivial loop. However, when we consider this space embedded
in R2 as a subspace, it is not an entire path component in R2 , so our theorem does not
apply.
We now prove the theorem:
74 3. DISTINGUISHING SPACES

Proof. Define ϕ : π1 (A, x0 ) → π1 (X, x0 ) by:

ϕ([f ]A ) = [f ]X
We claim that ϕ is a group isomorphism, i.e. that ϕ is a bijection and a group homo-
morphism. In other words, we need to show that ϕ is injective, surjective and that if
a, b ∈ π1 (A, x0 ), then ϕ(ab) = ϕ(a)ϕ(b). For those who have not had algebra, the existence
of an isomorphism between two groups means that while the two groups may not have the
same elements, they have the same algebraic structure.
Before we actually prove that ϕ satisfies the properties of an isomoprhism, we have to
show ϕ is well defined because ϕ is defined in terms of equivalence classes. First, let f, g
be loops in A based at x0 ∈ A such that f ∼A g. Thus there exists a path homotopy in A,
F : I × I → A, from f to g. Recall the inclusion map i : A → X where i is the identity map
on X restricted to A and is trivially continuous. Thus we can extend F to the continuous
map i ◦ F : I × I → X, and it is easy to see that i ◦ F is a path homotopy in X from f to
g, so f ∼X g. Thus if [f ]A = [g]A , ϕ([f ]A ) = ϕ([g]A ), and ϕ is well defined.
We now prove ϕ is injective. Suppose [f ]A , [g]A ∈ π1 (A, x0 ) such that ϕ([f ]A ) = ϕ([g]A ) ⇒
[f ]X = [g]X . Then there exists F : I × I → X, a path homotopy from f to g in X. Suppose
toward a contradiction there exists (s0 , t0 ) ∈ I × I such that F (s0 , t0 ) ∈ / A. Now take a
restriction of F , α : [0, s0 ] → X defined by α(s) = F (s, t0 ). Since [0, s0 ] ∼
= I, we can easily
reparameterize α over the unit interval as β : I → X, giving a path from x0 to F (s0 , t0 ),
since F (0, t0 ) = x0 . Consequently, α(s0 ) = β(1) ∈ A because A is a path component, so
we have a contradiction. Thus every point in F (I × I) is contained in A, so we can define
G : I × I → A by G(s, t) = F (s, t), so G is a homotopy from f to g in A. Consequently,
f ∼A g ⇒ [f ]A = [g]A . Thus ϕ is injective.
Next we prove ϕ is surjective. Let [f ]X ∈ π1 (X, x0 ), so f is a loop in X based at x0 . In
other words, f : I → X is a path containing x0 . Since A is a path component, f (I) ⊆ A
since we can easily use f to find a path between any two points in f (I). Consequently, f
is a loop in A based at x0 , so we may state ϕ([f ]A ) = [f ]X and ϕ is surjective.
Finally, we prove that ϕ is a homomorphism. Let [f ]A , [g]A ∈ π1 (A, x0 ). We see that:
ϕ([f ]A [g]A ) = ϕ([f ∗ g]A ) = [f ∗ g]X = [f ]X [g]X = ϕ([f ]A )ϕ([g]A )
and ϕ satisfies the definition of a homomorphism.
We have proved that ϕ is a bijective, homomorphism, and is therefore an isomorphism
between π1 (A, x0 ) and π1 (X, x0 ), so we conclude the two groups are isomorphic, as desired.


We now wish to prove a similar result:

Theorem 26. Let X be a topological space and let x, y ∈ X. Suppose f : I → X is a path
from x to y, then π1 (X, x) ∼
= π1 (X, y).
 7. LOOPS AND THE FUNDAMENTAL GROUP 75

Before providing the proof, we will define a key map which we will use again later in the
course:
Definition. Define uf : π1 (X, x) → π1 (X, y) by uf ([g]) = [f ∗ g ∗ f ].3

We now prove the theorem by showing that uf is an isomorphism from π1 (X, x) to π1 (X, y):

Proof. Again, we need to show that uf is a bijective homomorphism. As usual for

functions defined in terms of equivalence classes, we need to show that uf is actually well
defined.
Suppose that g, h are loops in X based at x such that g ∼ h, so there exists F : I × I → X,
a path homotopy from g to h in X. We want to show that:
f ∗g∗f ∼f ∗h∗f
Since we know trivally f ∼ f , g ∼ h and f ∼ f , using our inverse and identity lemmas for
multiplication of path homotopy classes, it is easy to see that f ∗ g ∗ f ∼ f ∗ h ∗ f , so uf is
well defined.
Next, we show that uf is injective. Suppose g, h are loops in X based at x such that
uf ([g]X ) = uf ([h]X ). Therefore:
[f ∗ g ∗ f ] = [f ∗ h ∗ f ] ⇒ f ∗ g ∗ f ∼ f ∗ h ∗ f
so by our important lemma for path homotopies and our inverse/identity lemmas, we have
that g ∼ h and uf is injective.
Now we wish to show that uf is surjective. Let [g] ∈ π1 (X, y). Then [f ∗ g ∗ f ] ∈ π1 (X, x),
and uf ([f ∗ g ∗ f ]) = uf ([f ∗ (f ∗ g ∗ f ) ∗ f ]), which by associativity, inverses, and identity,
is precisely [g]. Therefore uf is onto.
Finally, uf is a homomorphism: let g, h ∈ π1 (X, x). Then
uf ([g])uf ([h]) = [f ∗ g ∗ f ][f ∗ h ∗ f ]
= [f ∗ g ∗ h ∗ f ]
= uf ([g ∗ h])

Therefore, uf is a group isomorphism. 

Keep in mind that our overall goal is to show that the fundamental group is a topological
invariant. In general, a topological invariant is a “thing” that takes a value on topological
spaces such that every homeomorphic space takes the same value. These invariants are
used to distinguish topological spaces.
3Technically, f ∗ g ∗ f should have some indication of ordering of operations, but we dispense with these
designations because we previously proved that ∗ is associative.
76 3. DISTINGUISHING SPACES

7.2. Induced Maps.

Definition. Let φ : X → Y be continuous, and φ(x0 ) = y0 . Define φ∗ : π1 (X, x0 ) →

π1 (X, y0 ) by φ∗ ([f ]X ) = [φ(f )]Y .

We say that φ∗ is induced by φ.

Small Fact (about induced maps). Let φ : X → Y and φ(x0 ) = y0 . Then φ∗ is well
defined.

Proof. Let f and g be loops in X based at x0 such that f ∼ g. Then there exists
F : I × I → X, a path homotopy from f to g.

Consider φ(F ) : I × I → Y . We see that φ is a composition of continuous functions, so is

itself continuous.

Also, observe that

φ(F (s, 0)) = φ ◦ f (s)

φ(F (s, 1)) = φ ◦ g(s)
φ(F (0, t)) = φ(x0 ) = y0
φ ◦ F (1, t) = φ(x0 ) = y0

Thus φ(F ) is a homotopy between φ ◦ g and φ ◦ f . So φ ◦ g ∼ φ ◦ f – and φ∗ is well

defined. 

Lemma. Let φ : X → Y be continuous and φ(x0 ) = y0 . Then φ∗ is a homomorphism.

Proof. Let [f ], [g] ∈ π1 (X, x0 ). We want to show that φ∗ ([f ]X [g]X ) = φ∗ ([f ]X )φ∗ ([g]X .

Observe that

φx ([f ]X [g]X ) = φ∗ ([f ∗ g]X ) = [φ(f ∗ g)]Y

(
f (2s), s ∈ [0, 1/2]
φ ◦ (f ∗ g)(x) = φ∗
g(2s − 1) s ∈ [1/2, 1]
(
φ ◦ f (2s) s ∈ [0, 1/2]
=
phi ◦ g(2s − 1) s ∈ [1/2, 1]
= (φ ◦ f ) ∗ (φ ◦ g)(s).
 7. LOOPS AND THE FUNDAMENTAL GROUP 77

So φ ◦ f ∗ g = φ ◦ f ∗ φ ◦ g and, in particular,
φ(f ∗ g)]Y = [(φ ◦ f ) ∗ (φ ◦ g)]Y
= [φ ◦ f ]Y [φ ◦ g]Y
= φ∗ ([f ]X )φ∗ ([g]X )
This concludes the proof of the lemma. 
Theorem 27. Let φ : X → Y be a homeomorphism and φ(x0 ) = y0 . Then φ∗ is an
isomorphism.

With the previous lemma we showed that φ∗ is a homomorphism. It remains to show that
φ∗ is a bijection.

Proof. 1-1: Let [f ]X , [g]X ∈ π1 (X, x0 ) such that φ∗ ([f ]X ) = φ∗ ([g]X ). Then by
definition of φ∗ , [φ ◦ f ]Y = [φ ◦ g]Y .
Observe that φ ◦ f ∼Y φ ◦ g, so there exists a path homotopy F from φ ◦ f to φ ◦ g.
Also note that, as φ is a homeomorphism, φ−1 : Y → X is continuous.
Hence φ−1 ◦ F : I × I → Y is a path homotopy from φ−1 ◦ φ ◦ f to φ−1 φ ◦ g. Since
g is bijective, this is a path homotopy from f to g.
Onto: Recall that φ∗ : π1 (X, x0 ) → π1 (Y, y0 ). Let [f ] ∈ π1 (Y, y0 ). Then [φ−1 (f )]X ∈
π1 (X, x0 ), because φ−1 is continuous.
φ∗ ([φ−1 (f )]X ) = [φ ◦ φ−1 (f )]Y , and as φ is a bijection, this is [f ]Y .
This illustrates that φ∗ is bijective, and hence is an isomorphism between X and Y . 
Small Fact (about induced homomorphisms). (1) If φ : X → Y and ψ : Y → Z
are continuous, then (ψ ◦ φ)∗ = ψ∗ ◦ φ∗
(2) If i : X → X is the identity, then i∗ is the identity isomorphism
(3) Let φ : X → Y be continuous and f a path in X from p to q. Then φ∗ ◦ uf =
uφ(f ) ◦ φ∗ . (recall, uf : π1 (X, p) → π1 (X, q) by uf ([g]) = [f ∗ g ∗ f ])

Proof. (1) Note that (ψ ◦ φ)∗ : π1 (X, x0 ) → π1 (Z, (ψ ◦ φ)(x0 )), a mapping from
equivalence classes of loops in X based at x0 to the equivalence classes of loops in
Z based at (ψ ◦ φ)(x0 ). Let [f ]X ∈ π1 (X, x0 ). We then have that (ψ ◦ φ)∗ ([f ]X ) =
[(ψ ◦φ)(f )]Z . Similarly, we note that (ψ∗ ◦φ∗ )([f ]X ) = ψ∗ ([φ◦f ]Y ) = [ψ ◦φ◦f ]Z =
[(ψ ◦ φ)(f )]Z .
(2) Note that i∗ : π1 (X, x0 ) → π1 (X, x0 ). Let [f ]X ∈ π1 (X, x0 ). Then i∗ ([f ]X ) =
[i(f )]X = [f ]X .
78 3. DISTINGUISHING SPACES

(3) We want to show that the following diagram commutes:

uf
π1 (X, p) −−−−→ π1 (X, q)
 
φ φ
y ∗ y ∗
uφ(f )
π1 (Y, φ(p)) −−−−→ π1 (Y, φ(q))

Let [g]X ∈ π1 (X, p). Then note that

(φ∗ ◦ uf )([g]X ) = φ∗ ([f ∗ g ∗ f ]X )
= [φ ◦ (f ∗ g ∗ f )]Y
= [(φ ◦ f ) ∗ (φ ◦ g) ∗ (φ ◦ f )]Y
It is easy to show that φ ◦ f = φ ◦ f . To see this, note that (φ ◦ f )(s) = φ ◦ f (s) =
φ(f (1 − s)) = (φ ◦ f )(1 − s) = φ ◦ f (s). So continuing the above expression, we
have that:
[(φ ◦ f ) ∗ (φ ◦ g) ∗ (φ ◦ f )]Y = [(φ ◦ f ) ∗ (φ ◦ g) ∗ (φ ◦ f )]Y
= uφ◦f ([φ ◦ g]Y )
= (uφ(f ) ◦ φ∗ )([g]X )
We then conclude that φ∗ ◦ uf = uφ(f ) ◦ φ∗ .

Lemma. Suppose X is path-connected, and x0 ∈ X. Then π1 (X, x0 ) is trivial if and only
if ∀p, q ∈ X and paths f, g in X from p to q, then f ∼ g.

Proof.
(⇒) Suppose that π1 (X, x0 ) = h[ex0 ]i. Let p, q ∈ X, and f, g paths in X from p to q. Then
f ∗ g is a loop based at p. So it is trivial to see that π1 (X, p) ∼
= π1 (X, x0 ) by an earlier
theorem, from which we can see that f ∗ g ∼ ep . Using our multiplication and inverse
lemmas for path multiplication, we conclude that f ∼ g.

(⇐) Suppose that ∀p, q ∈ X and paths f, g from p to q, we have that f ∼ g. Let
p = q = x0 , let f = ex0 , and let g be a loop in X based at x0 . Then g ∼ ex0 . Hence,
π1 (X, x0 ) = h[ex0 ]i. 

7.3. A Technical Lemma. We’d like to try to prove that spaces that are homotopy
equivalent have isomorphic fundamental groups. First, we’ll need a technical lemma.
Lemma (A Technical Theorem). Let φ, ψ : X → Y be continuous, and φ ' ψ by a homotopy
F . Let x0 ∈ X, and a path f : I → Y be given by f (t) = F (x0 , t). Then uf ◦ φ∗ = ψ∗ .
 7. LOOPS AND THE FUNDAMENTAL GROUP 79

Proof. If [g] ∈ π1 (X, x0 ), then uf ◦ φ∗ ([g]X ) = uf ([φ ◦ g]X ) = [f ∗ (φ ◦ g) ∗ f ]Y . We

want to show that this equals ψ∗ ([g]X ). We’ll prove this using the “fishing rod” f , which
represents the image of the line {x0 } × I. We “reel in” the fishing rod, and create the path
homotopy between f ∗ (φ ◦ g) ∗ f and g.
More formally, we want to show that for all [g] ∈ π1 (X, x0 ), uf (φ([g]) = ψ([g]).

Let [g] ∈ π1 (X, x0 ). We want to
show that f ∗ φ(g) ∗
f ∼ ψ(g). More specifically, we will
do so by showing that f ∗ φ(g) ∗ f ∼ eψ(x0 ) ∗ ψ(g) ∗ eψ(x0 ) .
Define G : I × I → Y by

s ∈ 0, 4t
 


eψ(x0 )
s ∈ 4t , 14
 
f (4s − t)



s ∈ 41 , 12
 
G(s, t) = F (g(4s − 1), t)
s ∈ 12 , 2−t
  
f (2s − 1 + t)


 2
s ∈ 2−t

e  
ψ(x0 ) 2 ,1



To prove that f ∗ φ(g) ∗f ∼ eψ(x0 ) ∗ ψ(g) ∗eψ(x0 ) , we must show that G is path homotopy



from f ∗ φ(g) ∗ f to eψ(x0 ) ∗ ψ(g) ∗ eψ(x0 ) .
Continuous. G is defined differently over the five intervals. Over each interval, G is the com-
position of continuous functions and hence continuous. For G to be continuous
everywhere, the value of G at the intersection of any two intervals must agree.

– s = 4t : (f )(4 4t − t) = f (0) = ψ(x0 ) = eψ(x0 ) .

– s = 41 : f (4 14 − 1) = f (1 − t) = f (t).
F (g(4 41 − 1), t) = F (g(0), t) = F (x0 , t) = f (t).

– s = 21 : F (g(4 12 − 1), t) = F (g(1), t) = F (x0 , t) = f (t).

f (2 21 − 1 + t) = f (t).

– s = 2−t 2−t


2 : f (2 2 − 1 + t) = f (1) = ψ(x0 ) = eψ(x0 ) .
By the Pasting Lemma, the function G is continuous.
Homotopy. Consider G(s, 0):



eψ(x0 ) s ∈ [0, 0]
s ∈ 0, 14
  
f (4s)


s ∈ 14 , 12
 
G(s, 0) = F (g(4s − 1), 0)
 1 
f (2s − 1) s ∈ 2, 1





e
ψ(x0 ) s ∈ [1, 1]
80 3. DISTINGUISHING SPACES

Thus, G(s, 0) = f ∗ φ(g) ∗ f . Consider G(s, 1):


s ∈ 0, 14
 


 eψ(x0 )
s ∈ 14 , 41
 
f (4s − 1)


 1 1
G(s, 1) = F (g(4s − 1), 1) s ∈ 4 , 2
 1 1
f (2s) s ∈ ,


 12 2



e
ψ(x0 ) s ∈ 2, 1


Thus, G(s, 1) = eψ(x0 ) ∗ ψ(g) ∗ eψ(x0 ) .

Path: G(0, t) = ψ(x0 ), and G(1, t) = ψ(x0 ).

Hence G is a path homotopy from f ∗ φ(g) ∗ f to eψ(x0 ) ∗ ψ(g) ∗ eψ(x0 ) . Therefore,
uf (φ([g]) = ψ([g]) for all [g] ∈ π1 (X, x0 ) and uf ◦ φ∗ = ψ∗ . 

Corollary. Let φ : X → Y and ψ : Y → X be continuous such that φ ◦ ψ ' 1Y and

ψ ◦ φ ' 1X . Let φ(x0 ) = y0 . Then φ∗ : π1 (X, x0 ) → π1 (Y, y0 ) is an isomorphism.

Proof. We have already proven that φ∗ is a homomorphism. For φ∗ to be an isomor-

phism, it must also be a bijection.
Let F be a homotopy in Y from 1Y to φ ◦ ψ. Let G be a homotopy in Y from 1Y to φ ◦ ψ.
Let G be a homotopy in X from 1X to ψ ◦ φ. Let f : I → Y be f (t) = F (y0 , t) and let
g : I → X be g(t) = G(x0 , t). f is a path in Y from y0 to φ(ψ(y0 )), and g is a path in X
from x0 to ψ(φ(x0 )).
By the technical theorem proven above, uf ◦ 1Y∗ = (φ ◦ ψ)∗ and ug ◦ 1X∗ = (ψ ◦ φ)∗ .
Therefore, uf ◦ 1Y∗ = ψ∗ ◦ φ∗ , implying that uf = φ∗ ◦ ψ∗ . Similarly, ug = ψ∗ ◦ φ∗ . We
have already established that uf is an isomorphism and therefore onto — φ∗ must be onto
as well. Because ug is 1-1 and ug = ψ∗ ◦ φ∗ , φ∗ is 1-1. φ∗ is a bijection and hence an
isomorphism. 

8. Covering Maps

Our ultimate goal is to find a space with an interesting fundamental group.

Definition. Let X and X e be topological spaces, and p : X

e  X be a continuous sur-
jection. An open set U ⊆ X is said to be evenly covered by p if p−1 (U ) is the disjoint
union of open sets Vα , α ∈ A for some index set A, such that for all α ∈ A, p | Vα → U is
a homeomorphism.
In this case, we say that each Vα is a sheet covering U .
 8. COVERING MAPS 81

Example. Let X = D2 with the usual topology, and Xe = D2 × N with the usual product
topology. Define p : X e → X by p(x, n) = x. We can take U to be any open set in X;
∞
[
p−1 (U ) = p−1 (U ) ∩ Vi , where Vi = D2 × i.
i=1

...
f -1(U)

~
X

U
X
Example (a non-example). Let Xe = S 1 , and let X = S 1 ∨ S 1 . That is, X is two copies of
S 1 , which agree at a point.
Let ∼ be the equivalence relation on S 1 given by x ∼ y if and only if x, y ∈ {x1 , x2 } or
x = y. Let p the the quotient map from X e to X, corresponding to this relation, and let U
be an open set in X containing x1 , x2 , as shown. Is U evenly covered?
The answer is no. To see why, note that p−1 (U ) = V1 ∪ V2 , where V1 , V2 are disjoint open
e But p|V1 : V1 → U is not a homeomorphism because it is not onto.
sets in X.
Definition. Let p : X e  X be a continuous surjection. Suppose for all x ∈ X, there
exists an evenly covered open set U containing X. We say p is a covering map, X e is the
covering space and X is the base space.
Example (another non-example). Let X e = R2 , X = R, and P : R2 → R be defined by
p(x, y) = x. Then for each open U ⊆ X, p−1 (U ) = ∪α∈A Vα . (We can think of p−1 (U ) as a
horizontal stack of uncountably many copies of U ). For each α ∈ A, p | Vα : Vα → U is a
homeomorphism. However, each Vα is not open in X. e
Lemma (Important Lemma on Covering Maps). Let p : X e → X be a covering map. Let
−1
x ∈ X. Then the subspace topology on p ({x}) is the discrete topology.

Proof. Let y ∈ p−1 ({x}). We want to show that {y} is open in p−1 ({x}). Since p is
a covering map, there exists a evenly covered open set U containing x. Then p−1 (U ) =
82 3. DISTINGUISHING SPACES

∪α∈A Vα , where the Vα are pairwise-disjoint open sets such that p | Vα : Vα → U is a

homeomorphism for every α ∈ A. Hence, there exists α0 ∈ A such that y ∈ Vα0 . Now, Vα0
e Note that y ∈ Vα ∩ p−1 ({x}). Let y 0 ∈ Vα ∩ p−1 ({x}) be given. We know
is open in X. 0 0
that p | Vα0 is a homeomorphism, so it is injective. Since p(y 0 ) = x, and p(y) = x, we see
that y = y 0 . Thus, {y} = Vα0 ∩ p−1 ({x}), both of which are open in p−1 ({x}). So {y} is
open in p−1 ({x}) with the subspace topology. This completes the proof. 

The take-home message is that in a covering space, points in the pre-image of a single point
are “spread out.”
Example (Important). Let p : R → S 1 be defined by p(x) = (cos(2πx), sin(2πx)). (The
“slinky” space.)

0 1 2 3

For each s ∈ S 1 , an “open interval” around s is evenly covered, and so R is a covering

space.
e = S 1 , X = S 1 by p : X
Example. X e → X is p((cos(2πx), sin(2πx)) = (cos(4πx), sin(4πx)).

Not all quotient maps are covering maps. But we will prove that all covering maps are
quotient maps. (Remember our important example?)
e → X be a covering map. Then
Theorem 28. Let p : X
(1) p is an open map
(2) X is a quotient space, and p is a quotient map.
 8. COVERING MAPS 83

Proof. Let U be open in X. e We want to show that p(U ) is open. Let x ∈ p(U ). Then
there exists an evenly covered open set V containing X. Now, there exists y ∈ U such that
p(y) = x. Then p−1 (V ) = ∪α∈A Vα such that the Vα are disjoint open sets and p | Vα is a
homeomorphism for all α ∈ A. Let α0 ∈ A such that y ∈ Vα0 . There exists such an α0
because x ∈ V .

e Recall p | Vα : Vα → V is a
Because U and Vα0 are both open, U ∩ Vα0 is open in X. 0 0
homeomorphism, so

p(U ∩ Vα0 ) = p (Vα0 ∩ p(U ))

= V ∩ p(U )

where V ∩ p(U ) is open in V and U is open in X. Therefore, V ∩ p(U ) is open in X.

Because our x is in both V and p(U ), x ∈ V ∩ p(U ) ⊆ p(U ). That is, x is an element of an
open set contained in p(U ). Hence, p(U ) is open and p is an open map.

To show X has the quotient topology with regard to p, we want to show that FX =
{U ⊆ X | p−1 (U ) ∈ FX }.

(⊆) Let V ∈ FX . Because p is continuous, p−1 (V ) ∈ FXe . Thus, V ∈ {U ⊆ X |

p−1 (U ) ∈ FX }.

(⊇) Let U ⊆ X such that p−1 (U ) ∈ FXe . Because p is open, p(p−1 (U ) is open in X.
Because p is onto p(p−1 (U )) = U . Thus, U ∈ FX .

FX = {U ⊆ X | p−1 (U ) ∈ FX }, so p is a quotient map and X is a quotient space. 

8.1. Lifts.

Definition. Let p : X e → X be a covering map and f : Y → X be continuous. We define

a lift of f to be any continuous function fe : Y → X such that p ◦ fe = f .

Example. Let X e = R, X = S 1 and p(x) = (cos(2πx), sin(2πx)). Let f : I → S 1 by

f (x) = (cos(πx), sin(πx)). Define fe : I → R by f (x) = x2 . Then fe is a lift of f .
84 3. DISTINGUISHING SPACES

~
~ X 0 1
f

p
I
0 1

f
S1
Lemma. Uniqueness of Lifts. Let p : X e → X be a covering map and f : Y → X be a
covering and f : Y → X be continuous and Y be connected. Let fe0 and fe1 be lifts of f .
Suppose there exists y0 ∈ Y such that fe0 (y0 ) = fe1 (y0 ) then fe0 = fe1 .

Proof. Let Y 0 = {y ∈ Y | fe0 (y) = fe1 (y)}. Y 6= ∅. We want to show that Y 0 = Y , we

will accomplish by showing Y 0 is clopen in Y .
0 −1
[ y ∈ Y . There exists an evenly covered open set V containing f (y). p (V ) =
Open: Let
Vα such that Vα ’s are disjoint and open; also, p | Vα : V α → V is a homeo-
α∈A
morphism.
Let q = fe0 (y) = fe1 (y). There exists and α0 ∈ A such that q ∈ Vα0 . fe0−1 (Vα0 ) and
fe1−1 (Vα0 ) are open in Y , and y ∈ fe0−1 (Vα0 ) ∩ fe1−1 (Vα0 ). We claim that fe0−1 (Vα0 ) ∩
fe1−1 (Vα0 ) ⊆ Y 0 .
Let z ∈ fe0−1 (Vα0 ) ∩ fe1−1 (Vα0 ), implying fe0 (z) ∈ Vα0 and fe1 (z) ∈ Vα0 . Because
fe0 (z) and fe1 (z) are lifts of f , p ◦ fe0 (z) = f (z) and p ◦ fe1 (z) = f (z). Since p is 1-1
(because p is a homeomorphism) and fe0 (z) ∈ Vα0 and fe1 (z) ∈ Vα0 , fe0 (z) = fe1 (z).
Thus, z ∈ Y 0 .
Hence, fe0−1 (Vα0 ) ∩ fe1−1 (Vα0 ) is an open subset of Y 0 containing y, so Y 0 is open.
Closed: To show Y 0 is closed, we will show Y − Y 0 is open. Let y[
∈ Y − Y 0 . here exists an
−1
evenly covered open set V containing f (y). p (V ) = Vα such that Vα ’s are
α∈A
disjoint and open; also, p | Vα : V α → V is a homeomorphism.
Note that fe0 (y) 6= fe1 (y). There exists and α1 , α2 ∈ A such that fe0 (y) ∈ Vα1 and
fe1 (y) ∈ Vα2 . The set fe0−1 (Vα1 ) ∩ fe1−1 (Vα2 ) is open and contains y. We claim that
fe0−1 (Vα1 ) ∩ fe1−1 (Vα2 ) ⊆ Y − Y 0 .
 8. COVERING MAPS 85

Let z ∈ fe0−1 (Vα1 ) ∩ fe1−1 (Vα2 ), implying fe0 (z) ∈ Vα1 and fe1 (z) ∈ Vα2 . We now want
to show that fe0 (z) 6= fe1 (z). Recall that p | Vα1 is 1 − 1 and p ◦ fe0 (y) = f (y) =
p ◦ fe1 (y): because fe0 (y) 6= fe1 (y) where fe0 (y) ∈ Vα1 and fe1 (y) ∈ Vα2 , implying that
α1 6= α2 . Otherwise, fe0 (y) and fe1 (y) would be two points in Vα1 that both map
to f (y). Therefore, Vα1 ∩ Vα2 = ∅.

Therefore fe0 (z) 6= fe1 (z), implying that z ∈ Y − Y 0 . This implies y is contained in
the open set fe0−1 (Vα1 ) ∩ fe1−1 (Vα2 ) ⊆ Y − Y 0 , making Y − Y 0 open.
Therefore Y 0 is clopen in Y . Because Y 0 is non-empty and Y is connected, Y 0 must be all
of Y . By the definition of Y 0 , fe0 = fe1 . 
Lemma (Lebesgue Number Lemma). Let X be a compact metric space and let Ω be an
open cover of X. Then ∃ r > 0 such that ∀ A ⊆ X with lub{d(p, q)|p, q ∈ A} < r, A is
contained in a single element of Ω.
(r is said to be a Lebesgue Number for Ω)

We will now use this to prove the existence of lifts.

e → X be a
Theorem 29 (Very Important Homotopy Path Lifting Theorem). Let p : X
covering map. Then,
(1) Given a path f in X and a ∈ X e such that p(a) = f (0), then ∃! (exists unique)
e such that p ◦ fe = f and fe(0) = a.
path fe in X

(2) Given a continuous map F : I × I → X and a ∈ X e with p(a) = f (0, 0), ∃!

continuous map Fe : I × I → X
e such that p ◦ Fe = F and Fe(0, 0) = a.

Proof. (1) ∀ x ∈ f (I), ∃ Vx an evenly covered open set containing x. ∀ x ∈

f (I), f −1 (Vx ) is open in I, so {f −1 (Vx )|x ∈ f (I)} is an open cover of I. So,
∃ Lebesgue number r for this cover. ∃ n ∈ N such that n1 < r. ∀ k ≤ n,
[ k−1 k −1 (V ). So, ∃ {V , V , ..., V } ⊆ {V } such
n , n ] is contained entirely in some f x 1 2 n x
that ∀ k ≤ n, f ([ k−1 , k
]) ⊆ V k . First, V 1 is evenly covered and f (0) ∈ V1 , so
S n n
a ∈ p−1 (V1 ) = α∈A1 Vα . So, ∃ α1 ∈ A1 such that a ∈ Vα1 . p|Vα1 : Vα1 → V1 is
a homeomorphism. So ∀ s ∈ [0, n1 ], define fe(s) = (p|Vα1 )−1 f (s). Note that fe:
[0, n1 ] → X e continuous because (p|Vα )−1 is a homeomorphism. Now note that,
1
as above, p−1 (V2 ) = α∈A2 Vα . f ( n1 ) ∈ V2 by definition. ∃ α2 ∈ A2 such that
S

fe( n1 ) ∈ Vα2 . So, as above, define fe: [ n1 , n2 ] → X e by fe(s) = (p|Vα )−1 f (s). fe:
2
2
[0, n ] → X is therefore continuous by Pasting Lemma.
e

(2) ∀x ∈ F (I × I) ∃ evenly covered open set Vx . √{F −1 (Vx )} is an open cover of

I × I, so it has a Lesbegue number r. ∃ n > r2 . ∀ i ≤ n, let Ai = [ i−1 i
n , n ],
86 3. DISTINGUISHING SPACES

Bi = [ i−1 i
n , n ]. ∀ i, j, F (Ai × Bj ) ⊆ Vij for some Vij ∈ {Vx }. By Part 1, we can lift
F |(I × {0} ∪ {0} × I) to Fe : (I × {0} ∪ {0} × I) → X e such that Fe(0, 0) = a ∈ X.
e

evenly covered by hypothesis, and F (I1 ×J1 ) ⊆ V11 .

Begin by observing that V11 isS
This means that p−1 (V11 ) = α∈A11 Vα where the Vα are disjoint open sets and
A11 is some index set. Since F (0, 0) ∈ V11 , and since Fe(0, 0) = a, it follows that
there exists some α11 ∈ A11 such that a ∈ Vα11 .

Now we worry that this choice of Vα11 will agree with how we defined Fe on the
set L. Worry not! For L is connected, and L ∩ (I1 × J1 ) is connected, and since Fe
is continuous, Fe(L ∩ (I1 × J1 )) is connected. Since the Vα are open and disjoint,
we may therefore conclude that Fe(L ∩ (I1 × J1 )) ⊆ Vα11 (otherwise it would be
disconnected).
Since V1 1 was evenly covered, we know that p | Vα11 is a homeomorphism, so we
may define Fe : I1 × J1 → X
e by:

Fe(s, t) = (p | Vα11 )−1 ◦ F (s, t)

This is a composition of continuous functions, so is continuous. Furthermore,
Fe : L ∪ (I1 × J1 ) → X
e is continuous since we showed that Fe(L ∩ (I1 × J1 )) ⊆ Vα ,
11
so we apply the Pasting Lemma.
Now we want to extend Fe to the rest of I × I, and so we move to I2 × J1 . Here
we have an analogous situation as before: We want to choose the appropriate Vα
associated with V21 so that our extension of Fe agrees with what we had previously.
But again, Fe((I2 × J1 ) ∩ (L ∪ (I1 × J1 ))) is connected, so following the argument
from above there will be an appropriate choice of V α to make it “work”. So we
inductively define Fe : I ×I → Xe such that it is continuous as before, and p◦ Fe = F
and F (0, 0) = a. That F is unique follows from our Uniqueness of Lifts Lemma
e e
above.
We iterate this argument finitely many times for each tile Ii × Jj , and so inductively define
a unique lift of F : Fe : I × I → X
e such that Fe(0, 0) = a. 

The natural intuition is that our new function Fe is a path homotopy when F is a path
homotopy. This intuition provides a delightful segue to the next theorem:
Theorem 30 (Monodromy Theorem). Let p : X e → X be a covering map, and let a ∈ X. e
Let x1 , x2 ∈ X. Suppose that p(a) = x1 , and that f, g are paths in X from x1 to x2 . Let
fe, ge be the unique lifts of f, g beginning at a. Then if f ∼ g, fe(1) = ge(1) and fe ∼ ge.

Before beginning the proof, we observe with relish the etymology of monodromy. Mono
being the prefix for one, and dromy being some sort of Greek for a race track. E.g.
 8. COVERING MAPS 87

hippodrome, airdrome, palindrome (examples courtesy of dictionary.com). Let us race

towards the proof!

Proof. f ∼ g means that there exists a path homotopy F : I × I → X, and so by the

previous theorem there exists a unique lifting of F , whose name is Fe : I × I → X,e and Fe
has the property that F (0, 0) = a and p ◦ F = F . Now f , ge are lifts of f, g respectively.
e e e
Consider Fe | (I × {0}). This is a path in X
e from a to Fe(1, 0). Observe that:

p ◦ Fe | (I × {0}) = F | (I × {0}) = f

since F was a path homotopy, and on the other hand:

p ◦ Fe | (I × {1}) = F | (I × {1}) = g

The first observation allows us to conclude that Fe | (I × {0}) is a lift of f beginning at a.

By the uniqueness of lifts, we conclude that Fe | (I × {0}) = fe. We want to say the same
for Fe | (I × {1}), but we do not know that Fe(0, 1) = a, so we cannot immediately conclude
that this is equal to ge since it could possibly be a lift of g originating at some other point.

We claim: Fe | ({0} × I) = a. To see that this is the case, we know:

p ◦ Fe | ({0} × I) = F | ({0} × I) = x1 .

which implies that

Fe | ({0} × I) ⊆ p−1 (F | ({0} × I) = p−1 (x1 )

Now we know that since p is a covering map, p−1 (x1 ) has the discrete topology. Also,
Fe({0} × I) is connected, so must contain only a single point of p−1 (x1 ). Certainly a ∈
Fe({0} × I), so we may say that a = Fe({0} × I) as desired.

The previous consideration tells us that Fe | (I × {1}) = ge, since the left hand side is a lift
of g originating at a, and by the uniqueness of lifts this must be ge. To finish off proving
that Fe is a path homotopy, we need to show that the endpoints are constant as well. That
is, we want to show that Fe({1} × I) = a0 for some a0 ∈ X.
e But for this, the same argument
as above applies, replacing every instance of x1 with x2 . So we conclude that:

Fe(1, 0) = fe(1) = ge(1) = Fe(1, 1)

which was part of what we were trying to prove. All these considerations together tell us
that Fe is a path homotopy between fe and ge, so fe ∼ ge and we are done. 
 88 3. DISTINGUISHING SPACES

8.2. S 1 and Z. Apparently our ultimate goal is to prove that the fundamental group
of S 1 is isomorphic to Z (with addition). But we need just a teensy bit more machinery,
and introduce a new function.
Definition. Let p : R → S 1 be the covering map p(x) = (cos 2πx, sin 2πx). Let x0 = (1, 0),
let f be a loop in S 1 with base point x0 . Define the degree of f , denoted deg(f ), as fe(1)
where fe is the unique lift of f starting at 0.

First, observe that deg(f ) is well defined. That is, it doesn’t matter which lift we select,
because there is only one lift! Also, convince yourself that p−1 ({x0 }) = Z.
This leads us to the theorem we have been clamoring for:
Theorem 31. Let x0 ∈ S 1 . Then π1 (S 1 , x0 ) ∼
= (Z, +).

Proof. We assume WLOG that x0 = (1, 0) since S 1 is simply connected. Let φ : π1 (S 1 , s0 ) →

Z by φ([f ]) = deg(f ). Recall from last lecture that we defined deg(f ) = fe(1) where
fe was the lift of f based at 0 with respect to the covering map p : R → S 1 defined by
p(x) = (cos 2πx, sin 2πx). Our aim is to show that φ is an isomorphism.
Well-Defined: Suppose [f ] = [g]. Then f ∼ g are homotopic loops in S 1 based at x0 . By
the Monodromy theorem, we may say that fe(1) = ge(1) (from here on out we
assume that fe, ge are the lifts of f, g respectively based at 0). This implies that
deg(f ) = deg(g), which means that φ([f ]) = φ([g]) and so φ is well-defined.
1-1: Let [f ], [g] ∈ π1 (S 1 , x0 ) be such that φ([f ]) = φ([g]). This means that deg(f ) =
deg(g), and this means that fe(1) = ge(1). But R is simply connected, so since
fe, ge share an endpoint we may say that fe ∼ ge, and there exists a path homotopy
Fe : I × I → R such that Fe(0, t) = fe and Fe(1, t) = ge. Consider p ◦ Fe. This is
continuous because it is a composition of continuous functions, and p ◦ Fe(0, t) =
p ◦ fe = f ; p ◦ Fe(1, t) = p ◦ ge = g. Finally, we know that for all s ∈ I we have
Fe(s, 0) = fe(1), so p ◦ Fe(s, 0) = p ◦ fe(1) = f (1) = x0 , and the same goes for t = 1,
so p ◦ Fe goes to a loop homotopy F which takes f to g, but since these are loops
we may deduce that f ∼ g. This means that [f ] = [g] and we are done with
proving one to one.
Onto: Let n ∈ Z. Since R is path connected there exists a path fe from 0 to n. Then
p ◦ fe is a loop in S 1 based at x0 (since p(cos 2πn, sin 2πn) = (1, 0) = x0 ), and
deg(p ◦ fe) = n. Therefore φ([p ◦ fe]) = n. This proves that φ is onto.
Homo: Let [f ], [g] ∈ π1 (S 1 , x0 ). We want to show that φ([f ][g]) = φ([f ]) + φ([g]), and the
right hand side is equal to deg(f ) + deg(g), while the left hand side is equal to
deg(f ∗ g). So we want to show that deg(f ∗ g) = deg(f ) + deg(g). Let f] ∗ g be
 8. COVERING MAPS 89

the lift of f ∗ g beginning at 0. We want to show that f] ∗ g(1) = fe(1) + ge(1). Let
m = fe(1), n = ge(1). Let us define a function h : I → R by:
(
fe(2s) s ∈ [0, 21 ]
h(s) =
ge(2s − 1) + m s ∈ [ 12 , 1]

Now h is a path in R from 0 to m + n. This is because fe, ge are continuous, and

when s = 21 we have fe(1) = m, and ge(0) + m = 0 + m = m, so this is continuous
by the pasting lemma. Also, h(0) = 0 and h(1) = ge(1) + m = n + m. Also we
know: (
f (2s) s ∈ [0, 21
p ◦ h(s) =
p(eg (2s − 1) + m) s ∈ [ 12 , 1]
Now observe that the second half of this is:
p(e g (2s − 1) + m), sin 2π(e
g (2s + 1) + m) = (cos 2π(e g (2s − 1) + m))
g (2s − 1), sin 2πe
= (cos 2πe g (2s − 1))
g (2s − 1))
= p(e
= g(2s − 1)

Thus we see that p ◦ h = f ∗ g, which implies that h = f] ∗ g since h begins at 0 and we

know that lifts are unique w/r/t their starting points. Now h(1) = m + n = fe(1) + ge(1),
so putting it together:
deg(f ∗ g) = deg(f ) + deg(g) ⇒ φ([f ][g]) = φ([f ]) + φ([g])
So we conclude that φ is a homomorphism. This plus one to one and onto means that φ is
in fact an isomorphism, so we conclude that Z ∼
= π1 (S 1 , x0 ). 

We rejoice at the fact that we have now seen a non trivial fundamental group.

8.3. More Fundamental Groups. Check out this theorem:

Theorem 32. Let x0 ∈ X and p : X e → X be a covering map. If X

e is simply connected,
−1
then there exists a bijection from π1 (X, x0 ) to p ({x0 }).

Proof. We only needed simple connectedness in the proof that φ was 1 − 1 and onto,
so we use an analogous proof to show that φ : π1 (X, x0 ) → p−1 ({x0 }) by φ([f ]) = fe(1),
where fe is the unique lift originating at some y0 ∈ p−1 ({x0 }). 
Theorem 33. Let p : S 2 → RP2 be the quotient map. Then p is a covering map.
90 3. DISTINGUISHING SPACES

Proof. Recall that the equivalence relation for this quotient map was defined as x ∼ y
iff x = ±y. We use one of the most powerful proof techniques known to mathematics here:
proof by picture. Consider any open ball in RP2 containing some point. Then its pre image
will be a pair of disjoint balls such that p restricted to the balls will be a homeomorphism.
Just think about it... or look at the pictures below.

-1 S2
2 f

RP2 provides us with another example of a space with a nontrivial fundamental group.
From one of our theorems, we know that given x0 ∈ RP2 there is a bijection from
π1 (RP2 , x0 ) to p−1 ({x0 }) (since RP2 is simply connected), and we know that p−1 ({x0 })
has precisely two elements, so π1 (RP2 , x0 ) ∼= Z2 (alternatively written by algebraists as
Z/2Z).

9. R2 ∼
6= R3

Theorem 34. R2 ∼
6= R3

Proof. First, recall some previous results:

• Rn+1 \ {p} is homotopy equivalent to S n
• π1 (S 2 , x0 ) is trivial
• π1 (S 1 , x0 ) ∼
=Z
Now suppose that R2 ∼
= R3 . Then there exists some homeomorphism h : R2 → R3 .
Let p ∈ R2 and h(p) = q ∈ R3 . Therefore f = h|R2 \{p} : R2 \ {p} → R3 \ {q} is a
homeomorphism. Also, from the previous results:
• π1 (R2 \ {p}, x0 ) ∼
= π1 (S 1 , y0 ) ∼
=Z
• π1 (R3 , \{q}, z0 ) ∼
= π1 (S 2 , w0 ) ∼
= {1}
 10. ONE LAST THOUGHT 91

But f is a homeomorphism, so f∗ : π1 (R2 \ {p}, x0 ) → π1 (R3 , \{q}, z0 ) is an isomorphism.

However, Z ∼6 {1}, so this is a contradiction.
=
Therefore R2 ∼
6= R3 , as desired. 

10. One Last Thought

Example (Not all Fundamental Groups are Abelian). Let X = S 1 ∨ S 1 with wedge point
x0 . Then π1 (X, x0 ) is not abelian.

Proof. Define Y ⊆ R2 as Y = {(x, y) | x = 0 and/or y = 0}.

e as Y with a copy of S 1 wedged at every (z, 0) and (0, z) with z ∈ Z \ {0}
Define X
Define p : Xe → X such that (x, 0) goes to the first copy of S 1 for all x ∈ R, (0, x) goes to
the second, and (z, 0), (0, z) goes to x0 for all z ∈ Z. Copies of S 1 wedged at (z, 0) go with
the usual projection to the second copy of S 1 in X, and the copies wedged at (0, z) go the
the first copy in X, as shown in the image.

~
X X

Now define fe(t) = (t, 0), ge(t) = (0, t), and f = p ◦ fe, g = p ◦ ge. So f is a single loop on the
first circle and g is a single loop on the second circle.
Next, lift f ∗ g and g ∗ f at the origin. The construction of X e gives that f] ∗ g(1) = (1, 0)
and g] ∗ f (1) = (0, 1). So, by the Monodromy theorem, as this are lifts with the same fixed
point, f ∗ g 6∼ g ∗ f . Therefore [f ] and [g] satisfy [f ], [g] ∈ π1 (X, x0 ) and [f ][g] 6= [g][f ], so
π1 (X, x0 ) is not abelian, as desired. 

Well, that’s everything there is to know about topology.