Topology: Notes For W4051, Fall 2004, Columbia University

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 42

Topology

Notes for W4051, Fall 2004, Columbia University

Stanislav Jabuka
Contents

Chapter 1. Continuity on Rn via open sets 5


1. Continuity and convergence 5
2. Open and closed sets in Rn 5
3. Continuity in terms of open sets 6
4. Properties of open and closed sets in Rn 7
Chapter 2. Examples of topological spaces 9
1. The partition topology 9
2. The particular point topology 9
3. The excluded point topology 9
4. The finite complement topology 9
5. The countable complement topology 9
6. The Fort topology 10
7. The order topology 10
8. The lower-limit (upper-limit) topology on R 11
9. The topologist’s sine curve 11
10. The infinite broom 11
Chapter 3. Convergent sequences in topological spaces 13
1. Definition and examples 13
2. Uniqueness of limit and Hausdorff spaces 13
3. Convergent sequences and closed sets 15
4. Convergent sequences and continuous functions 16
Chapter 4. Separation axioms 17
1. The axioms 17
2. Examples 18
Chapter 5. Metric spaces and compactness 19
1. Definition and examples 19
2. Metrizability 19
3. Completions of metric spaces 20
4. The Sobolev spaces 20
5. Compactness for metric spaces 21
6. The one point compactification 23
7. Exercises 24
3
4 CONTENTS

Chapter 6. The Seifert-VanKampen theorem 25


1. Amalgamated products of groups 25
2. The Seifert-VanKampen theorem 27
3. Examples and applications 28
Chapter 7. The higher homotopy groups 35
1. Definitions 35
2. Properties 37
3. Examples and applications 39
4. Higher homotopy groups and suspensions 40
5. Eilenberg-MacLane spaces 41
CHAPTER 1

Continuity on Rn via open sets

1. Continuity and convergence


Recall that the distance function dist : Rn × Rn → R is defined as
p
dist(x, y) = (x1 − y1 )2 + ... + (xn − yn )2
where xi and yi are the coordinates of x and y. We define Bx (r) to be the ball of radius
r > 0 with center x ∈ Rn :
Bx (r) = {y ∈ Rn | dist(x, y) < r}
Definition 1.1. A function f : Rn → R is continuous at a ∈ Rn if for every
ε > 0 one can find a δ > 0 such that if dist(x, y) < δ then dist(f (x), f (y)) < ε. Said
differently, f maps the ball Bx (δ) into the ball Bf (x) (ε) :
f (Bx (δ)) ⊆ Bf (x) (ε)
Definition 1.2. A sequence ai ∈ Rn is convergent with limit a ∈ Rn if for every
ε > 0 there exists a natural number i0 such that for every i ≥ i0 we have dist(ai , a) < ε.
In other words, for i ≥ i0 all the elements ai lie in Ba (ε):
ai ∈ Ba (ε) for i ≥ i0
The following proposition links the concepts of continuity of functions and conver-
gence of sequences.
Proposition 1.3. Let ai ∈ Rn be a convergent sequence with lim ai = a. Let
f : Rn → R be a function which is continous at a. Then f (ai ) is also a convergent
sequence and lim f (ai ) = f (a).
2. Open and closed sets in Rn
For a subset Y ⊆ X we define the complement of Y in X as the set
X − Y = {x ∈ X | x ∈
/ Y}
Definition 2.1. A subset A ⊆ Rn is called closed if for every convergent sequence
ai ∈ A the limit lim ai also lies in A. A subset B ⊆ Rn is called open if Rn − B is
closed.
Examples: When n = 1, all closed intervals [a, b] are closed sets while open
intervals ha, bi are open sets in R. The sets R and ∅ are both closed and open while
the intervals [a, bi are neither closed nor open. Any finite subset of R is closed.
5
6 1. CONTINUITY ON Rn VIA OPEN SETS

When n = 2 examples of closed sets are the closed rectangle [a, b] × [c, d], the closed
circles {(x, y) ∈ R2 | x2 + y 2 ≤ r}, the upper half-plane {(x, y) ∈ R2 | y ≥ 0}, the graph
Γf of a continous function f : R → R: Γf = {(x, f (x)) | x ∈ R}. The main examples
of open sets are the open balls Bx (r).
Proposition 2.2. A subset U ⊆ Rn is open if and only if for every x ∈ U there
exists a ε > 0 such that Bx (ε) ⊆ U .
Proof. =⇒ Suppose U is open and let x ∈ U be an arbitrary element. Suppose
there where no ε > 0 for which Bx (ε) ⊆ U . In particular for ε = 1/i the ball Bx (1/i)
would have to intersect the complement V = Rn − U of U . Pick an element bi ∈
V ∩ Bx (1/i). Choosing i = 1, 2, 3, .... we get a sequence bi ∈ V with the property that
dist(x, bi ) < 1/i. This means that bi converges to x. However, since U is open, V must
be closed and so lim bi must lie in V . But clearly x ∈ / V . This is a contradiction and
so an ε with the property Bx (ε) ⊆ U must exist.
⇐= To prove that U is open we need to prove that V = Rn − U is closed. Supposed
it isn’t closed. Then there exists a convergent sequence bi ∈ V whose limit b = lim bi
lies in U . Pick an ε > 0 so that Bb (ε) ⊆ U . Since none of the bi lie in U , we see that
dist(bi , b) ≥ ε. This is impossible for a convergent sequence. This contradiction leads
us to conclude that V must be closed and hence that U is open. ¤

3. Continuity in terms of open sets


Theorem 3.1. (a) A function f : Rn → R is continuous if and only if for every
open set V ⊆ R the preimage f −1 (V ) of V under f is open in Rn .
(b) More specifically, f is continous at x ∈ Rn if and only if for every open set
V ⊆ R containing f (x), there is an open set U ⊆ Rn containing x with the property
that f (U ) ⊆ V .
Proof. We shall only give the proof of part (a) and leave part (b) as an exercise.
=⇒ Suppose that f is continuous and let V ⊆ R be an open set. Suppose that
U = f −1 (V ) is not open. Then Rn − U is not closed and so we can find a sequence
bi ∈ Rn − U which converges to b ∈ U . On the other hand, f (bi ) ∈ R − V and
f (b) ∈ V . Since V is open we can find a ε > 0 such that Bf (b) (ε) ⊆ V . Because f is
continuous, the sequence f (bi ) converges to f (b) which leads to a contradiction since
dist(f (bi ), f (b)) ≥ ε. Thus U must be open.
⇐= Suppose that f has the property that f −1 (V ) is an open set for every open set
V ⊆ R. Let x ∈ Rn be an arbitrary point. Choose an ε > 0 and let V = Bf (x) (ε).
Then U = f −1 (V ) is open and contains x. Pick a δ > 0 such that Bx (δ) ⊆ U . Then
f (Bx (δ)) ⊆ Bf (x) (ε) which means that f is continuous at x. Since x was arbitrary we
conclude that f is continous. ¤
Remark: The importance of theorem 2.1 is that it provides a definition of conti-
nuity that only relies on the notion of open sets. This notion will be easily generalized
to much more general situations.
4. PROPERTIES OF OPEN AND CLOSED SETS IN Rn 7

Theorem 3.2. A sequence xi ∈ Rn converges to x ∈ Rn if and only if for every


open set V ⊆ Rn containing x there exists a natural number i0 such that for i ≥ i0 we
get xi ∈ V .
The proof of theorem 3.2 is similar to the proof of theorem 2.1 and is left as an
exercise.

4. Properties of open and closed sets in Rn


In the proof of theorem 4.1 below, we shall use the following formulas knows as
DeMorgan’s laws. Let Ui ⊆ Rn be a family of subsets of Rn with i running through
some indexing set I. The following equalities of sets hold:
Rn − (∩i Ui ) = ∪i (Rn − Ui )
Rn − (∪i Ui ) = ∩i (Rn − Ui )
These relations say that
• The complement of the intersection is the union of the complements.
• The complement of the union is the intersection of the complements.
Theorem 4.1. The following are essential properties of open and closed subsets of
n
R :
1. The sets Rn and ∅ are both open and closed sets.
2. The union of any number of open sets is again an open set. The intersection
of any finite number of open sets is an open set.
3. The intersection of any number of closed sets is again a closed set. The union
of a finite number of closed sets is a closed set.
Proof. (1) This is evident from the definitions of open and closed sets.
(2) Let Ui ⊆ Rn be a family of open sets indexed by some set i ∈ I (and I may well
be infinite). Let U = ∪i Ui . We need to show that U is again an open set. Let x ∈ U .
Then x ∈ Ui for some i ∈ I. Find a ε > 0 such that Bx (ε) ⊆ Ui . But then Bx (ε) ⊂ U
since Ui ⊆ U . This shows that for every x ∈ U there is a ε > 0 such that Bx (ε) ⊆ U .
According to proposition 2.2 this means that U is an open set.
On the other hand, let Ui with i ∈ {1, 2, ..., m} be a finite family of open sets and
let V = ∩i Ui . To see that V is open, pick an arbitrary x ∈ V . Then x ∈ Ui for every
i ∈ {1, 2, ..., m} and so there exist numbers εi > 0 with the property that Bx (εi ) ⊆ Ui .
Let ε = min{ε1 , ..., εm }. Clearly then Bx (ε) ⊆ Ui for every i ∈ {1, ..., m}. In particular
Bx (ε) ⊆ V which means that V is open.
(3) This part of the theorem can be proved directly via the definition of closed
sets. An easier route is to use part (2) of the theorem (which we already proved) in
conjunction with DeMorgan’s laws above. Let Vi be a family of closed sets with i ∈ I
with I possibly infinite. Set Ui = Rn − Vi . Then all the sets Ui are open and so
8 1. CONTINUITY ON Rn VIA OPEN SETS

U = ∪i Ui is also open (by part (2) of the theorem). By DeMorgan’s second law we
know that
Rn − U = Rn − (∪i Ui ) = ∩i (Rn − Ui ) = ∩i Vi
Since U is open the set Rn − U is closed and so the above says that ∩i Vi is closed.
The part of (3) about finite unions of closed sets follows in a similar manner (using
DeMorgan’s first law) and is omitted here. ¤
There are many examples of infinite families of open sets whose intersection is not
an open set as well as infinite families of closed sets whose union is not a closed set.
Here are two instances:
• In R let Ui = h−1/i, 1/ii with i = 1, 2, 3, .... Each set Ui is an open set but the
intersection ∩i Ui = {0} is closed.
• In R let Vi = [1/i, 3 − 1/i] for i = 1, 2, 3, .... Each Vi is closed but their union
∪i Vi = h0, 3i is open.
CHAPTER 2

Examples of topological spaces

1. The partition topology


On a set X, a partition of X is a set P of mutually disjoint subsets of X whose
union is all of X. The partition topology TP is the set of unions of elements of P
together with the empty set.
For example, on X = R we can choose P = {[n, n + 1i | n ∈ Z}. The open sets of
the partition topology associated to this topology are intervals of the form [a, bi (with
a and b being integers) as well as [a, ∞i and h∞, bi (with a and b again being integers).

2. The particular point topology


On a set X let p ∈ X be an arbitrary point. We define a topology Tp as the set of
all subsets of X which contain p. We also add the empty set ∅ into Tp .
For example, on X = R pick p = 0. Then the open sets of T0 are all subsets of R
which contain the origin.

3. The excluded point topology


Similar to the previous example, we again pick an arbitrary point p ∈ X but now
we define the topology T p as the set of all subsets of X which do not contain p. We
also add in the empty set and X into T p .
As an example, consider again X = R and p = 0. Then the open sets of T 0 are the
closed sets of T0 from the previous example.

4. The finite complement topology


On a set X we define the finite complement topology Tf c as the set of all subsets
of X whose complement in X is finite (possibly empty). We also add in ∅ into Tf c .
On X = R, the closed sets in this topology are all finite subsets of R as well as ∅
and R.

5. The countable complement topology


The countable complement topology Tcc on a set X consists of all subsets of X with
countable complement (and as usual, add in the empty set). When X itself is a finite
set, the topologies Tf c and Tcc are the same.
On X = R, the closed sets of Tcc are the countable sets (in addition to R itself).
9
10 2. EXAMPLES OF TOPOLOGICAL SPACES

6. The Fort topology


Here we assume that X is an infinite set. Let p ∈ X be an arbitrary point in X.
We define the Fort topology TF,p as the set of all subsets of X which either have finite
complements or which don’t contain p (this is a mix of the finite complement topology
and the excluded point topology).
If we equip X = R with the Fort topology (with p = 0) then the closed sets are all
finite sets as well as all sets which contain 0.

7. The order topology


The order topology Tor is defined on an ordered set X. We say that X is ordered
if it carries a relation “≤” subject to the conditions
1. x ≤ x for any x ∈ X.
2. If x ≤ y and y ≤ x then x = y.
3. If x ≤ y and y ≤ z then x ≤ z.
We write x < y if x ≤ y and x 6= y. Given such an ordering on X, we define Tor as the
subsets of X obtained as finite intersections and arbitrary unions of sets of the form

Lx = {y ∈ X | y < x} and Rx = {z ∈ X | x < z}

with x ranging through all of X. Notice that the “intervals” hx, yi (where x < y)
defined as Rx ∩ Ly are open sets.
The order topology gives us a new topology on Rn . The sets Rn are ordered by the
so called lexicographical ordering in which x = (x1 , ..., xn ) < y = (y1 , ..., yn ) if at least
one of the following is true:

x1 < y 1
x1 = y1 and x2 < y2
x1 = y1 , x2 = y2 and x3 < y3
..
.
x1 = y1 , x2 = y2 , ..., xn−1 = yn−1 and xn < yn

When n = 1, this topology equals the usual Euclidean topology on R. When n = 2, we


get a topology on R2 different from the Euclidean topology. For example, the interval
ha, bi ⊆ R2 in Tor with a = (0, 0) and b = (1, 0) is the set (convince yourself of this!)

ha, bi = {(x, y) ∈ R2 | 0 < x < 1} ∪ {(0, y) ∈ R2 | y > 0} ∪ {(1, y) ∈ R2 | y < 0}

This is not an open set in the Euclidean topology on R2 .


10. THE INFINITE BROOM 11

8. The lower-limit (upper-limit) topology on R


The lower-limit topology Tll on R is the topology generated by intervals [a, bi with
a, b ∈ R arbitrary. In other words, the elements of Tll are set obtained as unions of
intervals of the form [a, bi. The upper-limit topology Tul is generated by ha, b].
Notice that every set that is open in the Euclidean topology is also open in Tll (as
well as in Tul ). The converse is not true since for example [a, bi is open in Tll but is
not open in the Euclidean topology on R.
9. The topologist’s sine curve
Let X ⊂ R2 be the set
X = {(x, sin(1/x)) | x ∈ h0, 1]} ∪ {(0, 0}
equipped with the induced topology coming from the Euclidean topology on R2 . This
space is called the topologists sine curve. As we shall see later in the course, this space
has a number of unusual properties.
10. The infinite broom
Let In ⊆ R2 be the closed line segment joining the origin (0, 0) to the point (1, 1/n)
where n ∈ N (here, as always, N denotes the natural numbers). We define X, the
infinite broom, to be the space
X = (∪n∈N In ) ∪ (h0, 1] × {0})
endowed with the induced topology coming from the Euclidean topology on R2 .

Many more examples of topological spaces can be found in the book “Counterex-
amples in Topology” by L. A. Steen and J. A. Seebach, Jr published by Springer Verlag
(1970) and Dover (1995).
CHAPTER 3

Convergent sequences in topological spaces

1. Definition and examples


Definition 1.1. Let (X, T ) be a topological space and xn ∈ X a sequence. We say
that the sequence xn converges to x0 ∈ X if for every open set U ⊆ X which contains
x0 there exists an n0 ∈ N such that for all n ≥ n0 the points xn lie in U .
Examples: 1. This definition of convergence agrees with the familiar definition
(from calculus) of convergence on Rn if we endow Rn with the Euclidean topology.
2. Let X = R be equipped with the partition topology TP associated to the
partition
P = {[4a − 2, 4a + 2i | a ∈ Z}
Then the sequence xn = (−1)n converges to 0. In fact it converges to any point in
[−2, 2i. Show that the sequence yn = sin n is also convergent.
3. If X is equipped with the trivial topology T = {∅, X} then any sequence in X
is convergent and its limit is any point in X.
4. Let X = R be given the finite complement topology. Convince yourself that
the sequence xn = n converges to x0 = 1. Does this still remain true if instead we let
X have the countable complement topology?
5. On X = R consider the particular point topology Tp with p = 0. Then the
only sequences converging to p = 0 are the sequences which are constant (and equal to
zero) after some n0 , that is there is some n0 such that for all n ≥ n0 we have xn = 0.
Describe all sequences which converge to x0 = 1!

2. Uniqueness of limit and Hausdorff spaces


As the examples above show, in general topological spaces limits of sequences may
not be unique. They are unique tough if X happens to have the Hausdorff property.
Definition 2.1. A topological space X is called Hausdorff if for every two points
a, b ∈ X there exist disjoint open sets Ua , Ub ⊆ X with a ∈ Ua and b ∈ Ub .
Theorem 2.2. Let X be a Hausdorff space and xn ∈ X a convergent sequence.
Then the limit lim xn is unique.
n→∞

Proof. Suppose that there are two (or more) limits, say a and b. Since X is
Hausdorff, we can find disjoint open sets Ua and Ub with a ∈ Ua and b ∈ Ub . Let
na ∈ N be such that xn ∈ Ua for all n ≥ na and nb ∈ N have the property that xn ∈ Ub
13
14 3. CONVERGENT SEQUENCES IN TOPOLOGICAL SPACES

for all n ≥ nb . Then for all n ≥ max{na , nb } we have that xn ∈ Ua ∩ Ub which is a


contradiction since Ua ∩ Ub = ∅. ¤
The converse of theorem 2.2 is not generally true. That is, there are examples
of topological spaces which are not Hausdorff even tough every of their convergent
sequences has a unique limit. Here is one such example.
Example 6: Consider X = R with the countable complement topology. Let us
first check that every convergent sequence xn ∈ X has a unique limit. Suppose not,
that is suppose that lim xn = a and lim xn = b with a 6= b. Let Ua be the open set
Ua = R − {xi | xi 6= a}
Clearly a ∈ Ua and so there must be some na ∈ N such that xn ∈ Ua for all n ≥ na .
But then xn = a for all n ≥ na since xn ∈ Ua ∩ {xi | i ∈ N} = {a}. A similar argument
shows that for some nb ∈ N all xn = b for n ≥ nb . But then xn = a and xn = b
for n ≥ max{na , nb } which is impossible since a 6= b. On the other hand, X is not
Hausdorff since every two non-empty open sets have nontrivial intersection.
There is still a broad class of topological spaces for which the converse of theorem
2.2 does hold true. To describe this class of spaces we first need the following definition.
Definition 2.3. Let (X, T ) be a topological space and x ∈ X a point. A collection
of open sets Bx is called a basis at x if for every open set U ⊆ X containing x
there is some element V ∈ Bx such that V ⊆ U . A topological space (X, T ) is called
first-countable if every point x ∈ X has a countable basis at x.
Note that if X is second-countable then it is first-countable (but not the other way
around). We now have the following converse of theorem 2.2.
Theorem 2.4. If X is a first-countable topological space and has the property that
every convergent sequence has a unique limit then X is Hausdorff.
Proof. Given two arbitrary points a, b ∈ X we need to find two disjoint open sets
of which one contains a and the other contains b.
Let Ba = {Uia ⊆ X | i ∈ N} and Bb = {Uib ⊆ X | i ∈ N} be countable bases at a and
b respectively. We define new open sets Via and Vib as
Via = U1a ∩ U2a ∩ ... ∩ Uia and Vib = U1b ∩ U2b ∩ ... ∩ Uib
These sets are open sets since they are obtained as finite intersections of open sets.
Furthermore, notice that Vja ⊆ Uia and Vjb ⊆ Uib for every j ≥ i and a ∈ Via and b ∈ Vib
for every i ∈ N.
If Via ∩Vib = ∅ for some i we have proved the theorem. Suppose thus that Via ∩Vib 6= ∅
for any i ∈ N. Let xi ∈ Via ∩ Vib be an arbitrary point. This yields a sequence in X
which we claim converges to a. To see this, let U be an open set containing a. Let
Una ∈ Ba be an open set such that Una ⊆ U . Then Vja ⊆ U for every j ≥ n, in particular
xj ∈ U for every j ≥ n. This means that lim xi = a. Repeating this same argument for
b shows also that lim xi = b. This is a contradiction since by assumption all convergent
sequences in X have a unique limit.
3. CONVERGENT SEQUENCES AND CLOSED SETS 15

We conclude that there is some i ∈ N for which Via ∩ Vib = ∅. Since a, b ∈ X were
chosen arbitrarily, X is a Hausdorff space. ¤

3. Convergent sequences and closed sets


Recall that we saw that a subset A ⊆ Rn (with the Euclidean topology) is closed if
and only if it contains the limits of all its convergent sequences (this was our original
definition of a closed subset of Rn ). This characterization of closed sets is only “half-
way”true in general topological spaces.
Theorem 3.1. If A ⊆ X is a closed subset of X, then A contains the limits of all
its convergent sequences.
Proof. Suppose an ∈ A is a convergent sequence with lim an = a0 . If a0 ∈ / A
then a0 ∈ X − A which is an open set. Since an is convergent, there must be some
n0 ∈ N such that an ∈ X − A for all n ≥ n0 . This is impossible since an ∈ A and
A ∩ (X − A) = ∅. ¤
That the converse of this theorem is not true in general topological spaces is illus-
trated by the following example.
Example 7: Let X = R be equipped with the countable complement topology
and let A ⊆ X be the set A = X − {0}. Notice that A is not closed (since X − A = {0}
is not open). But A contains the limits of all of its convergent subsequences. To see
this we only need to show that no sequence xn ∈ A can converge to 0. This is easy to
see since the set U = X −{x1 , x2 , ...} is an open set which contains zero but not element
of the sequence xn . Thus A contains the limits of all of its convergent sequences.
Nonetheless, as in the previous section, there is a class of topological spaces for
which the converse of theorem 3.1 is still true. Before stating the theorem we first need
the following lemma.
Lemma 3.2. Let X be a topological space and A ⊆ X a subset of X. If p ∈ Ā − A
then every open set U which contains p must intersect A.
Proof. Suppose not. Then there exists an open set U such that p ∈ U and
A ∩ U = ∅. But then the set Ā − U is a closed set and A ⊆ Ā − U ⊂ Ā. This is a
contradiction because Ā is the smallest closed set containing A. ¤
Theorem 3.3. Let X be a first-countable topological space and A ⊆ X a subset of
X with the property that it contains the limits of all of its convergent sequences. Then
A is closed.
Proof. We will show that A is closed by exhibiting that A = Ā. Suppose not,
then Ā − A is nonempty. Let p ∈ Ā − A and let Bp = {Ui ⊆ X | i = 1, 2, 3, ...} be a
countable basis at p. As in the proof of theorem 2.4 we define a new family Vi of open
sets as
Vi = U1 ∩ U2 ∩ ... ∩ Ui
16 3. CONVERGENT SEQUENCES IN TOPOLOGICAL SPACES

The sets Vi are open and Vj ⊆ Ui for all j ≥ i. Each Vi contains p. By lemma 3.2 each
set Vi ∩ A is non-empty. Pick an arbitrary element xi ∈ Vi ∩ A. Arguing as in the proof
of theorem 2.4, we find that xi is a convergent sequence with lim xi = p. Since xi ∈ A
and lim xi = p we conclude that p ∈ A. Thus Ā = A. ¤
This theorem together with example 7 shows that (R, Tcc ) (where Tcc is the count-
able complement topology) is not a first-countable space (and hence also not second-
countable). Thus (R, Tcc ) is not homeomorphic to (R, TE ) (where TE it the Euclidean
topology on R).
4. Convergent sequences and continuous functions
The relation between convergent sequences and continuous functions in general
topological spaces is the same as it is in Euclidean space:
Theorem 4.1. Let f : X → Y be a continuous function between two topological
spaces. If xi ∈ X is a convergent sequence with lim xi = x0 then f (xi ) ∈ Y is also a
convergent sequence with lim f (xi ) = f (x0 ).
Proof. Let V ⊆ Y be an open set containing f (x0 ). Since f is continuous, the set
U = f −1 (V ) ⊆ X is an open set in X containing x0 . Since lim xi = x0 there is some
n0 ∈ N such that xn ∈ U for all n ≥ n0 . But then f (xn ) ∈ V for all n ≥ n0 showing
that lim f (xi ) = f (x0 ). ¤
CHAPTER 4

Separation axioms

1. The axioms
The following categorization of topological spaces below is a measure of how “fine”or
how “coarse”the topology the space is. If a topological space X has any of these
properties then any space Y homeomorphic to X will also carry the same properties.
1. A topological space is said to be T0 or Kolmogorov if for any two points
x, y ∈ X there exists an open set U ⊆ X such that either x ∈ U and y ∈ / U , or
y ∈ U and x ∈ / U.
2. A topological space is said to be T1 or Fréchet if for any two points x, y ∈ X
there exist open sets Ux , Uy ⊆ X such that x ∈ Ux − Uy and y ∈ Uy − Ux .
3. A topological space is said to be T2 or Hausdorff if for any two points x, y ∈ X
there exist disjoint open sets Ux , Uy ⊆ X such that x ∈ Ux and y ∈ Uy .
4. A topological space is said to be T3 if for any closed set A ⊆ X and any point
y ∈ X − A there are disjoint open sets UA , Uy ⊆ X such that A ⊆ UA and
y ∈ Uy .
5. A topological space is said to be T4 if for any two disjoint closed sets A, B ⊆ X
there are disjoint open sets UA , UB ⊆ X with A ⊆ UA and B ⊆ UB .
6. A topological space is said to be T5 if for any two separated sets A, B ⊆ X (see
definition 1.1 below) there are disjoint open sets UA , UB ⊆ X with A ⊆ UA and
B ⊆ UB .
7. A topological space is called regular if it is both T0 and T3 .
8. A topological space is called normal if it is both T1 and T4 .
Definition 1.1. We say that two subsets A, B ⊆ X of the topological space X are
separated if Ā ∩ B = ∅ = A ∩ B̄.
It should be pointed out that there are examples of topological spaces which satisfy
property Ti but do not have property Ti+1 for i = 0, ..., 4. On the other hand, the
various Ti properties are not completely independent. For example property T1 implies
property T0 , property T2 implies T1 , T4 together with T2 implies T3 and T5 implies T4 .
However T3 does not imply property T2 . These dependencies are summarized in the
table:
T1 =⇒ T0
T2 =⇒ T1
T4 + T2 =⇒ T3
T5 =⇒ T4
17
18 4. SEPARATION AXIOMS

For a wealth of examples of topological spaces with their various properties de-
scribed in detail, see the excellent book “Counterexamples in Topology” by L. A.
Steen and J. A. Seebach, Jr published by Springer Verlag (1970) and Dover (1995).
2. Examples
1. Let X = R be given the partition topology TP associated to the partition
P = {[2a, 2a + 2i | a ∈ Z}. This space is not a Ti space for any i = 0, ..., 5.
Taking x = 0 and y = 1, there is no open set U which contains x but not y
or contains y but not x. Every open set of X either contains both x and y or
contains neither of the two.
2. Let X = R have the finite complement topology. This space is T1 (but not T2 ).
Let x, y ∈ X be arbitrary points and set Ux = X − {y} and set Uy = X − {x}.
These are clearly open sets and x ∈ Ux − Uy and y ∈ Uy − Ux . The space is not
T2 since every two non-empty open sets intersect. The same remains true for
the countable complement topology.
3. Consider X = R with the excluded point topology T p with p = 0. This space
is T0 but fails to be T1 since the only open set containing p is X.
4. Let X = R be given the lower limit topology Tll . Then X is Ti for every
i = 0, 1, ..., 5. To see this is suffices to show that it is T2 and T5 (all the others
follows from these two). It is clear that X is Hausdorff: given two points
x, y ∈ X with x < y, define Ux = [x, yi and Uy = [y, y + 1i. These are open
and disjoint sets containing x and y respectively.
To see that X is T5 , let A, B ⊆ X be two separated sets. Then X − B̄ is
an open set. We can therefore for each a ∈ A ⊆ X − B̄ find an xa ∈ X such
that [a, xa i ⊆ X − B̄ (since the half-open intervals are a basis for the topology).
Define UA = ∪a∈A [a, xa i. In the same vein define UB . These are open sets
(since they are a union of open sets) and clearly A ⊆ UA and B ⊆ UB . It
remains to see that they are disjoint. Suppose not, then UA ∩ UB 6= ∅. Thus
there is some a ∈ A and some b ∈ B so that [a, xa i ∩ [b, xb i 6= ∅. Suppose a < b
(the case b < a is treated analogously), then b ∈ [a, xa i (since b is the smallest
element of [b, xb i) but this is a contradiction since b ∈ B and [a, xa i ⊆ X − B̄.
Thus UA and UB must be disjoint. x
5. Convince yourself that the Euclidean spaces (Rn , TE ) are all Ti for i = 0, 1, 2, 3, 4, 5.
CHAPTER 5

Metric spaces and compactness

1. Definition and examples


Definition 1.1. Let X be a set and d : X × X → [0, ∞i a function subject to the
conditions:
1. d(x, x) = 0 for any x ∈ X and d(x, y) = 0 only if x = y.
2. d(x, y) = d(y, x) for all x, y ∈ X.
3. d(x, z) ≤ d(x, y) + d(y, z) for any x, y, z ∈ X.
Then the function d is called a metric on X and the pair (X, d) is called a metric
space. Each metric space comes equipped with the metric topology Td generated by
the basis Bd defined as
Bd = {Bx (r) | x ∈ X, r > 0}
As in the case of the Euclidean spaces, Bx (r) are the open balls of radius r centered at
x:
Bx (r) = {y ∈ X | d(x, y) < r}
Examples: 1. For any p ≥ 1, the function dp : Rn × Rn → [0, ∞i
p
dp (x, y) = p |x1 − y1 |p + .. + |xn − yn |p
with x = (x1 , ..., xn ) and y = (y1 , ..., yn ), is a metric on Rn . For p = 2 this becomes the
familiar Euclidean distance function. All of these metrics induce the same topology on
Rn , namely the Euclidean topology.
2. Let (X, d) be a metric space and let A be a subset of X. Then (A, d|A ) is also
a metric space. For example, any surface that embeds into R3 is a metric space - it
inherits any of the metrics dp from example 1.
Metric spaces share many nice properties with Euclidean spaces. For example, a
subset U of a metric space X is open if and only if for every x ∈ U there is some ε > 0
such that Bx (ε) ⊆ U . Likewise, a subset A is closed if and only if it contains the limits
of all of its convergent sequences (prove these two facts!).

2. Metrizability
Definition 2.1. We say that a topological space (X, T ) is metrizable if there is
a metric on X such that the metric topology Td agrees with T .
It is easy to see that metrizability is preserved by homeomorphisms. That is to say,
if (X, TX ) is a metrizable topological space and f : X → Y is a homeomorphism to a
19
20 5. METRIC SPACES AND COMPACTNESS

topological space (Y, TY ), then the latter is also metrizable. If dX is the metric on X
inducing the topology TX , then the function
dY (a, b) = dX (f −1 (a), f −1 (b)) a, b ∈ Y
is a metric on Y and it induces the topology TY (check that this is true!).
Since every metric space carries a topology (the metric topology), it is natural to
ask if in fact the converse is true. That is, is every topological space (X, T ) metrizable?
As we shall see, this is not so.
Theorem 2.2. A metrizable topological space (X, T ) is Hausdorff.
Proof. Let d be a metric on X such that T = Td . Let a, b ∈ X be two arbitrary
points and let D = d(x, y). Since x 6= y it must be that D 6= 0. But then the
sets Ua = Ba (D/2) and Ub = Bb (D/2) are two disjoint open sets containing a and b
respectively. ¤
Example 3: We already saw that the topological space (R, Tf c ) (where Tf c is the
finite complement topology) is not Hausdorff. By theorem 2.2 it cannot be metrizable.

3. Completions of metric spaces


Definition 3.1. Let (X, d) be a metric space. We say that a sequence xi ∈ X
is a Cauchy sequence if for every ε > 0 there is some n0 ∈ N such that for every
m, n ≥ n0 we get d(xm , xn ) < ε. We say that (X, d) is a complete metric space if
every Cauchy sequence in X is a convergent sequence.
For example, (Rn , d2 ) is a complete metric space but (h0, 1i, d2 ) is not: the sequence
xi = 1/(i+1) is Cauchy but has not limit in h0, 1i. This last example shows that subsets
of complete metric spaces are not necessarily themselves complete.
Theorem 3.2. For every metric space (X, d) there exists a complete metric space
(X , d0 ) with the properties:
0

1. X is contained in X 0 as a dense subset: X̄ = X 0 .


2. d0 extends d: if x, y ∈ X then d0 (x, y) = d(x, y).
Furthermore, if (X 00 , d00 ) is another such metric space, then there is a homeomorphism
F : X 0 → X 00 such that d00 (F (a), F (b)) = d0 (a, b) for every a, b ∈ X 0 .
A metric space (X 0 , d0 ) as in the theorem above is called a complete extension of
(X, d). The theorem tells us that complete extensions always exist and are unique (up
to metric preserving homeomorphisms) if we require that X be dense in the extension.

4. The Sobolev spaces


For n ∈ {0, 1, 2, ...}, let C n [a, b] be the set of all n times differentiable functions
from [a, b] to R. When n = 0, C 0 [a, b] is the set of all continuous function from [a, b]
5. COMPACTNESS FOR METRIC SPACES 21

to R. Given any real number p ≥ 1 and an integer q with 0 ≤ q ≤ n, we define metrics


dp,q on C n [a, b] as:
ÃZ q
!1/p
bX
dp,q (f, g) = |f (i) (x) − g (i) (x)|p dx
a i=0

where f (x) is the i-th derivative of f (x) and f (0) (x) = f (x).
(i)

Example 4: Consider the simplest case of dp,q where q = 0 = n and p = 2: Then


s
Z b
d2,0 (f, g) = (f (x) − g(x))2 dx
a

Show that d2,0 is a metric on C 0 [a, b].


These metric spaces of functions (C n [a, b], dp,q ) are not complete but by theorem
3.2 we know that they have a unique completion with C n [a, b] being dense.
Definition 4.1. The Sobolev space Lp,q (C n [a, b]) is the unique complete metric
space (X, D) which contains C n [a, b] as a dense subset and where D extends dp,q .
The various Sobolev spaces play an important role in theory of partial differential
equations. Observe that in the definition of the metrics dp,q it was essential that we
could subtract two functions f, g (since terms of the form f (i) (x) − g (i) (x) entered
the definition of dp,q ). These definitions don’t generalize for functions into arbitrary
topological spaces Y . However, we will see soon how one can in general put a topology
on the set C 0 (X, Y ) of continuous functions from X to Y using compact sets.
5. Compactness for metric spaces
Definition 5.1. We say that a metric space (X, d) is totally bounded if for
every ε > 0 there exist finitely many points xi ∈ X, i = 1, ..., N such that
N
[
X= Bxi (ε)
i=1
Typically the number N of points xi needed to cover X by radius ε balls will
increase as ε becomes small. But now matter how small ε gets, N stays finite.
Proposition 5.2. If a metric space (X, d) is totally bounded then X is separable.
Proof. Let εn = 1/n for n ∈ N. For each such εn , there is a finite set of points
An = {xn1 , ..., xn`n } such that
`n
[
X= Bxi (1/n)
i=1
Let A = ∪n∈N An . Clearly A is countable and we claim that A = X. Let x ∈ X be
an arbitrary point. For any n ∈ N, there is some yn ∈ An so that x ∈ Byn (1/n). This
gives a sequence yn ∈ A with d(x, yn ) < 1/n. Thus lim yn = x. This shows that every
x ∈ X lies in the closure of A and so A = X. ¤
22 5. METRIC SPACES AND COMPACTNESS

Proposition 5.3. A metric space (X, d) is second countable if and only if it sep-
arable.
In exercise 3 you are being asked to prove this proposition. Recall that we already
know that a second countable space is separable. The content of the proposition is
that for metric spaces the converse is also true.
Lemma 5.4. Let (X, d) be a metric space with the property that every sequence
xi ∈ X has a convergent subsequence. Then X is totally bounded.
Proof. If X is not totally bounded, then there is some ε > 0 so that X cannot
be covered by a finite number of balls of radius ε. We use this to construct a sequence
xi ∈ X: let x1 be arbitrary and choose consecutive elements of the sequence so that
/ ∪ij=1 Bxj (ε). Let yn = xin be some convergent subsequence of xi with lim yn = y0 .
xi+1 ∈
Let U = By0 (ε/2) and let yk , y` ∈ U be any two elements (which must exist by the
convergence property) with k < `. Then ik < i` and
d(xik , xi` ) ≤ d(xik , y0 ) + d(y0 , xi` ) < ε/2 + ε/2 = ε
This implies that xi` ∈ Bxik (ε) which is a contradiction (to the way we constructed the
sequence xi ). Thus X must be totally bounded. ¤
Theorem 5.5. A metric space (X, d) is compact if and only if every sequence
xi ∈ X has a convergent subsequence.
Proof. =⇒ Suppose X is compact and let xi ∈ X be an arbitrary sequence. If
there was no convergent subsequence of xi we could for each y ∈ X find an open set Uy
which contains only finitely many elements of the sequence xi . But then F = {Uy | y ∈
X} is a cover of X implying that there are finitely many points y1 , ..., ym ∈ X so that
X = Uy1 ∪ ... ∪ Uy1 . This would tell us that the set {xi | i ∈ N}. In particular, some
value must occur infinitely many times giving rise to a convergent subsequence.
⇐= Suppose that every sequence xi ∈ X has a convergent subsequence and let F
be a cover of X. Find a countable open cover F 0 = {Ui ⊆ X | Ui is open , i ∈ N} with
the property that for every Ui ∈ F 0 there is some V ∈ F with Ui ⊆ V (see exercise 2).
Claim: F 0 has a finite subcover.
To
Ãn−1prove
! the claim, suppose the opposite. Then by picking an arbitrary xn ∈ X −
[
Ui we obtain a sequence. By the assumption, there must be a subsequence
i−1
yi = xni with converges to y0 . Since F 0 is a cover, there is some m ∈ N with y0 ∈ Um .
But then yj ∈ / Um for all j ≥ m which is a contradiction. This concludes the proof of
the claim.
Let thus {U1 , .., UN } be a finite cover of X and let Vi ∈ F be such that Ui ⊆ Vi for
all i = 1, ...., N . The {V1 , ..., Vn } is also a finite cover of X.
¤
6. THE ONE POINT COMPACTIFICATION 23

Corollary 5.6. Every compact metric space X is totally bounded, separable and
second countable.
Theorem 5.7. Let (X, d) be a metric space. Then X is compact if and only if X
is complete and totally bounded.
Proof. =⇒ Suppose X is compact, we need to show that X is complete and
completely bounded. To see completeness, let xi ∈ X be any Cauchy sequence in X.
By theorem 5.5 there is a convergent subsequence of xi with limit x0 . It is then easy
to see that xi converges with limit x0 (see exercise 3). Thus X is complete.
On the other hand, consider the open cover Fε of X defined for any ε > 0 as
Fε = {Bx (ε) | x ∈ X}
By compactness of X there must be a finite subcover of Fε . This shows that X is
totally bounded.
⇐= Let X be complete and totally bounded. We will show that X is compact
by showing that any sequence xi ∈ X has a convergent subsequence (see theorem 5.5).
Let εk = 1/k for k ∈ N. For ε1 , there is a finite cover of X with balls of radius ε1 .
Since xi has infinitely many terms, one of the these radius εk balls contains infinitely
many elements of the said sequence. Let B1 be that ball. For ε2 there is again a finite
cover of X by ε2 balls. Since B1 contains infinitely many xi , there is some radius ε2
ball B2 such that B2 ∩ B1 contains infinitely many of the xi . One then continues this
process indefinitely producing a sequence of radius εk balls Bk so that Bk ∩bk−1 ∩...∩B1
contains infinitely many elements of the sequence xi .
Pick now indices n1 < n2 < n3 < ... such that xn` ∈ B` and set yi = xni . It is easy
to see that yi is Cauchy and so by the completeness assumption on X, it must have a
convergent subsequence. This completes the proof. ¤
Corollary 5.8. Consider Rn equipped with the Euclidean topology. Then a subset
A of Rn is compact if and only if it is closed and bounded.
Proof. By exercise 4, A is totally bounded and by exercise 5, A is complete. Now
use theorem 5.7. ¤

6. The one point compactification


Definition 6.1. The compactification of a topological space X is a compact
topological space Y containing X as a dense subset.
Given any non-compact topological space X, compactifications always exist, in fact
a compactification of any non-compact topological space can be achieved by adding
only one additional point to X and extending the topology in a suitable way. This
particular way of compactifying X is called the one-point compactification of X.
Here are the details:
24 5. METRIC SPACES AND COMPACTNESS

Given any non-compact topological space X, define Y = X ∪ {p} where p is some


abstract point (and p ∈ / X). Let TX be the existent topology on X. We define a new
topology TY on Y as follows: A subset U of Y is an open set of Y if either
1. p ∈/ U and U ∈ TX or
2. p ∈ U and X − U is a compact closed subset of X.
Let’s show first that X is dense in Y . By definition of the topology on Y , the set {p}
is not open (since X is non-compact) and so X ⊂ Y is not closed. Therefor X̄ 6= X
which forces X̄ = Y .
Now let’s show that Y is compact. Let F be an open cover of Y . There must be
some set U0 ∈ F which contains p. But then X − U0 is a compact set and so there
is a finite number of elements U1 , ..., Un ∈ F whose union is all of X − U0 . But then
F 0 = {U0 , U1 , ..., Un } is a finite subcover of F showing that Y is compact.
7. Exercises
1. Prove proposition 5.3.
2. Let (X, d) be a second-countable metric space. Then for every open cover F of
X there exists a countable open cover F 0 = {Ui ⊆ X | Ui is open , i ∈ N} with
the property that for every Ui ∈ F 0 there is some V ∈ F with Ui ⊆ V .
3. Let (X, d) be a metric space and xi ∈ X a Cauchy sequence which has a
convergent subsequence with limit x0 . Show that then xi is also convergent
with limit x0 .
4. Consider Rn equipped with the Euclidean topology. Show that A ⊆ Rn is
bounded if and only if it is totally bounded.
5. Let X be a complete topological space and let A be a subset of X endowed
with the relative topology. Then A is also complete.
CHAPTER 6

The Seifert-VanKampen theorem

1. Amalgamated products of groups


1.1. Free groups. In this section we will examine a family of examples of groups,
called the free groups. They will play a role in the discussion preceding the Seifert-
VanKampen theorem.
Let X = {xj | j ∈ J} be some nonempty set indexed by some indexing set J. Let
X −1 be the set of “formal inverses” of X, that is X −1 = {x−1
j | j ∈ J} with J as before.
−1
At this point, the power −1 in xj is just a part of the notation and does not represent
the inverse of xj . Both X and X −1 are only sets and carry no additional structure. We
demand however that X ∩ X −1 = ∅. Finally, let {1} be yet a third set disjoint from
X ∪ X −1 .
A word in X is a finite sequence of elements of X ∪ X −1 ∪ {1}. On the set W of
all words in X, we define an operation by concatenating words, that is

(a1 ...an ) · (b1 ...bm ) = a1 ...an b1 ...bm a1 , ..., an , b1 , ..., bm ∈ X ∪ X −1 ∪ {1}

By definition, this operation is automatically associative. We would like to make W


into a group with the word 1 acting as a unit and the word x−1j acting as the inverse
of the word xj . To achieve this, we simply impose these rules upon W: We shall say
that two words w1 , w2 ∈ W are equivalent if w2 can be obtained from w1 by a finite
number of the following transformations and their inverses:
• Replace an occurrence of 1x or x1 in a given word with x.
• Replace an occurrence of xj x−1 −1
j or xj xj in a given word with 1.

This notion of equivalence between words in W is an equivalence relation (check this!).


We will denote the set of equivalence classes by F(X). There is an obvious operation
on F(X), namely concatenation: [w1 ]·[w2 ] = [w1 ·w2 ] where [w] denotes the equivalence
class of the word w. You should check that this is a well defined operation, i.e. that
for any choices of w1 , w10 ∈ [w1 ] and w2 , w20 ∈ [w2 ] we get [w1 w2 ] = [w10 w20 ].

Theorem 1.1. The set F(X) with the operation of concatenation is a group with
unit element [1] and with [xj ]−1 = [x−1
j ]. It is called the free group generated by
the set X.

For X = {1, 2, ....n}, F(X) is often denoted by Fn and called the free group of rank
n. For example, F1 = Z and F0 is the trivial group.
25
26 6. THE SEIFERT-VANKAMPEN THEOREM

1.2. Generating sets. Let G be a group and G a subset of G (not necessarily a


subgroup of G). We say that G generates G (or that G is a generating set for G)
if every element g ∈ G can be obtained as a product of the elements from G and their
inverses. A group G is called finitely generated if it has a finite generating set. For
example, the groups Fn from the previous section are finitely generated.
A subset H of G is said to normally generate G (or that H is a normal gener-
ating set of G) if the set {ghg −1 |g ∈ G, h ∈ H} is a generating set of G. Having a
normal generating set H of G is often more desirable than just having a generating set
G since typically H can be taken to be a smaller set than G. If however G is Abelian,
then the two notion of a generating set and a normal generating set agree and there is
no distinction between them.
Recall that a subgroup H of G is called normal if ghg −1 ∈ H for every g ∈ G
and h ∈ H. Given a subset H of the group G, the subgroup H of G normally
generated by H is the smallest normal subgroup of G which contains H and it can
easily be seen that H is a normal generating set for H.

1.3. Presentations of groups. A presentation of a group is a system of short-


hand notation which specifies a given group up to isomorphism. It is particularly
well suited for stating the Seifert-VanKampen theorem 2.1 below. Here are first some
preliminaries.
Given a generating set G of G, let G −1 = {g −1 | g ∈ G}. Consider the free group
F(G) (where for X −1 we use the set G −1 ) and the group homomorphisms f : F(G) → G
uniquely determined by
f (g) = g ∀g ∈ G ∪ G −1 ∪ {1}
Let N = Ker(f ). Notice that N is always a normal subgroup of F(G) since it is the
kernel of a homomorphism. Let R ⊆ F(G) be a normal generating set for N . The set
R is called the set of relators for G (and like the set of generators, the set of relators
is by no means unique). The first isomorphism theorem for groups states that the
quotient group F(G)/N is isomorphic to G. The shorthand for writing the quotient
group F(G)/N is simply hG | Ri. Thus the above discussion leads to
(1) G∼
= hG | Ri
which is called a presentation of the group G (not to be confused with a represen-
tation of G). A group G is called finitely presented if both G and R can be chosen
to be finite sets.
Conversely, starting with an arbitrary set G and a subset R of F(G), let N be the
subgroup of F(G) normally generated by R. Then G and R determine a group G as
the quotient group G = F(G)/N which is for short again denoted by hG | Ri.
In conclusion, every group G has a presentation and in fact has typically many
different presentations. Conversely, every presentation uniquely (up to isomorphism)
determines a group.
2. THE SEIFERT-VANKAMPEN THEOREM 27

1.4. Amalgamated products. Let G,H and K be three groups and consider
presentations of G and H:
G∼ = hG | Ri H∼
= hH | Si
In addition, let φG : K → G and φH : K → H be two homomorphisms.
Definition 1.2. With the notation as above, the amalgamated product G ∗K H
of G and H over K associated to the homomorphisms φG and φH is the group whose
presentation is given by
(2) G ∗K H = hG ∪ H | R ∪ S ∪ {φG (k) · φ−1
H (k) | k ∈ K}i
If K is the trivial group then we simply write G ∗ H. The latter group is called the free
product of G and H.
The amalgamated product G ∗K H does depend on the particular homomorphisms
φG and φH but we suppress them from notation for simplicity. It is not hard to check
that the group G∗K H is independent of the particular presentations of G and H chosen
and only depends on the data G, H, K, φG and φH . For example, it is easy to see that
G ∗K H can alternatively be defined as
(G ∗ H)/N
where N is the subgroup of the free product G ∗ H normally generated by {φG (k) ·
φ−1
H (k) | k ∈ K}.

1.5. Abelianization of groups. Let G any group, its commutator subgroups


[G, G] is the group generated by the set of commutators: {ghg −1 h−1 | g, h ∈ G}. It is
always a normal subgroup (check this!). The quotient group G/[G, G] is is called the
Abelianization of G.
Theorem 1.3. If G ∼ = H are two isomorphic groups, then their Abelianizations are
also isomorphic:
G/[G, G] ∼= H/[H, H]
For example, the Abelianization of Fn is Zn , the free Abelian group of rank n.

2. The Seifert-VanKampen theorem


Theorem 2.1. Let X = U ∪ V be the union of two open and path connected subsets
U and V and assume in addition that U ∩ V is also path connected. Let i : U ∩ V → U
and j : U ∩ V → V be the inclusion maps. Then π1 (X) is the amalgamated product of
π1 (U ) and π1 (V ) over π1 (U ∩ V ) associated to i# and j# :
π1 (X) = π1 (U ) ∗π1 (U ∩V ) π1 (V )
Corollary 2.2. With the assumptions as in theorem 2.1, if U and V are simply
connected 1 then so is X.
1Recall that a path connected topological space Y is called simply connected if π1 (Y ) is the trivial
group.
28 6. THE SEIFERT-VANKAMPEN THEOREM

Corollary 2.3. With the assumptions as in theorem 2.1, if U ∩ V is simply


connected then π1 (X) is the free product of π1 (U ) and π1 (V ).

3. Examples and applications


3.1. Real projective space RP2 . We know that RP2 can be constructed by
gluing a 2-dimensional disk to a Möbius band via a homeomorphism. We shall make
the following choices for the sets U and V from theorem 2.1:
U = D2 V = Möbius band =⇒ U ∩ V ' S1
We already know the fundamental groups of these three spaces, namely:
π1 (U ) ∼
= ha |−i ∼
=Z π1 (V ) ∼
=1 π1 (U ∩ V ) ∼
= hb |−i ∼
=Z
The homomorphism i# : π1 (U ∩ V ) → π1 (U ) is given by i# (b) = a2 . This follows from
figures 1 and 2 and the comments therein. Applying theorem 2.1 yields
π1 (RP2 ) = ha | a2 i ∼
= Z2
This group is already Abelian and so it equals its Abelianization. In conclusion,

Figure 1. The pictures above depict the Möbius band M . In both


pictures, the generator a of π1 (M ) is the red curve running through the
center of the Möbius band. The blue curve represents the curve i# (b).
The left picture represents i(b) while in the right picture, we have started
to homotope i(b) towards the red curve. Thus, up to homotopy, we get
i(b) = a2 and so i# (b) = a2 .

Figure 2. These two figures represent the same as figure 1 above except
that the Möbius band has been cut open by a vertical cut. The two
pictures in figure 1 are obtained by glueing the left end to the right end
with a half-twist in the bands.
3. EXAMPLES AND APPLICATIONS 29

observe that RP2 is not homeomorphic to either S 2 or T 2 .

3.2. The figure 8 curve. Let X be the figure 8 and let the sets U, V ⊆ X be
chosen as in figure 3. Then

Figure 3. The figure 8 curve on the left and the sets U and V on the right.

π1 (U ) ∼
= ha|−i ∼
=Z π1 (V ) ∼
= hb|−i ∼
=Z π1 (U ∩ V ) ∼
=1
(since U ∩ V is contractible). Thus from theorem 2.1 it follows that
π1 (X) = ha, b | −i ∼
=Z∗Z
3.3. The Klein bottle. Let K be the Klein bottle. Recall that K is the identifi-
cation space of the rectangle associated to the partition indicated in figure 4. We will

Figure 4. The identification space which gives the Klein bottle.

decompose K as the union U ∪V where U = D1 (with D1 a disk) and V = K−D2 (where


D2 is a slightly smaller disk contained inside D1 ), see figure 5. Then U ∩V is an annulus
and thus homotopy equivalent to S 1 . We then already know that π1 (U ∩V ) ∼ = ha | i ∼
=Z

Figure 5. The decomposition of K into U (on the left) and V (on the
right). The boundary of the shaded circle in V represents j# (a).

and that π1 (U ) = 1. From figure 6 it is easy to see that V is homotopy equivalent to


= hb, c | i ∼
the figure 8 curve (from section 3.2) and thus π1 (V ) ∼ = Z ∗ Z. Also from figure
−1
6 one can easily see that j# (a) = c dcd. Assembling the pieces, we arrive at
π1 (K) = hc, d | c−1 dcdi
30 6. THE SEIFERT-VANKAMPEN THEOREM

Figure 6. This is a picture of V in which the shaded disk from figure


5 has been expanded. The boundary of the shaded region is the loop
a ∈ π1 (U ∩ V ). The generators c, d ∈ π1 (V ) correspond to the top and
left edges of the identification rectangle. Given this, it is now easy to see
that j# (a) = c−1 dcd

Notice that the Abelianization of π1 (K) is the simple group


π1 (K) ∼
= hc, d | d2 i ∼
= Z × Z2
[π1 (K), π1 (K)]
We conclude that the Klein bottle is not homeomorphic to RP2 , S 2 or T 2 .
3.4. The case of RP2 #T 2 . Let X be the the surface RP2 #T 2 . Recall that this
means that X is obtained by removing a disk from both RP2 and T 2 and gluing them
along the boundary (which is a circle S 1 ). Thus we can write X = U ∪ V with
U = RP2 − D2 and V = T 2 − D2 . Clearly U is just the Möbius band. From problem
2 on the graded homework assignment no. 5, we know that T 2 − D2 is homotopy
equivalent to the figure 8 curve from section 3.2. Thus π1 (T 2 − D2 ) ∼
= ha, b | −i ∼
=
Z ∗ Z. The generators a and b can be seen in figure 7. We also already know that

Figure 7. The two generators a and b of π1 (T 2 − D2 ) are the blue and


red curve in the above picture. The shaded region represents the removed
disk D2 whose boundary is the generator d of π1 (U ∩ V ).

π1 (U ) = hc |−i ∼
= Z and, since U ∩ V ' S 1 , we know that π1 (U ∩ V ) ∼= hd | −i. As in
the notation of theorem 2.1, let i and j be the two inclusion maps. From section 3.1
we know that i# (d) = c2 . From figure 7 it is not hard to see that j# (d) = a b a−1 b−1 .
Thus we arrive at
π1 (X) = ha, b, c | c−2 a b a−1 b−1 i
Observe that the Abelianization of π1 (X) is
π1 (X) ∼
(3) = ha, b, c |c−2 i ∼= Z2 × Z2
[π1 (X), π1 (X)]
3. EXAMPLES AND APPLICATIONS 31

It immediately follows that the RP2 #T 2 is not homeomorphic to RP2 , S 2 , T 2 or the


Klein bottle.

3.5. The genus g surface. In this section, let Σg be a genus g surface. Decompose
Σg as U ∪ V with U = Σg − D2 (see figure 8) and V = D2 . Then U ∩ V ' S 1 . Given
this, we know that π1 (V ) ∼
= 1 and π1 (U ∩ V ) = ha | −i ∼
= Z. It remains to find π1 (U )
and to determine the map i# : π1 (U ∩ V ) → π1 (U ). For reasons of exposition, let us

Figure 8. The genus 3 surface Σ3 with a small disk D2 (represented by


the shaded region) removed.

focus on the case of g = 3, the general case of arbitrary g isn’t that much more difficult
other than that one needs to keep track of more generators of π1 (Σg − D2 ).
We start by observing that Σ3 − D2 is homotopy equivalent to the one point union
of 6 circles (also called a wedge of 6 circles). More generally, Σg − D2 is homotopy
equivalent to the one point union of 2g circles. To see this in the case of g = 3, start
by enlarging the hole on Σ3 left behind by the removed disk D2 . Keep enlarging the
hole until you obtain a surface as in figure 9. The latter surface is homotopy equivalent
to a wedge of 6 circles, the homotopy equivalence is obtained by shrinking the width
of the bands down to zero, see figure 10. Thus the fundamental group of Σ3 − D2
is isomorphic to the fundamental group of the wedge of 6 circles (more generally, the
fundamental group of Σg − D2 is isomorphic to the fundamental group of the wedge
of 2g circles). Let Wn denote the wedge of n circles. In section 3.2 we saw that

Figure 9. This is a picture of what Σ3 looks like when the disk D2 is


sufficiently enlarged. The blue curves represent the generators a1 , a2 and
a3 while the red curves are b1 , b2 and b3 .

π1 (W2 ) ∼
= Z ∗ Z since W2 is simply the figure 8 curve. Using the same method one
can inductively find that π1 (Wn ) is the free product of n copies of Z. So in particular,
32 6. THE SEIFERT-VANKAMPEN THEOREM

Figure 10. These two homotopy equivalent pictures are obtained by


shrinking the width of the bands in figure 9 down to zero. On the left
we see the blue and red curves as they were in figure 9 while in the right
figure, they have been rearranged to make them appear more symmetri-
cally. In either case, we obtain a wedge of 6 circles.

π1 (W6 ) ∼
= Z ∗ Z ∗ Z ∗ Z ∗ Z ∗ Z. The free product of n copies of Z is typically written
∗n
as Z . To summarize, we now know that
π1 (Σ3 − D2 ) ∼
= π1 (W6 ) ∼
= ha1 , b1 , a2 , b2 , a3 , b3 |−i ∼
= Z∗6
The generators ai , bi with i = 1, 2, 3 can also be understood explicitly. In W6 they
correspond to the loops which go once around one of the circles in the wedge. This
circles in turn can be seen in Σ3 − D2 as in figure 9. We denote them by ai and bi
because they come in pairs: The blue curves represent a1 , a2 and a3 and the red curves
represent b1 , b2 and b3 .
One can also see from figure 9 what i# (a) is. Notice that i(a) is the boundary of
the surface in figure 9 which is homotopic to (convince yourself of this!)
a1 b1 a−1 −1 −1 −1 −1 −1
1 b1 a2 b2 a2 b2 a3 b3 a3 b3

Given two loops α and β at a point p, we define the commutator [α, β] of α and β
to be the loop
[α, β] = αβα−1 β −1
Using this notation, we have
i# (a) = [a1 , b1 ] · [a2 , b2 ] · [a3 , b3 ]
You might guess that in the case of arbitrary g, this formula becomes
i# (a) = [a1 , b1 ] · ... · [ag , bg ]
With all the necessary pieces in place, we can now plug everything into theorem 2.1 to
obtain
π1 (Σ3 ) ∼
= ha1 , a2 , a3 , b1 , b2 , b3 | [a1 , b1 ] · [a2 , b2 ] · [a3 , b3 ]i
and in general one obtains
g
Y
π1 (Σ3 ) ∼
= h{ai , bi | i = 1, ..., g} | [ai , bi ]i
i=1
3. EXAMPLES AND APPLICATIONS 33

While these groups are rather complicated, their Abelianizations are not. Observe that
in the Abelianization, each commutator [ai , bi ] is automatically trivial. Thus we arrive
at a free Abelian group of rank 2g:
π1 (Σg ) ∼
= Z2g
[π1 (Σg ), π1 (Σg )]
Recall from group theory that two free Abelian groups Za and Zb are isomorphic if and
only if a = b. An immediate corollary of the above calculation is then
Corollary 3.1. Let Σg and Σh be the surfaces of genera g and h. If Σg ∼ = Σh
then g = h.
CHAPTER 7

The higher homotopy groups

1. Definitions
Let I = [0, 1] be the closed unit interval and let I n = [0, 1] × ... × [, 1] be the product
of n copies of I. Notice that the boundary ∂I n is the set
∂I n = {(t1 , ..., tn ) ∈ I n | some ti is either 0 or 1 }
Given a topological space X, as in the case of the fundamental group, we shall pick
an arbitrary point p ∈ X (which once picked should be fixed) and refer to it as the
base point. Let Sn (X, p) (or simply Sn for short) be the set
Sn = {α : I n → X | α is continuous and α(t1 , ..., tn ) = p, ∀(t1 , ..., tn ) ∈ ∂I n }
Thus, Sn are all the continous functions from I n into X which map all of ∂I n to the
basepoint p. Observe that when n = 1 then ∂I = {0, 1} and so the set S1 agrees with
the loop space at the point p ∈ X.
We define an operation which takes two elements from Sn and produces a new
element in Sn : given α, β ∈ Sn we define α · β as
 £ ¤
 α(2t1 , t2 , t3 , ..., tn ) t1 ∈ 0, 21
(α · β)(t1 , ..., tn ) = £ ¤

β(2t1 − 1, t2 , t3 , ..., tn ) t1 ∈ 12 , 1
It is easy to check that α · β is continuous: when t1 = 1/2 then α(1, t2 , ..., tn ) = p
and β(0, t2 , ..., tn ) = p since (1, t2 , ..., tn ), (0, t2 , .., tn ) ∈ ∂I n . On the other hand, if
(t1 , ..., tn ) ∈ ∂I n then the points (2t1 , t2 , ..., tn ) and (2t1 − 2, t2 , ..., tn ) are also in ∂I n
and so (α · β)(t1 , ..., tn ) = p for any (t1 , ..., tn ) ∈ ∂I n . This shows that α · β ∈ Sn .
To make Sn with this operation into a group, we need to pass to homotopy equiv-
alence.
Definition 1.1. We say that two maps α, β ∈ Sn are homotopic if they are homo-
F
topic relative ∂I n . We will write α ' β rel ∂I n if there exists a map F : I n × I → X
such that
F (t1 , ..., tn , 0) = α(t1 , ..., tn )
F (t1 , ..., tn , 1) = β(t1 , ..., tn )
F (t1 , ..., tn , s) = p ∀s ∈ I and (t1 , ..., tn ) ∈ ∂I n
We already know that relative homotopy equivalence is an equivalence relation. Let
[α] denote the equivalence class of α ∈ Sn . Let πn (X, p) be the set
πn (X, p) = {[α] | α ∈ Sn (X, p)}
35
36 7. THE HIGHER HOMOTOPY GROUPS

We shall equip πn (X, p) with the operation


(4) [α] · [β] = [α · β]
Lemma 1.2. The operation defined by (2.2) on πn (X, P ) is well defined. That is,
given α1 , α2 ∈ [α] and β1 , β2 ∈ [β], there is a homotopy F between
F
α1 · β1 ' α2 · β2 rel ∂I n
.
Proof. Let G and H be the homotopies between the two α’s and β’s:
G H
α1 ' α2 rel ∂I n β1 ' β2 rel ∂I n
Then F is the homotopy given by the formula
 £ ¤
 G(2t1 , t2 , ..., tn , s) t1 ∈ 0, 21
F (t1 , ..., tn , s) = £ ¤

H(2t1 − 1, t2 , ..., tn , s) t1 ∈ 12 , 1
¤
Theorem 1.3. The set πn (X, p) equipped with the operation (2.2) is a group for any
n ≥ 1. The unit element is the homotopy class [ep ] of the constant map ep : I n → X
and the inverse of [α] is the homotopy class [α−1 ] where α−1 : I n → X is the map
(5) α−1 (t1 , t2 , ...tn ) = α(1 − t1 , t2 , ..., tn )
For n ≥ 2 these groups are referred to as the higher homotopy groups.
Proof. As in the case of the fundamental group, one needs to check the three
group axioms, each of which comes down to finding homotopies with certain properties.
We shall only focus on last group axiom - the existence of an inverse element. The
verification of the other two axioms is left as an exercise.
F
Thus, we need to check that α · α−1 ' ep rel ∂I n with α−1 as defined in (5). Such
a homotopy is easy to find, namely

 £ ¤

 p t1 ∈ 0, 21 s



 £1 ¤


 α(2t1 − s, t2 , ..., tn ) t1 ∈ 2
s, 12
F (t1 , , , .tn , s) = £1 ¤



 α−1 (2t1 + s − 1, t2 , ..., tn ) t1 ∈ 2
, 1 − 12 s



 £ ¤

p t1 ∈ 1 − 21 s, 1

You should convince yourself that F is indeed a homotopy with the needed properties.
¤
2. PROPERTIES 37

Remark: The homotopies that come up in the proof of theorem 1.3 are in fact the
same homotopies we encountered in the proof for the fundamental group. You should

Figure 1. The homotopy F between α · α−1 and the constant map ep .

notice that in the homotopy F above, not much happens in the coordinates t2 , ..., tn ,
all the action is in t1 . Indeed, F is simply the homotopy corresponding to the picture
in figure 1 with the horizontal direction representing the coordinate t1 and the vertical
direction representing the coordinate s.

2. Properties
Unlike the fundamental group π1 (X, p), the higher homotopy groups πn (X, p) are
always Abelian groups.
Theorem 2.1. For n ≥ 2, the group πn (X, p) is Abelian.
Proof. Given any two [α], [β] ∈ πn (X, p) we need to exhibit that there is a relative
homotopy F between
F
α · β ' β · α rel ∂I n
This homotopy can be taken to be of the form F (t1 , t2 , t3 , .., tn , s) = (G(t1 , t2 , s), t3 , ..., tn )
where G : I 2 × I → I 2 is the homotopy modeled on the sequence of pictures in figure
2. ¤

Figure 2. Each of the squares above represents I 2 with the horizontal


direction corresponding to t1 and the vertical direction to t2 . These pic-
tures represent 5 snapshots (think of the parameter s as being “time”)
of the homotopy G. As s goes from 0 to 1, the homotopy G interchanges
the left and right halves of I 2 by first shrinking both (second picture),
rotating them about each other (pictures 3 and 4) and finally expanding
each to their original size (picture 5). The black and white regions cor-
respond to the maps α and β respectively while all of the grey areas (in
pictures 2,3 and 4) get mapped to the basepoint p.
38 7. THE HIGHER HOMOTOPY GROUPS

Theorem 2.2. The groups πn (X, p)×πn (Y, q) and πn (X ×Y, (p, q)) are isomorphic.
Proof. Consider the function Φ : πn (X, p) × πn (Y, q) → πn (X × Y, (p, q)) defined
by
Φ([α], [β]) = [(α, β)]
It is easy to check that Φ is a homomorphism and a bijection, it is left as an exercise. ¤
As was the case with the fundamental group, the higher homotopy groups are
independent of the particular choice of basepoint. Namely, given two points p, q ∈ X
and given a path σ : I → X with σ(0) = p and σ(1) = q, there is again a homomorphism
σ# : πn (X, q) → πn (X, p) with σ# ([α]) being the homotopy class of the map σ(α) which
is defined (and explained) in figure 3. It is not hard to see that σ# is a homomorphism

Figure 3. The square above represents I n . The map σ(α) : I n → X is


constructed by using the map α on the smaller (shaded) square and using
the path σ along the radial lines connecting the boundary of the outer
square to the boundary of the inner (shaded) square. Convince yourself
that this construction in the case of n = 1 agrees with the construction
we used in the context of π1 .
(see figure 4) whose inverse is the homomorphism (σ −1 )# (where σ−1 is the inverse
path of σ).
Exercise: Give a homotopy between σ −1 (σ(α)) and α.
Thus, as in the case of the fundamental group, σ# is an isomorphism between πn (X, p)
and πn (X, q). Given this observation, if X is path-connected we will simply write
πn (X) to denote πn (X, p) for some p ∈ X.
Any map f : X → Y induces a homomorphism f# : πn (X, p) → πn (Y, f (p)) defined
as f# ([α]) = [f ◦ α].
Theorem 2.3. Let f : X → Y and g : Y → Z be two maps. Then
1. (idX )# = idπn (X)
2. (g ◦ f )# = g# ◦ f#
The first part of the theorem is obvious while the second part is proved in much the
same way as the analogous theorem was proved in the case of the fundamental group
and is omitted. An important corollary of the above theorem is
3. EXAMPLES AND APPLICATIONS 39

Figure 4. The two pictures above represent σ# (α)·σ# (β) and σ# (α·β).
The map α is represented by the grey region and β by the black region.
There is an obvious homotopy between them and thus σ# ([α])·σ# ([β]) =
σ# ([α] · [β]).

Corollary 2.4. If f : X → Y is a homeomorphism then f# : πn (X, p) →


πn (Y, f (p)) is an isomorphism. In particular, if X and Y are two path-connected,
homeomorphic spaces then πn (X) ∼= πn (Y ) for all n ≥ 1.
Thus the higher homotopy groups are topological invariants. As in the case of the
fundamental group, they too only depend on homotopy type of a space rather than on
its homeomorphism type:
Theorem 2.5. If X and Y are two path-connected, homotopy equivalent spaces
then πn (X) ∼
= πn (Y ) for all n ≥ 1.
Theorem 2.6. If X and Y are path connected and homotopy equivalent then
πn (X) ∼
= πn (Y ) for all n ≥ 1.

3. Examples and applications


The higher homotopy groups are harder to calculate than the fundamental group.
There is no analogue of the Seifert-VanKampem theorem for πn when n is greater
or equal to 2. One calculational tool for the higher homotopy groups is described in
section 4. Nonetheless, here are some examples with some partial proofs.
1. πn (Rm ) ∼ = 0 for any n, m ∈ N. This follows from the fact that Rm is homotopy
equivalent to a point (and clearly a point has trivial homotopy groups).
2. πn (S n+k ) ∼= 0 for all k ≥ 1. To prove this, we shall rely on a theorem whose proof
falls into the realm of differential topology:
Theorem 3.1. Suppose k ≥ 1 and let α : I n → S n+k be a map with α(∂I n ) = p
(where p is some chosen basepoint in S n+k ). Then there exists a non-surjective map
β : I n → S n+k which is homotopic to α rel ∂I n .
The upshot of the theorem is that any homotopy class [α] ∈ πn (S n+k ) has a non-
surjective representative β : I n → S n+k . Let q ∈ S n+k be any point not in the image
40 7. THE HIGHER HOMOTOPY GROUPS

of β. Then S n+k − {q} is homeomorphic to Dn+k and can be contracted to p. Thus,


any non-surjective map is null-homotopic and so πn (S n+k ) ∼
= 0.
3. πn (S n ) ∼
= Z. The outline of a proof of this fact is the subject of section 4. An
immediate and very important consequence is
Corollary 3.2. If Rn ∼
= Rm then n = m.
Proof. Suppose that n < m and that f : Rn → Rm is a homeomorphism. Then
f |Rn −{0} : Rn − {0} −→ Rm − {f (0)}
is also a homeomorphism and thus induces an isomorphism between πn (Rn − {0}) and
πn (Rm − {f (0)}). This however leads to a contradiction since Rn − {0} ' S n and
Rm − {f (0)} ' S m while πn (S n ) ∼
= Z and πn (S m ) ∼
= 0. The assumption n > m also
leads to a contradiction and so the only leftover possibility is n = m. ¤

4. πn (S 1 ) ∼
= 0 for n ≥ 2. This follows from the theory of covering spaces.
5. Unlike in the case of S 1 , the homotopy groups πn+k (S n ) for n ≥ 2 do not generally
vanish. Not all of them are known and calculating them in general is a difficult task.
Here are some samples results:
π3 (S 2 ) ∼
= Z π7 (S 4 ) ∼
= Z × Z12 π10 (S 4 ) ∼
= Z24 × Z3 π12 (S 5 ) ∼
= Z30

4. Higher homotopy groups and suspensions


Definition 4.1. Let X be some topological space. We shall define a new topolog-
ical space ΣX, called the suspension of X as the identification space of X × [−1, 1]
(equipped with the product topology) associated to the partition P:
P = {(x, t), X × {−1}, X × {1} | x ∈ X, t ∈ h−1, 1i}
Thus ΣX is the space X × [−1, 1] with X × {−1} and X × {1} each collapsed to a
point. The space X itself is contained in ΣX as X × {0}.

Examples: 1. ΣS n ∼ = S n+1 . This is easily seen in the case of n = 1 (see figure 5)


while in the case of n ≥ 2 one can give an explicit homeomorphism between ΣS n and
S n+1 .

Figure 5. The suspension of S 1 is homeomorphic to S 2 .


5. EILENBERG-MACLANE SPACES 41

2. ΣI n ∼
= I n+1 and ∂(ΣI n ) ∼
= Σ(∂I n ). Prove this!

In order to use suspensions to analyze the higher homotopy groups, we shall use an
equivalent but slightly different description of πn (X, p). Namely, rather than thinking
of elements of πn (X, p) as homotopy classes of maps α : I n → X with α(∂I n ) = p, we
shall regard πn (X, p) as homotopy classes of maps α : S n → X with α(P ) = p where
P ∈ S n is a marked (but otherwise arbitrary) point in S n . The connection between
the two descriptions is of course that the identification space I n /∂I n = S n and P ∈ S n
is the image of ∂I n under the quotient map. The entire formalism developed thus
far works equally well in this new setting. With this understood, we now proceed as
follows:
Any map f : X → Y induces a map Σf : ΣX → ΣY given by
Σf (x, t) = (f (x), t)
In particular, given a map α : S n → X with α(P ) = p, the map Σα : S n+1 → ΣX
sends (P, 0) to (p, 0) ∈ ΣX. This observation gives rise to a function φn : πn (X, p) →
πn+1 (ΣX, (p, 0)) defined as φn ([α]) = [Σα].
Lemma 4.2. The function φn : πn (X, p) → πn+1 (ΣX, (p, 0)) defined above is a group
homomorphism and is called the suspension homomorphism.
The importance of the suspension homomorphism lies in the following theorem.
Theorem 4.3. The suspension homomorphism φi : πi (S n ) → πi+1 (S n+1 ) is an
isomorphism for i < 2n − 1 and a surjection for i = 2n − 1.
Corollary 4.4. The suspension map φn : πn (S n ) → πn+1 (S n+1 ) is an isomor-
phism for n ≥ 2.
Thus knowing that π2 (S 2 ) ∼ = Z we get from the above corollary that πn (S n ) ∼
=Z
2
for all n ≥ 3. The group π2 (S ) can be calculated from the Hurewicz theorem relating
π2 (S 2 ) to the so called homology groups of S 2 which are very easy to calculate.
5. Eilenberg-MacLane spaces
Definition 5.1. Let G be a group and n ∈ N a natural number. A path-connected
topological space X is said to be a K(G, n) space if
πn (X) ∼
=G and πk (X) ∼= 1 for all k 6= n
If n ≥ 2, the group G has to be Abelian according to theorem 2.1. A space X is called
an Eilenberg-MacLane space if it is a K(G, n) space for some G and n.
The main result of this section is the following theorem:
Theorem 5.2. Let G be any group and n ∈ N any natural number. Assume further
that G is Abelian if n ≥ 2. Then K(G, n) spaces exist.
Corollary 5.3. Every group G is the fundamental group of some path-connected
topological space.
42 7. THE HIGHER HOMOTOPY GROUPS

The K(G, n) spaces whose existence is asserted by the above theorem are usually
rather abstract spaces. There are however some concrete examples. For instance, S 1
is a K(Z, 1), RP∞ is a K(Z2 , 1) and CP∞ is a K(Z, 2)1.
Given a K(G, n) and a K(H, n), the product K(G, n) × K(H, n) is a K(G × H, n)
(this fact follows directly from theorem 2.2). Thus the n-dimensional torus T n is an
example of a K(Zn , 1).

1The infinite dimensional real and complex projective spaces RP∞ and CP∞ are the identification
spaces of R∞ and C∞ associated to the partitions
PR = { {λ · x | λ ∈ R − {0}} | x ∈ R∞ }} and PC = { {µ · z | µ ∈ C − {0}} | z ∈ C∞ }}

You might also like