You are on page 1of 92

Part A

Metric and Topological Spaces


1
Lecture A1
Introduction to the Course
This course is an introduction to the basic ideas of topology and metric space
theory for rst-year graduate students. Topology studies the idea of continuity
in the most general possible context. As a separate subject, it takes its origin
from the pioneering papers of Poincar e at the end of the 19th century, and its
development is one of the central stories of 20th century pure mathematics. (The
history by Dieudonn e, A History of Algebraic and Differential Topology, 1900 -
1960, is available free to Penn State members online. but that history begins
roughly where this course leaves off.) But the idea of continuity is so central
that topological methods crop up all over mathematics in analysis of course
(including its computational twin, numerical analysis); in geometry; in pure
algebra; even in number theory. So, if youre a rst-year grad student, you most
likely need to take this course (and maybe even Math 528 as well).
In this lecture well just describe the course protocol and prerequisites well
get started properly next time.
There is a recommended course textbook Introduction to Topology by Gamelin
and Greene, which is available in a cheap Dover reprint for 15 bucks. However,
more important than that are the online course materials, which are available in
two ways:
Through the course website, which is
http://www.math.psu.edu/roe/527-FA10/index.html
Through the course page on ANGEL, Penn States course management sys-
tem, at http://cms.psu.edu
Registered students (i.e., those taking this course for a grade) will want to use
the ANGEL site which will contain homework assignments, quizzes and support-
ing material, which all need to be taken on time in order to receive credit. The AN-
GEL page also contains the detailed course syllabus, which contains University-
required information about exam schedules, academic integrity requirements, of-
ce hours, and suchlike. Please be sure to read this information carefully.
Detailed lecture notes for each lecture will be posted 24 hours in advance on
these sites. To succeed in the course you need to read and study the notes before
the lecture to which they refer. This is very important!
Most lectures will contain one or more in-class exercises. You are expected
to attempt these exercises before the next class session. Students may be called
2
on to present their solutions in class. You grade your own exercises using the
ANGEL system, and this self-assessment will form a (small) component of your
nal in-class grade.
Exercise A1.1. Here is your rst in-class exercise. Find the corresponding as-
sessment item on ANGEL (its called inclassAug23), and enter your response
(I fully understand this). You must complete this assignment before tomorrow
evening, when it will expire.
Nowfor some information about the prerequisites for the course, that is, things
that I will assume that you know. There are four groups of these.
1. Number systems and their properties: specically the natural numbers N,
integers Z, rational numbers Q, real numbers R, and complex numbers C. Courses
in analysis or foundations of mathematics often contain an account of procedures
for constructing all or some of these, but well just take them as given. The most
important of these number systems for us is the system R of real numbers. The
basic properties of the reals are summarized by
Proposition A1.2. R is an archimedean complete ordered eld.
What does this mean?
(a) A eld basic properties of additional, multiplication, subtraction, divi-
sion commutative, associative, distributive laws.
(b) ordered. There is an order relations < with the expected properties. In
particular x
2
0 for all x.
(c) Archimedean no innite or innitesimal real numbers. There is a
unique (injective) homomorphism of rings Z R taking 1 to 1. (Proof?)
The archimedean property is that every element of R is smaller than (the
image of) some element of Z.
(d) complete no holes, contrast Q.
There are various ways of expressing completeness. The standard way is the
Least Upper Bound Axiom. Let S be any nonempty set of real numbers. A number
a R is an upper bound for S if, for all x S, x a. Let U
S
be the set of all
upper bounds for S. The least upper bound axiom says that, if S and U
S
are both
nonempty, then U
S
has a least member. This is called the least upper bound for S
and written sup S.
3
Example A1.3. Let S = x R : x
2
< 2. Then U
S
= a R : a > 0, a
2
2.
U
S
has a least member, namely sup S =

2. The corresponding statement with


R replaced everywhere by Q would be false.
Exercise A1.4. (Cantors nested interval theorem) Let [a
n
, b
n
], n = 1, 2, . . . be a
sequence of closed intervals in R that are nested in the sense that [a
n+1
, b
n+1
]
[a
n
, b
n
] and whose lengths b
n
a
n
tend to zero. Show that the intersection

n
[a
n
, b
n
] contains one and only one real number c.
Exercise A1.5. Use the previous exercise to show that the real numbers are un-
countable, i.e. cannot be listed as c
1
, c
2
, . . .. Hint: Suppose such a listing is
possible. Pick an interval [a
1
, b
1
] of length at most 1 that doesnt contain c
1
, then
a subinterval [a
2
, b
2
] of length at most
1
2
that doesnt contain c
2
, and so on. Apply
the nested interval theorem. What can you say about the limit point c?
4
Lecture A2
Metric Spaces
We continue with our list of prerequisites for the course (we discussed number
1 last time.)
2. Set-theoretic language such as unions (A B), intersections (A B),
complements (AB), subsets (A B) and so on. In topology it is often necessary
to deal with innite unions and intersections. Let F be a family of sets (that is, a
set whose members are subsets of some other set). The intersection of the family,
written

FF
F or just

F, is the set of objects that belong to every member of


the family F: in symbols
x

F F F, x F.
Similarly the union of the family, written

FF
F or just

F, is the set of objects


that belong to some member of the family F: in symbols
x
_
F F F, x F.
De Morgans Laws extend to this context: if F is a family of subsets of some set
X, then
X

FF
F =
_
FF
X F
and the same with intersection and union reversed.
3. Quantied statements and their proofs. We already met some quantiers
in the previous section: things like and . Throughout pure mathematics, but
especially analysis and topology, one meets concepts that are built up of layers of
quantiers nested together in complicated ways. For instance, a function f : R
R is continuous at a given point x
0
if
> 0 > 0 x R[x x
0
[ < [f(x) f(x
0
)[ < .
It is necessary that you can grasp what such a complicated statement says, and
that you can understand how the form of the statement dictates the shape of its
proof. For instance the statement above begins > 0 so its proof has to begin
Let > 0 be arbitrary, followed by an argument that establishes the rest of the
statement for a given arbitrary value of . You could call a simple pattern like this
a proof skeleton corresponding to the for all quantier.
5
Exercise A2.1. Give some examples of other proof skeletons corresponding to
other parts of a statement, e.g. , , and so on.
Exercise A2.2. Write down in symbols (as above) what it is for the function f to
be not continuous at x
0
.
4. Axiomatic method This is the standard operating procedure of modern
mathematics. We understand specic examples by tting them into a general
theory. The general theory consists of a collection of axioms, such as those for a
group or a vector space. There are many examples which realize the axioms, and
we develop a theory that applies to all of them.
We will study topology from this point of view. Well begin with the axioms
for metric spaces.
Denition A2.3. A metric space is a set X equipped with a function d: XX
R (called a metric or distance function) such that:
(i) d(x, x

) 0 for all x, x

; moreover, d(x, x

) = 0 if and only if x = x

.
(ii) d(x, x

) = d(x

, x) for all x, x

(symmetry).
(iii) d(x, x

) d(x, x

) + d(x

, x

) for all x, x

, x

(triangle inequality).
The motivating example is the metric d(z, w) = [z w[ on C or R.
Denition A2.4. Let (X, d
X
) and (Y, d
Y
) be metric spaces. An isometry from X
to Y is a bijection f : X Y such that
d
Y
(f(x
1
), f(x
2
)) = d
X
(x
1
, x
2
)
for all x
1
, x
2
X.
For the usual metric on the plane, the isometries are just the congruences of
Euclidean geometry. Two metric spaces that are related by an isometry are equiv-
alent from the point of view of metric space theory.
Here are some more examples of metric spaces.
Example A2.5. The vector space R
n
can be made into a metric space by dening
the distance between points x = (x
1
, . . . , x
n
) and y = (y
1
, . . . , y
n
) to be
d(x, y) =
_
[x
1
y
1
[
2
+ +[x
n
y
n
[
2
_
1/2
.
(The proof that this does indeed satisfy the triangle inequality is a standard ex-
ercise.) The same formula also makes C
n
into a metric space. These are called
Euclidean spaces.
6
Example A2.6. One can also dene metrics on R
n
by
d

(x, y) =
_
[x
1
y
1
[ + +[x
n
y
n
[
_
and
d

(x, y) = sup
_
[x
1
y
1
[, . . . , [x
n
y
n
[
_
.
These are different metrics from the standard one (though, as well see later, they
are in a certain sense equivalent.)
Example A2.7. Let X be any set. The discrete metric on X is dened by
d(x, y) =
_
0 (x = y)
1 (x ,= y)
Example A2.8. Let A be a nite set (the alphabet) and consider the set A
n
of
n-tuples of elements of A, which we think of as n-letter words in the alphabet A.
Dene a distance on A
n
by
d(x, y) = #i : 1 i n, x
i
,= y
i
;
in other words, the number of positions in which the two words differ. This Ham-
ming distance was introduced in 1950 to give a technical foundation to the theory
of error-correcting codes (Hamming, Error-detecting and error-correcting codes,
Bell System Technical Journal 29(1950), 147160.)
Example A2.9. Let X be any metric space and let Y be a subset of X. Then the
distance function on X, restricted to Y , makes Y into a metric space; this metric
structure is called the subspace metric on Y .
Example A2.10. Here is an innite-dimensional example. Consider the collection
C[0, 1] of all continuous functions [0, 1] C. We can dene a metric by
d(f, g) = sup[f(t) g(t)[ : t [0, 1].
Convergence of a sequence of functions in this metric is called uniform conver-
gence.
Denition A2.11. A subset U of a metric space X is open if for every x U
there is > 0 such that the entire ball
B(x; ) := x

X : d(x, x

) <
is contained in U.
7
The triangle inequality shows that every ball is open, so there are plenty of
open sets. A set F whose complement X F is open is called closed. Note
carefully that closed does not mean the same as not open. Many sets are neither
open nor closed, and some may be both.
8
Lecture A3
Open and Closed Sets
We nished last time with an important denition: A subset U of a metric
space X is open if for every x U there is > 0 such that the entire ball
B(x; ) := x

X : d(x, x

) <
is contained in U.
Example A3.1. In a discrete metric space (Example A2.7), the open ball B(x,
1
2
)
is just the set x. Consequently, in a discrete space, every subset is open.
Example A3.2. Consider the three metrics on R
n
dened in examples A2.5 and A2.6.
Each open ball in any one of these metrics contains an open ball (with the same
center but possibly different radius) in any one of the other metrics. Consequently,
these three metrics all have exactly the same open sets. That is what is meant by
saying that they are equivalent.
Lemma A3.3. The union of any collection of open sets is open. The intersection
of a nite collection of open sets is open. The empty set , and the entire metric
space X, are open.
Proof. Let F be a collection of open subsets of a metric space X and let U =

F be the union of the family. If x U, then there is some V F such that


x V . Since V is open, there is > 0 such that B(x; ) V . Since V U,
we also have B(x; ) U. Thus for any x U there exists > 0 such that
B(x; ) U; which is to say that U is open.
Now let F = U
1
, . . . , U
n
be a nite collection of open sets and let U =

F = U
1
U
n
. If x U, then for each i = 1, . . . , n there is
i
> 0 such
that B(x;
i
) U
i
. Let = min
1
, . . . ,
n
> 0. Then B(x; ) U
i
for all i, and
thus B(x; ) U. Thus for any x U there exists > 0 such that B(x; ) U;
which is to say that U is open.
Equivalently, using de Morgans laws, we have
Lemma A3.4. The intersection of any collection of closed sets is closed. The
union of a nite collection of closed sets is closed. The empty set and the entire
metric space X are closed.
9
Let X be any metric space, and let S be a subset of X.
Denition A3.5. The interior of S (in X), denoted S

, is the union of all the open


subsets of X that are included in S.
In symbols, we have
S

=
_
_
U : U S and U open in X
_
.
The interior of S is an open set (because it is the union of a family of open
sets) and (from the denition) any open subset of X that is included in S is also
included in S

. Thus, the interior of S is just the biggest open subset of X that is


included in X. In particular, S is open if and only if it is equal to its own interior.
Dually, we have
Denition A3.6. The closure of S (in X), denoted S, is the intersection of all the
closed subsets of X that include S.
The closure of S is a closed set, and it is the smallest closed set that includes
S; in particular, S itself is closed iff it is equal to its own closure. Finally,
Denition A3.7. The boundary of S (in X), denoted X, is the set-theoretic
difference S = S S

.
The boundary of S is the intersection of two closed sets (S and X S

), and
it is therefore closed.
Example A3.8. Let X = R and let S =
_
(0, 1) Q
_
(2, 3]. Then S

= (2, 3),
S = [0, 1] [2, 3], and S = [0, 1] 2, 3.
Exercise A3.9. Show that if A, B are subsets of X, and A B, then A

and A B. Now show that for any subset S of X,


_
_
S
_

=
_
S
_

.
One more remark about open sets. Let X be a metric space and let Y be a
subset of X. As we said earlier, Y can be considered as a metric space in its own
right (with the metric that it inherits from X). What is then the relation between
the open subsets of the new space Y and the open subsets of the original space X?
10
Proposition A3.10. Let X be a metric space, Y a metric subspace (as above).
Then a subset V Y is open in Y iff it can be written V = U Y for some
U X open in X.
Proof. Notice the following relationship between balls in Y and in X:if y Y
then
B
Y
(y; r) = B
X
(y; r) Y.
Let V be open in Y . Then for each y V there is r
y
> 0 such that B
Y
(y; r
y
) V .
Let U =

yY
B
X
(y; r
y
). This is a union of open subsets of X, so open in X,
and
U Y =
_
_
yY
B
X
(y; r
y
)
_
Y =
_
yY
B
Y
(y; r
y
) = V.
This proves one direction of the if and only if statement in the proposition; the
other direction (which is easier) is an exercise.
The denition of continuity is translated in the natural way to the metric space
context.
Denition A3.11. A function f : X Y between metric spaces is continuous at
x X if for every > 0 there is > 0 such that d(f(x), f(x

)) < whenever
d(x, x

) < . It is continuous if it is continuous at every x X.


But there is a very important alternative characterization in terms of open sets.
Theorem A3.12. Let X and Y be metric spaces. Then f : X Y is continuous
iff, for every open U Y , the inverse image
f
1
(U) := x X : f(x) U
is open in X.
Proof. Suppose that f is continuous and let U Y be open. Let x f
1
(U);
then f(x) U so by denition of open there is > 0 such that B(f(x); ) U.
By denition of continuous there is > 0 such that if x

B(x; ) then f(x

)
B(f(x); ) U. But this means that B(x; ) f
1
(U). Thus the set f
1
(U) is
open.
Conversely, let x X, > 0 and suppose that f satises the condition in
the theorem. In particular, we may consider U = B(f(x); ), an open set such
that x f
1
(U). Our hypothesis tells us that f
1
(U) is open, which means that
there is a > 0 such that B(x; ) f
1
(U). We have shown that whenever
x

B(x; ), f(x

) B(f(x); ). This gives us continuity.


11
Lecture A4
Continuity and sequences
In the previous lecture we saw that a map f : X Y between metric spaces
is continuous iff f
1
(U) is open (in X) whenever U is open (in Y ).
Remark A4.1. It is equivalent to say that f
1
(F) is closed in X whenever F is
closed in Y . But in general nothing can be said about the behavior of the direct
image f(S) of an open or a closed set S under a continuous map.
Denition A4.2. A map f : X Y between metric spaces is called a homeo-
morphism if it is a bijection and both f and f
1
are continuous. (Equivalently,
f is a continuous map with a continuous inverse.) If there is a homeomorphism
between X and Y , then we say that these spaces are homeomorphic.
A homeomorphism is a topological equivalence (it is easy to check that the
relation of homeomorphism is an equivalence relation in the sense of algebra). In
topology, we consider homeomorphic spaces to be essentially the same. A fun-
damental question in topology is to devise processes for answering the question,
Are two concretely given spaces homeomorphic or not?
Example A4.3. The map x x(1 + x
2
)

1
2
is a homeomorphism from R to the
open interval 9 1, 1). Its inverse is y y(1 y
2
)

1
2
.
Example A4.4. The map t (cos 2t, sin 2t) is a continuous bijection from the
half-open interval (0, 1] onto the unit circle in R
2
, but it is not a homeomorphism.
Example A4.5. Let (X, d) be a metric space and let (X, d

) be the same set


equipped with the discrete metric. Then the identity map (X, d

) (X, d) is
a continuous bijection, but it is usually not a homeomorphism.
Example A4.6. Are R
n
and R
m
homeomorphic if m ,= n? This and related
questions were problematical in the early days of topology. (Note that there are
continuous maps of R
m
onto R
n
even if m < n space-lling curves.) But
the answer, as expected, is indeed no, as was shown by Brouwer and others at
the beginning of the 20th century.
Well now study the properties of sequences. A sequence in a metric space X
is a mapping from the natural numbers Nto X: well follow the usual abbreviation
of referring to the sequence (x
n
) rather than the sequence which maps the
natural number n to x
n
X.
12
Denition A4.7. Let (x
n
) be a sequence in the metric space X. We say (x
n
)
converges to x X if for every > 0 there is an integer N such that d(x, x
n
) <
whenever n > N. We write then x = lim
n
x
n
.
Exercise A4.8. Let T be the metric space 1,
1
2
,
1
3
, . . . , 0 (with the metric it in-
herits as a subspace of R). Show that a sequence (x
n
) in the metric space X is
convergent if and only if there exists a continuous function f : T X such that
f(1/n) = x
n
; and in that case f(0) = lim
n
x
n
.
Proposition A4.9. A function f : X Y between metric spaces is continuous
at x if and only if, whenever (x
n
) is a sequence converging to x, the sequence
(f(x
n
)) converges to f(x).
Proof. Suppose that f is continuous at x and let > 0 be given. By denition of
continuity there is > 0 such that d(x, x

) < implies d(f(x), f(x

)) < . By
denition of convergence there is N such that for all n > N, d(x, x
n
) < . Thus
d(f(x), f(x
n
)) < and f(x
n
) converges to f(x).
Suppose that f is not continuous at x. Then there is > 0 such that for all
> 0 there is x

with d(x, x

) < and d(f(x), f(x

) . Let x
n
be a value of x

corresponding to = 1/n. Then x


n
x, but d(f(x), f(x
n
)) for all n, hence
f(x
n
) does not converge to f(x).
Remark A4.10. This result illustrates how sequences can be used to probe the
topological properties of metric spaces. At the same time, it shows how the use-
fulness of these probes depends on the countable local structure of a metric
space specically, on the fact that any ball around x X contains one of
the countably many balls B(x; 1/n). When we come to consider more general
topological spaces, this countable local structure may not be available, and then
sequences will be of less use.
Denition A4.11. Let X be a metric space, A X. A point x X is a limit
point of A if it is the limit of a sequence of distinct points of A.
Lemma A4.12. x is a limit point of A if and only if every open set containing x
also contains innitely many points of A.
Proof. Suppose that x is a limit point of A. Then there is a sequence (x
n
) of
distinct points of A, converging to x. Let U be an open set containing x. Then
there is some > 0 such that B(x; ) U. By denition of convergence, there
13
is N > 0 such that x
n
B(x; ) U for all n > N. In particular, U contains
innitely many of the x
n
, all of which are members of A.
Conversely suppose that every open set containing x also contains innitely
many points of A. Dene a sequence (x
n
) inductively as follows: x
1
is any
point in B(x; 1) A, and, assuming that x
1
, . . . , x
n1
have been dened, let x
n
be any point in B(x; 1/n) A that is distinct from the (nitely many!) points
x
1
, . . . , x
n1
. Then (x
n
) is a sequence of distinct points of A and it converges to
x.
Proposition A4.13. The closure of a subset A is the union of A and the set of
all its limit points. In particular, A is closed if and only if it contains all its limit
points.
Proof. A point x belongs to the closure of Aiff it is not in the interior of XA. By
denition of interior, this is the same as to say that each ball B(x; ) must meet
A. By the previous lemma, then, any limit point of A must belong to A. Con-
versely, suppose that x A A. Dene a sequence (x
n
) inductively as follows:
x
1
is any point in B(x; 1) A, and, assuming that x
1
, . . . , x
n1
have been dened,
let x
n
be any point in B(x;
n
) A where
n
= mind(x
1
, x), . . . , d(x
n1
, x).
Then (x
n
) is a sequence of distinct points of A and it converges to x, so x is a
limit point of A.
Denition A4.14. An isolated point of A is a point of A that is not a limit point
of A. A closed subset of a metric space is perfect if it has no isolated points.
Exercise A4.15. Show that a A is isolated if and only if there is > 0 such that
B(a; ) A = a.
Exercise A4.16. A closed ball in a metric space is a set

B(x; r) := x

X : d(x, x

) r.
Show that a closed ball is closed. Must every closed ball be the closure of the
open ball of the same center and radius?
14
Lecture A5
Compactness
Ill take for granted the notion of subsequence of a sequence of points in a
matric space. (Formally speaking, if a sequence in X is a map N X, a subse-
quence of the given sequence is the result of composing it with a strictly mono-
tonic map N N.)
Lemma A5.1. Every sequence of real numbers has a monotonic subsequence.
Proof. Let (x
n
) be a sequence of real numbers. Suppose that it has no monotonic
increasing subsequence. Then there must be some n such that x
m
< x
n
for all
m > n. (Otherwise, we could build a monotonic increasing subsequence by
induction.)
But now take this n and call it n
1
. Apply the previous argument to the tail of
the original sequence beginning at n
1
, i.e. the subsequence x
n
1
+1
, x
n
1
+2
, . . .. This
also has no monotonic increasing subsequence, so by the same argument there is
n
2
> n
1
such that x
m
< x
n
2
for all m > n
2
. Continuing in this way by induction
we get a monotonic decreasing subsequence x
n
1
, x
n
2
, . . ..
It is a standard consequence of completeness that every bounded, monotonic
sequence of real numbers is convergent. Thus we get the Bolzano-Weierstrass
theorem: every bounded sequence of real numbers has a convergent subsequence.
Denition A5.2. A Cauchy sequence in a metric space is a sequence (x
n
) with
the following property: for every > 0 there exists N > 0 such that, for all
n, m > N, d(x
n
, x
m
) < . A metric space is called complete if every Cauchy
sequence in it converges.
A Cauchy sequence is necessarily bounded. Moreover, if a Cauchy sequence
has a convergent subsequence, then the whole sequence is in fact convergent.
These facts combine with the Bolzano-Weierstrass theorem to show that every
Cauchy sequence of real numbers converges: i.e., R is (Cauchy) complete.
Denition A5.3. A metric space X is (sequentially) compact iff every sequence
of points of X has a subsequence that converges in X.
Proposition A5.4. Every closed, bounded subset of R
n
or C
n
is sequentially com-
pact. (A subset is bounded if it is contained in some ball.)
15
Proof. For R this is just the Bolzano-Weierstrass theorem. For R
n
, this follows
from the fact (easily proved) that a sequence of points in R
n
is convergent if and
only if each of its coordinate sequences is convergent.
Proposition A5.5. If A is a subset of any metric space X, and A is compact (in
its own right), then A is bounded and closed in X.
Proof. Let x X be a limit point of A. Then there is a sequence (a
n
) in A
converging to x. But, by compactness, (a
n
) has a subsequence converging in A.
Thus x A, and A is closed.
Suppose that A is not bounded. Then one can construct by induction a se-
quence (a
n
) in A such that d(a
n
, a
0
) > 1 + d(a
n1
, a
0
) for n 1. The triangle
inequality shows that d(a
n
, a
m
) > 1 for n ,= m. Clearly (a
n
) has no convergent
subsequence.
Exercise A5.6. Let f : X Y be continuous and surjective. Show that if X is
compact then Y is compact also. What has this got to do with the familiar calculus
principle that a continuous function on a closed bounded interval is itself bounded
and attains its bounds?
Despite the above evidence, its not in general true that closed and bounded
equals compact. (Counterexample?) We shall analyze this in detail.
Denition A5.7. Let X be a metric space. An open cover U for X is a collection
(nite or innite) of open sets whose union is all of X. A Lebesgue number for
U is a number > 0 such that every open ball of radius is a subset of some
member of U .
Theorem A5.8. (Lebesgue) Every open cover of a (sequentially) compact metric
space has a Lebesgue number.
Proof. Suppose that U does not have a Lebesgue number. Then for every n
there is x
n
X such that B(x
n
; 1/n) is contained in no member of U . If X is
compact, the sequence (x
n
) has a subsequence that converges, say to x. Now x
belongs to some member U of U , since U is a cover. Thus there is > 0 such
that B(x; ) U. There is n > 2/ such that d(x
n
, x) < /2. But then
B(x
n
; 1/n) B(x
n
; /2) B(x; ) U
which is a contradiction.
16
Proposition A5.9. The following conditions on a metric space X are equivalent.
(a) Every sequence in X has a Cauchy subsequence.
(b) For every > 0 there is a nite cover of X by balls of radius .
In this case we say X is totally bounded (some writers say precompact).
Proof. (a) implies (b): Choose x
1
X, and then inductively choose x
n
X
such that d(x
n
, x
m
) when m < n, so long as this is possible. The process
must terminate because if it didnt it would produce a sequence with no Cauchy
subsequence. When it does terminate, with x
N
say, it does so because the balls
B(x
n
; ), n = 1, . . . , N, cover X.
(b) implies (a): Notice the following implication of (b): given any > 0, any
sequence in X has a subsequence all of whose members are separated by at most
(call this a -close subsequence).
Let (x
n
) be any sequence in X. Let (x
1
n
) be a 2
1
-close subsequence of (x
n
),
let (x
2
n
) be a 2
2
-close subsequence of (x
1
n
), and so on by induction. Then (x
n
n
) is
a Cauchy subsequence of the original sequence.
A metric space X is said to be (covering) compact if every open cover of X
has a nite subcover.
Proposition A5.10. The following conditions on a metric space X are equivalent:
(a) X is sequentially compact;
(b) X is complete and totally bounded;
(c) X is covering compact.
Proof. It is easy to see that (a) and (b) are equivalent.
Suppose (a). Let U be an open cover of X. Let > 0 be a Lebesgue number
for U (which exists because of Theorem A5.8). Since X is totally bounded it has
a nite cover by -balls. But each such ball is a subset of a member of U ; so U
has a nite subcover.
In the other direction, suppose (c). Let (x
n
) be a sequence without convergent
subsequence. Then for each x X there is some
x
such that x
n
/ B(x;
x
) for
all but nitely many n. The B(x;
x
) form a cover of X. Picking a nite subcover
we obtain the contradiction that x
n
/ X for all but nitely many n.
17
Remark A5.11. For general topological spaces, the notions covering compact
and sequentially compact are not equivalent, and covering compact is usually
the most appropriate one.
Exercise A5.12. A map f : X Y between metric spaces is uniformly contin-
uous if for each > 0 there is > 0 such that d(f(x), f(x

)) < whenever
d(x, x

) < . (The extra information beyond ordinary continuity is that does


not depend on x.) Show that if X is compact, every continuous f is uniformly
continuous. Hint: use Theorem A5.8.
18
Lecture A6
Compactness and Completeness
(The lecture will begin with a review of the basic results about compactness,
from the previous section).
Denition A6.1. Let X be a compact metric space. Then C(X) denotes the space
of all continuous functions X C. It is a metric space equipped with the metric
d(f, g) = sup[f(x) g(x)[ : x X.
(Compare Example A2.10.)
Compactness of X ensures that the supremum exists (why)? We can also
consider the space C
R
(X) of continuous real-valued functions.
Proposition A6.2. For a compact X, the metric spaces C(X) and C
R
(X) are
complete.
Proof. Most completeness proofs proceed in the same three-stage way: Given a
Cauchy sequence, identify a candidate for its limit; show that the candidate is in
the space in question; and show that the sequence approaches the candidate limit
in the metric of the space in question.
Let (f
n
) be a Cauchy sequence in C(X). Then, for each x X, f
n
(x) is a
Cauchy sequence in C. Since C is complete this sequence converges. Denote its
limit by f(x).
We show that f is a continuous function on X. Fix x

X and let > 0


be given. Because (f
n
) is Cauchy there is N such that for n

, n

N we have
[f
n
(x) f
n
(x)[ < /3 for all x. Take n

= N and let n

to nd that
[f
N
(x) f(x)[ /3 for all x. Now, f
N
is continuous at x

so there is > 0 such


that [f
N
(x) f
N
(x

)[ < /3 whenever [x x

[ < . It follows that, whenever


[x x

[ < ,
[f(x) f(x

)[
[f(x) f
N
(x)[ +[f
N
(x) f
N
(x

)[ +[f
N
(x

) f(x

)[ < .
This shows that f is continuous.
Finally to show that f
n
f in C(X) we must prove that for every > 0
there is N such that [f
n
(x) f(x)[ < for all x whenever n > N. But in fact we
already proved this in the previous paragraph.
19
It is an interesting and important question what are the compact subsets of
the complete space C(X). The answer is given by the Ascoli-Arzela theorem.
Closed and bounded is not enough.
20
Lecture A7
Applications of completeness
Recall that a metric space X is said to be complete if every Cauchy sequence
in X converges. The space (0, 1) is not complete, whereas the homeomorphic
space R is complete. This proves that completeness is not a topological property,
i.e., is not preserved under homeomorphism.
Exercise A7.1. A uniform homeomorphism between metric spaces is a homeo-
morphism f : X Y such that both f and f
1
are uniformly continuous. Show
that if X is complete and there is a uniform homeomorphism X Y , then Y is
complete.
Proposition A7.2. Let X be a complete metric space and let U X be an open
set. Then U is homeomorphic to a complete metric space (it is topologically
complete).
Proof. Let f : U R be the function dened by
f(x) =
1
infd(x, y) : y X U
.
This is a well-dened, continuous function with the property that f(x
n
)
whenever x
n
is a sequence in U that converges to some x X U.
Now let V be the metric space with the same points as U but with the metric
d
V
(x, x

) = d(x, x

) +[f(x) f(x

)[.
It is easy to see that this is indeed a metric. Moreover, it is complete: if (x
n
) is
a Cauchy sequence for the metric d
V
, then it is also a Cauchy sequence for the
metric d (so it converges in X, say to x X) and, in addition, the values f(x
n
)
are bounded (so the limit x in fact belongs to V ).
Finally I claim that the identity map U V is a homeomorphism. Suppose
x U and let > 0 be given. Since f is continuous, there is
1
> 0 such that
d(x, x

) <
1
implies [f(x) f(x

)[ < /2. Put = min


1
, /2). Then if
d(x, x

) < , we have
d
V
(x, x

) <

2
+[f(x) f(x

)[ < .
So the identity map U V is continuous, and the continuity of the inverse is
easy.
21
Remark A7.3. A subset of a metric space is called a G

-set (German: Gebeit-


Durschnitt) if it is the intersection of countably many open sets. For example, the
irrational numbers form a G

subset of R. Generalizing the above construction,


it can be shown that any G

subset of a complete metric space is topologically


complete. In particular, the topology on the irrational numbers can be given by a
complete metric. But the same is not true for the rational numbers, as will follow
from the Baire category theorem.
Remark A7.4. A countable union of closed sets is called a F

-set (French: ferm e,


somme). This is the only known example of dual mathematical concepts being
described using dual languages.
Denition A7.5. Let X be a metric space. A mapping f : X X is a (strict)
contraction if there is a constant a < 1 such that d(f(x), f(x

)) ad(x, x

) for
all x, x

X.
Note that a contraction must be continuous.
Theorem A7.6. (Banach) A contraction on a complete metric space has a unique
xed point (a point x such that f(x) = x).
Proof. A xed point is unique because if x, x

are two such then d(x, x

) =
d(f(x), f(x

)) ad(x, x

), which implies d(x, x

) = 0.
To prove existence, start with any x
0
X and dene x
1
= f(x
0
), x
2
= f(x
1
)
and so on. If d(x
0
, x
1
) = r then d(x
n
, x
n+1
) a
n
r and so
d(x
n
, x
n+k
) (a
n
+ + a
n+k1
)r
a
n
r
1 a
which tends to 0 as n . Thus (x
n
) is a Cauchy sequence, which converges to
a point x. We have
f(x) = limf(x
n
) = limx
n+1
= x
so x is a xed point.
Exercise A7.7. Banachs xed point theorem can be extended as follows: if
f : X X is a map and some power f
N
= f f is a strict contraction, then
f has a unique xed point. Prove this.
Banachs xed point theorem is one of the most important sources of existence
theorems in analysis. Another important source of existence theorems, which also
uses completeness, is the Baire category theorem.
22
Denition A7.8. A subset of a metric space is dense if its closure is the whole
space
Example A7.9. The rational numbers Qare dense in R. (It is often important that
R has a countable dense subset. A space with this property is called separable.)
Theorem A7.10. (Baire) In a complete metric space, the intersection of countably
many dense open sets is dense.
Proof. Let x X and let > 0. Let U
n
, n = 1, 2, . . ., be dense and open.
Choose x
1
U
1
B(x; /2). There is r
1
such that B(x
1
; r
1
) U
1
; without loss
of generality take r
1
< /4. Since U
2
is dense there is x
2
U
2
B(x
1
; r
1
).
Proceed inductively in this way choosing x
n
and r
n
such that
x
n
U
n
B(x
n1
; r
n1
), B(x
n
; r
n
) U
n
, r
n
< r
n1
/2.
Then (x
n
) is a Cauchy sequence, say converging to a point x

. Since x
n

B(x
j
; r
j
) for all j n, the point x

belongs to B(x
j
; r
j
) for every j, and hence to

U
n
. Moreover, d(x, x

) < . Since x and were arbitrary,

U
n
is dense.
Remark A7.11. The name category theorem comes from the following (tradi-
tional) terminology. A set A is nowhere dense if the complement of its closure is
dense. A set is of rst category if it is the countable union of nowhere dense sets.
Baires theorem then says that in a complete metric space the complement of a set
of rst category (sometimes called a residual set) is dense.
Remark A7.12. Notice that this gives us another proof that Ris uncountable (com-
pare Exercise A1.5). For the complement of a point in Ris a dense open set. More
generally, any complete, perfect (Denition A4.14) metric space is uncountable.
For in a perfect metric space, the complement of any point is dense.
Exercise A7.13. Consider the complete metric space C
R
[0, 1] of continuous real-
valued functions on [0, 1]. For a, b [0, 1] show that the set of functions f
which are nondecreasing on the interval [a, b] is nowhere dense in C
R
[0, 1]. Using
Baires theorem, deduce that there exist functions f C
R
[0, 1] that are nowhere
monotonic, that is, monotonic on no subinterval of [0, 1].
23
Lecture A8
Normed Spaces
Denition A8.1. Let V be a vector space (real or complex). A norm on V is a
function V R, denoted v |v|, such that
(a) |v| 0 for all v, and |v| = 0 iff v = 0.
(b) |v| = [[|v|, where is a scalar (real or complex).
(c) |v + v

| |v| +|v

|.
Clearly, d(v, v

) = |v v

| is then a metric. Because of (b), this metric has


an afne structure not present in a general metric.
Example A8.2. Let x = (x
1
, . . . , x
n
) be a vector in R
n
or C
n
. The expressions
|x|
1
= [x
1
[ + +[x
n
[
|x|
2
=
_
[x
1
[
2
+ +[x
n
[
2
_
1/2
|x|

= max[x
1
[, . . . , [x
n
[
all dene norms.
Exercise A8.3. Let a and b be positive real numbers and let p > 1. Show that
inft
1p
a
p
+ (1 t)
1p
b
p
: t (0, 1) = (a + b)
p
.
Hence (or otherwise) show that the expression
|x|
p
= ([x
1
[
p
+ +[x
n
[
p
)
1/p
is also a norm on R
n
or C
n
.
Example A8.4. If V = C(X), we may dene
|f| = sup[f(x)[ : x X.
This is a norm, and it gives rise to the usual metric on C(X). Since it is complete
(theorem A6.2) it is a Banach space as dened below.
24
Denition A8.5. If a normed vector space is complete in its metric, it is called a
Banach space.
Proposition A8.6. A linear mapping T : V W between normed vector spaces
is continuous if and only if there is a constant k such that
|Tv| k|v|
(one then says that T is bounded).
Proof. If T is bounded then |Tu Tv| k|u v| so T is continuous.
If T is continuous then there is some > 0 such that |u| < implies |Tu| <
1. Take k = 1/.
Denition A8.7. The best constant k in the above proposition, that is the quantity
sup|Tv| : |v| 1
is called the norm of T and denoted |T|.
Exercise A8.8. Showthat, with the above norm, the collection /(V, W) of bounded
linear maps from V to W is a normed vector space, and that it is a Banach space
if W is a Banach space.
Exercise A8.9. Showthat the normof linear maps is submultiplicative under com-
position, i.e. |S T| |S||T|.
25
Lecture A9
Differentiation in Normed Spaces
The basic idea of differentiation is that of best linear approximation. Normed
spaces provide a systematic way to express this.
Notation A9.1. Let f be a function dened on some ball B(0; ) in a normed
space V and having values in another normed space W. We shall write f(h) =
o(|h|) to mean that the limit lim
h0
_
|h|
1
|f(h)|
_
exists and equals zero. Sim-
ilarly for expressions like f(h) = g(h) + o(h).
Exercise A9.2. Suppose that T : V W is a linear map. Show that T(h) =
o(|h|) if and only if T = 0.
Now let f : V W be a continuous (need not be linear) map between normed
vector spaces. In fact, we need only suppose f is dened on some open subset U
of V .
Denition A9.3. With above notation, let x V . We say that f is differentiable
at x if there is a bounded linear map T : V W such that
f(x + h) = f(x) + T h + o(|h|).
By the exercise, T is unique if it exists. It is called the derivative of f at x and
written Df(x).
Example A9.4. If V = W = R then every linear map V W is multiplication
by a scalar, i.e. we identify /(V, W) = R. Under this identication our denition
of the derivative corresponds to the usual one from Calculus I.
Example A9.5. If V = R
m
and W = R
n
then /(V ; W) is the space of n m
matrices. The matrix entries of the derivative of f are the partial derivatives of the
components of f as dened in Calculus III. Remark: The existence of the partial
derivatives does not by itself imply differentiability in the sense of our denition
above. This just shows that our denition is right and partials are wrong!
Example A9.6. If V = R and W is arbitrary, the space of (bounded) linear maps
/(V ; W) can be identied with W itself. Under this identication the derivative
of f is given by the usual formula
Df(x) = lim
h0
f(x + h) f(x)
h
.
26
Exercise A9.7. Show that the derivative of a linear map is always equal to the
map itself.
Proposition A9.8. (Mean value theorem) Suppose that the function f is differ-
entiable throughout a ball B(x
0
; r) V , and that |Df(x)| k for all x
B(x
0
; r). Then
|f(x) f(x
0
)| k|x x
0
|
for all x B(x
0
; r).
Proof. Let > 0. By denition of the derivative each y B(x
0
; r) has a neigh-
borhood U
y
B(x
0
; r) such that
|f(y

) f(y)| (k + )|y

y| y

U
y
.
Now we will creep along the line segment [x
0
, x] B(x
0
; r). Let x
t
= (1
t)x
0
+ tx, t [0, 1], and let T be the set of all those t [0, 1] such that |f(x
s
)
f(x
0
)| (k+)|x
s
x
0
| for all s t. Clearly, T is an interval and the displayed
equation above shows that it contains [0, ) for some > 0. Let = sup T
(0, 1] and let y = x

. If < 1 then there exists > 0 such that


|f(x
t
) f(y)| (k + )|x
t
y| t ( 2, + 2).
For any c [0, 1] we have c T and so
|f(x
+c
) f(x
0
)|
|f(x
+c
) f(x

)| +|f(x

) f(x
c
)| +|f(x
c
) f(x
0
)|
(k + ) (|x
+c
x

| +|x

x
c
| +|x
c
x
0
|) =
= (k + ) (|x
+c
x
0
|) .
The nal equality is because the three vectors appearing here are positive multiples
of the same vector. If follows that + T, contradicting the denition of as the
supremum. Therefore it is impossible that < 1 and the theorem is proved.
Remark A9.9. From the proof one sees that the theorem is true if we replace the
ball by any region that is star-shaped about x
0
.
It is easy to see that the derivative of f g is Df Dg. Another familiar
calculus result that generalizes easily is the chain rule.
27
Proposition A9.10. (Chain rule) Let V, W, X be normed vector spaces, and let
x V . Let f : V W be differentiable at x and let g : W X be differentiable
at y = f(x). Then g f is differentiable at x and
D(g f)(x) = Dg(y) Df(x).
Note that f need only be dened near x, and g need only be dened near y.
Proof. Let k(h) = f(x + h) f(x) = Df(x) h + o(|h|). Note that there is a
constant A such that |k(h)| A|h| for small |h|. Now write
(g f)(x + h) (g f)(x) = g(y + k(h)) g(y)
= Dg(y) k(h) + o(|k(h)|)
= Dg(y) Df(x) h + o(|h|)
giving the result.
Remark A9.11. Note that the derivative of f, Df, is itself a function having values
in a normed space (namely the space /(V, W)). Thus we can differentiate it and
dene second and higher derivatives if we require.
Exercise A9.12. (Open-ended) Formulate precisely the advanced calculus result
that the mixed derivatives are symmetric, that is
2
f/xy =
2
f/yx, and
prove it (under suitable hypotheses) for maps between normed vector spaces.
Let F be a function R V V , where V is a normed vector space. Let
[a, b] R. A function f : [a, b] V is a solution to the differential equation
dened by F if it is differentiable and f

(t) = F(t, f(t)) for all t [a, b]. The


value f(a) V is called the initial condition for the solution.
Theorem A9.13. Suppose that V is a Banach space and suppose also that the
function F is continuous and satises a Lipschitz condition, that is, there exists a
constant C > 0 such that
|F(t, v) F(t

, v

)| C|v v

|.
Then for any a R and any v
0
V there exists a nontrivial interval [a, b] on
which there is a solution to the differential equation f

(t) = F(t, f(t)) with initial


condition f(a) = v
0
; moreover, this solution is unique.
28
Proof. We need to use some facts about integration for V -valued functions. It
isnt the job of this course to teach you about integration theory for that see
Math 501 so I will just state the various facts that are needed. Moreover, in a
rst course on integration these facts are usually proved only for real or complex
valued functions; you will need to take it on faith that the same facts are true for
functions having values in a normed vector space V . Finally, we shall use the
fact that if V is a Banach space (that is, complete) then the space of continuous
functions [a, b] V , equipped with the sup norm, is also complete. This is a fact
that we have already proved for real or complex valued functions, and the proof
in the general case is exactly the same.
We will use the fundamental theorem of calculus: a function f C[a, b] is a
solution of the differential equation if and only if it is a solution of the equivalent
integral equation
f(s) = v
0
+
_
s
a
F(t, f(t)) dt, t [a, b].
We will show that a solution to this equation exists (for b sufciently close to a)
by applying Banachs contraction mapping theorem (A7.6) to the space C[a, b] of
continuous functions on [a, b] (with values in V , if you like). Let I : C[a, b]
C[a, b] be dened by the right hand side of the display above, that is,
(If)(s) = v
0
+
_
s
a
F(t, f(t)) dt, t [a, b].
If f, g C[a, b] then the integrands in If and Ig differ by C|f g| at most,
where C is the Lipschitz constant, and therefore
|If Ig| C(b a)|f g|
by standard estimates on integrals. Therefore, if (b a) < C
1
, the map I is a
strict contraction and Banachs xed point theorem provides a unique solution to
If = f, which is a solution to the differential equation with the specied initial
condition.
29
Lecture A10
The inverse function theorem
Let V, V

be normed vector spaces. The space /(V, V

) is then a normed
vector space also. Say T /(V, V

) is invertible if it has an inverse which is a


bounded linear map from V

to V .
Proposition A10.1. Let V, V

be Banach spaces. Then the invertibles form an


open subset of /(V, V

), and the inverse operation T T


1
is continuous (on
this subset) from /(V, V

) to /(V

, V ).
Proof. Without loss of generality V

= V . Look rst at a nbhd of the invertible


operator I. Suppose |S| < 1. For y V consider the map

y
: V V, v y Sv.
Since |S| < 1 this map is a strict contraction, so it has a unique xed point x
(Theorem A7.6). This xed point is an x such that (I + S)x = y. By the triangle
inequality,
|y| (1 |S|)|x|, so |x| (1 |S|)
1
|y|
and the map y x is bounded. We have shown that if |S| < 1, I+S is invertible.
Moreover
|y (I + S)
1
y| = |S(1 + S)
1
y| |S|(1 |S|)
1
|y|
which shows that the map S (I + S)
1
is continuous at S = 0.
We have shown that the identity is an interior point of the set of invertibles
and that the inverse is continuous there. However, left multiplication by a xed
invertible is a homeomorphism from the set of invertibles to itself; so the same
results apply to any point of the set of invertibles.
Exercise A10.2. Let V be a normed space and let W be the normed space /(V, V ).
Show that the mapping i : T T
1
is differentiable (where dened) on W, and
that its derivative is
Di(T) H = T
1
HT
1
.
30
Denition A10.3. A map f from an open subset of a normed space V to a normed
space W is continuously differentiable or C
1
if it is differentiable (everywhere)
and the map x Df(x) is continuous.
The inverse function theorem says that under suitable conditions, if Df(x) is
invertible then f is locally invertible near x.
Theorem A10.4. Let f be as above, dened near x V , and suppose that f
is continuously differentiable and that Df(x) /(V, W) is invertible. Suppose
moreover that V, W are Banach spaces. Then there is an open set U containing x
such that f is a bijection of U onto f(U), which is an open set containing f(x),
and such that its inverse g : f(U) U is also continuously differentiable.
Proof. It is an application of Banachs xed point theorem A7.6. We will con-
struct the inverse function by looking at the xed points of a suitable map.
Let A = Df(x)
1
. For y W dene a map
y
: V V by

y
(z) = z + A (y f(z)).
A xed point of
y
is a solution to f(z) = y. We have
D
y
(z) = I A Df(z) = A
_
Df(x) Df(z)
_
.
Since Df is continuous there is r > 0 such that if z U = B(x; r) then
|D
y
(z)| <
1
2
. From the mean value theorem we conclude that if F is a closed
subset of U, and if we know for some reason that
y
maps F to F, then
y
is
actually a contraction of this complete metric space and so has a unique xed
point.
Fix z
0
U and let y
0
= f(z
0
). Thus y
0
f(U) and we want to prove
that there is some > 0 such that B(y
0
; ) f(U); this will show that f(U) is
open. Choose > 0 such that the closed ball B(z
0
; ) is contained in U. Choose
=
1
2
|A|
1
. If z B(z
0
; ) and y B(y
0
; ) then we may compute
|
y
(z) z
0
| = |
y
(z)
y
0
(z
0
)|
|
y
(z)
y
(z
0
)| +|
y
(z
0
)
y
0
(z
0
)|

1
2
|z z
0
| +|A(y y
0
)|
1
2
+
1
2
=
using the mean value theorem A9.8. We conclude that, for these y, the map
y
is
a contraction of the complete metric space B(z
0
; ) and thus has a (unique) xed
point there. We have shown that each y
0
f(U) has an -neighborhood contained
31
in f(U); thus, f(U) is open. The contraction property of the
y
ensures that each
y f(U) has only one inverse image in U; thus f is a bijection of U onto f(U).
Now to show that the inverse function g = f
1
: f(U) U is differentiable
at y = f(x) write g(y + h) g(y) = u(h), say. The calculations of the previous
paragraph show that |u(h)| 2|A||h| for |h| small. Then
h = f(g(y + h)) y = f(g(y) + u(h)) y = Df(x) u(h) + o(|h|).
Apply A = Df(x)
1
to get
u(h) = A h + o(|h|).
This gives differentiability of g with Dg(f(x)) = Df(x)
1
. Continuous differen-
tiability now follows from the continuity of the inverse operation on the space of
bounded linear maps (Proposition A10.1).
32
Lecture A11
Topological Spaces
We have been emphasizing that up to homeomorphism properties of metric
spaces depend only on their open sets. This motivates a further abstraction which
removes the numerical notion of metric entirely.
Denition A11.1. A topology on a set X is a family T of subsets of X with the
following three properties:
(i) If U

is any family of members of T , then the union

is also a member
of T ;
(ii) If U
1
, . . . , U
n
is a nite family of members of T then the intersection

n
i=1
U
i
is also a member of T ;
(iii) The sets and X are members of T .
A set equipped with a topology is called a topological space. The members of T
are called open subsets of the topological space X.
Notice that the open subsets of a metric space automatically satisfy (i)(iii);
thus, every metric space carries a topology (called its metric topology). Different
metrics can however give rise to the same topology (as we have already seen);
moreover, there are topologies that do not arise from any metric.
Example A11.2. The discrete topology on X has T = P(X): every subset is
open. This topology arises from the discrete metric.
Example A11.3. The indiscrete topology on X has just two open sets, and X.
This topology does not arise from a metric as soon as X has more than one point.
(Why not?)
Example A11.4. The conite topology on X consists of together with all the
conite subsets of X (a subset is conite if its complement is nite).
It is natural to say that a closed set is one whose complement is open, so that
the closed sets in the conite topology are just the nite sets together with the
whole space. Notice that this topology has many fewer open (or closed) sets
than the topologies with which we are familiar. For example, in the conite topol-
ogy on an innite set, any two nonempty open sets have a nonempty intersection.
A similar but more sophisticated example is the next one.
33
Example A11.5. Let K be an algebraically closed eld (for example, the complex
numbers) and let A = K
n
. A subset V of A is called an afne algebraic variety if
there is a set S K[X
1
, . . . , X
n
] of polynomials in n variables such that
(x
1
, . . . , x
n
) S p(x
1
, . . . , x
n
) = 0p S;
in other words, a variety is the set of common zeroes of a collection of polynomi-
als. (The Hilbert basis theorem says that S can always be taken to be nite.) It is
a simple exercise (for an algebraist!) to show that any intersection of varieties is a
variety, and a nite union of varieties is a variety. In other words, the varieties can
be considered as the closed sets of a topology on A, called the Zariski topology.
Example A11.6. Let T be a family of topologies on the set X. Then

T is
also a topology on X. In particular, let F be any family of subsets of X at all.
Let T be the family of all topologies T such that T F (there always is one
such topology, the discrete topology). The intersection

T is then the smallest


or weakest topology containing F. It is called the topology generated by F
(youll also see F called a subbasis for T ).
Exercise A11.7. Show that the topology on X generated by F can be concretely
expressed as follows: a set U is open in the generated topology if and only if it
can be written as a union

of some family of sets U

, where each U

is the
intersection of some nite family (depending on ) of members of F. (The empty
set and the whole space X are included among these sets by convention: we take
the intersection of an empty family to be X and the union of an empty family to
be .)
Remark A11.8. The following language is conventional in this context: the fam-
ily B of subsets of X is called a basis if, for every point x X, there is some
member of B containing x and every nite intersection of members of B con-
taining x includes some member of B containing x. Thus the topology generated
by a basis is just the collection of all unions of families of members of the basis.
(For example, the metric balls form a basis for the topology of a metric space.)
Moreover, the topology generated by a subbasis F is the topology generated by
the associated basis consisting of nite intersections of members of F.
34
Lecture A12
Compactness and Hausdorffness
Denition A12.1. Let X be a topological space, with topology T , and let Y X.
Then the family of subsets of Y given by Y U : U T denes a topology
on Y . It is called the relative topology or subspace topology induced by the given
topology on X.
(Compare Proposition A3.10.)
Any metric space notion which is dened simply in terms of open sets can
be taken over without change to the more general context of topological spaces.
Thus, for example, we can dene the interior, closure, and boundary of a subset
just as in the metric space context.
Exercise A12.2. Let X be a topological space and Y a subspace of X, equipped
with the subspace topology. Let S be a subset of X, with closure S. Show that
the closure of S Y in the subspace topology of Y is a subset of S Y , and that
these two sets are equal if S itself is a subset of Y ; but that they need not be equal
in general.
Just as in metric spaces we can also dene a continuous function f : X Y
(X and Y being topological spaces) as one which has the property that, for every
open subset U of Y , the inverse image f
1
(U) is open in X. The proof of the
following lemma is then obvious:
Lemma A12.3. The composite of continuous functions is continuous.
Exercise A12.4. Let f : X Y be a map of spaces (from now on, space will
mean a topological space, unless otherwise specied). Show that f is continuous
as a map from X to Y if and only if it is continuous as a map from X onto f(X),
where f(X) Y is equipped with its subspace topology.
Denition A12.5. A homeomorphism is a continuous bijection whose inverse is
also continuous.
Denition A12.6. A space X is compact if every open cover of X has a nite
subcover.
35
This is the same as the denition of covering compact that was given in
Lecture 5. We wont make use of the notion sequentially compact outside the
realm of metric spaces.
Example A12.7. Any nite topological space is compact. A discrete topological
space is compact if and only if it is nite.
Proposition A12.8. Let f : X Y be a continuous map. If X is compact, then
f(X) Y is also compact.
Proof. By Exercise A12.4 it sufces to consider the case where f(X) = Y . Let
U be an open cover of Y . The sets f
1
(U), U U , are then open in X and
they form a cover since for each x X, the image f(x) belongs to some member
of U . Since X is compact, there is a nite subcover: there is a nite subset
U
1
, . . . , U
n
of U such that the sets f
1
(U
i
) cover X. But for every y Y , there
is x X such that f(x) = y; so x f
1
(U
i
) for some i and therefore y U
i
.
That is, the U
1
, . . . , U
n
cover Y and we have shown that U has a nite subcover
as required.
Proposition A12.9. Let X be a compact space and let Y be a closed subset of X.
Then Y is also compact (in its relative topology).
Proof. We must show that every open cover of Y (in the relative topology) has a
nite subcover. Let U be such a cover. Each open set U U is (by denition
of the relative topology) of the form V
U
Y , where V
U
is some open subset of
X. The union of all the sets V
U
includes Y ; therefore, the open sets V
U
, together
with the open set X Y , form an open cover of X. By hypothesis this cover has
a nite subcover, comprising V
U
1
, . . . , V
Un
, together possibly with X Y . Then
U
1
, . . . , U
n
cover Y .
In metric spaces, compact subsets are necessarily closed (Proposition A5.5).
This need not be true in general topological spaces. Indeed, let X be any topo-
logical space. From Example A12.7, any nite subset of X in particular, any
one-point subset must be compact in its subspace topology. But we saw several
examples in the last lecture of topological spaces in which one-point subsets are
not closed.
What is missing is the Hausdorff property.
Denition A12.10. A topological space X is Hausdorff if, for any two distinct
points x, y X, there exist open sets U, V with x U, y V , and U V = .
36
Any metric space is Hausdorff (take U = B(x, ) and V = B(y, ) for 0 <
< d(x, y)/2). The Hausdorff property ensures a large supply of open sets.
Proposition A12.11. Let X be a Hausdorff space and let Y X be compact (in
the subspace topology). Then Y is a closed subset of X.
Proof. We will show that X Y is open. Let x X Y . For each y Y there
exist disjoint open sets U
y
containing x and V
y
containing y, by the Hausdorff
property. The sets V
y
Y form an open cover of Y (in the subspace topology)
which therefore has a nite subcover, say V
y
1
Y, . . . , V
yn
Y . Let U
x
= U
y
1

U
yn
. Then U
x
is an open set, contains x, and does not meet Y . The set XY
itself is the union of all these open sets U
x
, so it is open.
In particular, a one-point subset of a Hausdorff space is closed.
Corollary A12.12. Let f : X Y be a continuous bijection, where X is a com-
pact space and Y is a Hausdorff space. Then f
1
is continuous (i.e., f is a
homeomorphism).
Proof. To show that f
1
is continuous it is enough to show that for any closed
subset K of X, the image f(K) is closed in Y . But K is compact (Proposi-
tion A12.9); therefore f(K) is compact (Proposition A12.8); therefore f(K) is
closed (Proposition A12.11).
37
Lecture A13
Regular and Normal Spaces
Let A and B be disjoint subsets of a topological space X. One says that A and
B are separated by open sets if there exist disjoint open sets U and V such that
A U, B V . Thus the Hausdorff property says that any two distinct points
can be separated by open sets.
This is in fact just one (the most important one) of a series of separation prop-
erties that a topological space may have.
Denition A13.1. A Hausdorff topological space X is regular if disjoint subsets
A and B of X can be separated by open sets whenever A is a point and B is a
closed set. It is normal if any two disjoint closed subsets can be separated by open
sets.
Remark A13.2. A regular Hausdorff space is also called a T
3
space and a normal
Hausdorff space is a T
4
space; a plain vanilla Hausdorff space is a T
2
space. The
T stands for the German Trennungsaxiom (separation axiom). There are also T
0
,
T
1
, T
2
1
2
, T
3
1
2
, T
5
, and T
6
spaces, and maybe some others that I forgot.
Example A13.3. Every metric space is regular, and indeed normal. To see this we
need to make use of the distance function from a closed set. Let (X, d) be a metric
space and let K be a closed subset. Dene a real valued function d
K
: X R
+
by
d
K
(x) = infd(x, k) : k K.
It is easy to see that d
K
is a continuous function and that d
K
(x) = 0 iff x K.
Now let Aand B be disjoint closed sets in X and consider the continuous function
f(x) = d
B
(x) d
A
(x) on X. The sets U = f
1
((0, )) and V = f
1
((, 0))
are open (by continuity), disjoint, and A U, B V . This shows that X is
normal.
Proposition A13.4. Every compact Hausdorff space is regular, and indeed nor-
mal.
Proof. Its best to do this proof in two installments: rst regularity, then normality.
Let X be a compact Hausdorff space, p a point of X, and B a closed subset not
containing p. Then B is compact (Proposition A12.9). For each x B there
38
are disjoint open sets U
x
containing p and V
x
containing x. The V
x
form an open
cover of B, so there is a nite subcover, say V
x
1
, . . . , V
xn
. Then
U
p
= U
x
1
U
xn
, V
p
= V
x
1
V
xn
are disjoint open sets, the rst containing p, the second including V . Thus X is
regular.
To prove normality, we iterate the argument. Let A and B be disjoint closed
(hence compact) sets. By regularity, for each p A there are disjoint open U
p
containing p and V
p
including B. The U
p
form an open cover of A and have a
nite subcover say U
p
1
, . . . , U
pn
. Then
U = U
p
1
U
pn
, V = V
p
1
V
pn
are disjoint open sets including A and B respectively.
In Example A13.3 we used suitably chosen continuous functions to establish
regularity, and indeed normality, for metric spaces. Its also possible to proceed
in the other direction: starting with normality, construct a supply of continuous
real-valued functions.
Lemma A13.5. Let X be a normal Hausdorff space, F a closed subset of X, and
W an open subset of X including F. Then there is an open subset V of X such
that F V V W.
Proof. Let E = X W. The closed sets E and F are disjoint, so by normality
there are disjoint open sets U and V with E U and F V . Since V X U
which is closed, V XU also. In particular, V doesnt meet E, so it is included
in W.
Theorem A13.6. (Urysohns lemma) Let E, F be disjoint closed subsets of a nor-
mal Hausdorff space X; then there exists a continuous function f : X [0, 1]
such that f(x) = 0 for x E, f(x) = 1 for x F.
Warning: It is not asserted that the sets E, F are exactly equal to f
1
0,
f
1
1 respectively; in general this stronger statement is not true. We will use the
following exercise.
Exercise A13.7. In order that a function f : X R be continuous, it is necessary
and sufcient that for every x X and every > 0 there exists an open set W
x
containing x such that
f(W
x
) (f(x) , f(x) + ).
39
Proof. Its an induction based on lemma A13.5. Set U
1
= X F; then U
1
is
an open set containing E. We are going to construct by induction a collection
of open sets U
t
parameterized by dyadic rational numbers t (0, 1], having the
properties that U
r
U
s
whenever r < s and E U
t
for all t. At the nth stage
of the induction we assume that the sets U
t
have been constructed for t = m2
n
,
m = 1, . . . , 2
n
. Let us make the convention that U
0
refers to the closed set E
(even though we have not dened a set U
0
). By lemma A13.5, for each m there is
an open set V such that
U
m2
n V V U
(m+1)2
n;
call the set V so dened U
(m+
1
2
)2
n. We have now dened the sets U
t
for all
t = m

2
(n+1)
and the induction continues.
Now dene f : X [0, 1] by f(x) = inft : x U
t
1. Clearly f
maps X [0, 1] with f(E) = 0 and f(F) = 1, so we must show that f is
continuous, and we will use Exercise A13.7. Let x X with f(x) = t; suppose
that 0 < t < 1 (the cases t = 0 and t = 1 are handled by similar arguments). Let
> 0. There exist dyadic rationals r and s such that
t < r < t < s < t + .
Put W
x
= U
s
U
r
. Then W
x
is an open set and f(W
x
) [r, s] (t , t +) as
required.
Of course, [0, 1] can be replaced in the statement of Urysohns lemma by any
closed interval [a, b].
40
Lecture A14
Spaces of Continuous Functions
Let X be a set, Y X, and let f : Y Z be a function. A function g : X
Z is an extension of f if g(y) = f(y) for all y Y ; in other words, if f = g
|Y
.
When X, Y, Z are topological spaces we will naturally be interested in continuous
f and g.
Theorem A14.1. (Tietze extension theorem) Let X be a normal Hausdorff space,
and let Y be a closed subspace. Let f : Y R be a bounded continuous func-
tion. Then f can be extended to a bounded continuous function g : X R, with
sup [g[ = sup [f[.
Proof. Let us temporarily call a bounded continuous function h: Y R ex-
tendible if it extends to a bounded continuous function X R with the same
sup norm. What we have to prove is that all bounded continuous functions are
extendible.
If h: Y [c, c] is a continuous function, let E and F be the disjoint subsets
h
1
([c, c/3]) and h
1
([(c/3, c]) of Y . They are closed in Y (by continuity),
and Y is closed in X, so they are closed in X (Exercise A12.2). Thus by Urysohns
lemma there is a continuous function k: X [c/3, c/3] with k(E) = c/3,
k(F) = c/3. Notice that the supremum norm|hk| is at most 2c/3. We have
proved the following statement: for each bounded continuous h: Y R there
exists an extendible bounded continuous k: Y R with |k|, |hk| 2|h|/3.
Now we proceed inductively. Let f : Y Rbe given and suppose inductively
that extendible functions g
1
, . . . , g
n
have been constructed with |g
i
| (2/3)
i
|f|
and
|f g
1
. . . g
n
| (2/3)
n
|f|
. Apply the italicized claim to h = f g
1
. . . g
n
and let g
n+1
be the k that is
constructed. The function g
n+1
then has the desired properties so the induction is
completed.
By the Weierstrass M-test the series

i=1
g
i
converges uniformly on X to a
function g. But the display above shows that, when restricted to Y , f = g. Thus
f is extendible as required.
Recall that a subset of a topological space is dense if its closure is the whole
space. A classical method in analysis is to prove a result rst for a dense subset of
some space, and then to extend it to the whole space by continuity.
41
How can we nd dense subsets of C(X), the continuous real-valued functions
on a compact Hausdorff space X?
Denition A14.2. Acollection Lof continuous real-valued functions on a set X is
a lattice if, whenever it contains functions f and g, it also contains their pointwise
maximum and minimum usually written f g and f g in this context.
Proposition A14.3. (Stone) Let L be a lattice of continuous functions on a com-
pact space X. If, for all x, x

X and any a, a

R, there is a function f L
having f(x) = a and f(x

) = a

, then L is dense in C(X).


The property appearing in the statement is called the two point interpolation
property.
Proof. Let h C(X) be given and let > 0. We are going to approximate h
within by elements of L. By hypothesis, for each x, x

X there exists f
xx
L
such that f
xx
(x) = h(x) and f
xx
(x

) = h(x

). Fixing x for a moment, let


V
x
= y X : h(y) < f
xx
(y).
These sets are open (why?) and x

V
x
so they cover X. Take a nite subcover
and let g
x
be the (pointwise) maximum of the corresponding functions f
xx
L.
Because L is a lattice, g
x
L and by construction,
h(y) < g
x
(y) y, h(x) = g
x
(x).
So we have approximated h from one side by members of L.
Now we play the same trick again from the other direction: let
W
x
= y : g
x
(y) < h(y) + .
Again these form an open cover of X; take a nite subcover and let g be the
(pointwise) minimum of the corresponding g
x
. Then g L and by construction
h < g < h + , as required.
Exercise A14.4. Use Stones theoremto give another proof of the Tietze extension
theorem in the case of compact Hausdorff spaces. (Show that the collection of
extendible functions is a lattice. . . )
There is a more classical formulation which makes use of algebraic rather than
order-theoretic operations.
42
Lemma A14.5. There is a sequence of polynomials p
n
(x) on [1, 1] converging
uniformly to [x[.
Proof. Writing [x[ =
_
1 + (x
2
1), this follows from the fact that the binomial
series for (1 + t)
1/2
converges uniformly for t [1, 1].
One says that a subset of C(X) is a subalgebra if it is closed under pointwise
addition, subtraction, multiplication of functions, and multiplication by scalars.
Lemma A14.6. A closed subalgebra of C
R
(X) that contains the constant func-
tions is a lattice.
Proof. Let A be the given subalgebra and f, g A; let h = f g. There is
no loss of generality in assuming (by rescaling) that [h[ 1 everywhere. Any
polynomial in h belongs to A and hence so does [h[ = limp
n
(h), by closure and
lemma A14.5. But now
f g =
f + g [h[
2
, f g =
f + g +[h[
2
belong to A as well.
A subalgebra of C(X) is said to separate points if for every x, x

X there is
a function in the subalgebra taking different values at x and at x

.
Theorem A14.7. (Stone-Weierstrass) A subalgebra of C
R
(X) which contains the
constants and separates points is dense.
Proof. The closure of the subalgebra is a closed subalgebra, contains the con-
stants, and separates points. Using the algebraic operations, it has the two point
interpolation property. The previous lemma shows that it is a lattice. Hence by
Stones theorem A14.3 it is all of C(X).
The original result of Weierstrass was
Corollary A14.8. (Weierstrass) Every continuous function on a closed bounded
interval is the uniform limit of polynomials.
Proof. The polynomials form an algebra which contains the constants and sepa-
rates points.
43
Exercise A14.9. In the complex case the Stone-Weierstrass theorem takes the
following form: a -subalgebra of C(X) which contains the constants and sepa-
rates points is dense, where the -condition means that the algebra is closed under
(pointwise) complex conjugation. Prove this. Try to give an example to show that
the complex theorem is not valid without the extra condition (we shall discuss this
in detail later).
44
Lecture A15
Connectedness
What does it mean that a topological space is all in one piece?
Proposition A15.1. Let X be a topological space. The following properties of X
are equivalent:
(a) The only subsets of X that are both open and closed are the empty set and X
itself.
(b) The only continuous functions from X to a discrete topological space are
constant.
(c) The only equivalence relation on X with open equivalence classes is the triv-
ial one (every point is equivalent to every other point).
Proof. It is clear that (b) and (c) are equivalent, since if f : X D is a continuous
function to a discrete space, then the relation x y iff f(x) = f(y) is an
equivalence relation with open equivalence classes, and conversely.
Assuming (c), if Ais an open-and-closed subset of X, then the relation x y
iff either none, or both, of x, y belong to A is an equivalence relation whose
equivalence classes A and X A are open. Conversely, assuming (a), let be
an equivalence relation with open equivalence classes, and let A be a nonempty
equivalence class. A is open; but, since the complement of A is a union of equiva-
lence classes and therefore open, it is also true that A is closed. Thus A = X and
the equivalence relation is trivial.
Denition A15.2. A topological space is connected if it satises the equivalent
properties of Proposition A15.1. A subset of a topological space is called a con-
nected subset if it is a connected space in the relative topology. A space (or subset)
that is not connected is called disconnected.
Proposition A15.3. Let f : X Y be continuous and surjective. If X is con-
nected, then Y is connected.
Proof. Let g : Y D be a continuous map to a discrete space D. Then f
g : X D is continuous, hence constant because X is connected. But now since
f is surjective, g must be constant itself.
45
Proposition A15.4. Let X be a space and let X

A
be a family of subsets
each of which is connected. If

,= then

is connected.
Proof. Let Y =

and let p

. Let f : Y D be a continuous map


to a discrete space. Since each X

is connected, the restriction of f to X

takes
a constant value (say d

D); since p X

for every , d

= f(p) for each .


Thus f takes the constant value f(p) on Y .
Proposition A15.5. Every interval in R is connected, and every connected subset
of R is an interval.
Proof. We begin by proving that every compact interval [a, b] is connected. Sup-
pose that is an equivalence relation with open equivalence classes; then the
equivalence classes form an open cover U for the compact interval [a, b]. By the
Lebesgue covering theorem (Theorem A5.8)there is > 0 such that any ball of
radius less than is included in some member of the cover. In particular, any two
points that are less than apart belong to the same equivalence class. But any
pair of points in the interval [a, b] can be joined by a chain of points whose suc-
cessive members are less than apart; so, any two points in [a, b] are equivalent.
This proves the connectedness of closed intervals; other kinds of intervals can be
written as the increasing union of suitable families of closed intervals, so we may
apply Proposition A15.4.
Conversely, let S be a subset of R that is not an interval. Then there is some
a / S such that S has members which are less than a and also members that are
greater than a. So, S (, a) and S (a, ) are disjoint open subsets of S
whose union is S; so S is not connected.
Corollary A15.6. (Intermediate value theorem) Let f : [a, b] R be continuous.
Then the range of f includes all values between f(a) and f(b).
Proof. The range of f, being the image of a connected space under a continuous
function, is connected. Hence it is an interval. It must contain f(a) and f(b) so,
being an interval, it must also contain everything in between.
If X is a space, the relation x y iff there is a connected subset of X
containing both x and y is an equivalence relation (by Proposition A15.4). The
equivalence classes are called the connected components of X. The connected
component of x is the largest connected subset of X that contains x.
Exercise A15.7. Show that the closure of a connected subset of X is connected,
and deduce that the connected components of X are closed. Give an example
where they are not open.
46
Exercise A15.8. The quasicomponent of x X is the intersection of all the
open-and-closed subsets that contain x. Show that the connected component of x
is included in the quasicomponent. Can you give an example where they are not
equal?
Denition A15.9. A path in a topological space X is a continuous map [0, 1]
X. The start of the path is the point (0) and the end of the path is the point (1).
We say that is a path from its start to its end.
Lemma A15.10. Let X be a topological space,
1
a path in X from a to b,
2
a
path in X from b to c. Then the map : [0, 1] X dened by
(t) =
_

1
(2t) (t
1
2
)

2
(2t 1) (t
1
2
)
is a path in X from a to c.
Proof. All that needs to be checked is the continuity of at t =
1
2
. Let U be
an open set in X containing b. By continuity of
1
, there is
1
> 0 such that if
[s1[ <
1
, s [0, 1] then
1
(s) U; by continuity of
2
there is
2
> 0 such that
if [s[ <
2
, s [0, 1] then
2
(s) U. Now let =
1
2
min
1
,
2
; if t
1
2
[ < ,
t [0, 1], then (t) U. This gives the required continuity.
Corollary A15.11. The relation There is a path in X from a to b is an equiva-
lence relation.
Proof. The lemma gives transitivity. Reexivity and symmetry are easy.
The equivalence classes under this equivalence relation are called path com-
ponents of X.
Proposition A15.12. Every path component is connected. Consequently, the path
component of x X is included in the component of x.
Proof. By Proposition A15.3, the image of a path from a to b is a connected set
containing a and b. Consequently, the path component of X is the union of a
family of connected sets all of which contain x; so it is connected by Proposi-
tion A15.4.
47
A space which has only one path component (i.e., any two points can be joined
by a path) is called path connected. By the preceding proposition, a path con-
nected space is connected. The converse is false, however, as is shown by the
following classical example.
Exercise A15.13. Let Z be the set of points (x, y) R
2
such that either x = 0
and y [1, 1] or 0 < x 1 and y = sin(1/x). Show that Z is a compact
connected metric space with 2 path components.
The same example shows that, in contrast to components, path components
need not be closed in general.
48
Lecture A16
Local Properties of Spaces
Let X be a topological space.
Denition A16.1. An open neighborhood of x X is an open subset of X that
contains x. A neighborhood of x X is a subset that includes an open neighbor-
hood of x, i.e. contains x in its interior.
Suppose that P is some kind of topological property (e.g. connectedness).
Denition A16.2. We say that a space X has the property P locally if, for every
x X, every open neighborhood of x includes another open neighborhood of x
that has property P.
For instance, a space X is locally connected if every open neighborhood of
each x X includes an open connected neighborhood. Informally we may say
that x has arbitrarily small connected neighborhoods.
Proposition A16.3. In a locally connected space, the connected components are
open.
(It follows that the connected components are equal to the quasicomponents,
see Exercise A15.8.)
Proof. Let C be a connected component and let x C. There is a connected
neighborhood N of x. Then N C and so N

. Since x N

, x is an
interior point of C. Since x was arbitrary, C is equal to its own interior and so is
open.
Proposition A16.4. In a locally path-connected space, the connected components
are equal to the path components (and are open).
Proof. The same argument as in the previous proposition shows that the path com-
ponents in a locally path-connected space are open. Since the complement of
each path component is a union of other path components, they are also closed.
Now the connected component of x must be included in any open-and-closed set
that contains x, so in particular it must be included in the path component of x.
The path component is always included in the connected component, so they are
equal.
49
The most important example of a locally connected and locally path-connected
space is furnished by an open subset of R
n
(or any normed vector space).
Exercise A16.5. There is a weaker local path connectedness notion, sometimes
called semilocal path connectedness: X is semilocally path connected if for each
x X and each open neighborhood V of x, there is an open neighborhood U V
of x such that for each y U there is a path : [0, 1] V starting at x and
ending at y. Show that a space which is locally connected and semilocally path
connected must be locally path connected. Can you give an example of a space
which is semilocally path connected but not locally path connected?
We now turn to consider local compactness. Here it is necessary to modify
the denition somewhat: we say that a space is locally compact if every open
neighborhood of each point contains a compact neighborhood (which need not be
open - there may be very few compact, open sets). But in fact there is a simpler,
equivalent denition.
Proposition A16.6. A Hausdorff space is locally compact iff each point has a
compact neighborhood.
Proof. Local compactness says that each point has arbitrarily small compact
neighborhoods, a stronger condition than that of the proposition. Suppose that x
X has a compact neighborhood K, and let N be any neighborhood of x. We need
to nd a compact neighborhood that is included in N. Let U = K

. Then
U is a neighborhood of x, and U and U are closed subsets of a compact space
(namely K), hence are compact. By regularity (or, more exactly, by the proof of
regularity) there exists an open subset V of U, containing x, whose closure does
not meet the compact set U. Then V U is the desirec compact neighborhood.
Let X be a compact Hausdorff space and x
0
X. Then X x
0
is a locally
compact Hausdorff space (since each point of X distinct from x
0
has a closed,
hence compact, neighborhood that does not meet x
0
). In fact, every locally com-
pact Hausdorff space arises in this way.
Proposition A16.7. Let X be a locally compact Hausdorff space. Let X
+
be the
union of X and a disjoint point, denoted . The following sets form a topology
on X
+
:
(a) The open subsets of X in its original topology.
50
(b) The subsets of the form (X K) where K X is compact.
The topology so dened on X
+
is compact and Hausdorff and the identity map
X X
+
is a homeomorphism onto its image.
The space X
+
is called the one-point compactication of X. We will need the
following easy exercise.
Exercise A16.8. Let Y be a Hausdorff space. If K, L are compact subspaces of
Y then so is K L.
Proof. Let T be the family of subsets of X
+
described in the proposition. We
show it is a topology.
(i) Let U

be a family of members of T . If they are all of type (a), then their


union is an open subset of X and is of type (a) also. If one of them is of type
(b), say equal to (XK), then their union is of the form(XL)
where L is a closed subset of K, hence compact; so the union is of type (b).
(ii) Let U
1
, . . . , U
n
be a nite family of members of T . If one of them is of
type (a), then their intersection if of type (a). If they are all of type (b), their
intersection is the complement of a nite union of compact sets, which is
compact by Exercise A16.8 above. Thus the intersection id of type (b).
(iii) Clearly and X
+
belong to T .
Thus T is a topology, and the relative topology that it induces on X is the same
as the original topology of X. It remains to show that this topology is compact
and Hausdorff. To show that T is compact, let W be a family of members of
T covering X
+
. One of them (say W
0
) must contain , and hence must be of
the form (X K), K X compact. Now the remaining W

intersect
X in open sets which cover K, so some nite subcollection covers K also; this
subcollection, together with W
0
, forms a nite subcover of W .
Finally to show that T is Hausdorff: Any two points not at innity can be
separated by open sets of type (a) (by the Hausdorff property of X). Thus it
sufces to show that can be separated from a point x X. But by local
compactness, x X has a compact neighborhood, K say; then K

and
(X K) are disjoint members of T containing x and respectively.
51
Lecture A17
Some set theory
Further topics in our discussion will require some more advanced background
from set theory. We will take a rather naive approach to set theory. To learn
about the paradoxes this gives rise to, and how to avoid them, you need a course
in Logic.
Denition A17.1. A partial order on a set S is a relation which is reexive,
antisymmetric, and transitive: that is, for all x, y, z S, x x, x y and y x
imply x = y, and x y and y z together imply x z.
A set with a partial order is called a (partially) ordered set. If S is such a set
then
(a) An upper bound for a subset T S is an element x such that y x for all
y T.
(b) S is directed if every two-element subset has an upper bound (equivalently,
every nite subset has an upper bound).
(c) S is totally ordered if for all x, y, either x y or y x.
(d) A subset of S that is totally ordered (in the induced ordering) is called a chain.
(e) S is inductively ordered if every chain in S has an upper bound; it is strictly
inductively ordered if every chain has a least upper bound.
(f) A maximal element for S is an element m S such that x m implies
x = m (i.e., there isnt anything strictly greater than m.) Notice that this is a
weaker notion than that of upper bound.
(g) S is well-ordered if it is totally ordered and every non-empty subset has a least
member (necessarily unique).
Exercise A17.2. Give simple examples to illustrate all these notions.
Let S be an ordered set. A map f : S S will be called an inator (I just
made this word up) if f(x) x for all x S.
52
Lemma A17.3. (Bourbaki fundamental lemma) Any inator on a nonempty strictly
inductively ordered set has a xed point, i.e. a point x such that f(x) = x.
Well prove this in a moment. For now, lets draw some consequences.
Corollary A17.4. (Weak form of Zorns lemma) Every nonempty strictly induc-
tively ordered set contains a maximal element.
Proof. Let S be strictly inductively ordered and suppose it contains no maximal
element. Then for each x S there is a y
x
S with y
x
> x. Consider the map
1
f
sending x to y
x
. Then f is an inator with no xed point, contradicting Bourbakis
lemma.
Corollary A17.5. (Hausdorff maximality principle) Let S be a nonempty partially
ordered set. Each chain in S is included in a maximal chain (i.e., a chain which
is not itself included in any larger chain).
Proof. Let C be the collection of all chains in S, partially ordered by inclusion.
Then C is partially ordered, and indeed is strictly inductively ordered (with least
upper bound for a chain in C given by the union of its members). Now ap-
ply A17.4.
Corollary A17.6. (Zorns lemma) Every nonempty inductively ordered set (not
necessarily strict) contains a maximal element.
Proof. Let S be such a set and let C be a maximal chain in S (provided by A17.5).
Let m be an upper bound for C. Then m is maximal, since if x > m then C x
is a chain larger than the supposedly maximal chain C.
Now let us see to proving the Bourbaki lemma A17.3. Let S be nonempty and
strictly inductively ordered and f : S S an inator. Pick a point a S (xed
throughout the discussion). Well say that a subset F of S is closed if a F, if
f(x) F whenever x F, and if, whenever C F is a chain, the least upper
bound of C belongs to F.
Obviously, the intersection of any family of closed sets is closed. Moreover,
S itself is closed. Thus, there is a minimal closed set M, namely, the intersection
of all the closed sets that there are. The strategy of our proof is to show that M is
totally ordered, i.e., a chain. This will complete the proof because if x is the least
1
For those in the know, this is where we use the axiom of choice.
53
upper bound of M, then x M and f(x) M also, hence f(x) x; since f is
an inator, it follows that f(x) = x.
Note that the collection of all x a is closed. Thus, a is the least element of
the minimal closed set M.
Let us call u M unavoidable if, whenever x M and x < u, we have
f(x) u. If u is an unavoidable point, then the set
x M : either x u or f(u) x
is readily seen to be closed. Since it is a subset of M, it must be the whole of M,
by minimality.
We can now prove that every point is unavoidable. Let U be the set of unavoid-
able points in M. It contains a. We will show that it is closed; then, by minimality,
it must equal M itself. Suppose then that u is unavoidable and consider f(u). If
x < f(u) then, by the paragraph above, either x < u, or x = u, or f(u) x, and
the last possibility cannot occur. The other two possibilities yield f(x) f(u) by
the unavoidability of u. So, f(U) U. A similar argument shows that the least
upper bound of any chain in U belongs to U, so U is closed.
We now know that each point of M is unavoidable. It is easy to deduce that
M is totally ordered. For, let x, y M. Since x is unavoidable, either y x or
f(x) y, but the latter certainly implies that x y becasue f is an inator. As
already remarked, this completes the proof of Lemma A17.3.
Proposition A17.7. (Well ordering principle) Every nonempty set can be well-
ordered.
Proof. Let X be a set. Let Y be the set of pairs (S, _) where S is a subset of
X and _ is a well-ordering of S. Y is nonempty and is inductively ordered by
inclusion. By Zorns lemma, there is a maximal element in Y , say (S, _). But
now we must have S = X; if not, we can choose some x X S and well-order
S x by insisting that x is greater than every element of S. This contradicts
maximality.
54
Lecture A18
Countability and Metrization
We now briey discuss the theory of cardinal numbers. Let X and Y be sets.
Proposition A18.1. There exists an injection from X to Y if and only if there
exists a surjection from Y to X.
Proof. Without loss of generality assume that X ,= . Let i : X Y be an
injection. Let I = i(X) Y . Fix an element x
0
X. Dene s: Y X as
follows: if y I, then s(y) is the unique x X such that i(x) = y; otherwise,
s(y) = x
0
. This is a surjection.
Conversely, suppose that s: Y X is a surjection. For each x X choose
some y
x
Y such that s(y
x
) = x. Let i be the map that sends x to y
x
. Then i is
an injection.
Exercise A18.2. The last part of this argument uses the Axiom of Choice directly.
Reformulate it to use Zorns Lemma.
Theorem A18.3. (Schr oder-Bernstein) If there exists an injection from X to Y
and an injection from Y to X, then there exists a bijection from Y to X.
Proof. (K onig) Assume (wlog) that X and Y are disjoint sets and let Z be their
disjoint union. let f : X Y and g : Y X be injections and let h be the union
of these maps, considered as a map Z Z. A subset S of Z is stable if z Z
iff h(z) Z. Let S
z
be the minimal stable set containing z. Explicitly, S
z
is the
set h
n
(z) : n Z, where we note that h
1
(p) denotes an element q such that
h(q) = p; such an element may not exist, but it is unique if it does exist. By the
injectivity of h, stable sets either coincide or are disjoint; so they partition Z.
It sufces now to establish a bijection S
z
X S
z
Y for each S
z
separately.
The stable sets S
z
are of four types:
(a) X-stopping: h
n
(z) is dened only for n = k, k + 1, . . . and h
k
(z) X;
(b) Y -stopping: h
n
(z) is dened only for n = k, k + 1, . . . and h
k
(z) Y .
(c) Two way innite: the h
n
(z) are all distinct for different values of n Z.
(d) Cyclic: h
m
(z) = z for some m (necessarily even).
55
In the rst case, f gives a bijection S
z
X S
z
Y . In the second case, g gives
a bijection S
z
Y S
z
X. In the third and fourth cases, either f or g will
do.
To each set X associate a number, its cardinality card X. We dene card X =
card Y iff there is a bijection between X and Y (one says that X and Y are
equinumerous). We dene card X card Y iff there is an injection X Y . The
Schr oder-Bernstein theorem shows that this denes a partial order on the cardi-
nalities.
Denition A18.4. The cardinality of the natural numbers is denoted
0
(read:
aleph-zero). A cardinal that is
0
is called countable. (We also call a set
countable if its cardinality is countable.)
Example A18.5. The integers Z and the rationals Q are countable sets.
Proposition A18.6. For every cardinal there is a strictly larger cardinal.
Proof. Let X be a set and let P(X) (the power set) be the set of all subsets of X.
We will show that there is no surjection X P(X). Indeed, suppose that f is
such a surjection. Consider
S = x X : x / f(x).
This is a subset of X. If S = f(y) for some y X we obtain (from the denition
of S) the contradiction
y f(y) y / f(y).
Thus f is not surjective after all. On the other hand, there is an obvious injection
X P(X) (as one-element subsets). It follows that card(P(X)) > card(X).
We denote the cardinality of P(X) by 2
card X
. More generally, the cardinality
of the set of all maps X Y is denoted (card Y )
card X
. The cardinal 2

0
is
denoted c (the continuum).
Once can do arithmetic with cardinals: if X, Y have cardinalities x, y respec-
tively then xy is the cardinality of XY , x +y is the cardinality of X.Y , and so
on. Arithmetic laws like x
yz
= (x
y
)
z
then follow from corresponding set-theoretic
properties (a map fromY Z to X is the same as a map fromZ to the set of maps
from Y to X).
56
Example A18.7. Let us prove that c

0
= c. Its a one-liner:
c

0
=
_
2

0
_

0
= 2

0
= 2

0
= c.
Proposition A18.8. The set of real numbers has cardinality c.
Proof. The set of all sequences of rational numbers has cardinality
0

0
= c. The
subset of convergent sequences must then have cardinality c, and this subset
maps onto R, so card R c. On the other hand, the set of all sequences a = a
n

of zeroes and ones also has cardinality c, and the assignment a

n=1
3
n
a
n
is
an injection from this set of sequences to R, so card R c. Schroder-Bernstein
now completes the proof.
Let X be a topological space.
Denition A18.9. X is separable it it has a countable dense subset. It is sec-
ond countable if there is a countable basis for the topology of X: i.e., there is a
countable family B of open sets such that every open set is a union of sets from
B.
Lemma A18.10. A second-countable space is separable. A separable metric
space is second-countable.
However, there are in general examples of separable spaces that are not second
countable. For instance, let X be an uncountable set equipped with the conite
topology. Then X is separable (any countable subset of X meets every nonempty
open set) but it is not second countable (a countable family of conite sets can
omit only countably many points in total, so there must be some conite set that
does not contain any member of the family).
Proof. Let X be a second countable space. Let U
n
be a countable base for the
topology, and select a point x
n
from each U
n
. The set S = x
n
is countable, and
for any x X and any open set U containing x, there is some U
n
U containing
x; so U S contains x
n
and is in particular nonempty. Thus S is dense.
Let X be separable and metrizable, with S a countable dense set. Then the set
B of balls B(s, t) where s S and t Q is countable. Moreover, it forms a base
for the topology: if x X belongs to an open set U, there is a ball B(x, r) U;
there is an s S such that d(x, s) < r/2; there is a rational t with d(x, s) < t <
r/2; the ball B(s, t) belongs to B, contains x, and is included in U.
57
A neighborhood base at a point x X is a family of open sets containing x
such that each open set containing x must include a member of the family.
Denition A18.11. A topological space X is rst countable if each point has a
countable neighborhood base.
Every metric space is rst countable. Every second countable space is rst
countable but the converse is false: e.g. consider the discrete topology on an
uncountable set.
Exercise A18.12. Show that a compact metric space is separable and second
countable.
Theorem A18.13. (Urysohn metrization theorem) A second countable normal
Hausdorff space is metrizable, i.e. the topology can be given by a metric.
Proof. Let X be such a space and let B be a countable basis for the topology.
Whenever U, V Bwith U V there exists, by Urysohns lemma, a continuous
function f : X [0, 1] that is equal to 0 on U and is equal to 1 on X V . The
collection F of all such functions is countable: denote it f
1
, f
2
, . . ..
Dene d: X X R
+
by
d(x, y) =

n=1
2
n
[f
n
(x) f
n
(y)[.
We will show that d is a metric dening the topology of X.
It is clear that d(x, x) = 0, that d(x, y) = d(y, x), and that d satises the
triangle inequality. If d(x, y) = 0, then f
n
(x) = f
n
(y) for all n. But, if x ,= y,
then there is a V B containing x but not y (by the Hausdorff property); by
normality there is an open set U containing x with U V , and we may take
it that U B also; so then the function f
n
corresponding to this pair (U, V ) has
f
n
(x) = 0 but f
n
(y) = 1, and this is a contradiction. We conclude that d(x, y) = 0
iff x = y, so d is a metric.
Being the sum of a uniformly convergent series of continuous functions, d is
continuous (relative to the original topology on X). It follows that every d-open
ball is open in the original topology, so every d-open set is open in the original
topology.
It remains to show that every set open in the original topology is d-open. Let
W be open in the original topology, and x W. There is a set V B such that
58
x V and V W, and there is a set U B such that x U and U V . Let f
n
be the function constructed above corresponding to the sets U and V . Let = 2
n
.
Then, if y B
d
(x; ), f
n
(y) < 1 and so y V ; in particular y W. Thus W
contains a d-ball around each of its points, so it is d-open, as required.
59
Lecture A19
Convergence and Nets
Let x
n
be a sequence in a topological space X. (Okay, I know that I told
you that I was not going to use sequences outside metric spaces, but I lied.) We
say that x
n
converges to a point x X if, for every open neighborhood U of x,
there is some N such that x
n
U for n N.
If X is rst countable then sequences in x contain all the information that
there is about the topology of X. One way to see this is the following
Lemma A19.1. In a rst countable space X, a subset A is closed if and only if
the limit of every convergent sequence in A is itself a member of A.
Proof. Suppose that A X. A point x belongs to the closure of A iff every open
neighborhood of x contains points of A. But, since X is rst countable, this will
be true iff every member of a countable neighborhood base U
n
at x contains
points of A. Picking such a point a
n
A U
n
for each n = 1, 2, 3 . . . gives a
sequence (a
n
) in A converging to x. The converse statement (if x is the limit of a
convergent sequence in A, then x belongs to the closure of A) is easy.
For another example, try the following exercise.
Exercise A19.2. Let X and Y be rst countable. Show that f : X Y is con-
tinuous if and only if the sequence f(x
n
) f(x) in Y whenever the sequence
x
n
x in X. (Hint: The proof is exactly the same as the one we already dis-
cussed for metric spaces.) Do you need rst countability for both spaces, or only
for one of them (and, if so, which)?
But in general sequences are not enough to probe the structure of a topolog-
ical space.
Example A19.3. Any totally ordered set X can be equipped with the order topol-
ogy for which a basis is the half open intervals
(a, b] := x X : a < x b.
Let X be uncountable, and well-order it (using the axiom of choice). Let X
be the smallest element that has uncountably many predecessors (a.k.a. the rst
uncountable ordinal); this exists because of well-ordering. Let C = x X :
x < be the set of predecessors of . Then belongs to the closure of C, but
nevertheless there is no sequence in C converging to .
60
This motivates the following denition.
Denition A19.4. A net in a set X is a map from a directed set to X. (Recall that
a directed set is a partially ordered set in which any two elements have an upper
bound.)
Let S be a subset of X.
Denition A19.5. A net x
i
is eventually in S if there is i
0
such that x
i
S for
all i i
0
; it is frequently in S if it is not eventually in X S.
Now suppose that X is a topological space.
Denition A19.6. A net x
i
converges to x X if it is eventually in every
neighborhood of x.
These denitions match the denitions of sequence and convergent se-
quence when we take the net simply to be the natural numbers N. On the other
hand, in Example A19.3 above, we could take the collection C itself in its natural
ordering to be a net, and it is then almost a tautology that C converges to even
though we have seen there is no sequence that will do this.
Proposition A19.7. In any topological space X, a subset A is closed if and only
if the limit of every convergent net in A is itself a member of A.
Proof. Emulating the proof of Lemma A19.1, what we have to show is that if x
is a point of the closure of A, then there is some net in A that converges to x. A
net has to parameterized by a directed set. The trick is to pick the collection of all
neighborhoods of x as the directed set! In fact, suppose that U and V are open
neighborhoods of x. We dene U V if U V (notice that the inclusion is
reversed!). This makes the collection N = U open : x U a partially ordered
set, and in fact a directed set: an upper bound for U and V is the intersection UV .
Since x belongs to the closure of A, for every U N there exists a
U
A U.
The mapping
2
from N to A given by U a
U
is a net. Moreover, it converges
to x by denition: if V is any neighborhood of A then there is an element of N
(namely V itself) such that if U V then a
U
V .
Similarly we have
Proposition A19.8. A map f : X Y between topological spaces is continuous
if and only if the net f(x
i
) f(x) in Y whenever the net x
i
x in X.
2
Using the axiom of choice here!
61
Proof. The proof that a continuous map has this property is essentially the same
as the corresponding proof for sequences.
Suppose that f has the property in question but is not continuous. Then there
exist a point x X, y = f(x), and an open neighborhood W of y, such that
f
1
(W) is not a neighborhood of x. It follows that for every open neighborhood
U of x there exists x
U
U W. Organize the x
U
into a net parameterized by the
directed set N of neighborhoods of x, as in the previous proposition. Then, by
construction, x
U
converges to x; but f(x
U
) does not converge to y, since it never
lies in the neighborhood W of y.
Exercise A19.9. Suppose that all we know about the topology of X is that it is
generated by some family F of subsets. Show that a net x
i
converges to x iff,
for every U F which contains x, the net x
i
is eventually in U. (This exercise
should help you understand why we insist that a net should be parameterized by a
directed set.)
One might ask whether we can similarly generalize the denition of com-
pactness, which in the metric space case is expressed as every sequence has a
convergent subsequence. To do this, we must arrive at the correct denition of
subnet.
Denition A19.10. Let D and D

be directed sets. A function h: D

D is
called nal if, for any i
0
D, there is j
0
D

such that j j
0
implies h(j) i
0
.
Given a net D X, the result of composing it with a nal function h is called a
subnet or renement of the original one.
Every subnet of a convergent net is convergent (with the same limit); the de-
nition of nal function is cooked up to make this true.
Denition A19.11. A net x
i
in X is called universal if, given any subset S of
X whatsoever, either x
i
is eventually in S or else eventually in X S.
Lemma A19.12. Every net has a universal subnet.
Proof. This depends unavoidably on the axiom of choice.
Let A : D X be a net in X, parameterized by a directed set D. A collection
F of subsets of X will be called an A-lter if A is frequently in every member
of F, and if F is closed under nite intersections and the formation of supersets.
Such objects exist: for example, X is an A-lter. The collection of A-lters is
partially ordered by inclusion, and every chain in this partially ordered set has an
62
upper bound (the union). Thus Zorns Lemma provides a maximal A-lter; call
it F
0
.
Suppose that S X has the property that A is frequently in A S for every
A F
0
. Then the union of F
0
with the set of all sets A S, A F
0
, is again an
A-lter. By maximality we deduce that S itself belongs to F
0
.
We will use this property to construct the desired universal renement. Let D

be the collection of pairs (A, i) with A F


0
, i D, and A(i) A; it is a directed
set under the partial order
(B, j) (A, i) B A, j i.
The map (A, i) i is nal so denes a renement A

of the net A. We claim


that this renement is universal.
Let S X have the property that A

is frequently in S. Let A F
0
and let
i be arbitrary. By denition, there exist B F
0
, B A, and j i, such that
A(j) = A

(B, j) B S A S. We conclude that A is frequently in A S


for every A F
0
and hence, as observed above, that S itself belongs to F
0
.
Now let S be arbitrary. It is not possible that A

be frequently both in S and


in X S, for then (by the above) both S and X S would belong to F
0
, and then
their intersection, the empty set, would do so as well, a contradiction. Thus A

fails to be frequently in one of these sets, which is to say that it is eventually in


the other one.
Proposition A19.13. The following properties of a topological space X are equiv-
alent:
(i) X is compact;
(ii) (the nite intersection property) if T is a family of closed subsets of X, and
the intersection of any nite number of members of T is nonempty, then in
fact the intersection of all the members of T is nonempty;
(iii) every universal net in X converges;
(iv) every net in X has a convergent subnet.
Proof. Suppose (i). Let T be a family as in (ii) and let | = X F : F T.
If

T is empty then | is an open cover of X, hence it has a nite subcover,
so the intersection of some nite subcollection of T is empty. This proves (ii).
Conversely, if we assume (ii) and let | be an open cover that is supposed to have
63
no nite subcover, we reach a similar contradiction by considering the family of
closed sets T = X U : U |. Thus (i) and (ii) are equivalent.
Suppose (ii), and let x

be a universal net in X. If it does not converge, then


each x X has a neighborhood U
x
for which x

is not eventually in U
x
, hence
(by universality) x

is eventually in F
x
= X U
x
. It follows that x is eventually
in any nite intersection of the F
x
. In particular no nite intersection of the F
x
is
empty. By (ii) the intersection of all the F
x
is nonempty, and this is a contradiction
since x / F
x
for each x.
(iii) implies (iv) by Lemma A19.12.
Suppose (iv). Let T be a family of closed sets with the nite intersection
property. Let T be the directed set comprised of nite intersections of members
of T (directed by reverse inclusion) and for each D T let x
D
be a point of D.
Then x
D
is a net in X; choose a subnet convergent to x X. By denition, any
neighborhood of x meets every F T; since F is closed, we deduce that x F.
Thus the intersection of all the F contains x. This proves (ii).
Now let (X

) be a family of topological spaces, parameterized by A


where A is some index set (possibly innite or even uncountable). Let X be the
Cartesian product

: that is, an element of X is a mapping x: x

,
where x

. For each there is a projection map

: X X

, which sends
the point x to its coordinate x

.
Denition A19.14. The product topology on X is the topology generated by all
the sets
1

(U

), where runs over A and U

runs over all the open subsets of


X

.
Exercise A19.15. Let x
i
be a net in the product X dened above. Show that x
i
converges to x X if and only if

(x
i
) converges to

(x) for all A.


Theorem A19.16. (Tychonoff) The product of any family of compact spaces is
compact.
Proof. We use the characterization (iii) of compactness in Proposition A19.13
above.
Let x
i
be a universal net in X. Then

(x
i
) is a universal net in X

, hence
convergent, say to x

. Let x X be the unique point such that

(x) = x

.
Then, by Exercise A19.15, x
i
x in X. Thus X is compact.
64
Lecture A20
Quotient Spaces
Let X be a set and an equivalence relation on it. Recall that the notation
X/ denotes the set of equivalence classes of the equivalence relation. There is
a canonical map : X X/ that sends each x X to its equivalence class.
Now suppose that X is a topological space.
Denition A20.1. The quotient topology on X/ is dened by declaring that a
subset U X/ is open iff the inverse image
1
(U) is open in X.
Of course, there is something to check here, namely that the sets U identied
by this denition do satisfy the axioms for a topology. But this is easy. Clearly,
the quotient topology makes continuous, and it is the largest topology (most
open sets) that does so. Moreover, a function f : (X/ ) Y is continuous iff
f : X Y is continuous.
Remark A20.2. Asubset of X is called saturated (with respect to ) if it is a union
of equivalence classes. Thus, the open sets of the quotient topology correspond
to the saturated open subsets of X. The saturation of an arbitrary subset S is the
union of all the equivalence classes that meet S, i.e. the smallest saturated set that
includes S.
Example A20.3. Let X = [0, 1] and let be the equivalence relation that makes
0 and 1 equivalent and whose only other equivalence classes are points. Then
[0, 1]/ is homeomorphic to the unit circle S
1
. To see this, let g : X S
1
be
the continuous map g(t) = e
2it
= (cos 2t, sin 2t), and dene f : X/ S
1
by f((t)) = g(t); this is well-dened since g(0) = g(1). By the remark above,
f is a continuous map from X/ to S
1
, which is in fact bijective. We must show
that f is also open, and for this it sufces to show that f takes each neighborhood
of each point (x) of X/ to a neighborhood of f((x)) = g(x). That is, we
must show that any saturated open set containing x maps to a neighborhood of
g(x). When x , 0, 1 this is easy, since any saturated open set containing x
includes an interval (x , x + ). What about saturated open sets containing 0
or 1? Any such must include U = [0, ) (1 , 1] for some > 0, and then
f(U) = (cos 2t, sin 2t) : t (, ) is a neighborhood of (1, 0) in S
1
. So
we are done.
The ideas underlying this example can be summarized in the next two propo-
sitions. The rst is the topological analog of the rst isomorphism theorem in
65
algebra. Let us call a continuous map f : X Y -respecting if f(x) = f(x

)
whenever x x

. Then we have
Proposition A20.4. A continuous g : X Y is -respecting iff g = f for
some continuous f : X/ Y .
Proof. If g is -respecting then a function f can be dened by setting f(c), for an
equivalence class c, equal to f(x) for any representative x of c. Then g = f ,
and f is continuous by denition of the quotient topology.
Proposition A20.5. Let X be a compact space and Y a Hausdorff space and
suppose that g : X Y is continuous and surjective. Let be the equivalence
relation on X dened by x x

iff g(x) = g(x

). Then the map f : X/ Y


induced by g is a homeomorphism.
Proof. Tautologically, g is -respecting, so there is a continuous map f as de-
scribed, which is in fact a bijection X/ Y . But X/ is a compact space (as
the image of a compact space under the continuous map , Proposition A12.8),
and Y is a Hausdorff space, so this bijection is in fact a homeomorphism by Corol-
lary A12.12.
It might be helpful to think of , in this proposition, as the kernel of the
map g.
One problem with quotient spaces is that they are frequently not Hausdorff.
An obvious necessary condition for Hausdorffness is that the equivalence classes
should be closed: after all, each equivalence class is the inverse image of a point
in X/ under the continuous map , and we know that one-point sets are closed
in a Hausdorff space. But this is by no means a sufcient condition:
Example A20.6. Let X be the unit square [0, 1] [0, 1], which is a compact Haus-
dorff space, and let be the equivalence relation which makes (x
1
, y
1
) equivalent
to (x
2
, y
2
) iff either x
1
= x
2
and y
1
= y
2
or y
1
= y
2
> 0. Then any two saturated
neighborhoods of (0, 0) and (1, 0) must intersect; so the corresponding quotient
space cannot be Hausdorff.
One can even give an example of this kind where all the equivalence classes
except one consist of single points.
Exercise A20.7. Let X be the real line, with the topology generated by all the
open sets of the usual topology together with the complement of the set A =
1/n : n = 1, 2, 3, . . .. Show that X is a Hausdorff space, but that the quotient
66
space obtained by identifying all the points of A to a single point (this is usually
denoted X/A) is not. (This is also an example of a Hausdorff space that is not
regular.)
We will give some useful conditions for Hausdorffness of quotients. Let be
an equivalence relation on X. The graph of is the subset
G = (x
1
, x
2
) : x
1
x
2
X X.
Suppose that X/ is Hausdorff. Then whenever x
1
, x
2
there are disjoint open
subsets U
1
, U
2
of X/ with (x
1
) U
1
and (x
2
) U
2
. Thus
1
(U
1
)

1
(U
2
) is open in X X, contains (x
1
, x
2
), and does not meet G. We conclude
that if X/ is Hausdorff, then the graph G is closed. (This statement includes
our previous requirement that the equivalence classes should all be closed, but it
is stronger.)
The previous exercise shows that the converse is not necessarily the case.
However, there are additional conditions under which the closure of the graph
is both necessary and sufcient for the Hausdorffness of the quotient.
Proposition A20.8. Let X be a Hausdorff space and let be an equivalence
relation on X whose graph is closed. If in addition the projection : X X/
is an open map then X/ is Hausdorff also.
Proof. Let G denote the graph of . Suppose that is open. Let x
1
, x
2
X with
(x
1
) ,= (x
2
); then the point (x
1
, x
2
) X X does not meet the closed set
G, so there is an open rectangle of the form U
1
U
2
which contains (x
1
, x
2
) and
does not meet G. But then (U
1
) and (U
2
) are disjoint open subsets of X/
containing x
1
and x
2
respectively.
Proposition A20.9. Let X be a compact Hausdorff space and let be an equiva-
lence relation on X whose graph is closed. Then X/ is Hausdorff also. More-
over, the projection : X X/ is a closed map (it need not be open).
Proof. We rst showthat the saturation of any closed subset of X is closed (equiv-
alently, : X X/ is a closed map). Let C be a closed subset of S. We may
write the saturation S of C as
S =
2
((C X) G) X,
where
2
: X X X is the second projection map of the product space. Now,
C X and G are closed subsets of X X, which is compact (by the Tychonoff
67
theorem). Thus (C X) G is a closed subset of a compact space, so compact.
It follows that
2
((C X) G) is a compact subset of a Hausdorff space, so it is
closed. This proves our assertion.
Now suppose that x
1
, x
2
X with (x
1
) ,= (x
2
), and let C
1
and C
2
be the
equivalence classes of x
1
and x
2
respectively. Let us show that there are open sets
U
1
, U
2
, including C
1
, C
2
respectively, whose saturations do not meet that is to
say, (U
1
U
2
) G = . This is the same argument as the proof that compact
Hausdorff spaces are normal. First, we prove the analog of regularity. Fix x C
1
.
For each y C
2
there are open sets P
y
containing x and Q
y
containing y such
that (P
y
Q
y
) G = . Because C
2
is compact there is a nite subcover of C
2
say by Q
y
1
, . . . , Q
yn
. Then P =

n
j=1
P
y
j
and Q =

n
j=1
Q
y
j
are open sets, one
containing x and one including C
2
, such that P QG = . Repeat the argument
using the compactness of C
1
to obtain the desired open sets U
1
C
1
, U
2
C
2
,
with (U
1
U
2
) G = .
Consider now the closed set X U
i
, and let S
i
be its saturation; S
i
is closed
(by the rst thing we proved above) and does not meet C
i
, so the complement
X S
i
= V
i
is a saturated open set with C
i
V
i
U
i
. Now (V
i
) and (V
2
) are
disjoint open subsets containing x
1
and x
2
respectively.
68
Lecture A21
Basics of Topological Groups
Let G be a group. We can think of it as a set provided with a distinguished
element (the identity e) and two distinguished maps, the multiplication map
GG G, (x, y) xy
and the inversion map
G G, x x
1
.
Denition A21.1. A topological group is a group provided with a topology for
which the multiplication and inversion maps are continuous.
There are many examples: the (real or complex) numbers under addition,
the nonzero (real or complex) numbers under multiplication, the matrix groups
GL(n, R) and GL(n, C) as well as their various subgroups SL(n, R), SL(n, C),
O(n), U(n), etc. Any group can be thought of as a topological group when en-
dowed with the discrete topology.
In a topological group the left and right multiplication maps by a xed element
g, dened by L
g
(x) = gx, R
g
(x) = xg, are homeomorphisms G G. Conse-
quently, every translate gU or Ug of an open subset U is again open (and every
translate of a closed set is closed).
Proposition A21.2. A topological group G is Hausdorff if and only if the one-
point set e is closed.
Proof. Suppose that e is closed. By translation, every one-point set is closed.
If g
1
,= g
2
there exists an open set V = G g
2
that contains g
1
and does not
contain g
2
. By continuity of the multiplication map at the point (g
1
, e), together
with the denition of the product topology, there are open sets U
1
, U
2
that contain
g
1
and e respectively such that U
1
U
2
V . But now U
1
and g
2
U
1
2
are open
neighborhoods of e and g respectively. I claim they dont intersect: if they did,
then u
1
= g
2
u
1
2
with u
1
U
1
, u
2
U
2
, and then g
2
= u
1
u
2
U
1
U
2
V , a
contradiction.
This is one example of the homogeneity of topological groups.
Let Gbe a topological group and H a subgroup. We denote by G/H the space
of left cosets xH for x G, or in other words the space of equivalence classes
69
for the equivalence relation x y x
1
y H; similarly we denote by HG
the space of right cosets Hx, which are equivalence classes for the equivalence
relation x y xy
1
H. These are called (left and right) homogeneous
spaces of G by H, and we give them the quotient topology.
Exercise A21.3. If H is a normal subgroup of G then the left and right coset
spaces are the same. In this case we know from algebra that G/H has the structure
of a group. We have also just equipped it with a topology. Show that G/H is a
topological group, i.e., multiplication and inversion are continuous.
Proposition A21.4. The quotient maps G G/H and G HG are open
maps.
Proof. This is equivalent to the statement that the saturation of an open subset of
G is open. But the saturation of an open set U is
_
uU
uH =
_
hH
Uh;
the second representation exhibits it as a union of open sets, hence open.
Theorem A21.5. Let G be a Hausdorff topological group and let H be a closed
subgroup. Then the homogeneous spaces G/H and HG are Hausdorff.
Proof. By A20.8 and A21.4, it sufces to show that the graph of the equivalence
relation x y x
1
y H is a closed subset of GG.
Suppose that (x, y) / ; then x
1
y G H which is an open set. By
continuity of multiplication there exist open sets U x
1
, V y with U V
G H. But then U
1
V is an open neighborhood of (x, y) that does not meet .
It follows that the complement of is open, so is closed.
Denition A21.6. Let X, Y be topological spaces. A continuous, surjective map
f : X Y is called a local homeomorphism if every x X has a neighborhood
U such that the restriction of f to a map U f(U) is a homeomorphism.
Local homeomorphisms will play a very important role in the study of cov-
ering spaces, a key part of algebraic topology. The classic example of a local
homeomorphism is the exponential map t e
2it
from the real line to the circle.
This can be envisaged as a special case of the next result:
70
Proposition A21.7. Let G be a topological group and suppose that H is a closed
subgroup which is discrete (in the induced topology). Then the quotient map
: G G/H is a local homeomorphism.
Proof. There is a neighborhood V of e such that H V = e. By the same
kind of argument (continuity of multiplication) that we have already used several
times, there exists a neighborhood U of the identity with U
1
U V . But now
the restriction of : G G/H to the set U is injective: if (u
1
) = (u
2
) then
u
1
1
u
2
V H = e, and so u
1
= u
2
. Thus the restriction of to a map
from U to (U) is continuous, bijective, and open (A21.4): in other words, it
is a homeomorphism. We have proved that is a homeomorphism near e; by
translation, the same holds near any point.
A group G acts on a set X if there is given a map G X X, written
(g, x) gx, such that ex = x for all x X and g(hx) = (gh)x for all g, h G,
x X. If G is a topological group and X is a topological space we dene a
continuous action by requiring that the associated map GX X be continuous.
An action is transitive if for all x, y X there exists g G with gx = y.
Example A21.8. The rotation group SO(n) acts transitively on the unit sphere
S
n1
.
Proposition A21.9. Suppose that G acts transitively on the Hausdorff space X.
For any x X let G
x
denote the stabilizer
G
x
= g G : gx = x,
which is a closed subgroup of G. Then the formula
[g] g x
denes a continuous bijection G/G
x
X. It is a homeomorphism if G is com-
pact.
Proof. The continuous map G X given by g gx respects the equivalence
relation given by the cosets of G
x
. Thus it passes to a continuous map G/G
x

X, which is injective by construction, and is surjective because of the transitivity
of the action. The nal statement follows because G/G
x
is compact and X is
Hausdorff.
For example, considering the stabilizer of a single point (the North pole) gives
us the identication S
n1

= SO(n)/SO(n 1) (using the fact that SO(n) is
compact).
71
Lecture A22
Homotopy
Hereafter map means continuous map.
Consider the unit square X = I
2
= (x, y) R
2
: 0 x, y 1. Imagine
two continuous paths
1
,
2
: [0, 1] X, with
1
beginning at (0, 0) and ending
at (1, 1), and
2
beginning at (0, 1) and ending at (1, 0). It seems obvious that
these two paths must cross somewhere in the interior of the square. But this is
hard (maybe impossible?) to prove just with the tools that we have developed so
far. We need a little homotopy theory.
Denition A22.1. Let X and Y be topological spaces and let f
0
, f
1
: X Y be
continuous maps. A homotopy between f
0
and f
1
is a continuous map
h: X [0, 1] Y
with h(x, 0) = f
0
(x) and h(x, 1) = f
1
(x) for all x X. If there exists a homo-
topy between f
0
and f
1
we say that these maps are homotopic.
Proposition A22.2. Homotopy is an equivalence relation on the class of maps
X Y .
Proof. Clearly any map f is homotopic to itself via the constant homotopy h(x, t) =
f(x). If h is a homotopy between f
0
and f
1
, then
k(x, t) = f(x, 1 t)
is a homotopy between f
1
and f
0
. Finally, homotopies can be concatenated like
paths: if h

is a homotopy from f
0
to f
1
and h

is a homotopy between f
1
and f
2
then
h(x, t) =
_
h

(x, 2t) (t
1
2
)
h

(x, 2t 1) (t
1
2
)
is a homotopy between f
0
and f
2
.
Exercise A22.3. Check in detail that the homotopy h dened in the last part of
this proof is continuous.
72
Remark A22.4. This proof is obviously closely related to the proof that path-
connectedness is an equivalence relation, Lemma A15.10. In fact, path-connectedness
is just homotopy of maps where X is a point. Conversely, one can reduce the no-
tion of homotopy to path-connectedness in the space of maps X Y if one
denes a suitable topology on this space. But we wont do that here.
Proposition A22.5. If f
0
, f
1
are homotopic maps X Y , and g
0
, g
1
: Y Z are
homotopic maps, then g
0
f
0
, g
1
f
1
are homotopic maps X Z. (In particular,
homotopic maps remain homotopic after composition on the left or the right with
a xed map.)
There is a related notion, that of relative homotopy. Let X be a space and A
a closed subspace (one says sometimes that (X, A) is a pair). Then a homotopy
relative to A from X to Y is a homotopy h: X [0, 1] Y such that h(a, t) is
independent of t for all a A a homotopy that stays constant on A. This also
denes an equivalence relation by the same reasoning as above.
Example A22.6. Let X be the unit disc in Euclidean space. Then the identity map
X X is homotopic to the map that sends every point to the origin. A homotopy
between these maps is dened by h(x, t) = (1 t)x. A similar argument can be
applied to any star-shaped subset of Euclidean space.
Example A22.7. Let X, Y be spaces and let f : X Y be a continuous map.
The mapping cylinder of f is the space
C
f
= X [0, 1] .
f
Y
obtained from the disjoint union X[0, 1] .Y by identifying (x, 0) with f(x) for
all x X (and equipped with the quotient topology). As we saw earlier, there is
a continuous map F : C
f
Y dened by F(x, s) = f(x), F(y) = y. Moreover,
the identity map C
f
C
f
is homotopic, relative to Y , to the map C
f
Y C
f
dened by composing F with the standard inclusion Y C
f
. A homotopy h
between these maps can be dened by the formula
h((x, s), t) = (x, (1 t)s), h(y, t) = y.
Notice that the second example actually includes the rst: the disc is the map-
ping cylinder of the constant map from a sphere to a point.
The geometric situation dened in these examples occurs often enough to have
a special name.
73
Denition A22.8. Let X be a space and A a closed subspace. A retraction of X
onto A is a continuous map X A which is the identity on A. A deformation
retraction of X onto A is a homotopy, relative to A, from the identity map on X
to a retraction X A. If these exist one says that A is a retract or deformation
retract of X.
The examples above establish that a point is a deformation retract of a disk,
and that Y is a deformation retract of the mapping cylinder. Deformation retrac-
tion is a special case of the general notion of homotopy equivalence.
Denition A22.9. Let X and Y be spaces and let f : X Y and g : Y X
be maps. One says that f and g are homotopy inverses if the composite g f is
homotopic to the identity map on X and f g is homotopic to the identity map
on Y . A map that has a homotopy inverse is called a homotopy equivalence, and
one says that the spaces X and Y are homotopy equivalent if there is indeed a
homotopy equivalence between them.
For example, if A is a deformation retract of X, then the inclusion A X
and retraction X A form a pair of homotopy inverse maps, so X and A are
homotopy equivalent.
Proposition A22.10. The relation of homotopy equivalence is, indeed, an equiv-
alence relation.
Proof. Suppose that f

: X Y and f

: Y Z have homotopy inverses g

and
g

respectively. Then it follows from Proposition A22.5 that g

is a homotopy
inverse for f

.
74
Lecture A23
Homotopy Equivalence
The examples above establish that a point is a deformation retract of a disk,
and that Y is a deformation retract of the mapping cylinder. Deformation retrac-
tion is a special case of the general notion of homotopy equivalence.
Denition A23.1. Let X and Y be spaces and let f : X Y and g : Y X
be maps. One says that f and g are homotopy inverses if the composite g f is
homotopic to the identity map on X and f g is homotopic to the identity map
on Y . A map that has a homotopy inverse is called a homotopy equivalence, and
one says that the spaces X and Y are homotopy equivalent if there is indeed a
homotopy equivalence between them.
For example, if A is a deformation retract of X, then the inclusion A X
and retraction X A form a pair of homotopy inverse maps, so X and A are
homotopy equivalent.
Proposition A23.2. The relation of homotopy equivalence is, indeed, an equiva-
lence relation.
Proof. Suppose that f

: X Y and f

: Y Z have homotopy inverses g

and
g

respectively. Then it follows from Proposition A22.5 that g

is a homotopy
inverse for f

.
Denition A23.3. A space is called contractible if it is homotopy equivalent to a
point.
The cone CX on a space X is the quotient X [0, 1]/X 0, or in other
words the mapping cylinder of the constant map from X to a point. Any mapping
cylinder deformation retracts onto the range of the map from which it is dened
(as we have seen), so the cone on any space is contractible.
Homotopy equivalent spaces can be rather different topologically. For exam-
ple, the 1-dimensional spaces given by the theta curve, the gure 8, and the
dumb-bell are all homotopy equivalent, although no two of themare homeomor-
phic. However, there are topological properties that are preserved by homotopy
equivalence.
Example A23.4. Suppose X and Y are homotopy equivalent. If X is path-
connected, then so is Y .
75
Proof. Let f : X Y and g : Y X be mutually inverse homotopy equiva-
lences. Note that for each y Y , the points y and f(tg(y)) belong to the same
path component: indeed, if h is a homotopy between f g and the identity, then
t h(y, , t) denes a path joining y to f(g(y)). Now let y, y

Y and suppose
that X is path connected. Then there is a path in X joining g(y) to g(y

). Now
f is a path joining f(g(y)) to f(g(y

)). Thus f(g(y)) and f(g(y

)) are in the
same path component. It follows from the earlier discussion that y and y

are in
the same path component; that is, Y is path connected.
Suppose that X is a space and A a closed subspace. If A is contractible, we
might say that up to homotopy Ais the same as a point. If Areally were a point,
the quotient space X/A would be just the same as X. So we might be led to hope
that if Ais contractible, X/Ais up to homotopy the same as X, i.e., the quotient
map X X/A is a homotopy equivalence. Unfortunately this is not true.
Exercise A23.5. Let Z be the topologists sine curve of Exercise A15.13, and
let A Z be the interval (0, y) : 1 y 1 on the y-axis. Clearly A is
contractible. Show that Z/Ais path-connected. Since Exercise A15.13 shows that
Z itself is not path-connected, this proves that Z and Z/A cannot be homotopy
equivalent.
What is missing in this example is the homotopy extension property (HEP).
Denition A23.6. Let X be a space and A a closed subspace. One says that
(X, A) has the homotopy extension property if any map from the subspace
X 0 A [0, 1] X [0, 1]
to another space Y can be extended to a map from the whole of X [0, 1] to Y .
The way to think of this is that the initial data consist of a map X Y
together with a homotopy of the restriction of that map to the subspace A; the
HEP says that this homotopy can be extended to a homotopy dened on the whole
space X. Most nice pairs of spaces have the HEP: proofs are often inductive
based on the following example.
Example A23.7. The pair (D
n
, S
n1
) has the homotopy extension property. To
see this, it is enough to construct geometrically a retraction
r : D
n
[0, 1] D
n
0 S
n1
[0, 1]
(which can be done for instance by projection from the point (0, 3/2) in R
n+1
);
then the desired extension can be dened by composing with r.
76
Proposition A23.8. Suppose that (X, A) has the HEP and that A is contractible.
Then the quotient map : X X/A is a homotopy equivalence.
Proof. The rst order of business is to nd a potential homotopy inverse. Let
p A be a point and let h: A [0, 1] A be a homotopy between the identity
map and the constant map that sends A to p. By the HEP, the homotopy h extends
to a homotopy H between the identity on X and some map G: X X, such that
G(a) = p for all a A. Note that G respects the equivalence relation dening the
quotient space, so there is a map g : X/A X such that g = G. This map g
is, I claim, a homotopy inverse for .
To see this we must show that g and g are homotopic to the identity on
X and X/A respectively. The rst is easy: g = G is homotopic to the identity
by construction. What about the other? The homotopy H has H(a, t) A for
all a A and all t. Therefore, H(a, t) X/A is constant for all a A and
t [0, 1]; that is, H respects the equivalence relation. It follows that there is a
map
k: (X/A) [0, 1] (X/A)
such that k((x), t) = (H(x, t)) for all x X, t [0, 1]. In particular, k gives
the desired homotopy between g and the identity on X/A.
77
Lecture A24
The fundamental group
For the next few lectures we are going to consider pointed topological spaces.
Such an object is just a topological space X together with a choice of a base point
x
0
X. A map of pointed spaces (X, x
0
) (Y, y
0
) is just a map from X to Y
that takes x
0
to y
0
. Sometimes, to save space, well omit explicit mention of the
base point.
Example A24.1. We regard the unit circle S
1
as the quotient space [0, 1]/0, 1.
It is a pointed space whose base point is the equivalence class 0, 1.
Denition A24.2. A loop in a pointed space (X, x
0
) is a map of pointed spaces
(S
1
, ) (X, x
0
). In other words, it is a path : [0, 1] X such that (0) =
(1) = x
0
.
The natural notion of homotopy for maps of pointed spaces is based homotopy:
a based homotopy from (X, x
0
) to (Y, y
0
) is a homotopy h: XI Y such that
h(x
0
, t) = y
0
for all t I. In particular we can consider the collection of all based
homotopy classes of loops in (X, x
0
). This object is denoted
1
(X, x
0
) and it is
called the fundamental group of (X, x
0
). As we shall see, there is indeed a natural
way to make it into a group.
Lemma A24.3. Let be a based loop in (X, x
0
). Let g : [0, 1] [0, 1] be a
continuous map with g(0) = 0 and g(1) = 1. Then and g represent the same
element of
1
(X, x
0
). (The composite g is called a reparameterization of .)
Proof. The map
h(s, t) = ((1 t)s + tg(s))
denes a homotopy between and the reparameterized path g.
Let

and

be paths in X. Recall that the concatenation of these paths is


dened by

(s) =
_

(2s) (s
1
2
)

(2s 1) (s
1
2
)
If

and

are (based) loops then so is

. Moreover, if

or

is var-
ied by a (based) homotopy, then so is

. Thus the operation passes to a


multiplication operation on
1
(X, x
0
).
78
Proposition A24.4. Under this operation,
1
(X, x
0
) becomes a group.
Proof. There are three things that must be veried: associativity, the existence of
an identity, and the existence of inverses. The proofs of all three make extensive
use of the freedom to deform loops by homotopies and, in particular, to reparam-
eterize them.
For associativity, we just need to note that the paths (
1

2
)
3
and
1
(
2

3
)
are related by the reparameterization
s
_

_
2s s
1
4
s +
1
4
1
4
s
1
2
1
2
+
1
2
s
1
2
s 1
The constant loop (e(s) = x
0
for all s [0, 1]) denes an identity element.
Indeed, it is easy to see that e and e are simply reparameterizations of ,
for any loop .
Finally suppose that is a loop and let
1
be the loop dened by
1
(s) =
(1 s). Then
1
is the path s (g(s)) where g(s) = 2s for s
1
2
and
g(s) = 2 2s for s
1
2
. Then
h(s, t) = ((1 t)g(s))
is a homotopy of based loops from
1
to the identity path.
The effect of the choice of basepoint is summed up in the following.
Lemma A24.5. Let X be a path-connected space and let x
0
, x
1
X. Then the
groups
1
(X, x
0
) and
1
(X, x
1
) are isomorphic.
Proof. Let q : [0, 1] X be a path from x
0
to x
1
, and let q
1
(s) = q(1 s), as
usual. Then, if is a loop based at x
0
, the path
() := q
1
q
is a loop based at x
1
, and the assignment () gives a homomorphism of
groups from
1
(X, x
0
) to
1
(X, x
1
). The assignment

q
1
gives the
inverse homomorphism
1
(X, x
1
) to
1
(X, x
0
), so these two groups are isomor-
phic.
79
Beware that the isomorphism in the lemma above depends on the choice of the
path q. One says that the groups
1
(X, x
0
) and
1
(X, x
1
) are isomorphic but not
canonically isomorphic.
Let f : (X, x
0
) (Y, y
0
) be a (basepoint-preserving) map. If is a based loop
in X, the composite f is a based loop in Y . Moreover, the assignment []
[f ] passes to homotopy classes and denes a homomorphism
1
(X, x
0
)

1
(Y, y
0
).
Denition A24.6. The homomorphism so dened is called the induced homomor-
phism of the map f, and it is denoted by f

:
1
(X, x
0
)
1
(Y, y
0
).
Proposition A24.7. The induced homomorphism has the following properties:
(a) The identity map induces the identity homomorphism.
(b) Suppose that f and g are maps (X, x
0
) (Y, y
0
), which are homotopic by
a homotopy h. Let u
1
(Y, y
0
) be dened by the loop t h(x
0
, t). Then
f

(v) = ug

(v)u
1
for all v
1
(X, x
0
). In particular, based homotopic
maps induce the same homomorphism.
(c) If f : X Y and g : Y Z are based maps, then (f g)

= f

.
Item (c) is called functoriality: it is an extremely important property in many
mathematical contexts.
Proof. (a) and (c) are immediate consequences of the denitions. Part (b) would
be easier if we restricted our attention to based homotopies, but we need the gen-
eral case. Consider the map from the unit square to Y dened by
(s, t) h((s), t).
Going along the bottom and right sides of this square denes the path f u, going
along the left and top sides denes u g. But these are homotopic by a homotopy
that goes diagonally across the square. Explicitly, we want to dene
H(r, s) =
_
h(2r(1 s), 2rs) (r
1
2
)
h((1 s)(2r 1) + s, s(2r 1) + 1 s) (r
1
2
)
which gives a homotopy from H(, 0) = f u to H(, 1) = u g.
Corollary A24.8. A homotopy equivalence induces an isomorphism of fundamen-
tal groups.
80
Proof. Let f : X Y be a homotopy equivalence, with homotopy inverse g.
Then fg and gf are homotopic to the identity, and therefore by (a) and (b) above
they induce isomorphisms on fundamental groups. Thus by (c) above, f

and
g

are isomorphisms, which implies that f

and g

are isomorphisms too.


In particular
Corollary A24.9. If X is contractible, then its fundamental group is trivial.
81
Lecture A25
Fibrations and the fundamental group
Recall the homotopy extension property for a pair X, A), which we can also
think of as a property of the inclusion map i : A X. The pair has the HEP (or,
as one also says, the inclusion i is a cobration) if, given any homotopy of maps
A Y , together with an extension of the initial data of the homotopy to a map
X Y , we can always extend the homotopy to a homotopy of maps X Y ,
respecting the given initial data.
The notion of extension of a map can be expressed in diagrammatic terms
using the inclusion i:
X

A
?

i
OO
//
Y
Here the solid arrows represent the given data and the dotted arrows represent
the extension to be constructed. Dual to the notion of extension is the notion of
lifting: suppose that p: E B is a surjective map and f : Y B is any map,
then a lifting of f (over p) is a map g : Y E that makes the following diagram
commute:
E
p

Y
>>
//
B
in other words p g = f.
Denition A25.1. We say that p: E B has the homotopy lifting property or is
a bration if, given any homotopy of maps Y B together with a lifting of the
initial data of the homotopy over p, the entire homotopy can be lifted over p.
In other words, p has the HLP if, given the data expressed by the solid arrows
in the diagram below
Y 0

_

//
E

Y [0, 1]
//
<<
B
82
the dotted arrow can be lled in in such a way as to make the diagram commuta-
tive.
Remark A25.2. The homotopy extension property can be expressed in a similar
diagrammatic way, but to do so we have to consider a homotopy of maps A Y
as a single map from A to the space Y
I
of maps I Y . We havent discussed the
topology of mapping spaces, so we cant write down the full details of this, but it
should not be hard to gure out the general idea.
For now we are not going to worry too much about where brations might
come from. Instead we will concentrate on what they might tell us about the
fundamental group.
Denition A25.3. Let X be a space. If X is path-connected and
1
(X, x
0
) = 1
for some (and hence any) basepoint x
0
, we will say that X is simply connected.
Contractible spaces, for example, are simply connected (Corollary A24.9).
But there are many other examples.
Exercise A25.4. Show that the n-sphere S
n
is simply connected for n 2. Hints:
First show that it is enough if one knows that any loop in S
n
is homotopic to
one that is not surjective. Then prove this statement by any one of a number
of approximation techniques, e.g. by showing that any loop is homotopic to a
piecewise linear loop made up of great circle arcs.
Denition A25.5. Let p: E B be a bration, and let B have a xed basepoint
b
0
. The inverse image F = p
1
b
0
is called the ber of the bration.
Suppose now that p: E B is a bration and let b
0
B be a base point. Let
be a based loop in B and let x F be a point of the ber F = p
1
(b
0
). We may
consider the path as a homotopy (of maps of a point into E) and, if we do this,
and x together provide exactly the data for a homotopy lifting problem:
0

_

x
//
E

[0, 1]

//

>>
B
Filling in the dotted arrow provides a lifting of the path , but this lifting is
not guaranteed to be a loop! All we know is that p( (1)) = (1) = b
0
, that is,
83
(1) is a point of thee ber F. It is not uniquely determined by this construction
but we shall see in a moment that the path component of the ber in which it
lies is uniquely determined. The proof begins to build up the connection between
brations and the fundamental group.
Lemma A25.6. Let p: E B be a bration, as above, and let be a based loop
in B. If is nullhomotopic (that is, it represents the identity element in
1
(B, b
0
)),
then, for any lift : [0, 1] E of , (0) and (1) are in the same path component
of the ber F.
(Its obvious that (0) and (1) are in the same path component of E they
are joined by the path , after all! so the signicance of the lemma is that a
path joining them can be found that lies wholly in F.)
Proof. We apply the HLP to a homotopy between and the constant path. Let
h(s, t) be such a homotopy with h(s, 0) = (s) and h(s, 1) = b
0
. Then h together
with provide the data for a homotopy lifting problem:
[0, 1] 0

_


//
E

[0, 1] [0, 1]
h
//
H
<<
B
The bottom, right-hand and top sides of the map of the unit square dened by h
are constant maps with value x
0
. Therefore the bottom, right-hand and top sides
of the lifted homotopy H are paths in the ber F. The concatenation of these three
is a path in F from (0) to (1).
Corollary A25.7. If

are two liftings of the same loop, starting at the same


point

(0) =

(0), then their end points are in the same path component of F.
Proof. The concatenation (

)
1

is a lifting of a nullhomotopic loop.


To help keep track of this business of path components lets introduce some
notation.
Denition A25.8. For any topological space F, let
0
(F) denote the set of path
components of F.
84
Suppose that p: E B is a bration with ber F, as discussed above. For
g
1
(B, b
0
) and c
0
(F), let us dene c
g
(read: c acted on by g) as follows:
choose a based loop representing g, and a point x F representing c, let be
any lift of starting at x, and dene c
g
to be the path component containing (1).
The lemma and corollary above show that this is well dened.
Proposition A25.9. The construction above denes an action of the fundamental
group
1
(B, b
0
) on the set
0
(F). Moreover, if E is path connected, then this
action is transitive; and if in addition E is simply connected, the action is free.
Let us explain the terminology. A (right) action of a group G on a set C is
just a map C G C, written (c, g) c
g
, such that c
e
= c and (c
g
)
h
= c
gh
.
An action is free if c
g
= c implies g = e, and it is transitive if for all pairs c, c

of
elements of C, there exists g G such that c
g
= c

. When a group action is both


free and transitive, the mapping g c
g
, for xed c, is a bijection from the group
G to the set C.
Proof. The action law (c
g
)
h
= c
gh
follows from the fact that a lifting of a concate-
nation of loops is a concatenation of liftings of those loops. And c
e
= c follows
from the fact that a constant loop can be lifted to a constant.
Suppose now that E is path connected. Then given any two points x, x

of
the ber there is a path in E joining them. Composing with p gives a based loop
in B which (tautologically) lifts to a path in E from x to x

. Thus the action is


transitive.
Suppose additionally that E is simply connected. Suppose that c
g
= c for some
c
0
(F) and g
1
(B, b
0
); this means that g is represented by a based loop
in B which lifts to such that (0) and (1) are in the same path component
of F. Construct a loop in E by concatenating with a path in F from (0) to
(1). The composite p is the concatenation of with a constant path, so it still
represents g
1
(B, b
0
). But, on the other hand, is nullhomotopic in E (say
by a homotopy H) because E is simply connected. Then p H is a nullhomotopy
of the path p , which therefore represents the identity element of
1
(B, b
0
) as
claimed.
85
Lecture A26
Covering Spaces
Let p: E B be a surjective map of topological spaces.
Denition A26.1. The map p above is called a covering map (and E is called a
covering space of B) if there is a discrete topological space F (the ber) such that
each point x B has an open neighborhood U for which there is a homeomor-
phism p
1
(U)

= U F making the diagram
p
1
(U)

=
//
p
##
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
U F
pr
1

U
commute.
In other words, a covering map is locally the projection of a product with
a discrete space. Some authors (e.g. Hatcher) give a denition which appears
more general than this, but it turns out that the denitions are equivalent for path-
connected base spaces B. A neighborhood U with the property described in the
denition will be called a trivializing neighborhood for the bration.
Example A26.2. A homeomorphism is a covering map with 1-point ber.
Example A26.3. The map R S
1
dened by x x mod 1 (if we think of
the circle as [0, 1] with endpoints identied) or x e
2ix
(if we think of the unit
circle in C) is a covering map with ber Z.
Example A26.4. The map S
1
S
1
dened by x nx mod 1 (or z z
n
in
the complex pictures) is a covering map with ber an n-point space.
Example A26.5. There are many different possible covering spaces of the gure
8 space; Ill draw some pictures in class.
Example A26.6. Let B be a compact oriented surface of genus 2 (i.e. a the
surface of a 2-holed donut). Then B can be obtained by a suitable identication
of the edges of a regular octagon. One can tessellate the hyperbolic plane by
such octagons with vertex angle of 45 degrees. From this tessellation we obtain a
covering map from the plane to B.
86
Proposition A26.7. A covering map is a bration, i.e., it satises the homotopy
lifting property.
Proof. Remember what we have to show: for each space Y we can ll in the
homotopy lifting diagram
Y 0

_

f
//
E

Y [0, 1]
h
//
H
<<
B
First of all lets consider the case when Y is a single point (the path lifting
property). By denition, B has an open cover consisting of trivializing neigh-
borhoods. The inverse image of this open cover by the map h is then an open
cover of the compact space [0, 1], which has a positive Lebesgue number by
the Lebesgue covering theorem. Partition [0, 1] into nitely many subintervals
0 t
1
t
2
1 of length less than the Lebesgue number. We will
construct the desired lifting by induction over these subintervals.
Suppose then than a lifting H has been constructed on [0, t
k
]. We have h([t
k
, t
k+1
])
U for some trivializing neighborhood U, and in particular we may identify p
1
(U)
with U F where F is the (discrete) ber. By connectedness, pr
2
(H(t)) must be
constant for any lift H. Therefore the unique continuous lift H of h on [t
k
, t
k+1
]
that agrees with the lift already assumed to have been constructed on [0, t
k
] is
H(t) = (h(t), pr
2
(H(t
k
))) U F

= p
1
(U) E.
By induction, we can construct H on [0, 1]. Indeed, we have done more than
claimed: we have shown that there is a unique possible choice for H.
Now we return to the case of general Y . Because of the uniqueness of path
lifting, which we just proved, there is no question how to dene H for general Y :
namely, for each xed y Y , H(y, t) should be the unique lifting of the path t
h(y, t) that begins at f(y). The only question is whether the function H dened
by this process is continuous as a function of y and t. To see this, x y
0
Y and
consider the following fact: for each t [0, 1] there is a product neighborhood
V
t
W
t
of (y
0
, t) such that h(V
t
W
t
) lies in a trivializing neighborhood in B.
By compactness, nitely many of these product neighborhoods V
t
W
t
cover
87
y
0
[0, 1]. Let V be the intersection of the nitely many V
t
s that appear and
let be a Lebesgue number for the covering of [0, 1] by the nitely many W
t
s.
Partition [0, 1] into nitely many subintervals 0 t
1
t
2
1 of length
less than . Notice that for each k, h(V [t
k
, t
k+1
]) is contained in a trivializing
neighborhood.
Were going to prove by induction over subintervals that H is continuous on
V [0, 1]. Suppose then that we already know that H is continuous on V [0, t
k
];
in particular, H(y, t
k
) is a continuous function of y for y V . But by the formula
above, for t [t
k
, t
k+1
],
H(y, t) = (h(y, t),
2
(H(y, t
k
))) U F

= p
1
(U) E,
and this is a continuous function of (y, t). So the induction is complete and we
have shown (in particular) that H is continuous at (y
0
, t) for all t [0, 1]. Since
y
0
was arbitrary, this sufces to complete the proof.
88
Lecture A27
Applications
So, after all this machinery, we can nally compute
1
(S
1
).
Theorem A27.1. The group
1
(S
1
) is isomorphic to Z.
Proof. We make use of the covering space R S
1
described above (Exam-
ple A26.3). The ber F = Z is discrete and the total space E = R is contractible,
so according to Proposition A25.9 the action of
1
(S
1
) on
0
(F) = F is free and
effective. We can therefore dene a bijection
1
(S
1
) Z by sending g G to
0
g
Z. All that is necessary is to show that this is a group isomorphism.
For this, notice that S
1
itself is a topological group. In any topological group
G, the group operation in
1
(G) can be dened by pointwise multiplying paths
(using the group operation in G), instead of by concatenating them. For the ex-
pression
H(s, t) =

(max((2 t)s, 1))

(min(0, 1 (2 t)(1 s)))


denes a homotopy fromthe concatenation of

and

to their pointwise product.


However, the lift of a pointwise product of paths (in S
1
) is clearly their pointwise
sum (in R); in particular, the action of
1
(S
1
) on the ber Z satises 0
gh
= 0
g
+0
h
.
This proves that the bijection
1
(S
1
)
0
(Z) = Z is indeed an isomorphism of
groups.
Exercise A27.2. Using a similar argument to the above, show that if G is any
topological group (not necessarily abelian), then
1
(G) is an abelian group.
Remark A27.3. The identity map S
1
S
1
is a generator of the cyclic group

1
(S
1
) = Z, as the above proof shows. The integer n
1
(S
1
) is represented by
the map z z
n
, which wraps the unit circle n times around itself.
Here are some (standard) applications of this calculation.
Proposition A27.4. (No-retraction theorem) There is no retraction of the disk
D
2
= z C : [z[ 1 onto its boundary circle S
1
= z C : [z[ 1.
89
Proof. A retraction is a map r : D
2
S
1
that is a left inverse to the inclusion
i : S
1
D
2
, that is. the diagram
S
1
p

A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
=
//
S
1
D
2
r
OO
should commute. But if this is so then the induced diagram

1
(S
1
)
=
//
##
G
G
G
G
G
G
G
G
G
G
G
G
G
G
G
G

1
(S
1
)

1
(D
2
)
r
OO
should commute also, and this is impossible since
1
(S
1
)

= Z while
1
(D
2
) is
trivial.
Proposition A27.5. Any continuous map f : D
2
D
2
has a xed point.
Proof. Suppose for a contradiction that f : D
2
D
2
has no xed point. Then
for each x D
2
, let u = u(x) be the unit vector (x f(x))/|x f(x)|. Let
= (x) be the non-negative root of the quadratic equation
|x + u|
2
=
2
+ 2u x +|x|
2
= 1.
Then g(x) = x +u depends continuously on x and it is the point on S
1
obtained
by continuing the ray from f(x) through x until it hits the unit circle. Clearly, if
x S
1
, then g(x) = x. Thus g is a retraction of D
1
onto S
1
, contradicting the
previous result.
Let a C and let : S
1
C a be a loop in the complex plane that does
not pass through a. The map
a
(z) = (z a)/[z a[ is a homotopy equivalence
between C a and the unit circle. The element of
1
(S
1
) = Z represented by
the composite

a
f : S
1
S
1
is called the winding number of f about a, and is denoted n(f; a). It is a homotopy
invariant. (It does not depend on the choice of base point because
1
(S
1
) = Z is
abelian.)
90
Lemma A27.6. (Rouch es theorem) If f : S
1
C0, g : S
1
C, and [g(z)[ <
[f(z)[ for all z S
1
, then n(f; 0) = n(f + g; 0).
Proof. The map h(z, t) = f(z) + tg(z) is a homotopy between f and f + g in
C 0.
Theorem A27.7. (Fundamental theorem of algebra) Every non-constant polyno-
mial over C has a root.
Proof. Suppose not and let p(z) = z
n
+ a
n1
z
n1
+ + a
0
be a polynomial
without root. Let p
r
: S
1
C 0 be the loop dened by p
r
(e
i
) = p(re
i
).
These are homotopic loops in C 0. But for r = 0 the winding number n(p
r
, 0)
is zero, whereas for large r we may write
p(re
i
) = f(re
i
) + g(re
i
), f(z) = z
n
and simple estimates show[f(re
i
)[ > [g(re
i
)[, so by Rouch es theoremn(p
r
, 0) =
n(z
n
, 0) = n. Contradiction.
Lemma A27.8. Let f : S
1
S
1
be a loop with is antipodal, that is, f(z) =
f(z). Then f represents an odd multiple of the generator of
1
(S
1
).
Proof. Let g(z) = zf(z); then g(z) = g(z) so g respects the equivalence rela-
tion z z. The quotient of S
1
by this equivalence relation is again a copy of S
1
,
and the quotient map is identied with the squaring map s(z) = z
2
. Consequently,
g factors through s and thus g

factors through s

1
(S
1
) = Z
s
//

1
(S
1
) = Z
//

1
(S
1
) = Z .
But s

is multiplication by 2, so the winding number of g is even. The winding


number of f is 1 less than that of g, so it is odd.
Theorem A27.9. (Borsuk-Ulam) For every continuous map f : S
2
R
2
there
exists x S
2
such that f(x) = f(x).
Proof. Suppose not and let f be a counterexample. Then the map
g(x) = (f(x) f(x)[/|f(x) f(x)|
sends S
2
to S
1
and satises g(x) = g(x). Consider the restriction of g to the
equator on S
1
. This is an antipodal map S
1
S
1
, so it represents an odd multiple
of the generator; in particular it is not nullhomotopic. This is a contradiction since
it factors through the simply connected space S
2
.
91
Corollary A27.10. (Ham sandwich theorem) Any three bounded measurable sub-
sets of R
3
(e.g. ham, bread, cheese) can be simultaneously volume-bisected by a
common plane cut.
Proof. For each unit vector u S
2
let P
u
be the plane having u as normal vector
that volume-bisects one of the ingredients, say the bread. (The existence of such
a plane follows from the intermediate value theorem; in the general case one has
to maneuver a little to deal with the possible non-uniqueness of such a plane; Im
sweeping this under the rug.) Dene a map S
2
R
2
by sending u S
2
to the
vector (h(u), c(u)), where h(u) and c(u) are the fractions of the ham and cheese,
respectively, that lie on the u side of the plane P
u
. Then h(u) = 1 h(u)
and c(u) = 1 c(u). The Borsuk-Ulam theorem gives us a point such that
h(u) = h(u) and c(u) = c(u). This implies that h(u) = c(u) =
1
2
; thus P
u
perfectly divides the ham and the cheese (as well as the bread).
92

You might also like