Romyar SharifiLECTURE NOTES
READING EDITIONPDF

LECTURE NOTES / Chapter 3

Point-Set Topology

Romyar Sharifi

Chapter 3 Connected and compact spaces

Book contents

Chapter 3
Connected and compact spaces

3.1. Connectedness and path connectedness

Definition 3.1.1.

A topological space X is connected if it is not a disjoint union of any two nonempty open subspaces. Otherwise, X is said to be disconnected.

Lemma 3.1.2.

A topological space X is connected if its only subsets that are both open and closed are and X.

Proof.

If A is a subset of X that is both open and closed, then so is Ac, and then X = 𝐴∐Ac. Conversely, if X = U ∐𝑉 with U and V nonempty and open in X, then U and V are also closed.

Example 3.1.3.

Topological spaces with one element are always connected, but discrete topological spaces with more than one element are disconnected.

Example 3.1.4.

The union (,0)(0,) of intervals in is disconnected as a subspace of , as both (,0) and (0,) are open and closed in the subspace topology. However, itself is connected.

Remark 3.1.5.

If X is the disjoint union of subspaces A and B, then A and B are their own closures, so A and B contain no limit points of each other. In fact, if A and B are any two disjoint subsets of X with union X such that A contains no limit points of B and vice-versa, then A and B are open and closed in X.

Lemma 3.1.6.

If X = U ∐𝑉 for subspaces U and V, and A is a connected subset of X, then either A U or A V.

Proof.

We have that AU and AV are open and closed in A, so A is the disjoint union of these intersections. If A is connected, this forces one of the AU and AV to be empty, and then the other is A.

Proposition 3.1.7.

The closure of a connected subset of a topological space is connected.

Proof.

Let X be a topological space and A a connected subset. Suppose that A¯ = U ∐𝑉 for disjoint subspaces U and V in X. By Lemma 3.1.6, we have that A is contained in either U or V: without loss of generality, we suppose A U. As U is closed, we have that A¯ U as well, and therefore V = .

We also have the following statements.

Proposition 3.1.8.

The image of a connected space under a continuous map is connected.

Proof.

Suppose X is connected and f : X Y is continuous. If Y is the disjoint union of (open) subspaces U and V, then X is the union of the disjoint open subsets f1(U) and f1(V ), hence equal to their disjoint union as topological spaces.

Proposition 3.1.9.

Let X = iIAi for a collection of connected subsets Ai of X for i I, and suppose that iIAi is nonempty. Then X is connected.

Proof.

Let a iIAi. If X = U ∐𝑉 for subspaces U and V, then without loss of generality we may suppose a U. Since Ai is connected and contains a, we must then have Ai U for all i I, forcing V = .

The following is an immediate corollary of Proposition 3.1.9.

Corollary 3.1.10.

The union of all connected subsets of a topological space X that contain a given point x X is connected.

Proposition 3.1.11.

The relation on a topological space X given by x y for x,y X if y lies in the connected component of x is an equivalence relation on X.

Proof.

Clearly is reflexive. For symmetry, it’s enough show that for x,y X, the connected components A of x and B of y are either equal or disjoint. Suppose that z AB. Then AB is connected by Proposition 3.1.9, but in that A (resp., B) is the largest connected subset of X containing x (resp., y), we must have A = AB = B. Finally, if x y and y z for x,y,z X, then we’ve just seen that x and y have the same connected component, as do y and z, so x and z do as well.

By Corollary 3.1.10, we can always find a largest connected subset containing a given point.

Definition 3.1.12.

A connected component of a topological space X is a connected subset of X that is not properly contained in any larger connected subset of X. The connected component of a point x X is the unique connected component of X containing x.

Lemma 3.1.13.

A space X is the union of its distinct connected components, which are closed and disjoint. If every connected component of X is open, then X is the disjoint union of them. In particular, if X has only finitely many connected components, then it X is the disjoint union of its connected components.

Proof.

We know that X is the union of its connected components, which are disjoint by Proposition 3.1.11. It follows from Proposition 3.1.7 that connected components are closed, being the largest connected subsets containing a given point. The second statement follows from the definition of a disjoint union of topological spaces. If X has finitely many connected components, then any union of all but one of them is closed, and then they are all open as complements of these unions.

Example 3.1.14.

Consider the subspace A = {0}{1nn 1} of . Every {1n} is both open and closed in A, so is a connected component (in that it is connected). The set {0} is also then a connected component, being that it is not contained in any larger connected subset. However, it is closed but not open, so A is not the disjoint union of its connected components.

Example 3.1.15.

Consider as a subspace of . It is disconnected as is the union of its intersections with the intervals (,π) and (π,), for instance (as π is irrational). In fact, since there exists an irrational number between any two distinct rational numbers, the connected components of are just its singleton subsets.

Remark 3.1.16.

The property of being in the same connected component gives an equivalence relation on the points of a topological space, and the connected components are the equivalence classes.

Definition 3.1.17.

A topological space is said to be totally disconnected if its connected components are its singleton subsets.

Example 3.1.18.

Let A = {0,1} with the discrete topology, and consider X = n=1A with the product topology. Given any two distinct points x = (xn)n1 and y = (yn)n1 in X, there exists n 1 such that xnyn. Letting πn denote the nth projection map, we have that U = πn1(xn) and V = πn1(yn) are disjoint basic open sets in X with union X, so X = U ∐𝑉, and x and y lie in distinct connected components. Thus, X is totally disconnected.

Definition 3.1.19.

A path γ from a point x to a point y in a topological space X is a continuous function γ : [0,1] X such that γ(0) = x and γ(1) = y. The points x and y are called the endpoints of γ: in particular, x is its initial endpoint and y is its final endpoint.

Definition 3.1.20.

Let X be a topological space.

a.

We say that two points x and y in a topological space X can be connected by a path if there exists a path γ from x to y.

b.

We say that a topological space X is path connected if every two points x and y can be connected by a path.

c.

The path component of a point x X is the set of all points y X such that x and y are connected by a path.

Remark 3.1.21.

The relation of there exists a path from a point x to a point y on a topological space X is an equivalence relation on a topological space X. Thus, X is a disjoint union of its path components.

Proposition 3.1.22.

Every path connected space is connected.

Proof.

Let X be a path connected space. Suppose that X is a disjoint union of nonempty open subspaces U and V, and let x X and y Y. Let γ be a path from x to y, and let A = γ([0,1]). Then A is the disjoint union of the nonempty sets AU and AV, so A is disconnected, but it is the image of a connected space under a continuous function.

Example 3.1.23.

Consider the subset A = {(x,sin(1x))0 < x 1} of 2. We have A¯ = A{(0,y)1 y 1}. Note that A¯ is connected as a subspace of 2 since A is. On the other hand, A¯ is not path connected. Actually, this is easily reduced to proving that the topologist’s sine curve B = A{(0,0)}A¯ is connected, but not path connected. Let us explain this.

Suppose that γ is a path from (0,0) to (1,sin(1)) on B. We may assume without loss of generality that γ(t)(0,0) for t > 0. By definition, limt0+γ(t) = (0,0). On the other hand, for every 𝜖 > 0, there exists an x < 𝜖 such that sin(1x) = 1. By the continuity of γ, for every δ > 0, there then exists t < δ such that the second coordinate of γ(t) equals 1. But this contradicts that limt0+γ(t) = (0,0).

3.2. Compactness

Definition 3.2.1.

Let 𝒰 be a cover of a subset A of a topological space X. A subcover of 𝒰 is a a subset of 𝒰 that covers A.

Definition 3.2.2.

A topological space X is compact if every open cover of X has a finite subcover.

Examples 3.2.3.

a.

Any finite topological space is compact.

b.

Any topological space with the trivial topology is compact.

c.

The real line is not compact, since the collection of open intervals of length 1 is an open cover with no finite subcover.

d.

The interval (0,1] is not compact, since the connection of intervals (𝜖,1] with 𝜖 > 0 has no finite subcover.

Proposition 3.2.4.

Any closed interval in is compact.

Proof.

As all closed intervals of finite length are homeomorphic, we can and will consider the interval [0,1]. Let 𝒰 be an open cover of [0,1], and let A be the subset of [0,1] consisting of those x such that [0,x] has a finite subcover by elements in 𝒰. Let b be the supremum of of the elements of A. If b < 1, then let U be an element 𝒰 containing b. Since U contains an interval (b𝜖,b+𝜖) for some 𝜖 > 0 with b+𝜖 1, we can find a finite subcover 𝒱 of [0,b𝜖 2] inside 𝒰. Ten 𝒱{U} is a finite subcover of [0,b+ 𝜖 2] in 𝒰. This means that b+ 𝜖 2 A, contradicting the fact that b is the supremum of all elements of A. Thus b = 1, and therefore 𝒰 has a finite subcover.

Let’s establish a few basic statements regarding compact spaces.

Lemma 3.2.5.

A subspace A of a topological space X is compact if and only if every open cover of A in X has a finite subcover.

Proof.

Given an open cover 𝒱 of A by open sets in A, we can find a set 𝒰 of open sets in X covering A such that 𝒱 = {AUU 𝒰}. Conversely, given a collection 𝒰 of open sets in X covering A, we may define a cover 𝒱 of A by open sets in A by taking intersections with A as in the latter formula. Any (finite) subset 𝒞 of 𝒰 covers A if and only if the (finite) subset {AUA 𝒞} of 𝒱 covers A.

Proposition 3.2.6.

Every closed subset of a compact space is compact.

Proof.

Let X be compact, and let A X be closed. Let 𝒰 be an open cover of A in X. Then 𝒰{Ac} is an open cover of X, so it has a finite subcover 𝒱. If Ac 𝒱, then 𝒱{Ac} is an open cover of A in X, and otherwise, 𝒱 is an open cover of A in X.

Lemma 3.2.7.

Let A be a compact subset of a Hausdorff space, and let x Ac. Then there exist disjoint open sets U and V with A U and x V.

Proof.

For each a A, choose open disjoint neighborhoods Ua of a and Va of x in X. The collection {Uaa A} is an open cover of A, and it has a finite subcover, say by Ua1,,Uan. Then U = i=1nUa i and V = i=1nVa i are the desired open subsets of X.

Proposition 3.2.8.

Every compact subset of a Hausdorff space is closed.

Proof.

Let X be Hausdorff, and let A X be compact. By Lemma 3.2.7, for each x Ac, there exists an open neighborhood Vx of x in Ac. The union of the Vx is Ac, so Ac is open, and thus A is closed.

Proposition 3.2.9.

Let f : X Y be a continuous map. If X is compact, then so is the image of f.

Proof.

Let 𝒱 be an open cover of f(X) in Y. Then 𝒰 = {f1(V )V 𝒱} is an open cover of X. Since X is compact, it has a finite subcover {f1(V1),,f1(Vn)} with each Vi 𝒱, and then {V1,,Vn} is an open cover of f(X).

We have seen that continuous bijections need not be homeomorphisms. However, continuous bijections from compact spaces are.

Theorem 3.2.10.

Let f : X Y be a continuous surjection. If X is compact and Y is Hausdorff, then f is closed and a quotient map.

Proof.

To see that f is a quotient map, we may show that B Y is closed if f1(B) is closed. Since f is continuous and surjective, we may write B = f(f1(B)), so it suffices to show that f is a closed map.

Let A be a closed set in X, which is necessarily compact. As f is continuous, its image f(A) is compact as well. As Y is Hausdorff, we then have that f(A) is closed.

Corollary 3.2.11.

Every continuous bijection from a compact space to a Hausdorff space is a homeomorphism.

We next prove that a finite product of compact spaces is compact. First, we require the following lemma.

Lemma 3.2.12.

Let X and Y be topological spaces, and suppose that X is compact. For any y Y and open set W in X ×Y containing X ×{y}, there exists an open neighborhood V of y in Y such that X ×V W.

Proof.

For each x X, let Ux be an open neighborhood of x in X and Vx be an open neighborhood of y in Y such that Ux×Vx W, which exist since (x,y) W and W is open. The collection {Uxx X} covers X so has a finite subcover, say consisting of Ux1,,Uxn. If we set V = i=1nVx i, then V is an open neighborhood of y in Y, and since each Uxi ×V is contained in W, so is their union X ×V.

Theorem 3.2.13.

Let X1,,Xn be compact spaces for some n 1. Then i=1nXi is compact.

Proof.

It suffices by recursion to consider the case n = 2, so the product of compact spaces X and Y. Let 𝒲 be an open covering of X ×Y. For any y Y, the set X ×{y} is compact, being homeomorphic to X via the projection map, and so there exist Z1,,Zm 𝒲 that together cover X ×{y}. Set Wy = i=1mZi. We then have by Lemma 3.2.12 that there exists an open neighborhood of Vy of y such that X ×Vy Wy. The collection {Vyy Y} then covers Y, and it has a finite subcover as Y is compact. But then X ×Y is covered by the finitely many Wy, and each of these is in turn a union of finitely many sets in 𝒲. Thus, 𝒲 has a finite subcover.

Proposition 3.2.14.

A subspace of n is compact if and only if it is closed and bounded.

Proof.

Fix a subset A of n. We may consider the cover of A by all open balls in n of radius 1. If A is unbounded, then it cannot have a finite subcover, since the union of any finite number of balls of radius 1 is bounded. Therefore, only bounded subsets can be compact.

Now suppose that A is bounded. It is then contained in a direct product of closed intervals, which is compact as a finite product of compact sets. If A is also closed, then it is compact by Proposition 3.2.6. On the other hand, if A is not closed, then it cannot be compact by Proposition 3.2.8.

We give another criterion for a topological space to be compact.

Definition 3.2.15.

A collection 𝒜 of subsets of X is said to have the finite intersection property, or FIP, if for every n 1 and A1,,An 𝒜, we have that i=1nAi is nonempty.

Theorem 3.2.16.

A topological space X is compact if and only if every collection 𝒜 of closed subsets of X having the finite intersection property satisfies A𝒜A is nonempty.

Proof.

Let 𝒜 be a collection of closed subsets of X with the finite intersection property. Then 𝒰 = {AcA 𝒜} is a collection of open subsets of X with the property that no finite subset of 𝒰 covers X. If X is compact, this forces 𝒰 not to cover X, which is exactly to say that the intersection of elements of 𝒜 is nonempty.

Conversely, if X is not compact, there exists an open cover 𝒰 of X which has no finite subcover, which is to say that the collection 𝒜 = {UcU 𝒰} of closed sets has the finite intersection property but has empty intersection.

Let us give an example of the use of this theorem.

Theorem 3.2.17.

Let X be a nonempty compact Hausdorff space with no singleton open sets. Then X is uncountable.

Proof.

If X is finite Hausdorff, then it is discrete, so has singleton open sets. So, suppose by way of contradiction that X is countably infinite. Label its points xi for i 1. Let X = U0, which is an open neighborhood of x1. By Lemma 1.5.13, we may recursively let Ui be an open neighborhood of xi+1 contained in Ui1 with xiUi¯ for each i 1.

The intersection of any finite subcollection of {Uii 1} is some Un for n 1 since these Ui are nested. Since xn+1 Un, this collection satisfies the FIP. In particular, the collection {Ui¯i 1} of closed sets does as well. Now, X being compact, we then have that i=1Ui¯ is nonempty. At the same time, since xiUi¯, this intersection does not contain any of the points of X = {xii 1}, providing the desired contradiction.

We can use this to give a proof of the uncountability of the real numbers.

Corollary 3.2.18.

The set of real numbers is uncountable.

Proof.

By Theorem 3.2.17, the interval [0,1] is uncountable, and therefore so is .

3.3. Sequential and limit point compactness

We next introduce related notions to compactness.

Definition 3.3.1.

A space X is said to be limit point compact if every infinite subset of X has a limit point.

Proposition 3.3.2.

Compact spaces are limit point compact.

Proof.

Let X be a compact space, and let A be a subset without a limit point. Then A is necessarily closed. Moreover, for each a A, there exists an open set Ua containing a and no other point of A. Then X has an open cover by {Uaa A}{Ac}. Since X has a finite subcover, A is contained in a finite union of sets Ua1,,Uan, which implies that A = {a1,,an}, so A is finite.

Definition 3.3.3.

A space X is said to be sequentially compact if every sequence in X has a convergent subsequence.

The following is fairly immediate from the definitions.

Proposition 3.3.4.

Sequentially compact spaces are limit point compact.

Proof.

Let A be an infinite set in X, and let (an)n1 be a sequence of distinct elements in A. It has a convergent subsequence, say with limit a. Then every neighborhood of a contains some point of the subsequence not equal to a. Thus, a is a limit point of A.

To be limit point compact is a weaker notion than being either compact or sequentially compact.

Example 3.3.5.

Let Y = {a,b} the the two-point space with the trivial topology, and consider X = ×Y, where has the discrete topology. Then any open neighborhood of a point (n,a) in X contains (n,b) and conversely, so every nonempty set in X has a limit point. In particular, X is limit point compact. However, X is not compact, and it is covered by the disjoint open sets {n}×Y with n . It is also not sequentially compact, as the sequence of points (n,a) has no convergent subsequence.

In general, neither compactness nor sequential compactness implies the other. However, for metric spaces, all three of these notions of compactness are equivalent, as we shall show.

Definition 3.3.6.

Let A be a bounded subset of a metric space X. The diameter of A is the supremum of the distances between points in A.

Definition 3.3.7.

Let 𝒰 be an open cover of a space X. Its Lebesgue number is the supremum of all 𝜖 > 0 such that every subset A of X of diameter less than 𝜖 is contained in an element of 𝒰, if such an 𝜖 exists, and otherwise, we say 𝒰 has infinite Lebesgue number.

Lemma 3.3.8.

Let X be a sequentially compact metric space. Then every open cover of X has finite Lebesgue number.

Proof.

Suppose that some open cover 𝒰 of X has infinite Lebesgue number. For each n 1, there exists a subset An of X of diameter less than 1n not contained in any element of 𝒰. For each such An, choose xn An. The sequence (xn)n1 has a convergent subsequence (xnk)k1, say with limit x. Let U 𝒰 be an open neighborhood of x, and let B(x,δ) be an open ball inside of it. Let k be sufficiently large such that 1n k < δ 2 and d(x,xnk) < δ 2. By the triangle inequality, we then have Ank B(x,δ) U, providing the desired contradiction.

Definition 3.3.9.

We say that a metric space X is totally bounded if for every 𝜖 > 0, there exists a finite cover of X by open balls of radius 𝜖.

Lemma 3.3.10.

Every sequentially compact metric space is totally bounded.

Proof.

Let X be a metric space, and suppose that 𝜖 > 0 is such that X cannot be covered by finitely many balls of radius 𝜖. Choose x1 X and then recursively choose xn in the complement of i=1n1B(xi,𝜖). The sequence (xn)n1 satisfies d(xm,xn) 𝜖 for all m > n, and consequently it cannot have a convergent subsequence (since every convergent subsequence is necessarily Cauchy). Thus, X not sequentially compact.

Theorem 3.3.11.

Let X be a metrizable space. Then the following are equivalent:

i.

X is compact,

ii.

X is limit point compact,

iii.

X is sequentially compact.

Proof.

Let X be a limit point compact space. Let (xn)n1 be a sequence in X. If the set A = {xnn 1} of values of the sequence is finite, then (xn)n1 has a constant, hence convergent, subsequence. Otherwise, A is infinite, so has a limit point x X. Inductively, we have that every ball among the B(x, 1 k) for k 1 contains some xnk with nk nk1 if k 2. The sequence (xnk)k1 then converges to x. Thus, X is sequentially compact.

Now suppose that X is sequentially compact, and let 𝒰 be an open cover of X. Let 𝜖 > 0 be such that every subset of diameter less than 𝜖 is contained in an element of 𝒰. Choose a finite open cover of X by open balls of radius 𝜖3, and note that they have diameter at most 2𝜖 3, hence are each contained in some element of 𝒰. Then 𝒰 has a finite subcover by these elements, and therefore X is compact.

3.4. Tychonoff’s theorem

We briefly recall a few notions from set theory. In particular, recall that a relation on a set X is a subset of X ×X, and if R is such a relation, we often write 𝑎𝑅𝑏 to denote (a,b) R. We have already used the notion of an equivalence relation earlier in the notes without comment. Another useful sort of relation is known as a partial ordering.

Definition 3.4.1.

A partial ordering on a set X is a relation on X that satisfies the following properties.

i.

(reflexivity) For all x X, we have x x.

ii.

(antisymmetry) If x,y X satisfy x y and y x, then x = y.

iii.

(transitivity) If x,y,z X satisfy x y and y z, then x z.

A set X together with a partial ordering is referred to as a partially ordered set.

Definition 3.4.2.

A total ordering on a set X is a partial ordering such that for all x,y X, one has either x y or y x. In this case, X together with is called a totally ordered set.

Examples 3.4.3.

a.

The relation on is a total ordering, as is .

b.

The relation < on is not a partial ordering, as it is not reflexive.

c.

The relation on the set of subsets 𝒫X of any set X, which is known as the power set of X, is a partial ordering. It is not a total ordering if X contains more than one element.

d.

The relation = is a partial ordering on any set.

Given a partial ordering on a set X, we can speak of minimal and maximal elements of X.

Definition 3.4.4.

Let X be a set with a partial ordering .

a.

A minimal element in X (under ) is an element x X such that if z X and z x, then z = x.

b.

A maximal element y X is an element such that if z X and y z, then z = y.

Minimal and maximal elements need not exist, and when they exist, they need not be unique. Here are some examples.

Examples 3.4.5.

a.

The set has no minimal or maximal elements under .

b.

The interval [0,1) in has the minimal element 0 but no maximal element under .

c.

The power set 𝒫X of X has the minimal element and maximal element X under .

d.

Under = on X, every element is both minimal and maximal.

e.

Consider the set S 𝒫X of nonempty subsets of a set X, with the partial ordering . The minimal elements of S are exactly the singleton sets in X.

One can ask for a condition under which maximal (or minimal) elements exist. To phrase such a condition, we need two more notions.

Definition 3.4.6.

Let X be a set with a partial ordering . A chain in X is a subset of X that is totally ordered under .

Definition 3.4.7.

Let X be a set with a partial ordering . Let A be a subset of X. An upper bound on A under is an element x X such that a x for all a A.

Examples 3.4.8.

a.

The subset [0,1) of has an upper bound 1 under . In fact, any element x 1 is an upper bound for [0,1). The subset [0,1] has the same upper bounds.

b.

The subset of has no upper bound under .

We now come to Zorn’s lemma, which is equivalent to the axiom of choice. We omit the proof of this fact.

Theorem 3.4.9 (Zorn’s lemma).

Let X be a nonempty set with a partial ordering , and suppose that every chain in X has an upper bound. Then X contains a maximal element.

We use Zorn’s lemma to prove the following.

Theorem 3.4.10 (Alexander subbase theorem).

A space X is compact if and only if there exists a subbase 𝒮 for its topology such that every open cover of X by elements of 𝒮 has a finite subcover.

Proof.

Suppose that X is not compact, and let 𝒮 be a subbase of X. We need to show that 𝒮 contains an open cover 𝒰 that does not have a finite subcover.

Let Q be the set of all open covers of X that have no finite subcover, and note that Q by the noncompactness of X. Let 𝒞 be a chain in Q, and let 𝒲 = 𝒱𝒞𝒱 be the union of all covers in 𝒞. Any finite collection of elements of 𝒲, being each contained in some element of 𝒞, are all contained in the largest such element 𝒱 under inclusion. Being that 𝒱 has no finite subcover, such a finite collection cannot cover X, so 𝒲 has no finite subcover, which is to say that 𝒲 Q. Thus, every chain in Q has an upper bound, and so by Zorn’s lemma, Q has a maximal element M.

Now consider the subset 𝒰 = 𝒮M of 𝒮. Being a subset of M, no finite subset of 𝒰 covers M. We claim that 𝒰 covers X, which will finish the proof. Let x X, let U M with x U. Let V1,,Vn 𝒮 be such that x V1 Vn U, which exist as 𝒮 is a subbase. If ViM for all 1 i n, we can find by the maximality of M finite subsets 𝒩i of M such that the collections 𝒩i{Vi} cover X. Then 𝒩 = i=1n𝒩i and V1 Vn together cover X, so 𝒩{U}M covers X as well, contradicting the fact that M has no finite subcover. Thus, there exists i such that Vi 𝒰, and Vi contains x by definition. Thus 𝒰 is a cover of X.

Theorem 3.4.11 (Tychonoff’s theorem).

Any product of compact spaces is compact under the product topology.

Proof.

Let {Xii I} be a collection of compact spaces, and let X = iIXi. For each i I, let 𝒯 i be the topology on Xi, and let

𝒮 = iI{πi1(U i)Ui 𝒯 i},

where πi: X Xi is the ith projection map. Then 𝒮 is a subbase for the topology on X. We claim that every open cover of X by elements in 𝒮 has a finite subcover, which by the Alexander subbase theorem will finish the proof.

Let 𝒰 𝒮 be an open cover of X. Then

𝒰 = iI{πi1(U i)Ui 𝒰i}

for some collections 𝒰i of open sets in Xi for each i I. If no 𝒰i covers Xi, then for each i I, we may by the axiom of choice find xi Xi such that xi Ui𝒰iUi. Set x = (xi)iI. Then xU for any U 𝒰, for any such U has the form U = πi1(Ui) for some Ui 𝒰i for some i I, and xi = πi(x)Ui. Thus there exists i I such that 𝒰i covers Xi. It has a finite subcover 𝒱, and the finite set {πi1(V )V 𝒱} then covers X.

3.5. Local connectedness and compactness

Local properties of a space are those which happen within a small enough neighborhood of a point. Here are the definitions of local connectedness and local compactness.

Definition 3.5.1.

A topological space X is called locally connected at x X if every neighborhood of x contains a connected open neighborhood of x. A topological space is called locally connected if it is locally connected at each of its points.

Proposition 3.5.2.

A topological space X is locally connected if and only if every connected component of every open set in X is open in X.

Proof.

Suppose X is locally connected. Let U be open in X, and let A be a connected component of U. If x A, then there is a connected open neighborhood V of x that is contained in U, and V is contained in A by its connectedness. Since x was arbitrary, A is open.

Conversely, suppose that all connected components of open sets in X are open in X. Let x X, and let V be an open neighborhood of x. Let A be the connected component of x in V. Note that A is open by assumption, so it is a neighborhood of x. Thus, X is locally connected.

Example 3.5.3.

Let X = n=1{0,1} with the product topology, as in Example 3.1.18. Its basic open sets have the form n=1NUi×m=N+1{0,1}, which are in particular infinite. As its connected components of X are singletons, X is not locally connected.

We mention in passing that we have a similar notion of local path connectedness.

Definition 3.5.4.

A topological space X is called locally path connected at x X if every neighborhood of x contains a path connected open neighborhood of x. A topological space is called locally path connected if it is locally path connected at each of its points.

Of course, locally path connected spaces are locally connected, but the converse does not hold in general.

Definition 3.5.5.

A topological space X is said to be locally compact at x X if x has a compact neighborhood. A topological space is locally compact if it is locally compact at all of its points.

Compact spaces are of course locally compact, as are many other spaces.

Example 3.5.6.

The space n is locally compact, since every closed ball of positive radius about a point is compact. However, n=1 is not locally compact, since its basic open sets all have closure that is a product of finitely many closed intervals with infinitely many copies of , and these are not compact as is not.

Definition 3.5.7.

A compactification of a topological space X is a pair (Y,ι) consisting of a compact topological space Y and an embedding ι of X in Y.

Remark 3.5.8.

We may view X as a subspace of a compactification Y via identifying it with its homeomorphic image. Two compactifications Y and Z of a space X are then said to be equivalent if there exists a homeomorphism f : Y Z that restricts to the identity map on X. This gives an equivalence relation on the compactifications of X.

Topological spaces can be compactified by adding in a single point.

Theorem 3.5.9.

Let X be a topological space. The set Y containing X and one additional element called has a topology consisting of the open sets in X and the complements Y A of closed, compact subsets A of X. Moreover, Y is a compact space under this topology, and any other compactification Z of X with ZX a singleton set is equivalent to Y.

Proof.

Note that arbitrary intersections of closed, compact subsets A of X are closed, and then compact, and the union of an open subset U of X and the complement Y A of a closed, compact subset A of X is the complement in Y of a smaller closed, and then compact, subset of X. It follows that the collection 𝒯 of open sets defined in the theorem is closed under arbitrary unions. Similarly, finite unions of closed, compact subsets of X are closed and compact, and the intersection of U and Y A as above is an open subset of X. Hence, 𝒯 is closed under arbitrary intersections, so 𝒯 is a topology.

Given an open cover 𝒱 of Y, we may choose an element in it of the form Y A with A closed, compact in X, as such a set is needed to cover . Then 𝒱{Y A} is an open cover of A, which has a finite subcover as A is compact, so 𝒱 has a finite subcover as well. Thus, Y is compact.

Finally, for any other space Y as in the theorem, with its additional point, is in canonical bijection with Y via the map f : Y Y such that f(x) = x for x X and f() = . If V is open in Y and Y, then f(V ) = V is open in X, hence in Y. If V is an open neighborhood of in Y, then its complement A is closed, hence compact in Y, and of course also contained in X. Then f(V ) = Yf(A) is the complement of a compact, closed subset of X, so is open in Y as well. Thus, f is continuous, and then f1 is continuous too, as we did not distinguish Y and Y.

Definition 3.5.10.

For a space X, the compact space of Theorem 3.5.9 is called the one-point compactification of X.

Proposition 3.5.11.

A topological space X is locally compact and Hausdorff if and only if its one-point compactification is Hausdorff.

Proof.

Let Y be the one-point compactification of X. Suppose that X is locally compact and Hausdorff. To show that Y is Hausdorff, it suffices to consider some x X and . Since X is locally compact, we can find a compact neighborhood A of x, which then contains some open neighborhood U. Then U is disjoint from the open neighborhood Y A of .

If X is not Hausdorff, then there exist points u,v X not contained in disjoint open sets in X. No two complements of compact sets in X are disjoint, since they contain the added point . Moreover, if uA and v V for a compact subset A of X and an open set V in X, then the open complement X A contains u and is not disjoint from V, so Y A is not disjoint from v either. Thus, u and v are not contained in disjoint open subsets of Y, and Y is not Hausdorff.

If X is not locally compact, then it is not locally compact at . Given an open neighborhood Y A of , where A is closed in X and compact, and an open neighborhood U of x in X, we cannot have U A, since if this were the case, then U¯ A would be a compact neighborhood of x. Thus, Y is again not Hausdorff.

Remark 3.5.12.

The one-point compactification of a locally compact Hausdorff space X has the property of being a quotient space of each Hausdorff compactification of X via the unique surjection that is the identity on X.

Example 3.5.13.

The one-point compactification of n is homeomorphic to Sn. To see this, view Sn as the unit sphere centered at = (0,,0,1) n+1. This is realized via the embedding n Sn sending x = (x1,,xn) to the unique point in Sn{} on the line between (x1,,xn,0) and .

Our definition of local compactness differs from our definition of local connectedness, as we only ask for a compact neighborhood, not one contained in an arbitrarily small neighborhood. For Hausdorff spaces, these notions are the same.

Proposition 3.5.14.

A Hausdorff space X is locally compact if and only if every open neighborhood U of a point x X contains a compact neighborhood A with A U.

Proof.

As the other direction is immediate, we may suppose that X is locally compact. Let U be open in X, and let x U. Let Y be the one-point compactification of X, which is compact Hausdorff since X is locally compact Hausdorff. Let A = Y U, which is closed in Y, hence compact. By Lemma 3.2.7, we may find an open neighborhood V of x and an open set W containing A that are disjoint. The closure of V in Y is compact and disjoint from A, so is contained in U.

Corollary 3.5.15.

Any open or closed subspace of a locally compact Hausdorff space is locally compact.

Proof.

Let X be locally compact Hausdorff. If A is closed in X and a A, then there is a compact neighborhood C of a in X, and CA is then a compact neighborhood of a in A. If U is open in X and x U, then by Proposition 3.5.14, we can find a compact neighborhood B of x in U.

Corollary 3.5.16.

A Hausdorff space X is locally compact if and only if it is homeomorphic to an open subspace of a compact Hausdorff space.

Proof.

If X is locally compact but not compact, take its one-point compactification, in which X is open. If X is open in a compact Hausdorff space, then it is locally compact by Corollary 3.5.15.

Definition 3.5.17.

A map f : X Y of topological spaces is said to be a local homeomorphism if it is continuous, open, and for each x X, there exists an open neighborhood U of x such that the restriction of f to U is a homeomorphism onto its image.

Example 3.5.18.

The inclusion map of an open set in a topological space is a local homeomorphism.

Example 3.5.19.

The map f : S1 given by f(x) = (cosx,sinx) is a local homeomorphism, as is its restriction to any open interval. Note that there is, however, no local homeomorphism g: S1 .

Find in the notes