Book 7A

© 2026 NeckPinch · www.neckpinch.com · All rights reserved.

Course 7Book 7A: Topology and the Fundamental GroupChapter 2

Compactness and Compactification

Hausdorff spaces, Tychonoff, and the sphere as ℝⁿ plus a point.

18 min read · Updated Oct 3, 2026

Read with Lee, Introduction to Topological Manifolds, chapter 4 (connectedness and compactness), concentrating on what is new for general spaces: the closed map lemma, compactness in Hausdorff spaces, local compactness, proper maps and the one-point compactification. The metric-space material repeats [[2B.3]].

In this chapter · 7 sections
  1. 2.1Mapping the poles
  2. 2.2Compactness in topological spaces
  3. 2.3Products and Tychonoff's theorem
  4. 2.4Local compactness and proper maps
  5. 2.5The one-point compactification
  6. 2.6History
  7. 2.7Exercises

Compactness, for metric spaces, was the subject of 2B.3 Compactness: every sequence has a convergent subsequence, every open cover has a finite subcover, and the two are equivalent there. In general topological spaces the open-cover definition is the right one, and it combines with the Hausdorff condition to give the single most useful lemma for working with quotients: a continuous bijection from a compact space to a Hausdorff space is a homeomorphism. With it, identifying a glued-together space with a familiar one becomes a matter of writing down a continuous map.

The chapter's second theme is adding a point at infinity. The plane plus one point is a sphere; three-dimensional space plus one point is the 3-sphere S3S^3, the space the Poincaré conjecture is about. This is the one-point compactification, and stereographic projection makes it concrete.

By the end of this chapter you will be able to:

  • use the open-cover definition of compactness, and prove that compact subsets of Hausdorff spaces are closed;
  • prove and use the closed map lemma for continuous maps from compact to Hausdorff spaces;
  • state Tychonoff's theorem and say where the axiom of choice enters;
  • define local compactness and proper maps;
  • construct the one-point compactification and prove Rn∪{∞}≅Sn\mathbb{R}^n\cup\{\infty\} \cong S^n by stereographic projection.

Mapping the poles

In the world In use Polar stereographic sea-ice maps

Maps of Arctic sea ice, such as those distributed by the US National Snow and Ice Data Center (NSIDC) from satellite measurements, use a polar stereographic projection: each point of the Earth is projected onto a plane tangent to (or cutting) the globe near the North Pole, along the line through the South Pole. The NSIDC's standard grid for the Arctic, "NSIDC Sea Ice Polar Stereographic North" (EPSG code 3413 in its current form), uses a plane that cuts the globe so that scale is exact at latitude 70°70° N, close to where most of the ice edge lies. The projection is conformal (5A.1 Holomorphic Functions Are Conformal), so the shapes of small features such as ice floes and leads are preserved, and the North Pole sits at the centre of the map.

Projected this way, the whole Earth except the South Pole fills the entire plane: the Equator is a circle, the Southern Hemisphere fills the outside of that circle, and the South Pole goes "to infinity". Run backwards, the projection says that the plane together with one point at infinity is a sphere. That single added point is what turns the open, unbounded plane into a compact space (Figure 2.1).

Compactness in topological spaces

Definition 2.1 Compactness

A topological space XX is compact if every open cover of XX has a finite subcover. A subset is compact if it is compact in the subspace topology.

Several facts from 2B.3 Compactness hold with the same proofs: a closed subset of a compact space is compact, and a continuous image of a compact space is compact, so a continuous real function on a compact space attains its maximum and minimum (2A.9 Continuous Functions). What changes is the relation with sequences: in general spaces sequential compactness and compactness are different (each can hold without the other), which is why the open-cover definition is taken as basic. For metric spaces, and so for subsets of manifolds, the two agree.

Hausdorff spaces add one more fact, which is false without the separation condition.

Proposition 2.2 Compact subsets of Hausdorff spaces are closed

If XX is Hausdorff and K⊆XK \subseteq X is compact, then KK is closed.

Proof. Let x∉Kx \notin K. For each y∈Ky \in K, choose disjoint open sets Uy∋xU_y \ni x and Vy∋yV_y \ni y. The VyV_y cover KK, so finitely many do, say Vy1,…,VymV_{y_1}, \dots, V_{y_m}. Then U=Uy1∩⋯∩UymU = U_{y_1}\cap\dots\cap U_{y_m} is an open neighbourhood of xx disjoint from Vy1∪⋯∪Vym⊇KV_{y_1}\cup\dots\cup V_{y_m} \supseteq K. So the complement of KK is open.

In the cofinite topology on an infinite set, every subset is compact (Exercise 2.7) but only finite sets and the whole space are closed: the proposition genuinely needs the Hausdorff condition.

Theorem 2.3 The closed map lemma

Let f:X→Yf : X \to Y be continuous, with XX compact and YY Hausdorff. Then ff is a closed map (it sends closed sets to closed sets). Consequently:

  1. if ff is a bijection, it is a homeomorphism;
  2. if ff is surjective, it is a quotient map: YY carries the quotient topology from XX, so Y≅X/∼Y \cong X/{\sim} where x∼x′x \sim x' iff f(x)=f(x′)f(x) = f(x');
  3. if ff is injective, it is an embedding (a homeomorphism onto its image).

Proof. A closed subset CC of XX is compact; its image f(C)f(C) is compact, hence closed in the Hausdorff space YY. For (1), the inverse of a closed bijection is continuous: (f−1)−1(C)=f(C)(f^{-1})^{-1}(C) = f(C) is closed for every closed CC. For (2), a set V⊆YV \subseteq Y with f−1(V)f^{-1}(V) open has complement Y∖V=f(X∖f−1(V))Y\setminus V = f(X\setminus f^{-1}(V)), which is closed, so VV is open. (3) is (1) applied to f:X→f(X)f : X \to f(X).

This lemma is the workhorse for quotients. To identify X/∼X/{\sim} with a known space YY, it is enough to find a continuous surjection f:X→Yf : X \to Y, with XX compact and YY Hausdorff, whose fibres are exactly the equivalence classes. For example [0,1]→S1[0, 1] \to S^1, t↦e2πitt \mapsto e^{2\pi it}, shows [0,1]/(0∼1)≅S1[0, 1]/(0 \sim 1) \cong S^1 (7A.1 Topological Spaces and Quotients); and the map that sends the square to the doughnut surface by the usual angle coordinates shows that the glued square is the doughnut (Exercise 2.9).

Products and Tychonoff's theorem

Theorem 2.4 Tychonoff's theorem

Any product of compact spaces, with the product topology, is compact.

For finitely many factors the proof is elementary, by the tube lemma: if XX is compact and WW an open set in X×YX \times Y containing X×{y}X\times\{y\}, then WW contains a whole "tube" X×VX\times V with VV a neighbourhood of yy (Exercise 2.8). From it, X×YX \times Y is compact when both factors are, and by induction any finite product. So closed bounded subsets of Rn\mathbb{R}^n, being closed subsets of products of intervals, are compact, which is the Heine–Borel theorem.

For infinitely many factors the theorem needs the axiom of choice, and is in fact equivalent to it (John Kelley, 1950), in the sense of 2A.8 Infinite Sets. It is used in functional analysis, for example in the proof of the Banach–Alaoglu theorem (4A.6 Weak Convergence and the Direct Method), but not for manifolds, which are locally products of finitely many lines.

Local compactness and proper maps

Rn\mathbb{R}^n is not compact, but every point has a compact neighbourhood (a closed ball). A Hausdorff space is locally compact if every point has a neighbourhood whose closure is compact. Every manifold is locally compact, since it is locally homeomorphic to Rn\mathbb{R}^n (7A.3 Manifolds and Surfaces). Infinite-dimensional normed spaces are not (4A.1 Banach Spaces and Bounded Operators).

A continuous map f:X→Yf : X \to Y is proper if the preimage of every compact set is compact. Proper maps are the maps that "send infinity to infinity": as a point leaves every compact set of XX, its image leaves every compact set of YY. A proper map into a locally compact Hausdorff space is closed. The polynomial map R→R\mathbb{R} \to \mathbb{R}, x↦x2x \mapsto x^2, is proper; x↦exx \mapsto e^x is not (the preimage of [0,1][0, 1] is (−∞,0](-\infty, 0]). Covering maps with finitely many sheets are proper (7A.6 Covering Spaces), and in Riemannian geometry a complete manifold is one whose distance functions are proper (9A.3 Geodesics and the Exponential Map).

The one-point compactification

Definition 2.5 One-point compactification

Let XX be a locally compact Hausdorff space that is not compact. Its one-point compactification is X∗=X∪{∞}X^* = X\cup\{\infty\}, where ∞\infty is a new point, with open sets:

  • the open subsets of XX, and
  • the sets (X∖K)∪{∞}(X\setminus K)\cup\{\infty\} with K⊆XK \subseteq X compact.

Neighbourhoods of ∞\infty are the complements of compact sets: a sequence converges to ∞\infty exactly when it eventually leaves every compact subset of XX. Then X∗X^* is a compact Hausdorff space containing XX as an open dense subspace (Exercise 2.10). Compactness: any open cover contains a set around ∞\infty, which misses only a compact KK, and finitely many other sets cover KK. Hausdorff: two points of XX are separated in XX, and a point x∈Xx \in X is separated from ∞\infty by a neighbourhood of xx with compact closure U‾\overline U, and (X∖U‾)∪{∞}(X\setminus\overline U)\cup\{\infty\}. This is where local compactness is used.

Theorem 2.6 Rn∪{∞}≅Sn\mathbb{R}^n\cup\{\infty\} \cong S^n

Stereographic projection from the north pole N=(0,…,0,1)N = (0, \dots, 0, 1) of the unit sphere Sn⊂Rn+1S^n \subset \mathbb{R}^{n+1},

σ(x1,…,xn+1)=(x1,…,xn)1−xn+1,\sigma(x_1, \dots, x_{n+1}) = \frac{(x_1, \dots, x_n)}{1 - x_{n+1}},

is a homeomorphism Sn∖{N}→RnS^n\setminus\{N\} \to \mathbb{R}^n, and extends, by σ(N)=∞\sigma(N) = \infty, to a homeomorphism Sn→(Rn)∗S^n \to (\mathbb{R}^n)^*.

Proof. The point σ(x)\sigma(x) is where the line from NN through xx meets the plane xn+1=0x_{n+1} = 0 (Figure 2.1). The inverse is

σ−1(y)=(2y,∣y∣2−1)∣y∣2+1,\sigma^{-1}(y) = \frac{(2y, |y|^2 - 1)}{|y|^2 + 1},

and both maps are continuous (Exercise 2.11). As x→Nx \to N on the sphere, ∣σ(x)∣→∞|\sigma(x)| \to \infty; precisely, ∣σ(x)∣2=1+xn+11−xn+1|\sigma(x)|^2 = \frac{1 + x_{n+1}}{1 - x_{n+1}}, so the neighbourhoods {xn+1>1−δ}\{x_{n+1} > 1 - \delta\} of NN correspond to the complements of the compact balls {∣y∣≤R}\{|y| \leq R\}. So the extension is a bijection that maps neighbourhoods of NN to neighbourhoods of ∞\infty; it is continuous, and since SnS^n is compact and (Rn)∗(\mathbb{R}^n)^* Hausdorff, it is a homeomorphism by the closed map lemma.

Figure 2.1. Stereographic projection in cross-section. The line from the north pole NN through a point xx of the sphere meets the equatorial plane at σ(x)\sigma(x). The southern hemisphere goes inside the unit circle, the northern hemisphere outside it, and points near NN go far out: NN itself corresponds to the one added point ∞\infty. (Sea-ice maps project from the South Pole, so the North Pole is at the centre.)
Where this goes Four pictures of the 3-sphere

The Poincaré conjecture is about S3S^3, and four descriptions of it are used interchangeably in this guide. (1) The unit sphere {x∈R4:∣x∣=1}\{x \in \mathbb{R}^4 : |x| = 1\}. (2) R3∪{∞}\mathbb{R}^3\cup\{\infty\}, by this chapter, which is how knots in R3\mathbb{R}^3 become knots in S3S^3 (7A.5 Computing π₁, 10A.1 A Zoo of Three-Manifolds). (3) Two solid balls glued along their boundary spheres, the two hemispheres, or, less obviously, two solid tori glued along their boundary tori, the simplest Heegaard splitting (Exercise 2.12, 10A.2 Building Three-Manifolds). (4) The group of unit quaternions, SU(2)SU(2), which double covers the rotation group (7A.6 Covering Spaces, 8A.5 Lie Groups and Group Actions). Each picture is suited to a different job, and moving between them is part of the fluency this book aims at.

History

The open-cover formulation of compactness was introduced by Pavel Alexandroff and Pavel Urysohn in the 1920s; Andrey Tychonoff proved his product theorem in 1930 (for products of intervals) and 1935 in general, and John Kelley showed in 1950 that it implies the axiom of choice. Alexandroff introduced the one-point compactification in 1924. Stereographic projection is ancient, used by Hipparchus and Ptolemy for star charts and astrolabes; its conformality was proved in the seventeenth century, by Thomas Harriot and later Edmond Halley.

Recall Where we stand

Compactness means every open cover has a finite subcover. Closed subsets of compact spaces and continuous images of compact spaces are compact; compact subsets of Hausdorff spaces are closed. A continuous map from a compact space to a Hausdorff space is closed, so a continuous bijection between them is a homeomorphism and a continuous surjection is a quotient map: the tool for identifying glued spaces. Finite products of compact spaces are compact by the tube lemma; Tychonoff's theorem for infinite products needs choice. Locally compact Hausdorff spaces have one-point compactifications, and stereographic projection shows Rn∪{∞}≅Sn\mathbb{R}^n\cup\{\infty\} \cong S^n. 7A.3 Manifolds and Surfaces defines manifolds and classifies the compact surfaces.

Exercises

Exercise 2.7 Compactness without Hausdorff

Show that every subset of an infinite set with the cofinite topology is compact. Deduce that Proposition 2.2 fails without the Hausdorff condition.

Solution

Given an open cover of a subset AA, pick one non-empty member UU; it misses only finitely many points of the whole space, and each of those in AA is covered by one more member. So finitely many members cover AA. Any infinite proper subset (such as the even numbers in N\mathbb{N}) is compact but not closed, since closed sets are finite or everything.

Exercise 2.8 The tube lemma

Let XX be compact, y∈Yy \in Y, and WW open in X×YX\times Y with X×{y}⊆WX\times\{y\} \subseteq W. (a) For each x∈Xx \in X choose a basic open set Ux×Vx⊆WU_x\times V_x \subseteq W containing (x,y)(x, y), and use compactness to find VV with X×V⊆WX\times V \subseteq W. (b) Deduce that X×YX\times Y is compact when XX and YY are.

Exercise 2.9 The square, the torus and the doughnut

Define F:[0,1]2→R3F : [0, 1]^2 \to \mathbb{R}^3 by

F(s,t)=((2+cos⁡2πt)cos⁡2πs, (2+cos⁡2πt)sin⁡2πs, sin⁡2πt).F(s, t) = \big((2 + \cos2\pi t)\cos2\pi s,\ (2 + \cos2\pi t)\sin2\pi s,\ \sin2\pi t\big).

Show that FF is continuous, that its image is the doughnut surface, and that F(s,t)=F(s′,t′)F(s, t) = F(s', t') exactly when the two points are identified in the torus gluing of 7A.1 Topological Spaces and Quotients. Conclude, by the closed map lemma, that the glued square is homeomorphic to the doughnut.

Exercise 2.10 Properties of the one-point compactification

(a) Check that the sets in Definition 2.5 form a topology. (b) Show that (0,1)∗≅S1(0, 1)^* \cong S^1 and that N∗\mathbb{N}^* (with N\mathbb{N} discrete) is homeomorphic to {1,12,13,… }∪{0}⊂R\{1, \frac12, \frac13, \dots\}\cup\{0\} \subset \mathbb{R}. (c) What is the one-point compactification of the open disc? Of the open annulus? (The annulus answer is not a manifold: what goes wrong at ∞\infty?)

Solution

(b) (0,1)≅S1∖{1}(0, 1) \cong S^1\setminus\{1\} via t↦e2πitt \mapsto e^{2\pi it}, and the extension sending ∞↦1\infty \mapsto 1 is a continuous bijection from a compact space to S1S^1. N∗\mathbb{N}^*: send n↦1nn \mapsto \frac1n and ∞↦0\infty \mapsto 0; neighbourhoods of ∞\infty are cofinite sets plus ∞\infty, which correspond to neighbourhoods of 00. (c) The open disc gives S2S^2. The open annulus has two "ends", the inner and outer boundaries, and both are squeezed to the single point ∞\infty: the result is a sphere with two points identified, and near ∞\infty it looks like two discs touching at their centres, not like a single disc.

Exercise 2.11 Stereographic formulas

(a) Verify that σ−1(y)=(2y,∣y∣2−1)∣y∣2+1\sigma^{-1}(y) = \frac{(2y, |y|^2 - 1)}{|y|^2 + 1} lies on the unit sphere and inverts σ\sigma. (b) Show ∣σ(x)∣2=1+xn+11−xn+1|\sigma(x)|^2 = \frac{1 + x_{n+1}}{1 - x_{n+1}}. (c) For n=2n = 2, show that the equator goes to the unit circle, and the southern hemisphere to the open unit disc.

Solution

(a) ∣2y∣2+(∣y∣2−1)2=(∣y∣2+1)2|2y|^2 + (|y|^2 - 1)^2 = (|y|^2 + 1)^2. With x=σ−1(y)x = \sigma^{-1}(y), 1−xn+1=2∣y∣2+11 - x_{n+1} = \frac{2}{|y|^2 + 1}, so σ(x)=2y/(∣y∣2+1)2/(∣y∣2+1)=y\sigma(x) = \frac{2y/(|y|^2 + 1)}{2/(|y|^2 + 1)} = y. (b) ∣σ(x)∣2=1−xn+12(1−xn+1)2=1+xn+11−xn+1|\sigma(x)|^2 = \frac{1 - x_{n+1}^2}{(1 - x_{n+1})^2} = \frac{1 + x_{n+1}}{1 - x_{n+1}}, using ∑i≤nxi2=1−xn+12\sum_{i\leq n}x_i^2 = 1 - x_{n+1}^2. (c) x3=0x_3 = 0 gives ∣σ∣=1|\sigma| = 1, and x3<0x_3 < 0 gives ∣σ∣<1|\sigma| < 1.

Exercise 2.12 Rehearsal: the 3-sphere as two solid tori

Let S3={(z,w)∈C2:∣z∣2+∣w∣2=1}S^3 = \{(z, w) \in \mathbb{C}^2 : |z|^2 + |w|^2 = 1\}, and A={∣z∣≤∣w∣}A = \{|z| \leq |w|\}, B={∣z∣≥∣w∣}B = \{|z| \geq |w|\}. (a) Show that A∩B={∣z∣=∣w∣=12}A\cap B = \{|z| = |w| = \frac{1}{\sqrt2}\} is a torus S1×S1S^1\times S^1. (b) Show that AA is homeomorphic to the solid torus D2×S1D^2\times S^1: check that (z,w)↦(2 z,w∣w∣)(z, w) \mapsto \big(\sqrt2\,z, \frac{w}{|w|}\big) maps AA bijectively and continuously onto {∣u∣≤1}×S1\{|u| \leq 1\}\times S^1, and use the closed map lemma. (c) Conclude that S3S^3 is two solid tori glued along their boundary tori, with the meridian circle of each glued to the longitude of the other. This is the genus-1 Heegaard splitting of S3S^3; changing the gluing gives lens spaces and S2×S1S^2\times S^1 (10A.2 Building Three-Manifolds).

Solution

(a) On A∩BA\cap B, ∣z∣=∣w∣|z| = |w| and ∣z∣2+∣w∣2=1|z|^2 + |w|^2 = 1 give ∣z∣=∣w∣=12|z| = |w| = \frac{1}{\sqrt2}, so the set is {12eiα}×{12eiβ}\{\frac{1}{\sqrt2}e^{i\alpha}\}\times\{\frac{1}{\sqrt2}e^{i\beta}\}. (b) On AA, ∣z∣2≤12|z|^2 \leq \frac12, so ∣2z∣≤1|\sqrt2z| \leq 1, and w≠0w \neq 0. The map is continuous on the compact set AA; it is injective because ∣w∣=1−∣z∣2|w| = \sqrt{1 - |z|^2} is determined by zz, so ww is recovered from w∣w∣\frac{w}{|w|} and zz; and it is onto. The closed map lemma makes it a homeomorphism. (c) Similarly for BB with the roles of zz and ww exchanged. On the common torus, the circle along which zz varies and ww is fixed bounds a disc in AA (a meridian of AA) but runs parallel to the core of BB (a longitude of BB), and the circle along which ww varies does the opposite.

© 2026 NeckPinch (www.neckpinch.com). All content in the guidebook (text, mathematics, figures and exercises) is protected by copyright. All rights reserved. No part may be copied, republished or redistributed without written permission.