Read compact/Hausdorff spaces, normality and metric compactness first. “Compactification” below includes density of the original space, so the noncompactness hypothesis will not be optional.
Motivation
A sequence can escape every compact region without having a finite limit. Adding one point should make all such escape directions converge to the same place. Its neighborhoods must therefore contain everything outside a compact set. This prescription determines the topology, but it does not automatically separate the new point from an old point.
Local compactness supplies exactly the separation that is missing: place the old point inside a neighborhood with compact closure, and give infinity the complement of that closure. Compactness of the extension is easier and comes from a different argument: one set covers infinity, leaving only a compact remainder to cover finitely.
Definitions
Definition 12.1.1
(Compact neighborhoods and local compactness)
A compact neighborhood of is a compact set containing an open set that contains ; itself need not be open. A Hausdorff space is locally compact here if every point has such a neighborhood. We will prove that it then has a basis of open sets with compact closure.
Definition 12.1.2
(Hausdorff and one-point compactifications)
A Hausdorff compactification of is a compact Hausdorff space containing a homeomorphic copy of as a dense subspace. It is a one-point compactification if the complement consists of one point, denoted . Copies are identified with when no confusion is possible.
Definition 12.1.3
(The space and proper maps)
For a Hausdorff , put . Declare the original open sets open, and take
as neighborhoods of infinity for compact . Compact sets are closed in , a fact needed in checking that this prescription is a topology. A continuous map is proper here when the inverse image of every compact set is compact.
Theorem/proposition
Theorem 12.1.4
(Local compactness gives compact-closure neighborhoods)
In a locally compact Hausdorff space, for every with open, there is open with
This is stronger than merely finding one compact neighborhood at each point.
Theorem 12.1.5
(Exact existence and uniqueness of the one-point compactification)
A space has a Hausdorff one-point compactification exactly when it is noncompact, locally compact and Hausdorff. Under these hypotheses the displayed topology on is that compactification, unique up to a homeomorphism fixing and infinity. If is already compact Hausdorff, the same prescription gives an isolated extra point and fails the required density.
Theorem 12.1.6
(Maps and scalar functions at infinity)
For noncompact locally compact Hausdorff spaces , a continuous extends continuously with exactly when it is proper. A continuous extends with exactly when
The latter property is called vanishing at infinity.
Concept lensStructural
Functions vanishing at infinity are continuous functions on the compactification
The scalar criterion identifies two function spaces. Extension by zero, with , is a bijection from the continuous functions vanishing at infinity onto , with restriction to as inverse; it preserves sums, products and the supremum norm. Statements about a noncompact space thereby become statements about a compact one: every function vanishing at infinity is bounded, and if it is not identically zero it attains its maximum absolute value at a point of .
Proof sketch or proof idea
Proof
(Separate a point from a compact remainder)
First prove the neighborhood statement. Choose a compact neighborhood of and open with . The compact set misses . In a Hausdorff space a point and a disjoint compact set have disjoint open neighborhoods: separate the point from each element, cover the compact set by finitely many of its chosen neighborhoods, then intersect the point-side neighborhoods and unite the others. Let contain and contain , disjointly. Put . Since is closed, ; since misses , . Thus . It is a closed subset of compact , hence compact. If is empty, choose , .
Proof
(Topology, compactness, separation and uniqueness)
To check the topology, observe
for original-open . Finite unions of compact sets are compact and compact sets are closed, so these sets and the original opens form a basis. An open set containing infinity has compact complement: it contains some , its complement is a closed subset of because the trace of every basic set on is open, and hence is compact. Conversely each compact complement is one of the declared neighborhoods. The topology induced on is exactly the original one, and is open in .
Every open cover of has a member containing infinity. Its complement is compact in , so finitely many other members cover after restriction to . Together they give a finite subcover. To prove Hausdorffness, separate old points by original open sets. To separate from infinity, use an open with compact closure and the disjoint open set . Here local compactness enters; it was not needed in the finite-subcover argument.
If is noncompact, no compact equals , so every infinity neighborhood meets . Consequently is dense. If is compact, is open, so density fails. This is why adjoining an isolated point to a compact space is an extension, not a compactification under our definition.
Conversely suppose is a compact Hausdorff compactification. Its subspace is Hausdorff and open, since the omitted singleton is closed. Compactness of would make it closed in , contradicting density and the extra point. For , regularity of compact Hausdorff gives an open neighborhood whose closure in lies inside . That closure is a compact neighborhood in , proving local compactness. Finally an infinity-open set has compact complement in because that complement is closed in compact . Conversely every compact subset of remains compact, hence closed, in Hausdorff . Its complement is infinity-open. Thus has exactly the constructed topology, proving uniqueness as well as necessity.
Proof X-ray
Compact remainder and local separation are distinct steps
The set covering infinity leaves a compact remainder, which proves compactness of the extension. Separating infinity from an old point additionally needs a compact neighborhood of that point. Density needs noncompactness. These three conclusions therefore have different hypotheses; none should be inferred just from the phrase “add one point.”
Proof
(Inverse images of neighborhoods of infinity)
For maps, a basic infinity neighborhood in has complement compact . Its inverse image under the prescribed extension is
It is open at infinity exactly when is compact. On original open sets, continuity is already that of , proving the proper-map equivalence in both directions. The value at infinity matters: a map might extend with a finite value without being proper.
For scalar functions, continuity at the new point with value zero says that every interval has an infinity-open inverse image. Its complement is exactly the displayed level set. This proves necessity. Conversely, compactness of each level set puts inside that inverse image, proving continuity at infinity; original-point continuity was assumed. A finite limiting value is handled by applying this criterion to . Every continuous compactly supported function vanishes at infinity, since each positive level set is a closed subset of its compact support; the converse about support need not hold.
Worked examples
Worked example 12.1.7
(Euclidean space closes to a sphere)
For , define
The identity puts the image on the unit sphere. It misses , and the inverse there is for a sphere point . Moreover . Thus outside a sufficiently large compact Euclidean ball, the image lies in any chosen neighborhood of . Extending gives a continuous bijection from compact to Hausdorff , hence a homeomorphism. In particular the line gains a circle and the plane gains a sphere; all directions of escape share the one new point.
Worked example 12.1.8
(An infinite discrete space and cofinite infinity neighborhoods)
Every point of a discrete space has a compact singleton neighborhood. Its compact subsets are exactly the finite sets, because the singleton cover of an infinite subset has no finite subcover. Thus for infinite discrete , infinity neighborhoods in are exactly complements of finite subsets. If is countably infinite, enumerate it as and map , . Singleton neighborhoods at old points and cofinite neighborhoods at zero prove a homeomorphism onto . The uncountable case is still compact Hausdorff, but cannot have a countable neighborhood base at infinity, as proved in the exercise on uncountable discrete spaces.
Worked example 12.1.9
(The rationals fail the local hypothesis)
The usual rational line is Hausdorff but not locally compact. If a compact neighborhood of existed, it would contain for some real . Choose an irrational and rational points in that interval converging to . Compact metric would give a subsequence converging to a rational point of , whereas every real subsequence still tends to , impossible. Therefore the rationals have no Hausdorff one-point compactification. The complement-of-compact prescription still gives a compact extension, but the necessary Hausdorff separation cannot hold.
Common mistakes
Common mistake
(Compact neighborhoods need not be open)
A compact neighborhood need not be open. In the Hausdorff setting we proved the stronger open compact-closure basis rather than silently using it as an alternative definition. Removing Hausdorffness would also remove the general fact that compact subsets are closed, which the construction uses.
Counterexample mode
Without local compactness the extra point cannot be separated
Claim: every noncompact Hausdorff space has a Hausdorff one-point compactification. Take . A neighborhood of in contains for some , and a neighborhood of contains for a compact . If the two were disjoint, would contain , hence its closed subset , which would then be compact. It is not: rationals in that interval converging in to an irrational point of it have no convergent subsequence in . So and cannot be separated in . By the uniqueness argument, any Hausdorff one-point compactification of would have to carry exactly this topology, so none exists. The repair is local compactness, which lacks for the same reason.
Common mistake
(Escaping to infinity means leaving every compact set)
Going to infinity means escaping every compact set, not merely being unbounded along some subsequence. Infinity-preserving extensions require properness; arbitrary continuous maps do not all extend in that way. Conversely a scalar function can vanish at infinity while having noncompact support.
Common mistake
(One point at infinity forgets direction)
Finally, a two-ended compactification of the line adds different left and right endpoints. Its topology differs from this one-point construction, where both directions approach the same point. The construction does not remember a preferred direction of escape.
Summary
Local compactness in a Hausdorff space yields neighborhoods with compact closure. Complements of compact sets then define one added point, producing a compact Hausdorff extension, dense exactly when the original space is noncompact. Every Hausdorff one-point compactification has this topology. Proper maps preserve infinity, while scalar vanishing at infinity is detected by compact positive level sets.
Exercises
Quick check
(Prove open and closed subspaces of a locally compact Hausdorff space are locally compact Hausdorff. If is closed and noncompact, prove its inclusion extends to a closed embedding when is noncompact.)
Solution · Answer
For a closed subspace , intersect a compact neighborhood with ; it is compact and contains the appropriate relative open neighborhood. For an open subspace , use the proved compact-closure basis with closure contained in . Hausdorffness is inherited in both cases. If is closed and noncompact, its inclusion is proper: is closed in any compact , hence compact in . The extension criterion gives a continuous injection . Its domain is compact and codomain Hausdorff, so it is a homeomorphism onto a closed image. No assertion that arbitrary subspaces are locally compact is being made.
Quick check
(Decide whether and , as maps , extend continuously with infinity mapped to infinity. Use compact inverse images.)
Solution · Answer
The square map is proper. If is compact real, its inverse image is closed and bounded, since forces when ; an empty inverse image is harmless. Heine–Borel gives compactness. The exponential map is not proper: the inverse image of compact is , not compact. Thus only the first has the required infinity-preserving extension.
Quick check
(Show vanishes at infinity on but has noncompact support. Show does not vanish at infinity.)
Solution · Answer
For , ; for it is empty. These are compact, so the function vanishes at infinity. It is nonzero everywhere, hence its support is all , noncompact. The set is closed and unbounded, so noncompact; the level-set criterion fails.
Quick check
(Prove a sequence in converges to infinity in exactly when it eventually leaves every compact subset. For , relate this to norms and distinguish it from unboundedness.)
Solution · Answer
The basic neighborhoods of infinity are precisely , so the assertion is the definition of sequence convergence applied to that basis. In Euclidean space compact sets are bounded, and every closed norm ball is compact; hence the condition is equivalent to . An unbounded sequence need not satisfy it: take on even indices and zero on odd indices in the line. It meets the compact singleton infinitely often.
Quick check
(Prove for uncountable discrete is not first countable at infinity, and deduce that this compact Hausdorff space is not metrizable.)
Solution · Answer
If there were a countable neighborhood base, its members would have complements finite. Their union is countable, so choose . Every base member contains , but the neighborhood would have to contain some base member, impossible. Every metric space is first countable, using radius- balls. Thus is not metrizable despite its proved compactness and Hausdorffness.