Compactification

Why compactification matters

• Riemann surfaces and complex analysis. Adding a point at infinity to ℂ produces the Riemann sphere ℂℙ¹ ≅ S². Holomorphic functions on the Riemann sphere are constants; meromorphic functions are rational functions. Closed contour integrals can wrap ∞, residue theorem extends, Liouville's theorem follows from compactness. The complex plane was always being studied compactified.
• Projective varieties in algebraic geometry. Affine varieties V ⊂ ℂⁿ become projective varieties V̄ ⊂ ℂℙⁿ by adding their points at infinity. Bézout's theorem (two plane curves of degrees d, e meet in de points counted with multiplicity, including at infinity) only makes sense in ℂℙ². Compactness gives strong global theorems — every projective variety is compact, so every map to a Hausdorff space is closed.
• Dynamical systems and infinity. Phase spaces of differential equations often have orbits escaping to infinity. Compactifying the phase space — e.g., adding a sphere at infinity in ℝⁿ via Poincaré's compactification — makes "escape" a behaviour at a finite point on the boundary, allowing fixed-point analysis at infinity and global classification of orbits.
• Banach algebras and Gelfand duality. The Gelfand spectrum of a commutative C*-algebra A is a compact Hausdorff space; for the algebra C₀(X) of continuous functions vanishing at infinity, the spectrum is X⁺. So 1-point compactification is dual to adjoining a unit to a non-unital C*-algebra. Stone-Čech corresponds in the same way to bounded continuous functions C_b(X).
• Stone duality and ultrafilters. Stone-Čech βℕ has the structure of a left-topological semigroup under ultrafilter convolution; idempotent ultrafilters in βℕ underlie Hindman's theorem (any finite colouring of ℕ has a monochromatic IP-set). Compactification turns combinatorial Ramsey-type theorems into fixed-point statements about semigroup actions on compact spaces.
• Distributions and measures. The space of probability measures M(X) on a compact metric space X is itself compact (in the weak-* topology). Compactifying the underlying space lets one apply Prokhorov's theorem to extract weak limits of probability measures — central in probability theory and ergodic theory (Krylov-Bogolyubov).

Detailed constructions

One-point (Alexandroff) compactification

Hypothesis: X locally compact Hausdorff. Construction: set X⁺ = X ⊔ {∞}; open sets are (a) the open sets of X, and (b) {∞} ∪ (X ∖ K) for K ⊆ X compact. Then X⁺ is compact Hausdorff and X embeds as a dense open subset. Universal property: characterized as the smallest compactification — every other compactification of X surjects onto X⁺. Worked examples: ℝ⁺ ≅ S¹, (ℝⁿ)⁺ ≅ Sⁿ, ℂ⁺ ≅ S² (Riemann sphere), ℕ⁺ = ℕ ∪ {∞} (the convergent-sequence space), (open Möbius strip)⁺ ≅ ℝℙ².

Stone-Čech compactification

Hypothesis: X Tychonoff. Construction: embed X via the evaluation map e: X → [0,1]^{C(X, [0,1])}, x ↦ (f(x))_f; let βX be the closure of e(X). Universal property: every continuous f: X → K (K compact Hausdorff) extends uniquely to a continuous βf: βX → K. Equivalent description for X = ℕ: βℕ is the set of ultrafilters on ℕ with topology generated by clopen sets {U : A ∈ U} for A ⊆ ℕ. Cardinality |βℕ| = 2^c, vastly larger than ℕ. βℕ ∖ ℕ has rich algebraic and combinatorial structure.

Projective compactification

Embed ℝⁿ ↪ ℝℙⁿ via x ↦ [x : 1]. The complement is the hyperplane at infinity ℝℙ^(n-1), one point for each direction. ℝℙⁿ is compact, smooth (a manifold), and the embedding is dense. For ℂ, ℂℙ¹ ≅ ℂ ∪ {∞} ≅ S². For ℂ², ℂℙ² has a complex projective line at infinity ℂℙ¹ ≅ S². Algebraic geometry lives in projective space because intersection multiplicities behave better and Bézout's theorem holds.

Common misconceptions

• Always Hausdorff. The one-point compactification of ℚ is not Hausdorff because ℚ is not locally compact: any neighbourhood of ∞ is the complement of a compact subset of ℚ, hence is dense, hence meets every other open set. Locally compact Hausdorff is the necessary hypothesis for X⁺ to be Hausdorff.
• Stone-Čech is unique. It is — up to homeomorphism over X — by the universal property. But concretely, βX is enormous; for X = ℕ, βℕ has cardinality 2^c and is utterly intractable to describe set-theoretically. βX is unique but "uncomputable."
• Compactification preserves all properties. No. Connectedness can fail (the two-point compactification of ℝ is the closed interval, not S¹); metrisability often fails (βℕ is not metrisable; it has cardinality 2^c, exceeding any second-countable space). Smoothness, manifold structure, dimensions can all change. Compactification is a coarse one-way operation.
• One-point compactification is "the" compactification. Many compactifications exist: 1-point, projective, end compactification (one point per topological end), Stone-Čech (the maximum). They are different topological spaces. ℝ has 1-pt compactification S¹ but two-point end-compactification (interval [−∞, ∞]); the choice depends on what you want.
• Adding "infinity" is always possible. Stone-Čech requires Tychonoff; 1-point requires locally compact Hausdorff. A non-Tychonoff space (e.g., a quotient identifying all reals to one point with the indiscrete topology) has no Hausdorff compactification at all because continuous maps to compact Hausdorff are very limited.
• βℕ ∖ ℕ is countable. Far from it. βℕ ∖ ℕ has cardinality 2^c and is one of the most complicated compact Hausdorff spaces in ordinary mathematics. Walter Rudin (1956) proved βℕ ∖ ℕ has 2^c open sets, no isolated points, and a P-point structure dependent on set-theoretic axioms beyond ZFC.