All Posts
Mathematics

The Heine-Borel Theorem: Why Compactness Matters in Analysis

We prove the Heine-Borel theorem — that a subset of Euclidean space is compact if and only if it is closed and bounded — and explore why compactness is one of the most powerful concepts in all of analysis.

The Theorem

Heine-Borel Theorem

A subset KRnK \subseteq \mathbb{R}^n is compact if and only if it is closed and bounded.

Here compact means that every open cover of KK has a finite subcover: if KαIUαK \subseteq \bigcup_{\alpha \in I} U_\alpha with each UαU_\alpha open, then there exist finitely many indices α1,,αN\alpha_1, \ldots, \alpha_N such that KUα1UαNK \subseteq U_{\alpha_1} \cup \cdots \cup U_{\alpha_N}.


Why Compactness Matters

Compactness is the topological generalization of "finiteness." It allows us to reduce infinite problems to finite ones. Here is a small sample of theorems that require compactness:

  • Extreme Value Theorem: A continuous function on a compact set attains its maximum and minimum.
  • Uniform Continuity: A continuous function on a compact set is uniformly continuous.
  • Bolzano-Weierstrass: Every sequence in a compact set has a convergent subsequence.
  • Arzelà-Ascoli: Characterizes compact subsets of C(X)C(X).
  • Tychonoff's Theorem: Products of compact spaces are compact.
  • Brouwer Fixed-Point Theorem: Every continuous self-map of a compact convex set has a fixed point.

Without compactness, analysis would lose most of its powerful existence theorems.


Definitions

Open Cover

An open cover of KK is a collection {Uα}αI\{U_\alpha\}_{\alpha \in I} of open sets whose union contains KK:

KαIUαK \subseteq \bigcup_{\alpha \in I} U_\alpha

Finite Subcover

A finite subcover is a finite subcollection that still covers KK:

KUα1Uα2UαNK \subseteq U_{\alpha_1} \cup U_{\alpha_2} \cup \cdots \cup U_{\alpha_N}

Compact

KK is compact if every open cover has a finite subcover.


Examples and Non-Examples

Compact

  • [0,1][0, 1] — closed and bounded
  • {1/n:n1}{0}\{1/n : n \geq 1\} \cup \{0\} — closed and bounded
  • Sn1={xRn:x=1}S^{n-1} = \{x \in \mathbb{R}^n : |x| = 1\}
  • Any finite set

Not Compact

  • (0,1)(0, 1) — bounded but not closed
  • [0,)[0, \infty) — closed but not bounded
  • R\mathbb{R} — neither
  • Q[0,1]\mathbb{Q} \cap [0,1] — bounded but not closed

Proof: Compact \Rightarrow Closed and Bounded

Proof that compact implies bounded.

Cover KK with the open balls B(0,n)B(0, n) for n=1,2,3,n = 1, 2, 3, \ldots:

Kn=1B(0,n)=RnK \subseteq \bigcup_{n=1}^{\infty} B(0, n) = \mathbb{R}^n

By compactness, finitely many suffice: KB(0,n1)B(0,nN)=B(0,maxni)K \subseteq B(0, n_1) \cup \cdots \cup B(0, n_N) = B(0, \max n_i).

Therefore KK is bounded. \square


Proof that compact implies closed.

We show RnK\mathbb{R}^n \setminus K is open. Let xKx \notin K. For each yKy \in K, choose disjoint open sets UyyU_y \ni y and VyxV_y \ni x (possible since Rn\mathbb{R}^n is Hausdorff).

The sets {Uy}yK\{U_y\}_{y \in K} cover KK, so by compactness there exist y1,,yNy_1, \ldots, y_N with KUy1UyNK \subseteq U_{y_1} \cup \cdots \cup U_{y_N}.

Let V=Vy1VyNV = V_{y_1} \cap \cdots \cap V_{y_N}. Then VV is open, xVx \in V, and VK=V \cap K = \emptyset (since VUyj=V \cap U_{y_j} = \emptyset for each jj).

Therefore xx is an interior point of RnK\mathbb{R}^n \setminus K, so KK is closed. \square


Proof: Closed and Bounded \Rightarrow Compact

This is the deeper direction. We prove it for R\mathbb{R} first, then extend.

Proof for [a,b]R[a, b] \subset \mathbb{R} (by bisection)

Let {Uα}\{U_\alpha\} be an open cover of [a,b][a, b]. Suppose for contradiction that no finite subcover exists.

Step 1 — Bisect. Divide [a,b][a, b] into [a,m][a, m] and [m,b][m, b] where m=(a+b)/2m = (a+b)/2. At least one half has no finite subcover. Call it [a1,b1][a_1, b_1].

Step 2 — Iterate. Repeat: bisect [a1,b1][a_1, b_1] and choose a half with no finite subcover. This gives a nested sequence:

[a,b][a1,b1][a2,b2][a, b] \supset [a_1, b_1] \supset [a_2, b_2] \supset \cdots

with bnan=(ba)/2nb_n - a_n = (b-a)/2^n and no [an,bn][a_n, b_n] finitely coverable.

Step 3 — Nested intervals. By the nested intervals property, there exists a unique point:

cn=0[an,bn]c \in \bigcap_{n=0}^{\infty} [a_n, b_n]

Step 4 — Contradiction. Since c[a,b]c \in [a, b], there exists UαU_\alpha containing cc. Since UαU_\alpha is open, there exists ε>0\varepsilon > 0 with (cε,c+ε)Uα(c - \varepsilon, c + \varepsilon) \subset U_\alpha.

For large enough nn, bnan<εb_n - a_n < \varepsilon, so [an,bn](cε,c+ε)Uα[a_n, b_n] \subset (c-\varepsilon, c+\varepsilon) \subset U_\alpha.

But then {Uα}\{U_\alpha\} is a finite subcover of [an,bn][a_n, b_n] — contradicting our assumption. \square

Extension to Rn\mathbb{R}^n

For Rn\mathbb{R}^n, any closed and bounded set is contained in a box [a1,b1]××[an,bn][a_1, b_1] \times \cdots \times [a_n, b_n]. The proof extends by bisecting all nn coordinates simultaneously (dividing into 2n2^n sub-boxes at each step) and applying the same nested argument.

Alternatively, one proves that [a,b]n[a, b]^n is compact (using Tychonoff's theorem or repeated bisection), and then uses the fact that a closed subset of a compact set is compact.


Equivalent Formulations

In Rn\mathbb{R}^n, the following are equivalent for a set KK:

  1. KK is compact (every open cover has a finite subcover).
  2. KK is closed and bounded (Heine-Borel).
  3. KK is sequentially compact (every sequence has a convergent subsequence in KK).
  4. KK is limit point compact (every infinite subset has a limit point in KK).
  5. KK is complete and totally bounded (as a metric space).

In general metric spaces, (1), (3), and (5) are equivalent, but (2) is not — Heine-Borel is special to Rn\mathbb{R}^n.


Failure in Infinite Dimensions

In infinite-dimensional spaces, the Heine-Borel theorem fails. The closed unit ball in 2\ell^2:

B={x2:x1}\overline{B} = \{x \in \ell^2 : \|x\| \leq 1\}

is closed and bounded but not compact. The sequence e1,e2,e3,e_1, e_2, e_3, \ldots (standard basis vectors) has no convergent subsequence since enem=2\|e_n - e_m\| = \sqrt{2} for nmn \neq m.

This failure is one of the main challenges in functional analysis and PDEs, where compactness arguments require more sophisticated tools (weak compactness, Rellich-Kondrachov, etc.).


Applications

Extreme Value Theorem

Theorem. If f:KRf: K \to \mathbb{R} is continuous and KK is compact, then ff attains its maximum and minimum.

Proof sketch. The image f(K)f(K) is compact in R\mathbb{R} (continuous image of compact is compact), hence closed and bounded by Heine-Borel. A bounded subset of R\mathbb{R} has a supremum, and closedness ensures the supremum belongs to f(K)f(K). \square

Uniform Continuity

Theorem. If f:KRmf: K \to \mathbb{R}^m is continuous and KK is compact, then ff is uniformly continuous.

Proof sketch. For each xKx \in K and ε>0\varepsilon > 0, choose δx\delta_x from pointwise continuity. The balls B(x,δx/2)B(x, \delta_x/2) cover KK; extract a finite subcover; and take δ=minδxi/2\delta = \min \delta_{x_i}/2. \square


Historical Note

Eduard Heine proved that continuous functions on closed bounded intervals are uniformly continuous (1872). Émile Borel proved the finite subcover property of [0,1][0,1] in 1895. Henri Lebesgue generalized this to Rn\mathbb{R}^n in 1904. The combined "Heine-Borel theorem" bears their names, though the synthesis into a single result came gradually.


Summary

KRn is compactK is closed and boundedEvery sequence in K has a convergent subsequence in K (applications)Extreme values exist, uniform continuity, fixed points, \begin{aligned} &K \subseteq \mathbb{R}^n \text{ is compact} \\[6pt] &\Updownarrow \\[6pt] &K \text{ is closed and bounded} \\[6pt] &\Updownarrow \\[6pt] &\text{Every sequence in } K \text{ has a convergent subsequence in } K \\[6pt] &\Downarrow \text{ (applications)} \\[6pt] &\text{Extreme values exist, uniform continuity, fixed points, \ldots} \end{aligned}

References