Compactness: Explained with Examples, Concepts and Examples

A topological space is compact if every open cover has a finite subcover. An open cover is any collection of open sets whose union is the whole space, and a finite subcover means finitely many of those sets already cover everything. However you cover the space with infinitely many open patches, a finite selection always suffices.

Compactness matters because it turns infinite problems into finite ones. On a compact set, every continuous real-valued function attains its maximum and minimum, and every sequence has a convergent subsequence. The fact that a continuous function on a closed interval [a, b] is bounded and reaches its extremes is compactness at work.

Key Concepts

• Open cover: a family of open sets whose union equals the whole space X.
• Finite subcover: finitely many sets from the cover whose union is still X.
• Heine-Borel theorem: in R^n, a set is compact exactly when it is closed and bounded.
• Sequential compactness: every sequence has a convergent subsequence; in metric spaces this is equivalent to compactness.
• Closed subsets: a closed subset of a compact space is compact.
• Continuous image: the continuous image of a compact set is compact.

How It Works

To test compactness, take an arbitrary open cover and extract a finite subcover. Compactness survives the main constructions: continuous images of compact sets are compact, closed subsets of compact sets are compact, and finite products of compact spaces are compact. In metric spaces the test is concrete: compact means complete and totally bounded, which in R^n is just closed and bounded (Heine-Borel). The classic proof that [0, 1] is compact bisects the interval repeatedly: without a finite subcover, bisection would converge to a point missed by every covering set, which is impossible.

Worked Example

Show that (0, 1) is not compact using the open cover U_n = (1/n, 1) for n = 1, 2, 3, …. Every x in (0, 1) lies in some U_n: pick n larger than 1/x. Take any finite subcollection and let N be its largest index. The point x = 1/(2N) is in (0, 1), but for every n up to N we have 1/n at least 1/N, which exceeds 1/(2N), so x is in none of the chosen sets. No finite subfamily covers (0, 1), so it is not compact. By contrast [0, 1] is closed and bounded in R, hence compact by Heine-Borel.

Common Misconceptions

• "Compact means closed and bounded." This equivalence (Heine-Borel) holds in R^n, but in general topological spaces closed and bounded does not imply compact, and the words "bounded" may not even make sense.
• "Compact sets are always closed." True in Hausdorff spaces, but in a non-Hausdorff space a compact subset need not be closed.
• "Infinite sets cannot be compact." The interval [0, 1] contains infinitely many points and is compact; finiteness refers to the subcover, not to the set itself.

Real-World Applications

• Optimization: a continuous cost function on a compact set of parameters is guaranteed to attain a minimum, which is why training losses are minimized over compact (closed, bounded) parameter regions in machine learning.
• Economics: existence proofs for Nash equilibria apply fixed-point theorems to compact convex strategy sets.
• Control theory: reachable sets of controlled systems are often shown compact to guarantee that optimal controls exist.

Compactness means every open cover has a finite subcover, turning infinite problems into finite ones. Remember Heine-Borel in R^n, preservation under continuous maps and closed subsets, and that maxima exist and sequences cluster on compact sets.