the Arzelà-Ascoli theorem
/ ar-dze-LAH a-SKO-lee /
When is a collection of functions compact — that is, when can you always pull a uniformly convergent subsequence out of any sequence in it? For ordinary numbers, Bolzano-Weierstrass answers this with one word: boundedness. For functions the answer needs two words, because a function can stay bounded yet wiggle ever faster and refuse to settle down. The Arzela-Ascoli theorem identifies the exact two conditions that rescue compactness: the functions must be bounded, and they must not be able to change too fast in a coordinated way (equicontinuity).
Precisely: a family of continuous functions on a compact set K is relatively compact in the uniform (sup) norm — every sequence has a uniformly convergent subsequence — if and only if it is uniformly bounded and equicontinuous. Uniformly bounded means a single number M caps every |f| on K. Equicontinuous means a single modulus of continuity works for the whole family: for every error epsilon there is a single distance delta such that |x - y| < delta forces |f(x) - f(y)| < epsilon for every f in the family at once. The proof builds the subsequence by a diagonal argument over a countable dense set of points, then uses equicontinuity to upgrade pointwise convergence on those points into uniform convergence everywhere on K.
In complex analysis this is the abstract heart of Montel's theorem. The holomorphic setting supplies the two hypotheses almost for free: local boundedness is assumed, and the Cauchy estimates turn it into a derivative bound, which IS equicontinuity. So 'bounded holomorphic family -> normal' is really 'Arzela-Ascoli with the equicontinuity supplied by Cauchy'. Honest caveat: equicontinuity is genuinely needed and is the subtle half. The sequence f_n(x) = x^n on [0,1] is uniformly bounded but NOT equicontinuous, and indeed it has no uniformly convergent subsequence (its pointwise limit is discontinuous). Boundedness alone never suffices for function compactness.
On [0,1], the sequence f_n(x) = x^n is uniformly bounded (|f_n| <= 1) but not equicontinuous: near x = 1 the functions steepen without limit. Arzela-Ascoli therefore does not apply, and indeed no subsequence converges uniformly — the pointwise limit jumps from 0 to 1 at x = 1. Adding a uniform bound on f_n' (say |f_n'| <= 3) would restore equicontinuity and hence compactness.
Bounded but not equicontinuous: x^n shows why boundedness alone fails for function compactness.
The two hypotheses are independent: equicontinuity controls the rate of change, boundedness controls the size. Both are essential; drop either and the conclusion can fail. In the holomorphic case the Cauchy estimates supply equicontinuity automatically.