06.01.14 · riemann-surfaces / complex-analysis

Normal families and Montel's theorem

shipped3 tiersLean: none

Anchor (Master): Montel 1912 *Sur les suites infinies de fonctions* (Annales scientifiques de l'É.N.S.); Stieltjes 1894 *Recherches sur les fractions continues* (precursor compactness arguments); Vitali 1903 *Sopra le serie di funzioni analitiche* (convergence theorem); Arzelà 1895 *Sulle funzioni di linee* and Ascoli 1883 *Le curve limiti di una varietà data di curve* (originating equicontinuity compactness); Schottky 1904, Landau 1904 (Picard-type bounds underlying the fundamental normality test); Carathéodory-Landau 1911; Ahlfors *Complex Analysis* Ch. 5; Schiff *Normal Families* (Springer Universitext, 1993)

Intuition [Beginner]

A family of holomorphic functions is a collection of complex-valued functions on the same open set $Ω$ . You may have infinitely many of them — one for each parameter, one for each step of a construction, or one for each candidate in an extremal problem. The question Montel answers is when such a family is compact: when can you extract from any infinite sequence in the family a subsequence that converges to a holomorphic limit, with the convergence well-behaved on every closed bounded piece of $Ω$ ?

The answer is striking. You need only one bound: every compact piece $K$ of $Ω$ has a single constant $M_{K}$ such that $∣ f (z) ∣ \leq M_{K}$ for every function $f$ in the family and every point $z \in K$ . That alone, with no smoothness or regularity beyond holomorphicity, forces compactness. The proof uses the Cauchy integral formula to extract control of derivatives from control of values, and then a general topological lemma (Arzelà-Ascoli) to extract a convergent subsequence.

Compactness of families is the engine of the modern proof of the Riemann mapping theorem and of a wide network of existence results in complex analysis. The picture is: once you bound the values uniformly, the family becomes a compact metric space, and every continuous functional on it attains an extremum.

Visual [Beginner]

A schematic of the open set $Ω$ with a nested family of compact subsets $K_{1} \subset K_{2} \subset \dots$ exhausting $Ω$ . Above each compact, a stack of value plots for functions $f_{1}, f_{2}, \dots$ in the family. All plots lie inside a uniform bound on each $K_{j}$ . The picture indicates a selected subsequence converging uniformly on each $K_{j}$ .

Worked example [Beginner]

Consider the family $F$ of all holomorphic functions $f : D \to C$ on the open unit disc with $∣ f (z) ∣ \leq 1$ for every $z \in D$ . The bound $1$ is uniform across the whole disc and across the whole family. Pick any compact $K \subset D$ , say the closed disc of radius $r = 1/2$ centred at the origin. Every function in $F$ takes values inside the unit disc on $K$ , and the distance from $K$ to the boundary of $D$ is $1/2$ .

Now apply the Cauchy estimate for derivatives. For a holomorphic $f$ on $D$ with $∣ f ∣ \leq 1$ , the derivative at any $z_{0} \in K$ has size bounded by the value bound divided by the distance from $z_{0}$ to the boundary of $D$ : $∣ f^{'} (z_{0}) ∣ \leq 1/ (1/2) = 2$ . Every function in $F$ has its derivative bounded by $2$ on $K$ . Bounded derivatives mean equicontinuity: if $∣ z - w ∣ < ε /2$ with $z, w \in K$ , then $∣ f (z) - f (w) ∣ \leq 2∣ z - w ∣ < ε$ uniformly in $f \in F$ .

What this tells us: the boundedness of values together with the Cauchy derivative estimate forces uniform equicontinuity on each compact. By the Arzelà-Ascoli compactness lemma (every uniformly bounded and uniformly equicontinuous family of continuous functions on a compact set has a uniformly convergent subsequence), every sequence in $F$ has a subsequence converging uniformly on $K$ . A diagonal argument over a sequence of compacts exhausting $D$ then produces a single subsequence converging uniformly on every compact subset of $D$ . The family $F$ is normal.

Check your understanding [Beginner]

Exercise (easy, multiple choice).

A family $F$ of holomorphic functions on an open set $Ω$ is normal when which of the following holds?

A. Every function in $F$ is uniformly continuous on $Ω$ . B. Every infinite sequence drawn from $F$ has a subsequence converging uniformly on each compact subset of $Ω$ , with holomorphic limit (allowing the constant $\infty$ in the meromorphic generalisation). C. Every function in $F$ has the same Taylor series around the origin. D. The family $F$ is finite.

Hint

The defining feature is sequential compactness in the uniform-on-compacta topology, not pointwise or algebraic conditions on individual functions.

Answer

B. Feedback-correct: yes; a normal family is by definition a sequentially compact subset of $O (Ω)$ under the uniform-on-compacta topology. Feedback-wrong: A, C, D miss the compactness essence (uniform continuity of individuals does not control the family as a whole; Taylor-series agreement is far stronger; finiteness is sufficient but not the defining condition).

Formal definition [Intermediate+]

Let $Ω \subseteq C$ be a non-empty open set and let $O (Ω)$ denote the ring of holomorphic functions on $Ω$ , equipped with the topology of uniform convergence on compact subsets (also called the topology of compact convergence): a sequence $f_{n} \in O (Ω)$ converges to $f \in O (Ω)$ iff $sup_{z \in K} ∣ f_{n} (z) - f (z) ∣ \to 0$ for every compact $K \subset Ω$ . This topology is induced by the metric $d (f, g) = \sum_{k \geq 1} 2^{- k} min (1, sup_{K_{k}} ∣ f - g ∣)$ on any exhaustion $Ω = ⋃_{k} K_{k}$ by compacts with $K_{k} \subset int (K_{k + 1})$ .

Definition (normal family). A family $F \subseteq O (Ω)$ is normal when every sequence ${f_{n}} \subseteq F$ admits a subsequence ${f_{n_{k}}}$ that converges uniformly on every compact subset of $Ω$ to a holomorphic limit $f \in O (Ω)$ . Equivalently, the closure $\overline{F}$ in the compact-convergence topology is sequentially compact. ^{[Ahlfors Ch. 5 §5.1]}

The meromorphic generalisation allows the constant function $\infty$ as a limit: a family $F$ of meromorphic functions on $Ω$ is normal in the spherical sense when every sequence admits a subsequence converging uniformly on compact subsets either to a meromorphic limit or to the constant function $\infty$ , with convergence measured in the spherical metric on the Riemann sphere $C_{\infty} = P^{1} (C)$ .

Definition (locally bounded family). A family $F \subseteq O (Ω)$ is locally bounded when for every compact $K \subset Ω$ there exists $M_{K} < \infty$ with $∣ f (z) ∣ \leq M_{K}$ for every $f \in F$ and every $z \in K$ . Equivalently, every point of $Ω$ has a neighbourhood on which $F$ is uniformly bounded.

Definition (equicontinuity). A family $F$ of continuous complex-valued functions on $Ω$ is equicontinuous at $z_{0} \in Ω$ when for every $ε > 0$ there exists $δ > 0$ (depending on $z_{0}$ and $ε$ but not on $f \in F$ ) such that $∣ z - z_{0} ∣ < δ \Rightarrow ∣ f (z) - f (z_{0}) ∣ < ε$ for every $f \in F$ . The family is equicontinuous on a set $E \subseteq Ω$ when this holds at every $z_{0} \in E$ , uniformly equicontinuous on $E$ when $δ$ depends only on $ε$ (not on $z_{0}$ ).

Counterexamples to common slips

Local boundedness is not pointwise boundedness. The family ${f_{n} (z) := n 1_{∣ z ∣ < 1/ n}}$ would be pointwise bounded at every $z \neq = 0$ but is not locally bounded near $z = 0$ . (Such $f_{n}$ are not holomorphic, but the analogous holomorphic family $f_{n} (z) = n / (n z + 1)$ on $D$ illustrates the same gap.)
Normality is uniform-on-compacta, not pointwise. Pointwise-convergent sequences of holomorphic functions need not converge uniformly on compacta; the Vitali theorem below identifies the additional hypothesis (local boundedness plus pointwise convergence on a set with accumulation point) under which they do.
Normality does not require the family to be uniformly bounded on $Ω$ . On $C$ , the family ${f_{n} (z) := z^{n} / n!}$ is locally bounded — on each compact $∣ z ∣ \leq R$ one has $∣ f_{n} (z) ∣ \leq R^{n} / n! \to 0$ , so the family converges uniformly on compacts to the zero function and is normal — yet $sup_{z \in C} ∣ f_{n} (z) ∣ = \infty$ for each $n$ , so the family is not uniformly bounded on $C$ as a whole.

Key theorem with proof [Intermediate+]

Theorem (Montel, 1912). A family $F \subseteq O (Ω)$ of holomorphic functions on an open set $Ω \subseteq C$ is normal if and only if it is locally bounded. ^{[Montel 1912]}

The proof splits into the implications "normal $\Rightarrow$ locally bounded" (the easy direction, by sequential compactness) and "locally bounded $\Rightarrow$ normal" (the substantive direction, by Arzelà-Ascoli plus Cauchy estimate plus diagonal extraction).

Proof of "normal $\Rightarrow$ locally bounded". Suppose for contradiction that $F$ is normal but not locally bounded. Then there exists a compact $K \subset Ω$ and a sequence ${f_{n}} \subseteq F$ with $sup_{z \in K} ∣ f_{n} (z) ∣ \to \infty$ . Pick $z_{n} \in K$ realising the supremum (or close to it). By normality, extract a subsequence ${f_{n_{k}}}$ converging uniformly on $K$ to a holomorphic limit $f \in O (Ω)$ . Then $sup_{z \in K} ∣ f_{n_{k}} (z) ∣$ converges to $sup_{z \in K} ∣ f (z) ∣ < \infty$ (uniform convergence preserves $sup$ ). This contradicts $sup_{z \in K} ∣ f_{n_{k}} (z) ∣ \to \infty$ . So $F$ must be locally bounded.

Proof of "locally bounded $\Rightarrow$ normal". Assume $F$ is locally bounded. The strategy: derive equicontinuity from local boundedness via the Cauchy integral estimate on derivatives; then apply Arzelà-Ascoli to extract a uniformly convergent subsequence on each compact; then use a diagonal argument over an exhaustion of $Ω$ by compacts to produce a single subsequence converging uniformly on every compact.

Step 1 — derivative estimate from local boundedness. Fix a compact $K \subset Ω$ . Let $d := \frac{1}{2} min (1, dist (K, \partial Ω))$ if $Ω \neq = C$ , else $d := 1$ . The compact $K_{d} := {z \in C : dist (z, K) \leq d}$ lies in $Ω$ (for $Ω = C$ this is automatic; otherwise $dist (K_{d}, \partial Ω) \geq d$ ). By local boundedness there exists $M = M_{K_{d}}$ with $∣ f (z) ∣ \leq M$ for all $f \in F$ and all $z \in K_{d}$ .

For any $z_{0} \in K$ , the closed disc $\overline{B_{d} (z_{0})} \subseteq K_{d}$ , so $∣ f ∣ \leq M$ on the circle $∣ z - z_{0} ∣ = d$ . The Cauchy integral formula for the derivative (06.01.02) gives

f^{'} (z_{0}) = \frac{1}{2 π i} \oint_{∣ z - z_{0} ∣ = d} \frac{f ( z )}{( z - z _{0} ) ^{2}} d z .

Estimating the integral: $∣ f^{'} (z_{0}) ∣ \leq (2 π)^{- 1} \cdot 2 π d \cdot M / d^{2} = M / d$ . The bound is uniform in $f \in F$ and in $z_{0} \in K$ . Set $C_{K} := M / d < \infty$ .

Step 2 — equicontinuity from the derivative bound. For any $z, w \in K$ with $∣ z - w ∣ < d$ , the straight-line segment from $z$ to $w$ lies in $K_{d}$ (since both endpoints are in $K$ and the path stays within distance $d$ of $K$ ). By the fundamental theorem of calculus along this segment,

∣ f (z) - f (w) ∣ = \int_{0}^{1} f^{'} (w + t (z - w)) \cdot (z - w) d t \leq C_{K} ∣ z - w ∣.

Every $f \in F$ is $C_{K}$ -Lipschitz on $K$ . The family is uniformly equicontinuous on $K$ : for any $ε > 0$ , choosing $δ := min (d, ε / C_{K})$ gives $∣ z - w ∣ < δ \Rightarrow ∣ f (z) - f (w) ∣ < ε$ for every $f \in F$ .

Step 3 — Arzelà-Ascoli on each compact. Restricted to $K$ , the family $F$ is uniformly bounded (by $M$ ) and uniformly equicontinuous (by Step 2). The Arzelà-Ascoli theorem on the compact metric space $K$ (02.01.05 supplies the metric-space context for the classical statement: every uniformly bounded and equicontinuous family of continuous functions on a compact metric space is precompact in $C (K)$ with the supremum norm). Every sequence in $F ∣_{K}$ has a subsequence converging uniformly on $K$ .

Step 4 — diagonal extraction over an exhaustion of $Ω$ . Choose an exhaustion of $Ω$ by compacts: a sequence $K_{1} \subset K_{2} \subset K_{3} \subset \dots$ with $K_{n} \subset int (K_{n + 1})$ and $Ω = ⋃_{n} K_{n}$ . (Such an exhaustion exists for any open $Ω \subseteq C$ : take $K_{n} := {z \in Ω : ∣ z ∣ \leq n, dist (z, \partial Ω) \geq 1/ n}$ .) Given a sequence ${f_{n}} \subseteq F$ :

By Step 3 applied to $K_{1}$ , extract a subsequence ${f_{n}^{(1)}}$ converging uniformly on $K_{1}$ .
By Step 3 applied to $K_{2}$ (with ${f_{n}^{(1)}}$ as the starting sequence), extract a sub-subsequence ${f_{n}^{(2)}}$ converging uniformly on $K_{2}$ , hence also on $K_{1}$ .
Iterate: ${f_{n}^{(j + 1)}}$ is a subsequence of ${f_{n}^{(j)}}$ converging uniformly on $K_{j + 1}$ .

The diagonal subsequence $g_{n} := f_{n}^{(n)}$ is a subsequence of ${f_{n}}$ and, for every fixed $j$ , is eventually a subsequence of ${f_{n}^{(j)}}$ , hence converges uniformly on $K_{j}$ . The limit $g : Ω \to C$ is well-defined and continuous on each $K_{j}$ , and uniform limits of holomorphic functions on a compact disc are holomorphic on its interior (Weierstrass theorem on uniform limits, a direct consequence of the Cauchy integral formula passing to the limit). Hence $g \in O (Ω)$ and $g_{n} \to g$ uniformly on every compact subset of $Ω$ .

This completes the extraction of a uniformly-on-compacta convergent subsequence; $F$ is normal. $□$

Bridge. Montel's theorem builds toward the Riemann mapping theorem (06.01.06), where it appears again in its sharpest application: the existence of a biholomorphism $Ω \to D$ for any simply connected proper plane domain $Ω$ is established by selecting the extremal function on the family $F := {f : Ω \to D : f (z_{0}) = 0, f injective}$ maximising $∣ f^{'} (z_{0}) ∣$ . The family is locally bounded (by $1$ , uniformly on $Ω$ ), hence normal by Montel; a maximising sequence has a uniformly-on-compacta convergent subsequence by Montel; Hurwitz's theorem (06.01.13) preserves injectivity at the limit; the maximum principle plus Schwarz lemma (06.01.12) force the limit to be surjective onto $D$ . This is exactly the foundational reason normal families are central — the extremal-problem method requires sequential compactness, and Montel supplies the only natural compactness theorem in $O (Ω)$ .

The Montel-Hurwitz combination generalises the Bolzano-Weierstrass / Arzelà-Ascoli pattern from real analysis to the holomorphic setting: putting these together identifies the foundational principle as the bridge between bounded values and compact families. The central insight is that holomorphicity converts a single uniform bound on values into a hierarchy of derivative bounds via the Cauchy integral formula, generating the equicontinuity that Arzelà-Ascoli requires. The pattern recurs in Vitali's convergence theorem below (where pointwise convergence on an accumulation set plus local boundedness forces uniform-on-compacta convergence) and in the fundamental normality test (where omission of two values replaces local boundedness via the Schottky-Picard machinery). The bridge is that local boundedness is the holomorphic analogue of equicontinuity on a metric space — once seen as such, every existence theorem in conformal mapping, complex dynamics, and value-distribution theory becomes a direct consequence.

Exercises [Intermediate+]

Exercise 3 (medium, symbolic).

Prove the Weierstrass theorem on uniform limits: if ${f_{n}} \subseteq O (Ω)$ converges uniformly on compact subsets to $f : Ω \to C$ , then $f$ is holomorphic on $Ω$ and $f_{n}^{'} \to f^{'}$ uniformly on compact subsets.

Hint

Pass to the limit in the Cauchy integral formula $f_{n} (z) = (2 π i)^{- 1} \oint_{γ} f_{n} (ζ) / (ζ - z) d ζ$ on a small circle around any point of $Ω$ .

Answer

Fix $z_{0} \in Ω$ and a closed disc $\overline{B_{r} (z_{0})} \subset Ω$ . For every $z \in B_{r} (z_{0})$ and every $n$ , the Cauchy integral formula gives $f_{n} (z) = (2 π i)^{- 1} \oint_{∣ ζ - z_{0} ∣ = r} f_{n} (ζ) / (ζ - z) d ζ$ . Uniform convergence on the compact circle $∣ ζ - z_{0} ∣ = r$ lets the limit pass through the contour integral: $f (z) = (2 π i)^{- 1} \oint f (ζ) / (ζ - z) d ζ$ , which exhibits $f$ as the Cauchy integral of its boundary values on the disc, hence holomorphic on $B_{r} (z_{0})$ . Differentiating both representations under the integral sign and using the same uniform-limit argument gives $f_{n}^{'} (z) \to f^{'} (z)$ , with uniform convergence on $B_{r /2} (z_{0})$ supplied by a second Cauchy estimate. A covering of any compact $K \subset Ω$ by finitely many such half-radius discs gives uniform convergence of $f_{n}^{'} \to f^{'}$ on $K$ . $□$

Exercise 4 (medium, symbolic).

State and prove the Vitali convergence theorem: let ${f_{n}}$ be a locally bounded sequence in $O (Ω)$ with $Ω$ connected. If $f_{n}$ converges pointwise on a set $E \subset Ω$ with an accumulation point in $Ω$ , then $f_{n}$ converges uniformly on every compact subset of $Ω$ to a holomorphic limit. ^{[Vitali 1903]}

Hint

Use Montel's theorem to extract a uniformly-on-compacta convergent subsequence, then use the identity theorem to show any two such limits must agree.

Answer

By Montel's theorem, the locally bounded family ${f_{n}}$ is normal, so every subsequence has a further subsequence converging uniformly on compacta to a holomorphic limit. Let ${f_{n_{k}}}$ and ${f_{m_{j}}}$ be two such convergent subsequences with limits $g$ and $h$ respectively. On the set $E$ with accumulation point, $f_{n_{k}} \to f^{*}$ pointwise (by hypothesis) and uniformly-on-compacta to $g$ ; the two limits agree on $E$ , so $g ∣_{E} = f^{*} ∣_{E}$ . Similarly $h ∣_{E} = f^{*} ∣_{E}$ . Hence $g = h$ on $E$ . The set $E$ has an accumulation point in $Ω$ , so by the identity theorem for holomorphic functions on the connected $Ω$ , $g = h$ on all of $Ω$ .

Every uniformly-on-compacta convergent subsequence of ${f_{n}}$ has the same limit. Suppose $f_{n}$ did not converge uniformly on some compact $K \subset Ω$ ; then there exists $ε > 0$ and a subsequence ${f_{n_{k}}}$ with $sup_{K} ∣ f_{n_{k}} - g ∣ \geq ε$ . By normality this subsequence has a sub-subsequence converging uniformly on $K$ to a limit, which must equal $g$ — contradicting $sup_{K} ∣ f_{n_{k}} - g ∣ \geq ε$ . So $f_{n} \to g$ uniformly on every compact $K \subset Ω$ . $□$

Exercise 6 (hard, symbolic).

State the fundamental normality test: a family of holomorphic functions on a domain $Ω$ that uniformly omits two finite values is normal. Indicate how the proof reduces to Montel via a Schottky-type bound.

Hint

Compose with a Möbius transformation to assume the omitted values are $0$ and $1$ ; then map the doubly-punctured plane $C ∖ {0, 1}$ to the unit disc via the modular function $λ^{- 1}$ , lifting holomorphic functions to bounded holomorphic functions on the disc.

Answer

Statement. Let $F$ be a family of holomorphic functions on a domain $Ω \subseteq C$ , and suppose there exist two distinct values $a, b \in C$ such that $f (z) \neq = a$ and $f (z) \neq = b$ for every $f \in F$ and every $z \in Ω$ . Then $F$ is normal (in the spherical sense on the Riemann sphere; in particular, normal in the classical sense if additionally locally bounded). ^{[Montel 1912]}

Proof sketch. Compose with the Möbius transformation $z \mapsto (z - a) / (b - a)$ to reduce to omission of $0$ and $1$ . The doubly-punctured plane $C ∖ {0, 1}$ has universal cover the upper half-plane $H$ via the modular function $λ : H \to C ∖ {0, 1}$ (Picard's first proof, 1879, via the modular function); equivalently $D$ via composition with the Cayley transform. Each $f \in F$ omits ${0, 1}$ , so locally lifts to a holomorphic $\tilde{f} : Ω \to D$ via $λ$ . The lifted family is uniformly bounded by $1$ , hence normal by Montel. Composing back with $λ$ — a continuous map $D \to C_{\infty}$ extending continuously to $\partial D$ with boundary values in ${0, 1, \infty}$ — extracts a uniformly-on-compacta convergent subsequence of $F$ in the spherical metric on $C_{\infty}$ , with possible spherical limit at $\infty$ . The Schottky bound ( $f$ holomorphic on $D$ with $f (0) = a_{0}$ and omitting $0, 1$ has a uniform bound $∣ f (z) ∣ \leq S (∣ a_{0} ∣, r)$ on $∣ z ∣ \leq r < 1$ ) makes the bound explicit and recovers the lifted boundedness without invoking the modular function as an explicit object. $□$

Exercise 7 (hard, symbolic).

Use Montel plus Hurwitz to prove the existence half of the Riemann mapping theorem: every simply connected proper plane domain $Ω ⊊ C$ admits a biholomorphism to the open unit disc $D$ .

Hint

Fix $z_{0} \in Ω$ and let $F := {f : Ω \to D : f holomorphic and injective, f (z_{0}) = 0, f^{'} (z_{0}) > 0}$ . Show $F$ is non-empty (using a holomorphic square root and a Möbius rescaling), then maximise $f^{'} (z_{0})$ over $F$ via Montel.

Answer

Step A — non-emptiness. Since $Ω \neq = C$ , pick $a \in / Ω$ . The function $z \mapsto z - a$ omits $0$ on $Ω$ ; on a simply connected $Ω$ a holomorphic logarithm exists, hence a holomorphic square root $z - a$ . This square root is injective (a holomorphic non-vanishing square root composed with itself recovers $z - a$ , injective, so the square root is injective). The image avoids some open disc (the image of a non-empty open set under a non-constant holomorphic function is open, and the image of any point and its antipode under the square root are distinct, so the image and its negative are disjoint open sets). Compose with a Möbius transformation that maps the image complement into $D$ , then a rotation to make $f^{'} (z_{0}) > 0$ ; the resulting map $f_{1} \in F$ witnesses non-emptiness.

Step B — local boundedness. Every $f \in F$ maps into $D$ , so $∣ f ∣ \leq 1$ on $Ω$ . The family is uniformly bounded by $1$ ; by Montel, $F$ is normal.

Step C — extremal sequence. The functional $f \mapsto f^{'} (z_{0})$ is bounded on $F$ (Schwarz lemma on the disc through which $f$ factors gives $∣ f^{'} (z_{0}) ∣ \leq C$ for a constant $C$ depending only on $z_{0}$ and $Ω$ ). Let $μ := sup_{F} f^{'} (z_{0})$ . Choose a sequence ${f_{n}} \subset F$ with $f_{n}^{'} (z_{0}) \to μ$ . By Montel, extract a uniformly-on-compacta convergent subsequence with holomorphic limit $f^{*} \in O (Ω)$ .

Step D — properties of the limit. Uniform convergence on a compact circle around $z_{0}$ plus the Weierstrass theorem on derivatives gives $f^{*}^{'} (z_{0}) = μ > 0$ (the limit derivative is the limit of derivatives, and $f^{*}^{'} (z_{0}) > 0$ follows from $f_{n}^{'} (z_{0}) > 0$ converging to $μ > 0$ ). The limit $f^{*}$ maps $Ω$ into the closed disc $\overline{D}$ ; by the maximum principle (06.01.12) applied to $f^{*}$ (non-constant since $f^{*}^{'} (z_{0}) > 0$ ), $f^{*}$ maps $Ω$ into the open disc $D$ . Each $f_{n}$ is injective; by Hurwitz's theorem (06.01.13), the limit $f^{*}$ is either injective or constant; constancy would give $f^{*}^{'} (z_{0}) = 0$ , contradicting $μ > 0$ , so $f^{*}$ is injective. Hence $f^{*} \in F$ .

Step E — surjectivity. Suppose for contradiction that $f^{*} (Ω) ⊊ D$ . Pick $α \in D ∖ f^{*} (Ω)$ . The Möbius automorphism $ϕ_{α} (z) := (z - α) / (1 - \overset{α}{ˉ} z)$ of $D$ sends $α$ to $0$ ; the composition $ϕ_{α} \circ f^{*}$ is non-vanishing on $Ω$ , so on the simply connected $Ω$ admits a holomorphic square root $g$ . Set $\tilde{f} := ϕ_{g (z_{0})} \circ g$ , rotated so $\tilde{f}^{'} (z_{0}) > 0$ . A direct computation shows $\tilde{f}^{'} (z_{0}) > μ$ (the square root strictly increases the derivative when $α \neq = 0$ ), contradicting maximality of $μ$ . So $f^{*}$ is surjective; $f^{*} : Ω \to D$ is a biholomorphism. $□$

The proof is the canonical Carathéodory-Koebe-Fejér 1907–1916 argument. The two Montel inputs — normality of $F$ and uniform-on-compacta convergence of derivatives — are what makes the extremal problem solvable.

Exercise 8 (hard, symbolic).

State and sketch the proof of the Carathéodory kernel theorem: a sequence of simply connected domains $Ω_{n} ∋ z_{0}$ converges in the kernel sense to a domain $Ω ∋ z_{0}$ if and only if the corresponding Riemann maps $f_{n} : D \to Ω_{n}$ (normalised by $f_{n} (0) = z_{0}$ , $f_{n}^{'} (0) > 0$ ) converge uniformly on compact subsets of $D$ to the Riemann map $f : D \to Ω$ .

Hint

The kernel of a sequence ${Ω_{n}}$ at $z_{0}$ is the largest domain $Ω ∋ z_{0}$ such that every compact subset of $Ω$ lies in $Ω_{n}$ for all sufficiently large $n$ . The proof combines Montel (locally bounded family of injective maps into the unit disc) with Hurwitz (injectivity preserved in the limit) and an explicit identification of the limit Riemann map.

Answer

Kernel definition. The kernel of ${Ω_{n}}$ at $z_{0}$ is the connected component containing $z_{0}$ of the interior of $⋂_{N} ⋃_{n \geq N} Ω_{n}$ — the largest domain $Ω ∋ z_{0}$ all of whose compact subsets eventually sit inside every $Ω_{n}$ . The sequence converges in the kernel sense to $Ω$ when $Ω$ is the kernel and every subsequence has the same kernel.

Statement. If $Ω_{n} ∋ z_{0}$ are simply connected proper plane domains with Riemann maps $f_{n} : D \to Ω_{n}$ , $f_{n} (0) = z_{0}$ , $f_{n}^{'} (0) > 0$ , then $f_{n} \to f$ uniformly on compacta in $D$ for some $f : D \to Ω$ if and only if $Ω_{n} \to Ω$ in the kernel sense, and in that case $f$ is the Riemann map of $Ω$ .

Proof sketch. Forward direction: assume $f_{n} \to f$ uniformly on compacta. By Hurwitz, $f$ is injective. The image $f (D)$ contains every point $w \in D$ such that $w$ is the value of $f$ at some point, and the open-mapping property gives that $f (D)$ is open and equals the kernel of ${f_{n} (D)} = {Ω_{n}}$ at $z_{0}$ .

Reverse direction: assume $Ω_{n} \to Ω$ in the kernel sense. The family ${f_{n}}$ maps $D$ into a uniformly bounded set (each $Ω_{n}$ contains a fixed disc around $z_{0}$ for large $n$ ; by Koebe's distortion theorem, the $f_{n}$ are uniformly bounded on compacta of $D$ ). By Montel, extract a uniformly convergent subsequence $f_{n_{k}} \to g$ . By Hurwitz, $g$ is injective; the image $g (D)$ is the kernel, namely $Ω$ . By uniqueness of the Riemann map (normalised at $z_{0}$ with positive derivative), $g$ equals the Riemann map of $Ω$ . Since every convergent subsequence has the same limit, the whole sequence converges. $□$

The Carathéodory kernel theorem is the foundational continuity statement for the dependence of the Riemann map on its target domain, and is used throughout complex dynamics (Julia-set continuity, hyperbolic component theory) and in the iterated-Riemann-map approach to the measurable Riemann mapping theorem. ^{[Schiff Normal Families]}

Lean formalization [Intermediate+]

lean_status: none. Mathlib has the Arzelà-Ascoli theorem in topological-space form via ArzelaAscoli, the Cauchy integral formula via Complex.circleIntegral, and the uniform-on-compacta topology via UniformOnFun.compactConvergence, but the curriculum-facing Montel theorem and Vitali theorem are not assembled. Specifically missing: a NormalFamily predicate, Montel.locallyBounded_iff_normal, the derivative-bound extraction Complex.deriv_le_of_locallyBounded, the diagonal-extraction lemma Filter.subseq_diagonal_of_exhaustion, the Vitali convergence theorem Complex.vitali_convergence, and the fundamental normality test FundamentalNormalityTest.omits_two_values_iff_normal. The lean_mathlib_gap field records the precise contribution roadmap. The unit ships without a lean_module while these upstream gaps remain open; the human-reviewer gate covers correctness in the interim.

Advanced results [Master]

Vitali convergence theorem. Let $Ω \subseteq C$ be a connected open set, ${f_{n}} \subset O (Ω)$ a locally bounded sequence. If $f_{n}$ converges pointwise on a set $E \subset Ω$ with an accumulation point in $Ω$ , then $f_{n}$ converges uniformly on every compact subset of $Ω$ to a holomorphic limit. Vitali's 1903 paper Sopra le serie di funzioni analitiche (Rendiconti del R. Istituto Lombardo, Serie II, Vol. XXXVI) ^{[Vitali 1903]} established the result independently of Montel's later normality framework; the modern proof (Exercise 4) packages Montel plus the identity theorem cleanly. Vitali is the holomorphic analogue of the Banach-Steinhaus / Vitali-Hahn-Saks principle in functional analysis: a uniform regularity hypothesis (local boundedness) combined with a weak convergence hypothesis (pointwise on a set with accumulation) upgrades to a strong convergence (uniform on compacta).

Fundamental normality test. A family of holomorphic functions on a domain $Ω$ uniformly omitting two finite values is normal in the spherical sense. The proof (Exercise 6) reduces to Montel via the Schottky bound or, equivalently, via the modular function $λ$ uniformising the doubly punctured plane $C ∖ {0, 1}$ . The fundamental normality test extends to uniform omission of three values on $C_{\infty}$ , where the third value is allowed to be $\infty$ and the framework is meromorphic. The test is the foundational input for Picard's great theorem (a non-constant entire function omits at most one value in any neighbourhood of an essential singularity).

Montel's three-value theorem. A family of meromorphic functions on $Ω$ that uniformly omits three values on $C_{\infty}$ is normal in the spherical sense on $Ω$ . This generalises the holomorphic two-value test by including $\infty$ as one of the omitted values. Montel established it in the same 1912 paper; the proof is the spherical-metric version of the modular-function argument.

Marty's theorem. A family $F$ of meromorphic functions on $Ω$ is normal in the spherical sense if and only if the spherical derivative $f^{#} (z) := ∣ f^{'} (z) ∣/ (1 + ∣ f (z) ∣^{2})$ is locally bounded on $Ω$ uniformly in $f \in F$ . Marty 1931 Recherches sur la répartition des valeurs d'une fonction méromorphe (Annales de la Faculté des sciences de Toulouse) ^{[Marty 1931]} gave the necessary-and-sufficient form, the meromorphic analogue of Montel's holomorphic local-boundedness criterion. The spherical-derivative bound is the natural quantity once one moves to $C_{\infty}$ , where uniform boundedness fails for sequences approaching $\infty$ .

Zalcman renormalisation lemma. Lawrence Zalcman 1975 A heuristic principle in complex function theory (American Mathematical Monthly, vol. 82, pp. 813–817) ^{[Zalcman 1975]} established the modern renormalisation principle: a family $F$ of meromorphic functions on $Ω$ is not normal at $z_{0}$ if and only if there exist sequences $z_{n} \to z_{0}$ with $z_{n} \in Ω$ , $ρ_{n} \to 0^{+}$ , and $f_{n} \in F$ such that the renormalised sequence $g_{n} (ζ) := f_{n} (z_{n} + ρ_{n} ζ)$ converges locally uniformly on $C$ to a non-constant meromorphic function $g : C \to C_{\infty}$ . The Zalcman lemma is the operational form of the Bloch principle: any property of holomorphic functions that prevents the existence of a non-constant entire (or meromorphic on $C$ ) function with that property forces every family with that property to be normal.

Generalisations to higher dimensions. The Montel framework extends to holomorphic functions of several complex variables on a domain $Ω \subseteq C^{n}$ : a locally bounded family is normal in the topology of uniform convergence on compact subsets. The proof passes through the multi-variable Cauchy integral formula on a polydisc and the same Arzelà-Ascoli plus diagonal-extraction argument. The fundamental normality test generalises to families omitting a hypersurface of codimension $1$ in $C^{n}$ , with the Kobayashi-hyperbolic complement playing the role of the doubly punctured plane.

Application to value-distribution theory. Nevanlinna's 1925 Zur Theorie der meromorphen Funktionen (Acta Mathematica, vol. 46, pp. 1–99) ^{[Nevanlinna 1925]} used the Schottky bound (the explicit form of the fundamental normality test) to derive the defect relation for meromorphic functions: the sum of defects of a meromorphic function $f$ over all values is bounded by $2$ . The proof combines the normal-family compactness with the Ahlfors-Shimizu characteristic, producing the modern value-distribution framework.

Synthesis. Montel's theorem and the normal-family framework supply the structural junction connecting compactness in functional analysis with the rigidity of holomorphic functions, and the synthesis runs in four directions. First, the Arzelà-Ascoli pattern: local boundedness is the holomorphic analogue of equicontinuity; the Cauchy derivative estimate is the load-bearing input that converts a single hypothesis (bounded values) into the pair of hypotheses (bounded values, bounded derivatives) needed for Arzelà-Ascoli. The pattern recurs across analysis: the Banach-Alaoglu theorem (weak- $*$ compactness of bounded sets in dual Banach spaces) is the linear-functional analogue; the Rellich-Kondrachov theorem (compact embedding of Sobolev spaces into Lebesgue spaces) is the PDE analogue.

Second, the extremal-problem method in conformal mapping: the Riemann mapping theorem (06.01.06) is the canonical application — Montel supplies sequential compactness on a family of candidate maps, Hurwitz (06.01.13) preserves injectivity at the limit, and the Carathéodory-Koebe extremal argument selects the unique biholomorphism. The pattern recurs in the Bieberbach conjecture (de Branges 1985: the coefficients of a normalised univalent map on $D$ satisfy $∣ a_{n} ∣ \leq n$ , proved by extremal-problem methods on a compact family of univalent maps), the measurable Riemann mapping theorem (Ahlfors-Bers 1960: extremal selection on quasiconformal maps), and the Loewner equation (every univalent map on $D$ embeds in a one-parameter family solving a controlled evolution equation).

Third, the dynamical-systems application via the Fatou-Julia framework: a meromorphic map $f : C_{\infty} \to C_{\infty}$ has its Fatou set $F (f)$ — the set of points around which the iterate family ${f, f^{\circ 2}, f^{\circ 3}, \dots}$ is normal — and its Julia set $J (f)$ — the complement, where the iterate family fails to be normal. The Julia set is the closure of repelling periodic points; the Fatou set decomposes into invariant components classified by the Sullivan no-wandering-domain theorem. The entire architecture of complex dynamics rests on normality of iterate families, and Montel's theorem is the analytic engine. The fundamental normality test gives the explicit Picard-like results: a Julia set contains either at most two exceptional values or the whole sphere.

Fourth, the higher-dimensional and Kobayashi-hyperbolic generalisation: a complex manifold $X$ is Kobayashi hyperbolic when its Kobayashi pseudometric is a genuine metric — equivalently, when every holomorphic map $C \to X$ is constant. Brody's 1978 theorem states that a compact complex manifold $X$ is Kobayashi hyperbolic if and only if every family of holomorphic maps $D \to X$ is normal — Montel's theorem upgraded to the manifold setting. Putting these together identifies the foundational principle as the dichotomy normal-or-Picard: a family of holomorphic maps is either compact (normal) or contains enough freedom to escape every reasonable bound (omits at most a small set of values, equivalently is "value-rich"). This is exactly the bridge between compactness and Picard-type omission theorems, and the central insight underlying every existence theorem in modern complex analysis.

Full proof set [Master]

Proposition (Cauchy derivative estimate). Let $f$ be holomorphic on a neighbourhood of the closed disc $\overline{B_{r} (z_{0})}$ with $∣ f (z) ∣ \leq M$ on the boundary circle $∣ z - z_{0} ∣ = r$ . Then $∣ f^{(k)} (z_{0}) ∣ \leq k! M / r^{k}$ for every integer $k \geq 0$ .

Proof. The Cauchy integral formula for the $k$ -th derivative (06.01.02) reads $f^{(k)} (z_{0}) = (k! /2 π i) \oint_{∣ z - z_{0} ∣ = r} f (z) / (z - z_{0})^{k + 1} d z$ . Estimating the integral: the length of the contour is $2 π r$ ; the integrand has modulus $\leq M / r^{k + 1}$ on the contour. Hence $∣ f^{(k)} (z_{0}) ∣ \leq (k! /2 π) \cdot 2 π r \cdot M / r^{k + 1} = k! M / r^{k}$ . $□$

Proposition (Arzelà-Ascoli, classical form). Let $K$ be a compact metric space and $F \subseteq C (K, C)$ a family of continuous complex-valued functions. Then $F$ is precompact in the supremum norm if and only if $F$ is uniformly bounded and equicontinuous.

Proof. $(\Rightarrow)$ A precompact family is bounded in $C (K)$ by total boundedness, and equicontinuous because $K$ is compact: any finite $ε /3$ -net for $F$ has equicontinuity uniform at each net function (by uniform continuity on the compact $K$ ) and uniform across $F$ within $ε$ of the net.

$(\Leftarrow)$ Cover $K$ by finitely many balls of radius $δ$ (compactness; $δ$ from the equicontinuity at $ε$ ). On the finite set of ball centres, $F$ values lie in a bounded subset of $C^{N}$ (where $N$ is the number of centres), hence precompact in $C^{N}$ . Extract a subsequence whose values on the centres converge. Equicontinuity then upgrades convergence on centres to uniform convergence on $K$ via the triangle inequality $∣ f_{n} (z) - f_{n} (c) ∣ < ε$ (equicontinuity) plus convergence on $c$ . $□$ ^{[Arzelà 1895; Ascoli 1883]}

Proposition (Weierstrass uniform-limit theorem for holomorphic functions). If ${f_{n}} \subset O (Ω)$ converges to $f$ uniformly on every compact subset of $Ω$ , then $f \in O (Ω)$ and $f_{n}^{(k)} \to f^{(k)}$ uniformly on every compact subset, for every $k \geq 0$ .

Proof. Holomorphy of $f$ is Exercise 3 (pass to the limit in the Cauchy integral formula). Convergence of derivatives: the Cauchy integral formula for the $k$ -th derivative on a circle of radius $r$ around any $z_{0} \in K$ gives $f_{n}^{(k)} (z_{0}) = (k! /2 π i) \oint_{∣ z - z_{0} ∣ = r} f_{n} (z) / (z - z_{0})^{k + 1} d z$ and similarly for $f$ . The difference $∣ f_{n}^{(k)} (z_{0}) - f^{(k)} (z_{0}) ∣ \leq (k! / r^{k}) sup_{∣ z - z_{0} ∣ = r} ∣ f_{n} (z) - f (z) ∣$ tends to zero uniformly in $z_{0} \in K_{r /2}$ (a slightly smaller compact, to keep the circle inside $Ω$ ). $□$

Theorem (Montel, full statement). A family $F \subset O (Ω)$ is normal if and only if it is locally bounded.

Proof. Stated and proved in the Key theorem section above. The implication "locally bounded $\Rightarrow$ normal" passes through the Cauchy derivative estimate, the equicontinuity bound, Arzelà-Ascoli on each compact, and diagonal extraction over an exhaustion of $Ω$ . The reverse implication uses sequential compactness directly: a non-locally-bounded family has a sequence escaping to infinity on some compact, contradicting any uniform convergence on that compact. $□$ ^{[Montel 1912]}

Theorem (Vitali, full statement). A locally bounded sequence ${f_{n}} \subset O (Ω)$ on a connected open set, converging pointwise on a set $E \subset Ω$ with an accumulation point in $Ω$ , converges uniformly on every compact subset of $Ω$ to a holomorphic limit.

Proof. Exercise 4 gives the proof: Montel produces convergent subsequences; the identity theorem forces all such subsequence limits to agree on $Ω$ (they agree on $E$ , hence on its accumulation point, hence on the connected $Ω$ ); a sub-subsequence argument upgrades subsequential convergence to convergence of the full sequence. $□$ ^{[Vitali 1903]}

Theorem (Fundamental normality test). A family $F$ of holomorphic functions on a domain $Ω$ uniformly omitting two finite values $a, b \in C$ is normal.

Proof. Exercise 6 gives the proof via reduction to a uniformly bounded family on a covering space (the unit disc, via the modular function $λ$ uniformising $C ∖ {0, 1}$ ), or equivalently via the explicit Schottky bound for holomorphic functions on $D$ omitting ${0, 1}$ . ^{[Montel 1912]}

Corollary (Picard's great theorem via normality). A non-constant entire function omits at most one finite value, and an entire function with an essential singularity at $\infty$ attains every finite value (with at most one exception) infinitely often in any neighbourhood of $\infty$ .

Proof. Suppose for contradiction $f$ entire omits two finite values $a, b$ . The family ${f_{n} (z) := f (n z)}$ on $D$ uniformly omits ${a, b}$ , so is normal by the fundamental normality test. Each $f_{n}$ has $f_{n} (0) = f (0)$ , a fixed value. By normality, extract a uniformly convergent subsequence ${f_{n_{k}}}$ on $D$ with limit $g$ . But $∣ f_{n_{k}}^{'} (0) ∣ = n_{k} ∣ f^{'} (0) ∣ \to \infty$ (if $f^{'} (0) \neq = 0$ ), contradicting the limit-derivative bound. If $f^{'} (0) = 0$ , repeat at any other point where the derivative is non-zero (which exists since $f$ is non-constant). Hence $f$ omits at most one value. The essential-singularity version follows by the same argument applied near $\infty$ via the substitution $z \mapsto 1/ z$ . $□$

Connections [Master]

Cauchy integral formula 06.01.02 — The derivative estimate at the heart of Montel's theorem is a direct application of the Cauchy integral formula for the derivative. The Cauchy formula $f^{(k)} (z_{0}) = (k! /2 π i) \oint f (z) / (z - z_{0})^{k + 1} d z$ converts a uniform bound on the values into a uniform bound on every derivative, which is the foundational reason local boundedness of values forces equicontinuity. Without the Cauchy formula, no analogue of Montel's theorem holds — and indeed real-analytic functions do not enjoy a Montel-type compactness statement.
Argument principle and Rouché 06.01.13 — Hurwitz's theorem (a corollary of the argument principle) is the second-half companion of Montel in the conformal-mapping existence argument: Montel produces a uniform-on-compacta limit of injective holomorphic maps, and Hurwitz confirms that the limit is itself injective (rather than degenerating to a constant). The combination "Montel for compactness, Hurwitz for injectivity preservation" is the analytic engine of the Riemann mapping theorem.
Riemann mapping theorem 06.01.06 — The existence half of the Riemann mapping theorem is the canonical application of Montel plus Hurwitz, exhibited in Exercise 7. The proof packages the family of normalised injective maps $Ω \to D$ as a normal family by Montel (uniform bound $1$ ), selects a derivative-maximising sequence, extracts a uniformly convergent subsequence by Montel's compactness, and preserves injectivity by Hurwitz. The resulting extremal map is the desired biholomorphism. The uniqueness half uses the Schwarz lemma (06.01.12).
Metric space 02.01.05 — Arzelà-Ascoli is a theorem in the abstract metric-space framework: a uniformly bounded and equicontinuous family of continuous functions on a compact metric space is precompact in the supremum norm. Montel's theorem applies Arzelà-Ascoli on each compact subset $K$ of $Ω$ , with the metric structure of $K \subset C$ supplying the necessary compactness and continuity language. The Lipschitz bound on holomorphic functions derived from the Cauchy estimate is the load-bearing equicontinuity input.
Maximum modulus and Schwarz lemma 06.01.12 — In the Riemann mapping theorem proof, the maximum principle is what confirms that the extremal limit $f^{*}$ maps into the open disc $D$ (rather than touching the boundary), and the Schwarz lemma bound $∣ f^{'} (0) ∣ \leq 1/ r$ on injective maps $D \to D$ supplies the upper bound that ensures the supremum of $∣ f^{'} (z_{0}) ∣$ over the normal family is finite. Without these companion theorems, the Montel-Hurwitz extraction would not produce a biholomorphism.
Picard's theorems [foundations.picard-theorems] (pending) — The little Picard theorem (a non-constant entire function omits at most one finite value) and the great Picard theorem (near an essential singularity, every value is attained infinitely often except at most one) follow from the fundamental normality test plus standard rescaling arguments. The normality framework is what converts "omission of two values" into "global rigidity" for holomorphic functions.
Complex dynamics — Fatou-Julia sets [foundations.complex-dynamics] (pending) — The Fatou set of a holomorphic self-map $f$ of $C_{\infty}$ is by definition the set of points around which the iterate family is normal. The Julia set is the complement: the locus of chaotic behaviour. The classification of Fatou components, the Sullivan no-wandering-domain theorem, and the entire architecture of complex dynamics rest on Montel's normality framework.

Historical & philosophical context [Master]

Paul Montel introduced the concept of a normal family in his 1907 doctoral thesis and developed the full theory in Sur les suites infinies de fonctions (Annales scientifiques de l'École Normale Supérieure, 3e série, t. 29, 1912, pp. 487–535) ^{[Montel 1912]}. The thesis title — Sur les suites infinies de fonctions — already states the central question: when does an infinite sequence of holomorphic functions admit a convergent subsequence? Montel's answer was the local-boundedness criterion now bearing his name, generalising and packaging earlier compactness arguments scattered across the late-nineteenth-century literature. Montel went on to write the comprehensive monograph Leçons sur les familles normales de fonctions analytiques et leurs applications (Gauthier-Villars, 1927), which fixed the modern terminology and remained the standard reference for half a century.

The precursors to Montel's theorem lie in two threads. The first is the Arzelà-Ascoli compactness lemma, established by Giulio Ascoli in Le curve limiti di una varietà data di curve (Atti della R. Accademia dei Lincei, Memorie della Cl. Sci. Fis. Mat. Nat., Serie III, Vol. XVIII, 1883, pp. 521–586) ^{[Ascoli 1883]} and refined by Cesare Arzelà in Sulle funzioni di linee (Memorie della R. Accademia delle Scienze di Bologna, Serie V, Vol. V, 1895, pp. 55–74) ^{[Arzelà 1895]}. Ascoli identified equicontinuity as the regularity hypothesis converting bounded sets of continuous functions into precompact sets in the supremum norm; Arzelà completed the converse and packaged the result as the modern criterion. The second thread is Thomas Joannes Stieltjes's 1894 Recherches sur les fractions continues (Annales de la Faculté des sciences de Toulouse, Tome VIII, 1894) ^{[Stieltjes 1894]}, which used a diagonal-extraction lemma to produce uniform-on-compacta convergent subsequences of holomorphic functions arising from continued-fraction expansions. Stieltjes's argument was the operational precursor to Montel's theorem, made systematic and applicable beyond the specific Padé-approximation setting.

The Vitali convergence theorem appeared in Giuseppe Vitali's Sopra le serie di funzioni analitiche (Rendiconti del R. Istituto Lombardo, Serie II, Vol. XXXVI, 1903, pp. 771–774) ^{[Vitali 1903]}, independently of Montel's later normality framework. Vitali proved the result by reduction to the Weierstrass uniform-limit theorem plus an analytic-continuation argument; the modern proof in terms of Montel-normality plus the identity theorem is a streamlined repackaging. The same theorem was established a year later by Erhard Schmidt in Über die Konvergenz von Reihen, deren Glieder analytische Funktionen sind (Mathematische Annalen 60, 1905), but Vitali's priority is universally credited.

The fundamental normality test — uniform omission of two finite values implies normality — has roots in two distinct nineteenth-century results. Émile Picard's 1879 Mémoire sur les fonctions entières (Annales scientifiques de l'É.N.S.) ^{[Picard 1879]} established that a non-constant entire function omits at most one finite value, by reducing to a modular-function argument on the doubly punctured plane. Friedrich Schottky's 1904 Über den Picard'schen Satz und die Borel'schen Ungleichungen (Sitzungsberichte der Königlich Preussischen Akademie der Wissenschaften zu Berlin, 1904) ^{[Schottky 1904]} supplied the quantitative bound now bearing his name — an explicit upper bound for $∣ f (z) ∣$ on $∣ z ∣ \leq r$ given $∣ f (0) ∣$ , for $f$ holomorphic on the unit disc omitting ${0, 1}$ . Montel's 1912 paper synthesised the qualitative Picard-omission theorem with the quantitative Schottky bound into the family-level normality test. Edmund Landau's 1904 contributions (Über eine Verallgemeinerung des Picardschen Satzes, Berlin Sitzungsberichte) ^{[Landau 1904]} sharpened the bound near the origin and connected the framework to Picard's later great theorem on essential singularities.

The pedagogical synthesis of this French-Italian-German nineteenth-century arc as the foundational chapter of normal-family theory is Lars Ahlfors's Complex Analysis Ch. 5 (1953, revised editions through 1979) ^{[Ahlfors Complex Analysis]}, which fixed the modern textbook treatment of Montel's theorem as the compactness engine behind the Riemann mapping theorem. Joel Schiff's Normal Families (Springer Universitext, 1993) ^{[Schiff Normal Families]} gives the comprehensive twentieth-century treatment, including the Zalcman renormalisation lemma, Marty's theorem, the Bloch principle, and the value-distribution-theoretic applications.

Bibliography [Master]

@article{Montel1912,
  author  = {Montel, Paul},
  title   = {Sur les suites infinies de fonctions},
  journal = {Annales scientifiques de l'\'Ecole Normale Sup\'erieure},
  series  = {3e s{\'e}rie},
  volume  = {29},
  year    = {1912},
  pages   = {487--535}
}

@book{MontelFamilles1927,
  author    = {Montel, Paul},
  title     = {Le{\c c}ons sur les familles normales de fonctions analytiques et leurs applications},
  publisher = {Gauthier-Villars},
  year      = {1927},
  address   = {Paris}
}

@article{Vitali1903,
  author  = {Vitali, Giuseppe},
  title   = {Sopra le serie di funzioni analitiche},
  journal = {Rendiconti del R. Istituto Lombardo di Scienze e Lettere},
  series  = {Serie II},
  volume  = {36},
  year    = {1903},
  pages   = {771--774}
}

@article{Stieltjes1894,
  author  = {Stieltjes, Thomas Joannes},
  title   = {Recherches sur les fractions continues},
  journal = {Annales de la Facult{\'e} des sciences de Toulouse},
  volume  = {8},
  year    = {1894},
  pages   = {1--122}
}

@article{Ascoli1883,
  author  = {Ascoli, Giulio},
  title   = {Le curve limiti di una variet{\`a} data di curve},
  journal = {Atti della R. Accademia dei Lincei, Memorie della Cl. Sci. Fis. Mat. Nat.},
  series  = {Serie III},
  volume  = {18},
  year    = {1883},
  pages   = {521--586}
}

@article{Arzela1895,
  author  = {Arzel{\`a}, Cesare},
  title   = {Sulle funzioni di linee},
  journal = {Memorie della R. Accademia delle Scienze dell'Istituto di Bologna},
  series  = {Serie V},
  volume  = {5},
  year    = {1895},
  pages   = {55--74}
}

@article{Schottky1904,
  author  = {Schottky, Friedrich},
  title   = {{\"U}ber den Picard'schen Satz und die Borel'schen Ungleichungen},
  journal = {Sitzungsberichte der K{\"o}niglich Preussischen Akademie der Wissenschaften zu Berlin},
  year    = {1904},
  pages   = {1244--1262}
}

@article{Landau1904,
  author  = {Landau, Edmund},
  title   = {{\"U}ber eine Verallgemeinerung des Picardschen Satzes},
  journal = {Sitzungsberichte der K{\"o}niglich Preussischen Akademie der Wissenschaften zu Berlin},
  year    = {1904},
  pages   = {1118--1133}
}

@article{Picard1879,
  author  = {Picard, {\'E}mile},
  title   = {M{\'e}moire sur les fonctions enti{\`e}res},
  journal = {Annales scientifiques de l'\'Ecole Normale Sup{\'e}rieure},
  series  = {2e s{\'e}rie},
  volume  = {9},
  year    = {1880},
  pages   = {145--166}
}

@article{Marty1931,
  author  = {Marty, Fr{\'e}d{\'e}ric},
  title   = {Recherches sur la r{\'e}partition des valeurs d'une fonction m{\'e}romorphe},
  journal = {Annales de la Facult{\'e} des sciences de Toulouse},
  series  = {3e s{\'e}rie},
  volume  = {23},
  year    = {1931},
  pages   = {183--261}
}

@article{Zalcman1975,
  author  = {Zalcman, Lawrence},
  title   = {A heuristic principle in complex function theory},
  journal = {American Mathematical Monthly},
  volume  = {82},
  year    = {1975},
  pages   = {813--817}
}

@article{Nevanlinna1925,
  author  = {Nevanlinna, Rolf},
  title   = {Zur Theorie der meromorphen Funktionen},
  journal = {Acta Mathematica},
  volume  = {46},
  year    = {1925},
  pages   = {1--99}
}

@book{Schiff1993,
  author    = {Schiff, Joel L.},
  title     = {Normal Families},
  series    = {Universitext},
  publisher = {Springer},
  year      = {1993}
}

@book{Ahlfors1979,
  author    = {Ahlfors, Lars V.},
  title     = {Complex Analysis},
  edition   = {3rd},
  publisher = {McGraw-Hill},
  year      = {1979}
}

@book{ConwayCV1,
  author    = {Conway, John B.},
  title     = {Functions of One Complex Variable I},
  edition   = {2nd},
  series    = {Graduate Texts in Mathematics},
  volume    = {11},
  publisher = {Springer},
  year      = {1978}
}

@book{SteinShakarchiCA,
  author    = {Stein, Elias M. and Shakarchi, Rami},
  title     = {Complex Analysis},
  series    = {Princeton Lectures in Analysis},
  volume    = {II},
  publisher = {Princeton University Press},
  year      = {2003}
}

Prerequisites

06.01.02
06.01.13
02.01.05

Tier anchors

beginner: Needham *Visual Complex Analysis* §9.III informal (families of holomorphic maps and limit extraction); Ahlfors *Complex Analysis* Ch. 5 §5.1 informal opening
intermediate: Ahlfors *Complex Analysis* Ch. 5 §5.1–§5.5; Conway *Functions of One Complex Variable I* §VII.1–§VII.3; Stein-Shakarchi Vol. II §8.3
master: Montel 1912 *Sur les suites infinies de fonctions* (Annales scientifiques de l'É.N.S.); Stieltjes 1894 *Recherches sur les fractions continues* (precursor compactness arguments); Vitali 1903 *Sopra le serie di funzioni analitiche* (convergence theorem); Arzelà 1895 *Sulle funzioni di linee* and Ascoli 1883 *Le curve limiti di una varietà data di curve* (originating equicontinuity compactness); Schottky 1904, Landau 1904 (Picard-type bounds underlying the fundamental normality test); Carathéodory-Landau 1911; Ahlfors *Complex Analysis* Ch. 5; Schiff *Normal Families* (Springer Universitext, 1993)

References

TODO_REF
Montel 1912 — Sur les suites infinies de fonctions · Annales scientifiques de l'É.N.S. 3e série, t. 29 (1912), pp. 487–535 (originator paper for normal families and Montel's theorem)
TODO_REF
Ahlfors — Complex Analysis · Ch. 5 §5.1–§5.5 (modern anchor: normal families, Arzelà-Ascoli compactness, the fundamental normality test, Vitali's theorem, Riemann mapping theorem extraction)
TODO_REF
Stieltjes 1894 — Recherches sur les fractions continues · Annales de la Faculté des sciences de Toulouse (1894), Tome VIII (precursor: uniform-on-compacta convergence and the diagonal-extraction lemma)
TODO_REF
Vitali 1903 — Sopra le serie di funzioni analitiche · Rendiconti del R. Istituto Lombardo, Serie II, Vol. XXXVI (1903), pp. 771–774 (Vitali convergence theorem)
TODO_REF
Arzelà 1895 — Sulle funzioni di linee · Memorie della R. Accademia delle Scienze di Bologna, Serie V, Vol. V (1895), pp. 55–74
TODO_REF
Ascoli 1883 — Le curve limiti di una varietà data di curve · Atti della R. Accademia dei Lincei, Memorie della Cl. Sci. Fis. Mat. Nat., Serie III, Vol. XVIII (1883), pp. 521–586
TODO_REF
Schiff — Normal Families · Springer Universitext (1993): full modern treatment, including the Zalcman renormalisation lemma and the Bloch principle
TODO_REF
Conway — Functions of One Complex Variable I · Springer GTM 11, §VII.1–§VII.3 (normal families, Vitali, the fundamental normality test)
TODO_REF
Stein-Shakarchi — Complex Analysis (Vol. II) · Princeton Lectures in Analysis II, §8.3 (normal families and the Riemann mapping theorem)

Reviewer

TBD

Estimated time

beginner: 16m
intermediate: 40m
master: 70m