Notes on Complex Analysis
Export PDFGitHub

§5.3 The Riemann Mapping Theorem

5.3 The Riemann Mapping Theorem

The Riemann Mapping Theorem is one of the most profound results in complex analysis; in the case of one dimension, it establishes sufficient conditions for the biholomorphic equivalence between two open subsets of the complex plane.

If there exists a biholomorphism 𝑓 between two regions, then the two regions are said to be conformally equivalent, holomorphically equivalent, or biholomorphically equivalent. As a required intermediate for the proof, we first introduce:

Definition 5.3.1 (Holomorphic Logarithms).

Suppose Φ:𝑈→ℂ∗=ℂ∖{0} is holomorphic, where 𝑈 is simply connected. Define the holomorphic logarithm of Φ(𝑧) to be a branch of

log(Φ(𝑧))=∫𝛾Φ′(𝜁)Φ(𝜁)d𝜁+log(Φ(𝑧0))

for any 𝑧0∈𝑈, where the integral is path-independent 𝛾⊂𝑈 is any piecewise 𝐶1 curve from 𝑧0 to 𝑧.

Similarly, define the holomorphic powers of Φ(𝑧) to be branches of (Φ𝛼)(𝑧)=e𝛼log(Φ(𝑧)), where log(Φ(𝑧)) is the holomorphic logarithm.

The path independence of the definition is provided by the simple connectivity of 𝑈. The result is the heuristic concatenation of several different branches of the complex logarithm, unique up to an additive factor of 2πi𝑘, where this additive factor is the same throughout.

Theorem 5.3.2 (Riemann Mapping Theorem).

Let 𝑈⊂ℂ (a proper subset, in other words, 𝑈≠ℂ) be a simply connected (nonempty) open region. Let 𝑧0∈𝑈 be arbitrary. Then there exists a unique biholomorphism 𝑓:𝑈→𝔻 such that 𝑓(𝑧0)=0 and 𝑓′(𝑧0)∈ℝ>0.
Proof.

First consider the case for when 𝑈 is a bounded region. In other words, ∃𝑅>0 such that 𝑈⊆𝐷(0,𝑅).

Define ℱ︀ to be the family of all univalent functions 𝛼:𝑈→𝔻 (not necessarily surjective) such that 𝛼(𝑧0)=0. This family is well-defined and nonempty.

To prove this assertion, observe that since 𝑧0∈𝐷(0,𝑅), it follows that ∀𝑧∈𝑈⊆𝐷(0,𝑅), |𝑧−𝑧0|<2𝑅, and consequently, |𝑧−𝑧02𝑅|<1. Therefore,

𝛼(𝑧)=𝑧−𝑧02𝑅

maps 𝑈 to 𝔻, and it is complex affine and univalent. This shows that 𝛼∈ℱ︀. It is easy to prove that ℱ︀ is infinite; any function in the form of 𝑧↦𝑧−𝑧0𝜁 for 𝜁≥2𝑅 also lies in ℱ︀.

Since ℱ︀ is uniformly bounded on 𝑈, by Montel’s Theorem (Theorem 5.2.6Theorem 5.2.6Theorem 5.2.6), ℱ︀ is a normal family. Let 𝑟>0 satisfy 𝐷(𝑧0,𝑟)‾⊂𝑈. Then by Cauchy’s Estimate (Theorem 3.2.3Theorem 3.2.3Theorem 3.2.3), ∀𝛼∈ℱ︀, |𝛼′|≤1𝑟 on 𝐷(𝑧0,𝑟)‾. Hence, we have

0<𝑀=sup𝛼∈ℱ︀|𝛼′(𝑧0)|≤1𝑟,(5.3.1)

where we can assure that 𝑀 is positive since each 𝛼∈ℱ︀ is univalent at 𝑧0 and by Lemma 5.1.2Lemma 5.1.2Lemma 5.1.2.

If 𝑀 is an accumulation point of {|𝛼′(𝑧0)|}𝛼∈ℱ︀, there exists a sequence {𝛼𝑛}𝑛∈ℕ⊆ℱ︀ such that {|𝛼𝑛′(𝑧0)|}𝑛∈ℕ converges to 𝑀. If 𝑀 is attained as a maximum or that |𝛼′(𝑧0)|=𝑀 for some 𝛼∈ℱ︀, we may let each 𝛼𝑛≡𝛼.

By the normality of ℱ︀, there exists a subsequence {𝛼𝑛𝑘(𝑧)}𝑘∈ℕ⊆{𝛼𝑛(𝑧)}𝑛∈ℕ such that {𝛼𝑛𝑘(𝑧)}𝑘∈ℕ is locally uniformly convergent in 𝑈 to a function 𝛼̃̃(𝑧) (holomorphy of which is provided by Theorem 4.1.1Theorem 4.1.1Theorem 4.1.1). By definition, |𝛼̃̃′(𝑧0)|=𝑀, and define a function sequence with 𝛼̃𝑛𝑘=𝛼𝑛𝑘|𝛼̃̃′(𝑧0)|𝛼̃̃′(𝑧0)∈ℱ︀, whose locally uniform limit is 𝑓. It follows that 𝑓 is a rotation of 𝛼̃̃ such that 𝑓′(𝑧0)=𝑀.

Let 𝜁1,𝜁2∈𝑈 be arbitrary and different. Choose 𝑟′>0 to satisfy 0<𝑟′<|𝜁1−𝜁2|, and let 𝜓𝑘(𝑧)=𝛼̃𝑛𝑘(𝑧)−𝛼̃𝑛𝑘(𝜁2). Since each 𝛼̃𝑛𝑘 is univalent in 𝑈, it follows that each 𝜓𝑘 is non-vanishing in 𝑈∖{𝜁2} and consequently, in 𝐷(𝜁1,𝑟′)‾. By Theorem 3.3.5Theorem 3.3.5Theorem 3.3.5, it follows that the locally uniform limit of 𝜓𝑘, or 𝜓=𝑓(𝑧)−𝑓(𝜁2), is either non-vanishing or is identically zero in 𝐷(𝜁1,𝑟′)‾. The latter is an impossibility since 𝜓′(𝑧0)=𝑀>0. Hence, 𝑓(𝑧)=𝑓(𝜁2) has no solutions for 𝑧∈𝐷(𝜁1,𝑟′)‾. In particular, 𝑓(𝜁1)≠𝑓(𝜁2). By the arbitrariness of 𝜁1 and 𝜁2, the univalence of 𝑓 follows.

Additionally, since ∀𝑘∈ℕ, |𝛼̃𝑛𝑘|<1 in 𝑈, it follows that 𝑓(𝑈)⊆𝔻‾. By the Open Mapping Theorem (Theorem 5.1.1Theorem 5.1.1Theorem 5.1.1), the condition becomes 𝑓(𝑈)⊆𝔻. Since 𝛼̃𝑛𝑘(𝑧0)=0 for all 𝑘∈ℕ and 𝛼̃𝑛𝑘(𝑧0)→0=𝑓(𝑧0), it follows that 𝑓∈ℱ︀.

Lastly, we aim to prove that 𝑓 maps 𝑈 to 𝔻 surjectively. For the sake of contradiction, assume that ∃𝜉∈𝔻∗ such that 𝜉∉𝑓(𝑈). Consider the unit disk automorphism 𝜑𝜉(𝑧)=𝑧−𝜉1−𝑧𝜉‾. Since 𝜑𝜉(𝑧) vanishes when 𝑧=𝜉, and since 𝑓(𝑧)=𝜉 has no solutions in 𝑈, there exists a holomorphic square root

𝜇(𝑧)=𝜑𝜉∘𝑓(𝑧)∈𝔻

for 𝑧∈𝑈. Let 𝜏=𝜇(𝑧0)=−𝜉, and let

𝜂(𝑧)=𝜑𝜏∘𝜇(𝑧),

where 𝜑𝜏=𝑧−𝜏1−𝑧𝜏‾. Since 𝜂(𝑧0)=𝜑𝜏(𝜏)=0, it follows that 𝜂∈ℱ︀. Let 𝜂̃=|𝜂′(𝑧0)|𝜂′(𝑧0)𝜂, which is also in ℱ︀. However, since 𝜂̃′=|𝜂′(𝑧0)|𝜂′(𝑧0)𝜂′, we have

𝜂̃′(𝑧0)=|𝑓′(𝑧0)(𝜑𝜏′∘𝜏)(𝜑𝜉′∘0)2𝜑𝜉∘0|=𝑀2|𝜉|1−𝜏‾𝜏(1−𝜏𝜏‾)21−𝜉‾𝜉(1−0𝜉‾)2=𝑀2|𝜉|1−|𝜉|21−|𝜉|=𝑀(1+|𝜉|)2|𝜉|.

Additionally, since

(|𝜉|−1)2>0⟺1+|𝜉|>2|𝜉|⟺1+|𝜉|2|𝜉|>1,

it follows that 𝜂̃′(𝑧0)>𝑀, which is a contradiction of (5.3.1)(5.3.1)(5.3.1).

Hence, 𝑓:𝑈→𝔻 is biholomorphic. To prove the uniqueness of 𝑓, suppose 𝑔:𝑈→𝔻 is an arbitrary biholomorphism such that 𝑔(𝑧0)=0 and 𝑔′(𝑧0)>0. Then, 𝜑=𝑓∘𝑔−1∈Aut(𝔻), and by Theorem 3.5.3Theorem 3.5.3Theorem 3.5.3, 𝜑(𝑧)=𝜑𝑎(𝑧ei𝜃) for some 𝑎∈𝔻 and 0≤𝜃<2π. Since 𝜑(0)=0, it follows that 𝑎=0. Since 𝜑′(0)=𝑓′(𝑧0)(𝑔−1)′(0)=𝑓′𝑧0𝑔′(𝑧0)>0, and 𝜑′(0)=𝜑0′(0)ei𝜃=ei𝜃>0, it follows that 𝜃=0. Hence, we have 𝜑(𝑧)=𝑧 and 𝑓≡𝑔.

Next, assume that 𝑈 is unbounded. It is easy to show that the boundary 𝜕𝑈 contains at least two points. Indeed, if 𝜕𝑈=⌀, 𝑈 would be closed because 𝜕𝑈⊆𝑈 and open by assumption. By Theorem 3.2.1.19Theorem 3.2.1.19Theorem 3.2.1.19, 𝑈 would either be equal to ⌀ or ℂ, both of which are impossibilities. Additionally, if 𝜕𝑈 comprises exactly one point 𝑎∈ℂ, then in subspace defined by 𝑋=ℂ∖{𝑎}, 𝑈 is clopen (by the same reason as before, open by assumption and closed because 𝑋∖𝑈=ℂ∖𝑈‾ is open). It follows that 𝑈=𝑋=ℂ∖{𝑎}, which is not simply connected.

Suppose 𝜉1 and 𝜉2 are two distinct points in 𝜕𝑈. Let us apply the linear transformation 𝜌(𝑧)=𝑧−𝜉1𝜉2−𝜉1 to 𝑈, and denote the resulting region by 𝑈′=𝜌(𝑈). It follows that 0,1∈𝜕𝑈′. Consider a branch 𝜓(𝑧) of the holomorphic square root 𝑧↦𝑧−1 (existent by the holomorphic logarithm from simple connectivity and the fact that 1∉𝑈′). Trivially, 𝜓 is univalent in 𝑈′.

In addition, we assert that 𝜓(𝑈′)∩(−𝜓(𝑈′))=⌀. If not, then ∃𝜉∈𝜓(𝑈′) such that −𝜉∈𝜓(𝑈′). By definition, ∃𝑧1,𝑧2∈𝑈′ such that 𝜓(𝑧1)=𝜉 and 𝜓(𝑧2)=−𝜉. It would then follow that 𝑧1−1=−𝑧2−1 and 𝑧1=𝑧2. It follows that 𝜉=0, which is obtained when 𝑧1=𝑧2=𝜓−1(𝜉)=1. Since 1∈𝜕𝑈′ and 𝑈′ is open, 1∉𝑈′, which is an impossibility.

Fix 𝜉∈𝜓(𝑈′) to be arbitrary. By the Open Mapping Theorem (Theorem 5.1.1Theorem 5.1.1Theorem 5.1.1), there exists an open neighborhood 𝐷(𝜉,𝜀)⊆𝜓(𝑈′). It follows that 𝐷(−𝜉,𝜀)∩𝜓(𝑈′)=⌀. Therefore, ∀𝑧∈𝑈′, |𝜓(𝑧)+𝜉|≥𝜀, and consequently, |1𝜓(𝑧)+𝜉|≤1𝜀. Hence, the function 𝜑(𝑧)=1𝑧+𝜉 maps 𝑈′ to a bounded region that lies within the compact disk 𝐷(0,1𝜀)‾. Denote 𝜑∘𝜓(𝑈′) by 𝑈̃.

It is easy to see that 𝑈̃ is simply connected. Let 𝑈̃=𝜑∘𝜓∘𝜌(𝑈). To prove this, it suffices to show that the line integral of any holomorphic function over any closed curve in 𝑈̃ vanishes. Let 𝑔:𝑈̃→ℂ be holomorphic, and let Γ⊂𝑈̃ be a closed piecewise 𝐶1 curve. Then

∮Γ𝑔(𝑤)d𝑤=∮𝜌−1∘𝜓−1∘𝜑−1(Γ)𝑔∘𝜑∘𝜓∘𝜌(𝑧)⋅(𝜑∘𝜓∘𝜌)′(𝑧)d𝑧=0,

by Cauchy–Goursat (Theorem 3.1.9Theorem 3.1.9Theorem 3.1.9), since 𝑈 is simply connected by assumption, 𝜌−1∘𝜓−1∘𝜑−1(Γ) is a closed piecewise 𝐶1 curve in 𝑈, and the integrand is holomorphic on 𝑈. Therefore, 𝑈̃ is simply connected.

Hence, we may use our previous result and establish a biholomorphism 𝑓̃:𝑈̃→𝔻, unique up to a transformation in Aut(𝔻). Let 𝑓=𝑓̃∘𝜑∘𝜓∘𝜌, which is a biholomorphism from 𝑈 to 𝔻. Similarly, it is unique up to a transformation in Aut(𝔻), and the same assertion follows. ⁠□

Remark.

It is natural that we require 𝑈≠ℂ; if there exists a univalent function 𝑓:ℂ→𝔻, then by Liouville’s Theorem (Theorem 3.2.5Theorem 3.2.5Theorem 3.2.5), 𝑓 would be a constant function.

As we will see in Section 11Section 11Section 11, this theorem and many other properties of one-variable holomorphic functions do not extend to functions of several complex variables.