An existence theorem with precise boundaries
The explicit maps for half-planes and sectors suggest a larger question: can every simply connected proper plane domain be mapped conformally onto the disk? An answer must supply a bijection, not just a holomorphic map with nonzero derivative. It must also say what is assumed about the boundary. The theorem below concerns the open domains; it does not assert that the map extends continuously or injectively to their boundaries.
Theorem 3.3.1
(Riemann mapping theorem)
Let be a nonempty simply connected domain, and let . There is a unique biholomorphic map such that and .
We use the open mapping and maximum-modulus theorems, Cauchy estimates, locally uniform convergence of holomorphic functions, Montel and Rouché, and Schwarz's lemma. Here simply connected means every closed curve is null-homotopic in the domain; we also use its standard consequence that a zero-free holomorphic function has a logarithm. An injective holomorphic map has nonzero derivative: otherwise its local Taylor expansion and Rouché's theorem give more than one nearby preimage, as proved in the note on injective entire maps. The proof occupies the next four sections and has three distinct tasks: construct at least one bounded injective map, obtain an injective extremal limit, and prove that the limit fills the disk; uniqueness follows from Schwarz's lemma.
Construct a bounded injective map
Lemma 3.3.2
(A normalized injective map into the disk exists)
The family of injective holomorphic maps with and is nonempty.
Proof
(A square root separates the image from its reflection)
Choose . The function is holomorphic and zero-free on , so it has a holomorphic logarithm . Put . Then , is nowhere zero, and is injective: implies .
Let . By the open mapping theorem, contains some disk . The sets and are disjoint: if , squaring gives , which would force , impossible. Hence avoids , and
is holomorphic, injective and bounded in modulus by . Multiply by a sufficiently small positive constant to place its image strictly inside , then compose with the disk automorphism , where is the rescaled value of . The result is an injective holomorphic map with . Its derivative at is nonzero by the injective-map lemma; multiplication by a rotation makes it positive.
This step is where properness is used: without a point outside the domain, the square-root construction has no starting point.
Counterexample mode
The whole plane is not conformally a disk
Claim: every simply connected domain is biholomorphic to . The plane is simply connected. If were biholomorphic, would be an entire function with , hence constant by Liouville's theorem, contradicting injectivity. The repaired statement requires , which is exactly what supplies the point above.
Worked example 3.3.3
(The square-root step for a slit plane)
Let , and . The principal square root satisfies and maps onto the right half-plane ; then is the left half-plane, disjoint from it, as the proof predicts. Here , and lies in the right half-plane, so works. Indeed , so maps injectively into with . Composing with gives a map sending to ; its derivative at is , and multiplying by makes it positive.
Take an extremal limit
If , Cauchy's estimate gives for every , since . Thus
Lemma 3.3.4
(The supremum is attained by an injective map)
There is with .
Proof
(Montel for a limit, Rouché for injectivity)
Choose with . Montel's theorem gives a locally uniformly convergent subsequence; write its limit as . Cauchy's integral formula on a small circle gives along that subsequence, so and is nonconstant. Pointwise convergence gives . Since , the maximum-modulus principle makes everywhere.
Suppose and . Choose disjoint closed disks around inside whose boundaries contain no zeros of ; this is possible because is not identically zero, so its zeros are isolated. The continuous function has a positive minimum on each boundary. For large , uniform convergence gives on both boundaries, and Rouché's theorem gives a zero of in each disk. That contradicts injectivity of . Thus and attains the supremum.
Proof X-ray
Why the limit remains a candidate
Three properties of the limit must be recovered, and each uses a different tool. Holomorphy and the value come from locally uniform convergence. Membership in the open disk needs nonconstancy, which is where the extremal normalization pays off: rules out a constant limit, after which maximum modulus excludes at interior points. Injectivity is the non-obvious move. Rouché's theorem makes zeros of stable under small perturbation, so two distinct preimages of for the limit would persist as two preimages for . The two disks are disjoint, so this contradicts injectivity of . The upper bound from the Cauchy estimate is what makes the maximizing sequence normal in the first place.
Common mistake
(Assuming a limit of injective maps is injective)
A locally uniform limit of injective holomorphic maps may be constant: on is injective for each and converges to . The Rouché argument proves only that the limit is injective or constant. In the proof the constant alternative is excluded by ; without that normalization the extremal limit could fail to belong to .
A missing point would improve the derivative
Lemma 3.3.5
(The extremal map is onto)
If satisfies , then , and is a biholomorphism.
Proof
(A square root and an automorphism enlarge the derivative)
Suppose omits . Since , necessarily . The disk automorphism sends to zero. Therefore is zero-free on the simply connected domain and has a holomorphic square root :
The square root is injective: equality of two values implies equality of their squares, so , and injectivity of and gives . The map maps injectively into and sends to zero. Differentiating the square relation at and using gives
Since , the derivative of at has modulus
The strict inequality follows from . A rotation makes the derivative positive, producing a member of with derivative larger than , a contradiction. Thus the image is all of . An injective holomorphic map has nonzero derivative, so its inverse is holomorphic: is a biholomorphism.
Proof X-ray
Why a square root gains derivative
Simple connectivity is used here for the second time, to take a square root of the zero-free function ; the omitted value is exactly what makes it zero-free. The explicit computation hides a cleaner reason. Undo the construction: with . This maps into , fixes because , and is two-to-one, so it is not a rotation. Schwarz's lemma therefore gives , and the chain rule forces the new map to have the larger derivative. The key move is thus to factor the extremal map through a non-injective self-map of the disk, which is possible only because misses a value. The contradiction with maximality of closes the argument; as the gain factor tends to , but it is always strict.
Uniqueness and limitations
Proof
(Uniqueness of the normalized map)
If are two normalized biholomorphisms, then is a disk automorphism fixing zero. Schwarz's lemma applied to and makes a rotation. Its derivative at zero is the positive real number ; the only positive unit complex number is , so is the identity and .
A conformal map need not be an elementary formula. Nor does this existence proof settle boundary values: extending a map to the boundary requires additional hypotheses and separate results, such as Carathéodory's theorem under its own boundary hypotheses, which is not proved here.
Worked examples: normalized maps
The theorem determines a map uniquely once we choose its base point and require a positive derivative there.
Worked example 3.3.6
(Normalize at any point of the upper half-plane)
Let and , where . Consider
The reflected point lies outside , so the denominator does not vanish in the domain. For ,
Thus . Moreover, for , solving gives
The denominator is nonzero because , and direct calculation yields
This inverse proves that the image is the whole disk, not merely a subset. Finally and
The factor matters: also maps bijectively onto the disk and sends to zero, but its derivative at is , not a positive real number. Uniqueness therefore identifies as the normalized map. At it becomes , with derivative at .
Worked example 3.3.7
(When the domain is already the disk)
Let and . The automorphism is a biholomorphism of with and
So is the normalized Riemann map, and . For it is the identity with , which is Schwarz's lemma in extremal form: no injective self-map of the disk fixing has derivative larger than there.
Practice and solutions
Quick check
(Why is the point necessary in the initial bounded-map construction?)
What does need in order to have a square root?
Solution · Answer
It makes zero-free on , allowing a holomorphic logarithm and hence a square root on the simply connected domain. For no such point exists, and no biholomorphism onto exists either.
Quick check
(Why is the extremal limit not allowed to be constant?)
Compare derivatives at .
Solution · Answer
Cauchy's formula gives , whereas a constant has derivative zero.
Quick check
(If a normalized extremal map omits , why can not be zero? For , by what factor does the square-root construction enlarge the derivative?)
Evaluate the gain factor.
Solution · Answer
The map already takes the value zero at . For , the factor is .
Quick check
(Suppose two biholomorphisms both send to zero but no derivative normalization is imposed. How are they related?)
Compose one with the inverse of the other.
Solution · Answer
Their composition is a disk automorphism fixing zero, hence a rotation by Schwarz's lemma. Thus for some real .