The statement, in one breath
You have spent the last two guides building a machine you did not yet know the purpose of. You learned that a locally uniform limit of holomorphic functions stays holomorphic, and that a normal family — one where every sequence has a locally uniformly convergent subsequence — is exactly what Montel's theorem hands you for free whenever the family is locally bounded. That is a compactness engine. This guide tells you the magnificent thing it was built to power.
Here is the theorem. Take any simply connected open region D in the plane — a region with no holes — and suppose D is not the entire plane C. Then there exists a single conformal map carrying D bijectively onto the open unit disk: a map f that is holomorphic, one-to-one, onto, with a holomorphic inverse. A square, the inside of a heart-shaped curve, an infinite strip, the plane with a slit cut out of it — every one of them is, conformally, the very same disk.
Pause on how strong that is. The shapes look nothing alike — one is jagged, one is unbounded, one has a wound cut into it — yet a single holomorphic bijection irons each onto the round disk while preserving every angle. The boundaries can be wildly different lengths and roughnesses; the theorem does not care. The only invariants that survive a conformal map are angles and the topological fact of having no holes, and the Riemann mapping theorem says those are the only things that ever mattered.
Why this is the dream you have wanted all along
Way back in the conformal-mapping rung you met the strategy of transplanting a problem: a question about a hard region becomes easy if you can conformally carry it to the unit disk, solve it there with the disk's clean toolkit, and carry the answer home. The catch was always the same nagging worry — does such a carrying map even exist for MY region? You could write Möbius maps for half-planes and a Cayley transform for a few special cases, but for a genuinely awkward shape you had nothing.
The Riemann mapping theorem dissolves the worry completely. As long as your region has no holes and is not the whole plane, the carrying map is guaranteed to exist — you do not have to find it to know it is there. Every Dirichlet problem, every steady heat distribution, every ideal fluid flow posed on a simply connected region can in principle be pulled back to the disk, where the Poisson integral solves it outright, then pushed forward. The theorem is the formal license that makes the entire transplant philosophy legitimate rather than a lucky trick for special domains.
Three fences you may not climb over
The hypotheses are not decoration; each one is load-bearing, and removing any single one breaks the theorem. The first fence is no holes — D must be simply connected. An annulus, the ring between two circles, has a hole, and it is genuinely not conformally a disk. The cleanest way to see this: a conformal map onto the disk would let you build, on the annulus, a function that the hole forbids. In fact two annuli are conformally equivalent only when their inner-to-outer radius ratios agree, so the annuli form a whole continuum of distinct shapes — proof that holes carry real, unremovable information.
The second fence is not the whole plane. The plane C itself is simply connected, so you might hope it slips in — but it is firmly excluded, and Liouville's theorem is the reason. A conformal map of C onto the disk would be a bounded, non-constant entire function (its values all sit inside the disk, hence |f| < 1 everywhere). Liouville says no such function exists: a bounded entire function must be constant. So the whole plane stubbornly refuses to be conformally a disk, and the exclusion 'D is not all of C' is not fussiness but a hard wall.
The third fence is the quiet one: the conclusion is about INTERIORS only. The map is a bijection between the open region and the open disk, and says nothing on its own about the boundary — whether a point on the jagged edge of your region matches up cleanly with a point on the unit circle. For many applications that boundary correspondence is exactly what you need, and it can fail when the boundary is wild. Guide 5 of this rung takes up boundary behaviour in earnest; for now, just file away that 'D maps onto the disk' is a statement about the insides.
Pinning down THE map: uniqueness by one normalization
Strictly speaking there is not one Riemann map but infinitely many, because once you have a conformal map onto the disk you can follow it with any rotation or any disk automorphism and get another perfectly good one. So how do we ever speak of 'the' Riemann map? The answer is beautifully tidy: the leftover freedom is exactly the freedom of the disk's own symmetries, no more, and you can spend all of it with one small normalization.
- Choose a basepoint z_0 anywhere inside D — a single interior point you care about, often the place you want to be the 'centre' of the disk picture.
- Demand f(z_0) = 0: the chosen point must land at the centre of the disk. That uses up two real degrees of freedom (a complex value is two real numbers).
- Demand f'(z_0) > 0, a positive real number: this fixes which way the map is rotated at z_0, spending the one remaining real degree of freedom.
- With those three real conditions imposed, the map is now unique — there is exactly one conformal f from D onto the disk meeting them.
The arithmetic is the punchline. The disk's automorphism group is exactly three real dimensions wide — two to move the centre anywhere, one to rotate — and our normalization imposes exactly three real conditions. They match perfectly, so they pin the map down with no slack and no over-constraint. The proof that this really forces uniqueness is a one-line Schwarz-lemma argument: if f and g both satisfy the normalization, then g composed with f-inverse is a disk automorphism fixing 0 with positive derivative there, and the Schwarz lemma you proved earlier in this chapter forces such a map to be the identity, so g = f.
A glance ahead at how it gets proved
It is fair to wonder how anyone manufactures a conformal map onto the disk when there is no formula to write. The next guide tells the full story, but here is the shape of it, so the existence claim does not feel like magic. You gather the family F of ALL injective holomorphic maps from D into the disk that send z_0 to 0, and you ask which member spreads D out most aggressively at z_0 — that is, which one maximizes |f'(z_0)|. The claim is that this greediest map is forced to be onto, and an onto, injective, holomorphic map to the disk is precisely a Riemann map.
This is where the engine from guides 1 and 2 finally fires. The family F is bounded by 1, hence locally bounded, hence NORMAL by Montel — so a sequence of maps whose derivatives climb toward the maximum has a locally uniformly convergent subsequence. Its limit is holomorphic (Weierstrass), it still has the maximal derivative, and it is still injective (Hurwitz keeps a limit of injective maps injective). The only delicate step is showing the maximizer is surjective, and that uses the Schwarz lemma to manufacture a contradiction if it weren't. Compactness gives existence; the extremal property forces every other quality you need.