From one-to-one to a finite area
In the previous guide you met the class S of univalent functions on the unit disk, normalized so that f(0) = 0 and f'(0) = 1, with Taylor series f(z) = z + a_2 z^2 + a_3 z^3 + .... The whole subject hangs on one deceptively simple constraint: f is injective, it never takes the same value twice. The art of geometric function theory is turning that single word 'one-to-one' into hard numbers about the coefficients a_n. This guide makes the very first such conversion, and the lever that does it is something you already trust completely — area cannot be negative.
Here is the picture. A univalent map sends the disk to some region with no overlaps — the image folds onto itself nowhere, so it has a well-defined area, and as the disk fills out, that area is a genuine, finite number. The trick is to make this area visible as a sum over the coefficients. To do that cleanly we do NOT work with f directly; instead we step OUTSIDE the disk and look at the complementary region, where the geometry of 'no overlap' turns into a single clean inequality. That detour is the heart of the area theorem.
The area theorem
Step outside the disk. Consider the class Sigma of functions g that are univalent on the OUTSIDE of the unit disk (the region |z| > 1, including the point at infinity) and fixed up there so that near infinity g looks like the identity with a tail: g(z) = z + b_0 + b_1/z + b_2/z^2 + .... This is just a Laurent series about infinity. Because g is one-to-one, its image misses a compact 'hole' E — the bounded set the map carves out of the plane. That hole has some area, and area is at least zero. The area theorem turns that single fact into a constraint on the b_n.
Computing the area of that hole by Green's theorem — integrate around the image of the circle |z| = r and let r shrink to 1 — collapses to a beautifully simple result. The area of the omitted set E equals pi times (1 minus the sum of n |b_n|^2 over n = 1, 2, 3, ...). Since that area cannot be negative, the bracket cannot be negative, and we land the area theorem: the sum of n |b_n|^2 is at most 1. Every single coefficient is squeezed by all the others sharing one unit of weight — an honest-to-goodness budget, paid for entirely by the geometry of not overlapping.
g(z) = z + b_0 + b_1/z + b_2/z^2 + ... (univalent on |z| > 1) Area of the omitted hole E = pi * ( 1 - sum over n>=1 of n |b_n|^2 ) >= 0 ==> sum over n>=1 of n |b_n|^2 <= 1 (the area theorem) In particular |b_1| <= 1 , with equality only when all other b_n = 0.
Read off the cheapest consequence first: taking just the n = 1 term gives |b_1| at most 1, and equality forces every other b_n to vanish. That extreme case is a map that omits the largest possible hole — a full line segment, the most spread-out shape of unit weight. Keep that equality case in mind; the entire flavor of this rung is that the SHARP bound is achieved by one special, maximally-stretched map, and we are about to meet its cousin on the inside of the disk.
Crossing back inside: the bound |a_2| <= 2
The area theorem lives outside the disk, but class S lives inside. We need a bridge. The bridge is a clever algebraic substitution that turns an inside-univalent f into an outside-univalent g, so that the budget on g's coefficients becomes a budget on f's. The standard one is the square-root transform: from f(z) = z + a_2 z^2 + ... build a new map by taking a single-valued branch of the square root of f(z^2)/z^2 and inverting. It is built precisely so the FIRST exterior coefficient b_1 turns out to equal -a_2/2.
Why is the square root even allowed? Because f(z)/z = 1 + a_2 z + ... is never zero on the disk (f vanishes only at the origin, and there z divides it out), so it has a well-defined holomorphic square root once you pick the branch with value 1 at the origin — no branch cut crosses the disk. This is the multivaluedness caveat handled correctly: we are NOT taking a careless square root of an arbitrary function; we checked the function avoids zero and chose the branch deliberately. With g constructed, the area theorem's |b_1| at most 1 reads directly as |a_2/2| at most 1.
Multiply through and you have the Bieberbach bound: for every f in S, |a_2| is at most 2. This is the founding theorem of the whole field — the first member of an infinite list of coefficient bounds — and it dropped straight out of 'area is non-negative,' translated across the bridge. Equality |a_2| = 2 forces b_1 to have modulus 1, which we just saw forces every other exterior coefficient to vanish, which pins f down to exactly one shape. That shape is the star of the next section.
The Koebe function: the extremal shape
Meet the function that achieves |a_2| = 2: the Koebe function k(z) = z/(1 - z)^2. Expand the geometric tail and it reads k(z) = z + 2 z^2 + 3 z^3 + 4 z^4 + ..., so its n-th coefficient is exactly n. In particular a_2 = 2 — it sits dead on the Bieberbach bound. The Koebe function is to this subject what the Koebe-the-extremal-example is to every later guide: the single map against which all sharpness is measured. Whenever a coefficient bound in S is tight, it is the Koebe function (or a rotation of it) that makes it tight.
What does it actually look like? The Koebe function maps the unit disk onto the entire plane with a single slit removed — the ray along the negative real axis from -1/4 all the way out to infinity. Picture taking the round disk, grabbing the boundary point z = -1, and stretching it outward forever while pinning the rest: the image is the whole plane minus that one cut. The image is enormous, yet it deliberately refuses to come any closer to the origin along the negative axis than the point -1/4. That number -1/4 is not a coincidence; it is the punchline of this whole guide.
The one-quarter theorem
Now the headline. The Koebe one-quarter theorem says that for EVERY f in S — not just the Koebe function — the image f(disk) contains the whole open disk of radius 1/4 centered at the origin. No matter how the univalent map stretches, twists, or slits the disk, its image cannot avoid any point within distance 1/4 of 0. The normalization f'(0) = 1 sets a scale, and one-to-one-ness forbids the image from squeezing too tightly near the origin: it must leave room for a disk of guaranteed radius 1/4.
The proof is a gorgeous little argument that rides entirely on the bound we just earned. Suppose f omits some value c — that is, f(z) never equals c on the disk. Then a single Mobius adjustment of f, namely the map c f(z)/(c - f(z)), is again normalized and STILL univalent, so it too belongs to S, so ITS second coefficient is also bounded by 2 in modulus. Spelling out that new second coefficient gives a_2 + 1/c, and demanding both |a_2| at most 2 and |a_2 + 1/c| at most 2 forces |1/c| at most 4, that is |c| at least 1/4. Every omitted value lies at distance at least 1/4 from the origin — so the disk of radius 1/4 is never omitted.
- Suppose f in S omits the value c (so c is NOT in the image). The goal is to prove |c| >= 1/4.
- Form h(z) = c f(z) / (c - f(z)). Since f never hits c the denominator never vanishes, and a Mobius map of a univalent function stays univalent; check h(0) = 0 and h'(0) = 1, so h is in S.
- Expand h to second order: its second coefficient comes out to a_2 + 1/c. Apply the Bieberbach bound to BOTH f and h, giving |a_2| <= 2 and |a_2 + 1/c| <= 2.
- Add the two bounds with the triangle inequality: |1/c| = |(a_2 + 1/c) - a_2| <= 2 + 2 = 4, so |c| >= 1/4. Every omitted value is at distance at least 1/4.
And the constant 1/4 is the best possible, because the Koebe function actually omits the point -1/4 (the tip of its slit) and nothing closer. So the theorem is SHARP: you cannot improve 1/4 to anything larger, since one map in S sits exactly on the edge. This is the recurring shape of the whole rung — a clean universal bound for all of S, with the Koebe function standing alone at the boundary to prove it cannot be tightened. Notice too where the Schwarz lemma sits relative to all this: Schwarz controlled maps that FIX the origin and stay inside the disk, while here univalence plus a fixed derivative controls how far the image must REACH. Both are rigidity statements — one constraint, forced consequences — and the next guide will sharpen this one into the full growth and distortion theorems.