What we are hunting, and why the disk deserves it
An automorphism of a region is a holomorphic map of that region onto itself that is one-to-one and has a holomorphic inverse — a symmetry of the region in the eyes of complex analysis. For the open unit disk D = { |z| < 1 } we want the whole list: every holomorphic bijection f : D to D. This guide answers that completely, and the answer is small, clean, and slightly surprising. The hero turning the crank is the Schwarz lemma you met in guide 1.
First, a warm-up that fixes the easy case. The Schwarz lemma says: if f : D to D is holomorphic and f(0) = 0, then |f(z)| is at most |z| everywhere, and equality at even one interior point forces f(z) = e^(i alpha) z, a pure rotation. So among automorphisms that pin the centre, there is nothing but rotations. The real work is what happens when the map is allowed to move the centre somewhere else — and to handle that we need a tool that can slide any chosen interior point back to 0.
The Blaschke factor: a slider for the disk
Fix a point a inside the disk, |a| < 1. The Blaschke factor for a is the Mobius transformation written below. It is built so that it carries a to 0 and 0 to a, and — the magic property — it maps the disk exactly onto itself, boundary circle to boundary circle. So each Blaschke factor is itself an automorphism of D, one that re-centres the disk at the chosen point a.
a - z
phi_a(z) = ----------- ( |a| < 1 )
1 - a-bar z
phi_a(a) = 0 phi_a(0) = a phi_a( phi_a(z) ) = z
On the boundary |z| = 1 : |phi_a(z)| = 1 (circle stays a circle)The boundary claim is worth verifying once by hand, because it is the entire reason the factor stays inside the disk. Take |z| = 1, so z z-bar = 1. Then |phi_a(z)| = |a - z| / |1 - a-bar z|. Multiply numerator and denominator of the inside by z-bar in the denominator's factor: |1 - a-bar z| = |z-bar| |z - a| ... more cleanly, |1 - a-bar z| = |z| |z-bar - a-bar| = |z-bar - a-bar| = |z - a| (the last step because a modulus is unchanged by conjugation). So the numerator and denominator have equal modulus and |phi_a(z)| = 1. By the maximum modulus principle a holomorphic map with boundary modulus 1 sends the inside to the inside, so phi_a really is a self-map of D — and since it is its own inverse, a bijection.
The classification theorem
Now the payoff. Every automorphism of the disk has the form f(z) = e^(i alpha) phi_a(z) for some real angle alpha and some a in D — a Blaschke slider followed by a rotation. Equivalently, up to a rotation factor, f(z) = e^(i alpha) (a - z)/(1 - a-bar z). The whole automorphism group of the disk is therefore parametrised by just two real numbers in a, plus one angle: three real parameters in all. A short, complete census.
The proof is a beautiful piece of judo that throws the general case onto the centre-fixing case. Suppose f is any automorphism, and let a = f^(-1)(0) be the point f sends to the origin. Form the composite g = f composed with phi_a (recall phi_a sends 0 to a, then f sends a to 0). Then g is again an automorphism of D, and now g(0) = 0. By the rotation form of the Schwarz lemma, g must be a pure rotation, g(z) = e^(i alpha) z. Unwinding the composition, f = g composed with phi_a^(-1) = g composed with phi_a (it is its own inverse), giving f(z) = e^(i alpha) phi_a(z). Done.
- Given an automorphism f, locate the point a = f^(-1)(0) that lands on the centre.
- Compose with the Blaschke factor phi_a to build g = f composed with phi_a; check g(0) = 0 and g is still an automorphism.
- Apply the Schwarz lemma's equality case to g: a centre-fixing automorphism must be a rotation, g(z) = e^(i alpha) z.
- Peel phi_a back off to recover f(z) = e^(i alpha) phi_a(z) — every automorphism is a rotation of a Blaschke factor.
Why the Schwarz lemma was strong enough to do this
Pause on what just happened, because it is the moral of the whole rung. We classified an infinite group of maps using a lemma whose entire content was a single inequality plus a tie-breaking equality case. The inequality (|f(z)| at most |z|) is never even used in its inequality form here; what does all the lifting is the rigidity in the equality case — "if you can't shrink, you must be a rotation." A bijective self-map of D cannot truly shrink, because its inverse would then have to expand, which the lemma forbids. So both f and f^(-1) live on the equality boundary, and rigidity pins them down exactly.
This is the recurring shock of complex analysis: a hypothesis that looks almost too weak (one complex derivative, or a modulus bound) pins down vastly more than its real-variable cousins ever could. A smooth real bijection of an interval onto itself has wildly many forms; the holomorphic bijections of the disk fit on a three-parameter list. Holomorphy is that rigid. It is the same rigidity behind Liouville's theorem and the identity theorem — local data straitjacketing the global map.
The same group, wearing the half-plane's clothes
The disk is not the only stage for this story. The upper half-plane H = { Im z > 0 } is conformally the very same region as the disk, glued to it by the Cayley transform w = (z - i)/(z + i), which carries H bijectively onto D. Symmetries travel along this bridge: if you conjugate a disk automorphism by the Cayley transform you get a half-plane automorphism, and vice versa. So the two automorphism groups are not merely similar — they are the same group, viewed through two different windows.
And in the half-plane's window the group wears a famous uniform: the automorphisms of the half-plane are exactly the real Mobius maps z mapsto (a z + b)/(c z + d) with a, b, c, d real and ad - bc > 0. That is the group usually written PSL(2, R) — the same object that runs number theory's modular forms and the dynamics of hyperbolic surfaces. We arrived at this rich group from nothing but a self-map-of-the-disk lemma. Keep this dictionary handy: the next guides measure distances with these maps, and you will often find a computation easier on whichever side, disk or half-plane, the fixed points sit more conveniently.