Skip to content
The book

Conformal maps, Möbius transformations, Schwarz lemma

Why this is asked: Know the dictionary of standard maps between disc, half-plane, strip and sector, and that Möbius maps send circles-and-lines to circles-and-lines and preserve cross-ratio.

In one sentence

Conformal means angle-preserving, which for holomorphic maps is exactly f′≠0f' \ne 0; Möbius transformations are the maps of the sphere to itself and send circles-or-lines to circles-or-lines; and Schwarz's lemma pins down every map of the disc to itself.

Why the exam asks it

Because the geometry is concrete enough to be examined without heavy machinery, and because two facts get misremembered in predictable directions: that a line is a kind of circle once ∞\infty is included, and that Schwarz's lemma needs a fixed point at the origin.

The idea

Conformal maps

ff is conformal at aa if it preserves angles between curves through aa, including orientation. For holomorphic ff this happens exactly where f′(a)≠0f'(a) \ne 0 — the derivative acts as multiplication by a complex number, which rotates by arg⁡f′(a)\arg f'(a) and scales by ∣f′(a)∣|f'(a)|, and both operations leave angles alone.

Where f′(a)=0f'(a) = 0 conformality fails in a specific way: if f(z)−f(a)f(z) - f(a) vanishes to order mm, angles at aa are multiplied by mm. At z↦z2z \mapsto z^2 the origin sends a right angle to a straight one.

z‾\overline{z} preserves the magnitude of angles and reverses their sense; it is anticonformal, and not holomorphic. Options use it to test whether "preserves angles" was read as including orientation.

Möbius transformations

T(z)=az+bcz+d,ad−bc≠0.T(z) = \frac{az+b}{cz+d}, \qquad ad - bc \ne 0.

The condition ad−bc≠0ad-bc\ne0 is exactly what stops TT collapsing to a constant, and T′(z)=ad−bc(cz+d)2T'(z) = \dfrac{ad-bc}{(cz+d)^2}, so TT is conformal wherever it is defined.

Viewed on the Riemann sphere C∪{∞}\mathbb{C}\cup\{\infty\}, TT is a bijection of the sphere: T(−d/c)=∞T(-d/c) = \infty and T(∞)=a/cT(\infty) = a/c. Composition corresponds to matrix multiplication of (abcd)\begin{pmatrix}a&b\\c&d\end{pmatrix}, so the Möbius group is PGL2(C)PGL_2(\mathbb{C}) — the matrix is determined up to a scalar, which is why 2×22\times2 determinant facts transfer directly.

generator does fixes ∞? z + b translation yes az rotation and scaling yes 1 / z inversion no — swaps 0 and ∞ Every Möbius map is a composition of these, and each sends circles-or-lines to circles-or-lines. So the family is closed under that, and a line is just a circle through ∞ — which is why the two cases are one case on the sphere, and why 1/z can turn a line into a circle.
Every Möbius map is built from translation, rotation-scaling and inversion. Each sends circles-or-lines to circles-or-lines, so the whole family does.

Circles to circles. Every Möbius map is a composition of translations, rotation-scalings and the single inversion 1/z1/z; each of those sends circles-or-lines to circles-or-lines, so all of them do. On the sphere this is one statement, because a line is precisely a circle through ∞\infty. That is why 1/z1/z can turn a line into a circle: the line passed through ∞\infty, the image does not.

Three points determine the map. Given distinct z1,z2,z3z_1,z_2,z_3 and w1,w2,w3w_1,w_2,w_3, there is exactly one Möbius map sending each zkz_k to wkw_k. The cross-ratio is the invariant behind this, and it is preserved by every Möbius transformation.

Two maps worth knowing by name:

  • Cayley, z−iz+i\dfrac{z-i}{z+i}: upper half plane onto the unit disc. It sends 0↦−10 \mapsto -1, 1↦−i1 \mapsto -i, ∞↦1\infty \mapsto 1 — the real line onto the circle, as required.
  • Disc automorphisms: every conformal bijection of the unit disc to itself is φ(z)=eiθ z−a1−a‾z,∣a∣<1,\varphi(z) = e^{i\theta}\,\frac{z-a}{1-\overline{a}z}, \qquad |a|<1, which sends aa to 00 and preserves ∣z∣=1|z|=1. Every one of them; there are no others.

Schwarz's lemma

f:D→Df : \mathbb{D}\to\mathbb{D} holomorphic with f(0)=0f(0) = 0. Then ∣f(z)∣≤∣z∣|f(z)|\le|z| for all zz, and ∣f′(0)∣≤1|f'(0)|\le1. If equality holds in either at a single interior point, f(z)=czf(z) = cz with ∣c∣=1|c|=1.

The proof is the maximum principle applied to f(z)/zf(z)/z, which has a removable singularity at 00.

The rigidity is the examinable part: one equality anywhere forces the map to be a rotation everywhere. Nothing in real analysis behaves like that.

Without the fixed point at the origin the correct statement is Schwarz–Pick, which says every holomorphic self-map of the disc is non-expanding in the hyperbolic metric, with equality exactly for the automorphisms above.

Riemann mapping theorem

Every simply connected domain that is not all of C\mathbb{C} is conformally equivalent to the unit disc. Both exclusions matter: C\mathbb{C} itself is not (by Liouville, a bijection C→D\mathbb{C}\to\mathbb{D} would be bounded entire), and a domain with a hole is not, since conformal equivalence preserves simple connectedness.

Where intuition breaks

"Möbius maps send circles to circles and lines to lines." They mix the two. 1/z1/z sends the line Re⁡z=1\operatorname{Re}z = 1 to a circle through the origin. The correct statement joins the classes: circles-or-lines to circles-or-lines.

"A conformal map preserves angles, so z‾\overline{z} is conformal." It reverses orientation. Conformal includes preserving the sense of the angle, and z‾\overline{z} is not holomorphic anyway.

"f′≠0f' \ne 0 makes ff conformal, so it is injective." Locally conformal, not globally injective. eze^{z} again.

"Schwarz's lemma applies to any self-map of the disc." It needs f(0)=0f(0)=0. Without it, use Schwarz–Pick or pre-compose with an automorphism moving the relevant point to the origin — which is the standard trick and often the intended route.

"∣f′(0)∣≤1|f'(0)| \le 1, so ff is a rotation." Only if equality holds. Strict inequality says nothing beyond the bound.

"Every simply connected domain is conformally a disc." Except C\mathbb{C} itself. That exclusion is not a technicality; it is Liouville.

"C\mathbb{C} and D\mathbb{D} are conformally equivalent because both are simply connected." No — that is exactly the excluded case, and the reason the theorem has to name it.

The exam's angle

  1. To build a Möbius map, use three points. Send three convenient points where they must go and the map is forced; no other conditions are needed.
  2. Track ∞\infty. Where ∞\infty goes, and what maps to it, usually determines the geometry — whether a line becomes a circle, and where the image is centred.
  3. For self-maps of the disc, normalise to 00 first. Compose with z−a1−a‾z\frac{z-a}{1-\overline a z} so Schwarz applies.
  4. If an option asserts a bound at one point with equality, expect rigidity. Schwarz turns one equality into a complete description.
  5. Check simple connectedness before invoking Riemann, and remember C\mathbb{C} is excluded.

The night before

  • Conformal at aa   ⟺  \iff holomorphic with f′(a)≠0f'(a) \ne 0. A zero of order mm multiplies angles by mm.
  • z‾\overline z is anticonformal: same angles, reversed orientation, not holomorphic.
  • Möbius az+bcz+d\frac{az+b}{cz+d} with ad−bc≠0ad-bc\ne0; T′=ad−bc(cz+d)2T' = \frac{ad-bc}{(cz+d)^2}; group is PGL2(C)PGL_2(\mathbb{C}), composition is matrix multiplication.
  • Circles-or-lines to circles-or-lines. A line is a circle through ∞\infty.
  • Three points determine a Möbius map; cross-ratio is the invariant.
  • Cayley z−iz+i\frac{z-i}{z+i}: upper half plane →\to disc.
  • Disc automorphisms are exactly eiθz−a1−a‾ze^{i\theta}\frac{z-a}{1-\overline a z}.
  • Schwarz: f(0)=0f(0)=0, f:D→Df:\mathbb{D}\to\mathbb{D} ⇒∣f(z)∣≤∣z∣\Rightarrow |f(z)|\le|z|, ∣f′(0)∣≤1|f'(0)|\le1; equality anywhere forces a rotation. Without the fixed point, Schwarz–Pick.
  • Riemann mapping: simply connected and not C\mathbb{C} ⇒\Rightarrow conformally a disc.

See it move

A grid, deformed by a complex functioninteractive

This visual is part of Trap Analysis

z², 1/z, eᶻ, sin z — watch which ones preserve angles everywhere, and where each one doesn't.

Unlock interactive visuals

The trap here

“C\mathbb{C} and the unit disc are biholomorphic (both are simply connected)” — false

Liouville's theorem

A biholomorphism 𝔻 →C\to \mathbb{C} would invert to a bounded entire function. C\mathbb{C} is the sole exception in the Riemann mapping theorem.

More on this →

Check yourself

The image of the upper half-plane under the map w = (z − i)/(z + i) is

More in Zeros and Mappings

Take it further

Last revised . Found a mistake? Tell us — corrections are the fastest thing we act on.

Next: Subgroups, cosets, Lagrange, cyclic groups

Create a free account to keep your place and have this feed your study plan.

ShareWhatsAppTelegram

Open this in the full syllabus view · Unit 2