A holomorphic map of the unit disk into itself may move points in complicated ways. If it fixes the origin, however, it cannot move any point farther from the origin, and equality at one nonzero point forces a rigid rotation. The proof turns the fixed point into a removable quotient and uses the maximum-modulus principle.
The statement
Theorem 7.1.1
(Schwarz's lemma)
Let and let be holomorphic, with and for every . Then
Moreover, if equality holds in the first inequality at any nonzero point, or if , then for all with . Conversely, every such rotation satisfies equality. The target need only lie in the closed unit disk; no boundary values of on are assumed.
The two conclusions are linked: , so the derivative bound is the pointwise bound read at the origin. Equality in the derivative bound therefore means that the ratio , which never exceeds , reaches its supremum at .
Why the quotient is controlled
Proof
(The quotient is bounded by one)
Define for and . Since is holomorphic and , its local power series at 0 shows that is holomorphic on the whole disk. For any with , the closed disk lies inside the domain. On its boundary,
The maximum-modulus principle on that smaller disk gives the same bound inside it. Fix and let with . Then . Multiplying by proves the pointwise bound, while gives the derivative bound. We never evaluate on the unit circle.
Proof X-ray
Where each hypothesis enters
- The condition is what makes holomorphic at : the singularity of is removable only because the numerator vanishes there. Without it the quotient has a pole (see the counterexample below).
- The key move is to apply the maximum-modulus principle to rather than to . Applied to it only returns . Dividing by makes the boundary bound on equal to , and this sharpens as .
- Letting is how the proof closes without boundary values: every fixed interior point lies inside some circle with , so a bound on all such circles is a bound on the whole open disk.
- The bound is exactly the two conclusions: multiplying by gives , and gives .
Equality forces a rotation
Equality in either bound is rigid, and the same quotient explains why.
Proof
(Equality through the maximum-modulus principle)
If at a nonzero , then . If , then . In either case the holomorphic function , already bounded in modulus by 1, attains its maximum at an interior point of the connected disk. The maximum-modulus principle forces to be constant, say with . Hence throughout the disk. A rotation with clearly maps the open disk into itself and attains both equalities.
Examples and a missing hypothesis
Worked example 7.1.2
(A strict inequality)
The map fixes 0 and maps the disk into itself. For every nonzero , , while . Thus strict inequality is possible; the theorem does not say every disk map is a rotation.
Worked example 7.1.3
(A map with derivative one half at the origin)
The map satisfies in the disk, because , so it maps the disk into itself and fixes 0. Here , and .
Worked example 7.1.4
(A product of and a disk automorphism)
Let . For ,
so the fraction has modulus less than and maps into itself with . Then and . Both inequalities are strict, so the equality clause says is not a rotation, as is evident.
Counterexample mode
A disk map that does not fix zero
The constant function maps into itself, but . The claim fails if the fixed-origin hypothesis is omitted. Restoring repairs the statement under the other stated assumptions.
Checks and exercises
Common mistake
(A target disk of another radius)
If on with and , the correct bound is . For , which maps into the disk of radius , one has , so the unscaled inequality fails.
Quick check
(Why is defined as ?)
Consider the limit of as .
Solution · Answer
Because , the difference quotient tends to ; the local series shows the extension is holomorphic.
Quick check
(If , what is ?)
Compare with the bound in the lemma.
Solution · Answer
Since , equality rigidity gives .
Quick check
(Why is a bound on each circle sufficient?)
Think about a fixed interior point.
Solution · Answer
For any fixed interior point we may choose larger than its modulus and then let approach 1 from below.
Quick check
(Does force ?)
Look for an example among the maps above.
Solution · Answer
No: the example has derivative zero at 0 but is not the zero function.
The proof uses holomorphy on the open disk, the origin condition and the stated modulus bound.