Angles and local invertibility
A map is conformal at a point if it preserves oriented angles between smooth curves with nonzero tangents there. If is holomorphic near and , then . Multiplication by the nonzero complex number rotates by its argument and scales by its modulus, so angles are preserved. The holomorphic inverse function theorem also gives a conformal local inverse. At a critical point this argument fails: doubles angles at and is not locally one-to-one there.
Definition 3.1.1
(Conformal equivalence)
A conformal equivalence between domains is a bijective holomorphic map. Two domains joined by one are called conformally equivalent.
Such a map has nonzero derivative everywhere: if the first nonzero term of its Taylor expansion at had degree , nearby values would have preimages locally, contradicting injectivity. Its inverse is holomorphic by the local inverse theorem. Nonzero derivative alone does not guarantee global injectivity: has no critical points on but is periodic.
Möbius maps and the unit disk
Definition 3.1.2
(Möbius transformation)
A Möbius transformation has the form , with , regarded on the Riemann sphere; its finite pole maps to infinity when ; when , infinity maps to infinity.
Its inverse is again Möbius. It maps generalized circles (circles or straight lines, with lines completed by infinity) to generalized circles. This can be checked by writing a generalized circle as , with real , and substituting the inverse transformation.
For in the unit disk, define
Then , the denominator is nonzero on the closed disk, and
Thus the disk maps into itself, its boundary maps to the unit circle, and the inverse is . The boundary identity alone is not a proof of injectivity; here the explicit inverse supplies it.
Theorem 3.1.3
(Automorphisms of the unit disk)
Every automorphism of the unit disk has the form for some in the disk and real , and every such map is an automorphism.
Proof
(Reducing to an automorphism that fixes the origin)
The maps are automorphisms by the preceding paragraph. Conversely, let be an automorphism and . Then is an automorphism fixing . Schwarz's lemma applied to and to gives and , so equality holds in the lemma and . Hence .
Constructing maps by composition
Worked example 3.1.4
(Upper half-plane to disk)
The Cayley map sends to . Since , it sends the upper half-plane into the disk. Solving gives , which sends the disk into the upper half-plane. Both maps are holomorphic on their respective domains, proving a conformal equivalence. The point corresponds to infinity and is not part of the open disk.
Worked example 3.1.5
(A strip to the upper half-plane)
For , the exponential has argument in , so its image lies in the upper half-plane. Every upper-half-plane point has a unique logarithm whose imaginary part lies in ; hence the exponential is bijective on . Its derivative is never zero. Composing with gives a conformal map from the strip to the disk. Widening the strip past height destroys injectivity because the exponential is -periodic.
Worked example 3.1.6
(A sector to a half-plane)
On with , choose the branch of logarithm with argument in . Then maps the sector bijectively onto the upper half-plane. Specifying the branch and domain is essential; a complex power is not a globally single-valued holomorphic function on in general.
The strip example marks exactly how far injectivity of the exponential extends.
Counterexample mode
A nonvanishing derivative on a strip that is too wide
Claim: since never vanishes, is injective on every horizontal strip. Take the strip . The distinct points and both lie in it, and . A nonvanishing derivative gives only local injectivity. The repaired statement: exactly when , so is injective on an open horizontal strip precisely when its height is at most .
Constructing a Möbius map from three points
A boundary correspondence is often the most useful starting point for a mapping problem.
Theorem 3.1.7
(Three points determine a Möbius map)
Given distinct points and distinct points of the Riemann sphere, there is exactly one Möbius transformation with for .
Proof
(Normalizing both triples to 0, 1 and infinity)
For existence, first normalize the source triple to , then undo the same normalization for the target triple. For distinct finite , put
It sends to respectively. If one source point is infinity, take the corresponding limit; for instance, if , then . Given distinct target points , construct their normalization in the same way. The required map is .
Why is it unique? If is any such map, then fixes . A Möbius map fixing is affine, and fixing and forces it to be the identity. Hence the two maps coincide on the sphere. The construction requires distinct points in each triple: an injective Möbius map cannot send distinct points to the same target.
Worked example 3.1.8
(A circle onto the extended real line)
Send the circle through to the extended real line, with these points going to . The normalization gives
A Möbius map sends a generalized circle to a generalized circle. The image contains , so it is the extended real line. The original circle has center and radius . Since , its interior maps to the upper half-plane. To justify the side choice, is a homeomorphism of the sphere and maps the two complementary components of the circle onto the two components of the extended real line; the test point selects the upper one. Its pole lies on the boundary, not inside the disk, so the restriction to the disk is holomorphic and bijective.
Quick check
(Find the Möbius map taking to and determine the image of the upper half-plane.)
Use the normalization formula, then test the sign of .
Solution · Answer
The formula gives . Its real coefficients have determinant , so in the upper half-plane. Its inverse has the same property, so the image is the whole upper half-plane. Knowing only that the boundary maps to itself does not decide the side; also takes the real boundary to itself but reverses the two half-planes.
Every automorphism of the upper half-plane
The Cayley map does more than provide one example: together with the disk classification it determines every conformal self-equivalence of . This is useful when a construction specifies where a point or boundary point must go. It also explains why real coefficients alone are not enough; their determinant must have the correct sign.
Theorem 3.1.9
(Automorphisms of the upper half-plane)
The automorphisms of are exactly
Multiplying all four coefficients by the same nonzero real number does not change the map. In particular one may normalize the determinant to ; the two normalized matrices and describe the same transformation.
Proof
(Transport to the disk, then read off the coefficients)
Let be an automorphism. The Cayley map is a conformal equivalence from to the disk, with inverse . Thus is a disk automorphism, and by the classification above it is Möbius. Consequently is Möbius too. Only now do we use its extension to the Riemann sphere: the extended real line is carried by to the unit circle, which a disk automorphism carries onto itself. Hence maps the extended real line onto itself.
Its preimages of are three distinct points of that extended line. There is a real-coefficient Möbius map with these same three values: first send the three preimages to using the usual cross-ratio formula, interpreting a factor at infinity by its limit. Its coefficients are real. Two Möbius maps agreeing at three distinct points agree everywhere, so has real coefficients. For real coefficients direct calculation gives
The denominator cannot vanish in : if its zero is real, while if then . As has positive imaginary part, .
Conversely, suppose the coefficients are real and . The displayed identity shows that maps into itself. Its inverse is ; its real coefficient matrix has the same positive determinant. The same identity therefore puts the inverse image of every in . Thus is a biholomorphism of onto itself. This also covers , when is an affine map and fixes infinity.
Three different facts enter this proof, at three different moments.
Proof X-ray
Möbius form, real coefficients, positive determinant
The proof never assumes that is Möbius; it imports that from the disk classification through the conjugate . Only then does the extension to the sphere make sense, and the boundary information it yields, that preserves the extended real line, is used once: a Möbius map is determined by three values, and three real points with real images can be matched by a normalization with real coefficients. Real coefficients still allow both signs of , so the last step uses an interior point: selects the positive sign. The converse rests on the single identity for , applied to and to its inverse.
Worked example 3.1.10
(An automorphism fixed by one interior and one boundary value)
The map has determinant , so it is an automorphism. It sends to and sends the boundary point to . These two prescriptions determine it uniquely: if another automorphism has the same values, form . Conjugate the resulting automorphism, which fixes and , to the disk using . The conjugate fixes the origin and a boundary point, so it is a rotation that fixes a point of the unit circle, hence the identity. One checks directly that .
Common mistake
(Preserving the boundary line does not decide the side)
By contrast, has real coefficients but determinant and sends to the lower half-plane. Preserving the extended real line as a set does not decide which side is preserved. Check the sign of , or the image of one interior test point such as .
Concept lensAlgebraic
Automorphisms of the half-plane as a matrix group
Composing with gives the map whose coefficient matrix is the product of the two matrices. With the determinant normalized to , the assignment is therefore a surjective group homomorphism from onto . Its kernel is : for all forces and , so . Hence , and questions about composites become matrix multiplication.
Quick check
(Determine which of and is an automorphism of .)
Write each map as with real coefficients.
Solution · Answer
The first has , determinant , and is an automorphism; the second has determinant and swaps the half-planes.
Quick check
(Find all automorphisms of fixing .)
Compare real and imaginary parts in .
Solution · Answer
gives and . After normalizing the positive determinant to , all solutions are
Indeed the displayed matrix has determinant and fixes . Parameters differing by give the same map because the matrices differ by a minus sign.
Worked construction: a half-disk to a half-plane
The upper half-disk combines two boundary conditions. A useful strategy is to transform each condition into a simple sign condition, then remove the corner with a power map. Set
The denominator is nonzero on . Multiplying by shows that
Both quantities are positive on , so maps it into the first quadrant. This is only half of the argument: to prove the image is the whole quadrant, solve to get . For and ,
Thus this inverse takes every point of the first quadrant back into . Squaring doubles arguments in and is a bijection from the first quadrant to the upper half-plane . The inverse of is therefore
where is the branch with argument in . Both and this inverse are holomorphic on their respective domains, so is a conformal equivalence. The branch is part of the answer: the other square root lies in the third quadrant and does not give the inverse on .
The boundary calculation explains the picture without replacing the interior proof. For , , so the diameter goes to the positive real axis under . If with , then lies on the positive imaginary axis, so the semicircle goes to the negative real axis under . The endpoints are excluded: , whereas has a pole at , corresponding to infinity. A boundary-to-boundary correspondence alone would not prove which side is covered or that the map is one-to-one.
Quick check
(Find a conformal equivalence from , the lower half of the unit disk, onto . Identify the image of .)
Reduce to the upper half-disk first.
Solution · Answer
First send to by , then apply . Thus . At , the intermediate value is , so . The inverse is the composition of the displayed inverse of with negation; consequently surjectivity and injectivity follow from the two proven bijections.
Existence versus an explicit formula
Theorem 3.1.11
(Riemann mapping theorem)
Every simply connected domain in the plane is conformally equivalent to the unit disk.
The theorem does not include : a biholomorphism from to the disk would be a bounded nonconstant entire function, contrary to Liouville's theorem. The theorem is stated here as a deep existence result, not proved. Nor does it automatically give a continuous one-to-one extension to an arbitrary boundary; boundary regularity is a separate question.
Practice and solutions
Quick check
(Find a disk automorphism carrying to and compute its inverse.)
Use .
Solution · Answer
Take . Solving for gives , which is .
Quick check
(Show that is not injective on the whole plane despite its nonzero derivative.)
Use the period of the exponential.
Solution · Answer
, so the distinct points and have the same image.
Quick check
(Map the first quadrant conformally to the upper half-plane.)
Double the argument.
Solution · Answer
The square map sends to . The unique square root with argument in supplies its inverse; its derivative is nonzero on the quadrant.