printed on 5/12/2014

Complex Analysis
Math 534-535-536 2014-2015

Donald E. Marshall

If you are not in my class, you are still welcome to view these notes. My only require-
ment is that you send me any typos you observe or suggestions for improvement you might
have.
WARNING: these notes are subject to change throughout the quarter. Check the
date at the top of this page. Additional information is available from our course web page:
http://www.math.washington.edu/∼marshall/math534-au14.html

i
ii
 CONTENTS

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. vii
Prerequisites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. ix
I. Preliminaries
1: Complex Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 1
2: Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 5
3: Stereographic Projection . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 8
4: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 10
II. Analytic Functions
1: Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 13
2: Power Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 15
3: Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 19
4: Elementary Operations with Analytic Functions . . . . . . . . . . . . . . . . . p. 23
5: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 27
III. The Maximum Principle
1: The Maximum Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 31
2: The Fundamental Theorem of Algebra and Partial Fractions . . . . . . . . . . . p. 32
3: Local Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 36
4: Growth on C and D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 39
5: Boundary Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 42
6: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 45
IV. Integration and Approximation
1: Integration on Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 49
2: Equivalence of Analytic and Holomorphic . . . . . . . . . . . . . . . . . . . . p. 54
3: Approximation by Rational Functions . . . . . . . . . . . . . . . . . . . . . p. 60
4: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 68
V. Cauchy’s Theorem and Applications
1: Cauchy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 71
2: Winding Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 73

iii
iv Contents

3: Removable Singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 79
4: Laurent Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 81
5: The Argument Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 85
6: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 88
VI. Elementary Maps
1: Linear Fractional Transformations . . . . . . . . . . . . . . . . . . . . . . . p. 94
2: Exp and Log . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 99
3: Power Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 102
4: The Joukovski Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 104
5: Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 108
6: Constructing Conformal Maps . . . . . . . . . . . . . . . . . . . . . . . . . p. 109
7: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 116
VII. Calculus of Residues
1: Contour Integration and Residues . . . . . . . . . . . . . . . . . . . . . . . p. 119
2: Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 121
3: Fourier, Mellin, and Inverse Laplace Transforms . . . . . . . . . . . . . . . . . p. 124
4: Series via Residues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 131
5: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 132
VIII. Normal Families
1: Normality and Equicontinuity . . . . . . . . . . . . . . . . . . . . . . . . . p. 135
2: Riemann Mapping Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . p. 144
3: Zalcman, Montel, and Picard . . . . . . . . . . . . . . . . . . . . . . . . . p. 146
4: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 151
IX. Series and Products
1: Mittag-Leffler’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 153
2: Weierstrass Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 160
3: Blaschke Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 167
4: The Gamma and Zeta Functions . . . . . . . . . . . . . . . . . . . . . . . . p. 171
5: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 177
X. Harmonic Functions
 Contents v

1: Mean Value Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 181

2: Cauchy-Riemann and Laplace Equations . . . . . . . . . . . . . . . . . . . . p. 187
3: Hadamard, Lindelöf, and Harnack . . . . . . . . . . . . . . . . . . . . . . . p. 190
4: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 196
XI. Conformal Maps and Harmonic Functions
1: Symmetry and Conformal Maps . . . . . . . . . . . . . . . . . . . . . . . . p. 199
2: The Geodesic Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 203
3: Conformal Maps to Polygonal Regions . . . . . . . . . . . . . . . . . . . . . p. 212
4: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 214
XII. Conformal Maps to Jordan Regions
1: Badly Behaved Jordan Curves . . . . . . . . . . . . . . . . . . . . . . . . . p. 217
2: Janiszewski’s Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 219
3: Jordan Curve Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 222
4: Caratheodory’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 226
5: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 230
XIII. Dirichlet Problem
1: Perron Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 233
2: Local Barriers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 235
3: Another Proof of the Riemann Mapping Theorem . . . . . . . . . . . . . . . . p. 235
4: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 235
XIV. Riemann Surfaces
1: Analytic Continuation . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 237
2: Monodromy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 237
3: Riemann Surface Definition . . . . . . . . . . . . . . . . . . . . . . . . . . p. 237
4: Simply Connected Covering Surface . . . . . . . . . . . . . . . . . . . . . . p. 237
5: Deck Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 237
6: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 237
XV. The Uniformization Theorem
1: Lindelöf’s Maximum Principle . . . . . . . . . . . . . . . . . . . . . . . . . p. 239
2: Green’s Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 240
vi Contents

3: Dipole Green’s Function . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 247

4: Simply Connected with a Green’s Function . . . . . . . . . . . . . . . . . . . p. ‘XV.4’
5: Simply Connected without a Green’s Function . . . . . . . . . . . . . . . . . . p. ‘XV.5’
6: Classification of All Riemann Surfaces . . . . . . . . . . . . . . . . . . . . . p. ‘XV.6’
7: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. ‘XV.7’
XVI. Elliptic Functions
1: . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 255
XVII. Fuchsian Groups
1: Convergence Type and Blaschke Products . . . . . . . . . . . . . . . . . . . . p. ‘XVII.1’
2: Hyperbolic Geometry on D . . . . . . . . . . . . . . . . . . . . . . . . . . p. ‘XVII.2’
XVIII. Equivalent Definitions of Analytic
1: Goursat’s proof of Cauchy’s Theorem . . . . . . . . . . . . . . . . . . . . . . p. 270
2: Weyl’s Lemma . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 272
3: Conditions Equivalent to Analytic . . . . . . . . . . . . . . . . . . . . . . . p. 272
4: Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . p. 272
 Preface

This course is a three quarter graduate level introduction to complex analysis. There
are four points of view for this subject due primarily to Cauchy, Weierstrass, Riemann
and Runge. Cauchy thought of analytic functions in terms of a complex derivative and
through his famous integral formula. Weierstrass instead stressed the importance of power
series expansions. Riemann thought of analytic functions as locally rigid mappings from
one domain to another, a more geometric point of view. Runge showed that analytic
functions are nothing more than limits of rational functions. The seminal modern text
in this area was written by Ahlfors, Complex Analysis, which stresses Cauchy’s point of
view. One aspect of the first year course in complex analysis is that the material has
been around so long that some very slick and elegant proofs have been discovered. The
subject is quite beautiful as a result, but some theorems then may seem mysterious. I’ve
decided instead to start with Weierstrass’s point of view for local behavior. Power series
are elementary and give you many non-trivial functions immediately. In many cases it is a
lot easier to see why certain theorems are true from this point of view. For example, it is
remarkable that a function which has a complex derivative actually has derivatives of all
orders. On the other hand, the derivative of a power series is just another power series and
hence has derivatives of all orders. Cauchy’s theorem is a more global result concerned
R 1
with integrals of analytic functions. Why integrals of the form z−a dz are important in
Cauchy’s theorem is very easy to understand using partial fractions for rational functions.
So we will use Runge’s point of view for more global results: analytic functions are simply
limits of rational functions.

As a dydactic device we will use the term “analytic” for local power series expansion
and “holomorphic” for possessing a continuous complex derivative. We will of course prove
that these concepts (and several others) are equivalent eventually, but in the early chapters
the reader should be alert to the different definitions. The connection between analytic
and harmonic functions is deferred until much later in the course. The emphasis in the
beginning is to view analytic functions as behaving like polynomials.

vii
viii

The exercises at the end of each chapter are divided by difficulty, though in some
cases they can be solved in more than one way. Exercises A are mostly straightforward,
requiring little originality and designed for practice with the material. The B exercises
require a good idea or non-routine use of the results in the chapter. Sometimes a creative
idea or the right insight can lead to a simple solution. The C exercises are usually much
more difficult. You can think of “C” as “challenge”. It is entirely possible to find solutions
to problems by searching the internet. It is also possible to solve some problems using more
advanced techniques or theorems than have been covered in the text. Both will defeat the
purpose of developing the ability to solve problems, a goal of this book. For that reason,
we also ask you to not tempt others by posting solutions.
 Prerequisites

You should be on friendly terms with the following concepts. If you have only seen the
corresponding proofs for real numbers and real-valued functions, check to see if the same
proofs also work when “real” is replaced by “complex”, after reading the first two sections
of Chapter I. As you read the text, check all details. If many of the concepts below are
new to you, then I would recommend that you first take a senior level analysis class.

Let {an }∞ ∞
n=0 , {bn }n=0 be sequences of real numbers and let {fn } be a sequence of

real-valued functions defined on some interval I ⊂ R.

1. {an } converges to a (Notation: an → a) “ǫ − δ ” version.
2. Cauchy sequence
P P
3. an converges, converges absolutely (Notation: |an | < ∞)
P
4. an converges implies an → 0, but not conversely.
5. lim supn→∞ an , lim inf n→∞ an
6. comparison test for convergence
7. Rearranging absolutely convergent series gives the same sum, but a similar statement
does not hold for for conditionally convergent series.
P∞ P∞
8. If n=0 an = A and n=0 bn = B then
∞
X
A+B = (an + bn )
n=0

∞
X
cA = can .
n=0
P Pn
If an converges absolutely and cn = k=0 ak bn−k then

∞
X
AB = cn .
n=0

9.
∞ X
X ∞ ∞ X
X ∞
an,k = an,k
n=0 k=0 k=0 n=0

provided at least one sum converges absolutely.

ix
x

10. Continuous function, uniformly continuous function

11. fn (x) → f (x) pointwise, fn (x) → f (x) uniformly.
12. Uniform limit of a sequence of continuous functions is continuous.
R  R
13.  I f (x)dx ≤ I |f (x)|dx
14. If fn → f uniformly on a bounded interval I then
Z Z Z
lim fn (x)dx = lim fn (x)dx = f (x)dx
I I I

15. Corollary:
∞ Z
X Z X
∞
fn (x)dx = fn (x)dx,
n=0 I I n=0
P
if the partial sums of fn converge uniformly on the bounded interval I.
16. open set, closed set, connected set, compact set, metric space.
17. f continuous on a compact set X implies f is uniformly continuous on X.
18. X ⊂ Rn is compact if and only if it is closed and bounded.
19. A metric space X is compact if and only if every infinite sequence in X has a limit
point in X. (This can fail if X not a metric space)
20. If f is continuous on a connected set U , then f (U ) is connected. If f is continuous on
a compact set K then f (K) is compact.
21. A continuous real-valued function on a compact set has a maximum and a minimum.
22. Green’s theorem. You should look at the proof you learned (if in fact you saw a proof)
and figure out exactly what the hypotheses are for that version. Many undergrad
books prove a special case, then wave their hands.

All of the above can be found in the undergraduate text Rudin, Principles of Math-
ematical Analysis as well as many other sources. Items #15 -#20 are also in Ahlfors,
Complex Analysis, Chapter 3 section 1 (pages 50-61).
 I

Preliminaries

§1. Complex Numbers.

The “complex numbers” C consist of pairs of real numbers: {(x, y) : x, y ∈ R}.

The complex number (x, y) can be represented geometrically as point in the plane R2 , or
viewed as a vector whose tip has coordinates (x, y) and whose tail has coordinates (0, 0).
The complex number (x, y) can be identified with another pair of real numbers (r, θ), called
the polar coordinate representation. The line from (0, 0) to (x, y) has length r and forms
an angle with the positive x-axis. The angle is measured by using the distance along the
corresponding arc of the circle of radius 1 (centered at (0, 0)). By similarity, the length of
the subtended arc on the circle of radius r is rθ.

(x, y)
y θ
r rθ

x 1

Figure I.1 Cartesian and polar representation of complex numbers.

Conversion between these two representations is given by

x = r cos θ, y = r sin θ

and
p y
r= x2 + y 2 , tan θ = .
x
Care must be taken to find θ from the last equality since many angles can have the
same tangent. However, consideration of the quadrant containing (x, y) will give a unique
θ ∈ [0, 2π), provided r > 0 (we do not define θ when r = 0).

1
2 I. Preliminaries

Addition of complex numbers is defined coordinatewise:

(a, b) + (c, d) = (a + c, b + d),

and can be visualized by vector addition.

(a + c, b + d)
(a, b)

(c, d)

Figure I.2 Addition.

Multiplication is given by:

(a, b) · (c, d) = (ac − bd, bc + ad)

and can be visualized as follows: The points (0, 0), (1, 0), (a, b) form a triangle. Construct
a similar triangle with corresponding points (0, 0), (c, d), (x, y). Then it is an exercise in
high school geometry to show that (x, y) = (a, b) · (c, d). By similarity, the length of the
product is the product of the lengths and angles are added.

(a, b) · (c, d)

(c, d)
(a, b)

(0, 0) (1, 0)

Figure I.3 Multiplication.

The real number t is identified with the complex number (t, 0). With this identifica-
tion, complex addition and multiplication is an extension of the usual addition and multipli-
cation of real numbers. For conciseness when t is real, t(x, y) means (t, 0) · (x, y) = (tx, ty).
The additive identity is 0 = (0, 0) and −(x, y) = (−x, −y). The multiplicative identity
is 1 = (1, 0) and the multiplicative inverse of (x, y) is (x/(x2 + y 2 ), −y/(x2 + y 2 )). It
 §1: Complex Numbers 3

is a tedious exercise to check that the commutative and associative laws of addition and
multiplication hold, as does the distributive law.

The notation for complex numbers becomes much easier if we use a single letter instead
of a pair. It is traditional, at least among mathematicians, to use the letter i to denote
the complex number (0, 1). If z is the complex number given by (x, y), then because
(x, y) = x(1, 0) + y(0, 1), we can write z = x + yi. If z = x + iy then the “real” part of
z is Rez = x and the “imaginary” part is Imz = y. Note that i · i = −1. We can now
just use the usual algebraic rules for manipulating complex numbers together with the
simplification i2 = −1. For example, z/w means multiplication of z by the multiplicative
inverse of w. To find the real and imaginary parts of the quotient, we use the analog of
“rationalizing the denominator”:

x + iy (x + iy)(a − ib) xa − i2 yb + iya − ixb xa + yb
ya − xb

= = 2 2
= 2 + 2 i.
a + ib (a + ib)(a − ib) a +b a + b2 a + b2

Here is some additional notation: if z = x + iy is given in polar coordinates by the

pair (r, θ) then
p
|z| = r = x2 + y 2

is called the modulus or absolute value of z. Note that |z| is the distance from the complex
number z to the origin 0. The angle θ is called the “argument” of z and written

θ = arg z.

The most common convention is that −π < arg z ≤ π, where positive angles are measured
counter-clockwise and negative angles are measured clockwise. The complex conjugate of
z is given by

z = x − iy.

The complex conjugate is the reflection of z about the real line R.

4 I. Preliminaries

It is an easy exercise to show

|zw| = |z||w|

|cz| = c|z| if c > 0,

z/|z| has absolute value 1,

zz = |z|2 ,

Rez = (z + z)/2,

Imz = (z − z)/(2i),

z + w = z + w,

zw = z · w,

z = z,

|z| = |z|,

arg zw = arg z + arg w modulo 2π,

arg z = − arg z = 2π − arg z modulo 2π.

The statement “modulo 2π” means that the difference between the left and right hand
sides of the equality is an integer multiple of 2π.

The identity a + (z − a) = z expressed in vector form shows that z − a is (a translate

of) the vector from a to z. Thus |z − a| is the length of the complex number z − a but it
is also equal to the distance from a to z. The circle centered at a with radius r is given by
{z : |z − a| = r} and the disk centered at a of radius r is given by {z : |z − a| < r}. The
open disks are the basic open sets generating the standard topology on C. We will use D
to denote the unit disk,
D = {z : |z| < 1}

and use ∂D to denote the unit circle,

∂D = {z : |z| = 1}.

Complex numbers were around for at least 250 years before good applications were
found. Cardano discussed them in his book Ars Magna (1545). Beginning in the 1800’s,
 §2: Estimates 5

and continuing today, there has been an explosive growth in their usage. Now complex
numbers are very important in the application of mathematics to engineering and physics.
It is a historical fiction that solutions to quadratic equations forced us to take complex
numbers seriously. How to solve x2 = mx + c has been known for 2000 years and can be
visualized as the points of intersection of the standard parabola y = x2 and the line
y = mx + c. As the line is shifted up or down by changing c, it is easy to see there are
two, or one, or no (real) solutions. The solution to the cubic equation is where complex
numbers really became important. A cubic equation can be put in the standard form

x3 = 3px + 2q,

by scaling and translating. The solutions can be visualized as the intersection of the
standard cubic y = x3 and the line y = 3px + 2q. Every line meets the cubic, so there will
always be a solution. By formal manipulations, Cardano showed that a solution is given
by:
p 1 p 1
x = (q + q 2 − p3 ) 3 + (q − q 2 − p3 ) 3 .

Bombelli pointed out 30 years later that if p = 5 and q = 2 then x = 4 is a solution,

but q 2 − p3 < 0 so the above solution doesn’t make sense. His “wild thought” was to use
complex numbers to understand the solution:

1 1
x = (2 + 11i) 3 + (2 − 11i) 3 .

He found that (2 ± i)3 = 2 ± 11i and so the above solution actually equals 4. In other
words, complex numbers were used to find a real solution.

§2. Estimates.

Here are some elementary estimates which the reader should check:

−|z| ≤ Rez ≤ |z|

−|z| ≤ Imz ≤ |z|

6 I. Preliminaries

and
|z| ≤ |Rez| + |Imz|.

Perhaps the most useful inequality in analysis is the

Triangle inequality.
|z + w| ≤ |z| + |w|,

and
 
|z + w| ≥ |z| − |w|.

The associated picture perhaps makes this result clear:

z+w w

Figure I.4 Triangle inequality

Analysis is used to give a more rigorous proof of the triangle inequality (and it is good
practice with the notation we’ve introduced):

Proof.
|z + w|2 = (z + w)(z + w)

= zz + wz + zw + ww

= |z|2 + 2Re(wz) + |w|2

≤ |z|2 + 2|w||z| + |w|2

= (|z| + |w|)2 .
To obtain the second part of the triangle inequality:

|z| = |z + w + (−w)| ≤ |z + w| + | − w| = |z + w| + |w|

and by subtracting |w|,

|z| − |w| ≤ |z + w|,
 §2: Estimates 7

and switching z and w:

|w| − |z| ≤ |z + w|,

so that
 
|z| − |w|≤ |z + w|. 

These estimates can be used to prove that {zn } converges if and only if both {Rezn }
P
and {Imzn } converge. The series an is said to converge if the sequence of partial sums
m
X
Sm = an
n=1
P
converges, and the series converges absolutely if |an | converges. A series is said to
diverge if it does not converge. Absolute convergence implies convergence because Cauchy
P
sequences converge. We sometimes write |an | < ∞ to denote absolute convergence
P
because the partial sums are increasing. It also follows that an is absolutely convergent
P P
if and only if both Rean and Iman are absolutely convergent. By comparing the nth
P
partial sum and the (n − 1)st partial sum, if an converges then an → 0. As with real
series, the converse statement is false.
Another useful estimate is the

Cauchy-Schwarz inequality.
 n   n  12 X
n  12
X  X
 
aj bj ≤ |aj | 2
|bj | 2
.

j=1 j=1 j=1

If v and w are vectors in Cn , the Cauchy-Schwarz inequality says that |hv, wi| ≤
||v||||w||, where the left side is the absolute value of the inner product and the right side
is the product of the lengths of the vectors.

Proof.
n n
1 XX
0≤ |ai bj − aj bi |2
2
j=1 i=1
Xn X n
1
= |ai |2 |bj |2 + |aj |2 |bi |2 − ai bj aj bi − ai bj aj bi
2 j=1 i=1
n
X n
X X
n X
n 
2 2
= |ai | |bj | − a i bi a j bj
i=1 j=1 i=1 j=1
8 I. Preliminaries

Thus  n 2 n n
X  X X
 
a j bj  ≤ |aj | 2
|bj |2 . 

j=1 j=1 j=1

The proof above also gives the error

n n
1 XX
|ai bj − aj bi |2 ,
2
j=1 i=1

and so equality occurs if and only if ai = cbi for all i and some (complex) constant c, or
bi = 0 for all i.

The reader can use Riemann integration to deduce the following, which is also called
the Cauchy-Schwarz inequality. For a complex valued function f defined on a real interval
Rb Rb Rb
[a, b], we define a f dx ≡ a Ref dx + i a Imf dx.

Corollary. If f and g are continuous complex-valued functions defined on [a, b] ⊂ R then

Z  Z ! 21 Z ! 21
 b  b b
 
 f (t)g(t)dt≤ |f (t)|2 dt |g(t)|2 dt .
 a  a a

This corollary can also be proved directly by expanding

Z b Z b
|f (x)g(y) − f (y)g(x)|2dxdy (2.1)
a a

in a similar way, giving a proof for f, g ∈ L2 . Moreover, the error term is half of the integral
(2.1) and equality occurs if and only if f = cg, for some constant c, or g is identically zero.

§3. Stereographic Projection.

A component of Riemann’s point of view of functions as mappings is that ∞ is like any

other complex number. But we cannot extend the definition of complex numbers to include
∞ and still have the usual laws of arithmetic hold. However, there is another “picture” of
complex numbers that can help us visualize this idea. The picture is called “stereographic
projection”. We identify the complex numbers with the plane {(x, y, 0) : x, y ∈ R} in R3 .
 §3: Stereographic Projection 9
c, b
If z = x + iy, let z ∗ be the unique point on the unit sphere in R3 which also lies on the
line from the “North Pole” (0, 0, 1) through (x, y, 0). Thus

z ∗ = (x1 , x2 , x3 ) = (0, 0, 1) + t[(x, y, 0) − (0, 0, 1)].

N = (0, 0, 1)
(x1 , x2 , x3 )

(x, y, 0)

Figure I.5 Stereographic projection.

Then
p
|z ∗ | = (tx)2 + (ty)2 + (1 − t)2 = 1,

which gives
2
t= ,
x2 + y 2 + 1
where 0 < t ≤ 2, and
 
∗ 2x 2y x2 + y 2 − 1
z = , , .
x2 + y 2 + 1 x2 + y 2 + 1 x2 + y 2 + 1

The reader is invited to find z = x + iy from z ∗ = (x1 , x2 , x3 ). The sphere is some-

times called the “Riemann sphere” and is denoted S 2 or C∗ . It is an explicit one-point
compactification of the complex plane. The North Pole corresponds to “∞”.

Theorem 3.1. Under stereographic projection, circles and straight lines in C correspond
precisely to circles on S 2 = C∗ .

Proof. A circle on S 2 is the intersection of a plane with the sphere. Indeed a circle
determines a plane. Conversely the sphere is invariant under rotation about the normal
10 I. Preliminaries

direction to a plane P , so that the intersection of the sphere and the plane must be a circle.
If the plane is given by
Ax1 + Bx2 + Cx3 = D

and if (x1 , x2 , x3 ) corresponds to (x, y, 0) under stereographic projection then

     2 
2x 2y x + y2 − 1
A 2 +B 2 +C = D.
x + y2 + 1 x + y2 + 1 x2 + y 2 + 1
Thus
(C − D)(x2 + y 2 ) + 2Ax + 2By = C + D.

If C = D, then this is the equation of a line, and all lines can be written this way. If
C 6= D, then by completing the square, we get the equation of a circle, and all circles can
be put in this form. 

So we will consider a line in C as just a special kind of “circle”.

The sphere S 2 inherits a topology from the usual topology on R3 generated by the
balls in R3 .

Corollary 3.2. The topology on S 2 induces the standard topology on C via stereographic
projection, and moreover a basic neighborhood of ∞ is of the form {z : |z| > r}.

For later use, we note that the chordal distance between two points on the sphere
induces a metric on C which is given by

χ(z, w) = |z ∗ − w∗ | = p 2|z − w|
p .
1 + |z|2 1 + |w|2

This metric is bounded (by 2).

§4. Exercises.
A

1. Check that item 4 of the prerequisites holds for complex an . Check that items 13, 14,
and 15 of the prerequisites hold for complex valued functions defined on an interval
I ⊂ R.
 §4: Exercises 11

2. Check the details of the high school geometry problem in the geometric version of
complex multiplication.
3. Prove the parallelogram equality:

|z + w|2 + |z − w|2 = 2(|z|2 + |w|2 ).

In geometric terms, the equality says that the sum of the squares of the lengths of the
diagonals of a parallelogram equals the sum of the squares of the lengths of the sides.
It is perhaps a bit easier to prove it using the complex notation of this chapter than
a proof using high school geometry.
4. Prove the Corollary at the end of section 2.
5. Check the formula for the chordal distance between two points on the unit sphere
given in section 3. The chordal distance is bounded by 2, by the triangle inequality.
Verify analytically that the formula for this distance given in the text is bounded by
2 using the Cauchy-Schwarz inequality, and also by directly multiplying it out using
complex notation and one of the estimates at the start of section 2.

6. Suppose that f is a continuous complex valued function on a real interval [a, b]. Let
Z b
1
A= f (x)dx
b−a a
be the average of f over the interval [a, b].
(a) Show that if |f (x)| ≤ |A| for all x ∈ [a, b], then f = A. Hint: rotate f so that A > 0.
Rb
Then a (A − Ref )dx/(b − a) = 0, and A − Ref is continuous and non-negative.
Rb
(b) Show that if |A| = (1/(b − a)) a |f (x)|dx, then arg f is constant on {z : f (z) 6= 0}.
7. Formally solve the cubic equation ax3 + bx2 + cx + d = 0, where x, a, b, c, d ∈ C, a 6= 0,
by the following reduction process:
(a) Set x = u + t and choose the constant t so that the coefficient of u2 is equal to zero.
(b) If the coefficient of u is also zero, then take a cube root to solve. If the coefficient
of u is non-zero, set u = kv and choose the constant k so that v 3 = 3v + r, for
some constant r.
12 I. Preliminaries

(c) Set v = z + 1/z and obtain a quadratic equation for z 3 . The map z + 1/z is
important for several reasons, including constructing what are called conformal
maps. It will be examined in more detail in Section VI.4.
(d) Use the quadratic formula to find two possible values for z 3 , and then take a cube
root to solve for z.

8. Prove that stereographic projection preserves angles between curves. In other words,
if two curves in the plane meet at an angle θ, then their lifts to the sphere meet
at the same angle. Hint: this can be done without any calculations by considering
intersecting planes. This problem will be revisited in the exercises for Chapter VI,
where it is used to find the Mercator projection, a map of tremendous economic
impact.
9. Stereographic projection combined with rigid motions of the unit sphere can be used
to describe some transformations of the plane.
(a) Map a point z ∈ C to S 2 , apply a rotation of the unit sphere, then map the
resulting point back to the plane. For a fixed rotation, find this map of the
extended plane to itself as an explicit function of z. Two cases are worth working
out first: rotation about the x3 axis and rotation about the x1 axis.
(b) Another map can be obtained by mapping a point z ∈ C to S 2 , then translat-
ing the sphere so that the origin is sent to (x0 , y0 , z0 ), then projecting back to
the plane. The projection to the plane is given by drawing a line through the
(translated) north pole and a point on the (translated) sphere and finding the
intersection with the plane {(x, y, 0)}. For a fixed translation, find this map as an
explicit function of z. In this case it is worth working out a vertical translation
and a translation in the plane separately. Then view an arbitrary translation as
a composition of these two maps. Partial answer: the maps in part (a) and (b)
are of the form (az + b)/(cz + d) with ad − bc 6= 0
For a movie of these maps, see http://www.ima.umn.edu/∼arnold/moebius/index.html,
but do the problem before viewing this link.
 II

Analytic Functions

§1. Polynomials.

This course is about complex-valued functions of a complex variable. We could think

of such functions in terms of real variables as maps from R2 into R2 given by

f (x, y) = (u(x, y), v(x, y)),

and think of the graph of f as a subset of R4 . But the subject becomes more tractable if
we use a single letter z to denote in the independent variable and write f (z) for the value
at z, where z = x + iy and f (z) = u(z) + iv(z). For example

f (z) = z n

is much simpler to write (and understand) than its real equivalent. Here z n means the
product of n copies of z.
The simplest functions are the polynomials in z:

p(z) = a0 + a1 z + a2 z 2 + . . . + an z n , (1.1)

where a0 , . . . , an are complex numbers. If an 6= 0, then we say that n is the degree of p.

Note that z is not a (complex) polynomial, and neither is Rez or Imz.
Let’s take a closer look at linear or degree 1 polynomials. For example if b is a (fixed)
complex number, then
g(z) = z + b

translates, or shifts, the plane. If a is a (fixed) complex number then

h(z) = az

13
14 II. Analytic Functions

can be viewed as a rotation and dilation. To see this, write z = reiθ where r = |z| > 0
and θ = arg z ∈ (−π, π]. Similarly, write a = Aeiα , where a > 0. Then by Chapter I and
Exercise I.1,
h(z) = Arei(θ+α)

so that h rotates the point z by the angle α and dilates, or scales, by a factor of A. A
linear function
f (z) = az + b

can then be viewed as a dilation and rotation followed by a translation. Equivalently,

writing f (z) = a(z + b/a) we can view f as a translation followed by a rotation and
dilation.
Another instructive example is the function

p(z) = z n = r n einθ .

Each pie slice  

 2πk  π

Sk = {z : arg z − < } ∩ {z : |z| < r},
n  n
k = 0, . . . , n − 1 is mapped to a slit disk

{z : |z| < r n } \ (−r n , 0).

Angles between straight line segments issuing from the origin are multiplied by n and for
small r, the size of the image disk is much smaller than the “radius” of the pie slice. See
Figure II.1

zn
r −r n 0
0

Figure II.1 The power map.

The function k(z) = b(z − z0 )n can be viewed as a translation by −z0 , followed by

the power function, and then a rotation and dilation. To put it another way, k translates
 §2: Power Series 15

a neighborhood of z0 to the origin, then acts like the power function z n , followed by a
dilation and rotation by b.
To understand the local behavior of a polynomial (1.1) near a point z0 , write z =
(z − z0 ) + z0 and expand (1.1) by multiplying out and collecting terms to obtain:

p(z) = p(z0 ) + b1 (z − z0 ) + b2 (z − z0 )2 + . . . + bn (z − z0 )n . (1.2)

Another way to see this is to note that p(z) − an (z − z0 )n is a polynomial of degree at

most n − 1, so (1.2) follows by induction on the degree. If b1 6= 0 then p(z) behaves
like the linear function p(z0 ) + b1 (z − z0 ) for z near z0 . If b1 = 0 then near z0 , p(z) is
closely approximated by p(z0 ) + bk (z − z0 )k , where bk is the first non-zero coefficient in
the expansion (1.2). Indeed, for small ε,

|p(z0 + εeiθ ) − [p(z0 ) + bk εk eikθ ]| ≤ Cεk+1 ,

for some constant C. Figure II.2 is sometimes called “walking the dog”, where the leash
has length s.

s p(z0 ) + bk εk eikθ
r
p(z0 )

Figure II.2 p(z0 + εeiθ ) lies in a small disk of radius s = Cεk+1 = ε |bCk | r < r.

For z near z0 then, p(z) behaves like a translation by z0 , followed by a power function,
a rotation and dilation, and finally a translation by p(z0 ).
More complicated functions are found by taking limits of polynomials.

§2. Power Series.

Here is the primary example:

∞
X
zn .
n=0
16 II. Analytic Functions

This series is important to understand because its behavior is typical of all power series
(defined shortly) and because it is one of the few series we can actually add up explicitly.
The partial sums
m
X
Sm = zn = 1 + z + z2 + . . . + zm
n=0

satisfy
(1 − z)Sm = 1 − z m+1 ,

as can be seen by multiplying out the left side and canceling. If z 6= 1 then

1 − z m+1
Sm = .
1−z

Notice that if |z| < 1, then |z m | = |z|m → 0 as m → ∞ and so Sm (z) → 1/(1 − z). If
|z| > 1, then |z m | = |z|m → ∞ and so the sum diverges for these z. If |z| = 1 but z 6= 1
then z n does not tend to 0, so the series diverges. Finally if z = 1 then the partial sums
satisfy Sm = m → ∞, so we conclude that if |z| < 1 then

∞
X 1
zn = , (2.1)
n=0
1−z

and if |z| ≥ 1, then the series diverges. It is important to note that the left and right sides
of (2.1) are different objects. They agree in |z| < 1, the right side is defined for all z 6= 1,
but the left side is defined only for |z| < 1.
The formal power series

∞
X
f (z) = an (z − z0 )n = a0 + a1 (z − z0 ) + a2 (z − z0 )2 + . . .
n=0

is called a convergent power series centered (or based) at z0 if there is an r > 0 so

that the series converges for all z such that |z − z0 | < r. Note: If we plug z = z0 into the
formal power series, then we always get a0 = f (z0 ) (more formally, the definition of the
summation notation includes the convention that the n = 0 term equals a0 , so that we are
not raising 0 to the power 0.) The requirement for a power series to converge is stronger
than convergence at just the one point z0 .
 §2: Power Series 17

A variant of the primary example is:

1 1 1
= = z−z0 .
z−a z − z0 − (a − z0 ) −(a − z0 )(1 − ( a−z0
))

Substituting
z − z0
w=
a − z0
into (2.1) we obtain, when |w| = |(z − z0 )/(a − z0 )| < 1,

∞
X
1 −1
= (z − z0 )n . (2.2)
z − a n=0 (a − z0 )n+1

In other words by (2.1), this series converges if |z − z0 | < |a − z0 | and diverges if |z − z0 | ≥

|a − z0 |. The region of convergence is an open disk and it is the largest disk centered at z0
and contained in the domain of definition of 1/(z − a). In particular, this function has a
power series expansion based at every z0 6= a, but different series for different base points
z0 .

Theorem 2.1 (Weierstrass M-Test). If |an (z − z0 )n | ≤ Mn for |z − z0 | ≤ r and if

P P∞
Mn < ∞ then n=0 an (z −z0 )n converges uniformly and absolutely in {z : |z −z0 | ≤ r}.

Proof. If M > N then the partial sums Sn (z) satisfy

M M
 X  X
|SM (z) − SN (z)| =  an (z − z0 )n ≤ Mn .
n=N+1 n=N+1

P PM
Since Mn < ∞, we deduce n=N+1 Mn → 0 as N, M → ∞, and so {Sn } is a Cauchy
sequence converging uniformly. The same proof also shows absolute convergence. 

Note that the convergence depends only on the “tail” of the series so that we need
only satisfy the hypotheses in the Weierstrass M-test for n ≥ n0 to obtain the conclusion.
The primary example (2.1) converges on a disk and diverges outside the disk. The
next result says that disks are the only kind of region in which a power series can converge.
18 II. Analytic Functions
P
Theorem 2.2 (Root Test). Suppose an (z − z0 )n is a formal power series. Let
1 1
R = lim inf |an |− n = 1 ∈ [0, +∞].
n→∞ lim sup |an | n
n→∞
P∞ n
Then n=0 an (z − z0 )
(a) converges absolutely in {z : |z − z0 | < R},
(b) converges uniformly in {z : |z − z0 | ≤ r} for all r < R, and
(c) diverges in {z : |z − z0 | > R}.
diverge

converge
R z0

Figure II.3 Convergence of a power series.

P
Proof. The idea is to compare the given series with the example (2.1), z n . If |z − z0 | ≤
1
r < R, then choose r1 with r < r1 < R. Thus r1 < lim inf |an |− n , and there is an n0 < ∞
1
so that r1 < |an |− n for all n ≥ n0 . This implies that |an (z − z0 )n | ≤ ( rr1 )n . But by (2.1),
∞ 
X n
r 1
= <∞
n=0
r1 1 − r/r1
since r/r1 < 1. Applying Weierstrass’s M-test to the tail of the series (n ≥ n0 ) proves
(b). This same proof also shows absolute convergence (a) for each z with |z − z0 | < R. If
1
|z − z0 | > R, fix z and choose r so that R < r < |z − z0 |. Then |an |− n < r for infinitely
many n and hence  n
|z − z0 |
n
|an (z − z0 ) | >
r
n
for infinitely many n. Since (|z − z0 |/r) → ∞ as n → ∞, (c) holds. 

The proof of the Root Test shows that if the terms an (z − z0 )n of the formal power
series are bounded when z = z1 then the series converges on {z : |z − z0 | < |z1 − z0 |}.
The Root Test does not give any information about convergence on the circle of radius
R. The series can converge at none, some, or all points of {z : |z−z0 | = R}, as the following
examples illustrate.
 §3: Analytic Functions 19

Examples.

X∞ X∞ ∞
X ∞
X ∞
X
zn zn 2 2
(i) (ii) (iii) nz n (iv) 2n z n (v) 2−n z n
n=1
n n=1
n2 n=1 n=1 n=1

The reader should verify the following facts about these examples. The radius of
convergence of each of the first three series is R = 1. When z = 1, the first series is the
harmonic series which diverges, and when z = −1 the first series is an alternating series
whose terms decrease in absolute value and hence converges. The second series converges
uniformly and absolutely on {|z| = 1}. The third series diverges at all points of {|z| = 1}.
The fourth series has radius of convergence R = 0 and hence is not a convergent power
series. The fifth example has radius of convergence R = ∞ and hence converges for all
z ∈ C.
P
What is the radius of convergence of the series an z n where
(
3−n if n is even
an =
4n if n is odd?
This is an example where ratios of successive terms in the series does not provide sufficient
information to determine convergence.

§3. Analytic Functions

Definition. A function f is analytic at z0 if f has a power series expansion valid in a

P
neighborhood of z0 . This means that there is an r > 0 and a power series an (z − z0 )n
which converges in B = {z : |z − z0 | < r} and satisfies
∞
X
f (z) = an (z − z0 )n ,
n=0

for all z ∈ B. A function f is analytic on set Ω if f is analytic at each z0 ∈ Ω.

Note that we do not require one series for f to converge in all of Ω. The example
(2.2), (z − a)−1 , is analytic on C \ {a} and is not given by one series. Note that if f is
analytic on Ω then f is continuous in Ω. Indeed, continuity is a local property. To check
20 II. Analytic Functions

continuity near z0 , use the series based at z0 . Since the partial sums are continuous and
converge uniformly on a closed disk centered at z0 , the limit function f is continuous on
that disk.

A natural question at this point is: where is a power series analytic?

P
Theorem 3.1. If f (z) = an (z − z0 )n converges on {z : |z − z0 | < r} then f is analytic
on {z : |z − z0 | < r}.

Proof. Fix z1 with |z1 − z0 | < r. We need to prove that f has a power series expansion
based at z1 . By the binomial theorem

n  
X
n n n
(z − z0 ) = (z − z1 + z1 − z0 ) = (z1 − z0 )n−k (z − z1 )k .
k
k=0

Hence
∞ X
X n   
n n−k k
f (z) = an (z1 − z0 ) (z − z1 ) . (3.1)
n=0 k=0
k

Suppose for the moment, that we can interchange the order of summation, then

∞ X
X ∞   
n n−k
an (z1 − z0 ) (z − z1 )k
k
k=0 n=k

should be the power series expansion for f based at z1 . To justify this interchange of
summation, it suffices to prove absolute convergence of (3.1). By the root test

∞
X
|an ||w − z0 |n
n=0

converges if |w − z0 | < r. Set

w = |z − z1 | + |z1 − z0 | + z0 .

Then |w − z0 | = |z − z1 | + |z1 − z0 | < r provided |z − z1 | < r − |z1 − z0 |.

 §3: Analytic Functions 21

z
r − |z1 − z0 | z1 r

z0 w

Figure II.4 Proof of Theorem 3.1.

Thus if |z − z1 | < r − |z1 − z0 |,

∞
X
∞> |an ||w − z0 |n
n=0
X∞

n
= |an | |z − z1 | + |z1 − z0 |
n=0
X∞ X
n   
n n−k k
= |an | |z1 − z0 | |z − z1 |
n=0 k=0
k

as desired. 

Another natural question is: Can an analytic function have more than one power
series expansion based at z0 ?

Theorem 3.2 (Uniqueness). Suppose

∞
X ∞
X
n
an (z − z0 ) = bn (z − z0 )n ,
n=0 n=0

for all z such that |z − z0 | < r where r > 0. Then an = bn for all n.

P∞
Proof. Set cn = an − bn . The hypothesis implies that n=0 cn (z − z0 )n = 0 and we need
to show that cn = 0 for all n. Suppose cm is the first non-zero coefficient. Set
∞
X ∞
X
n −m
F (z) ≡ cn+m (z − z0 ) = (z − z0 ) cn (z − z0 )n .
n=0 n=m
22 II. Analytic Functions

The series for F converges in 0 < |z − z0 | < r because we can multiply the terms of the
series on the right side by the non-zero number (z − z0 )−m and not affect convergence. By
the root test, the series for F converges in a disk and hence in {|z − z0 | < r}. Since F is
continuous and cm 6= 0, there is a δ > 0 so that if |z − z0 | < δ, then

|F (z) − F (z0 )| = |F (z) − cm | < |cm |/2.

If F (z) = 0, then we obtain the contradiction | − cm | < |cm |/2. Thus F (z) 6= 0 when
|z − z0 | < δ. But (z − z0 )m = 0 only when z = z0 , and thus
∞
X
cn (z − z0 )n = (z − z0 )m F (z) 6= 0
n=0
P
when 0 < |z − z0 | < δ, contradicting our assumption on cn (z − z0 )n . 

Notice that the proof of Theorem 3.2 shows that if f is analytic at z0 then for some
δ > 0, either f (z) 6= 0 when 0 < |z − z0 | < δ or f (z) = 0 for all z such that |z − z0 | < δ. If
f (a) = 0, then a is called a zero of f . Recall that a region is a connected open set.

Corollary 3.3. If f is analytic on a region Ω then either f ≡ 0 or the zeros of f are

isolated.

Proof. Let E denote the set of non-isolated zeros of f . In the proof of the Uniqueness
theorem, we showed that if z0 is a non-isolated zero of f then f is identically zero in a
neighborhood of z0 . Thus the set of non-isolated zeros of f is open. Since f is continuous,
the set of zeros and hence the set of non-isolated zeros is closed, because we can find a
union of open disks containing only isolated zeros. By connectedness, either E = Ω or
E = ∅. 

There are plenty of continuous functions for which the corollary is false, for example
x sin(1/x). The corollary is true because near z0 , f (z) behaves like the first non-zero term
in its power series expansion about z0 .
 §4: Elementary Operations 23

§4. Elementary Operations with Analytic Functions

Theorem 4.1. If f and g are analytic at z0 then so are

f + g, f − g, cf (where c is a constant), and f g.

If h is analytic at f (z0 ) then (h ◦ f )(z) ≡ h(f (z)) is analytic at z0 .

Proof. The first three follow from the fact that the partial sums are absolutely convergent
near z0 , together with the associative, commutative and distributive laws applied to the
partial sums. Here we have used the fact that absolutely convergent complex series can be
rearranged, which follows from the same statement for real series by considering real and
imaginary parts. To prove that the product of two analytic functions is analytic, multiply
P P
f (z) = an (z − z0 )n and g = bn (z − z0 )n as if they were polynomials to obtain:

∞ ∞ ∞ n
!
X X X X
an (z − z0 )n bk (z − z0 )k = ak bn−k (z − z0 )n , (4.1)
n=0 k=0 n=0 k=0

which is called the Cauchy product of the two series. Why is this formal computation
valid? If the series for f and the series for g converge absolutely then because we can
rearrange non-negative convergent series
∞ ∞ ∞ n
!
X X X X
n k
∞> |an ||z − z0 | |bk ||z − z0 | = |ak ||bn−k | |z − z0 |n .
n=0 k=0 n=0 k=0

This says that the series on the right-hand side of (4.1) is absolutely convergent and
therefore can be arranged to give the left-hand side of (4.1). To put it another way, the
doubly indexed sequence an bk (z − z0 )n+k can be added up two ways: If we add along
diagonals: n + k = m, for m = 0, 1, 2, . . ., we obtain the partial sums of the right-hand
side of (4.1). If we add along partial rows and columns n = m, k = 0, . . . , m, and k = m,
n = 0, . . . , m − 1, for m = 1, 2, . . ., we obtain the product of the partial sums for the series
on the left-hand side of (4.1). Since the series is absolutely convergent (as can be seen by
using the latter method of summing the doubly indexed sequence of absolute values), the
limits are the same.
24 II. Analytic Functions

To prove that you can compose analytic functions where it makes sense, suppose
P P
f (z) = an (z − z0 )n is analytic at z0 and suppose h(z) = bn (z − a0 )n is analytic at
a0 = f (z0 ). The sum
∞
X
|am ||z − z0 |m−1 (4.2)
m=1

converges in {z : 0 < |z − z0 | < r} for some r > 0 since the series for f is absolutely
convergent, and |z − z0 | is non-zero. By the root test (set k = m − 1), this implies that
the series (4.2) converges uniformly in {|z − z0 | ≤ r1 }, for r1 < r, and hence is bounded in
{|z − z0 | ≤ r1 }. Thus there is a constant M < ∞ so that

∞
X
|am ||z − z0 |m ≤ M |z − z0 |,
m=1

if |z − z0 | < r1 , and so

∞
X X
∞ m ∞
X
n
|bm | |an ||z − z0 | ≤ |bm |(M |z − z0 |)m < ∞,
m=0 n=1 m=0

for |z − z0 | sufficiently small, by the absolute convergence of the series for h. This proves
absolute convergence for the composed series, and thus we can rearrange the doubly-
indexed series for the composition so that it is a (convergent) power series. 

As a consequence, if f is analytic at z0 and f (z0 ) 6= 0 then composing with the function

1/z which is analytic on C \ {0}, we conclude 1/f is analytic at z0 . A rational function
r is the ratio
p(z)
r(z) =
q(z)
where p and q are polynomials. The rational function r is then analytic on {z : q(z) 6= 0}.
Rational functions and their limits are really what this whole course is about.

Definition. If f is defined in a neighborhood of z then

f (w) − f (z)
f ′ (z) = lim
w→z w−z
 §4: Elementary Operations 25

is called the (complex) derivative of f , provided the limit exists.

The next Theorem says that you can differentiate power series term-by-term.

P
Theorem 4.2. If f (z) = an (z − z0 )n converges in B = {z : |z − z0 | < r} then f ′ (z)
exists for all z ∈ B and
∞
X ∞
X
′ n−1
f (z) = nan (z − z0 ) = (n + 1)an+1 (z − z0 )n ,
n=1 n=0

for z ∈ B. Moreover the series for f ′ based at z0 has the same radius of convergence as
the series for f .

Proof. If 0 < |h| < r then

P∞ n X∞ X∞
f (z0 + h) − f (z0 ) n=0 an h − a0 n−1
− a1 = − a1 = an h = an+1 hn .
h h n=2 n=1
P
By the root test, the series an+1 hn converges in a disk and hence converges uniformly
P
in {h : |h| ≤ r1 }, if r1 < r. In particular, an+1 hn is continuous at 0 and hence
∞
X
lim an+1 hn = 0.
h→0
n=1

This proves that f ′ (z0 ) exists and equals a1 .

By Theorem 3.1, f has a power series expansion about each z1 with |z1 − z0 | < r given
by
∞ X
X ∞   
n n−k
an (z1 − z0 ) (z − z1 )k
k
k=0 n=k

Therefore f ′ (z1 ) exists and equals the coefficient of z − z1

X∞   X∞
′ n n−1
f (z1 ) = an (z1 − z0 ) = an n(z1 − z0 )n−1 .
n=1
1 n=1
1
By the root test and the fact that n n → 1, the series for f ′ has exactly the same radius of
convergence as the series for f . 

Since the series for f ′ has the same radius of convergence as the series for f , we obtain
the following corollary.
26 II. Analytic Functions

Corollary 4.3. An analytic funtion f has derivatives of all orders. Moreover if f is equal
to a convergent power series on B = {z : |z − z0 | < r} then the power series is given by

X∞
f (n) (z0 )
f (z) = (z − z0 )n ,
n=0
n!

for z ∈ B.

By definition of the symbols, the n = 0 term in the series is f (z0 ).

P∞
Proof. If f (z) = n=0 an (z − z0 )n , then we proved in Theorem 4.2 that a1 = f ′ (z0 ) and

∞
X
f ′ (z) = nan (z − z0 )n−1 .
n=1

Applying Theorem 4.2 to f ′ (z), we obtain 2a2 = (f ′ )′ (z0 ) ≡ f ′′ (z0 ) and by induction

n!an = f (n) (z0 ). 

If f is analytic in a region Ω with f ′ (z) = 0 for all z ∈ Ω then by Corollary 4.3, f

is constant. In fact, if f is non-constant then by Theorem 4.2, f ′ is analytic and so the
zeros of f ′ must be isolated.

A closer examination of the idea of the proof of Theorem 4.2 shows that power series
satisfy a stronger notion of differentiability at a point. Corollary 4.4 will be used in the
next chapter for understanding the local behavior of power series.

P
Corollary 4.4. If f (z) = an (z − z0 )n converges in B = {z : |z − z0 | < r} then

f (z) − f (w)
f ′ (z0 ) = lim .
z,w→z0 z−w

Proof. Set z = z0 + h and w = z0 + k. Then for h − k 6= 0 and ε = max(|h|, |k|) < r,

X∞ X∞ n−1
X
f (z0 + h) − f (z0 + k) hn − k n
− a1 = an = an hj k n−j−1 . (4.3)
h−k n=2
h−k n=2 j=0
 §5: Exercises 27

But
M
X n
X M
X
lim |an | |h|j |k|n−j−1 ≤ lim |an |nεn−1 = 0
N,M →∞ N,M →∞
n=N j=0 n=N

by the root test. Because of uniform convergence, the right-hand side of (4.3) is a contin-
uous function of (h, k) when ε < r, vanishing at (0, 0) and hence

f (z) − f (w) f (z0 + h) − f (z0 + k)

lim = lim = a1 = f ′ (z0 ). 
z,w→z0 z−w h,k→0 h−k

P
Corollary 4.5. If f (z) = an (z − z0 )n converges in B = {z : |z − z0 | < r} then the
power series
X∞
an
F (z) = (z − z0 )n+1
n=0
n + 1

converges in B and satisfies

F ′ (z) = f (z),

for z ∈ B.

The series for F has the same radius of convergence as the series for f , by Theorem
4.2 or by direct calculation.

§5. Exercises

1. Check that examples (i)-(v) in section 2 are correct.

2. For what values of z is
X∞  n
z
n=0
1+z

convergent? Draw a picture of the region.

3. Prove the product and chain rules for analytic functions and find the derivatives of
z n , (z − a)−n , where n is a positive integer, and f /g, where f and g are analytic.
4. (a) Prove that f has a power series expansion about z0 with radius of convergence
f (z)−f (z0 )
r > 0 if and only if g(z) = z−z0 has a power series expansion about z0 , with the
28 II. Analytic Functions

same radius of convergence. (How must you define g(z0 ), in terms of the coefficients
of the series for f to make this a true statement?)
(b) It follows from (a) that if f has a power series expansion at z0 with radius of
convergence R and if |z − z0 | ≤ r < R then there is a constant C so that
P
|f (z)−f (z0 )| ≤ C|z −z0 |. Use the same idea to show that if f (z) = an (z −z0 )n
then
m
X
|f (z) − an (z − z0 )n | ≤ Dm |z − z0 |m+1 ,
n=0

where Dm is a constant and |z − z0 | ≤ r < R.

P∞ n
5. Define ez = n=0 zn! . Show
(a) this series converges for all z ∈ C
(b) ez ew = ez+w
(c) Define cos θ = 21 (eiθ + e−iθ ) and sin θ = 1
2i
(eiθ − e−iθ ), so that eiθ = cos θ + i sin θ.
Using the series for ez show that you obtain the same series expansions as you
learned in calculus for sin and cos. Check by multiplying out the definitions that
cos2 θ + sin2 θ = 1, so that eiθ is a point on the unit circle corresponding to the
cartesian coordinate (cos θ, sin θ).
(d) |ez | = eRez .
(e) ez = 1 only when z = 2πki for some integer k.
d z
(f) dz e = ez .
1
6. Prove that z n is analytic in B = {z : |z − 1| < 1} in the following sense: there is
a convergent power series f in B with the property that f (z)n = z and f (1) = 1.
P
Hint: write z = 1 + w, |w| < 1, and let g(w) = ak wk be the (formal) Taylor series
1
for (1 + w) n . Then prove |ak | ≤ 1/k so that g is analytic in |w| < 1. Use Taylor’s
theorem to show that g(x) = (1 + x)1/n for −1 < x < 1, and then use the Uniqueness
Theorem 3.2 to show that g(w)n = 1 + w. Alternatively, you can prove that z 1/n can
be defined so that it has derivatives of all orders and prove that Taylor’s theorem is
true for complex differentiable functions using complex integration.

B
 §5: Exercises 29

7. (a) If w is a complex number, find n solutions to z n = w using polar coordinates.

(b) The quadratic formula gives two solutions to the equation az 2 + bz + c = 0, when
a, b, and c are complex numbers with a 6= 0 because completing the square is
a purely algebraic manipulation of symbols, and there are two complex square
roots of every non-zero complex number by part (a). Check the details.
(c) In Exercise I.6, we found the solution to the general cubic, which has exactly
three solutions, as we will see in the next chapter. But the process appears to
generate more solutions, if we use two solutions to the quadratic and all three
cube roots. Moreover there might be more than one valid choice for the constants
used to reduce to a simpler equation. Explain.
P∞ 2
P∞ j
8. Suppose j=0 |aj | < ∞. Show f (z) = j=0 aj z is analytic in {z : |z| < 1}.

Compute (and prove your answer):

Z 2π
dθ
lim |f (reiθ )|2 .
rր1 0 2π
9. Suppose f has a power series expansion at 0 which converges in all of C. Suppose also
R
that C |f (x + iy)|dxdy < ∞. Prove f ≡ 0.
10. Suppose f is analytic in a connected open set U such that for each z ∈ U , there exists
an n (depending upon z) such that f (n) (z) = 0. Prove f is a polynomial.
11. Let f be analytic in a region U containing the point z = 0. Suppose |f (1/n)| < e−n
for n ≥ n0 . Prove f (z) ≡ 0.

12. Suppose f is analytic in a convex open set U . Suppose that for each z, w ∈ U there
exists a point ζ on the line segment between z and w with
f (z) − f (w)
= f ′ (ζ).
z−w
Prove f is a polynomial of degree at most 2. (The point is that you have to be careful:
not all calculus theorems extend to similar complex versions.)
P∞
13. Let f (z) = n=0 an z n have radius of convergence 1 and suppose an ≥ 0 for all n.
Prove that z = 1 is a singular point of f . That is, there is no function g analytic in a
neighborhood U of z = 1 such that f = g on U ∩ D.
30 II. Analytic Functions

14. (a) Define n−z = e−z ln(n) . Prove that

∞
X
ζ(z) = n−z
n=1

converges uniformly and absolutely in {z : Rez > c} for c > 1.

(b) Show that for Rez > 1,
∞ 
X Z n+1 
1 −z −z
ζ(z) − = n − x dx
z − 1 n=1 n

and prove that the sum converges uniformly and absolutely on compact subsets
of {z : Rez > 0}. Also evaluate the integral on the right-hand side to obtain a
formula for the extension of ζ(z) to Rez > 0. Here we define x−z = e−z ln x .

Probably the most famous problem in all of mathematics is to prove that if ζ(z) = 0
and 0 < Rez < 1, then Rez = 1/2.
 III

The Maximum Principle.

§1. The Maximum Principle.

We can apply the same local (“walking the dog”) analysis to an analytic function that
we applied to polynomials in Section II.1. If
∞
X
f (z) − f (z0 ) = an (z − z0 )n , (1.1)
n=k

where ak 6= 0, then setting z = z0 + h,

|f (z0 + h) − [f (z0 ) + ak hk ]| ≤ C|h|k+1 , (1.2)

for |h| sufficiently small by Exercise II.4(b). This inequality says that the value f (z) lies
inside a disk centered at f (z0 ) + ak hk of radius at most C|h|k+1 , which is much smaller
than |ak hk |. See Figure II.2. In other words,

f (z) ≈ f (z0 ) + ak (z − z0 )k . (1.3)

The next result is perhaps the most important elementary result in complex analysis.
A region is a connected open set in the plane.

Theorem 1.1 (Maximum Modulus Principle). Suppose f is analytic in a region Ω.

If there exists a z0 ∈ Ω such that

|f (z0 )| = sup |f (z)|

z∈Ω

then f is constant in Ω.

In particular a non-constant analytic function has no local maximum absolute value.

Proof. If f has a series expansion about z0 given by (1.1), write h = εeit , ε > 0. Then
ak hk = ak εk eikt traces the circle of radius |ak |εk , k times, as t increases from 0 to 2π. By

31
32 III. The Maximum Principle.

(1.2), the image f (z0 + εeit ), for ε sufficiently small, traces a curve that winds k times
around f (z0 ), and which approximately lies on the circle of radius |ak |εk centered at f (z0 ).
In particular, the image of the circle of radius ε contains points with larger absolute value
than |f (z0 )|. See Figure II.2 again.
Thus if |f | is maximal at z0 , then there is no first non-zero coefficient in the series
expansion of f (z) − f (z0 ) about z0 and so f (z) ≡ f (z0 ) in a neighborhood of z0 . If

E = {z ∈ Ω : f (z) = f (z0 )}

then we have proved E is open. But E is also closed in Ω since f is continuous. Because
E is connected, E = Ω and f is constant. 

If f (z0 ) 6= 0 then the same argument shows that |f (z0 )| is not a local minima if f is
non-constant. This fact can also be derived from the statement of the maximum principle
by considering the function 1/f which is analytic off the zeros of f .
Another form of the Maximum Principle is:

Corollary 1.2. If f is a non-constant analytic function in a bounded region Ω and if f is

continuous on Ω then
max |f (z)|
z∈Ω

occurs on ∂Ω but not in Ω.

The reader should verify the alternate form: if f is analytic on Ω then

lim sup |f (z)| = sup |f (z)|.

z→∂Ω Ω

If Ω is unbounded, then ∞ must also be viewed as a point in ∂Ω. The function f (z) = e−iz
is analytic in the upper half plane H = {z : Imz > 0}, continuous on {z : Imz ≥ 0} and
has absolute value 1 on the real line R, but is not bounded by 1 in H. See Exercise 1. A
more analytical proof of the Maximum Principle is given in Exercise 8.

§2. The Fundamental Theorem of Algebra and Partial Fractions.

The maximum principle gives an easy proof of an important result you’ve seen in some
form or another since high school.
 §2: The Fundamental Theorem of Algebra and Partial Fractions 33

Theorem 2.1 (Fundamental Theorem of Algebra). Every non-constant polynomial

has a zero.

This remarkable result says that by extending the real numbers to the complex num-
bers via the solution to the equation z 2 + 1 = 0 then every polynomial equation has a
solution.

Proof. Suppose p(z) = an z n + an−1 z n−1 + . . . + a1 z + a0 , n ≥ 1, is a polynomial which

has no zeros and for which an 6= 0. Then 1/p(z) is analytic in C and
1 1
= an−1 1 .
p(z) (an z n )(1 + an z + . . . aan0 z1n )

Since 1/z k → 0 as |z| → ∞, we conclude that given ε > 0, then for |z| = R with R
sufficiently large we have  
 1 
 
 p(z) ≤ ε.
By the Maximum Principle  
 1 

sup  ≤ ε.
|z|≤R p(z) 
1
Letting ε → 0 we conclude p(z) = 0 for all z, which is a contradiction. 

Corollary 2.2. If p is a polynomial of degree n, then there are complex numbers z1 , . . . , zn

and a complex constant c so that
n
Y
p(z) = c (z − zk ).
k=1

Corollary 2.2 does not tell us how to find the zeros, but it does say that there are
exactly n zeros.

Proof. The proof is by induction. First note that

z n − bn = (z − b)(bn−1 + zbn−2 + . . . z n−2 b + z n−1 ).

Pn
So if p(z) = k=0 ak z k and p(b) = 0, then
n k−1
p(z) p(z) − p(b) X X
q(z) ≡ = = ak ( bk−1−j z j ). (2.1)
z−b z−b j=0 k=1
34 III. The Maximum Principle.

The coefficient of z n−1 in (2.1) is an so q is a polynomial of degree n − 1. Repeating this

argument n times proves the Corollary. 

For example the polynomial z n − 1 has n zeros. If z n = 1 then 1 = |z n | = |z|n so

that |z| = 1. Write z = cos t + i sin t = eit , by Exercise II.5. Then z n = eint = 1 so that
nt = 2πk for some integer k, and thus t = 2πk/n. The n distinct zeros of z n are then
ei2πk/n , k = 0, 1, . . . , n − 1 which are equally spaced around the unit circle.
Recall that a rational function r is the ratio of two polynomials. By the Fundamental
Theorem of Algebra we can write r in the form

p(z)
r(z) = QN .
j=1 (z − zj )nj

The next Corollary, also probably familiar to you, allows us to write a rational function
in a form that is easier to analyze. The form is also of practical importance because it
allows us to solve certain differential equations that arise in Engineering problems using
the Laplace transform and its inverse.

Corollary 2.3 (Partial Fraction Expansion). If p is a polynomial then there is a

polynomial q and constants ck,j so that
nj
N X
p(z) X ck,j
QN = q(z) + . (2.2)
j=1 (z − zj )nj j=1 k=1
(z − zj )k

Proof. There are two initial cases to consider: If p is a polynomial then

p(z) p(a)
= q(z) + . (2.3)
z−a z−a

where q(z) = (p(z) − p(a))/(z − a) is a polynomial, as in (2.1). Secondly, if a 6= b, we can

write
1 A B
= + , (2.4)
(z − a)(z − b) z−a z−b
for some constants A and B. For if this equation is true, then we can multiply each term
on the right by z − a and let z → a to obtain A on the right. The same process on the left
 §2: The Fundamental Theorem of Algebra and Partial Fractions 35

yields 1/(a − b), and hence A = 1/(a − b). Similarly B = 1/(b − a). Now substitute these
values for A and B into (2.4) and check that equality holds. The full theorem now follows
by induction. Suppose the Corollary is true if the degree of the denominator is at most d.
If we have an equation of the form (2.2) then we can divide the equation by z − a. The
right side consists of lower degree terms to which the induction hypothesis applies, with
one exception: if the denominator of the left side of (2.2) is (z − b)d . If b = a then the
induction hypothesis applies to q(z)/(z − a) and the last term on the right after division
is c/(z − a)d+1 . If b 6= a, then we could have applied the inductive assumption to the
decomposition of
p(z)
(z − b)d−1 (z − a)
and then divided the result by z − b. 

The proof above also suggests an algorithm for computing the coefficients {ck,j }. First
apply (2.3) with a = z1 . Multiply each term of the result by 1/(z − b) where b is one of the
zeros of the denominator in (2.3) and apply either (2.3) or (2.4) to each of the resulting
terms on the right side. Rinse and repeat. The algorithm can be speeded up because we
know the form of the solution. For example if powers in the denominator nj are all equal
to one and if the numerator has smaller degree than the denominator, then the form is

N
X
p(z) cj
QN = . (2.5)
j=1 (z − zj ) j=1
z − zj

If we multiply each term of the right side by z −z1 then let z → z1 , we obtain c1 . If we
multiply the left side by the same factor, it cancels one of the terms in the denominator and
letting z → z1 we obtain the value of the remaining part of the left side at z1 . This quickly
gives c1 and can be repeated for c2 , . . . , cN . This method is sometimes called the “cover-up
method” because it can be done with less writing by observing that cj is the value of the left
side at zj when you cover z − zj with your hand. If the denominator has terms with degree
bigger than one, first use a denominator with all terms of degree one as above then as in
the proof, multiply everything by 1/(z −b) and simplify all terms on the right, repeating as
often as needed. If the degree of the numerator at any stage is not less than the degree of the
36 III. The Maximum Principle.

denominator, use polynomial division to reduce the degree. Engineering problems typically
have rational functions with real coefficients. See the Exercises for a similar technique that
decomposes rational functions with real coefficients into terms whose denominators are
either powers of linear terms with real zeros or powers of irreducible quadratics with real
coefficients.

§3. Local Behavior.

The maximum principle allows us to give an improved description of the local mapping
property of analytic functions.

Corollary 3.1. A non-constant analytic function defined on a region is an open map.

In other words, if f is analytic and non-constant on a region Ω and if U ⊂ Ω is open

then f (U ) is an open set.

Proof. Suppose f has a power series expansion which converges on {z : |z − z0 | < R}.
Pick r < R and set
δ= inf |f (z) − f (z0 )|.
|z−z0 |=r

Since the zeros of f − f (z0 ) are isolated, we may suppose that δ > 0 by decreasing r if
necessary. If |w − f (z0 )| < δ/2 and if f (z) 6= w for all z such that |z − z0 | ≤ r, then
1/(f − w) is analytic in |z − z0 | ≤ r and
 
 1  1 1 2
 ≤ < =
 f (z) − w  |f (z) − f (z0 )| − |w − f (z0 )| δ − δ/2 δ
on |z − z0 | = r. By the maximum principle the inequality persists in |z − z0 | < r. But
evaluating this expression at z0 we obtain the contradiction 2/δ < 2/δ. Thus the image of
the disk of radius r about z0 contains a disk of radius δ/2 about f (z0 ). This implies that
the image of an open set contains a neighborhood of each of its points. 

Definition. A function is one-to-one if f (z) = f (w) only when z = w.

A more enlightened terminology, proposed by one of my teachers Richard Arens, is

two-to-two since two points go to two points. “One-to-one” should really be the definition
of a function, but the inertia of common usage is too large to overcome.
 §3: Local Behavior. 37
p
Proposition 3.2. If f is analytic at z0 with f ′ (z0 ) 6= 0, then f is one-to-one in a suffi-
ciently small neighborhood of z0 .

Proof. By Corollary II.4.4, if zn , wn → z0 with f (zn ) = f (wn ) then f ′ (z0 ) = 0. 

Corollary 3.1 and Proposition 3.2 show that if f is analytic at z0 with f ′ (z0 ) 6= 0,
then f is a homeomorphism of a neighborhood of z0 onto a neighborhood of f (z0 ). That
is, f has a continuous inverse function in a neighborhood of f (z0 ).

Returning to our local analysis, if f ′ (z0 ) 6= 0 then by (1.3), for small |z − z0 |, the map
f approximately translates by −z0 , dilates by the factor |f ′ (z0 )| and rotates by arg f ′ (z0 ),
then translates by f (z0 ), and this approximation becomes more and more accurate as
|z − z0 | → 0.
σ
f
f (z0 ) f (σ)
α
γ α
z0

f (γ)

Figure III.1 Conformality.

To put it another way, if γ and σ are two curves passing through z0 with angle α from
γ to σ, then f (γ) and f (σ) will be curves passing through f (z0 ) and the angle from f (γ)
to f (σ) will also be equal to α.

Definition. We say that f is locally conformal if it preserves angles (including direc-

tion) between curves.

An analytic function f with f ′ 6= 0 on a region Ω is locally conformal on Ω, though not

necesarily one-to-one. Some books use the term “conformal” instead of locally conformal.
But modern usage of the phrase “conformal map” usually means a one-to-one and analytic
map. For that reason we add the qualifier “locally”.
 38 III. The Maximum Principle.
1 −
We can now picture the local correspondence given by an analytic function. By
repeated division as in Exercise II.4, if f is analytic at z0 , we can write

f (z) − f (z0 ) = an (z − z0 )n g(z),

where an is the first non-zero power series coefficient after a0 and g is analytic at z0 with
1
g(z0 ) = 1. Choose a so that an = an . By Exercise II.6, we can define z n to be analytic in
1
a neighborhood of 1. Set F (z) = g(z) n . Then

f (z) = f (z0 ) + [a(z − z0 )F (z)]n .

By Proposition 3.2, a(z − z0 )F (z) is one-to-one in a neighborhood of z0 so that we have

the following picture.

a(z − z0 )F (z) z n + f (z0 )

tn
t
z0 0 f (z0 )

Figure III.2 Local Behavior of an Analytic Function.

In Figure III.2, the case when n = 3 is illustrated. The composed function is equal to
f (z) near z0 . Each of the three regions in the left-hand figure is mapped one-to-one onto
the slit disk in the right-hand figure. Asymptotically (as the radius tends to 0) the map
“looks” like z 3 , translated and dilated. Note that Figure III.2 was constructed right-to-left.
The left side is the preimage of the right side.

P∞
Corollary 3.3. Suppose f is analytic at z0 . Then f − f (z0 ) = m=n am (z − z0 )m with
an 6= 0 if and only if for ε is sufficiently small, there exists δ > 0 so that f (z) − w has
exactly n distinct roots in {z : 0 < |z − z0 | < ε}, provided 0 < |w − f (z0 )| < δ.

The condition in Corollary 3.3 states that f (Bε (z0 )) covers Bδ (w0 ) \ {w0 } exactly n
times, where Br (ζ) is the ball centered at ζ with radius r. In particular f is one-to-one in
a neighborhood of z0 if and only if f ′ (z0 ) 6= 0.
 §4: Growth on C and D 39

§4. Growth on C and D.

In this section we will use the Maximum Principle to draw some conclusions about
the growth of analytic functions defined on the plane C or on the unit disk D.

Corollary 4.1 (Liouville’s Theorem). If f is analytic in C and bounded, then f is

constant.

Proof. Suppose |f | ≤ M < ∞. Set g(z) = (f (z) − f (0))/z. Then g is analytic and |g| → 0
as |z| → ∞. By the maximum principle g ≡ 0 and hence f ≡ f (0). 

Here is a typical use of Liouville’s Theorem. If p is a polynomial of degree n, then

|p(z)| ≤ C|z|n for sufficiently large |z|. If f is analytic in C and if |f (z)| ≤ C|z|n for
Pn
|z| > M , then let p(z) = k=0 ak z k be the terms of the power series expansion of f at 0
up to degree n. Then g(z) = (f (z) − p(z))/z n is analytic in C and bounded. By Liouville’s
Theorem, g is constant and hence f must be a polynomial. In fact, f = p since g(0) = 0.

Corollary 4.2 (Schwarz’s Lemma). Suppose f is analytic in D = {z : |z| < 1} and

suppose |f (z)| ≤ 1 and f (0) = 0. Then

|f (z)| ≤ |z|, (4.1)

for all z ∈ D and

|f ′ (0)| ≤ 1. (4.2)

Moreover, if equality holds in (4.1) or (4.2) then f (z) = cz where c is a constant with
|c| = 1.

In some sense, Schwarz’s Lemma says that a bounded analytic function can’t grow
too fast in the disk.

Proof. By Exercise II.4, the function g given by


 f (z) if z ∈ D \ {0}

g(z) = z

 ′
f (0) if z = 0
40 III. The Maximum Principle.

is analytic in D and
1
sup |g(z)| ≤ .
|z|=r r

Fix z0 ∈ D, then for r > |z0 |, the maximum principle implies |g(z0 )| ≤ 1r , so that, letting
r → 1, we obtain (4.1) and (4.2). If equality holds in (4.1) or (4.2) then g(z) has a
maximum at z0 and hence is constant. 

Corollary 4.3 (Invariant form of Schwarz’s Lemma). Suppose f is analytic in D =

{z : |z| < 1} and suppose |f (z)| ≤ 1. If z, a ∈ D then
   
 f (z) − f (a)   z − a 
   
 1 − f (a)f (z) ≤  1 − az , (4.3)

and
|f ′ (z)| 1
2
≤ . (4.4)
1 − |f (z)| 1 − |z|2

Proof. If |c| < 1, the function

z−c
Tc (z) =
1 − cz
is analytic except at z = 1/c. Moreover for z = eit ∈ ∂D,
 it 
 e − c  |eit − c|
it 
|Tc (e )| =  = = 1.
1 − ceit  |e−it − c|

By the maximum principle (or direct computation), |Tc | ≤ 1 on D. Setting c = f (a), the
composition Tc ◦ f is analytic on D and bounded by 1. Furthermore
  
Tc ◦ f (z) f (z) − f (a) 1 − az
=
Ta (z) 1 − f (a)f (z) z−a

is analytic on D and
 
 Tc ◦ f (z) 
lim sup = lim sup |Tc ◦ f (z)| ≤ 1,
|z|→1 Ta (z)  |z|→1

By the maximum principle, (4.3) holds. Inequality (4.4) follows by dividing both sides of
(4.3) by z − a and letting z → a. 
 §4: Growth on C and D 41

An alternative proof is to apply Schwarz’s Lemma to g(w) = Tc ◦ f ◦ T−a (w) then set
w = Ta (z). The details are almost the same.

Corollary 4.4. If f is analytic on D, |f | ≤ 1, and f (zj ) = 0, for j = 1, . . . , n then

n 
Y 
z − zj
f (z) = g(z),
1 − zj z
j=1

where g is analytic in D and |g(z)| ≤ 1 on D.

Proof. If f (a) = 0, then by the proof of Corollary 4.3, g(z) = f (z)/Ta (z) is analytic in
the disk and bounded by 1. Repeating this argument n times proves the Corollary. 

By the Uniqueness theorem, if f is analytic on D, then the zeros of f do not cluster

in D, unless f is identically zero. If we have more restrictions on f , such as boundedness,
then the zeros cannot approach the unit circle too slowly by the next Corollary.

Corollary 4.5. If f is non-constant, bounded and analytic in D and if {zj } are the zeros
of f then
X
(1 − |zj |) < ∞.
j

Proof. We may suppose |f | ≤ 1, by dividing f by a constant if necessary. If f (0) 6= 0,

then using the notation of the proof of Corollary 4.4,
n
Y n
Y
|f (0)| = ( |zj |)|g(0)| ≤ |zj |,
j=1 j=1

so that by taking logarithms (base e),

X n X n
1 1
log ≥ log ≥ (1 − |zj |).
|f (0)| j=1 |zj | j=1

If f (0) = 0, then write f (z) = z k h(z) where h(0) 6= 0. Applying the preceeding argument
to h, we obtain
n
X 1
(1 − |zj |) ≤ log + k.
j=1
|h(0)|
42 III. The Maximum Principle.

The Corollary follows by letting n → ∞. 

Much of what we’ll do in this class takes place on D or on C, which look like rather
special domains, but we know a power series converges on a disk, and by translating and
scaling the domain, we can assume it is D or C. Also in the third quarter, we’ll prove the
uniformization theorem which says in some sense the only analytic functions we need to
understand can be defined on D or C.

§5. Boundary Behavior.

We conclude this chapter with some examples and a theorem about boundary behavior
of analytic functions on the unit disk.

The first example is

z+1
I(z) = e z−1 .

Since I is the composition of an analytic function on C \ {1} and the exponential function,
which is analytic in C, I is analytic on C \ {1}, by Exercise II.5. Moreover |ez | = eRez so
that by a computation
|z|2 −1
|I(z)| = e |z−1|2 .

Thus |I(z)| ≤ 1 on D. On the unit circle I(eit ) = e−i cot(t/2) , for 0 < t < 2π. In particular
if ζ ∈ ∂D \ {1}, then

lim I(z)
z→ζ

r+1
exists and has absolute value 1. However, for 0 < r < 1, I(r) = e r−1 → 0 as r → 1. On
the unit circle I(eit ) is spinning rapidly as t → 0. Hence I(z) does not have a limit as
z → 1. The proper way to take limits in the disk is through cones. For ζ ∈ ∂D and α > 1,
define

Γα (ζ) = {z ∈ D : |z − ζ| ≤ α(1 − |z|)}.

to be the Stolz angle or cone at ζ.

 §5: Boundary Behavior 43
1 − |z|

ζ
z |z − ζ|
Γα (ζ)
0
2 sec−1 α

Figure III.3 Stolz angle Γα (ζ).

The precise shape of Γα is not important except that it is symmetric about the line
segment [0, ζ] and forms an angle less than π at ζ. For z ∈ Γα (1) we have
−(1−|z|) (1+|z|) 1
|I(z)| = e |z−1| |z−1| ≤ e− α|z−1| → 0

as z ∈ Γα (1) → 1. Thus I(z) → 0 in every cone, and ∪α Γα (1) = D, but there is still no
limit as z → 1.

Another example is the function given by

∞
X n
L(z) = z2 .
n=0

This series converges uniformly and absolutely in D. It is sometimes called a lacunary

series since the spacing between nonzero coefficients increases as n increases. The function
k
L does not extend continuously to any point of ∂D. For example if ζ 2 = 1 then
k−1
X ∞
X
2n n
L(rζ) = (rζ) + r2 .
n=0 n=k

The second sum is clearly positive and increasing to ∞ as r → 1. Since the 2k roots of 1
are evenly spaced around ∂D, if eit ∈ ∂D, then we can find a ζ as close to eit as we like
k
with ζ 2 = 1. Thus in any neighborhood of eit , f is unbounded.

The next Theorem gives a connection between Fourier series and analytic functions
in D.
44 III. The Maximum Principle.
P∞
Theorem 5.1 (Abel’s Limit Theorem). If ζ = eit ∈ ∂D and if n=0 an eint converges,
P∞
then f (z) = n=0 an z n converges in D and if Γ = Γα (ζ) is any Stolz angle at ζ then

∞
X
lim f (z) = an eint .
z∈Γ→ζ
n=0

The convergence in Theorem 5.1 is called nontangential convergence. Abel’s The-

orem says that you get what you would expect for a limit, but only nontangentially. Note
that the limit does not depend on which Stolz angle you choose. The power series

1  3n 
X∞
n
f (z) = z − (z 2 )3
n=1
n

k
converges at z = 1 and hence converges in |z| < 1 by the root test. Set ζk = eiπ/3 . Then
|f (rζk )| → ∞ as r ↑ 1, so we may choose zk = rk ζk → 1 so that |f (zk )| → ∞. This says
that we cannot conclude unrestricted convergence in Abel’s Theorem.

Proof. Replacing f (z) by f (ζz), we may suppose ζ = 1 and by subtracting a constant

P∞
from a0 we may suppose n=0 an = 0. By the root test, the series for f converges in D.
Suppose z ∈ Γα (1). Then
∞
X ∞
X X∞ X
n 
f (z) n k
= z ak z = ak z n .
1 − z n=0 n=0
k=0 k=0

Pn
Set sn = k=0 ak . Then

N−1   ∞ 
X  X 

|f (z)| ≤ |1 − z| n 
sn z +|1 − z| n
sn z .
n=0 n=N

Given ε > 0, there exists N < ∞ so that |sn | < ε for n ≥ N . Thus

N−1
X ∞
X N−1
X |1 − z||z|N ε
|f (z)| ≤ |1 − z| |sn | + |1 − z|ε |z|n = |1 − z| |sn | + .
n=0 n=0
1 − |z|
n=N

We can choose δ > 0 so that if z ∈ Γα (1) and |z − 1| < δ,

|f (z)| ≤ ε + αε.
 §6: Exercises 45

Since ε > 0 was arbitrary,

lim |f (z)| = 0. 
z∈Γα →1

P∞
For example the series n=1 z n /n converges at z = −1 by the alternating series test.
By Taylor’s Theorem this series converges to ln(1 − x) for −1 < x < 1. By Abel’s Limit
P∞
Theorem, n=0 (−1)n /n = ln 2.

A Fourier Series is a function of the form

∞
X
f (t) = an eint .
n=−∞

if f (t) converges then

∞
X ∞
X
n
f (z) = an z and g(z) = a−n z n
n=0 n=0

converge and are analytic on D by the root test. By Abel’s Theorem, f (reit ) + g(reit )
converges to F (t) at each t where the series for F converges. Thus the function f + g
“extends” F to D, and the infinitely differentiable functions defined on [0, 2π] by

fr (t) = f (reit ) and gr (t) = g(reit )

satisfy
fr + g r → F

as r → 1, at each t where the series for F converges.

Fourier Series arose from attempting to solve certain differential equations. Each
square integrable function on [0, 2π] has a Fourier Series, and it was a famous problem for
many years to prove that the Fourier series converges almost everywhere on [0, 2π]. The
proof that was eventually found remains perhaps one of the hardest proofs in analysis, but
it used this connection between Fourier Series and analytic functions on the disk, and the
relation between convergence on the circle and nontangential convergence.

§6. Exercises.
46 III. The Maximum Principle.

1. (a) Prove Corollary 1.2.

(b) Prove that if f is analytic on a region Ω then

sup |f (z)| = lim sup |f (z)|.

z∈Ω z→∂Ω

We say that a sequence tends to ∂Ω if it is eventually outside each compact subset

of Ω. The lim sup is then the largest subsequential limit of the values of |f | over
all sequences tending to ∂Ω. If Ω is unbounded, we view our sets as lying on
the Riemann sphere, so that the boundary includes the North Pole (the point
“at ∞”). Equivalently we can measure the distance to the boundary using the
chordal metric given at the end of Section I.3.
2. (a) If p is a polynomial with real coefficients, prove that p can be factored into a
product of linear and quadratic factors, each of which has real coefficients, and so
that the quadratic factors are non-zero on R. Most Engineering problems involving
polynomials only need polynomials with real coefficients.
(b) For rational functions with real coefficients, such as those that typically occur in
applications, it is sometimes preferable to use a partial fraction expansion without
complex numbers in the expression. The cover-up method can also be used in
this case. Here is an example to illustrate the idea. If a, b, c, d, and e are real
show that
z 2 + dz + e A B(z − b) + D
2 2
= + ,
(z − a)((z − b) + c ) z−a (z − b)2 + c2
where A, B, and D are real. Here we have completed the square for the irreducible
quadratic factor. Notice also that we have written the numerator of the last term
as B(z − b) + D, not Bz + D. We can find A by the usual cover-up method.
Then to find B and D, we multiply by (z − b)2 + c2 and let it tend to 0. Thus
z → b ± ic. Cover up the quadratic factor in the denominator on the left and
let z → b + ic. On the right side, when we multiply by the quadratic factor, the
first term will tend to 0, the denominator of the second term will be cancelled
and B(z − b) + D tends to Bic + D. Thus the real part of the result on the left
 §6: Exercises 47

equals D and the imaginary part equals Bc, and then we can immediately write
down the coefficients B and D. Try this process with two different irreducible
quadratic factors in the denominator, and you’ll see how much faster and accurate
it is than solving many equations with many unknowns. The choice of the form
of the numerator at B(z − b) + D instead of Bz + D made this computation a bit
easier. It also turns out that it makes it a bit easier to compute inverse Laplace
transforms of these rational functions, because the resulting term is a shift in the
domain of a simpler function.
3. Find the series expansion of
z + 2i
(z − 2)(z 2 + 1)
about the point 1.
4. Suppose f is analytic in a connected open set U . If |f (z)| is constant on U , prove that
f is constant on U . Likewise, prove that f is constant if Ref is constant.
5. Suppose f and g are analytic in C and |f (z)| ≤ |g(z)| for all z. Prove there exist a
constant c so that f (z) = cg(z) for all z.
6. Prove that if f is non-constant and analytic on all of C then f (C) is dense in C.
7. Let f be analytic in D and suppose |f (z)| < 1 on D. Let a = f (0). Show that f does
not vanish in {z : |z| < |a|}
8. (a) Prove the Mean Value Property for analytic functions: If f is analytic at z0 and
if r < R, where R is the radius of convergence of its power series at z0 , then
Z 2π
dt
f (z0 ) = f (z0 + reit ) .
0 2π

(b) Use (a) and Exercise I.6(a) to prove the Maximum Principle for analytic functions.
A colleague calls this the “everyone less than average” proof (cf. Lake Wobegon).

9. Prove that if f is a one-to-one (two-to-two!) analytic map of an open set Ω onto f (Ω)
and if zn ∈ Ω → ∂Ω, then f (zn ) → ∂f (Ω), in the sense that f (zn ) eventually lies
outside each compact subset of f (Ω). Another way to state the problem is to view
48 III. The Maximum Principle.

the sets as lying on the Riemann sphere, so that the boundary can include the North
Pole (the point at “∞”).
10. (a) Prove that ϕ is a one-to-one analytic map of D onto D if and only if
 
z−a
ϕ(z) = c ,
1 − āz

for some constants c and a, with |c| = 1, and |a| < 1. What is the inverse map?
(b) Let f be analytic in D and satisfy |f (z)| → 1 as |z| → 1. Prove f is rational.
11. (a) Suppose p is a polynomial with all its zeros in the upper half plane H = {z : Imz >
0}. Prove that all of the zeros of p′ are contained in H. Hint: Look at the partial
fraction expansion of p′ /p.
(b) Use (a) to prove that if p is a polynomial then the zeros of p′ are contained in the
(closed) convex hull of the zeros of p. (The closed convex hull is the intersection
of all half planes containing the zeros.)
12. Suppose f is analytic in D and |f (z)| ≤ 1 in D and f (0) = 1/2. Prove that |f (1/3)| ≥
1/5.
13. Suppose f is analytic and non-constant in D and |f (z)| ≤ M on D. Prove that the
number of zeros of f in the disk of radius 1/4, centered at 0, does not exceed
 
1  M 
log .
log 4 f (0) 

14. Suppose f is bounded and analytic in the right half-plane {z : Rez > 0}, and
lim supz→iy |f (z)| ≤ M for all iy on the imaginary axis. Prove |f (z)| ≤ M on the
right half-plane. Check that f (z) = ez satisfies all the hypotheses above, except for
boundedness, and fails to be bounded in the right half-plane.
 IV

Integration and Approximation

§1. Integration on Curves.

Definition. A curve is a continuous mapping of an interval I ⊂ R into C.

Different curves can have the same image. For example if γ(t) : [0, 1] → C then
γ(t2 ) : [0, 1] → C and both curves have the same image. Many times we will write

Figure IV.1 A curve γ.

and call the curve γ. But we really mean a particular choice (though unstated) of the
parameterization. Arrows show how γ(t) traces the image as t ∈ I increases in Figure
IV.1.

Definition. A curve γ is called an arc if it is one-to-one (two-to-two!).

Definition. A curve γ : [a, b] → C is called closed if γ(a) = γ(b).

Definition. A closed curve γ : [a, b] → C is called simple if γ restricted to [a, b) is

one-to-one.

A simple closed curve γ : [0, 2π] → C can also be viewed as a one-to-one continuous
mapping of the unit circle given by ψ(eit ) = γ(t).

49
50 IV. Integration and Approximation

Definition. A curve γ = x + iy is called piecewise continuously differentiable if

γ ′ (t) = x′ (t) + iy ′ (t)

exists and is continuous except for finitely many t and x′ and y ′ have one-sided limits at
the exceptional points.

From now on, all curves will be assumed to be piecewise continuously differentiable
unless stated otherwise. In particular, by the Fundamental Theorem of Calculus if γ
is piecewise continuously differentiable then
Z t2 Z t2
′
γ(t2 ) − γ(t1 ) = (x(t2 ) − x(t1 )) + i(y(t2 ) − y(t1 )) = x (t)dt + i y ′ (t)dt
t1 t1

Definition. A curve ψ : [c, d] → C is called a reparameterization of a curve γ : [a, b] →

C if there exists an increasing function α : [a, b] → [c, d] such that ψ(α(t)) = γ(t).

For example ψ(t) = t2 + it4 , for 0 ≤ t ≤ 1, is a reparameterization of γ(t) = t + it2 ,

1
for 0 ≤ t ≤ 1, with α(t) = t 2 . If σ : [0, 1] → C is a curve then the curve β, defined
by β(t) = σ(1 − t), is not a reparameterization of σ because 1 − t is decreasing. If
ψ is a reparameterization of γ with ψ(α(t)) = γ(t), where α is piecewise continuously
differentiable then ψ ′ (α(t))α′ (t) = γ ′ (t), by the chain rule applied to the real and imaginary
parts, or by taking limits of difference quotients.

Definition. If γ : [a, b] → C is a curve and if f is a continuous complex-valued function

defined on (the image of) γ then

Z Z b
f (z)dz ≡ f (γ(t))γ ′(t)dt.
γ a
The reader can check that a reparameterization of γ will not change the integral by
the chain rule, and so the integral really depends on the image of γ, not the choice of
R
parameterization and for that reason we use the notation γ f dz. Note, however, that the
direction of the image curve is important. If ψ(−t) = γ(t), then
Z Z
f (z)dz = − f (z)dz.
ψ γ
 §1: Integration on Curves. 51
R
Another reason for the notation γ
f dz is the following.
Suppose a = t0 < t1 < t2 < . . . < tn = b and set γ(tj ) = zj . Then

n−1
X n−1
X
f (zj )(zj+1 − zj ) = f (γ(tj ))[γ(tj+1 ) − γ(tj )]
j=0 j=0
n−1
X
≈ f (γ(tj ))γ ′ (tj )[tj+1 − tj ],
j=0

and the latter is a Riemann sum using the independent variable t for
Z b Z
′
f (γ(t))γ (t)dt = f (z)dz.
a γ

R R R
Integration is linear γ (f (z) + g(z))dz = γ f (z)dz + γ g(z)dz, and if C is constant
R R
then γ Cf (z)dz = C γ f (z)dz. If a closed curve γ can be written as the union of two
R R R
curves γ1 and γ2 , then γ f (z)dz = γ1 f (z)dz + γ2 f (z)dz, so that as far as integration
is concerned, it doesn’t matter if we parametrize γ as γ1 followed by γ2 or as γ2 followed
by γ1 . So it does not matter which point of a closed curve is considered as the “starting
point”.

Definition. If γ : [a, b] → C is a curve and if f is a continuous complex-valued function

defined on (the image of) γ then
Z Z b
f (z)|dz| = f (γ(t))|γ ′(t)|dt.
γ a

Thus Z  Z 
   b 
 f (z)dz  =  f (γ(t))γ (t)dt ′
  
γ a

Z b Z
′
≤ |f (γ(t))||γ (t)|dt = |f (z)||dz|.
a γ

Note that if zj = γ(tj ) then

X
f (zj )|zj+1 − zj |
R Rb
is a Riemann sum for γ
f (z)|dz| = a
f (γ(t))|γ ′(t)|dt.
52 IV. Integration and Approximation

Definition. If γ : [a, b] → C is a curve then the length of γ is defined to be

Z
ℓ(γ) = |γ| = |dz|.
γ

Note that we are assuming all curves are piecewise C 1 on a closed interval in R so
that all curves have finite length.
The following important estimate follows immediately from the definitions:
Z   
 
 f (z)dz ≤ sup |f (z)| ℓ(γ). (1.1)
  γ
γ

Consequently, if fn converges uniformly to f on γ and if ℓ(γ) < ∞ then

Z Z
lim fn (z)dz = f (z)dz.
n γ γ

To prove this just use the estimate:

Z  
| (fn − f )dz| ≤ sup |fn − f | ℓ(γ).
γ γ
Pn
Definition. If γ1 , . . . , γn are curves, then we denote their union by j=1 γj , so that for
any continuous function f ,
Z n Z
X
Pn f (z)dz = f (z)dz. (1.2)
γj j=1 γj
j=1

If γ1 ends where γ2 begins then γ1 + γ2 can be parametrized as a curve, but we also

allow unions of curves which are not connected in the definition above.
γ1 (1) = γ2 (0)

γ1
γ2

Figure IV.2 A curve γ1 + γ2 .

Definition. A cycle γ is a finite union of closed curves γ1 , . . . , γn .

If γ is a cycle as in the definition above, we can write γ = γ1 + . . . + γn . Equation

(1.2) then holds for cycles.
 §2: Equivalence of Analytic and Holomorphic. 53
z)
Definition. If γ : [a, b] → C is a curve, then −γ : [−b, −a] → C is the curve defined by

−γ(t) = γ(−t).

The curve −γ has the same geometric image as γ, but it is traced in the opposite
direction. If f is continuous on γ, then
Z Z
f (z)dz = − f (z)dz.
−γ γ

R R R
In particular γ−γ
f (z)dz = γ
f (z)dz − γ
f (z)dz = 0. This idea can be used to
simplify some integrals. For example, the integral around two adjacent squares, each in
the counter-clockwise direction, is equal to the integral around the boundary of the union
of the squares.

∂(S 1 ∪ S 2 )

S1 S2

Figure IV.3 Integrals around squares.

In the left side of Figure IV.3, the boundary of the squares, ∂S1 and ∂S2 , are param-
eterized in the counter-clockwise direction and
Z Z Z
f (z)dz + f (z)dz = f (z)dz,
∂S1 ∂S2 ∂(S1 ∪S2 )

for every continuous function f defined on ∂S1 ∪ ∂S2 . This can be seen by writing the
integrals around each square as the sum of integrals on the bounding line segments. The
common boundary edge is traced in opposite directions, so the corresponding integrals will
54 IV. Integration and Approximation

cancel. A similar argument applies to a finite union of squares, so that after cancellation,
the sum of the integrals around the boundaries of all the squares is equal to the integral
around the boundary of the union of the squares.

§2. Equivalence of Analytic and Holomorphic.

Definition. A complex valued function f is said to be holomorphic on an open set U if

f (w) − f (z)
f ′ (z) = lim
w→z w−z

exists for all z ∈ U and is continuous on U . A complex valued function f is said to be

holomorphic on a set S if it is holomorphic on an open set U ⊃ S.

There are various apriori weaker conditions for analyticity. For example, many books
do not require continuity of the derivative in the definition of a holomorphic function.
In almost every situation encountered in practice though, verifying that the derivative is
continuous, once you have proved it exists, is not hard. However, it was an important
advance in partial differential equations to consider “weak” derivatives in the sense of
distributions. Indeed, it led to the development of Functional Analysis. We will treat
these more advanced topics in Chapter XVIII where weaker versions of “holomorphic” are
considered.
As we saw in section II.4, analytic functions are holomorphic. In particular polyno-
mials are holomorphic. A rational function is holomorphic except where the denominator
is zero. Linear combinations of holomorphic functions are holomorphic. The reader is
invited to verify that the chain rule for complex differentiation holds for the composition
of two holomorphic functions, and so the composition of two holomorphic functions is
holomorphic, wherever the composition is defined.
It also follows from the usual chain rule applied to real and imaginary parts that if
γ : [a, b] → C is a piecewise continuously differentiable curve and if f is holomorphic on a
neighborhood of γ then f ◦ γ is a piecewise continuously differentiable curve and

d
f (γ(t)) = f ′ (γ(t))γ ′(t).
dt
 §2: Equivalence of Analytic and Holomorphic. 55

Then by the Fundamental Theorem of Calculus,

Z Z b Z b
′ ′ ′ d
f (z)dz = f (γ(t))γ (t)dt = f (γ(t))dt = f (γ(b)) − f (γ(a)).
γ a a dt
For example the line segment from z to ζ can be parameterized by γ(t) = z + t(ζ − z),
for t ∈ [0, 1], so that if f is holomorphic in a neighborhood of γ, then
Z 1
f (ζ) − f (z) = f ′ (z + t(ζ − z))(ζ − z)dt. (2.1)
0

Corollary 2.1. If γ : [a, b] → C is a closed curve and if f is holomorphic in a neighborhood

of γ then Z
f ′ (z)dz = 0.
γ

Proof. Because γ is closed, f (γ(b)) − f (γ(a)) = 0. 

R
For example, if γ is a closed curve then γ
p(z)dz = 0 for every polynomial p.
P∞
Corollary 2.2. If f (z) = n=0 an (z − z0 )n converges in B = {z : |z − z0 | < r} and if
γ ⊂ B is a closed curve then Z
f (z)dz = 0.
γ

Proof. The series expansion for f converges uniformly on γ, so Corollary 2.2 follows by
interchanging the order of the integral and the sum. 

Much of this chapter and the next center around extending Corollary 2.2 to larger
sets than B.

If γ is a closed curve and a ∈

/ γ then by Corollary 2.1, if n > 1
Z
1
n
dz = 0.
γ (z − a)
By the partial fraction expansion, Corollary III.2.3, in order to integrate a rational function
R
along γ we only need to be able to compute γ (z − a)−1 dz for various values of a. The
next example will be key to understanding integrals of analytic functions. If r > 0 set

Cr = {z0 + reit : 0 ≤ t ≤ 2π},

then we have
56 IV. Integration and Approximation

Proposition 2.3. 
Z  1 if |a − z0 | < r
1 1
dz =
2πi Cr z−a 
0 if |a − z0 | > r.

Proof. Suppose |a − z0 | < r. Then Cr′ (t) = ireit and

Z Z 2π
1 1 1 1
dz = ireit dt
2πi Cr z−a 2πi 0 reit − (a − z0 )

Z 2π
1 1
= dt
2π 0 1 − ( a−z
reit
0
)

Z ∞ 
2π X n
1 a − z0
= dt
2π 0 n=0
reit

X∞ Z 2π
(a − z0 )n 1
= n
e−int dt = 1.
n=0
r 2π 0

Interchanging the order of summation and integration is justified since |(a − z0 )/(reit )| < 1
implies uniform convergence of the series.
If |a − z0 | > r, then write
  ∞ ∞
reit reit 1 reit X r n eint X r n eint
= it = − = − ,
reit − (a − z0 ) re
z0 − a 1 − a−z a − z0 n=0 (a − z0 )n n=1
(a − z0 )n
0

so that
Z Z 2π X∞ Z 2π
1 1 1 reit rn 1
dz = it
dt = − n
eint dt = 0.
2πi Cr z−a 2π 0 re − (a − z0 ) n=1
(a − z0 ) 2π 0


An immediate consequence of Corollary 2.1 and Proposition 2.3 is that there is no

function f defined in a neighborhood of ∂D satisfying f ′ (z) = 1/z.

Theorem 2.4. If f is holomorphic on {z : |z − z0 | ≤ r} then for |z − z0 | < r,

Z
1 f (ζ)
f (z) = dζ.
2πi Cr ζ − z
 §2: Equivalence of Analytic and Holomorphic. 57

where Cr is the circle of radius r centered at z0 , parameterized in the counter-clockwise

direction.

Theorem 2.4 shows that it is possible to find the values of a holomorphic function
inside a disk from the values on the bounding circle.

Proof. By (2.1) and Corollary 2.1, for |z − z0 | < r,

Z Z Z 1
f (ζ) − f (z)
dζ = f ′ (z + t(ζ − z))dtdζ
Cr ζ −z Cr 0

Z 1 Z
= f ′ (z + t(ζ − z))dζdt
0 Cr

Z 1 Z
d dt
= lim f (z + t(ζ − z))dζ = 0.
ε→0 ε Cr dζ t

Thus Z Z
1 f (ζ) 1 dζ
dζ = f (z) · = f (z),
2πi Cr ζ −z 2πi Cr ζ −z
by Proposition 2.3. 

A consequence of Theorem 2.4 is the converse to Theorem II.4.2.

Corollary 2.5. A complex-valued function f is holomorphic on a region Ω if and only if

f is analytic on Ω. Moreover, the series expansion for f based at z0 ∈ Ω converges on the
largest open disk centered at z0 and contained in Ω.

Proof. If f is analytic in Ω then f is holomorphic in Ω, by Theorem II.4.2. To prove the

converse, suppose f is holomorphic on {z : |z − z0 | ≤ r}. If |z − z0 | < r, then by II.1.2
Z Z X
∞ 
1 f (ζ) 1 1 n
f (z) = dζ = n+1
(z − z0 ) f (ζ)dζ
2πi Cr ζ −z 2πi Cr n=0
(ζ − z0 )

X∞  Z 
1 f (ζ)
= n+1
dζ (z − z0 )n .
n=0
2πi Cr (ζ − z0 )
58 IV. Integration and Approximation

Interchanging the order of the summation and integral is justified by the uniform con-
vergence of the series for z fixed. Thus f has a power series expansion convergent in
{z : |z − z0 | ≤ r}, provided this closed disk is contained in Ω. 

In particular, if f is analytic in C then f has a power series expansion which converges in

all of C. Such functions are called entire. From now on, we will use the words analytic
and holomorphic interchangably.

The proof of Corollary 2.5 yields a bit more information. Not only can we find the
values of an analytic function inside a disk from it values on the boundary, but we also
have a formula for each of its derivatives in the disk.

Corollary 2.6 (Cauchy’s Estimate). If f is analytic in {z : |z − z0 | ≤ r} and Cr (z0 ) =

{z0 + reit : 0 ≤ t ≤ 2π}, then
Z
f (n) (z0 ) 1 f (ζ)
= dζ, (2.2)
n! 2πi Cr (z0 ) (ζ − z0 )n+1

and  (n) 
 f (z0 )  supCr (z0 ) |f |
 
 n! ≤ rn
. (2.3)

Proof. Equation (2.2) follows from Corollary II.4.3, the proof of Corollary 2.5, and the
Uniqueness Theorem II.3.2. Inequality (2.3) follows from (2.2) by using inequality (1.1).

The equivalence of analytic and holomorphic makes it easy to prove that the inverse
of a one-to-one analytic function is analytic.

Corollary 2.7. If f is analytic and one-to-one in a region Ω then the inverse of f , defined
on f (Ω), is analytic.

Proof. Since analytic functions are open by Corollary III.3.1, f has a continuous inverse.
If z0 ∈ Ω, then by Corollary III.3.3, f ′ (z0 ) 6= 0 and f maps small neighborhoods of z0 onto
 §2: Equivalence of Analytic and Holomorphic. 59

small neighborhoods of w0 = f (z0 ). If |h| sufficiently small, then we can find z1 near z0 so
that f (z1 ) = w0 + h. This implies

f −1 (w0 + h) − f −1 (w0 ) z1 − z0 1
= → ′ ,
h f (z1 ) − f (z0 ) f (z0 )

as h → 0 and hence f −1 has a complex derivative at w0 equal to 1/f ′ (f −1 (w0 )). This
derivative is continuous, so that f −1 is holomorphic and hence analytic. 

As an application of the second sentence of Corollary 2.5, we obtain a local version of

a theorem we will encounter in the next chapter, called Cauchy’s Theorem.

Corollary 2.8. If f is analytic in an open disk B and if γ ⊂ B is a closed curve, then

Z
f (z)dz = 0.
γ

Proof. By Corollary 2.5, f has a power series expansion which converges on B. Now
apply Corollary 2.2. 

Morera’s Theorem is a useful converse to Corollary 2.8.

Theorem 2.9 (Morera). If f is continuous in an open disk B and if

Z
f (ζ)dζ = 0
∂R

for all closed rectangles R ⊂ B with sides parallel to the axes, then f is analytic on B.

Proof. We may suppose B = D. Define

Z
F (z) = f (ζ)dζ,
γz

where γz is a curve from 0 to z consisting of a horizontal line segment followed by a vertical

line segment.
60 IV. Integration and Approximation

z+h
σ
z

0
Figure IV.4 Proof of Morera’s Theorem.

If |h| < 1 − |z| then γz+h = γz + σ + ∂R where σ is a curve from z to z + h consisting

of a horizontal line segment followed by a vertical line segment and R ⊂ B is a closed
R
rectangle. See Figure IV.4. Because ∂R f (ζ)dζ = 0, we have
Z Z Z Z
F (z + h) − F (z) = f (ζ)dζ − f (ζ)dζ = f (ζ)dζ = f (ζ)dζ.
γz+h γz σ σ

By the Fundamental Theorem of Calculus, since the identity function has derivative equal
R
to 1, σ dζ = z + h − z = h, and so.
Z
F (z + h) − F (z) 1
− f (z) = (f (ζ) − f (z))dζ.
h h σ

By (1.1),  Z 
1  √
 (f (ζ) − f (z))dζ  ≤ 2 sup |f (ζ) − f (z)|,
h 
σ ζ∈σ
√
because |σ| ≤ 2|h|. Since f is continuous, letting h → 0 proves that F is holomorphic
on B with F ′ = f . By Corollary 2.5, F is analytic on B and by Theorem II.4.2, f = F ′ is
analytic on B. 

One consequence of Morera’s Theorem is that the definition of holomorphic does not
need to include the continuity of the derivative. We put the proof in Chapter XVIII
because we do not know of a good application. The proof can be read now since it only
depends on material that we have covered.

§3. Approximation by Rational Functions.

 §3: Approximation by Rational Functions. 61

In this section we will show that Theorem 2.4 also holds if the circle Cr is replaced by
the boundary of a square, and then use it to prove Runge’s theorem that analytic functions
can be uniformly approximated by rational functions.

Proposition 3.2. If S is an open square with boundary ∂S parameterized in the counter-

clockwise direction then


Z  1 if a ∈ S
1 1
dz =
2πi ∂S z−a 
0 if a ∈ C \ S.

Proof. If a ∈ C \ S, then we can find a disk B which contains S and does not contain a.

a
B S

Figure IV.5 If a is outside S.

By II.2.2 and Corollary 2.8,

Z
1
dz = 0.
∂S z−a

If a ∈ S, then let C be the circumscribed circle to ∂S parameterized in the clockwise

direction.
62 IV. Integration and Approximation

c2

s2

c3 s3 s1 c1

a B1
s4

c4
Figure IV.6 The square S and its circumscribed circle C.

Then we can write

∂S = s1 + s2 + s3 + s4 ,

where sj , j = 1, . . . , 4, are the sides of ∂S and

C = c1 + c4 + c3 + c2 ,

where cj , j = 1, . . . , 4 are the arcs of C subtended by the corresponding sides of ∂S. Then
sj + cj is a closed curve contained in a disk Bj with a ∈
/ Bj , for j = 1, . . . , 4. By Corollary
2.8,
Z
1
dz = 0, (3.2)
sj +cj z−a

for j = 1, . . . , 4. Adding the four integrals (3.2) we obtain

Z Z Z
1 1 1
dz + dz = dz = 0.
∂S z−a C z−a ∂S+C z−a

But we also have by Proposition 2.3

Z
1
dz = −2πi,
C z−a

because −C is the circle parameterized in the counter-clockwise direction. This proves

Proposition 3.2. 
 §3: Approximation by Rational Functions. 63

Proposition 3.2 can also be proved by explicit computation, but we chose this proof
because the idea will be used later to compute the integral of 1/(z − a) for other curves.

Theorem 3.3. If f is analytic in a neighborhood of the closure S of an open square S,

then for z ∈ S, Z
1 f (ζ)
f (z) = dζ,
2πi ∂S ζ −z
where ∂S is parameterized in the counter-clockwise direction.

Proof. The proof of Theorem 3.3 is exactly like the proof of Theorem 2.4 except that
Proposition 3.2 is used instead of Proposition 2.3. 

Theorem 3.4 (Runge). If f is analytic on a compact set K and if ε > 0 then there is a
rational function r so that
sup |f (z) − r(z)| < ε.
z∈K

Proof. Suppose f is analytic on U open, with U ⊃ K. Let

d = dist(∂U , K) = inf{|z − w| : z ∈ ∂U, w ∈ K}.

Construct a grid of squares with side length d/2.

Figure IV.7 A union of closed squares covering K and contained in U .

64 IV. Integration and Approximation

Shade each square in the grid whose closure intersects K. Note that none of the
√
shaded squares intersect ∂U because they have diameter d/ 2. Let {Sk } be the collection
of shaded squares and let Γ denote the boundary of the union of the closed shaded squares
 
Γ = ∂ ∪j Sj ,

formed from ∪∂Sj , each parameterized in the counter-clockwise direction, by canceling

edges which are traced in opposite directions. Then Γ is a cycle, a finite union of closed
curves. If z is in the interior of one of the squares, Sj0 , then f (ζ)/(ζ − z) is analytic as a
function of ζ on Sj , for j 6= j0 . Apply Theorem 3.3 to Sj0 and Corollary 3.1 to all of the
other squares, then cancel edges traced in opposite directions to obtain
Z
1 f (ζ)
f (z) = dζ. (3.3)
2πi Γ ζ − z
Both sides of (3.3) are continuous functions of z, for z ∈
/ Γ, so that equality holds for all z
in the region bounded by Γ. Fix z0 ∈ K. Choose a Riemann sum for the integral on the
right side of (3.3) to obtain
 m 
 X f (ζ j )(ζ j+1 − ζ j ) 
f (z0 ) − < ε.
 2πi(ζ j − z0 ) 
j=1

We can in fact choose the partition so that the inequality remains true for all refinements
of the partition. By uniform continuity, this inequality remains true for all z in a disk
containing z0 and all refinements of the partition, if we replace ε with 2ε. See Exercise 4.
Cover K by finitely many such disks, and take a common refinement. 

Definition. If r is a rational function, by the Fundamental Theorem of Algebra we write

r(z) = p(z)/q(z) where p and q are polynomials with no common zeros. The zeros of q are
called the poles of the rational function r.

If b is a pole of r(z) then |r(z)| → ∞ as z → b. A rational function is analytic

everywhere in the plane, except at its poles.

It is possible to improve the statement of Runge’s theorem by restricting where the

poles of the rational function need to be placed, using the next Lemma.
 §3: Approximation by Rational Functions. 65

Lemma 3.5. Suppose U is open and connected, and suppose b ∈ U . Then a rational
function with poles only in U can be uniformly approximated on C \ U by a rational
function with poles only at b.

Proof. Suppose a, c ∈ U and suppose |c − a| < dist(a, ∂U ). If z ∈ C \ U then |a − z| ≥

dist(a, ∂U ) ≥ |c − a|, so that
∞
X
1 1 1 (c − a)n
= = c−a = ,
z−c z − a − (c − a) (z − a)(1 − ( z−a )) n=0 (z − a)n+1

where the sequence of partial sums approximate 1/(z − c) uniformly on C \ U . By taking

products we can also approximate (z − c)−n for n ≥ 1 on C \ U , and by taking finite linear
combinations, we can uniformly approximate on C \ U any rational function with poles
only at c by rational functions with poles only at a.
Write c ∈ Rd if every rational function with poles only at c can be uniformly approx-
imated on C \ U by rational functions with poles only at d. This relation is transitive: if
c ∈ Rd and d ∈ Re then c ∈ Re . Set E = {a ∈ U : a ∈ Rb }. By transitivity and the
argument above, if a ∈ E then E contains a disk centered at a with radius dist(a, ∂U).
Thus E is open. Moreover if an ∈ E converges to a∞ ∈ U , then we can choose n so large
that for all z ∈ ∂U

|z − an | ≥ |z − a∞ | − |an − a∞ | ≥ dist(a∞ , ∂U ) − |an − a∞ | > |an − a∞ |,

so that a∞ ∈ Ran and by transitivity, a∞ ∈ E. This proves E is closed in U and by

connectedness E = U .
Now suppose that r is rational with poles only in U and fix b ∈ U . Each term 1/(z−c)k
in the partial fraction expansion of r can be approximated by a rational function with poles
only at b. Adding the approximations, gives an approximation of r by a rational function
with poles only at b. 

Corollary 3.6. Suppose U is connected and open and suppose {z : |z| > R} ⊂ U for some
R < ∞. Then a rational function with poles only in U can be uniformly approximated on
C \ U by a polynomial.
66 IV. Integration and Approximation

Proof. By Lemma 3.5 we need only prove that if |b| > R, then a rational function with
poles at b can be uniformly approximated by a polynomial on C \ U . But
∞  n
1 1 1X z
= =− ,
z−b −b(1 − zb ) b n=0 b

where the sum converges uniformly on |z| ≤ R. As in the proof of Lemma 3.5, we
can approximate (z − b)−n for n ≥ 1 and by taking finite linear combinations, we can
approximate any rational function with poles only at b by a polynomial, uniformly on
{z : |z| ≤ R} ⊃ C \ U . 

Theorem 3.4, Lemma 3.5 and Corollary 3.6 combine to give the following improvement
of Runge’s Theorem.

Theorem 3.7 (Runge). Suppose K is a compact set. Choose one point an in each
bounded component Un of C \ K. If f is analytic on K and ε > 0, then we can find a
rational function r with poles only in the set {an } such that

sup |f (z) − r(z)| < ε.

z∈K

If C \ K has no bounded components, then we may take r to be a polynomial.

For example if K1 and K2 are disjoint compact sets such that C\(K1 ∪K2 ) is connected
and ε > 0 then we can find a polynomial p so that |p| < ε on K1 and |p − 1| < ε on K2
because the function which is equal to 0 on K1 and equal to 1 on K2 is analytic on K1 ∪K2 .

Corollary 3.8. If f is analytic on an open set Ω then there is a sequence of rational

functions rn with poles in C \ Ω so that rn converges to f uniformly on compact subsets
of Ω.

Proof. Set
1
Kn = {z ∈ Ω : dist(z, ∂Ω) ≥ and |z| ≤ n}.
n
Then Kn is compact, ∪Kn = Ω and each bounded component of C \ Kn contains a
point of C \ Ω. By Theorem 3.7 and Corollary 3.6, we can choose the rational functions
approximating f to have poles only in C \ Ω. 
 §3: Approximation by Rational Functions. 67

The improvement of Corollary 3.8 over Theorem 3.4 is that the poles of rn are outside
of Ω, not just outside the compact subset of Ω on which rn is close to f .

Corollary 3.8 says that every analytic function is a limit of rational functions, uni-
formly on compact subsets. We complete this chapter by showing that a uniform limit of
rational functions is always analytic.

Theorem 3.9 (Weierstrass). Suppose {fn } is a collection of analytic functions on a

region Ω such that fn → f uniformly on compact subsets of Ω. Then f is analytic on Ω.
Moreover fn′ → f ′ uniformly on compact subsets of Ω.

Lemma 3.10. If G is continuous on γ and then

Z
G(ζ)
g(z) ≡ dζ
γ ζ −z

is analytic in C \ γ and
Z
′ G(ζ)
g (z) = dζ.
γ (ζ − z)2
Repeated application of Lemma 3.10 shows that formula (2.2) in Cauchy’s Estimate
holds for all z such that |z − z0 | < r.

Proof. There are at least two ways to prove this Lemma. One way is to write out a
power series expansion for 1/(ζ − z) based at z0 , where z0 ∈
/ γ, then interchange the order
of summation and integration to obtain a power series expansion for g based at z0 . The
derivative of g can be found by differentiating the series term by term. The second proof
is to write
Z Z
g(z + h) − g(z) G(ζ) h
− dζ = G(ζ) dζ,
h γ (ζ − z)2 γ (ζ − z)2 (ζ− (z + h))

which → 0 as h → 0. The proof shows that

Z
′ G(ζ)
g (z) = dζ. 
γ (ζ − z)2
68 IV. Integration and Approximation

Proof of Theorem 3.9. Analyticity is a local property, so to prove the first statement
we may suppose D is a disk with D ⊂ Ω. Then by Theorem 2.4, if z ∈ D,
Z
1 fn (ζ)
fn (z) = dζ.
2πi ∂D ζ −z

Set Z
1 f (ζ)
F (z) = dζ.
2πi ∂D ζ −z
Then
|fn (z) − F (z)| → 0

for each z ∈ D, because fn → f uniformly on ∂D. Thus F = f on D and by Lemma 3.10,

F is analytic on D. By Theorem 2.4 and Lemma 3.10
Z
1 fn (ζ)
fn′ (z) = dζ.
2πi ∂D (ζ − z)2

and Z
′ ′ 1 f (ζ)
f (z) = F (z) = dζ.
2πi ∂D (ζ − z)2
Again, since fn → f uniformly on ∂D, we have that fn′ converges uniformly to f ′ on
compact subsets of D. Thus fn′ converges uniformly to f ′ on closed disks contained in Ω.
Given a compact subset K of Ω, we can cover K by finitely many closed disks contained
in Ω and hence fn′ converges uniformly on K to f ′ . 

This proof of Theorem 3.9 is easier than the original proof of Weierstrass which used
doubly indexed sums. The proof is implicitly using the equivalence of analytic and holo-
morphic.

§4. Exercises.

A
1. Let U be an open set in C. Define an equivalence relation on the points of U by: a ∼ b
if and only if there is a polygonal arc contained in U with edges parallel to the axes
and with endpoints a and b. Show that each equivalence class is open and closed in U
 §4: Exercises 69

and connected and that there are at most countably many equivalence classes. The
equivalence classes are called the components of U .
2. Prove the following version of Weierstrass’s theorem: Suppose {Un } is an increasing
sequence of regions and suppose fn is defined and analytic on Un . If the sequence
fn converges uniformly on compact subsets of U ≡ ∪n Un to a function f , then f is
analytic on U and the sequence fn′ converges to f ′ uniformly on compact subsets of
U (even though perhaps none of the fn are defined on all of U ).
3. (a) Use Cauchy’s estimate to prove Liouville’s Theorem.
(b) Use Cauchy’s estimate (2.3) to compute a lower bound on the radius of conver-
gence of the power series representation of a holomorphic function.
4. At the end of the proof of Runge’s Theorem 3.4 we stated that an inequality remains
true for all z in a disk containing z0 and for all refinements of the partition. Supply the
details. Hint: Using Cauchy’s integral formula prove that |f (z) − f (z0 )| ≤ C|z − z0 |,
1
for all z with |z − z0 | < 2 dist(z0 , Γ), where C depends only on sup |f |, ℓ(Γ), and
dist(z0 , Γ). Do the same for the difference of the Riemann sums at z and z0 (with the
same partition).

5. Suppose Ω is a region which is symmetric about R. Set Ω+ = Ω ∩ H and Ω− =

Ω ∩ (C \ H). If f is analytic on Ω+ , continuous on Ω+ ∪ (Ω ∩ R) and Imf = 0 on Ω ∩ R
then the function defined by

 f (z) for z ∈ Ω \ Ω−
F (z) =

f (z) for z ∈ Ω−

is analytic on Ω. Hint: Divide a rectangle meeting R into two rectangles. This is

the analytic version of the Schwarz Reflection Principle, which we will encounter in
Chapter XI.
6. Show that the function ζ(z) given in Exercise II.14(a) is analytic in Rez > 1. Write
R n+1 R n+1 R x −z−1
n−z − n x−z dx = n n
zt dtdx and use it to prove that (z − 1)ζ(z) has a
unique analytic extension to {z : Rez > 0}.
70 IV. Integration and Approximation

7. Show that there is a constant C < ∞ so that if f is analytic on D, then

Z
′
|f (z)| ≤ C |f (x + iy)|dxdy
D

for all |z| ≤ 1/2.

8. Prove that there exists a sequence of polynomials pk such that

 1 if Re(z) > 0



lim pk (z) = 0 if Re(z) = 0
k→∞ 



−1 if Re(z) < 0.

9. (Universal entire function). Prove that there is an entire function f with the property
that if g is entire and if K is compact and ε > 0 then there is a number r > 0,
depending on g, K, and ε, so that supK |g(z) − f (r + z)| < ε.

10. Suppose f is a function defined on D with the property that given any three points
a, b, c ∈ D, there is an analytic function g (possibly depending on a, b, c) so that
|g| ≤ 1 on D and g(a) = f (a), g(b) = f (b) and g(c) = f (c). Prove that f has a
complex derivative at each point of D and |f | ≤ 1. Hint: Use the two point version
to prove f is continuous first.
 V

Cauchy’s Theorem and Applications

§1. Cauchy’s Theorem.

Recall that if γ is a closed curve then

Z
q(z)dz = 0
γ

for all polynomials q and Z

1
dz = 0,
γ (z − a)k
when k ≥ 2 and a ∈
/ γ, because the integrand in each case is the derivative of an analytic
function. See Corollary IV.2.1. If r is a rational function, then it has a partial fraction
expansion
nk
N X
X ck,j
r(z) = + q(z)
(z − pk )j
k=1 j=1

where q is a polynomial. So if γ is a closed curve which does not intersect any of the poles
of r, then
Z N
X Z
1
r(ζ)dζ = ck,1 dζ. (1.1)
γ γ (ζ − pk )
k=1

Theorem 1.1 (Cauchy). Suppose γ is a cycle contained in a region Ω and suppose

Z
dζ
=0 (1.2)
γ ζ −a

for all a ∈
/ Ω. If f is analytic on Ω then
Z
f (ζ)dζ = 0.
γ

Proof. By Runge’s Theorem, we can find a sequence of rational function rn with poles in
C \ Ω so that rn converges to f uniformly on the compact set γ ⊂ Ω. By (1.1) and (1.2),
R
r (z)dz = 0. But then
γ n
Z Z
| f (z)dz| = | (f (z) − rn (z))dz| ≤ sup |f − rn |ℓ(γ) → 0. 
γ γ γ

71
72 V. Cauchy’s Theorem

We can use Cauchy’s Theorem to extend Theorems IV.2.4 and IV.3.3.

Theorem 1.2 (Cauchy’s Integral Formula). Suppose γ is a cycle contained in a region

Ω and suppose Z
dζ
=0
γ ζ −a
for all a ∈
/ Ω. If f is analytic on Ω and z ∈ C \ γ then
Z Z
1 f (ζ) 1 1
dζ = f (z) · dζ.
2πi γ ζ −z 2πi γ ζ −z

Proof. For each z ∈ Ω, the function g(ζ) = (f (ζ) − f (z))/(ζ − z) extends to be analytic
on Ω and by Cauchy’s Theorem it has integral over γ equal to 0. Theorem 1.2 follows by
splitting the integral of g along γ into two pieces. 

z
γ

Figure V.1 Integration on a cycle.

The cycle γ in Figure V.1 consists of two circles, parameterized in opposite directions
as indicated. If f is analytic on the closed region bounded by the two circles, then by
Proposition IV.2.3 the hypotheses in Cauchy’s Integral Formula are satisfied, and
Z
1 f (ζ)
f (z) = dζ,
2πi γ ζ −z

when z is between the two circles, again using Proposition IV.2.3. When z is outside the
larger circle or inside the inner circle, the integral is equal to 0 by Cauchy’s Theorem,
because f (ζ)/(ζ − z) is an analytic function of ζ on the region between the two circles, if
z is not in this region.
 §2: Winding Number. 73

§2. Winding Number.

The important integrals (1.2) have a geometric interpretation which we will next
explore.

Lemma 2.1. If γ is a cycle and a ∈

/ γ, then
Z
1 1
dζ
2πi γ ζ − a
is an integer.

Proof. Without loss of generality we may suppose γ is a closed curve parameterized by a

continuous, piecewise differentiable function γ(t) : [0, 1] → C. Define
Z x
γ ′ (t)
h(x) = dt.
0 γ(t) − a

Then h′ (x) exists and equals γ ′ (x)/(γ(x) − a), except at finitely many points x. Then
d −h(x)
e (γ(x) − a) = −h′ (x)e−h(x) (γ(x) − a) + e−h(x) γ ′ (x)
dx

= −γ ′ (x)e−h(x) + γ ′ (x)e−h(x) = 0,

except at finitely many points. Since e−h(x) (γ(x) − a) is continuous, it must be constant.
Thus
e−h(1) (γ(1) − a) = e−h(0) (γ(0) − a) = 1 · (γ(1) − a).

Since γ(1) − a 6= 0, e−h(1) = 1 and h(1) = 2πki, where k is an integer. Thus

Z
1 dζ h(1)
dζ = = k,
2πi γ ζ − a 2πi
an integer. 

Definition. If γ is a cycle, then the index or winding number of γ with respect to a

(or about a) is Z
1 dζ
n(γ, a) = ,
2πi γ ζ −a
74 V. Cauchy’s Theorem

for a ∈
/ γ.

Note the following properties of n(γ, a):

(a) n(γ, a) is an analytic function of a, for a ∈
/ γ, by Lemma 3.10. In particular it
is continuous and integer-valued by Lemma 2.1. Thus n(γ, a) is constant in each
component of C \ γ.
(b) n(γ, a) → 0 as a → ∞. Thus n(γ, a) = 0 in the unbounded component of C \ γ.
(c) n(−γ, a) = −n(γ, a).
(d) n(γ1 + γ2 , a) = n(γ1 , a) + n(γ2 , a).
(e) If γ(t) = eikt , for 0 ≤ t ≤ 2π, where k is an integer, then
Z Z 2π
1 dz 1 ikeikt
n(γ, 0) = = dt = k.
2πi γ z 2πi 0 eikt
The curve γ in (e) “winds” k times around 0.
If a ∈ C, let
La = {z : Imz = Ima}

be the horizontal line through a. Suppose that γ1 is a curve in the half-plane {z : Imz >
Ima} with endpoints b and c on La such that Reb < Rea < Rec. See Figure V.2a.

γ1

γ2
La
b γ3 a γ4 c
Figure V.2a The Winding of a curve in a half-plane.

For δ > 0 small, let γ2 be the semi-circular arc in the half-plane centered at a and
radius δ. Let γ3 be the segment on La between b and γ2 and let γ4 be the segment on
La between γ2 and c. Then we can choose directions for γ2 , γ3 , and γ4 so that γ =
γ1 + γ2 + γ3 + γ4 is a closed curve. Note that a is in the unbounded component of C \ γ
(indeed, we can draw a vertical line from a to ∞ which does not meet γ) and hence
Z
dζ
=0 (2.1)
γ ζ −a
 §2: Winding Number. 75

by (b).

γ1

La
a b c
Figure V.2b The Winding of a curve in a half-plane.

If γ1 is a curve in the half-plane {z : Imz > Ima} with endpoints b and c on La

such that Rea < Reb ≤ Rec, then we can add the line segment between b and c to form
a closed curve γ. See Figure V.2b. Then a is in the unbounded component of C \ γ
so that (2.1) holds. A similar construction can be made if the endpoints b and c of γ
satisfy Reb ≤ Rec < Rea. Similar constructions can be performed starting with a curve γ1
contained in the lower half-plane {z : Imz < Ima}.
Now suppose that γ is a closed curve such that a ∈
/ γ. If γ does not intersect La ∩
{Rez > Rea} then a is in the unbounded component of C \ γ and hence n(γ, a) = 0.
Otherwise, we may suppose that γ begins on La ∩ {Rez > Rea}. Write γ \ La as a
finite union of curves Aj . Each Aj is contained in the upper half-plane or the lower half-
plane. Choose δ > 0 so small that the disk of radius δ centered at a does not intersect
γ. Notice that if γ1 as in Figure V.2a is one of the components Aj , then we can replace
γ1 by −γ2 − γ3 − γ4 without affecting the winding number about a and without changing
the beginning and endpoints of the component. Similarly, if γ1 as in Figure V.2b can be
replaced by the line segment from b to c without affecting the winding number about a or
the endpoints. By applying the appropriate case to each Aj we create a new curve closed
curve σ with n(γ, a) = n(σ, a). Moreover, we can write σ = I + ∪ I − ∪ C where I + is a
union of intervals contained in La ∩ {Rez > Rea}, I − is a union of intervals contained
in La ∩ {Rez < Rea} and C is a union of semicircles contained in Ca = {|z − z0 | = δ}.
Since σ is closed, I + can be written as a closed curve and n(I + , a) = 0, since a is in the
unbounded component of C \ I + . Similarly I − is closed and n(I − , a) = 0. Thus C is closed
and n(γ, a) = n(C, a).
76 V. Cauchy’s Theorem

If α is a semicircular arc contained in Ca then

Z
1 dζ 1
dζ = ± ,
2πi α ζ − a 2

where the sign is + if α is traced in the counter-clockwise direction and otherwise the sign
is −. The winding number of C, and hence γ, about a can then be found by counting + 21
or − 12 for each semi-circular arc depending on the direction it is parameterized.
More informally, if we trace γ starting at a point b ∈ La with Reb > Rea, then we can
compute n(γ, a) by counting the “winding” of γ about a adding ± 12 if the endpoints are
on opposite sides of a in La , depending on the direction of travel, and no change to the
count if the endpoints are on the same side of a on La .
We can remove the assumption that the line La is horizontal by rotating the plane
about a without altering the winding number. But there is an even easier way to compute
winding numbers. Let Ra be any ray, or half line, from a to ∞. Suppose that γ is a cycle
such that a ∈
/ γ, and suppose that γ ∩ Ra consists of finitely many points zj . Because γ
is oriented, we can count the “winding” of γ about a by adding +1 if the crossing at zj is
counter-clockwise and −1 if the crossing is clockwise. See Exercise 1.

−1 1 2

Figure V.3 Calculating the winding number.

In Figure V.3 n(γ, −1) = 0, n(γ, 1) = 2, n(γ, 2) = 0 and n(γ, i) = 1.

Definition. Closed curves γ1 and γ2 are homologous in a region Ω if n(γ1 − γ2 , a) = 0

for all a ∈
/ Ω and we write
γ1 ∼ γ2 .

Homology is an equivalence relation on the curves in Ω (Exercise 2). A closed curve γ ⊂ Ω

is homologous to 0 in Ω if n(γ, a) = 0 for all a ∈
/ Ω. In this case we write γ ∼ 0.
 §2: Winding Number. 77

Cauchy’s Theorem says that if γ ∼ 0 in Ω and if f is analytic in Ω then

Z
f (z)dz = 0,
γ

and if γ1 is homologous to γ2 in Ω, then

Z Z
f (z)dz = f (z)dz.
γ1 γ2

The most common application of Cauchy’s Integral Formula is when γ ⊂ Ω with γ ∼ 0

and n(γ, z) = 1. Then for f analytic on Ω,
Z
1 f (ζ)
f (z) = dζ.
2πi γ ζ −z

0 1

Figure V.4 n(γ, 0) = n(γ, 1) = 0.

Figure V.4 shows a closed curve σ contained in Ω = C \ {0, 1} with the property that
n(σ, a) = 0 for all a ∈
/ Ω. If you are familiar with homotopy (we will treat this subject later
in the course), this curve shows that homotopy and homology are different since this curve
cannot be shrunk to a point while remaining in Ω. By Cauchy’s theorem if f is analytic
R
on Ω, then σ f (z)dz = 0.
If Ω is a bounded region in C bounded by piecewise differentiable curves, then we can
parameterize ∂Ω so that as you trace each boundary component, the region Ω lies on the
left. In other words iγ ′ (t) is the “inner normal”, rotated counter-clockwise by π/2 from
the tangential direction γ ′ (t). In Exercise 3 you are asked to show that n(∂Ω, a) = 1 for
all a ∈ Ω and n(∂Ω, a) = 0 for all a ∈
/ Ω. We call this the positive orientation of ∂Ω.
78 V. Cauchy’s Theorem

Thus for all such regions, ∂Ω ∼ 0 in any region containing Ω. So by Cauchy’s Theorem, if
Ω is a bounded region, bounded by piecewise differentiable curves and if f is analytic on
Ω, then Z
f (ζ) dζ
f (z) = , (2.2)
∂Ω ζ − z 2πi
for all z ∈ Ω, where ∂Ω has positive orientation. However, if Ω = C \ D, then ∂D is not
homologous to zero in a region containing Ω and (2.2) does not hold.

Definition. A region Ω ⊂ C is called simply connected if S 2 \ π(Ω) is connected in S 2 ,

where π : C → S 2 is the stereographic projection.

Simply connected essentially means “no holes”. For example, the unit disk D is simply
connected. The vertical strip {z : 0 < Rez < 1} is simply connected. The punctured plane
C \ {0} is not simply connected. More generally, a connected open subset of S 2 is called
simply connected, if its complement in S 2 is connected. Thus the set that corresponds to
C \ D together with ∞ (the “North Pole” on the sphere) is simply connected, but C \ D
is not simply connected. The open set inside a “figure 8” curve is not simply connected
because it is not connected.

Theorem 2.2. A region Ω ⊂ C is simply connected if and only if every closed curve in Ω
is homologous to 0 in Ω.

The point of Theorem 2.2 is that simply connected is a geometric condition which is
sufficient for Cauchy’s Theorem to apply.

Proof. Suppose Ω is simply connected and suppose γ is a cycle contained in Ω and suppose
/ Ω. Since Ωc is connected, it must be contained in one component of the complement
a∈
of γ. Because ∞ ∈ Ωc , viewed as a subset of the Riemann sphere, a must be in the
unbounded component of C \ γ, and n(γ, a) = 0 by (b).
Conversely, suppose that S 2 \ π(Ω) = A ∪ B where A and B are non-empty closed sets
in S 2 with A∩B = ∅. Without loss of generality ∞ ∈ B. Since A is closed, a neighborhood
of ∞ does not intersect A and hence π −1 (A) is bounded. Pick a0 ∈ A. We’ll construct a
curve γ0 ⊂ Ω such that n(γ0 , a0 ) 6= 0, proving Theorem 2.2.
 §3: Removable Singularities. 79

The construction is the same construction used to prove Runge’s theorem. Let

d = dist(π −1 (A), π −1 (B)) = inf{|a − b| : a ∈ π −1 (A), b ∈ π −1 (B)} > 0.

Pave the plane with squares of side d/2 such that a0 is the center of one of the squares.
Orient the boundary of each square in the positive, or counter-clockwise direction (like all
storms in the Northern hemisphere). Shade each square Sj with Sj ∩ A 6= ∅. Let γ0 denote
the cycle obtained from ∪∂Sj after performing all possible cancellations. Then γ0 ⊂ Ω
since γ0 does not intersect either A or B, and n(γ0 , a0 ) = 1. 

§3. Removable Singularities.

The main result in this section is in some sense a generalization of Exercise II.4.

Corollary 3.1 (Riemann’s Theorem on Removable Singularities). Suppose f is

analytic in Ω = {z : 0 < |z − a| < δ} and suppose

lim (z − a)f (z) = 0.

z→a

Then f extends to be analytic in {z : |z − a| < δ}.

In other words, we can “remove” the “singularity” of f at a.

Proof. Fix z ∈ Ω and choose ε and r so that 0 < ε < |z − a| < r < δ. Let Cε and
Cr denote the circles of radius ε and r centered at a, oriented in the counter-clockwise
direction as in Figure V.1. The cycle Cr − Cε is homologous to 0 in Ω so that by Cauchy’s
Integral Formula Z Z
1 f (ζ) 1 f (ζ)
f (z) = dζ − dζ.
2πi Cr ζ −z 2πi Cε ζ −z
But Z 
 f (ζ)  1
 dζ ≤ max |f (ζ)| 2πε.
 ζ −z ζ∈Cε |z − a| − ε
Cε

But if ζ ∈ Cε then |f (ζ)|ε = |f (ζ)||ζ − a| → 0 as ε → 0 and hence

Z
1 f (ζ)
f (z) = dζ. (3.1)
2πi Cr ζ − z
80 V. Cauchy’s Theorem

By Lemma IV.3.10, the right side of (3.1) is analytic in {z : |z − a| < r}. Thus if we define
f (a) as the value of the right side of (3.1) when z = a, then this extension is analytic at a
and we have extended f to be analytic in {z : |z − a| < δ}. 

The most important special case of Riemann’s Removable Singularity Theorem is: if
f is bounded and analytic in a punctured neighborhood of a then f extends to be analytic
in a neighborhood of a.

We say that a compact set E has one-dimensional Hausdorff measure equal to

0 if for every ε > 0 there are finitely many disks Dj with radius rj so that

E ⊂ ∪j Dj

and
X
rj < ε.
j

Corollary 3.2 (Painlevé). Suppose E is a compact set with one-dimensional Hausdorff

measure 0. If f is bounded and analytic on U \ E, where U is open and E ⊂ U , then f
extends to be analytic on U .

Proof. As in the proof of Runge’s Theorem and Theorem 2.2, we can find a curve γ ⊂ U \E
so that n(γ, b) = 1 for all b ∈ E, and n(γ, b) = 0 for b ∈ C \ U . Cover E by finitely many
P
disks Dj of radius rj so that rj < ε. For small ε, the disks Dj will not intersect γ. Let


V = {z : n(γ, z) = 1}, let σ = ∂ ∪Dj , and let Ω = V \ ∪Dj . Then γ + σ = ∂Ω, which we
parametrize so that ∂Ω has positive orientation. Then as in (2.2), γ + σ ∼ 0 in U \ ∪Dj ,
so that by Cauchy’s Theorem
Z Z
1 f (ζ) 1 f (ζ)
f (z) = dζ + dζ,
2πi γ ζ −z 2πi σ ζ −z

for z ∈ V \ ∪Dj . Since ℓ(σ) ≤ ℓ(∪Dj ) < 2πε and since f is bounded, the second integral
tends to 0 as ε → 0, exactly as in the proof of Riemann’s Theorem. Thus
Z
1 f (ζ)
dζ
2πi γ ζ − z
 §4: Laurent Series. 81

provides an analytic extension of f to E, by Lemma IV.3.10. 

Painlevé asked for geometric conditions on a compact set E to be removable for

bounded analytic functions in 1888. Removable means that the second sentence of Corol-
lary 3.2 holds. A major accomplishment in complex analysis within the last fifteen years
was to solve this problem, though there is some debate about whether the solution is really
“geometric”.

We can use the Cauchy integral formula and Riemann’s Theorem to give a formula
for the inverse of a one-to-one analytic function. Suppose f is analytic and one-to-one on
a region Ω. Let γ be a cycle in Ω such that n(γ, a) = 0 for all a ∈ Ωc . Suppose f (z) − w
has exactly one zero in {z : n(γ, z) = 1}. Then

f ′ (z)z
lim (z − f −1 (w)) = f −1 (w).
z→f −1
(w) f (z) − w

By Riemann’s Removable Singularity Theorem

f ′ (z)z f −1 (w)
h(z) = −
f (z) − w z − f −1 (w)

extends to be analytic in Ω, and by Cauchy’s Theorem

Z
h(z)dz = 0.
γ

Since n(γ, f −1 (w)) = 1, we obtain

Z
−1 1 f ′ (z)z
f (w) = dz.
2πi γ f (z) − w

We can use this formula to show that f −1 is analytic, by imitating the proof of Lemma
IV.3.10.

§4. Laurent Series.

An annulus is the region between two concentric circles. If f is analytic on the

annulus A = {z : r < |z − a| < R} then by Runge’s Theorem, we can approximate f by
82 V. Cauchy’s Theorem

a rational function with poles only at a. The Laurent Series is another version of this
result, similar to a power series expansion.

Theorem 3.3 (Laurent Series). Suppose f is analytic on A = {z : r < |z − a| < R}.

Then
∞
X
f (z) = an (z − a)n ,
n=−∞

where the series converges uniformly and absolutely on compact subsets of A. Moreover
Z
1 f (ζ)
an = dζ, (3.2)
2πi Cs (ζ − a)n+1
where Cs is the circle centered at a with radius s, r < s < R, oriented counter-clockwise.

One consequence of Theorem 3.3 is that an analytic function f on A can be written

as f = f1 + f2 where f1 is analytic in |z| < R and f2 is analytic in |z| > r. Since an does
not depend on s, the integrals (3.2) do not depend on s, so long as r < s < R.

Proof. Without loss of generality, a = 0. Set

Z
1 f (ζ)
fs (z) = dζ
2πi Cs ζ − z
where Cs = {seit : 0 ≤ t ≤ 2π}. Then by Lemma IV.3.10, fs is analytic off Cs . If
r < |z| < s1 < s2 < R then Cs2 − Cs1 ∼ 0 with respect to A and n(Cs2 − Cs1 , z) = 0. By
Cauchy’s Integral Formula, fs2 (z) − fs1 (z) = 0. This says that fs (z) does not depend on
s, so long as r < |z| < s < R.
1
Expanding ζ−z in a power series expansion about 0, and interchanging the order of
summation and integration, as we have done before, we conclude that fs has a power series
expansion
∞
X
fs (z) = an z n ,
n=0
valid in |z| < s, where an satisfies (3.2). Likewise fs (z) does not depend on s so long
1
as s < |z| < R. Expanding ζ−z in a power series expansion about ∞, i.e. in powers of
1/z, and interchanging the order of summation and integration we get that fs has a power
series expansion
∞
X
fs (z) = − a−n z −n ,
n=1
 §4: Laurent Series. 83

valid in |z| > s, where a−n satisfies (3.2).

If r < s1 < |z| < s2 < R, then Cs2 − Cs1 ∼ 0 in A, so that by Cauchy’s Integral
Formula Z X∞
1 f (ζ)
f (z) = dζ = fs2 (z) − fs1 (z) = an z n . (3.3)
2πi Cs2 −Cs1 ζ −z n=−∞

For example the function

z 2 1
r(z) = + −
z2 + 4 (z − 3) 2 z−4
is analytic in C except at ±2i, 3, 4 and hence is analytic on four regions centered at 0:
A1 = {z : |z| < 2}, A2 = {z : 2 < |z| < 3}, A3 = {z : 3 < |z| < 4}, and A4 = {z : 4 < |z|}.
Of course A1 is a disk, so that r has a power series expansion in A1 . We can use
∞
X
1 1 −1 n
= z = z ,
z−a −a(1 − a ) n=0 an+1

provided |z| < |a|, and

∞
X
1 1 an−1
= = ,
z−a z(1 − az ) n=1 z n
provided |z| > |a|. These formulae give the Laurent expansion of 1/(z − 4) in each region.
Differentiating a series of this form with a = 3 will give the series for 1/(z − 3)2 . Setting
a = −4 and replacing z with z 2 will give the series for 1/(z 2 + 4), and multiplying by z will
give the series for z/(z 2 + 4). If a rational function is given as a ratio of two polynomials,
p/q, then use the techniques of Section III.2 to find its partial fraction expansion, and then
use the series expansion for 1/(z − a) as above. Sometimes it might be easier to find the
partial fraction expansion for 1/q, then multiply the resulting series by the polynomial p
and collect together similar powers of z.

Laurent Series are useful for analyzing the behavior of an analytic function near an
isolated singularity. We say that f has an isolated singularity at b if f is analytic in
0 < |z − b| < ε for some ε > 0. Write
∞
X
f (z) = an (z − b)n .
n=−∞
84 V. Cauchy’s Theorem

Note the following:

(i) If an = 0 for n < 0 then f extends to be analytic at b with f (b) = a0 . In this case we
say that f has a removable singularity at b.

(ii) If an = 0 for n < n0 with n0 > 0 and an0 6= 0, then we can write
∞
X
f (z) = (z − b)n0 an0 +n (z − b)n = an0 (z − b)n0 + an0 +1 (z − b)n0 +1 + . . . .
n=0
In this case b is called a zero of order n0 .

(iii) If an = 0 for n < −n0 with n0 > 0 and a−n0 6= 0 then we can write
X∞
−n0 a−n0 a−n0 +1
f (z) = (z − b) a−n0 +n (z − b)n = n
+ + ....
n=0
(z − b) 0 (z − b)n0 −1
In this case b is called a pole of order n0 , and |f (z)| → ∞ as z → b.
In each of the above cases there is a unique integer k so that

lim (z − b)k f (z)

z→b

exists and is non-zero, and

(z − b)k f (z)

is analytic and non-zero in a neighborhood of b.

(iv) If an 6= 0 for infinitely many negative n, then b is called an essential singularity.
For example
X∞
1 (−1)n −2n
f (z) = e− z2 = z
n=0
n!
has an essential singularity at 0, even though it is infinitely differentiable on the real line.
If f is analytic in {z : |z| > R}, then f (1/z) has a isolated singularity at 0, and we
say that f has an isolated singularity at ∞. We classify this singularity at ∞ as a zero,
pole or essential singularity if f (1/z) has a zero, pole or (respectively) essential singularity
at 0. In terms of the Laurent expansion of f in |z| > R,
∞
X
f (z) = bn z n ,
n=−∞
f has an essential singularity at ∞ if bn 6= 0 for infinitely many positive n. The reader can
supply the corresponding statement for zeros and poles and their orders. For example, a
polynomial of degree n has a pole of order n at ∞.
 §5: The Argument Principle. 85

Definition. A zero or pole is called simple if the order is 1.

Definition. If f is analytic in a region Ω except for isolated poles in Ω then we say

that f is meromorphic in Ω. A meromorphic function in C is sometimes just called
meromorphic.

If f is meromorphic in Ω, then 1/f is meromorphic in Ω, and a zero of order k

becomes a pole of order k, and a pole of order k becomes a zero of order k. If f and g are
meromorphic in Ω then the order of a zero or pole for the meromorphic function f /g at b
can be found by factoring the appropriate power of z − b out of f and g then taking the
difference of these powers.
The next result gives an idea of the behavior near an essential singularity.

Theorem 3.4. If f is analytic in U = {z : 0 < |z − b| < δ} and if b is an essential

singularity for f then f (U ) is dense in C.

In other words, every (punctured) neighborhood of an essential singularity has a dense

image.

Proof. If not there exists A ∈ C and ε > 0 so that |f (z) − A| > ε for all z ∈ U . Then

1
f (z) − A

is analytic and bounded by 1/ε on U . By Riemann’s Theorem, 1/(f (z) − A) extends to be

analytic in U ∪{b} and so f (z) −A is meromorphic in U ∪{b} and so f is also meromorphic
in U ∪ {b}. Thus the Laurent expansion for f has at most finitely terms with a negative
power of z − b, contradicting the assumption that b is an essential singularity. 

§5. The Argument Principle.

The next result is useful for locating zeros and poles of meromorphic functions.
86 V. Cauchy’s Theorem

Theorem 4.1(Argument Principle). Suppose f is meromorphic in a region Ω with

zeros {zj } and poles {pk }. Suppose γ is a cycle with γ ∼ 0 in Ω and suppose {zj } ∩ γ = ∅
and {pk } ∩ γ = ∅. Then
Z X X
1 f ′ (z)
n(f (γ), 0) = dz = n(γ, zj ) − n(γ, pk ). (4.1)
2πi γ f (z)
j k

In the statement of the Argument Principle, if f has a zero of order k at z, then z occurs
k times in the list {zj }, and a similar statement holds for the poles. For example, if γ
is a simple closed curve in Ω which is homologous to 0 in Ω, then the number of zeros
“enclosed” by γ minus the number of poles “enclosed” by γ is equal to the winding number
of the image curve f (γ) about zero.

Proof. The first equality in (4.1) follows from the change of variables w = f (z). Note that
γ ∼ 0 and γ ⊂ Ω implies that n(γ, a) = 0 if a is sufficiently close to ∂Ω. Thus n(γ, zj ) 6= 0
for only finitely many zj and for only finitely many pj because there are no cluster points
of {zj } or {pk } in Ω. This implies that the sums in (4.1) are finite.
If b is a zero or pole of f then we can write

f (z) = (z − b)k g(z)

where g is analytic in a neighborhood of b and g(b) 6= 0. Then

f ′ (z) = k(z − b)k−1 g(z) + (z − b)k g ′ (z)

and
f ′ (z) k g ′ (z)
= + .
f (z) z−b g(z)
Since g(b) 6= 0, g ′ /g is analytic in a neighborhood of b and hence f ′ /f − k(z − b)−1 is
analytic near b. Thus
f ′ (z) X 1 X 1
− + (4.2)
f (z) z − zj z − pk
is analytic at all points z where n(γ, z) 6= 0. In the sums, we repeat zj and pk according
to their multiplicity. By Cauchy’s Theorem integrating (4.2) over γ gives (4.1). 
 §5: The Argument Principle. 87

Corollary 4.2 (Rouché). Suppose γ is a closed curve in a region Ω with γ ∼ 0 in Ω and

n(γ, z) = 0 or = 1 for all z ∈ Ω. If f and g are analytic in Ω and satisfy

|f (z) + g(z)| < |f (z)| + |g(z)| (4.2)

for all z ∈ γ, then f and g have the same number of zeros enclosed by γ.

Equation (4.2) says that strict inequality holds in the triangle inequality. The points
“enclosed” by γ are those z ∈ Ω for which n(γ, z) = 1. The number of zeros of f and g
are counted according to their multiplicity.

f
Proof. The function g is meromorphic in Ω and satisfies
   
f   
 + 1<  f +1. (4.3)
g  g

By (4.2), f 6= 0 and g 6= 0 on γ, so that the hypotheses of the Argument Principle are

satisfied.

w
|w + 1|
|w|
−1 0
1

Figure V.5 Proof of Rouché’s Theorem.

The left side of (4.3) is the distance from w = f (z)/g(z) to −1. But |w − (−1)| = |w| + 1
f
if and only if w ∈ [0, ∞). See Figure V.5. Thus the assumption (4.2) implies that g
(γ)
omits the half-line [0, ∞) and hence does not wind around 0. By the argument principle,
f
the number of zeros of g equals the number of poles and hence the number of zeros of f
equals the number of zeros of g, counting multiplicity. 

Example. f (z) = z 9 − 2z 6 + z 2 − 8z − 2.
88 V. Cauchy’s Theorem

How many zeros does f have in |z| < 1? The biggest term is −8z, so comparing f and
−8z:
|f (z) + 8z| = |z 9 − 2z 6 + z 2 − 2| ≤ 1 + 2 + 1 + 2 = 6 < |8z|

on |z| = 1. By Rouché’s Theorem, f and 8z have the same number of zeros in |z| < 1,
namely one. How many zeros does f have in |z| < 2? In this case we compare f with z 9 :

|f (z) − z 9 | = | − 2z 6 + z 2 − 8z − 2| ≤ 128 + 4 + 16 + 2 = 150 < |z 9 | = 512.

Therefore f and z 9 have the same number of zeros in |z| < 2, namely 9. Thus 8 of the
zeros of f lie in 1 < |z| < 2, and the remaining zero lies in |z| < 1.
To find the number of zeros of z 4 − 4z + 5 in |z| < 1 we compare with the constant
function 5 on |z| = 1:

|z 4 − 4z + 5 − 5| = |z 4 − 4z| ≤ 5 ≤ |5| + |z 4 − 4z + 5|.

If equality holds in the first inequality, then 5 = |z 4 − 4z| = |z 3 − 4|. But then z 3 = −1,
and so z 4 − 4z + 5 = −z − 4z + 5 = 5(−z + 1). Since z 6= 1, |z 4 − 4z + 5| > 0 and we have

|z 4 − 4z + 5 − 5| < |5| + |z 4 − 4z + 5|.

By Rouché’s Theorem, z 4 − 4z + 5 and 5 have the same number of zeros in |z| < 1.
A more elaborate process for locating zeros of polynomials is in Exercise 12.

§6. Exercises.

1. Suppose γ is a closed curve and a ∈

/ γ. Suppose Ra is a ray (half-line) from a to
∞ such that γ ∩ Ra consists of finitely many points. Prove that n(γ, a) equals the
number of counterclockwise crossings of Ra minus the number of clockwise crossings.
The set γ \ Ra consists of finitely many pieces, but not all endpoints are necessarily
“crossings”.
2. Prove that homology is an equivalence relation on curves in a region Ω.
 §6: Exercises 89

3. Suppose Ω is a bounded region whose boundary consists of finitely many disjoint

piecewise differentiable simple closed curves. Orient ∂Ω so that the region lies on
the left for each boundary component (the inner normal is i times the unit tangent
vector). Prove n(∂Ω, a) = 1 if a ∈ Ω and n(∂Ω, a) = 0 if a ∈ C \ Ω. You may assume
that Ω is formed from a simply connected region by removing finitely many simply
connected sub-regions.
4. Prove the uniqueness of the Laurent series expansion, that is if
∞
X ∞
X
n
an z = bn z n
n=−∞ n=−∞

for r < |z| < R then an = bn for all n. Convergence of the series on the region is part
of the assumption.
5. Notice that in the proof of Laurent Series expansions we proved that a function f
which is analytic on r < |z| < R can be written as f = f1 + f2 where f1 is analytic in
|z| < R and f2 is analytic in |z| > r and f2 (z) → 0 as |z| → ∞. Suppose that Ω is a
bounded region in C such that ∂Ω is a finite union of disjoint (piecewise continuously
differentiable) closed curves Γj , j = 1, . . . , n. Suppose that f is analytic on Ω. Prove
P
that f = fj where fj is analytic on the component of C \ Γj which contains Ω.

6. Prove that if a sequence of analytic polynomials converges uniformly on a region Ω

then the sequence converges uniformly on a simply connected region containing Ω.
7. We define the singularity at ∞ of f (z) to be the singularity at 0 of g(z) = f (1/z).
Find the singularity at ∞ of the following functions. If the singularity is removable,
give the value, if the singularity is a zero or pole, give the order.
z 2 +1 1 1 z1
a. ez
b. e1/z −1
−z c. ez/(1−z) d. ze1/z e. z 2 − z f. z3
e
8. Find the expansion in powers of z for

z
(z 2 + 4)(z − 3)2 (z − 4)

which converges in 3 < |z| < 4.

90 V. Cauchy’s Theorem

9. Prove that the number of roots of the equation

z 2n + αz 2n−1 + β 2 = 0,

where α, β are real and nonzero, and n is a natural number, that have positive real
part is equal to n if n is even. If n is odd, their number is n − 1 for α > 0 and n + 1
for α < 0.
Hint: See what happens to

z 2n + αz 2n−1 + β 2

as z traces the boundary of a large half-disk.

10. How many zeros does p(z) = 3z 5 + 21z 4 + 5z 3 + 6z + 7 have in D ? How many zeros
in {z : 1 < |z| < 2} ?
11. Prove that all of the zeros of the polynomial

p(z) = z n + cn−1 z n−1 + . . . + c1 z + c0

lie in the disk centered at 0 with radius

p
R= 1 + |cn−1 |2 + . . . + |c1 |2 + |c0 |2 .

12. This problem gives an algorithm for counting the number of zeros of a polynomial
in a disk. First translate and dilate the disk to the unit disk. Suppose p(z) =
an z n + an−1 z n−1 + . . . + a0 is a polynomial with an 6= 0. Set
 
∗ 1n
p (z) = z p = a0 z n + a1 z n−1 + . . . + an
z

and let q(z) = a0 p(z) − an p∗ (z).

Case (a): If |an | < |a0 | set p1 = q.
Case (b): If |an | > |a0 |, set p1 = q ∗ /z = (a0 p∗ − an p)/z.
 §6: Exercises 91

Case (c): If |an | = |a0 | and q 6= 0 on |z| = 1, set p1 = q.

Let N (p) denote the number of zeros of p in D = {z : |z| < 1}. Then prove

 N (p1 ) in Cases (a) and (c)
N (p) =

N (p1 ) + 1 in Case (b).

Repeat the procedure until it stops. Try it for p(z) = 2z 4 − 3z 3 + z 2 + z − 1. If

none of the cases apply, one possibility is to work with p((1 − ε)z).
The following problem has been open since the 1930’s and is of current interest in
ergodic theory. Suppose p is a polynomial with integer coefficients

p(z) = an z n + . . . + a0

with |an | = |a0 | = 1. Let z1 , z2 , . . . , zk be the roots in |z| < 1. If p has at least one
Q
root in |z| < 1, how big can |zj | be? It is conjectured that

k
Y
|zj | ≤ .8501371...
p=1

This value is achieved for the polynomial

p(z) = z 10 + z 9 − z 7 − z 6 − z 5 − z 4 − z 3 + z + 1.

Eight zeros lie on the unit circle, one zero z1 is outside the unit circle and the last is
1/z1 which is inside the unit disk.
92
 VI

Elementary Maps

In this chapter we will study the mapping properties of the elementary functions and
their compositions. The emphasis will be on the behavior of linear fractional transfor-
mations and the power, trigonometric, and exponential functions related to the familiar
functions of a real variable. These functions are all built from linear functions a + bz,
P n
ez = z /n!, and its locally defined inverse log z using algebraic operations and compo-
sition.
2
4

1.5
3

2 1

1 0.5

0 0

−1
−0.5

−2
−1

−3
−1.5

−4
−2
−3 −2 −1 0 1 2 3 4 5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure VI.1 A rational function.

To facilitate our study, we will illustrate these functions using color pictures. The
right-hand plot in Figure VI.1 shows a polar grid on the plane, where rays are colored
using a standard color wheel in counter-clockwise order beginning along the negative reals:
red, yellow, green, cyan, blue, magenta, red. Circles of radius (1 + ε)n , n = −6, . . . , 6 are
also plotted using a gray scale, increasing in darkness with the modulus except for the
unit circle which is plotted in black with a thicker line width for emphasis. A “picture” of
a complex valued function f can be given by plotting points z using the same color that
f (z) has on the polar grid. For example, the left-hand plot in Figure VI.1 shows the plot
of a rational function. The rational function is a map from the left-hand picture to the

93
94 VI. Elementary Maps

polar grid on the right. Notice that the colors near z = 3 cycle twice around in the same
order as in the polar grid on the right. This means that there is a zero of order 2 at z = 3.
The colors near 2i and near −2i cycle once in the opposite or clockwise order. This means
that the function has poles of order one at ±2i. In fact, it is a picture of the function
(z − 3)2 /(z 2 + 4). The preimage of the unit circle is black.

§1. Linear Fractional Transformations.

A Linear Fractional Transformation (or LFT) is a non-constant rational function

of the form
az + b
T (z) = , (1.1)
cz + d
where a, b, c, d are complex constants. It is also called a Möbius Transformations. The
representation is not unique since the same function is obtained if a, b, c, d are replaced
by ta, tb, tc, td with t 6= 0. Various normalizations are used, depending on the application.
The four basic types of LFTs are:

Translation: T (z) = z + b,

Rotation: T (z) = eiθ z,

Dilation: T (z) = az for a > 0, and

1
Inversion: .
T (z) =
z
The translation above shifts every point by the vector b. The rotation rotates the
plane by an angle θ. The dilation expands (if a > 1) or contracts (if a < 1). The inversion
is best understood by writing z = reit . Then the argument of 1/z is −t and the length of
1/z is the reciprocal of the length of z.
An LFT can be build out of these examples using composition. If c = 0 in (1.1) then
a b
T = z+ . (1.2)
d d
 §1: Linear Fractional Transformations. 95
 
 
In this case T is a dilation by  ad , a rotation by arg ad followed by a translation by b
d. If
c 6= 0, then we can rewrite (1.1) as

bc − ad 1 a
T (z) = 2 d
+ . (1.3)
c (z + c ) c
 
d  , a rotation by
In this case, T is a translation by c
, an inversion, a dilation by  bc−ad
c 2

bc−ad a
arg c2
followed by a translation by c
.
Note that by (1.2) and (1.3), T is non-constant if and only if bc − ad 6= 0.

Proposition 1.1. The LFTs form a group under composition.

Proof. If S is an LFT, then for constants a, b, eiθ , it is not hard to check that S + b, eiθ S,
aS and 1/S are all LFTs, so by (1.2) and (1.3) LFTs are closed under composition. The
inverse of (1.1) is easily found to be

dw − b
z= . 
−cw + a

If T is given by (1.1) or (1.3) with c 6= 0, then T is meromorphic in C with a simple

pole at − dc . Moreover T extends to be analytic at ∞ with T (∞) = ac . Thus T extends to
be a one-to-one map of the extended plane, or the Riemann sphere onto itself. The next
proposition says that LFTs are the only such maps.

Theorem 1.2. If f is analytic on C \ {z0 } and one-to-one then f is an LFT.

Proof. Using a translation, we may assume z0 = 0. Then f has a Laurent series expansion
∞
X
f (z) = an z n .
n=−∞

If f has an essential singularity at 0, then by Theorem V.3.4 the image of every (punctured)
neighborhood of 0 is dense in C. In particular, if B = {z : |z −1| < 21 } then there is a ζ ∈
/B
with f (ζ) in the open set f (B). But then there is z ∈ B with f (z) = f (ζ), contradicting
the assumption that f is one-to-one. If f has a pole of order n at 0, then 1/f has a zero of
96 VI. Elementary Maps

order n at 0. Since 1/f is one-to-one, we must have n ≤ 1 by Corollary III.3.3. Applying

the same argument to f (1/z) we conclude that

a
f (z) = + b + cz,
z

for some constants a, b, c. If a 6= 0 and c 6= 0, then

cz 2 + (b − w)z + a = 0

has two roots in C \ {0}, counting multiplicity, contradicting the assumption that f is
one-to-one. Thus either a = 0 or c = 0, but not both since f is non-constant. In either
case, f is an LFT. 

In particular, Theorem 1.2 shows that if f is entire and one-to-one, then f (z) = cz +b,
for some constants c and b. Another consequence of Theorem 1.2 is that these maps are
“onto” maps. We have now identified the analytic automorphisms of the plane, the sphere,
and the disk (Exercise III.10). In this context, an analytic automorphism of the sphere
is a function satisfying the hypotheses of Theorem 1.2. As we will see in Chapter XVI,
subgroups of these automorphisms will allow us to transplant the study of analytic and
meromorphic functions on any domain, indeed on any Riemann surface, to the study of
functions on the disk, plane or sphere.

The LFTs are characterized geometrically by the next proposition. A “circle” or

generalized circle is a circle or a straight line. As we saw in Section I.3, “circles”
correspond precisely to the circles on the Riemann sphere. Lines lift to circles through the
north pole. Similarly a “disk” is a region (in C∗ ) bounded by a “circle”.

Theorem 1.3. LFTs map “circles” onto “circles” and “disks” onto “disks”.

Proof. We need only check this for the four basic types of LFTs. If |z − c| = r then
|az − ac| = |a|r, and |(z − b) − (c − b)| = r so that rotations, dilations and translations
map circles to circles. The equation of a straight line is given by

Re(c(z − b)) = 0,
 §1: Linear Fractional Transformations. 97

since we can translate the line so it passes through 0 then rotate it to correspond to the
imaginary axis. Rotations, dilations and translations map lines to lines exactly as in the
case of circles. To check that inversions preserve “circles”, suppose |z − c| = r and set
w = 1/z. Multiply out | w1 − c|2 = r 2 . If r 2 = |c|2 , then Re 2cw = 1, the equation of a line.
If r 2 6= |c|2 , then by completing the square we obtain
 2  2
 c  r
w −  = ,
 r 2 − |c|2  r 2 − c2

which is the equation of a circle. A similar reasoning for the image of a line is left to the
reader to verify.
The equation of a disk is found by replacing the equal sign in the equation for a circle
with < or >, so that the proof of the statement for “disks” follows in a similar way. 

2 2

1.5 1.5

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1

−1.5 −1.5

−2 −2
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure VI.2 The Cayley Transform.

The most common example is the conformal map from the upper half-plane H onto
the disk D, sometimes called the Cayley Transform (see Figure VI.2):

z−i
C(z) = .
z+i

Notice the order of the coloring around i and −i. Level curves (where C has constant
modulus) and curves where C has constant argument are circles, as Theorem 1.3 guaran-
tees. The real line is colored black. Indeed, the distance from x ∈ R to i is equal to the
98 VI. Elementary Maps

distance to −i, so that C maps R ∪ {∞} into ∂D, and therefore by Theorem 1.3 onto ∂D.
Since C(i) = 0, the image of H must be D by Theorem 1.3. Another way to see that the
image of H is D is to note that if z ∈ H then the distance from z to i is less than the
distance from z to −i, so that |C(z)| < 1 and similarly |C| > 1 on the lower half-plane.
Since C maps C∗ onto C∗ , the image of H must be D. The Cayley transform can be used,
for example, to transform an integral on R to an integral on ∂D and vice-versa.

Conformality can also be used to determine the image of an LFT. For example the
LFT given by T (z) = (z − 1)/(z + 1) is real-valued on R so it maps the unit circle to a
“circle” which is orthogonal to R at the image of 1, namely 0, and passes through ∞, the
image of −1. Thus it maps the unit circle to the imaginary axis, and since T (0) = −1, it
maps D onto the left half-plane {Rez < 0}.

The proof of the next proposition is useful for constructing LFTs.

Theorem 1.4. Given z1 , z2 , z3 distinct points in C∗ , and w1 , w2 , w3 distinct points in C∗ ,

there is a unique LFT, T , such that

T (zi ) = wi , (1.4)

for i = 1, 2, 3.

In the statement of Theorem 1.4, C∗ is the extended plane, so that ∞ is included in

the possibilities.

Proof. First suppose that w1 = 0, w2 = ∞ and w3 = 1. Set

  
z − z1 z3 − z2
T (z) = .
z − z2 z3 − z1

Then T (zi ) = wi , i = 1, 2, 3. For the general case, choose LFTs R and S so that R(z1 ) =
S(w1 ) = 0, R(z2 ) = S(z2 ) = ∞ and R(z3 ) = S(w3 ) = 1, then

T = S −1 ◦ R
 §2: Exp and Log. 99

satisfies (1.4). If U is an LFT satisfying (1.4) then

az + b
S ◦ U ◦ R−1 =
cz + d
maps 0 to 0 so that b = 0. It maps ∞ to ∞ so that then c = 0, and it maps 1 to 1 so that
a = d, and U (z) = S −1 ◦ R(z). 

One additional property of LFTs that is sometimes useful for determining their image
is: if T is an LFT and if z1 , z2 , z3 are three points on the boundary of a “disk” D such
that D lies to the left of ∂D as ∂D is traced from z1 to z2 to z3 , then T (D) lies to the left
of T (∂D) as it is traced from T (z1 ) to T (z2 ) to T (z3 ). For example, the unit disk D lies to
the left of the unit circle as it is traced from 1 to i to −1. If T (z) = (z − 1)/(z + 1) then
T (1) = 0, T (i) = i and T (−1) = ∞. Thus the image of ∂D is the imaginary axis, since it is
the unique “circle” through 0, i, ∞ and the image T (D) must lie to the left of the imaginary
axis as it is traced from 0 to i to ∞. Thus T (D) is the left half-plane {z : Rez < 0}. The
reason for this is that conformal maps preserve angles (including direction) between curves
and LFTs are conformal everywhere, in particular on the boundary of D.

§2. Exp and Log.

In Exercise II.5, we encountered the function

ez = ex eiy = ex (cos y + i sin y).

2
4

1.5
3

2 1

1 0.5

0 0

−1
−0.5

−2
−1

−3
−1.5

−4
−2
−4 −3 −2 −1 0 1 2 3 4 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure VI.3 The maps ez and log z.

100 VI. Elementary Maps

It maps the horizontal line y = c onto the ray arg z = c from 0 to ∞, and it maps
each segment of length 2π in the vertical line x = c onto the circle |z| = ec . By Exercise
d z
II.5(f), dz
e = ez , which is non-zero, so ez has a (local) inverse in a neighborhood of each
point of C \ {0}. This section is about the inverse of ez .
We begin with a fundamental consequence of Cauchy’s Theorem.

Corollary 2.1. Suppose f is analytic on a simply connected domain Ω and f (z) 6= 0 for
all z ∈ Ω. Then we can define g(z) = log f (z) to be analytic on Ω.

The conclusion of Corollary 2.1 is that there is an analytic function g such that

f (z) = eg(z) . (2.1)

Note that g(z) + 2πi is another solution, so it is not unique. We are not claiming that log z
can be defined on the range f (Ω) of f on Ω and then compose this function with f . The
function g is locally the composition of a function log z and f , but there might not be a
function h defined on all of f (Ω) such that z = eh(z) . For example, the LFT (z − 1)/(z + 1)
maps the disk D onto the left half-plane {Rez < 0} and so the function f (z) = e(z−1)/(z+1)
maps the D onto D \ {0} and is non-vanishing. Thus f satisfies the hypotheses of Corollary
2.1 and indeed g(z) = log f (z) = (z − 1)/(z + 1) works. However, we cannot define log z
as an analytic function on f (D) = D \ {0} since arg z = Im log z will increase by 2π as a
circle is traced counterclockwise about 0.

Proof. If g satisfies (2.1) then g ′ = f ′ /f . Fix z0 ∈ Ω and define

Z
f ′ (w)
g(z) = dw,
γz f (w)

where γz is any curve in Ω beginning at z0 and ending at z. If γ1 and γ2 are curves in Ω

from z0 to z, then γ1 − γ2 is a closed curve in Ω which is homologous to 0 by Theorem
V.2.2. Since f 6= 0 on Ω, f ′ /f is analytic on Ω and so by Cauchy’s Theorem
Z Z Z
f ′ (w) f ′ (w) f ′ (w)
0= dw = dw − dw.
γ1 −γ2 f (w) γ1 f (w) γ2 f (w)
 §2: Exp and Log. 101

Thus g(z) does not depend on the choice of γz . If z1 ∈ Ω and |h| is small then we may
take γz1 +h = γz1 + σh , where σh ⊂ Ω is the straight line segment from z1 to z1 + h. Thus
   Z  ′  
 g(z1 + h) − g(z1 ) f ′ (z1 )   1 f (w) f ′ (z1 ) 
 −  =  − dw 
 h f (z1 )   h σh f (w) f (z1 ) 

 ′ 
 f (w) f ′ (z1 )  ℓ(σh )
≤ sup  −  <ε
w∈σh f (w) f (z1 )  |h|

if |h| is small, since f ′ /f is continuous. Thus g ′ exists and g ′ = f ′ /f . By Corollary IV.2.4

g is analytic.
To compare f and eg , set
f
F = = f e−g .
eg
Then
F ′ = f ′ e−g − f g ′ e−g = f ′ e−g − f ′ e−g = 0.

This implies F is a constant, and evaluating at z0 , F (z0 ) = f (z0 ). Thus

f (z) = f (z0 )eg(z) = eg(z)+a ,

where a is any complex number with ea = f (z0 ), proving Corollary 2.1. 

Note that if g and h satisfy f = eg = eh on a region Ω and if g(z0 ) = h(z0 ) for some
z0 ∈ Ω, then g = h in a neighborhood of z0 because log z has a unique inverse in a (small)
neighborhood of f (z0 ) with log f (z0 ) = g(z0 ). By the Uniqueness Theorem, g = h in Ω.
For example, the function z is non-zero on the simply connected domain C \ (−∞, 0].
Then log z, with log 1 = 0, is the function given by

log z = log |z| + i arg z, (2.2)

where −π < arg z < π. If instead we specified that log 1 = 2πi then (2.2) holds with
π < arg z < 3π. However, if Ωγ = C \ γ where γ is the spiral given in polar coordinates
by r = eθ , −∞ < θ < ∞, then Ωγ is simply connected and Im log z is unbounded on Ωγ .
In this case we can still specify, for example, log(−1) = πi and this uniquely determines
102 VI. Elementary Maps

the function log z on Ωγ . Figure VI.3 probably gives more insight into the log function
than the inverse image of a polar grid. If Ω is any simply connected region contained in
the right-hand plot of Figure VI.3 which omits the point w = 0, then log z maps it to a
corresponding region on the left-hand plot. A horizontal strip of height 2π in the left-hand
plot of Figure VI.3 is the image of the polar grid, slit along a ray, by the function log z.
Any region contained in the left-hand plot of Figure VI.3 with the property that each
vertical line hits the region in a segment of length at most 2π is the image of a simply
connected region by the map log z. The complement of the spiral r = eθ in the right-hand
plot is mapped by log z to a strip in the left-hand plot lying between two lines parallel to
y = x in the left-hand plot. A vertical shift by an integer multiple of 2π is also the image
of the same region by log z, but with a different choice of log(−1).

Notice that with the definition (2.2), log(zw) 6= log(z) + log(w) if arg(z) + arg(w) is
not in the interval (−π, π).

The hypothesis that Ω is simply connected is essential in Corollary 2.1. If Ω is not

simply connected, then there exists a point a ∈ C \ Ω and a curve γ ⊂ Ω so that n(γ, a) =
1
R dz
2πi γ z−a 6= 0, by Theorem V.2.2. The function f (z) = z − a is non-zero on Ω and if

log(z − a) could be defined as an analytic function, then f ′ (z) = 1/(z − a) by the chain
R
rule. By the Fundamental Theorem of Calculus γ f ′ (z)dz = 0, contradicting n(γ, a) 6= 0.
Throughout this course log denotes the natural log, not base 10. Some books call log z a
“multiple-valued function” which is a bit of a contradiction in terms. We will only consider
log z on domains where it can be defined as a function, or we will define log f as an analytic
function, as in Corollary 2.1.

§3. Power Maps

If Ω is a simply connected region not containing 0, and if α ∈ C, we define

z α = eα log z .

where log z can be specified by giving its value at one point z0 ∈ Ω. Then z α is an analytic
function on Ω. For example suppose Ω = {z : Rez > 0} and define log 1 = 0. If z = reit ,
where −π/2 < t < π/2, then z 1/4 = r 1/4 eit/4 .
 §3: Power Maps 103
2 2

1.5 1.5

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1

−1.5 −1.5

−2 −2
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2 −1 −0.5 0 0.5 1 1.5 2 2.5 3

Figure VI.4 Maps z 4 and z 1/4 .

The image of the sector {z : | arg z| < β} is the sector {z : | arg z| < β/4}. Points z on the
circle |z| = r are mapped to points on the circle |z| = r 1/4 . The map z 1/4 is conformal
in Ω, but it is not conformal at 0. Indeed angles are multiplied by 1/4 at 0. If we define
log 1 = 2πi instead of 0, then the image sector is rotated by the angle 2π/4. Test your
understanding by showing that there are exactly four possible definitions of z 1/4 in any
simply connected region not containing 0.
How would you define log z on C \ [0, +∞) and what would be the image of this region
by the map z 1/π ? There are in fact infinitely many definitions of z 1/π on this region.

A slightly more complicated function is

1+i
z 2

defined on Ω0 = C \ (−∞, 0]. If log(1) = 0, then the function ϕ(z) = log z is analytic on
Ω0 and has image equal to the horizontal strip Ω1 = {z : |Imz| < π}. The image of Ω1 by
1+i
the map 2 z is the strip

Ω2 = {x + iy : x − π < y < x + π}.

The function ez maps the line y = x onto a spiral S given in polar coordinates by r = eθ .
The image of y = x + c is the rotation of S by the angle c. Thus ez is analytic and
1+i
one-to-one on Ω2 with image Ω3 = C \ {−S}. The composition of these maps is z 2 ,
104 VI. Elementary Maps
1+i
which is then a one-to-one analytic map of Ω0 onto Ω3 . Notice that z 2 does not extend
continuously to (−∞, 0]. In Figure VI.5 we show the inverse function z 1−i since it will
map to the polar grid on Ω0 = C \ (−∞, 0], displaying the spirals as the preimages of rays.
The logarithm can be defined in the complement of a spiral colored red, or on C \ [0, ∞)
(as the computer chooses). The resulting function has continuous argument in C \ {0},
but the absolute value increases by e2π ≈ 535 around any circle centered at 0.

15 2

1.5
10

5
0.5

0 0

−0.5
−5

−1
−10

−1.5

−15
−2
−15 −10 −5 0 5 10 15 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure VI.5 The map z 1−i , the inverse of z (1+i)/2 .

The function z α has been used to explain one of M.C. Escher’s lithographs. See
Exercise 4.

§4. The Joukovski Map

The next function we will consider in this chapter is

1 1
w(z) = z+ ,
2 z

which is analytic in C \ {0}. By the quadratic formula

p
z =w± w2 − 1 (3.1)

so that w is two-to-one unless w2 = 1. Since w(z) = w(1/z), the two roots in (3.1) are
reciprocals of each other, one inside D and one outside D or else complex conjugates of
each other on ∂D.
 §4: The Joukovski Map. 105
4 2

3 1.5

2 1

1 0.5

0 0

−1 −0.5

−2 −1

−3 −1.5

−4 −2
−4 −3 −2 −1 0 1 2 3 4 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure VI.6 The map (z + 1/z)/2.

If we write z = reit , then

1 1 1 1 i 1
w= z+ = r+ cos t + r− sin t.
2 z 2 r 2 r

If we also write w = u + iv then

 2  2
u v
1 + 1 = 1,
2
(r + 1r ) 2
(r − 1r )

which is the equation of an ellipse, unless r = 1. Thus for each r 6= 1, the circles of radius
r and 1/r are mapped onto the same ellipse. The circle of radius r = 1 is mapped onto
the interval [−1, 1]. We leave as an exercise for the reader to show that the image of a ray
from the origin to ∞ which is not on a coordinate axis is a branch of a hyperbola which is
perpendicular to each ellipse given above, by conformality.
106 VI. Elementary Maps
2 2

1.5 1.5

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1

−1.5 −1.5

−2 −2
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2
2 2

1.5 1.5

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1

−1.5 −1.5

−2 −2
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure VI.7 The inverses of (z + 1/z)/2.

The function w is a one-to-one analytic map of C \ D onto C \ [−1, 1], and a one-to-one
analytic map of D\{0} onto C\[−1, 1]. The function w is also analytic on H = {z : Imz > 0}
and one-to-one, since Im1/z < 0 if z ∈ H. The image of H by the map w is the region
Ω2 = C \ {(−∞, −1] ∪ [1, ∞)}.
The inverse of w is given by (3.1) in each case, but we must make the correct choice
for the square root on the image region. If U is any simply connected domain with ±1 ∈
/U
then we can define log(w2 − 1) so as to be analytic in Ω by Corollary 2.1, and thus

p 1 2
w+ w2 − 1 = w + e 2 log(w −1)

is analytic and one-to-one on U with inverse function 21 (z + 1/z).

 §5: Trigonometric Functions. 107

For example if U = C∗ \ [−1, 1], write

r
p 1
w2 − 1 = w 1− .
w2
p
The function 1 − 1/w2 is analytic and non-zero on U . Define ϕ(w) = 1 − 1/w2 so that
ϕ(∞) = 1. More concretely, the image of U by the map 1−1/w2 is contained in C\(−∞, 0]
p
so if log ζ = log |ζ| + i arg ζ with −π < arg ζ < π then 1 − 1/w2 is the composition of
ζ = 1 − 1/w2 and exp( 12 log ζ), and therefore analytic. Then
 r 
1
ψ(w) = w 1 + 1− 2
w

is an inverse to (z + 1/z)/2 on U . By the preceeding discussion, there are two inverses to

(z + 1/z)/2 on U . Since ψ(w) → ∞ as w → ∞, ψ must be the inverse to (z + 1/z)/2 with
range C \ D. See Figure VI.7, upper-left. The inverse with range D \ {0} is
 r 
1
ψ1 (w) = w 1 − 1 − 2
w

which tends to 0 as w → ∞. See Figure VI.7, upper-right. Notice that the pictures are
not continuous across (−1, 1), but are continuous across (−∞, −1) ∪ (1, ∞). The reader is
invited to test their understanding by finding the inverses on C \ {(−∞, −1] ∪ [1, +∞)}.
See Figure VI.7, lower-left and lower-right, both of which continuous across (−1, 1) but
not (−∞, −1) ∪ (1, ∞). Notice also the ellipses and hyperbolas where the functions have
constant modulus and constant argument.

In the next section we will use the function (z +1/z)/2 and the closely related function
(z − 1/z)/2 which can be understood as a composition
 
1 1 1 1
z− = −i · (iz) + .
2 z 2 (iz)
1
It is the composition of rotation by π/2 followed by 2 (z + 1/z) followed by rotation by
−π/2. Thus circles and lines through 0 are mapped to ellipses and orthogonal hyperbolas.
The ellipses have semi-major axis along the imaginary axis. The unit circle is mapped to
the interval [−i, i].
108 VI. Elementary Maps

§5. Trigonometric Functions.

We define

eiz + e−iz eiz − e−iz

cos z = and sin z = .
2 2i

Then cos z and sin z are entire functions satisfying

cos z + i sin z = eiz ,

(cos z)2 + (sin z)2 = 1,

d d
cos z = − sin z and sin z = cos z.
dz dz

These functions agree with their usual calculus definitions when z is real. However, we
know by Liouville’s Theorem that they cannot be bounded in C. The function cos z is
best understood by viewing it as the composition of the maps iz, ez and 12 (z + 1/z). For
example, the vertical strip {z : |Rez| < π} is rotated to the horizontal strip {z : |Imz| < π}
by the map iz. This horizontal strip is mapped onto C \ (−∞, 0] by the map ez . The
composition eiz maps vertical lines to rays from 0 to ∞ and maps horizontal lines to
circles. Rays and circles are mapped by 12 (z + 1/z) to branches of hyperbolas and ellipses,
as we saw in section 3. Other trigonometric functions are defined using sin and cos, for
example

sin z
tan z = .
cos z

Hyperbolic trigonometric functions are also defined using the exponential function:

ez + e−z ez − e−z
cosh z = and sinh z = .
2 2
 §6: Constructing Conformal Maps. 109
6 2

1.5
4

2
0.5

0 0

−0.5
−2

−1

−4
−1.5

−6 −2
−6 −4 −2 0 2 4 6 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure VI.8 The map cos(z).

The inverse trigonometric functions can be found by working backward. For example
to find arccos z, set z = (eiw + e−iw )/2, multiply by eiw and obtain a quadratic equation
in eiw , so that by the quadratic formula
p
eiw = z ± z 2 − 1.

Thus
p
arccos z = w = −i log(z ± z 2 − 1).
√
/ Ω then f (z) = z ± z 2 − 1 is an analytic
If Ω is a simply connected domain such that ±1 ∈
√
function, as seen in Section 4. If z ± z 2 − 1 = 0 then z 2 = z 2 − 1 which is impossible.
Thus f is a non-vanishing function on Ω and by Corollary 2.1, we can define log f as an
analytic function on Ω. Thus arccos is analytic on any simply connected region which does
π
not contain ±1. To find the arccos with arccos(0) = 2, it is best to write it in the form
p
arccos z = w = −i log(z + i 1 − z 2 ).
√
with 1 = 1 and log(i) = iπ/2.

§6. Constructing Conformal Maps.

In this section we will use the functions we’ve studied in this chapter to construct
conformal maps. In modern usage, the phrase conformal map means a one-to-one ana-
lytic map. The entire function f (z) = ez is conformal everywhere, since its derivative is
110 VI. Elementary Maps

everywhere non-zero, but it is not a conformal map on C because it is not one-to-one. It

is a conformal map on {z : |Imz| < π}, for example. The list below is not exhaustive, but
rather is meant to illustrate the techniques which can be used at this point in the course.

D onto D with f (z0 ) = 0 and f ′ (z0 ) = eit : As we saw in Exercise III.10, the conformal
maps of D onto D are given by
 
z−a
f (z) = c ,
1 − az

where a and c are constants with |a| < 1 and |c| = 1. Setting a = z0 and choosing the
appropriate c will work.

Sector: A conformal map of a sector Ω = {z : a < arg z < b} onto D can be constructed
in steps.

f (z) = z α = eα log z

where α = π/(b − a) will map Ω onto a sector with opening π, a half-plane. The choice of
log z is already given in the description of Ω. A rotation z → eit z will map the half-plane
onto H, and the Cayley transform (z − i)/(z + i) will map H onto D. It is usually sufficient
to describe a conformal map as a composition of a sequence of simpler conformal maps.

Intersection of Disks: If Ω is the intersection of two disks, then in order to map Ω onto
D, find the two points c, d where the bounding circles meet. The map

z−c
z−d

will map each disk onto a “disk” with 0 and ∞ on its boundary, and hence the image of
Ω is the intersection of two half-planes forming a sector at 0. Now apply the Sector map
constructed above. Note that the region outside the union of two disks is the intersection
of two “disks” in the extended plane if we add the point at ∞.
 §6: Constructing Conformal Maps. 111
2

1.5
1 1

1
0.5 0.5
0.5

0 0 0

−0.5
−0.5 −0.5

−1

−1 −1
−1.5

−1.5 −1.5 −2
−1.5 −1 −0.5 0 0.5 1 1.5 −1.5 −1 −0.5 0 0.5 1 1.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure VI.9 Four intersections of disks meeting at angle π/4.

Figure VI.9 shows four regions each of which has bounding circles meeting at 0 and
1 +i at an angle of π/4. The point at ∞ is viewed as an interior point of one of the regions.
The map w = z/(z − (1 + i)) maps each region to a sector. The map ζ = w4 will map each
sector onto the right half plane, and (ζ − 1)/(ζ + 1) maps the right half plane onto D.

Half-plane with a slit: The region Ω = H \ I, where I is the segment [0, i] on the
imaginary axis, can be mapped onto H by the map
p
z 2 + 1,
√
where −1 = i. Indeed the image of Ω by the map z 2 + 1 is the slit plane C \ [0, +∞). The
√
“branch” of the square root is uniquely determined by the requirement that −1 = i. It
can be given more explicitly as exp( 21 log z) where 0 < arg z = Im log z < 2π. Notice that
the two “sides” of the slit [0, i] correspond to the intervals [−1, 0] and [0, 1] in the closure
of the upper half plane. The interval [0, +∞) in the boundary of the slit plane corresponds
to the interval [1, +∞) in the image region, the upper half plane.

Half-plane minus a circular arc: The region Ω = H \ A, where A is an arc lying on a

circle C which is orthogonal to R at 0 can be mapped to D by first applying a conformal
map az/(1 − z/b) of H onto H where b is the other point of intersection of C and R, and
a > 0. The image of A must lie on a “circle” through ∞ and be perpendicular to R since
LFTs are conformal. Thus the image of A is an interval [0, ic] on the imaginary axis. We
can choose a > 0 so that c = 1 and then the slit half-plane example applies. The circular
slit is “opened up” to two intervals in R and the real line is mapped to the remaining
portion of R.
112 VI. Elementary Maps
1.5 1.8
2.5
1.6

1.4
2

1 1.2

1 1.5

0.8

1
0.5 0.6

0.4
0.5

0.2

0 0 0

−0.2
−0.2 0 0.2 0.4 0.6 0.8 1 1.2 −1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1 −1.5 −1 −0.5 0 0.5 1 1.5

Figure VI.10 Mapping a circularly slit half plane

Between branches of a hyperbola: To map H onto the region between two branches
of a hyperbola, first map H onto a sector symmetric about the y-axis using eit z α with the
proper choice of the rotation angle t and opening α. Then apply 12 (z + 1/z). See Section
4. In Figure VI.11, we used the standard coloring scheme on the domain, instead of the
range. So in fact these pictures were drawn using the inverses of the maps involved, in the
reverse order.

1.5
2.5
2.5

1
2
2

0.5
1.5
1.5

0
1
1

−0.5
0.5
0.5

−1
0
0

−1.5
−1.5 −1 −0.5 0 0.5 1 1.5 −1.5 −1 −0.5 0 0.5 1 1.5 −1.5 −1 −0.5 0 0.5 1 1.5

Figure VI.11 Region between branches of a hyperbola.

Exterior of an Ellipse: To map C \ D onto the exterior of an ellipse, apply the map
1
z → rz, then 2 (z + 1/z). See Section 4. In Figure VI.12, as in Figure VI.11, the domain
is given the standard (polar) coloring, so the inverses of the two maps are used.
 §6: Constructing Conformal Maps. 113
3 4 3

3
2 2

1 1
1

0 0 0

−1
−1 −1

−2

−2 −2
−3

−3
−3 −2 −1 0 1 2 3 −4
−4 −3 −2 −1 0 1 2 3 4 −3
−3 −2 −1 0 1 2 3

Figure VI.12 Exterior of an ellipse

Below a Parabola: To map H onto the region below the parabola y = x2 first apply the
map z → −iaz + b where a > 0 and b > 0. The image of H is the half-plane {Rez > b}.
The image of this region under the map z 2 is the exterior of a parabola. Applying the map
z → ciz + id with c < 0 will result in a map to the region below a parabola. The reader
can check that a proper choice of a, b, c will give the desired map.

3
2.5

2
2

1.5

−1

0.5

−2

−3
−1.5 −1 −0.5 0 0.5 1 1.5 −3 −2 −1 0 1 2 3

Figure VI.13 Below a parabola

To map D or H onto any of regions bounded by a conic section other than those given
in 8, 9, and 10 will be covered later in the course.

Strip: To map the strip {z : 0 < Rez < 1} onto D, first apply the map eπiz . The image of
{Rez = c}, 0 < c < 1, is the ray reiπc , r > 0, so the image of the strip is H. Now apply
the Cayley transform.
114 VI. Elementary Maps

Half-Strip: To map the half-strip {z : 0 < Imz < π, Rez < 0} onto D, first apply the
map ez which has image D ∩ H. Now apply the sector example, or use the map 12 (z + 1/z)
and a map of a half plane to the disk.

3.5

2
3

2.5 1.5

2
1

1.5

0.5
1

0.5 0

0
−0.5

−3 −2.5 −2 −1.5 −1 −0.5 0 0.5 1 −1.5 −1 −0.5 0 0.5 1 1.5

Figure VI.14

Many other examples can be constructed by using combinations of the above ideas.
The conformal map in each of the examples above is a composition of a sequence of simpler
conformal maps. The inverse map can be found by composing the inverses of the simpler
functions in the reverse order. To find the conformal map of D onto a region Ω, it is usually
easier to discover the map from Ω onto D, then compute its inverse.

A natural question at this point is how unique are these maps? A conformal map of
Ω onto D can be composed with the linear fractional transformations of the form given in
the first example of this section and still map D onto D.

Proposition 6.1. If there exists a conformal map of a region Ω onto D, then given any
z0 ∈ Ω, there exists a unique conformal map f of Ω onto D such that

f (z0 ) = 0 and f ′ (z0 ) > 0.

Proof. If g is a conformal map of Ω onto D, set a = g(z0 ). Then

z−a
h(z) = c
1 − az
 §6: Constructing Conformal Maps. 115

maps D onto D if |c| = 1, and f = h ◦ g maps Ω onto D with h(z0 ) = 0. Then

f ′ (z0 ) = cg ′ (z0 )/(1 − |a|2 ),

so that the proper choice of the argument of c will give h′ (z0 ) > 0. If k maps Ω onto D
with k(z0 ) = 0 and k ′ (z0 ) > 0, then H = k ◦ h−1 maps D onto D with H(0) = 0. By
Schwarz’s Lemma, |H(z)| ≤ |z| and |H −1 (z)| ≤ |z| so that |H(z)| = |z| and H(z) = cz,
with |c| = 1. Since H ′ (0) = k ′ (0)/h′ (0) > 0, we must have c = 1 and h = k. 

For example, to find a conformal map ϕ of H onto D such that ϕ(z0 ) = 0 and ϕ′ (z0 ) >
0, first apply the Cayley transform, then apply a linear fractional transformation of the
form given in the first example of this section. Here it is actually easier to first apply the
map
z − z0
f (z) =
z − z0
which maps H onto D because |f (z)| = dist(z, z0 )/dist(z, z0 ) < 1, when z ∈ H, and
is > 1 when z ∈ C \ H so that the image of H is D by Theorem 1.3. Then f ′ (z0 ) =
limz→z0 f (z)/(z − z0 ) = 2i/Imz0 . So −if will work.
Another natural question is: what regions can be mapped conformally onto D? The
next Proposition gives a necessary condition.

Proposition 6.2. If ϕ is a conformal map of a region Ω onto D, then Ω must be simply

connected.

Proof. Suppose γ is a closed curve contained in Ω and suppose a ∈ C \ Ω. Let f = ϕ−1 ,

which is analytic on D by Corollary 4.4. Then
Z Z
1 dw 1 f ′ (z)
n(γ, a) = = dz = 0,
2πi γ w − a 2πi ϕ(γ) f (z) − a

by Cauchy’s Theorem since f ′ /(f − a) is analytic in D. By Theorem V.2.2 Ω is simply

connected. 

In the next chapter we will prove that any simply connected region, other than C can
be conformally mapped onto D.
116 VI. Elementary Maps

We conclude this chapter by giving a picture of the function exp((z +1)/(z −1)) which
we encountered a couple of times so far. It is a composition of an LFT which maps the
unit disk onto the left half plane followed by the exponential function. The preimage of
the unit disk is black. Each level line is a circle through z = 1, where it has an essential
singularity. The image of each level line covers the corresponding circle in the standard
plane infinitely many times.
1.5 2

1.5
1

0.5
0.5

0 0

−0.5
−0.5

−1

−1
−1.5

−1.5 −2
−0.5 0 0.5 1 1.5 2 2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2

Figure VI.15 exp ((z + 1)/(z − 1)).

§7. Exercises.

1. Provide details of the proof of the statement about “disks” in Theorem 1.3. Also
provide the details for the image of a line by an inversion.
2. Suppose that C is a circular arc in D which is orthogonal to ∂D at a, b ∈ ∂D. Prove
there is a conformal map of D onto D which maps the interval (−1, 1) onto C so the
the image of 0 is the closest point on C to 0. No computations are needed to solve
this problem.
3. The proof of Corollary 1.2 should remind you of the proof that the winding number
is an integer. Explain this using the increase in the argument of a continuation of
log(z − a) along γ.
4. Draw a picture in the rectangle {z = x + iy : 0 < x < π, 0 < y < π}. Reflect the
picture about the line y = 0 and then reflect the union about x = 0. The picture
 §7: Exercises 117
1+i
can be extended to the plane by repeated reflections. Now apply the map e 2 z

and describe the image. Use the matlab program fcn.m provided with this course to
visualize this map and its inverse. A similar idea was used by M.C. Escher in one of
his lithographs. See R. de Smit and H.W. Lenstra, Jr., Notices A.M.S. 50(4), April
2003, 446-451.
5. The region Ω = {z : Imz > 0 and |z| > 1} is the intersection of two disks. Map
this region to a sector then to H. Find a conformal map of H onto H so that the
composition of these maps fixes 1, −1 and ∞. You should get (z + 1/z)/2. Why?

Construct a one-to-one analytic map f of the following regions Ω onto D. If z0 is

given, find the map with f (z0 ) = 0 and f ′ (z0 ) > 0. You may leave your answer as an
explicit sequence of maps, but you must show that each map does what you claim it does.
6. Ω = C \ (−∞, 0] z0 = 1
7. Ω = {z : 0 < Imz < 1}
8. Ω = {z : 0 < Imz < 1 and Rez > 0}
9. Ω = D ∩ {z : Imz > 21 } z0 = 34 i
10. Ω = {z : |z| < 1} ∩ {z : |z − 1| < 1} ∩ {z : Imz > 0}.
11. Ω = C∗ \ [0, 1] z0 = ∞. Here f ′ (∞) means limz→∞ z(f (z) − f (∞)).
12. Prove there is no one-to-one analytic mapping of {z : 0 < |z| < 1} onto {z : 0 < |z| <
∞}.

13. (Rain = complement of the Seattle Umbrella). Map C \ D onto

Ω = C∗ \ ({z : |z| ≤ 1 and Imz ≥ 0} ∪ {it : −1 ≤ t ≤ 0})

so that ∞ is mapped to ∞. Extra credit: put a handle on the umbrella (your choice
of a handle).
p
14. The function g(z) = log(1/(1 − z 2 )) can be defined so that it is analytic and one-to-
p
one on D as follows: Set f (z) = log[1/(1 − z)]/z), and g(z) = zf (z 2 ). Show that g is
118 VI. Elementary Maps

a one-to-one analytic map on D with Img continuous on ∂D but Reg not continuous.
In fact Reg is not bounded. Draw a picture. (For Fourier Series enthusiasts the
imaginary part of the power series for g gives a Fourier series of a continuous function
whose conjugate Fourier series is not even bounded).

The next two problems are not hard, if you get the right idea.
15. Suppose C and D are tangent circles, one inside the other. Find a circle C1 which
is tangent to both C and D, then for n ≥ 2 find Cn tangent to C, D and Cn−1 ,
with {Cj , j = 1, . . . , n} bounding disjoint open disks. Prove that this process can be
continued indefinitely and that the points of tangency of the Cn all lie on a circle.
16. Suppose C and D are non-intersecting circles, one inside the other. Choose C1 tangent
to both C and D. Again choose Cn tangent to C, D and Cn−1 , with {Cj , j = 1, . . . , n}
bounding disjoint open disks. Under some circumstances, the chain of circles may
“close up” with Cn tangent also to C1 . Show that for each n ≥ 3, there is a choice of
C and D so that the chain always closes up. Given circles C and D, show that the
property of “closing up” does not depend on the choice of C1 .
It is known that given a collection of disks in a simply connected region whose interiors
are disjoint, there is collection of disks in the unit disk with the same pattern of tangencies.
If the disks are small enough, the map from one set of disks to the other approximates the
conformal map between the region and the unit disk. See Circle Packing by K. Stephenson
for an introduction to the subject.
See Penrose and Rindler, Spinors and Space-Time, Camb. Univ. Press 1984, for the
connection between LFTs and Lorentz transformations in the theory of relativity.
 VII

Calculus of Residues

§1. Contour Integration and Residues.

In this chapter we will learn one of the main applications of complex analysis to
engineering problems, but we’ll concentrate on the math, not the engineering. Cauchy’s
Theorem says that if f is analytic in a region Ω and if γ is a closed curve in Ω which is
R
homologous to 0, then γ f (z)dz = 0. What happens if f has an isolated singularity at
a ∈ Ω, but otherwise is analytic? Expanding f in its Laurent series about a, we have
∞
X
f (z) = bn (z − a)n
n=−∞

∞
b−1 d X bn
= + (z − a)n+1 .
z − a dz n=−∞ (n + 1)
n6=−1

So if ∆ is a disc centered at a and contained in Ω then by the Fundamental Theorem of

Calculus, Z Z
dz
f (z)dz = b−1 = 2πib−1 , (1.1)
∂∆ ∂∆ z−a
provided ∂∆ is oriented in the positive or counterclockwise direction.

Definition. If f is analytic in {0 < |z − a| < δ} for some δ > 0, then the residue of f at
a, written Resa f , is the coefficient of (z − a)−1 in the Laurent expansion of f about z = a.

Theorem 1.1 (Residue Theorem). Suppose f is analytic in Ω except for isolated

singularities at a1 , . . . , an . If γ is a cycle in Ω with γ ∼ 0 and aj ∈/ γ, j = 1, . . . , n, then
Z X
f (z)dz = 2πi n(γ, ak )Resak f.
γ n

Usually the Residue Theorem will be applied to curves γ such that n(γ, ak ) = 0 or 1,
so that the sum on the right is 2πi times the sum of the residues of f at points “enclosed”

119
120 VII. Calculus of Residues

by γ. If f has infinitely many zeros clustering only on ∂Ω then we can shrink Ω slightly
so that it contains only finitely many aj and still have γ ∼ 0.

Proof. Let ∆k be a disc centered at ak , k = 1, 2, . . . , n, such that ∆m ∩ ∆k = ∅ if m 6= k.

Orient ∂∆k in the positive or counterclockwise direction. Then
X
γ− n(γ, ak )∂∆k ∼ 0
k

in the region Ω \ {a1 , . . . , an }. By Cauchy’s Theorem

Z n
X Z
f (z)dz − n(γ, ak ) f (z)dz = 0.
γ k=1 ∂∆k

Then Theorem 1.1 follows from (1.1). 

Here are some examples illustrating several useful techniques for computing residues.

e3z
(1) f (z) = has a simple pole at z = 2 and hence
(z − 2)(z − 4)
e6
Res2 f = lim (z − 2)f (z) = .
z→2 −2
The residue at z = 4 can be calculated similarly.

e3z
(2) g(z) = . We expand e3z in a series expansion about z = 2:
(z − 2)2
X∞
e6 e3(z−2) e6 3n n e6 3e6
g(z) = = (z − 2) = + + ...,
(z − 2)2 (z − 2)2 n=0 n! (z − 2)2 z−2

so that
Res2 g = 3e6 .

In this case lim(z − 2)2 g(z) is not the coefficient of (z − 2)−1 and lim(z − 2)g(z) is
infinite. Of course the full series for e3z was not necessary to compute the residue.
We can find the appropriate coefficient in the series expansion for e3z by computing
the derivative of e3z . More generally, if G(z) is analytic at z = a then

G(z) G(n−1) (a)

Resa = .
(z − a)n (n − 1)!
 §2: Examples 121

(3) Another trick that can be used with simple poles, when pole is not already written as
a factor of the denominator, is illustrated by the example h(z) = eaz /(z 4 + 1). Then
h has simple poles at the fourth roots of −1. If ω 4 = −1, then

(z − ω)eaz eaω
Resω h = lim = 4 +1 .
z→ω z4 + 1 limz→ω zz−ω

Note that the denominator is the limit of difference quotients for the derivative of
z 4 + 1 at z = ω and hence

eaz eaω ωeaω

Resω 4 = =− .
z +1 4ω 3 4

(4) Another method using series is illustrated by the example

π cot πz
k(z) =
z2

To compute the residue of k at z = 0, note that cot πz has a simple pole at z = 0 and
hence k has a pole of order 3, so that

π cot πz b−3 b−2 b−1

2
= 3 + 2 + + b0 + . . . .
z z z z

Then  
sin πz
π cos πz = (b−3 + b−2 z + b−1 z 2 + b0 z 3 + . . . .
z
Inserting the series expansions for cos and sin we obtain

π2 2 π3 2
π(1 − z + . . .) = (π − z + . . .)(b−3 + b−2 z + b−1 z 2 + . . . ,
2 6

Equating coefficients

π3 π3
π = πb−3 − = − b−3 + πb−1 ,
2 6
2
and Res0 k = − π3 .

§2. Examples.
122 VII. Calculus of Residues

In this section we compute some integrals using the Residue Theorem. The emphasis
is on the techniques instead of proving general results that can be quoted.

(a) If γ is the circle centered at 0 with radius 3, then

Z
e3z e6
dz = −2πi ,
γ (z − 2)(z − 4) 2

by the Residue Theorem and Example (1) above.

Z ∞
dx
(b) .
−∞ x4+1

We construct the contour γ consisting of the interval [−R, R] followed by the semi-
circle CR in H of radius R, with R > 1.

CR

z1 z2

Figure VII.1 Half-disc contour.

By the Residue Theorem,

Z R Z
f (z)dz + f (z)dz = 2πi(Resz1 f + Resz2 f ), (1.2)
−R CR

where z1 and z2 are the roots of z 4 + 1 = 0 in the upper half-plane H. Note that
Z  Z 2π
  Rdθ
 f (z)dz ≤ →0
 R4 − 1
CR 0
R
as R → ∞. Since the integral R (x4 + 1)−1 dx is absolutely convergent, it equals
RR
limR→∞ −R (x4 + 1)−1 dx, so that by (1.2), and Example (3) above
Z ∞
1 2π π
dx = − (z1 + z2 ) = √ .
−∞ x4 +1 4 2
 §2: Examples 123

The technique in Example (b) can be used to compute the integral of any rational
function with no poles on R if the degree of the denominator is at least 2 plus the
degree of the numerator. This latter condition is needed for the absolute convergence
of the integral.
Z 2π
1
(c) dθ.
0 3 + sin θ

Set z = eiθ . Then

Z 2π Z Z
1 1 dz 2dz
dθ = 1 = .
0 3 + sin θ |z|=1 (3 + 2i (z − 1/z)) iz |z|=1 z2 + 6iz − 1
√ √
The roots of z 2 + 6iz − 1 occur at z1 , z2 = i(−3 ± 8). Only i(−3 + 8) lies inside
|z| = 1. By the Residue Theorem and method for computing residues in Example (1)
or Example (3),
Z 2π
1 2 2π
dθ = 2πiResi(−3+√8) 2 =√ .
0 3 + sin θ z + 6iz − 1 8

The technique in Example (c) can be used to compute

Z 2π
R(cos θ, sin θ)dθ,
0

where R(cos θ, sin θ) is a rational function of sin θ and cos θ, with no poles on the unit
circle. An integral on the circle as in Example (c), can be converted to an integral on
the line using the Cayley transform z = (i − w)/(i + w) of the upper half plane onto
the disc. It is interesting to note that you obtain the substitution x = tan 2θ which
you might have learned in calculus.
Z ∞
cos x
(d) dx.
−∞ x2 + 1

A first guess might be to write cos z = (eiz + e−iz )/2, but if y = Imz then cos z ∼
e|y| /|z|2 for large |z|. This won’t allow us to find a closed contour where the part off
the real line has small contribution to the integral. Instead, we use eiz /(z 2 + 1) then
124 VII. Calculus of Residues

take real parts of the resulting integral. Using the same contour as in Example (b),
we have the estimate
Z  Z
 eiz  e−y πR
 dz ≤ |dz| ≤ 2 →0
 2  2
CR z +1 CR R − 1 R −1

as R → ∞, where y = Imz > 0. By the Residue Theorem and the method in Example
(3),
Z ∞ X
eix eiz ei·i π
2
dx = 2πi Res a 2 = 2πi = .
−∞ x +1 z +1 2i e
Ima>0

In this particular case, we did not have to take real parts, because the integral itself
is real since sin x/(x2 + 1) is odd.

§3. Fourier, Mellin and Inverse Laplace Transforms.

The technique in Example (d) can be used to compute

Z ∞
f (x)eiλx dx (1.3)
−∞

for λ > 0, provided f is meromorphic in the closed upper half-plane H ∪ R with no

poles on R and |f (z)| ≤ K/|z|1+ε for some ε > 0 and all large |z| with Imz > 0. If the
integral (1.3) is desired for all real λ, then for negative λ use a contour in the lower
half-plane, provided is meromorphic and |f (z)| ≤ 1/|z|1+ε in Imz < 0. As the reader
can deduce, if f is meromorphic in C satisfies this inequality for all |z| large, then
f is rational and the degree of the denominator is at least 2 plus the degree of the
numerator. The integral in (1.3) is called the Fourier Transform of f , as a function
of λ.
Z ∞
x sin λx
(e) 2
dx = πe−λ .
−∞ x +1

Example (e) cannot be done by the method in Example (d) because the integrand
R
does not decay fast enough to prove CR |f (z)||dz| → 0, where f (z) = zeiλz /(z 2 + 1).
Indeed it is not even clear that the integral in Example (e) exists.
 §3: Fourier, Mellin and Inverse Laplace Transforms. 125

Let γ = γ1 + γ2 + γ3 + γ4 where γ1 = [−A, B], γ2 = {B + iy : 0 ≤ y ≤ A + B},

γ3 = {z = x + i(A + B) : B ≥ x ≥ −A}, and γ4 = {z = −A + iy : A + B ≥ y ≥ 0},
where γ is oriented as indicated in Figure VII.2
γ3
B + i(A + B)

γ4 γ2

−A γ1 B

Figure VII.2 Rectangle contour.

To prove convergence of the integral in Example (e), we will let A and B tend to ∞
independently, and use the estimate |z/(z 2 + 1)| ≤ |z|/(|z|2 − 1) ≤ 2/|z| when |z| > 2.
For A and B large,
Z  Z B
 zeiλz  2 2e−λ(A+B)
 dz ≤ e−λ(A+B)
dx ≤ (A + B) → 0,
 z2 + 1 
γ3 −A A + B A+B

as A + B → ∞. Also
Z  Z A+B
 zeiλz  2 −λy 2 (1 − e−λ(A+B) )
 dz ≤ e dy ≤ → 0,
 z2 + 1  B B λ
γ2 0

as B → ∞. A similar estimate holds on γ4 as A → ∞. By the Residue Theorem

Z B
xeiλx zeiλz 2πi · ie−λ
lim 2
dx = 2πi Res i 2
= = iπe−λ . (1.4)
A,B→∞ −A x +1 z +1 2i

Indeed by our estimates, the integrals over γ2 , γ3 , and γ4 tend to 0 as A and B tend
to ∞ so that the limit on the left side of (1.4) exists and (1.4) holds. Example (e)
follows from Proposition 1.2 by taking the imaginary parts.
The technique in Example (e) can be used to compute (1.3) with the (weaker) as-
sumption |f (z)| ≤ K/|z| for large |z|.
Z ∞
sin x
(f) dx.
−∞ x
126 VII. Calculus of Residues

The main difference between Example (e) and Example (f) is that the function f (z) =
eiz /z has a simple pole on R. The function sin x/x is integrable near 0 since sin x/x →
1 as x → 0, but f (x) is not integrable. However the imaginary part of f is odd so
that Z Z
−δ 1
eix
lim + dx
δ→0 −1 δ x
exists.

Definition. If f is continuous on γ \ {a}, where γ is a smooth curve (γ ′ is continuous),

then the Cauchy Principal Value of the integral of f along γ is defined by
Z Z
PV f (z)dz = lim f (z)dz,
γ δ→0 γ∩{|z−a|≥δ}

provided the limit exists.

Note that we have deleted points in a ball centered at a. If the limit exists for all balls
containing a, then the usual integral of f exists. For example
Z 1
cos x
PV dx = 0,
−1 x

because the integrand is odd, but the integral itself does not exist.

Proposition 1.2. Suppose f is meromorphic in {Imz ≥ 0} such that

K
|f (z)| ≤
|z|

when Imz ≥ 0 and |z| > R. If λ > 0 then

Z ∞ X X 1
PV f (x)eiλx dx = 2πi Resa eiλz f (z) + 2πi Resa eiλz f (z). (1.5)
−∞ 2
Ima>0 Ima=0

Part of the conclusion of Proposition 1.2 is that the integral exists.

Proof. Note that f has at most finitely many poles in {Imz ≥ 0} because |f (z)| ≤ K/|z|,
so that both sums in the statement of Proposition 1.2 are finite. Construct a contour similar
to the rectangle in Figure VII.2, but avoiding the poles on R using small semicircles Cj of
radius δ > 0 centered at each pole aj ∈ R. See Figure VII.3.
 §3: Fourier, Mellin and Inverse Laplace Transforms. 127
γ3
B + i(A + B)

γ4 γ2

Cj

−A aj γ1 B

Figure VII.3 Contour avoiding poles on R.

The integral of f (z)eiλz along the top and sides of the contour tend to 0 as A, B → ∞
as in the previous example.
The semi-circle Cj centered at aj can be parameterized by z = aj + δeiθ , π > θ > 0
so that if
bj
f (z)eiλz = + gj (z)
z − aj
where g is analytic in an neighborhood of aj then
Z Z 0 Z
iλz bj iθ
f (z)e dz = iθ
δie dθ + g(z)dz.
Cj π δe Cj

Because gj is continuous at aj and the length of Cj is δ, we have

Z
lim f (z)eiλz dz = −iπbj .
δ→0 Cj

By the Residue Theorem

Z X X
PV f (z)eiλz dz − iπ bj = 2πi Resa f (z)eiλz
R j Ima>0

and (1.5) holds. 

One way to remember the conclusion of Proposition 1.2 is to think of the real line as
cutting the pole at each aj in half. The integral contributes half of the residue of f at aj .
In our example Z ∞
eix
PV dx = πi, (1.6)
−∞ x
so that by taking imaginary parts
Z ∞
sin x
dx = π.
−∞ x
128 VII. Calculus of Residues
R
Note that sin x/x → 1 as x → 0 so that sin x/x exists as an ordinary Riemann integral.
For this reason we can drop the PV in front of the integral.

Mellin Transforms areZa useful tool in Physics.

∞
xα
(g) Mellin Transforms dx,
0 x2 + 1

where 0 < α < 1. Here we define z α = eα log z in C \ [0, +∞) where 0 < arg z < 2π and set
f (z) = 1/(z 2 + 1). Part of the difficulty here is constructing a closed contour that will give
the desired integral. Consider the “keyhole” contour γ consisting of a portion of a large
circle CR of radius R and a portion of a small circle Cδ of radius δ, both circles centered
at 0, along with two line segments between Cδ and CR at heights ±ε, oriented as indicated
in Figure VII.4.
CR

R
Cδ
2ε

Figure VII.4 Keyhole contour.

By the Residue Theorem, for R large and δ small,

Z
z α f (z)dz = 2πi(Resi z α f (z) + Res−i z α f (z))
γ

 α log i 
e eα log −i iπα 3πα
= 2πi + = π(e 2 − ei 2 ). (1.7)
2i −2i

We will first let ε → 0, then R → ∞ and δ → 0. Even though the integrals along the
horizontal lines are in opposite directions, they do not cancel as ε → 0. For ε > 0

lim (x + iε)α f (x + iε) = eα log |x| f (x),

ε→0

and
lim (x − iε)α f (x − iε) = eα(log |x|+2πi)f (x),
ε→0
 §3: Fourier, Mellin and Inverse Laplace Transforms. 129

because of our definition of log z. Thus the integral over the horizontal line segments tends
to Z R
(1 − e2πiα )xα f (x)dx. (1.8)
δ

For R large Z  Z
  2π
 Rα
 z f (z)dz ≤
α
Rdθ → 0, (1.9)
CR 0 R2 − 1
as R → ∞. Similarly Z  Z
  2π
 δα
 z f (z)dz ≤
α
δdθ → 0, (1.10)
Cδ 0 1 − δ2
as δ → 0. By (1.7), (1.8), (1.9), and (1.10)
Z ∞ iπα 3πα
xα e 2 − ei 2 π
dx = π = .
0 x2 + 1 1 − e2πiα 2 cos απ/2

This line of reasoning works for meromorphic f satisfying |f (z)| ≤ C|z|−2 for large |z|
and with at worst a simple pole at 0. The function z α can be replaced by other functions
which are not continuous across R, such as log z. In this case real parts of the integrals
along [0, ∞) will cancel, but the imaginary parts will not. Mellin transforms are used in
applications to signal processing, image filtering, stress analysis and other areas.

(h) Inverse Laplace Transforms

The Laplace Transform of a function F defined on (0, +∞) is given by

Z ∞
L(f )(z) = f (t)e−zt dt.
0

The Laplace Transform can be used to convert a differential equation for a function f into
an algebra problem whose solution is the Laplace of f . It roughly converts differentiation
into multiplication. Indeed, an integration by parts for reasonably behaved functions f
shows that L(f ′ ) = zL(f ) − f (0), and a similar result holds for higher order derivatives.
The algebra problem is generally easy to solve, but to find the solution to the original
differential equation, we need find a function whose Laplace transform is the solution to the
algebra problem. In other words, given a function F , we’d like to find f so that L(f ) = F.
In this example, we will find an integral equation for the inverse Laplace Transform.
130 VII. Calculus of Residues

Theorem 1.3. Suppose F is analytic in {Rez > a} and satisfies

C
|F (z)| ≤
(1 + |z|)p

when Rez > a, for some constant C and some p > 1. For b > a define
Z ∞
1
f (t) = F (b + iy)e(b+iy)t dy.
2π −∞

Then the definition of f does not depend on b, |f (t)| ≤ C1 ebt where C1 is a constant
depending only on C and p, and

L(f )(ζ) = F (ζ),

for Reζ > a.

Proof. For t > 0, z = b + iy, and b > a,

Cebt
|F (z)ezt | ≤
(1 + |y|)p

and so the integral defining f converges and |f (z)| ≤ C1 ebt for Rez = b > a, where C1
depends only on p and C. Thus the integral defining L(f ) converges absolutely in Rez > a
and by Cauchy’s Theorem, it is independent of the choice of b > a. By Fubini’s Theorem,
we can write
Z ∞ Z ∞ Z
1 (b+iy)t −ζt 1 F (z)
L(f )(ζ) = F (b + iy) e e dtdy = − dz,
2π −∞ 0 2πi Rez=b z−ζ

where the integral along {Rez = b} is in the upward direction. By the assumptions on the
growth of |F (z)|, if CR is the semi-circle in the right half-plane with radius R, centered at
b, then Z  
 F (z)  C
 
 ζ − z |dz| ≤ Rp πR → 0,
CR

as R → ∞. Applying the Residue Theorem to the integral over the closed contour con-
sisting of CR together with the portion of {Rez = b} between the endpoints of CR , and
letting R → ∞, we obtain
L(f )(ζ) = F (ζ),
 §3: Series via Residues. 131

for Reζ > b, and since b > a was arbitrary, this holds for Rez > a. 

∞
X 1
(i) .
n=0
n2 +1

§4. Series via Residues.

We can turn this game around. Instead of using sums to compute integrals, we can
1
use integrals to compute sums. Set f (z) = z 2 +1 and consider the meromorphic function
f (z)π cot πz. Then π cot πz has a simple pole at z = 0 with residue 1, and since cot π(z −
n) = cot πz, for any integer n, it has a simple pole with residue 1 at each integer. Since the
poles are simple, f (z)π cot πz has a simple pole with residue f (n) at z = n. We consider
the contour integral of f (z)π cot πz around the square SN with vertices (N + 12 )(±1 ± i).
The function π cot πz is uniformly bounded on SN , independent of N . To see this write

e2πiz + 1
cot πz = i .
e2πiz − 1

The linear fractional transformation (ζ +1)/(ζ −1) is bounded in the region |ζ −1| > δ and
|e2πiz − 1| > 1 − e−2πN on SN , as can be seen by considering the horizontal and vertical
segments separately. Since |f (z)| ≤ C|z|−2 , we have
Z
f (z)π cot πzdz → 0.
Sn

By the Residue Theorem

∞
X
0 = Resi f (z)π cot πz + Res−i f (z)π cot πz + f (n),
−∞

and hence  
∞
X 1 π eπ + e−π
2+1
= π − e−π
−1.
n=0
n 2 e

This technique can be used to compute

∞
X
f (n),
n=−∞
132 VII. Calculus of Residues

provided f is meromorphic with |f (z)| ≤ C|z|−2 for |z| large. If some of the poles of f occur
at integers, then the residue calculation at those poles is slightly more complicated because
the pole of f (z)π cot πz will not have order 1. If only the weaker estimate |f (z)| ≤ C|z|−1
holds, then f has a removable singularity at ∞ and so g(z) = f (z) + f (−z) satisfies
|g(z)| ≤ C|z|−2 for large |z|. Applying the techique to g, we can find the symmetric limit

N
X
lim f (n).
N→∞
n=−N

§2. Exercises.

1. Find all non-zero residues of the following functions:

z−1
(a)
(z + 1)2 (z − 2)
z 2 − 2z
(b)
(z + 1)2 (z 2 + 4)
(c) ez csc2 z
2
(d) ze−1/z
cot πz
(e)
z6

2. Let C be the circle of radius 3 centered at 0, oriented in the positive sense. Find
Z
eλz
dz.
C (z + 4)(z − 1)2 (z 2 + 4z + 5)

3. Suppose that f and g are analytic in a neighborhood of the closed unit disc. Let {an }
denote the zeros of f , including multiplicity, and suppose that |an | < 1 for all n. Prove:
Z 2π X
eiθ g(eiθ )f ′ (eiθ ) dθ
= g(an ).
0 f (eiθ ) 2π n
 §5: Exercises 133

In all of the problems below, be sure you have defined your functions carefully. Prove
all claims about integrals. Draw a picture of any contour you use.

4. Find the Fourier transform of x3 /(x2 + 1)2 . (Note that the integral is not absolutely
convergent, so part of the problem is to prove that the integral converges, i.e. let the limits
of integration tend to ±∞ independently.)

5. For 0 < α < 1, find Z ∞

xα
dx.
0 x(x + 1)

6. Find the inverse Laplace transform of

3z 2 + 12z + 8
F (z) =
(z + 2)2 (z + 4)(z − 1)

using the residue theorem.

7. Verify:
∞
X (−1)n π3
= .
n=0
(2n + 1)3 32

8. Compute
X∞
1
ζ(6) =
n=1
n6

by calculus of residues. You can use your answer to problem 1(e). (Remark: This process
can be used to compute ζ(2n). The value of ζ at an odd integer is another story.)

9. By means of the calculus of residues, evaluate

Z ∞√
x log x
dx.
0 (1 + x2 )

10. Find: Z ∞
x3 + 8
dx,
0 x5 + 1
134 VII. Calculus of Residues

using a contour integral of (log z)(z 3 + 8)/(z 5 + 1).

11. Find Z 1
1
p dx.
0 x(1 − x)
Put a “dog bone” around the interval [0, 1] and add a large circle. Carefully define the
integrand so that it is analytic at ∞, then the integral over the large circle can be found
from the series expansion at ∞. Redo the problem by first making the substitution x = 1/w.
 VIII

Normal Families

Our next major goal is to prove the Riemann mapping theorem, that every simply
connected plane domain (except the plane itself) can be mapped, one-to-one and analyt-
ically, onto the unit disk. The original approach to this problem, due to Riemann, was
to minimize a certain integral (with side conditions) over a collection of functions. This
method had the advantage of a physical interpretation, but it assumed that there was a
minimizer. Weierstrass then showed that there are similar minimization problems with no
minimizer. Existence proofs of this and other related problems dominated mathematical
thinking in analysis for the next 50 or so years. Around 1900 Hilbert finally patched up the
difficulties. See Courant, Dirichlet Principle, and J. Grey [reference XX] for a discussion
of some of the history. It is possible to base a proof on constructive methods such as the
Geodesic Algorithm, given in Chapter XI, but here we will use a different approach that
is rather elegant and which is a useful technique for solving other problems. It is based on
a notion of convergence of analytic functions, called normal convergence.

§1. Normality and Equicontinuity .

Definition 1.1. A collection, or family, F of continuous functions on a region Ω ⊂ C is

said to be normal on Ω provided every sequence {fn } ⊂ F contains a subsequence which
converges uniformly on compact subsets of Ω.

A normal family F is not necessarily countable, a limit function is continuous but not
necessarily in F and if F is a sequence which is normal, it does not necessarily converge.
For example, F = {z n } is normal on D but does not converge uniformly on D. The
only limit function is the zero function which is not in F . A simple example of a normal
sequence which does not converge is F = {gn } where g2n ≡ 1 and g2n+1 ≡ 0.
The first lemma says that normality is local.

135
136 VIII. Normal Families

Lemma 1.2. A family of continuous function F is normal on a region Ω if and only if F

is normal on each disk ∆ whose closure is contained in Ω.

Proof. If F is normal on Ω it is normal on each disk ∆ ⊂ Ω by definition. Suppose

F is normal on each disk ∆ with ∆ ⊂ Ω. Write Ω = ∪∞
j=1 ∆j where ∆j ⊂ Ω. Suppose
(1) (1)
{fn } ⊂ F . Then there is a subsequence {fn } ⊂ {fn } such that {fn } converges uniformly
(2) (1)
on ∆1 . Likewise there is a sequence {fn } ⊂ {fn } which converges uniformly on ∆2 , and
(k) (k−1)
indeed there is a sequence {fn } ⊂ {fn } which converges uniformly on ∆k . Then the
“diagonal” sequence
(k)
{fk }
(k)
converges uniformly on each ∆j , since it is a subsequence of {fn } for each k. Since each
compact set K ⊂ Ω can be covered by finitely many ∆j , the diagonal sequence converges
uniformly on K. 

The proof of the above lemma shows that if for each compact subset, there is a
subsequence which converges uniformly, then there is a single subsequence which converges
uniformly on all compact subsets.
We can define a metric on the space C(Ω) of continuous functions on a region Ω as
follows. Write
Ω = ∪∞
j=1 ∆j ,

where each ∆j is a disk with ∆j ⊂ Ω. If f, g ∈ C(Ω), set

|f (z) − g(z)|
ρj (f, g) = sup ,
z∈∆j 1 + |f (z) − g(z)|

and
∞
X
ρ(f, g) = 2−j ρj (f, g).
j=1

Then ρ is a metric:
If ρ(f, g) = 0, then f = g on each ∆j and hence f = g on Ω.
ρ(f, g) = ρ(g, f ) for f, g ∈ C(Ω),
ρ(f, g) ≤ ρ(f, h) + ρ(h, g) for f, g, h ∈ C(Ω).
 §1: Normality and Equicontinuity 137

The triangle inequality follows from the observation that if a and b are non-negative num-
bers then
a+b a b
≤ + .
1+a+b 1+a 1+b
Note that ρj ≤ 1 and ρ ≤ 1.

Proposition 1.3. A sequence {fn } ⊂ C(Ω) converges uniformly on compact subsets of

Ω to f ∈ C(Ω) if and only if
lim ρ(fn , f ) = 0.
n

In other words, the space C(Ω) with the topology of uniform convergence on compact
subsets is a metric space. Compactness and sequential compactness are the same for this
topology. A family is normal if and only if its closure is compact in this topology.

Proof. If ρ(fn , f ) → 0, then ρj (fn , f ) → 0 for each j. This implies fn converges uniformly
to f on ∆j for each j. Since each compact set K ⊂ Ω can be convered by finitely many
∆j , the sequence fn converges uniformly on K. Conversely if fn converges to f uniformly
on each ∆j , then given ε > 0, choose nj so that for z ∈ ∆j
ε
|fn (z) − f (z)| < , whenever n ≥ nj ,
2
for j = 1, 2, 3, . . .. Then choose m so that
∞
X ε
2−j < .
2
j=m

Then for n ≥ max{n1 , . . . nm−1 }

m−1
X ∞
−j ε X −j
ρ(fn , f ) ≤ 2 + 2 < ε. 
j=1
2 j=m

Normality is related to a strong concept of continuity, called uniform equicontinuity.

Definition 1.4. A family of functions F is called uniformly equicontinuous on a set

E if for all ε > 0 there exist a δ > 0 so that if |z − w| < δ, then |f (z) − f (w)| < ε for all
f ∈ F.

For example if FM is the family of analytic functions on D such that |f ′ | ≤ M < ∞ for
all z ∈ D, then writing f as the integral of its derivative, we obtain |f (z)−f (w)| ≤ M |z−w|.
Thus |f (z) − f (w)| < ε whenever |z − w| < δ = ε/M .
138 VIII. Normal Families

Continuity, uniform continuity and uniform equicontinuity are all related to the state-
ment:

|z − w| < δ implies |f (z) − f (w)| < ε.

For continuity, δ is allowed to depend on ε, the function f and the point w. For uni-
form continuity, δ depends on ε and the function f , but works for all w. For uniform
equicontinuity, δ depends on ε but works for all w and for all f ∈ F .
About half of the books on my shelf use the term equicontinuous to denote this concept
of uniformly equicontinuous. The other half uses a weaker definition of equicontinuity for
a family F on a set E: Given ε > 0 and w ∈ E, there is a δ > 0 so that if |z − w| < δ
then |f (z) − f (w)| < ε. Here δ is allowed to depend on w. A family is equicontinuous on a
region Ω in this latter sense if and only if it is uniformly equicontinuous on each compact
subset of Ω. When the family has only one function, this is a familiar result, moreover the
proof is the same. To emphasize the uniform convergence on compact sets we will use the
phrase uniformly equicontinuous on compact subsets in the statement of the next theorem.
I do not know of a good application of the weaker definition of equicontinuity where the
stronger does not hold.

Theorem 1.5 (Arzela-Ascoli). A family F of continuous functions is normal on a region

Ω ⊂ C if and only if
(i) F is uniformly equicontinuous on every compact subset of Ω, and
(ii) there is a z0 ∈ Ω so that the collection {f (z0 ) : f ∈ F } is a bounded subset of C.

Proof. Suppose F is normal. If F is not uniformly equicontinuous on a compact set

K ⊂ Ω, then there is an ε > 0 and zn , wn ∈ K and fn ∈ F such that |zn − wn | < 1/n and

|fn (zn ) − fn (wn )| ≥ ε.

By normality, we may suppose fn → f uniformly on K. Thus

|fn (zn ) − fn (wn )| ≤ |fn (zn ) − f (zn )| + |f (zn ) − f (wn )| + |f (wn ) − fn (wn )|. (1.1)
 §1: Normality and Equicontinuity 139

For sufficiently large n, the first and third terms on the right side of (1.1) are less than
ε/3 by uniform convergence on K, and the second term is less than ε/3 because the limit
function f is uniformly continuous on K. This contradiction shows that F is uniformly
equicontinuous on K. If z0 ∈ Ω then S = {f (z0 ) : f ∈ F } is a bounded set by the
normality of F , arguing by contradiction.
Conversely suppose (i) and (ii) hold. We first show that (ii) holds at each z ∈ Ω. By
uniform equicontinuity, each w ∈ Ω is contained in a disk ∆w ⊂ Ω such that

|f (z) − f (w)| < 1 for all z ∈ ∆w and all f ∈ F . (1.2)

If z1 ∈ Ω then we can find an arc γ connecting z0 to z1 . Cover γ by finitely many ∆w .

Since γ is connected, we can order these disks so that ∆wj intersects

j−1
∪k=0 ∆w k

for j = 1, . . . , n and where w0 = z0 and wn = z1 . By (1.2) and (ii), {f (z1 )} is a bounded

set.
Now suppose {fn } ⊂ F . Let D = {zk } be a countable dense subset of Ω. Since
(1) (1)
{f (z1 ) : f ∈ F } is a bounded set, we can find a subsequence {fn } such that {fn (z1 )}
(2) (1) (2)
converges. Likewise we can find a sequence {fn } ⊂ {fn } such that {fn (z2 )} converges
(k) (k−1) (k)
and indeed there is a sequence {fn } ⊂ {fn } such that {fn (zk )} converges. Then as
in the proof of Lemma 1.2, the sequence

{fn(n) }

converges on {zk }, for each k. We may relabel this sequence as {fn }.

By (i), if K is compact and ε > 0 then we can find δ > 0 so that if f ∈ F and if
z, w ∈ K with |z − w| < δ then

|f (z) − f (w)| < ε/3. (1.3)

Cover K by finitely many disks ∆1 , ∆2 , . . . , ∆M of radius < δ/2. For each j, choose
zk(j) ∈ D ∩ ∆j . Then find nj so that

|fm (zk(j) ) − fp (zk(j) )| < ε/3, (1.4)

140 VIII. Normal Families

for m, p ≥ nj . Choosing N = max{n1 , n2 , . . . , nM }, we have that for z ∈ ∆j , j = 1, . . . , M

and m, p > N

|fm (z) − fp (z)| ≤ |fm (z) − fm (zk(j) )| + |fm (zk(j) ) − fp (zk(j) )| + |fp (zk(j) ) − fp (z)|. (1.5)

The first and third term on the right side of (1.5) are at most ε/3 by (1.3) and the middle
term is at most ε/3 by (1.4). Thus for m, p > N ,

|fm (z) − fp (z)| < ε

on K. Since ε > 0 is arbitrary, this proves that {fn } is a Cauchy sequence in the uniform
topology on K. Thus {fn } converges uniformly on compact subsets of Ω and hence F is
normal. 

Definition 1.6. A family F of continuous functions is said to be locally bounded on

Ω if there exist disks ∆j ⊂ ∆j ⊂ Ω = ∪j ∆j and Mj < ∞, j = 1, 2, . . ., such that

|f (z)| ≤ Mj for all z ∈ ∆j and all f ∈ F .

An equivalent definition is that F is locally bounded on Ω if and only if for each

compact K ⊂ Ω, the set {f (z) : f ∈ F , z ∈ K} is bounded.

Theorem 1.7. A family F of analytic functions on a region Ω is normal if and only if F

is locally bounded on Ω.

Proof. Suppose F is normal. In the course of the proof of the Arzela-Ascoli Theorem,
we showed that (ii) holds for each z ∈ Ω by covering a curve with a finite number of disks
∆wj . Similarly, if K is any closed disk contained in Ω, we can cover K with finitely many
such disks. Thus the same argument shows that {f (z) : f ∈ F , z ∈ K} is bounded, and so
F is locally bounded.
Conversely, if F is locally bounded on Ω, and if ∆ ⊂ ∆ ⊂ Ω, is a disk then by Cauchy’s
Theorem Z
1 f (ζ)
f (z) = dζ.
2πi ∂∆ ζ −z
 §1: Normality and Equicontinuity 141

If B ⊂ ∆ is a smaller disk, then for z, w ∈ B

Z 
 z − w dζ 
|f (z) − f (w)| =  f (ζ) ≤ C|z − w|,
∂∆ (ζ − z)(ζ − w) 2πi 
where C = M · length(∂∆)/(2πdist(∂∆, B)2). So if |z − w| < ε/C, then |f (z) − f (w)| < ε
for all f ∈ F . We can cover any compact subset of Ω by finitely many of these smaller
disks, and so it follows that F is uniformly equicontinuous on compact subsets of Ω and
so by the Arzela-Ascoli Theorem, F is normal. 

In the discussion above, we used the Euclidean distance |α−β| to measure the distance
between α = f (z) and β = g(z). We could similarly consider families of continuous
functions with values in a complete metric space. For example we could consider continuous
functions with values in the Riemann sphere using the chordal distance between any two
points on the sphere. In this case we must allow the north pole (0, 0, 1) as a possible value
of a function. Equivalently we can consider functions with values in the extended plane
C∗ = C ∪ {∞} with the metric

2|α−β|
 √1+|α|2 √1+|β|2
 α, β ∈ C
χ(α, β) =

√ 2
if β = ∞.
1+|α|2

Then to say that a function f is “continuous” at z0 with f (z0 ) = ∞ means that for all
R < ∞ there is a δ > 0 so that |f (z)| > R for all z with |z − z0 | < δ.
The Euclidean metric is usually used for families of analytic functions, because by
Weierstrass’s Theorem if {fn } is a sequence of analytic functions on a region Ω, converging
to f uniformly on compact subsets of Ω then the limit function f is analytic. The chordal
metric is used for families of meromorphic functions. We need the following result due to
Hurwitz.

Theorem 1.8 (Hurwitz). Suppose gn are analytic and gn (z) 6= 0 for all z in a region Ω.
If gn converges uniformly to g on compact subsets of Ω, then either g is identically zero in
Ω or g is non-zero in Ω.

Proof. The limit function g is analytic on Ω by Weierstrass’s Theorem. In particular, if g

is not identically zero, then the zeros of g are isolated. If γ ∼ 0 is a simple curve on which
142 VIII. Normal Families

g 6= 0, then by Weierstrass’s Theorem, gn′ converges to g ′ and hence gn′ /gn converges to
g ′ /g uniformly on γ. By the Argument Principle, for n sufficiently large, the number of
zeros of g enclosed by γ is the same as the number of zeros of gn enclosed by γ, and since
gn is never zero, the theorem follows. 

Lemma 1.9. If {fn } is a sequence of meromorphic functions which converges uniformly

on compact subsets of a region Ω in the chordal metric then the limit function is either
meromorphic on Ω or identically equal to ∞. If {fn } is a sequence of analytic functions
which converges uniformly on compact subsets of Ω in the chordal metric then the limit
function is either analytic or identically equal to ∞. Moreover, if the limit function is
analytic then the convergence is uniform in the Euclidean metric on compact subsets of Ω.

Proof. Note that

χ(α, β) = χ( 1 , 1 ). (1.6)
α β

If {fn } is a sequence of meromorphic functions which converges uniformly on compact

subsets of Ω in the χ metric then the limit function f is continuous as a map into the
extended plane C∗ . If f (z0 ) = ∞, then 1/fn is bounded in a neighborhood of z0 , for
n sufficiently large, and hence analytic. By Weierstrass’s Theorem, 1/f is analytic in a
neighborhood of z0 and hence f is meromorphic in a neighborhood of z0 or 1/f is identically
0 in a neighborhood of z0 . Since Ω is connected, either 1/f ≡ 0 or f is meromorphic in Ω.
If the sequence {fn } consists of analytic functions, and if the limit function f satisfies
f (z0 ) = ∞ then applying Hurwitz’s Theorem to 1/fn in a neighborhood of z0 , we conclude
1/f ≡ 0 in this neighborhood. Since Ω is connected 1/f ≡ 0 in Ω. Thus if f is not
identically equal to ∞, then f is analytic. Since f is bounded on compact subsets of Ω,
the convergence is also uniform in the Euclidean metric. 

For example, by (1.6)

 
1
z − 1/n

is a normal family in C in the χ metric with limit function 1/z.

 §1: Normality and Equicontinuity 143

One of the exercises is to prove that a family F of analytic functions is normal on a

region Ω if and only if the family is bounded at one point of Ω and the family of derivatives
is normal. To make a similar statement about meromorphic functions, we need to use the
spherical derivative.

Theorem 1.10 (Marty). A family F of meromorphic functions on a region Ω ⊂ C is

normal in the chordal metric if and only if the expressions

2|f ′ (z)|
f # (z) =
1 + |f (z)|2

are locally bounded.

The spherical distance d(p, q) between p, q ∈ C∗ is the length of the shortest curve on
the sphere containing p and q. The quantity f # is the limit
χ(f (z), f (w)) d(f (z), f (w))
f # (z) = lim = lim
w→z |z − w| w→z |z − w|

and is sometimes called the spherical derivative. Note that (1/f )# = f # so that f # (z) is
finite for each z ∈ Ω.

Proof. Suppose F is normal in the chordal metric on Ω and ∆ is a closed disk contained
in Ω. If {f # (z) : f ∈ F } is not bounded on ∆, then we can find a sequence fn ∈ F and
zn ∈ ∆ such that fn# (zn ) → ∞. Taking a subsequence, we may suppose that zn → z∞ ∈ ∆.
Since F is normal, we can find a subsequence of {fn } which converges uniformly (in the
chordal metric) on compact subsets of Ω to a function f which is either meromorphic or
identically ∞, by Lemma 1.9. The limit function is uniformly continuous in the chordal
metric, so that replacing fn with 1/fn if necessary, we may suppose that f is bounded
in a neighborhood of z∞ , and hence by Weierstrass’s theorem fn# converges to f # in the
Euclidean metric on a closed disk centered at z∞ . Thus fn# is bounded in a neighborhood
of z∞ , contradicting our choice of fn and zn .
Conversely suppose the spherical derivatives are bounded by M on a disk ∆ ⊂ ∆ ⊂ Ω.
Then Z
χ(f (z), f (w)) ≤ d(f (z), f (w)) ≤ inf f # (ζ)|dζ| ≤ M |z − w|,
γ γ
144 VIII. Normal Families

where the infimum is taken over all curves γ ⊂ Ω connecting z to w. This can be seen
by considering Riemann sums for the integral using the remark just after the statement of
Marty’s Theorem. Covering a compact subset of Ω with finitely many such disks proves
that F is uniformly equicontinuous on compact subsets of Ω. Condition (ii) in the Arzela-
Ascoli Theorem always holds for the chordal metric because the chordal distance is always
bounded by 2. Thus by the Arzela-Ascoli Theorem, F is normal. 

§2. Riemann Mapping Theorem.

Theorem 2.1 (Riemann Mapping Theorem). Suppose Ω ⊂ C is simply connected

and Ω 6= C. Then there exist a one-to-one analytic map f of Ω onto D = {z : |z| < 1}. If
z0 ∈ Ω then there is a unique such map with f (z0 ) = 0 and f ′ (z0 ) > 0.

Proof. Fix z0 ∈ Ω. The idea of the proof is to show that there is conformal map of Ω into
D, vanishing at z0 , with largest possible derivative at z0 . This forces the image region to
be as large as possible and hence all of D.
First we prove that there is a one-to-one map into D. By assumption there is a point
z1 ∈ C \ Ω. Since Ω is simply connected we can define log(z − z1 ) so as to be analytic on
Ω. Set
√ 1
f1 (z) = z − z1 = e 2 log(z−z1 ) .

If f1 (z) = f1 (w) then by squaring z −z1 = w−z1 and hence z = w, so that f1 is one-to-one.
Note that f1 (z) covers a neighborhood of f1 (z0 ) and hence f1 (z) omits a neighborhood of
−f1 (z0 ), since f1 (z) and −f1 (z) have the same square. Thus

C
f2 (z) =
f1 (z) + f1 (z0 )

is analytic and one-to-one on Ω and bounded by 1 for C sufficiently small.

Set

F = {f : f is one-to-one, analytic and |f (z)| < 1 on Ω with f (z0 ) = 0}.

 §3: Zalcman, Montel, and Picard 145

The function
f2 (z) − f2 (z0 )
g(z) =
1 − f2 (z0 )f2 (z)
is the composition of f2 with a linear fractional transformation, and hence one-to-one with
g(z0 ) = 0. Thus F is non-empty. By Theorem 1.7, F is normal. Let {fn } ⊂ F such that

lim |fn′ (z0 )| = M = sup{|f ′ (z0 )| : f ∈ F }. (2.1)

n→∞

Since F is normal we may suppose that fn converges uniformly on compact subsets. By

Weierstrass’s Theorem, the limit function f is analytic. Moreover {fn′ } is also normal
and |f ′ (z0 )| = M . In particular f is not constant, and M < ∞. By Hurwitz’s Theorem
(Exercise 3) f is one-to-one.
Next we show that f must map Ω onto D. Suppose not. Then there is a point ζ0 ∈ D
such that f 6= ζ0 on Ω. But then

f (z) − ζ0
g1 (z) = = T1 ◦ f (z)
1 − ζ0 f (z)

is a non-vanishing function on the simply connected domain Ω and hence it has an analytic
p
square root, by Corollary VI.2.1. Set g2 (z) = g1 (z) and

g2 (z) − g2 (z0 )
g(z) = = T2 ◦ g2 (z).
1 − g2 (z0 )g2 (z)

Then g(z) ∈ F . Define

ϕ = T1−1 ◦ z 2 ◦ T2−1 .

Then ϕ is an analytic map of D onto D with ϕ(0) = 0, and

f (z) = ϕ ◦ g(z).

By Schwarz’s lemma, |ϕ′ (0)| < 1, so that |f ′ (0)| = |ϕ′ (0)||g ′ (0)| < |g ′ (0)|, contradicting
(2.1), and hence f maps D onto D.
Replacing the map f by λf where |λ| = 1 we may suppose that f ′ (z0 ) > 0. If f1
is another such map then h = f1 ◦ f −1 is a one-to-one analytic map of D onto D with
h(0) = 0 and h′ (0) = 1. By Schwarz’s lemma, h(z) = z and so f1 = f . 
146 VIII. Normal Families

§3. Zalcman, Montel, and Picard.

In this section we will prove Zalcman’s clever lemma characterizing non-normal fam-
ilies, and use it to prove far-reaching extensions of Liouville’s Theorem and Riemann’s
Theorem on removable singularities.

Theorem 3.1 (Zalcman’s Lemma). A family F of meromorphic functions on a region

Ω is not normal in the chordal metric if and only if there exists a sequence {zn } converging
to z∞ ∈ Ω, a sequence of positive numbers ρn converging to 0, and a sequence {fn } ⊂ F
such that
gn (ζ) = fn (zn + ρn ζ) (3.1)

converges uniformly on compact subsets of C to a nonconstant function g which is mero-

morphic in C with the property that

g # (ζ) ≤ g # (0) = 1, (3.2)

for all ζ ∈ C.

It is remarkable that non-normality can be described in terms of a convergent sequence.

If fn were convergent then the functions gn given by (3.1) would converge to a constant,
since the radii ρn tend to 0. Zalcman’s Lemma says that arbitrarily small disks can be
found where fn is close to a non-trivial meromorphic function, after rescaling.
For example the family F = {z n } is not normal on 2D = {z : |z| < 2}. The lack of
normality comes from the lack of uniform equicontinuity in a neighborhood of any point
on the unit circle. So set zn = 1 and ρn = 1/n. Then for ζ ∈ C

ζ n
f (zn + ρn ζ) = (1 + ) → eζ ,
n

as n → ∞, as can be seen by taking logarithms. It is easy to check that the spherical

derivative of eζ satisfies (3.2), and therefore is not constant, the main point of the Theorem..
 §3: Zalcman, Montel, and Picard 147

Proof. One direction of Zalcman’s Lemma is easy. If F is normal then any sequence
{fn } contains a convergent subsequence, which we can relabel as {fn }, and call the limit
function f . If zn → z∞ ∈ Ω and ρn → 0, then

gn (ζ) = fn (zn + ρn ζ) → f (z∞ ),

for each ζ ∈ C, as n → ∞ since the family {fn } is uniformly equicontinuous. Thus the
limit of gn is constant.
Conversely suppose F is not normal. By Marty’s Theorem, there exists wn → w∞ ∈ Ω
and fn ∈ F such that
fn# (wn ) → ∞.

Without loss of generality we may suppose w∞ = 0 and {|z| ≤ r} ⊂ Ω. Then

Mn = max(r − |z|)fn# (z) = (r − |zn |)fn# (zn ),

|z|≤r

for some |zn | < r, since fn# is continuous. Note that Mn → ∞, since wn → 0. Then

gn (ζ) = fn (zn + ζ/fn# (zn )),

is defined on {|ζ| ≤ Mn }, since |zn + ζ/fn# (zn )| ≤ |zn | + Mn /fn# (zn ) = |zn | + r − |zn | = r.
Fix ζ ∈ C. Then for Mn > |ζ|,

fn# (zn + ζ/fn# (zn ))

gn# (ζ) =
fn# (zn )

Mn r − |zn |
≤
r − |zn + ζ/fn# (zn )| Mn

r − |zn |
≤
r − |zn | − |ζ|/fn# (zn )

1
= → 1,
1 − |ζ|/Mn

as n → ∞. Since the spherical derivatives of {gn } are locally bounded, the family {gn }
contains a convergent subsequence by Marty’s Theorem. Relabeling the subsequence,
148 VIII. Normal Families

(3.1) holds with ρn = 1/fn# (zn ). The limit function g satisfies (3.2) and is meromorphic
by Lemma 1.9. Since g # (0) = 1, g is nonconstant. Since {zn } is contained in a compact
subset of Ω, we can arrange that zn → z∞ ∈ Ω by taking subsequences. 

The first consequence of Zalcman’s Lemma is an improvement of the locally bounded

condition for normality.

Theorem 3.2 (Montel). A family F of meromorphic functions on a region Ω that omits

three fixed values a, b, c is normal in the chordal metric.

Proof. Normality is local so we may assume Ω = D. Composing with a linear fractional

transformation, we may suppose a = 0, b = 1 and c = ∞. Without loss of generality, we
may assume that F is the family of all analytic functions on D which omit the values 0
and 1. Set
−m
Fm = {f analytic on D : f 6= 0 and f 6= e2πik2 , k = 1, . . . , 2m }.

Then
F = F0 ⊃ F1 ⊃ F2 ⊃ . . . .
1
If f ∈ Fm then f is analytic and f 6= 0, so that we can define f 2 so as to be analytic.
1
Moreover f 2 ∈ Fm+1 . If F is not normal then there exists a sequence {fn } ⊂ F with
1
no convergent subsequence. Moreover, {fn2 } is then a sequence in F1 with no convergent
subsequence. By induction, each Fm is not normal. Thus for each m we can construct a
limit function hm as in Zalcman’s lemma. The functions hm are entire by Lemma 1.9 and
nonconstant since h#
m (0) = 1. By Marty’s Theorem {hm } is a normal family. If h is a limit

of a subsequence, uniformly on compact subsets of C, then h is entire by Lemma 1.9 and

nonconstant since h# (0) = 1. By Hurwitz’s Theorem, h omits the 2m roots of 1 for each
m. These points are dense in the unit circle. Since h(C) is connected and open, either
h(C) ⊂ D or h(C) ⊂ C \ D. Thus either h or 1/h is bounded. By Liouville’s Theorem, h
must be constant, which contradicts h# (0) = 1, and hence F is normal. 

We remark that the families Fm in the proof of Montel’s Theorem are not closed.
Indeed constant functions are in the closure. For example 2 + z n omits 0 and 1 in D and
 §3: Zalcman, Montel, and Picard 149

tends to 2 in D as n → ∞. However, Zalcman’s lemma yields limit functions which are

not constant, and therefore by Hurwitz’s Theorem, in the family Fm .
The next consequence of Zalcman’s Lemma can be viewed an improved version of
Riemann’s Theorem on removable singularities.

Theorem 3.3 (Picard’s Great Theorem). If f is meromorphic in Ω = {z : 0 <

|z−z0 | < δ}, and if f omits three (distinct) values in C∗ , then f extends to be meromorphic
in Ω ∪ {z0 }.

The reader can verify that an equivalent formulation of Picard’s great theorem is
that an analytic function omits at most one complex number in every neighborhood of an
essential singularity. The function f (z) = e1/z does omit the value 0 in every neighbhood
of the essential singularity 0, so that Picard’s theorem is the strongest possible statement
in terms of the range of an analytic function. The weaker statement that a non-constant
entire function can omit at most one complex number is usually called Picard’s Little
Theorem, and can be viewed as an extension of Liouville’s Theorem.

Proof. As before we may assume a = 0, b = 1, c = ∞, and z0 = 0. Let εn → 0. If f

omits 0 and 1 in a punctured neighborhood of 0, then by Montel’s Theorem the family
{f (εn z)} is normal on compact subsets of C \ {0}. Relabelling a subsequence, we may
suppose f (εn z) converges uniformly on compact subsets of C \ {0} to a function g analytic
on C \ {0} or to g ≡ ∞. If g is analytic, then |g(z)| ≤ M < ∞ on |z| = 1 and hence
|f | ≤ M + 1 on |z| = εn for n ≥ n0 . But then by the maximum principle, |f | ≤ M + 1 on
εn+1 < |z| < εn for n ≥ n0 . Thus |f | ≤ M + 1 on 0 < |z| < εn0 , so that by Riemann’s
Theorem on removable singularities f extends to be analytic in a neighborhood of 0.
If g ≡ ∞, we can apply a similar argument to 1/f (εn z) to conclude that 1/f extends
to be analytic at 0 and hence f is meromorphic in a neighborhood of 0. 

Picard’s Little Theorem follows from Picard’s Great Theorem by considering f (1/z).
Picard’s Little Theorem also follows from exactly the same proof by letting εn → ∞, using
Liouville’s Theorem instead of Riemann’s Theorem, to conclude that f is constant. Test
your understanding of the proof by writing out the details. Zalcman’s Lemma has been
150 VIII. Normal Families

used in many situations with the heuristic principle that if P is a property of meromorphic
functions then
{f meromorphic on Ω : f has P}

is normal on Ω if and only if no meromorphic function on C has P. For example, P could

be “omits three values”.

We end this chapter with a comment about normal families and some open problems.
Normal families can be used to prove results like (see the exercises):

Koebe’s Theorem. There is a K > 0 so that if f is analytic and one-to-one on D with

f (0) = 0 and f ′ (0) = 1, then f (D) ⊃ {z : |z| < K}.

Landau’s Theorem. There is a constant L > 0 so that if f is analytic on D with f ′ (0) = 1

then f (D) contains a disk of radius L.

Bloch’s Theorem. There is a constant B > 0 so that if f is analytic on D with f ′ (0) = 1

then there is a region Ω ⊂ D such that f is one-to-one and analytic on Ω and f (Ω) is a
disk of radius B.

In each theorem, we already know that there is a constant for each f . Normal families
can be used to prove that the constants can be taken to be independent of the functions f ,
but this method of proof does not give any information about what the constants are. The
largest K is known to be 1/4. The conformal map of D onto C \ [1/4, ∞) is an example
where 1/4 is the best possible. The largest L is called Landau’s constant, and it is known
that    
Γ 13 Γ 56
.5 < L ≤ L0 =   = 0.544..
Γ 16

It is conjectured that Landau’s constant is L0 . The largest B is called Bloch’s constant

and it is known that
   
√ Γ 1 11
3 1 3 Γ 12
= 0.433... < B ≤ B0 = p √   = 0.472...
4 1+ 3 Γ 41

and it is conjectured that B = B0 .

 §4: Exercises 151

§4. Exercises

1. If a normal family has only one normal limit, then it converges. This simple obser-
vation is sometimes a useful bootstrap technique to prove convergence of a sequence:
find enough properties of a subsequential limit function to prove it is unique.
2. Give details for the proof of Picard’s little theorem mentioned in the text.
3. If a sequence of one-to-one analytic functions {fn } on a region Ω converges uniformly
on compact subsets of Ω to f , then either f is constant or f is one-to-one and analytic.
(This is sometimes also called Hurwitz’s Theorem).
4. (a) Prove that a family F of analytic functions on a region Ω is normal if and only if
the family F ′ = {f ′ : f ∈ F } is normal and for some z0 ∈ Ω, the set {f (z0 ) : f ∈ F }
is bounded.
(b) Find an example of a sequence {fn } which is normal on C using the chordal
metric, but {fn′ } is not normal on C using the chordal metric.
5. Prove the following analytic version of Zalcman’s Lemma: Suppose F is a family of
analytic functions on a region Ω whose members vanish at z0 ∈ Ω. Then F is not
normal (in the Euclidean metric) if and only if there is a sequence {zn } converging to
z∞ ∈ Ω, a sequence of positive numbers ρn converging to 0, and a sequence {fn } ⊂ F
such that gn (ζ) = fn (zn + ρn ζ) − fn (zn ) converges uniformly on compact subsets of
C to the function g(ζ) = cζ, with |c| = 1.

6. Let {fn } be a sequence of analytic functions on a region Ω with |fn | ≤ 1 on Ω. Let

K be compact and contained in Ω. Suppose {fn } converges at infinitely many points
in K. Then is it true or false that {fn } necessarily converges at every point of Ω?
7. Let FM be the set of functions analytic on the (open) unit disk D and continuous on
the closed unit disk which satisfy
Z 2π
|f (eiθ )|dθ ≤ M.
0
152 VIII. Normal Families

Show FM is a normal family on D with respect to the Euclidean metric.

8. Let B be the set of functions f which are analytic on the unit disk D and satisfy both
f (0) = 0 and f (D) ∩ [1, 2] = ∅. Prove B is a normal family (as maps from D into the
complex plane with the Euclidean metric) which contains all of its limit functions.
9. Suppose f and g are entire functions with the property that

f n + g n = 1.

If n = 2, prove that there is an entire function h so that f = cos h and g = sin h.

Prove that if n > 2 then f and g are constant.
10. Prove that the set of one-to-one analytic functions on D with the property that f (0) =
0 and f ′ (0) = 1 is normal in the Euclidean metric. Is the conclusion true if we eliminate
the requirement f ′ (0) = 1, but use the chordal metric?
11. Prove Koebe’s Theorem. Hint: see Exercise 10.
12. Prove Landau’s Theorem. Hint: see Exercise 5.
13. (a) Prove that a non-constant entire function covers disks of arbitrarily large radii.
Hint: see Exercise 12.
(b) If g is analytic on a simply connected region Ω such that g omits the two values
 p 
i 2
1 and −1, then prove that we can define h(z) = − π log g(z) + g(z) − 1 so
as to be analytic on Ω. Furthermore show that h omits the integers.
(c) Repeat the idea in part (b) by constructing an analytic function k with k(z) =
 p 
− πi log h(z) + h(z)2 − 1 . Prove that k does not cover arbitrarily large disks.
(d) Deduce Picard’s little theorem from (c).

14. Prove Bloch’s Theorem. Hint: see Exercise 5.

15. Give a proof of the Koebe, Landau, or Bloch Theorem with an explicit constant.
 IX

Series and Products

§1. Mittag-Leffler’s Theorem.

So far we’ve learned a little bit about polynomials, rational functions, exponentials
and logarithms, as well as their real and imaginary parts as harmonic functions. Now we’ll
look at more “complex” functions.

Suppose f is meromorphic in a region Ω with a pole at b ∈ Ω. Then, recalling the

Laurent expansion,

cn cn−1 c1
f (z) = n
+ n−1
+...+ + a0 + a1 (z − b) + a2 (z − b)2 + . . . ,
(z − b) (z − b) (z − b)

for z near b. The sum of the first n terms

cn cn−1 c1
Sb (z) = n
+ n−1
+...+
(z − b) (z − b) (z − b)

is called the singular part of f at b. If f is rational, then by a partial fraction expansion

n
X
f (z) = Sbk (z) + p(z),
k=1

where p is a polynomial and {bk } are the poles of f . If f is meromorphic in a region Ω

with only finitely many poles {bk } and singular parts Sbk , k = 1, . . . , n, then

n
X
f (z) = Sbk (z) + g(z),
k=1

P
where g is analytic in Ω. This follows because f (z) − Sbk (z) is analytic at each bk
and therefore in all of Ω. In this section, we’ll find a similar expansion for meromorphic
functions in Ω with infinitely many poles. As before we say that an infinite sequence
bk ∈ Ω → ∂Ω as k → ∞ if each compact K ⊂ Ω contains only finitely many bk .

153
154 IX. Series and Products

Theorem 1.1 (Mittag-Leffler). Suppose bk ∈ Ω → ∂Ω. Set

nk
X cj,k
Sk (z) =
(z − bk )j
j=1

where each nk is a positive integer and cj,k ∈ C. Then there is a function meromorphic in
Ω with singular parts Sk at bk , k = 1, 2, . . ., and no other singular parts in Ω.

Proof. Let
1
Kn = {z ∈ Ω : dist(z, ∂Ω) ≥ and |z| ≤ n}.
n
Then Kn is a compact subset of Ω such that each component of C \ Kn contains a point
of C \ Ω and Kn ⊂ Kn+1 ⊂ ∪Kn = Ω. By Runge’s Theorem IV.3.7 we can find a rational
function fn with poles in C \ Ω so that

X
| Sk (z) − fn (z)| < 2−n
bk ∈Kn+2 \Kn+1

for all z ∈ Kn . Then

X X
Sk (z) − fn (z)
n≥m bk ∈Kn+2 \Kn+1

converges uniformly on Km to an analytic function on Km by the Weierstrass M -test and

Weierstrass’s Theorem IV.3.9. Thus

X ∞
X X
f (z) = Sk (z) + Sk (z) − fn (z) (1.1)
bk ∈K2 n=1 bk ∈Kn+2 \Kn+1

is the desired function. 

A more constructive proof of Mittag-Leffler’s Theorem is useful in some circumstances.

We will illustrate the technique in some special cases, and leave the more general case to
the reader.
Suppose we’d like to find a function meromorphic in C with singular part ak /(z − bk )
at {bk }, where |bk | → ∞. Then
∞
X ak
z − bk
k=1
 §1: Mittag-Leffler’s Theorem 155

will work, provided the sum converges. When does it converge? If |z| ≤ R < ∞, write
∞
X X X
ak ak ak
= + . (1.3)
z − bk z − bk z − bk
k=1 {k:|bk |≤2R} {k:|bk |>2R}

The first sum has only finitely many terms. For the second sum, |z| ≤ R < |bk |/2, so that
 
 1  2
 
 z − bk ≤ |bk | .

Thus if
∞
X |ak |
< ∞, (1.4)
|bk |
k=1

then the second sum in (1.3) converges uniformly on {|z| ≤ R} to an analytic function, by
Weierstrass again. The right side of (1.3) then is meromorphic in |z| < R with singular
part ak /(z − bk ) at {bk : |bk | < R}. Since R is arbitrary, the sum in (1.3) is meromorphic
in C.

What if |bk | tends to ∞ more slowly? If we examine the proof, we just need the
tail of the series in (1.3) to converge for each R. Mittag-Leffler’s idea was to subtract a
polynomial from each term so that the result converges. As we have seen before
∞  j
1 1 1 X z
= z = .
z − bk −bk (1 − bk ) −bk j=0 bk

So it is natural to subtract a few terms of the expansion to make it smaller:

   nk  j   ∞  j 
 1
 1 X z   1
=
X z 

 z − bk −
−bk bk   −bk
j=0 j=n +1
b k

k

  nk +1
 1  z  1
≤    z ,
bk bk 1−| bk |

provided |z| < |bk |.

For example the following proposition holds.

156 IX. Series and Products

Proposition 1.2. If bk → ∞ and if for some n < ∞

X |ak |
<∞
|bk |n+2
k

then
∞ 
X   n  j 
ak ak X z
f (z) = −
z − bk −bk j=0 bk
k=1
ak
is meromorphic in C with singular part Sbk (z) = z−bk at bk , k = 1, 2, . . ., and no other
poles.

To prove Proposition 1.2, set nk = n for all k. Then if |z| < R, split the sum into two
pieces: a finite sum of the terms with |bk | ≤ 2R and a convergent sum of the terms with
|bk | > 2R, with nk = n for all k.

Example 1.3.
X∞
π2 1
2 = (1.7)
sin πz n=−∞ (z − n)2
The right side of (1.7) converges uniformly on compact subsets of C and the limit is
meromorphic with singular parts Sn (z) = 1/(z − n)2 at z = n and has no other poles. To
see this follow the proof of Proposition 1.2: fix R < ∞ and split the sum into two pieces:
a finite sum of terms with |n| < 2R and the remaining infinite sum. In the second sum
when |z| ≤ R, the nth term is uniformly bounded by 1/(|n| − R)2 ≤ 4/n2 . Thus the second
(infinite) sum converges uniformly and absolutely on |z| ≤ R to an analytic function by
Weierstrass’s Theorem.
The function πz/ sin πz has a removable singularity at z = 0 and is an even function
so that
πz
= 1 + O(z 2 ),
sin πz
near 0. By squaring and dividing by z 2 we conclude that the singular part of π 2 / sin2 πz
at z = 0 is 1/z 2 . Since sin2 π(z − n) = sin2 πz, we conclude that the singular part of the
left side of (1.7) is the same as the singular part of the right side of (1.7) at z = n for each
n. Set
X∞
π2 1
F (z) = 2 − .
sin πz n=−∞ (z − n)2
 §1: Mittag-Leffler’s Theorem 157

Then F is entire and F (z + 1) = F (z).

Write z = x + iy and suppose 0 ≤ x ≤ 1. We claim that

|F (z)| → 0, (1.8)

as |y| → ∞. If so, F is bounded in the strip 0 ≤ x ≤ 1 and since F (z + 1) = F (z), the

function F is bounded in C. By Liouville’s Theorem, F is constant and by (1.8), F ≡ 0,
proving the Example. To see (1.8), first observe that

1 −πy+iπx
| sin πz| = |e − eπy−iπx | → ∞,
2

as |y| → ∞. Thus the left side of (1.7) tends to 0 as |y| → ∞, with 0 ≤ x ≤ 1. Likewise in
the strip 0 ≤ x ≤ 1, the right side of (1.7) is dominated by
 
X 1 
 
 (|n| − 1)2 + y 2 

R∞
which also tends to 0 as |y| → ∞ by comparison with the integral 1
dx/(x2 + y 2 ). This
proves (1.8) and the Example. 

Example 1.4.  
1 X 1 1
π cot πz = + + . (1.9)
z z−n n
n6=0

First observe that

d π2 X 1
π cot πz = − 2 =− .
dz sin πz (z − n)2

But
X 1
π cot πz 6= ,
z−n
because the latter sum does not converge. The difficulty is that for |z| ≤ R, the nth term
behaves like 1/n. However,

1 1 z 1
+ = ∼ 2,
z−n n (z − n)n n
158 IX. Series and Products
P
for large n and n−2 < ∞. To prove Example 1.4, suppose |z| ≤ R and split the sum on
the right side of (1.9) into the sum of the terms with |n| ≤ 2R and the sum of the terms
with |n| > 2R. The first sum is finite and the second sum has terms satisfying
   
 1 1   z  R 2R
   
 z − n + n =  (z − n)n ≤ n · n = n2 .
2
P
Since 2R/n2 < ∞, the right side of (1.9) is meromorphic in |z| ≤ R with poles only at
the integers and with prescribed singular parts 1/(z − n), for each R < ∞. Furthermore
by Weierstrass’s Theorem IV.3.9 and Example 1.3 the right side of (1.9) has derivative

1 X 1 d
− 2
− 2
= π cot πz.
z (z − n) dz
n6=0

Thus the two sides of (1.9) differ by a constant C. Convergence of the right side of
(1.9) is absolute, so that we can add the terms involving n and −n to obtain
∞
1 X 2z
+ , (1.10)
z n=1 z 2 − n2

which is clearly odd. Since π cot πz is also odd, we must have C = 0, proving (1.9), as well
as the equality with (1.10). 

A slightly subtle point is that the convergence of (1.10) is not the same as the con-
vergence of (1.9). Convergence of (1.10) is the same as convergence of symmetric partial
sums in (1.9), from −n to n which is apriori weaker than allowing the upper and lower
limits of the partial sums to tend to ∞ independently.
We will use the function π cot πz in the chapter on Calculus of Residues.

Example 1.5 (Weierstrass P Function).

Suppose w1 , w2 ∈ C with w1 /w2 not real. In other words w1 and w2 are not on
the same line through the origin. There are no non-constant entire function f satisfying
f (z + w1 ) = f (z + w2 ) = f (z) for all z, by Liouville’s Theorem, but there are meromorphic
functions with this property. The Weierstrass P function is defined by

1 X 1 1
P(z) = + − , (1.11)
z2 (z − mw1 − nw2 )2 (mw1 + nw2 )2
(m,n)6=(0,0)
 §1: Mittag-Leffler’s Theorem 159

where the sum is taken over all pairs of integers except (0, 0).
To prove convergence of this sum, we first observe that there is a δ > 0 so that
|mw1 + nw2 | ≥ δ unless m = n = 0, for if |mj w1 + nj w2 | → 0, then

 w1 nj 
 + → 0,
w2 mj

contradicting the assumption that w1 /w2 is not real. Thus {ζm,n = mw1 +nw2 : m, n ∈ Z}
forms a lattice of points in C with no two points closer than δ. If we place a disk of radius
δ/2 centered at each point of the lattice, then the disks are disjoint. The area of the
annulus k ≤ |ζ| ≤ k + 1 is (2k + 1)π so there are at most Ck lattice points in this annulus,
for some constant C depending on δ.
For |z| < R, we split the sum in (1.11) into a finite sum of terms with |ζm,n | ≤ 2R
and the sum of terms with |ζm,n | > 2R. Note that if |z| < R and |ζ| > 2R, then
   
 1 1   2zζ − z 2  R(2|ζ| + R) 10R
   
 (z − ζ)2 − ζ 2 =  ζ 2 (z − ζ)2 ≤ |ζ|2 |ζ/2|2 ≤ |ζ|3 .

We conclude that for K ≥ R

  ∞  
X  1 1  X X  1 1 
 
 (z − ζm,n )2 − ζ 2  =  (z − ζm,n )2 − ζ 2 
|ζm,n |>2K m,n k=2K k<|ζm,n |≤k+1 m,n

∞
X 10K
≤ Ck < ∞.
k3
k=2K

By Weierstrass’s Theorem, the Weierstrass P function is meromorphic in C with singular

part S(z) = 1/(z − mw1 − nw2 )2 at mw1 + nw2 and no other poles.
Next we show that P(z + w1 ) = P(z). By Weierstrass’s Theorem

2 X 2
P ′ (z) = − 3
− .
z (z − ζm,n )3
ζm,n6=0

By the same estimate, this series converges absolutely so that P ′ (z + w1 ) = P ′ (z), and
hence P(z + w1 ) − P(z) is a constant. The series for P is even, so P(z + w1 ) = P(z)
when z = −w1 /2, and thus P(z + w1 ) = P(z) for all z. A similar argument shows that
P(z + w2 ) = P(z).
160 IX. Series and Products

It is possible to replace the proof in the preceeding paragraph with an estimate of

the difference of the two partial sums. However, when z is replaced by z + w1 , the term
involving z will be paired with the wrong value at 0. An estimate of the error made by
replacing it with the correct value must be made, as well as an estimate of the terms in
the difference of the partial sums which do not cancel, since the summation is shifted by
w1 .
The Weierstrass P function is a basic example of an elliptic function. Elliptic functions
are important in the application of complex analysis to number theory. See Exercise 5,
Exercise XI.5, and Chapter XVI.

§2. Weierstrass Products.

Just as infinite sums are used to create meromorphic functions with prescribed poles,
infinite products are used to create analytic functions with prescribed zeros. If {fj } are
Qn
analytic and fj (zj ) = 0, then j=1 fj vanishes at z1 , z2 , . . . , zn . If we want a function to
vanish at infinitely many zj , it is natural to try
n
Y
lim fj .
n→∞
j=1

However this limit may not exist. Moreover, it is possible for the product of infinitely
many non-zero numbers to converge to zero, which might create additional zeros. First we
will treat the convergence of infinite products of complex numbers.

Qn
Proposition 2.1. Suppose pj ∈ C \ {0}. Then j=1 pj converges to a non-zero complex
number P as n → ∞ if and only if
∞
X
log pj
j=1

converges to a complex number S where log pj is defined so that −π < arg pj ≤ π. More-
over, if convergence holds then P = eS and lim pj = 1.

Pn Qn
Proof. Let Sn = j=1 log pj and Pn = j=1 pj . If Sn → S then Pn = eSn → eS ∈ C\{0}.
Conversely if Pn → P ∈ C \ {0}, then by altering p1 if necessary, we may suppose P > 0.
 §2: Weierstrass Products 161

Then log Pn → log P , where the logarithm is defined so that −π < arg Pn ≤ π. Note that

|Pn−1 ||pn − 1| = |Pn − Pn−1 | → 0,

so that pn → 1 and arg pn → 0. But

n
X
log Pn = log pj + 2πikn
j=1

for some integers kn , and since log pn → 0 and log Pn − log Pn−1 → 0, we must have
Pn
kn = kn0 for n0 sufficiently large and n ≥ n0 . Thus limn j=1 log pj exists and the
Proposition is proved. 

Q
One possibility for a definition of absolute convergence of an infinite product pj
Qn Qn n(n+1)
would be the convergence of the partial product j=1 |pj |. However, j=1 eij = ei 2
does not converge even though the product of the absolute values is 1. The notion of
absolute convergence was useful for infinite sums because it implies that the terms in the
sum can be rearranged without affecting convergence.

Q∞
Definition 2.2. If pj are non-zero complex numbers then we say that j=1 pj converges
P
absolutely if | log pj | converges.
Q∞
It is not hard to see that if pj are nonzero complex numbers such that j=1 pj con-
verges absolutely then any rearrangement of the terms in the infinite product will not
Qn
affect lim j=1 pj . The next Lemma is useful for determining absolute convergence.

Q∞
Lemma 2.3. If pj are nonzero complex numbers then j=1 pj converges absolutely if and
only if
∞
X
|pj − 1|
j=1

converges.

Proof. 
log z d 

lim = log z  = 1,
z→1 z − 1 dz 
z=1
162 IX. Series and Products

so that if pj → 1 then for j sufficiently large

 
 log pj  1
 − 1 <
 pj − 1  2

and thus
1 3
|pj − 1| ≤ | log pj | ≤ |pj − 1|,
2 2
and the Lemma follows. 

Q∞ P∞
Lemma 2.3 says that j=1 pj converges absolutely if and only if j=1 (pj −1) converges
j √
absolutely. The analogous statement for convergence is false. If pj = e(−1) / j then
P Q P
log pj converges by the alternating series test and hence pj converges, but j (pj − 1)
j P P
diverges. Similarly if pj = 1 + i (−1)
√
j
then (pj − 1) converges but log pj diverges and
Q
hence pj diverges. The proof of these statements is an exercise for the reader.

As mentioned at the start of this section, we would like to consider infinite products
Q
of analytic functions fj (z), but we would like to allow the fj to have zeros. For that
reason, we make the following definition.

Q∞
Definition 2.4. Suppose {fj } are analytic on a region Ω. We say that j=1 fj (z) con-
verges on Ω if
n
Y
lim fj (z)
n→∞
j=0

converges uniformly on compact subsets of Ω to an analytic function g which is not iden-

tically equal to 0.

Qn
If limn→∞ j=0 fj (z) exists, uniformly on compact subsets of Ω, then the limiting
function g is analytic by Weierstrass’s Theorem. It follows that if g 6≡ 0 then g has at most
finitely many zeros on each compact subset K ⊂ Ω. Choose compact sets Kn−1 ⊂ Kno ⊂
Kn ⊂ ∪Kn = Ω, so that fj 6= 0 on ∂Kn , for all j. Then fj 6= 0 on Kn for all j ≥ jn , for
some jn . Then
∞
Y
Fn = fj
j=jn
 §2: Weierstrass Products 163

converges uniformly on ∂Kn and by the maximum principle the convergence is also uniform
on Kn . Moreover Fn is non-vanishing on Kn by Hurwitz’s Theorem. Thus the function g
has zero set equal to the union of the zero sets of fj .
The next result is the analog of Mittag-Leffler’s Theorem for products.

Theorem 2.6 (Weierstrass). If {bj } ⊂ Ω with bj → ∂Ω, and if nj are positive integers,
then there exists an analytic function f on Ω such that f has a zero of order exactly nj at
bj , for j = 1, 2, . . . and no other zeros in Ω.

In Theorem 2.6 we view ∂Ω as a subset of C∗ so that if ∂Ω is unbounded we allow

(sub)sequences bj with |bj | → ∞.

Proof. We may suppose that ∂Ω is bounded by first composing with a map T (z) =
1/(z − b) for some b ∈ Ω if necessary. We then compose the constructed function with
T −1 .
As in the proof of Mittag-Leffler’s Theorem, let

1
Kn = {z ∈ Ω : dist(z, ∂Ω) ≥ }.
n

Then Kn is a compact subset of C∗ such that each component of C∗ \ Kn contains a point

of C∗ \ Ω and Kn ⊂ Kn+1 . Choose aj ∈ ∂Ω so that

dist(bj , ∂Ω) = |bj − aj |.

We can define log((z −bj )/(z −aj )) so as to be analytic in C∗ \[aj , bj ] where [aj , bj ] denotes
the line segment from aj to bj . In particular if bj ∈
/ Kn+1 then log(z − bj )/(z − aj ) is
analytic on Kn .
By Runge’s Theorem IV.3.7 we can find a rational function fn with poles in C \ Ω so
that    
 X z − bk 
 nk log −fn (z)< 2−n , (2.1)
 z − ak
bk ∈Kn+2 \Kn+1

for all z ∈ Kn . Then

X X  
z − bk
nk log −fn (z)
z − ak
n≥m bk ∈Kn+2 \Kn+1
164 IX. Series and Products

converges uniformly on Km to an analytic function on Km by Weierstrass’s Theorem

IV.3.9. Thus
Y  z − bk nk Y
∞ Y 
z − bk
nk
f (z) = e−fn (z)
z − ak n=1
z − ak
bk ∈K2 bk ∈Kn+2 \Kn+1

is the desired function. Note that if ∞ is an interior point of Ω then f is analytic at ∞. 

Here is a more constructive version in the plane.

Theorem 2.7. Suppose {an } ⊂ C \ {0} and suppose g is a nonnegative integer such that
∞
X 1
g+1
< ∞.
n=1
|a n |

Then
∞  Pg
Y z  1 z j
j ( an )
1− e j=1 (2.2)
n=1
a n

converges and represents an entire function with zeros at {an } and no other zeros.

The number g in Theorem 2.7 is called the genus of the infinite product. In the case
when g = 0 we interpret the sum as equal to 0. The an in Theorem 2.7 do not need to be
distinct, so that the Theorem can be used to construct functions with zeros of order more
than one.

Proof. By integrating the series for 1/(1 − w) we have

∞
X 1
log(1 − w) = − wj ,
j=1
j

where the sum converges and therefore is analytic for |w| < 1. If |z| < R and |a| > 2R,
then    j   g+1
  X
g
1 z  R
log 1 − z + ≤ C   ,
 a j a  a
j=1
 
Pg 1 j
where C is a constant. This follows from the fact that log(1 − w) + j=1 j w /wg+1 is
analytic on the unit disk and therefore bounded on the disk of radius 1/2. (The interested
reader can prove that C = 2 works by estimating the tail of the series.)
 §2: Weierstrass Products 165

If |z| ≤ R then
  j 
X   z  X g
1 z  X 1
log 1 − + ≤ CRg+1 < ∞.
 an j an  |an |g+1
|an |>2R j=1 |an |>2R

Thus by Weierstrass’s Theorem

Y  Pg
z  1 z j
j ( an )
1− e j=1
an
|an |>2R

is analytic and non-zero in {|z| < R}. The finite product with terms |an | ≤ 2R then gives
the zeros in |z| < R. 

Example 2.8.

The function
Y∞  
z z
f (z) = 1− en
n=1
n
P
converges since n−2 < ∞ (g = 1) and vanishes precisely at the positive integers. Thus
we can write
Y∞  
g(z) z z
sin πz = ze 1− en ,
n=−∞
n

where g is entire. Indeed sin πz has a simple zero at each integer and no other zeros, so if
we divide by z times the product, the resulting function is entire and never equal to 0 in C
and hence has an analytic logarithm. To find g, we take the logarithmic derivative f ′ /f :

1 X 1 1
π cot πz = + g ′ (z) + + .
z (z − n) n
n6=0

But then by (1.9), we must have g ′ (z) ≡ 0, and hence g is constant. Since

sin πz
lim = π,
z→0 z

we must have eg(0) = π and we obtain

Y∞  
z z
sin πz = πz 1− en .
n=−∞
n

We now give a few corollaries of the Weierstrass Product Theorem 2.6.

166 IX. Series and Products

Corollary 2.9. If Ω is a region then there is a function f analytic on Ω such that f does
not extend to be analytic in any larger region.

Proof. Take a sequence {an } → ∂Ω such that ∂Ω ⊂ {an }. By the Weierstrass Product
Theorem (Theorem 2.6) we can find f analytic on Ω, with f (an ) = 0 but f not identically
zero. If f extends to be analytic in a neighborhood of b ∈ ∂Ω then the zeros of the extended
function would not be isolated. 

In several complex variables, a similar result is not true. If Br = {(z, w) : |z|2 + |w|2 <
r 2 }, then any function which is analytic on B2 \ B1 extends to be analytic on B2 .

Corollary 2.10. Suppose Ω is a region and an → ∂Ω, with an 6= am when n 6= m, and

suppose {cn } are complex numbers, then there exists f analytic on Ω such that

f (an ) = cn .

Results like this corollary are usually called interpolation theorems. Corollary 2.12
has a version of this result when the zeros are not distinct.

Proof. By the Weierstrass Product Theorem, we can find G analytic on Ω with a simple
zero at each an . Let
G(z)
dn = lim = G′ (an ).
z→an z − an
Since the zero of G at an is simple, dn 6= 0. By Mittag-Leffler’s Theorem we can find F
meromorphic on Ω with singular part

cn /dn
Sn (z) =
z − an

at an and no other poles in C. Then f (z) = F (z)G(z) is analytic on Ω \ {an } and

G(z) cn
lim F (z)G(z) = lim (z − an )F (z) = dn = cn .
z→an z→an z − an dn

Thus the singularity of f at each an is removable and f extends to be entire with f (an ) =
cn . 
 §3: Blaschke Products 167

Corollary 2.11. If f is meromorphic in Ω then there are functions g and h, analytic on

Ω such that
g
f= .
h

Proof. Let {an } be the poles of f , where the list is written such that a pole of order k
occurs k times in this list. By the Weierstrass Product Theorem, there is a function h
analytic on Ω with zeros {an } and no other zeros. Then g = f h is analytic on Ω \ {an }
with removable singularity at each an . Thus g extends to be analytic in Ω and f = g/h.


The idea behind the proof of Corollary 2.11 can be used to prove the analog of Mittag-
Leffler’s Theorem for Taylor polynomials.

Corollary 2.12. Suppose pn is a polynomial of degree dn for n = 1, 2, . . ., and suppose

an → ∂Ω. Then there is a function f analytic on Ω such that for each n = 1, 2, . . .,

f (z) − pn (z − an )

has a zero of order dn + 1 at an . In other words the Taylor polynomial for f at an of degree
dn is the prescribed function pn (z − an ).

Proof. By the Weierstrass Product Theorem, we can find h analytic on Ω with zeros
an of order dn + 1. Let Sn (z) be the singular part of pn (z − an )/h(z). This means that
Sn (z) − pn (z − an )/h(z) is analytic in a neighborhood of an . By Mittag-Leffler’s Theorem,
we can find g meromorphic on Ω with singular part Sn (z) at an . Then g(z)−pn (z−an )/h(z)
extends to be analytic near an and so g(z)h(z) − pn (z − an ) is analytic near an with a zero
of order at least dn + 1 at an . This proves the corollary with f = gh. 

§3. Blaschke Products.

Weierstrass’s Theorem shows the zero set of an analytic function on the unit disc can
be any sequence tending to the unit circle. More information on the growth of the analytic
168 IX. Series and Products

function will give us more information on how quickly the zeros approach the circle. This
is the content of Jensen’s formula.

Theorem 3.1 (Jensen). Suppose f is meromorphic on |z| ≤ R with zeros a1 , . . . , an and

poles b1 , . . . , bm . Suppose also that 0 is not a zero or a pole of f . Then
Z π X X
1 R R
log |f (Reit )|dt = log |f (0)| + log − log .
2π −π |ak | |bj |

Proof. Replacing f (z) by f (Rz), we may assume R = 1. First suppose that f has no
poles or zeros on |z| = 1. Write
Q z−ak
k 1−ak z
f (z) = Q z−bj g(z),
j 1−bj z

Q Q

where g is analytic on |z| ≤ 1 and has no zeros. Then |f (0)| = j |aj |/ k |bk | |g(0)| and
log |f (eit )| = log |g(eit )| = Re log g(eit ), where log g(z) is analytic on |z| ≤ 1. Note that if
z = eit then dz/(iz) = dt. So by Cauchy’s Theorem
Z π Z π
1 it 1
log |f (e )|dt = Re log g(eit )dt
2π −π 2π −π

X X
= Re log g(0) = log |f (0)| − log |aj | + log |bk |.
j k

To prove Theorem 3.1 in the case where f has zeros or poles on |z| = 1, it suffices to
show that Z π
log |eit − eit0 |dt = 0, (3.1)
−π

by factoring out the zeros and poles on the circle. Furthermore, we can suppose t0 = 0 by
the change of variables s = t − t0 . Since log(1 − rz) is analytic on |z| ≤ 1 when r < 1, we
Rπ
have −π log |reit − 1|dt = 0. For δ > 0, write
Z π Z π  it 
 e −1 
it
log |e − 1|dt = 
log it dt
−π −π re − 1 

Z  it  Z  it 
 e −1   e −1 
= 
log it dt + log it dt. (3.2)
|t|<δ re − 1  |t|≥δ re − 1 
 §3: Blaschke Products 169

If |t| ≥ δ, then  it 
 e −1 

log it → 0
re − 1 
uniformly in t. If |t| < δ, then
  it     
  e − 1   2   2  1
log     
  reit − 1 ≤ log reit − 1 + log eit − 1 ≤ 2 log c|t|

2 π
since |reit − 1| ≥ | sin t| ≥ π |t| if δ ≤ 2. Thus
Z   it  Z
  e − 1  1 1
log dt ≤ 2 log dt ≤ c2 δ log → 0,
  reit − 1  c|t| δ
|t|<δ |t|<δ

as δ → 0. Equation (3.1) with t0 = 0 now follows from (3.2). 

Corollary 3.2. Suppose f is analytic in D, f 6≡ 0, with zeros {an }, and suppose

Z π
sup log |f (reit )|dt < ∞.
r<1 −π

Then
∞
X
(1 − |an |) < ∞.
n=0

Proof. Since f has only finitely many zeros at 0, we may divide them out and suppose
f (0) 6= 0. Then by Jensen’s formula

∞
X 1
log < ∞.
n=0
|an |

log x P 1
P
But limx→1 x−1
= 1, so that log |an |
< ∞ if and only if (1 − |an |) < ∞. 

We remark that
Z Z
it
log |f (re )| ≤ Cp |f (reit )|p dt

for some constant Cp depending on p > 0. So if |f (reit )| ∈ Lp then the hypothesis of

Corollary 3.2 holds. As a partial converse, we have the following Theorem.
170 IX. Series and Products
P
Theorem 3.3. If {an } ⊂ D such that (1 − |an |) < ∞ then

Y∞
|an | z − an
B(z) =
n=0
−an 1 − an z

converges uniformly and absolutely on compact subsets of D, where we define the conver-
gence factor |an |/(−an ) to be equal to 1 if an = 0. The function B is analytic on D, is
bounded by 1 and has zero set exactly equal to {an }.

The function in Theorem 3.3 is called a Blaschke product.

Proof. Without loss of generality an 6= 0 for all n. Then for |z| ≤ r

   
 |an | z − an  (1 − |an |)||an |z + an | 2(1 − |an |)
 
 −an 1 − an z −1= |an (1 − an z)|
≤
(inf |an |)(1 − r)
.

By Weierstrass’s M-test, B converges uniformly and absolutely on |z| ≤ r. Note that the
partial products for B are all bounded by 1 and analytic on D. 

Theorem 3.3 gives the following factorization result.

Theorem 3.4. If f is bounded and analytic on D with zero set {an } (counting multiplicity)
and if B is the Blaschke product with zero set {an } then

f (z) = B(z)eg(z)

for some analytic function g with

sup |eg(z) | = sup |f (z)|.

z∈D z∈D

Proof. Without loss of generality, |f | ≤ 1 on D. Write

YN
|an | z − an
BN (z) = .
n=1
−a n 1 − a n z

Then by Schwarz’s Lemma (Corollary II.4.3) |f (z)/BN (z)| ≤ 1. Choose rn → 1 so that

B 6= 0 on |z| = rn . Then f /BN converges uniformly to f /B on |z| = rn and by the
Maximum Principle the convergence is uniform on |z| ≤ rn . Now let rn → 1. Thus
 §4: The Gamma and Zeta Functions 171

h = f /B is bounded by 1 on D and non-vanishing. This implies g = log h can be defined

as an analytic function on D and Theorem 3.3 follows. 

It is difficult to overestimate the importance of Blaschke products in function theory

on D.

§4. The Gamma and Zeta Functions.

The simplest entire function with zeros at the integers is sin πz. By the same conver-
gence proof
∞
Y z −z/n
G(z) ≡ (1 + )e (4.1)
n=1
n
is the simplest entire function with zeros only at the negative integers. Moreover

1
zG(z)G(−z) = sin πz. (4.2)
π

Note that G(z − 1)/(zG(z)) has no zeros and hence

eγ(z) zG(z) = G(z − 1), (4.3)

where γ(z) is entire. By Weierstrass’s Theorem and the uniform convergence of the product
(4.1), we can take logarithmic derivatives and
∞   X ∞  
′ 1 X 1 1 1 1
γ (z) + + − = −
z n=1 z + n n n=1
z − 1 + n n

X∞  
1 1 1
= −1+ −
z n=1
z+n n+1

X∞   X ∞  
1 1 1 1 1
= −1+ − + − .
z n=1
z+n n n=1
n n+1

Thus
∞ 
X 
′ 1 1
γ (z) = −1 + − =0
n=1
n n+1
172 IX. Series and Products

and γ is a constant. By (4.1) and (4.3)

Ym  
γ γ1 −1/n 1 1 1
1 = G(0) = e G(1) = lim e 1+ e = lim eγ (m + 1)e−(1+ 2 + 3 +...+ m ) .
m→∞
n=1
n m→∞

Thus  
1 1 1
γ = lim 1 + + + . . . + − log(m + 1) ≈ 0.57722 . . . .
m→∞ 2 3 m
The proof above shows that this limit, called Euler’s constant, exists.
The Gamma Function is defined by
Y∞
1 −γz 1 1
Γ(z) = = e ez/n .
zG(z)eγz z n=1 1 + z/n

Then Γ is a non-vanishing meromorphic function with poles at the non-positive integers

and no other poles. Moreover by (4.3)

Γ(z + 1) = zΓ(z) (4.4)

and by (4.2)
π
Γ(z)Γ(1 − z) = . (4.5)
sin πz
It follows from (4.1) and (4.2) that Γ(1) = 1 and by induction

Γ(n) = (n − 1)!

1
By (4.5) at z = 1/2, we also obtain Γ( 12 ) = π 2 .

Theorem 4.1. If Rez > 0 then

Z ∞
Γ(z) = tz−1 e−t dt. (4.6)
0

Proof. The integral converges uniformly and absolutely on compact subsets of Rez > 0
because |tz−1 | = |t|Rez−1 is integrable near t = 0 and because |t|Rez−1 e−t ≤ e−t/2 for t
large. By Morera’s Theorem, the right side of (4.6) is analytic for Rez > 0.
Set Z  n
n
z−1 t
Γn (z) = t 1− dt.
0 n
 §4: The Gamma and Zeta Functions 173

Then we claim that for Rez = x > 0

Z ∞
lim Γn (x) = tx−1 e−t dt (4.7)
n→∞ 0

and
1 1
lim = . (4.8)
n→∞ Γn (z) Γ(z)
If (4.7) and (4.8) hold then Theorem 4.1 follows from the Uniqueness Theorem.

Lemma 4.2. If M < ∞ then

t n t
(1 − ) e increases to 1,
n

uniformly for 0 ≤ t ≤ M .

Proof. The slope of the secant line to the graph of log u with one end at u = 1 − t/n and
the other end at u = 1 decreases to 1 as n → ∞ since the graph of log u is concave. Thus
 
t log(1 − nt ) − log 1
n log(1 − ) + t = − 1 (−t) → 0, (4.9)
n (1 − nt ) − 1

uniformly for 0 ≤ t ≤ M . Lemma 4.2 now follows since −t ≤ 0 and es is an increasing

function of s. 

By Lemma 4.2, Z ∞
Γn (x) ≤ Γn+1 (x) ≤ tx−1 e−t dt,
0

and (4.7) holds for x > 0. Substitute t = ns in the definition of Γn (z) and obtain
Z 1
z
Γn (z)n = sz−1 (1 − s)n ds ≡ fn (z). (4.10)
0

Note that f0 (z) = 1/z and fn+1 (z) = fn (z) − fn (z + 1), so that by induction

n!
fn (z) = .
z(z + 1) · · · (z + n)

Thus

1
n
Y z
n
Y z Pn 1 1
−z
=n z (1 + ) = z (1 + )e−z/k ez( k=1 k −log n) → ,
Γn (z) k k Γ(z)
k=1 k=1
174 IX. Series and Products

as n → ∞, proving (4.8) and Theorem 4.1. 

The Riemann Zeta Function is defined by

∞
X
ζ(z) = n−z ,
n=1

where n−z = e−z log n . The zeta function is important because of its connection with the
prime numbers, as illustrated in Theorem 4.3 below. By the integral test, the series for ζ(z)
converges uniformly and absolutely on compact subsets of Rez > 1 and hence is analytic
in Rez > 1. The zeta function can be extended to Rez > 0 by comparing the sum to the
appropriate integral. For Rez > 1

∞
X Z n+1
1 −z
ζ(z) − = n − t−z dt
z − 1 n=1 n

∞ Z
X n+1 Z t
= zs−z−1 ds. (4.11)
n=1 n n

The estimate
Z n+1 Z t
|z| −Rez−1
|zs−z−1 |ds ≤ n
n n 2

and the Weierstrass M-test show that (4.11) converges uniformly and absolutely on com-
pact subsets of Rez > 0 to an analytic function. Thus ζ(z) extends to be meromorphic in
Rez > 0 with a simple pole at z = 1 and no other poles. The Riemann Hypothesis,
perhaps the most famous problem in mathematics, is that ζ(z) 6= 0 for Rez > 12 . The next
Theorem proves this statement for Rez > 1 and gives the connection between ζ(z) and the
prime numbers.
 §4: The Gamma and Zeta Functions 175

−1

−2

−3

−2 −1 0 1 2 3 4 5

Figure IX.1 ζ( 21 + iy), −40 ≤ y ≤ 40.

Theorem 4.3. Let {pn } be the primes, with p1 = 2, p2 = 3, . . . . For Rez > 1,
∞
Y
1
= (1 − p−z
n ). (4.12)
ζ(z) n=1

In particular, ζ(z) 6= 0 for Rez > 1.

Proof. The infinite product in Theorem 4.3 converges uniformly and absolutely because
for x = Rez > x0 > 1,
|p−z −x
n | = pn ≤ n
−x0
,

so that by the Weierstrass M-test, the right side of (4.12) is analytic in Rez > 1. Because
the series is absolutely convergent, we can rearrange the terms and obtain
X X X
ζ(z)(1 − 2−z ) = n−z − (2n)−z = m−z .
m odd

Similarly
X X X
ζ(z)(1 − 2−z )(1 − 3−z ) = m−z − (3m)−z = m−z .
m odd m odd m not divisible
by 2 or 3

By induction
k
Y X
ζ(z) (1 − p−z
n ) = m−z = 1 + p−z
k+1 + . . . .
n=1 m not div. by p1 ,...,pk
176 IX. Series and Products
P
Since pk+1 → ∞ and since |n−z | converges, the product converges and

∞
Y
ζ(z) (1 − pn−z ) = 1,
n=1

for Rez > 1. 

As one might expect from (4.10), there is a connection between Γ(z) and ζ(z). By
the same substitution used to establish (4.10),
Z ∞
z
Γ(z) = n sz−1 e−ns ds,
0

and
∞
X ∞ Z
X ∞
−z
n Γ(z) = sz−1 e−ns ds
n=1 n=1 0

Z ∞
e−s
= sz−1 ds.
0 1 − e−s
Thus
Z ∞
1 sz−1
ζ(z) = ds, (4.13)
Γ(z) 0 es − 1
for Rez > 1.
The Riemann zeta function is used to prove the celebrated prime number theorem
which describes the rate of growth of the number of primes which are at most x, as
x → ∞. One of the main ingredients is the next Theorem.

Theorem 4.4. The Riemann zeta function ζ(z) is non-zero for Rez ≥ 1.

Proof. We proved in Theorem 4.3 that ζ(z) 6= 0 in Rez > 1. By (4.12) for Rez > 1

∞ ∞
ζ ′ (z) X log pn X log pn
− = z
= Φ(z) + , (4.14)
ζ(z) p −1
n=1 n
p (pz − 1)
z
n=1 n n

where
X∞
log pn
Φ(z) = z
.
n=1
pn
 §5: Exercises 177
1
The last sum in (4.14) converges and is analytic for Rez > 2, so that Φ(z) extends to
1
be meromorphic in Rez > 2, with poles only at z = 1 and the zeros of ζ(z). Note that
ζ(z) = ζ(z), so that if ζ(z) has a zero of order M ≥ 0 at z = 1 + iα and a zero of order
N ≥ 0 at 1 + 2iα then

lim εΦ(1 + ε) = 1, lim εΦ(1 + ε ± iα) = −M, and lim εΦ(1 + ε ± i2α) = −N.
ε↓0 ε↓0 ε↓0

Multiplying the magical inequality

X2   X∞
4 log pn iα/2
Φ(1 + ε + irα) = 1+ε (pn + pn−iα/2 )4 ≥ 0 (4.15)
r=−2
2 + r n=1
pn

by ε > 0 and letting ε → 0 shows that 6 − 8M − 2N ≥ 0. Since M and N are non-negative

integers, we must have that M = 0, i.e. ζ(1 + iα) 6= 0. 

§5. Exercises.

Q∞ Q∞
1. (a) Prove that n=1 (1 + ni ) diverges but n=1 |1 + ni | converges.
√
(−1)j / j
P P
(b) If pj = e then log pj converges but (pj − 1) diverges.
√ P P
(c) If pj = 1 + i(−1)j / j prove (pj − 1) converges but log pj diverges.
2. Prove directly from the definition
Y∞
n2 + π
n=3
n2 − 1

converges. (Don’t just quote a theorem or proposition. It is practice with making esti-
mates.)

3. Find an explicit bounded analytic function f 6≡ 0 defined on the unit disk with zeros
{zn } such that every point of the unit circle is a cluster point of {zn } and such that
X
(1 − |zn |) < ∞.
178 IX. Series and Products

4. Find a meromorphic function with singular parts

n 4
Sn (z) = + ,
z − n (z − n)2

for n = 1, 2, 3, . . .. Prove convergence of your sum.

5. If w1 /w2 is purely imaginary, write P in terms of the inverse of a conformal map to a
rectangle.

6. Suppose {bk } → ∞ and |ak | ≤ M < ∞ for all k. Prove that

X a k   k  j 
ak X z
−
z − bk −bk j=0 bk
k

is meromorphic in C with singular part ak /(z − bk ) at bk and no other poles in C. The

point of this problem is that you can take the number of terms equal to k, no matter how
slowly bk → ∞. In practice it is usually best to take as few terms as possible, so in many
examples fewer terms are taken.

7. Find an explicit entire function g with g(n log n) = nπ , n = 1, 2, 3, . . .. Prove that your
function works. Hint: |(1 − t)|et is decreasing for 0 < t < 1 and increasing for t > 1. Use
this to compare the derivative of a product vanishing at n log n to one vanishing at n, for
integers n.

8. Find an entire function of least possible genus, with simple zeros at the Gaussian
integers:
{m + in : m, n integers },

and no other zeros. (prove your result).

9. If Ω is a region, then H(Ω) is the algebra of analytic functions on Ω. One of the

first questions that you might ask about an algebra are: what are its ideals? Suppose
g1 , g2 ∈ H(Ω) with no common zeros. Prove that there are functions f1 , f2 ∈ H(Ω) with

(∗) f1 (z)g1 (z) + f2 (z)g2 (z) = 1

 §5: Exercises 179

for all z ∈ Ω. If g1 and g2 had a common zero then we could not find f1 and f2 satisfying
(∗). This says that the ideal generated by g1 and g2 consists of all analytic functions on Ω
if and only if g1 and g2 have no common zeros.

Γ′ (z)
10. Let Ψ(z) = Γ(z)
.
(a) Show: Ψ′ (z) + Ψ′ (z + 12 ) = 4Ψ′ (2z).
(b) Integrate (a) twice to obtain

1
Γ(2z) = eaz+b Γ(z)Γ(z + ).
2

for some constants a and b. (Careful: log Γ(z) is only defined locally.)
(c) Find a and b. The resulting formula is called Legendre’s duplication formula.

11. Stirling’s formula for the gamma function, used in combinatorics, physics, and else-
where.
√
(a) Suppose x > 21 . Substitute t = ( x + v)2 in the integral formula (4.6) for Γ(x) to
show that √ Z ∞
Γ(x)ex x 2

x
=2 ϕx (v)e−v dv,
x −∞

where  √

 0 if v ≤ − x

ϕx (v) =  2x−1
 1 √

e−2vx 2
1+ √v if v ≥ − x.
x

(b) Prove that √ Z ∞

Γ(x)ex x −2v 2
√
lim = 2 e dv = 2π.
x→∞ xx −∞

Hint: find an upper bound for log ϕx (v) as a function of v.

(c) Use part (b) to estimate n!.

12. (a) Prove that for b ∈ Ω,

Ib = {f ∈ H(Ω) : f (b) = 0}
180 IX. Series and Products

is a closed maximal ideal in H(Ω) and all closed maximal ideals are of this form. The
topology is given by uniform convergence on compact subsets of Ω. Hint: First prove the
analog of problem 9 for n functions instead of 2. Then use that H(Ω) is closed under
uniform convergence on compact sets.

(b) The functions which are “eventually” zero on a sequence tending to the boundary is
an ideal in H(Ω) which is dense and not generated by a single function.

(c) Find the principal ideals.

 X

Harmonic Functions

Harmonic functions arise in applications as solutions of Laplace’s equation ∆u = 0.

They are useful for the study of analytic functions in two ways: if f is analytic then Ref
is harmonic and, more importantly, log |f | is harmonic where f 6= 0. The most useful
property of harmonic functions is the Mean Value Property, and its consequence, the
Maximum Principle. To avoid difficulties with proving differentiability properties, we take
the mean value property as definition, then give the connection with Laplace’s equation in
Theorem 2.3.

§1. The Mean Value Property and the Maximum Principle.

In this section, we define harmonic and subharmonic functions, prove the maximum
principle and then show how to find the values of a harmonic function on a disc from its
values on the bounding circle.

Definition 1.1. If u is a continuous real-valued function defined on a region Ω ⊂ C so

that for each z ∈ Ω there is an rz > 0 (depending possibly on z) and
Z 2π
1
u(z) = u(z + reit )dt (1.1)
2π 0

for all r < rz , then we say that u is harmonic on Ω.

Equation (1.1) is called the Mean Value Property. The Mean Value Property says
that the value of u at the center of the disk is equal to the average of its values on the
boundary of the disk. This is interesting because it says that if you know u on the circle,
then you can recover its value inside the disk, at least at the center. We’ll learn later how
to recover values at other points of the disk.

181
182 X. Harmonic Functions

Definition 1.2. If u is a continuous function defined on a region Ω ⊂ C, with values in

[−∞, +∞) so that for each z ∈ Ω there is an rz > 0 (depending possibly on z) and
Z 2π
1
u(z) ≤ u(z + reit )dt (1.2)
2π 0

for all r < rz , then we say that u is subharmonic on Ω.

Equation (1.2) is called the Mean Value Inequality. Note that we allow a subhar-
monic function to take the value −∞ but not +∞. The Mean Value Inequality is trivially
satisfied for those points z where u(z) = −∞.
The reader can easily verify that if u and −u are subharmonic then u is harmonic.
If u1 and u2 are harmonic then A1 u1 + A2 u2 is harmonic, for real numbers A1 , A2 , and
|A1 u1 | is subharmonic. If u is subharmonic then Au is subharmonic provided A > 0.
For example, an analytic function f satisfies the Mean Value Property, as can be
seen by expanding in a (local) power series and interchanging the order of integration and
summation. Then by taking real and imaginary parts, u = Ref and v = Imf are harmonic.
It also follows that |f | is subharmonic. Harmonicity, like analyticity, is a local property,
so that log |f | is harmonic on Ω \ {f = 0} since log |f | is the real part of log f , which can
be defined to be analytic in a neighborhood of any point where f 6= 0. This also shows
that log |f | is subharmonic on Ω. Moreover by Exercise 1, log |f | satisfies the Mean Value
Inequality on a disc whose closure is contained in Ω. Of these examples, perhaps the most
important for the study of analytic functions is the subharmonicity of log |f |. We will use
subharmonic functions to study harmonic functions as well as analytic functions.
The next result is perhaps the most important elementary result in complex analysis.

Theorem 1.3 (Maximum Modulus Principle). Suppose u is subharmonic on a region

Ω. If there exists z0 ∈ Ω such that

u(z0 ) = sup u(z) (1.3),

z∈Ω

then u is constant.

It follows that subharmonic functions do not have any local maxima, unless constant.
 §1: Mean Value Property 183

Proof. Suppose (1.3) holds. Set E = {z ∈ Ω : u(z) = u(z0 )}. Since u is continuous, E is
closed. By equation (1.3), the set E is non-empty. We need only show E is open, since Ω
is connected. If z1 ∈ E, then by (1.2)

Z 2π
1
[u(z1 ) − u(z1 + reit )]dt ≤ 0, (1.4)
2π 0

for r < rz1 . But the integrand in (1.4) is non-negative, by (1.3), and continuous and hence
identically 0 for all t and all r < rz1 . This proves E is open and hence equal to Ω. 

Another form of the Maximum Principle is:

Corollary 1.4. If u is a non-constant subharmonic function in a bounded region Ω and

if u is continuous on Ω then
max u(z)
z∈Ω

occurs on ∂Ω but not in Ω.

The reader should verify the alternate form: if u is continuous and subharmonic on Ω
then
lim sup u(z) = sup u(z).
z→∂Ω Ω

If Ω is unbounded, then ∞ must also be viewed as part of ∂Ω. The function u(z) = Rez is
harmonic on Ω = {z : Rez > 0} and satisfies u = 0 on ∂Ω ∩ C but u is not bounded by 0.

Theorem 1.5 (Schwarz). If g is real-valued and continuous on ∂D, set

Z 2π
1 1 − |z|2
u(z) = g(eit )dt,
2π 0 |eit − z|2

for z ∈ D. Then u is harmonic in D and

lim u(z) = g(ζ), (1.5)

z→ζ

for all ζ ∈ ∂D.

Proof. The function Z 2π

1 eit + z
G(z) = g(eit )dt
2π 0 eit − z
184 X. Harmonic Functions

is analytic on D as can be seen by expanding the kernel

X ∞
eit + z 1 + e−it z
= = 1 + 2 e−int z n
eit − z 1 − e−it z 1

and interchanging the order of summation and integration. The identity

 
1 − |z|2 eit + z
= Re (1.6)
|eit − z|2 eit − z

shows that u = ReG and hence u is harmonic. Note also that if g ≡ 1, then G ≡ 1, since
R −int
e dt = 0 if n 6= 0. Thus for all z ∈ D
Z 2π
1 1 − |z|2
dt = 1. (1.7)
2π 0 |eit − z|2

To prove (1.5) fix t0 and ε > 0 then choose δ > 0 so that |g(eit ) − g(eit0 )| < ε/2 if
t ∈ Iδ = {t : |t − t0 | < δ}. Then using (1.7)
 1 Z 2π 1 − |z|2 
it0  it it0 
|u(z) − g(e )| =  (g(e ) − g(e ))dt 
2π 0 |eit − z|2

Z Z
1 1 − |z|2 ε 1 1 − |z|2
≤ it 2
dt + 2 sup |g|dt.
2π Iδ |e − z| 2 2π {|t−t0 |≥δ} |eit − z|2

If we increase the first integral to include the whole circle then it is at most ε/2 by (1.7).
If z is sufficiently close to eit0 then the the second integral can be made smaller than
ε/2 because |z| is close to 1 and |eit − z| is not too small since |t − t0 | ≥ δ, and thus
|u(z) − g(eit0 )| < ε. 

The proof of Schwarz’s Theorem shows that we need only assume that g is integrable
on ∂D and continuous at ζ for (1.5) to hold. The kernel

1 1 − |z|2
Pz (t) =
2π |eit − z|2
R
is called the Poisson kernel and u = P I(g) ≡ Pz (t)g(eit )dt is called the Poisson
integral of g. If B is a disk and g is continuous on ∂B, then P IB (g) denotes the harmonic
function on B which equals g on ∂B, and is called the Poisson integral of g on B. The reader
 §1: Mean Value Property 185

can give a formula for P IB (g) by transplanting the problem to the unit disk using a linear
map. Note that by (1.7) if g is integrable and satisfies m ≤ g ≤ M then m ≤ P IB (g) ≤ M .
The next result shows how to find the values of a harmonic function in the disc from
its values on the boundary.

Corollary 1.6. If u is harmonic on |z| < 1 and continuous on |z| ≤ 1, then

Z 2π
1 1 − |z|2
u(z) = u(eit )dt. (1.8)
2π 0 |eit − z|2

Proof. Let U (z) denote the right side of (1.8). Then by Schwarz’s Theorem u − U is
harmonic on D, continuous on D and equal to 0 on ∂D. By the Maximum Principle
applied to u − U and U − u, we conclude u = U . 

The next Corollary shows how to recapture an analytic function from the boundary
values of its real part.

Corollary 1.7. If u is harmonic on |z| < 1 and continuous on |z| ≤ 1, then

Z 2π it
1 e +z
f (z) = it
u(eit )dt
2π 0 e − z

is the unique analytic function on D with Ref = u and Imf (0) = 0.

The kernel (eit + z)/(eit − z) is called the Herglotz kernel and the integral is called
the Herglotz integral.

Proof. By the first part of the proof of Schwarz’s Theorem 1.5, f is analytic. The real
part of f is then equal to u by (1.6) and Corollary 1.6. If g is another analytic function
with Reg = u, then f − g is purely imaginary and hence not open. Thus f − g is constant.
R
Finally note that f (0) = u(eit )dt is real. 

It follows from Corollary 1.7 that a harmonic function on a region Ω is the real part
of an analytic function on each disc contained in Ω. One consequence is that it is not hard
186 X. Harmonic Functions

to prove that if u is harmonic on a disk and if f is analytic then u ◦ f is harmonic. Indeed,

harmonicity is local and on a disk u = Reg by Corollary 1.7. Hence u ◦ f = Re(g ◦ f ) is
harmonic. If u = Rez and f = z 2 then f ◦ u is not harmonic. The function u = log |z|
is harmonic on Ω = {z : 0 < |z| < ∞} and is the real part of an analytic function on
each disc that does not contain 0, but it is not the real part of an analytic function on Ω
because arg z is not continuous on Ω.

Here is an application to analytic functions:

Corollary 1.8 (Jump Theorem). Suppose f is an integrable function on ∂D. Let

Z
f (ζ) dζ
F (z) = .
|ζ|=1 ζ − z 2πi

Then F is analytic on C \ ∂D and for |ζ| = 1,

  
1
lim F (z) − F = f (ζ)
z→ζ z

at all points of continuity of f .

The function F is called the Cauchy integral or Cauchy transform of f . The Jump
Theorem says that the analytic function F jumps by f (ζ) as z crosses the unit circle at
ζ. Notice that if f is analytic on D then F = f in D and F = 0 in |z| > 1 by Cauchy’s
Integral Formula, Theorem V.1.2.

Proof. We already proved that F is analytic off ∂D in Lemma IV.3.10. To prove the
Corollary, just manipulate the integrals:
  Z  
1 1 1 dζ
F (z) − F = f (ζ) −
z |ζ|=1 ζ −z ζ − 1/z 2πi

Z
z − 1/z dζ
= f (ζ)
|ζ|=1 (ζ − z)(ζ − 1/z) 2πi

Z 2π
1 − |z|2 dt
= f (eit ) .
0 |eit − z|2 2π
 §2: Cauchy-Riemann and Laplace Equations 187

Applying Schwarz’s Theorem 1.5 to the real and imaginary parts of f completes the proof.


§2. Cauchy-Riemann and Laplace Equations.

If f (z) = u(z) + iv(z) we will sometimes use the notation f (x, y) = u(x, y) + iv(x, y)
where z = x + iy with x, y real and u(x, y) and v(x, y) real-valued. If f is analytic then
by the definition of the complex derivative

f (x + h, y) − f (x, y)
f ′ (z) = lim = fx (x, y) = ux (x, y) + ivx (x, y)
h→0 h

f (x, y + k) − f (x, y) 1
= lim = fy (x, y) = vy (x, y) − iuy (x, y)
k→0 ik i
Thus
ux = vy and uy = −vx . (2.1)

Equations (2.1) are called the Cauchy-Riemann equations. If u and v satisfy the
Cauchy-Riemann equations then v is called a harmonic conjugate of u. In this case,
then −u is a harmonic conjugate of v. If v1 and v2 are harmonic conjugates of u on a region
Ω then v1 − v2 is constant. By the definition of analytic, f ′ is continuous and so fx and fy
are continuous. But f ′ is also analytic and hence all second order partial derivatives of f
exist and are continuous. By induction f has continuous partial derivatives of all orders.
In particular if u = Ref , then

uxx = (vy )x = (vx )y = (−uy )y = −uyy , (2.2)

and hence uxx + uyy = 0. Similarly vxx + vyy = 0.

A harmonic function on a region Ω is the real part of an analytic function on each disk
contained in Ω by Corollary 1.7. So harmonic functions have continuous partial derivatives
of all orders. A harmonic function is not always the real part of an analytic function on
all of Ω. However, the Cauchy-Riemann equations allow us to associate another analytic
function with a harmonic function: A continuously differentiable function u is harmonic
on a region Ω if and only if ux − iuy is analytic on Ω. Indeed, if B is a disk contained
188 X. Harmonic Functions

in Ω then by Corollary 1.7, u = Ref for some analytic function f on B. Moreover f ′ is

analytic and f ′ = ux + ivx = ux − iuy by the Cauchy Riemann equations. This proves
that ux − iuy is analytic.

Definition 2.1. The Laplace of u is the second order derivative given by ∆u = uxx + uyy .
We say that u satisfies Laplace’s equation on a region Ω if u has continuous second
order partial derivatives (including the mixed partials) and ∆u = 0 on Ω.

The next result relates Laplace’s equation to harmonicity, and shows that both are
essentially the same as the maximum principle.

Theorem 2.3. Suppose u is real-valued and continuous on a region Ω. Then the following
are equivalent:
(i) u is harmonic on Ω

(ii) u satisfies Laplace’s equation on Ω

(iii) If B is a disk with B ⊂ B ⊂ Ω and if v is harmonic on B, then u − v and v − u satisfy

the maximum principle on B.

Proof. We have already seen that (i) implies (ii). By the Maximum Principle (i) implies
(iii). If (iii) holds and if B ⊂ B ⊂ Ω, then let v be the Poisson integral of u|∂B on B. Then
v is harmonic on B and u − v and v − u are equal to 0 on ∂B by Schwarz’s Theorem 1.5 .
By (iii), u = v on B. This proves u is harmonic on each B ⊂ Ω, and hence (i) holds.
Finally we show that (ii) implies (i). Harmonicity is a local property so we may assume
Ω is a disk B centered, say, at z0 . Set g = ux − iuy . Now if R is a rectangle contained in
R
B, we claim that ∂R g(ζ)dζ = 0. To see this, note that
Z Z Z
(ux − iuy )(dx + idy) = ux dx + uy dy + i ux dy − uy dx.
∂R ∂R ∂R

By the Fundamental Theorem of Calculus applied to each segment in ∂R, the first integral
is zero. Also by the Fundamental Theorem of Calculus, integrating along horizontal lines
in R then vertical lines in R, the second integral can be rewritten as
Z
(uxx + uyy )dxdy,
R
 §2: Cauchy-Riemann and Laplace Equations 189

which is also equal to 0 by (ii). (We’ve just followed a proof of the simplest form of Green’s
theorem). By Morera’s Theorem, g is analytic. Since B is simply connected g = f ′ for
some analytic function f . In fact, this function was given in the proof of Morera’s Theorem.
Set w = Ref . Then by the Cauchy-Riemann equations and the definition of g, we have
that ux = wx and uy = wy so that u = w + c = Re(f + c) where c is a constant. This
proves that u is harmonic. 

There are several interesting consequences of the proof of Theorem 2.3 which are worth
noting.

Corollary 2.5. A function u is harmonic on a simply connected region Ω if and only if

u = Ref for some f analytic on Ω.

Proof. If u is harmonic then by the Cauchy-Riemann equations, ux − iuy is analytic.

R
Since Ω is simply connected, if γ is a closed curve then γ (ux − iuy )dz = 0 by Cauchy’s
R
Theorem. Fix z0 ∈ Ω. We can then define f (z) = σz (ux − iuy )dζ where σz is any curve
contained in Ω connecting z0 to z. This definition does not depend on the choice of σz
because the integral along any closed curve is zero. If B ⊂ Ω is a disk, then for z ∈ B we
can write f (z) as an integral from z0 to the center of B plus an integral along a horizontal
then vertical line segment from the center of B to z. By the proof of Morera’s Theorem
IV.2.9, f is analytic and Ref + c = u for some constant c as in the proof of Theorem 2.3.
As we have already seen, the converse follows from the Mean Value equality for analytic
functions. 

Corollary 2.6. Suppose u and v are real-valued and have continuous second partials
(including the mixed partials) on a region Ω. Then u and v satisfy the Cauchy-Riemann
equations (2.1) if and only if u + iv is analytic on Ω.

Proof. We have already proved that if u + iv is analytic then u and v satisfy the Cauchy
Riemann equations. Conversely if u and v satisfy the Cauchy Riemann equations, then
190 X. Harmonic Functions

by (2.2) u and v are harmonic. By the Cauchy Riemann equations ux − iuy = vy + ivx is
analytic. In any disk, ux − iuy is the derivative of an analytic function G. As in the proof
of Theorem 2.3, we deduce that u + iv = G + c, where c is a constant, by noting that both
sides of this equality have the same x and y derivatives. This proves that u + iv is analytic
on all of Ω. 

In Section XVIII.XX we will weaken the differentiability requirements in Theorem

2.3 and its Corollaries. It is also known that to prove a continuous function is harmonic
in a region Ω, it suffices to show that each point in Ω is the center of two circles, whose
corresponding disks are contained in Ω, such that the Mean Value Property holds for these
two circles. I think it is an open problem to prove a similar result for only one circle per
point. (Better check this).

§3. Hadamard, Lindelöf, and Harnack.

The maximum principle can be used to give more information than is immediately
apparent from its statement.

Theorem 3.1 (Hadamard’s Three Circles Theorem). Suppose f is analytic in the

annulus A = {z : r < |z| < R}. Let m = lim sup|z|→r |f (z)| and M = lim sup|z|→R |f (z)|.
Then
|f (z)| ≤ M ω(z) m1−ω(z) ,

where ω(z) = (log |z|/r)/ log R/r.

Theorem 3.1 is called the three circles theorem because ω(z) depends only on |z|, so
the theorem relates the max on the two outer circles to the max on another circle. Another
way to state the conclusion is that if M (s) = sup|z|=s |f (z)| then log |M (s)| is a convex
function of log |z|.

Proof. The function ω log M + (1 − ω) log m is harmonic on A and equal to log M on

|z| = R and equal to log m on |z| = r. Thus u = log |f | − (ω log M + (1 − ω) log m) is
subharmonic on A and lim supz→∂A u(z) ≤ 0. By the Maximum Principle, u ≤ 0 in A. 
 §3: Hadamard, Lindelöf, and Harnack 191

Fourier Series were first developed by Fourier because he wanted to solve the following
problem: Find a function u which satisfies 0 < u < 1 and ∆u = 0 in the half strip
S = {x + iy : x > 0, 0 < y < π} with u(0, y) = 1, 0 < y < π, and u(x, 0) = u(x, π) = 0 for
x > 0. If a thin plate of homogeneous material in the shape of a long strip is heated to one
temperature on the top and bottom, and another temperature on the end and allowed to
reach a state of equilibrium, then the resulting temperature satisfies ∆u = 0 on the strip.
Fourier’s idea was that enx sin ny = Im(enz ) is harmonic and equal to 0 on the top and
P∞
bottom edges of S. So set u(z) = n=1 cn enx sin ny and choose cn so that
∞
X
1 = u(0, y) = cn sin ny.
n=1

The coefficients cn can be determined by setting F = 1 for x ∈ (0, π) and F = −1 for

x ∈ (−π, 0) and integrating F against sin ny. There are convergence issues though.
Another way to solve Fourier’s problem is to note that ϕ(z) = (z +1/z) ◦ ez = ez +e−z
is a conformal map of the half strip onto the upper-half plane H, which maps the end of
z−b
the strip to the interval (−2, 2). If −∞ < a < b < ∞ then the function θ(z) = Im log z−a

is harmonic H and equal to the angle θ formed at z ∈ H between line segments from b to
z and from a to z.

a b
Figure X.2

The angle θ approaches π as z → (a, b) and approaches 0 as z → R \ (a, b). Thus

 
1 z−2
u(z) = arg ◦ (ez + e−z )
π z+2
solves Fourier’s problem. It is natural though to ask if this function is the only solu-
tion. The answer is provided by an extension to the maximum principle, called Lindelöf’s
Maximum Principle.
192 X. Harmonic Functions

Theorem 3.2 (Lindelöf ). Suppose Ω is open and suppose {ζ1 , . . . , ζn } is a finite subset
of ∂Ω, not equal to all of ∂Ω. If u is subharmonic on Ω with u ≤ M < ∞ on Ω and if

lim sup u(z) ≤ m,

z∈Ω→ζ

for all ζ ∈ ∂Ω \ {ζ1 , . . . , ζn } then u ≤ m on Ω.

Proof. First suppose that Ω is bounded and let d = diam(Ω). For ε > 0 set
n  
X  z − ζj 
uε (z) = u(z) + ε 
log  . (3.1)
d 
j=1

Then uε is subharmonic in Ω, uε ≤ u and uε → −∞ as z → ζj , for j = 1, . . . , n. Thus

lim supz→∂Ω uε (z) ≤ m, and so by the Maximum Principle uε ≤ m on Ω. Fix z and let
ε → 0 in (3.1) to obtain u ≤ m.
If Ω is not bounded, we may suppose that ζj 6= ∞ for j = 1, . . . , n by composing
with a Möbius map if necessary. Given ε > 0, we can choose R so that u(z) ≤ m + ε for
z ∈ Ω ∩ {|z| > R}. Now apply the bounded case to u − ε on Ω ∩ {|z| < R} to conclude
that u − ε ≤ m on Ω ∩ {|z| < R}. Let ε → 0 and R → ∞ to conclude u ≤ m on Ω. 

Lindelöf’s Maximum principle shows that there is a unique bounded harmonic func-
tion solving Fourier’s problem since the difference of two such solutions is bounded and
harmonic with boundary values equal to 0 except at three boundary points, 0, πi and ∞.
Similarly, if Ij = (aj , bj ), j = 1, . . . , n are disjoint intervals on R then
n
X  
cj z − bj
u= arg
π z − aj
j=1

is the unique bounded harmonic function on H with boundary values cj on Ij and 0 on

R \ ∪I j .
Another consequence of Lindelöf Maximum Principle is the Three Lines version of
Hadamard’s theorem, which plays an important role in complex interpolation theory of
operators. It follows from the proof of Hadamard’s Three Circles Theorem and Lindelöf’s
Maximum Principle.
 §3: Hadamard, Lindelöf, and Harnack 193

Corollary 3.3. If u is subharmonic on the strip S0 = {z = x + iy : 0 < x < 1} set

m0 = lim supRez→0 u(z) and m1 = lim supRez→1 u(z). If u ≤ M < ∞ on S0 then u(z) ≤
m1 x + m0 (1 − x).

For example, suppose g and h are continuous functions on a compact set X and suppose
dµ is a positive measure on X. Then the function
Z
F (z) = |g|pz |h|q(1−z) dµ
X

is analytic on S0 for p, q > 0. This follows from Morera’s Theorem after interchanging the
order of integration. Thus Z 
 
 pz
log  |g| |h| q(1−z)
dµ
X

is subharmonic on S0 , and bounded above. By Corollary 3.3, for 0 < x < 1,

Z Z Z
px q(1−x) p
log |g| |h| dµ ≤ x log |g| dµ + (1 − x) log |h|q dµ.
X X X

If 1/p + 1/q = 1 with p > 1 then set x = 1/p and exponentiate to obtain
Z Z  p1 Z  q1
p q
|gh|dµ ≤ |g| dµ |h| dµ ,
X X

which is called Hölder’s inequality. By an approximation, Hölder’s inequality holds for all
g ∈ Lp (dµ) and h ∈ Lq (dµ) and for any measure space X.
Positive harmonic functions cannot grow too quickly in the unit disk as the next result,
known as Harnack’s inequality, shows.

Theorem 3.4 (Harnack). Suppose u is a positive harmonic function on D. Then

   
1−r 1+r
u(0) ≤ u(z) ≤ u(0),
1+r 1−r

for |z| = r.

Proof. We may assume u is harmonic on D by replacing u with u(sz), s < 1 and then
letting s → 1. We first estimate the Poisson Kernel:

1−r 1 − r2 1 − |z|2 1 − r2 1+r

= 2
≤ it 2
≤ 2
= .
1+r (1 + r) |e − z| (1 − r) 1−r
194 X. Harmonic Functions

Because u is positive and satisfies the Mean Value Property

  Z 2π  
1−r 1 − |z|2 dt 1+r
u(0) ≤ it 2
u(eit ) ≤ u(0). 
1+r 0 |e − z| 2π 1−r

A similar estimate can be used to show that two positive harmonic functions on D
vanishing on a interval of ∂D must approach 0 at the same rate.

Theorem 3.5 (Boundary Harnack inequality). Suppose u and v are positive har-
monic functions on D which extend to be continuous and equal to 0 on an arc I ⊂ ∂D. Let
Uδ = {z ∈ D : dist(z, ∂D \ I) ≥ δ}. Then for z ∈ Uδ
   
δ2 u(0) u(z) 4 u(0)
≤ ≤ 2 .
4 v(0) v(z) δ v(0)

Proof. First suppose that u and v are continuous on D. By the Poisson integral formula
Z
u(eit )dt
u(z) ∂D\I |eit − z|2
4 u(0)
=Z it
≤ 2 .
v(z) v(e )dt δ v(0)
it 2
∂D\I |e − z|

The lower bound is similar. To remove the assumption that u and v are continuous on
∂D \ I, take a simply connected subset Ω ⊂ D with ∂Ω ∩ ∂D = I. For instance Ω can be
the intersection of D and another disk which is very close to D. If the conformal map ϕ
of D onto Ω with ϕ(0) = 0 extends to be continuous on Ω, then we can apply the special
case just proved to u ◦ ϕ and v ◦ ϕ. 

Corollary 3.6. Let K be a compact subset of a region Ω and let z0 ∈ K. Then there
exists a constant C depending only on Ω and K such that if u is positive and harmonic
on Ω then for all z ∈ K
1
u(z0 ) ≤ u(z) ≤ Cu(z0 ). (3.3)
C
The point of Corollary 3.6 is that the constant C can be taken to be independent of
the function u.
 §4: Exercises 195

Proof. Suppose B is a disk in Ω. Let 12 B denote the disk with the same center as B but
1/2 the radius. Let ϕ be a linear map of D onto B, then by Harnack’s inequality applied
1
to u ◦ ϕ we have that (3.3) holds on 2
B with z0 replaced by the center of B and C = 3.
Cover K by finitely many disks of the form 12 Bj with Bj ⊂ Ω. Add more disks if necessary
1 1
so that the union of the disks is connected. If 2 B1 and 2 B2 are two disks with centers
1
z1 and z2 and if z3 ∈ B
2 1
∩ 21 B2 then 1
3
u(z2 ) ≤ u(z3 ) ≤ 3u(z1 ) so that u(z2 ) ≤ 9u(z1 ).
Since there are only finitely many disks and since their union is connected, we can find a
S
universal bound on the values at the centers and therefore on 12 Bj . 

We end this section with simple consequence, which could also be proved using normal
families.

Theorem 3.7 (Harnack’s Principle). Suppose {un } are harmonic on a region Ω such
that un (z) ≤ un+1 (z) for all z ∈ Ω. Then either
(i) lim un (z) ≡ u(z) exists and is harmonic on Ω, or
n→∞

(ii) lim un (z) = +∞

n→∞

where convergence is uniform on compact subsets of Ω. In case (ii), this means that given
K compact and M < ∞ there is an n0 < ∞ so that un (z) ≥ M for all n ≥ n0 .

Proof. By assumption, if n > m then un − um ≥ 0, and by the maximum principle

un − um is strictly positive or identically 0. Fix z0 ∈ K where K is compact. By Corollary
3.6 there is a constant C so that
1
(un (z0 ) − um (z0 )) ≤ un (z) − um (z) ≤ C(un (z0 ) − um (z0 )),
C
for all z ∈ K. Thus {un (z0 )} is Cauchy if and only if {un } is uniformly Cauchy on K. Thus
if the increasing sequence {un (z0 )} converges, then {un } converges uniformly on compact
subsets of Ω. Similarly if un (z0 ) → ∞ then un (z) → ∞ uniformly on compact subsets of
Ω. The limit function u is harmonic by the Mean Value Property (see Exercise 5). 

The assumption that the sequence un is increasing is essential. Note that one only
needs to prove the sequence is bounded at one point to prove uniform convergence on
compact subsets.
196 X. Harmonic Functions

§4. Exercises.
A

1. If f is analytic on a region Ω then log |f | is subharmonic on Ω. Moreover if B is a

disk with B ⊂ Ω, then the mean value of log |f | on ∂B is greater than or equal to the
value of log |f | at the center of B. Hint: log |x| is integrable.
2. Check that Schwarz’s Theorem 1.5 holds for any integrable function g at points of
continuity of g.
3. (a) Prove the following version of Green’s Theorem: Suppose that Ω is a bounded
region, bounded by finitely many C 2 simple closed curves. If u and v are twice
continuously differentiable on Ω then
Z Z
∂v ∂u
(u − v )|dz| = (u∆v − v∆u)dxdy,
∂Ω ∂η ∂η Ω

where η(ζ) is the unit normal at ζ ∈ ∂Ω pointing out of the region Ω, |dz| is arc length
measure on ∂Ω and dxdy is area measure on Ω.
(b) Prove that (ii) implies (i) in Theorem 2.3 by directly verifying that the Mean Value
Property holds. Hint: in D \ B(0, ε) set v = log |z|, apply (a) and let ε → 0.
If you already know a proof of some version of Green’s theorem, you may use it to
derive the version in (a).
4. Show that the one variable analogs of harmonic and subharmonic are linear and convex
in two ways: using the second derivative and using the maximum principle on intervals.
4.’ (a) Prove that in polar coordinates (r, θ)
     
2 ∂ ∂ ∂ ∂
r ∆u = r r u+ u.
∂r ∂r ∂θ ∂θ

(b) Prove that if u is harmonic in C \ {0} and if u depends only on r (not θ) then
u = a log r + b where a and b are constants.
5’. Suppose un are twice continuously differentiable in a region Ω and satisfy ∆un = 0 in
Ω. If un converges to a function u uniformly on compact subsets of Ω then ∆u = 0.
Proving directly that the limit function is differentiable may not be very easy. Limits
of integrals are easier. Prove fn = un + ie
un also converges uniformly on compact
 §4: Exercises 197

subsets of D, where u
en is the harmonic conjugate of un vanishing at some fixed point
z0 ∈ Ω.

5. Prove that the following are equivalent for a real-valued twice continuously differen-
tiable function v on a region Ω.
(i) v is subharmonic
(ii) ∆v ≥ 0
(iii) If B is a disk with B ⊂ B ⊂ Ω and if u is harmonic on B then v − u satisfies the
Maximum principle.
Hints: If ∆v < 0 in a ball then the mean value over a circle is a decreasing function
of the radius. If ∆v ≥ 0, then ∆(v + εx2 ) ≥ ε > 0.
6. Prove a “Schwarz’s Lemma” for harmonic functions: If u is real-valued and harmonic
in D with |u| ≤ 1 and u(0) = 0, then

4
|u(z)| ≤ arctan |z|.
π

When can equality hold?

7. If u(x, y) is harmonic in the unit disk D = {z = x + iy : |z| < 1}, define

z z
f (z) = 2u( , ) − u(0, 0).
2 2i

In other words formally replace the real variables x and y by the complex variables α
and β then set α = z/2 and β = z/2i. If u is expanded in a (double) series in powers
xn y m where will this series converge? Prove that f is well-defined (i.e. that this
substitution makes sense) and analytic in D and Ref = u. Note that this procedure
does not work for the functions u = log(x2 + y 2 ) and Re z1 which are harmonic in
D \ {0}. However, if u is harmonic in a neighborhood of (x0 , y0 ) then for z0 = x0 + iy0 ,

f (z) = 2u((z − z0 )/2 + x0 , (z − z0 )/(2i) + y0 ) − u(x0 , y0 )

satisfies Ref = u. Try this procedure for a few familiar functions such as ex cos y and
Re(z n ).
198 X. Harmonic Functions

8. Suppose u is harmonic in C and satisfies

|u(z)| ≤ M |z|k ,

when |z| > R. Show that u is the real part of a polynomial of degree at most k.

9. Suppose u is harmonic in C and

M (r)
lim inf ≤ 0,
r→∞ log r

where M (r) = sup|z|=r u(z). Prove u is constant.

10. Suppose v is twice continuously differentiable and subharmonic in C. Prove that

M1 (r)
lim
r→∞ log r

exists (possibly infinite), where M1 (r) is the mean value of v on |z| = r.

C
11. Suppose u is harmonic in C and satisfies

u(z) ≤ A|z| + B

where A and B are constants. Prove u = ax + by + c where a, b, and c are constant. Note
that the assumption is about u not |u|.
 XI

Conformal Maps and Harmonic Functions

§1. Symmetry and Conformal Maps.

In this section we will explore the connection between symmetry in a simply connected
domain and symmetry of a conformal map from the region to the disk D.

We say that a region Ω is symmetric about R provided z ∈ Ω if and only if z ∈ Ω.

If Ω is symmetric about R and if f is analytic on Ω then f (z) is analytic on Ω. Indeed, the
power series expansion for f (z) based at b has coefficients which are the complex conjugates
of the coefficients in the power series for f at based at b. If f is also real-valued on an
interval in Ω ∩ R, then
f (z) ≡ f (z) (1.1)

by the Uniqueness Theorem II.3.2. Geometrically, (1.1) says that f maps symmetric points
to symmetric points. For one-to-one functions we have the following related result.

Proposition 1.1. If f is a conformal map of a simply connected symmetric region Ω onto

D such that f (x0 ) = 0 and f ′ (x0 ) > 0 for some x0 ∈ R ∩ Ω then f (z) = f (z).

Proof. Apply the uniqueness conclusion in the Riemann Mapping Theorem to the func-
tions f and f (z). 

Proposition 1.1 says that if the region is symmetric then we can take the mapping
function to be symmetric. We’d like to use this idea to help construct conformal maps.

Theorem 1.2 (Schwarz Reflection Principle). Suppose Ω is a region which is sym-

metric about R. Set Ω+ = Ω ∩ H and Ω− = Ω ∩ (C \ H). If v is harmonic on Ω+ , continuous
on Ω+ ∪ (Ω ∩ R) and equal to 0 on Ω ∩ R then the function defined by

 v(z) for z ∈ Ω \ Ω−
V (z) =

−v(z) for z ∈ Ω−

199
200 XI. Conformal Maps and Harmonic Functions

is harmonic on Ω. If also v(z) = Imf (z) where f is analytic on H ∩ Ω then the function

 f (z) for z ∈ Ω+
g(z) =

f (z) for z ∈ Ω−

extends to be analytic in Ω.

Note that there is no assumption about boundary values of the real part of f on Ω∩R.

Proof. The extended function V is continuous on Ω. To prove the first claim, we need
only prove that V satisfies the Mean Value Property for small circles centered on Ω ∩ R.
But for x0 ∈ R ∩ Ω, V (x0 + reit ) = −V (x0 + re−it ) so that the mean value over a circle
centered at x0 contained in Ω is zero, the value of V at x0 . Thus V is harmonic in Ω. If D
is a disk contained in Ω and centered on R, then V = Imh for some analytic function h on
D. It is uniquely defined by requiring that h = f on Ω+ ∩ D. By the symmetry discussion
before the statement of Theorem 1.2, h(z) = h(z) on D. Thus h = g on D ∩ Ω \ R and so
h provides the unique analytic extension of g to all of D. Thus g extends to be analytic
on Ω. 

Corollary 1.3. If the function f in the Schwarz Reflection Principle is one-to-one in Ω+

with Imf > 0 on Ω+ , then g is also one-to-one on Ω.

Proof. By definition h is one-to-one on Ω− with Imh < 0 on Ω− . So if h(z1 ) = h(z2 )

then z1 , z2 ∈ R. Since h is open, it maps a small disk centered at zj onto a neighborhood
of h(zj ) for j = 1, 2. Thus there are two points ζ1 , ζ2 ∈ Ω+ near z1 , z2 (respectively) with
h(ζ1 ) = h(ζ2 ), contradicting the assumption that f is one-to-one on Ω+ . 

As an application we outline how to find a conformal map of H to the region Ω = {x +

iy : y > x2 }. First find a map f1 of H onto the half-strip S = {z = x+iy : 0 < x < b, y > 0}
so that f1 (∞) = ∞ (i.e. lim|z|→∞ |f1 (z)| = ∞), where b is chosen later. Then the function
f2 (z) = z 2 is one-to-one on S with image equal to half of the interior region of a parabola.
The composed function f12 maps an interval Ia = (−∞, a] onto Ib2 = (−∞, b2 ] and [a, +∞)
 §1: Symmetry and Conformal Maps. 201

onto the half parabola. By the Schwarz Reflection Principle we can extend this function by
reflecting the domain across the interval Ia and the range across Ib2 to obtain a conformal
map of C \ [a, ∞) onto the interior of a parabola. The slit plane can be mapped onto H
√
using z − a and the parabola can be translated and rotated. With the proper choice of
b we obtain the desired region.
This same idea can be used to construct a conformal map from H to the region outside
one branch of a hyperbola. We now have the tools to map any region bounded by a conic
section onto D, with the exception of the interior of an ellipse. We will learn how to do
this latter problem after learning how to map the disk onto a rectangle in Section XI.3.
It is sometimes useful to be able to reflect across curves more general than line seg-
ments.

Definition 1.4. An (open) simple arc γ is called analytic if there exists a function g which
is one-to-one and analytic in a neighborhood N of the interval (0, 1) with g((0, 1)) = γ.
A simple closed curve σ is called analytic if there exists a one-to-one analytic function g
defined in a neighborhood of ∂D with g(∂D) = σ.

Defintion 1.5. An open analytic arc γ contained in the boundary of a region Ω is called
one-sided if there exists a function g which is one-to-one and analytic in a neighborhood
N of (0, 1) with g((0, 1)) = γ and g(N ∩ H) ⊂ Ω and g(N \ H) ⊂ Ωc .

Corollary 1.6. Suppose Ω is a simply connected region and suppose γ ⊂ ∂Ω is a one-

sided analytic arc. Let f be a Riemann map of Ω onto D. Then f extends to be analytic
and one-to-one is a neighborhood of Ω ∪ γ.

Proof. By Exercise III.9, |f (z)| → 1 as z → ∂Ω. Choose g and N as in Definition 1.5.

Then f ◦ g is analytic in N ∩ H. By shrinking N if necessary, we may suppose N ∩ H is
simply connected and f ◦ g 6= 0 on N . We can then define h(z) = −i log f ◦ g(z) to be
analytic in N ∩ H and

lim Imh(z) = lim − log |f ◦ g(z)| = 0.

z∈N→(0,1) z∈N→(0,1)

By Corollary 1.3, h extends to be one-to-one and analytic in N . Therefore f extends to

be one-to-one and analytic in Ω ∪ g(N ). 
202 XI. Conformal Maps and Harmonic Functions

More generally, if f is analytic in a region Ω1 whose boundary contains a one-sided

analytic arc γ and if f (z) tends to an analytic arc σ as z → γ, then f extends to be analytic
in a neighborhood of Ω ∪ γ. See Exercise 2. Care must be taken if γ is a “two-sided” arc
√
in ∂Ω. For example if Ω = C \ I, where I = (−∞, 0] then the function z defined in Ω
has an analytic extension across I from the upper half-plane, and across I from the lower
half plane, but these extensions doe not agree with the original function.
In Section IV.6 we constructed a conformal map of the upper half plane slit along
a circular arc onto the upper half plane. Using the material on harmonic functions in
Chapter XI, together with the Schwarz Reflection Principle, we can construct a conformal
map f of H onto Ω = H \ I where I is a straight line segment in the upper half plane
meeting R at 0 with angle πa. See Figure XI.1.
f
H H\I
I
πa
J 0
Figure XI.1

The Riemann mapping theorem says that f exists. We first use the Reflection Principle
to deduce boundary behavior of f . If C is a linear fractional transformation of H onto D
then ∂D \ {C(0)} is an analytic arc so that C ◦ f −1 ◦ C −1 extends by Corollary 1.6 to be
analytic and one-to-one across ∂D \ {C(0)}. The image of ∂D \ {C(0)} by the extended
map is a subarc of ∂D since it is connected. Thus we can choose f so that f −1 (z) → ∞ if
and only if z ∈ Ω → ∞ and so f −1 will be an analytic map of C \ (I ∪ I R ) onto C \ J where
J ⊂ R is an interval and I R is the reflection of I about the real line R. By composing with
a linear map, we may suppose that J = [0, 1]. Thus f extends to be a conformal map of
C \ J onto C \ (I ∪ I R ) and hence f (z) → I as z ∈ H → J. Then log f (z) is a one-to-one
analytic map of H onto a slit strip S = {z : 0 < Imz < π} \ {z : Imz = πa, Rez < x0 }.
This implies that u(z) = Im log f (z) is a harmonic function which extends continuously
to (−∞, 0) with boundary value π, continuously to (0, 1) with boundary value πa and
continuously to (1, ∞) with boundary value 0. It is also bounded, so that by Lindelöf’s
maximum principle it equals a arg(z − 1) + (1 − a) arg z, which has the same properties.
 §2: The Geodesic Algorithm. 203

This implies
f (z) = Cz 1−a (z − 1)a ,

where C is a positive constant chosen so that the length of the image segment I is cor-
rect. The preimage of the tip of the slit can be calculated by setting f ′ /f = 0. Another
representation of this map, found by pre-composing with the linear map z/C + a and
post-composing with the linear map Cz is

1−a a C 2 a(1 − a) 1 1
f (z) = (z + Ca) (z + C(a − 1)) = z − + O( 2 ).
2 z z
These maps are very important for a recent application of conformal mapping and
probability to statistical mechanics, called SLE and are the building blocks for solutions
to Löwner’s differential equation (Seee [XXXX]).

§2. The Geodesic Algorithm.

Conformal maps can be used to transfer problems on a complicated region to problems

on a simpler region, and using the inverse map transfer the solutions on the simpler region
back to the original region. For example, suppose Ω is a simply connected region such that
a conformal map ϕ of Ω onto D extends to be a one-to-one and continuous map of Ω onto
D. If f is a continuous function on ∂Ω, then we can find a harmonic function u on Ω which
extends to be continuous on Ω and equal to f on ∂Ω. To accomplish this, let g = f ◦ ϕ−1
on ∂D and let v be the Poisson integral of g. Then u = v ◦ ϕ is the desired solution.
How do you find a conformal map of D or H to a complicated region? Rather few
maps can be given explicitly by hand, so that a computer must be used to find the map
approximately. One reasonable way to describe a region numerically is to give a large
number of points on the boundary. One way to say that a computed map defined on H is
“close” to a map to the region is to require that the boundary of the image be uniformly
close to the polygonal curve through the data points. Indeed, the only information we may
have about the boundary of a region are these data points. In this section we will give
an algorithm for computing a conformal map of H onto a simply connected region whose
boundary contains a given set of boudary points and is analytic between the boundary
points. We will use harmonic functions to bound the curve between the data points.
204 XI. Conformal Maps and Harmonic Functions

The hyperbolic geometry on a simply connected region Ω is given by transplanting

the hyperbolic geometry from the unit disk to the region via a Riemann map f . The
shortest curve between two points z1 , z2 in the hyperbolic metric on a simply connected
region Ω is found by choosing the map f : Ω → D so that f (z1 ) = 0 and f (z2 ) = r > 0.
Then the shortest curve is f −1 ([0, r]). For that reason, the images of diameters of D by
conformal maps of D onto Ω are called geodesics. We can create a curve in C passing
through a given collection of points {zj }n0 by first taking a line segment from z0 to z1 to
form a region C∗ \ γ1 . By induction γj+1 is a curve from z0 to zj+1 found by taking γj
followed by the unique geodesic in the hyperbolic geometry of Ωj from zj ∈ ∂Ωj to zj+1 .
If we set zn+1 = z0 and let γ be γn followed by the geodesic between zn and zn+1 then γ
is a simple closed curve through the points {zj } provided the points are not in the wrong
order, in the sense that zj+1 does not belong to γj for j = 1, . . . , n.
The conformal map of D onto Ω0 can be constructed using the techniques of section
VI.6. Once we have constructed a conformal map fj of D onto Ωj , we can normalize it so
that −1 is mapped to zj and 0 is mapped to zj+1 . Then (−1, 0) is mapped to the geodesic
from zj to zj+1 . Composing this map with a map of D onto D \ (−1, 0] gives a conformal
map of D onto Ωj+1 . For the last map of D onto Ωn+1 we can arrange that −1 and 1 are
mapped to zn and zn+1 = z0 , respectively, by fn . Then fn restricted to a half disk maps
onto Ωn+1 . Composing with a map of D onto a half disk gives a conformal map f of D
onto a region Ωn+1 bounded by a simple closed curve passing through the points {zj }n0 .

We can carryout this idea using the upper half plane instead of the disk and thereby
avoid the repeated renormalization. The geodesic algorithm is based on the simple map
fa : H \ γ −→ H where γ is an arc of the (unique) circle from 0 to a ∈ H which is
orthogonal to R at 0. This map can be realized by a composition of a linear fractional
transformation, the square and the square root map as illustrated in Figure XI.2. See also
Figure VI.10. The orthogonal circle also meets R orthogonally at a point b = |a|2 /Rea
illustrated as the endpoint of the dotted circular arc in Figure XI.2. Most computers define
√ √
z for −π < arg z ≤ π, which is why we labelled the square root map as i −z. One way
to verify that z/(1 − z/b) maps H onto H is to note that it is real-valued on R and has
 §2: The Geodesic Algorithm. 205

derivative 1 at 0. Since the map is one-to-one and maps generalized disks onto generalized
disks, it must map H onto H.
fa
H\γ H

a
γ
b
0 −c 0 c
z
1 − z/b √
i −z
z 2 + c2

ic

0 0 c2
Figure XI.2. The basic map fa .

In Figure XI.2, c = |a|2 /Ima. Observe that the arc γ is opened to two adjacent intervals
at 0 with a, the tip of γ, mapped to 0. The inverse fa−1 can be easily found by composing
the inverses of these elementary maps in the reverse order.

Now suppose that z0 , z1 , . . . , zn are points in the plane. The basic maps fa can be
used to compute a conformal map of H onto a region Ωc bounded by a Jordan curve which
passes through the data points as illustrated in Figure XI.3.
206 XI. Conformal Maps and Harmonic Functions

Ωc
z1 ζ4
zn z3
z0 z2
0
p
ϕ1 = i (z − z1 )/(z − z0 )

ϕ3 = fζ3 ϕn = fζn

ζ2

0 ζn+1 0
 2
ϕ2 = fζ2 z
ϕn+1 =−
1 − z/ζn+1

H
ζ3
0 0
Figure XI.3. The Geodesic Algorithm.

The complement in the extended plane of the line segment from z0 to z1 can be
mapped onto H with the map r
z − z1
ϕ1 (z) = i
z − z0
and ϕ1 (z1 ) = 0 and ϕ1 (z0 ) = ∞. Set ζ2 = ϕ1 (z2 ) and ϕ2 = fζ2 . Repeating this process,
define
ζk = ϕk−1 ◦ ϕk−2 ◦ . . . ◦ ϕ1 (zk )

and
ϕk = fζk .

for k = 2, . . . , n. Finally, map a half-disc to H by letting

ζn+1 = ϕn ◦ . . . ◦ ϕ1 (z0 ) ∈ R
 §2: The Geodesic Algorithm. 207

be the image of z0 and set

 2
z
ϕn+1 =±
1 − z/ζn+1

The + sign is chosen in the definition of ϕn+1 if the data points have negative winding
number (clockwise) around an interior point of ∂Ω, and otherwise the − sign is chosen.
Set

ϕ = ϕn+1 ◦ ϕn ◦ . . . ◦ ϕ2 ◦ ϕ1

and

ϕ−1 = ϕ−1 −1 −1
1 ◦ ϕ2 ◦ . . . ◦ ϕn+1 .

Then ϕ−1 is a conformal map of H onto a region Ωc such that zj ∈ ∂Ωc , j = 0, . . . , n.

The portion γj of ∂Ωc between zj and zj+1 is the image of the arc of a circle in the upper
half plane by the analytic map ϕ−1 −1 −1
1 ◦ . . . ◦ ϕj , and hence is an analytic arc. Because fa

doubles angles at 0 and halves angles at ±c the boundary curve ∂Ωc has a continuously
turning tangent direction. In more picturesque language, after applying ϕ1 , we grab the
ends of the displayed horizontal line segment and pull, splitting apart or unzipping the
curve at 0. The remaining data points move down until they hit 0 and then each splits
into two points, one on each side of 0, moving further apart as we continue to pull.
Note that ϕ is a conformal map of the complement of Ωc , C∗ \ Ωc , onto the lower half
plane, C \ H where C∗ denotes the extended plane. Simply follow the unshaded region in
H in Figure XI.3.
The geodesic algorithm can be applied to any sequence of data points z0 , z1 , . . . , zn ,
unless the points are out of order in the sense that a data point zj belongs to the (computed)
arc from zk−1 to zk , for some k < j. We will next give a simple condition on the data
points z0 , z1 , . . . , zn which is sufficient to guarantee that the curve computed by the geodesic
algorithm is close to the polygon with vertices {zj }.

Definition 3.1. A disc-chain D0 , D1 , . . . , Dn is a sequence of pairwise disjoint open

discs such that ∂Dj is tangent to ∂Dj+1 , for j = 0, . . . , n − 1. A closed disc-chain is a
disc-chain such that ∂Dn is tangent to ∂D0 .
208 XI. Conformal Maps and Harmonic Functions

Any simple closed polygon P , for example, can be covered by a closed disc-chain with
arbitrarily small radii and centers on P . There are several ways to accomplish this, but
one straightforward method is the following: Given ε > 0, find pairwise disjoint discs {Bj }
centered at each vertex, and of radius less than ε. Then

[ [
P\ Bj = Lk
j

where {Lk } are pairwise disjoint closed line segments. Cover each Lk with a disc-chain
centered on Lk tangent to the corresponding Bj at the ends, and radius less than half the
distance to any other Li , and less than ε.

Figure XI.4. Disc-chain covering a polygon.

Another method for constructing a disc-chain is to use a Whitney decomposition

of a simply connected domain. Suppose Ω is a simply connected domain contained in
the unit square. The square is subdivided into 4 equal squares. Each of these squares is
subdivided again into 4 equal squares, and the process is repeated. If Q is a square, let
2Q denote the square with the same center, and sides twice as long. In the subdivision
process, if a square Q satisfies 2Q ⊂ Ω, then no further subdivisions are made in Q. Let
Un be the union of all squares Q obtained by this process with side length at least 2−n
for which 2Q ⊂ Ω. If z0 ∈ Ω, let Ωn be the component of the interior of Un containing
z0 . Then ∂Ωn is a polygonal Jordan curve. Note that ∂Ωn consists of sides of squares Q
with length 2−n . Thus we can form a disc chain by placing a disc of radius 2−n /2 at each
vertex of ∂Ωn . The points of tangency are the midpoints of each square with edge length
2−n on ∂Ωn .
 §2: The Geodesic Algorithm. 209

Yet another method for constructing a disc-chain would be to start with a hexagonal
grid of tangent discs, all of the same size, then select a sequence of these discs which form
a disc-chain. Disc-chains can be used to approximate the boundary of an arbitrary simply
connected domain.

Lemma 3.3 (Jørgensen). If ∆ is an open disc contained in a simply connected domain

Ω and if J is a geodesic, then J ∩ ∆ is connected.

Proof. Without loss of generality, Ω is bounded by a simple closed curve and ∆ ⊂ Ω. Let
f be a conformal map of Ω onto H such that f (J) is the positive imaginary axis which we
denote by I. If Jc is the subinterval of the imaginary axis from 0 to ic, then the conformal
√
map τ (z) = z 2 + c2 of H \ J onto H maps I \ Jc onto I. Replacing Ω with f −1 (H \ Jc ),
and replacing f with τ ◦ f , we may suppose that f −1 (iy) → z1 ∈ ∂Ω ∩ ∂∆ as y → 0.
Similarly we may suppose that f −1 (iy) → z2 ∈ ∂Ω ∩ ∂∆ as y → +∞. This corresponds
to cutting the region Ω along the geodesic J from ∂Ω to z1 and from ∂Ω to z2 . The point
is that the rest of J is still a geodesic in the cut region. The points zj divide ∂∆ into
two arcs α and ∂∆ \ α. Then f (α) and f (∂∆ \ α) are arcs in H connecting 0 to ∞. We
may suppose that f (α) lies to the left of f (∂∆ \ α). Let U be the component of H \ f (α)
containing f (∆).

f f (∆)
z2
∂∆ \ α f (∂∆ \ α)
α ∆ f (α)
z1
Ω

Figure XI.5. Proof of Jørgensen’s theorem.

210 XI. Conformal Maps and Harmonic Functions

We first construct an auxilliary function ω defined on the complement of α in the

extended plane. Let ω be the bounded harmonic function defined on C \ α with boundary
values 1 on the “inside” of α and 0 on the “outside” of α. More precisely

ω(z) → 1 as z ∈ f −1 (U ) → α◦

and
ω(z) → 0 as z ∈ Ω \ f −1 (U ) → α◦ .

The function ω can be found explicitly by mapping C∗ \ α onto C \ [0, +∞) via a linear
1
fractional transformation, then composing with arg z/(2π). Note that ω = 2
on (∂∆ \ α)◦ .
By comparison of boundary values and the Lindelöf Maximum Principle, arg z <
1
πω(f −1 (z)) for all z ∈ U . Since ω = 2 on (∂∆ \ α)◦ , we conclude

π
arg z <
2

on f (∂∆ \ α) ∩ H. Switching the definition of α and ∂∆ \ α then repeating the same

argument shows that
π
π − arg z <
2
on f (α) ∩ H. Thus f (∆) ⊃ I. 

Note that the proof of Jörgensen’s Lemma shows that if a geodesic J intersects ∂∆ at
two points z1 , z2 then the portion of J between z1 and z2 is contained in the interior ∆.

One interpretation of Jørgensen’s Lemma is that disks are convex in the metric on a
simply connected domain given by the geodesics.

If D0 , D1 , . . . , Dn is a closed disc-chain, set

zj = ∂Dj ∩ ∂Dj+1 ,

for j = 0, . . . , n, where Dn+1 ≡ D0 .

 §3: Conformal Maps to Polygonal Regions 211

Theorem 3.4. If D0 , D1 , . . . , Dn is a closed disc-chain, then the geodesic algorithm ap-

plied to the data z0 , z1 , . . . , zn produces a conformal map ϕ−1
c from the upper half plane
H to a region bounded by a C 1 curve γ with
n
[
γ⊂ (Dj ∪ zj ).
0

Proof. An arc of a circle which is orthogonal to R is a hyperbolic geodesic in the upper half
plane H. Let γj denote the portion of the computed boundary, ∂Ωc , between zj and zj+1 .
Since hyperbolic geodesics are preserved by conformal maps, γj is a hyperbolic geodesic in

j−1
C∗ \ ∪k=0 γk .

Theorem 3.4 then follows from Jörgensen’s Lemma by induction. 

Figure XI.6. Tenerife.

Figure XI.6 shows the conformal map of a grid on the disc to both the interior and
exterior of the island Tenerife (Canary Islands). The center of the interior is the volcano
Teide.
212 XI. Conformal Maps and Harmonic Functions

§3. Conformal Maps to Polygonal Regions.

In this section we will use the Schwarz Reflection Principle to find conformal maps
from the upper half plane onto regions bounded by polygons, called the Schwarz-Christoffel
formula or integral.
Suppose Ω is a bounded simply connected region bounded by a simple closed curve
consisting of finitely many straight line segments with vertices {vj }n1 . If we give ∂Ω a
positive (counter-clockwise) orientation, then at vj the tangent vector turns by an angle
παj , where −1 < αj < 1. By the Schwarz Reflection principle (Corollary 1.6), if ϕ is
a conformal map of Ω onto D, then ϕ extends analytically and one-to-one across (the
interior of) each boundary segment. If Bj is a small ball centered at vj , then the map
(z − vj )1/(1−αj ) is one-to-one and analytic in Ω ∩ Bj , and maps ∂Ω ∩ Bj onto a straight
line segment. The inverse of this map composed with ϕ then extends to be analytic and
one-to-one in a neighborhood of 0 again by the Schwarz Reflection Principle. Thus ϕ
extends to be a one-to-one and continuous map of Ω onto D. So if f (z) is a conformal
map of H onto Ω then f (z) extends to be one-to-one and continuous on H and analytic
on R except at xj = f −1 (vj ), j = 1, . . . , n. We may assume that the vertices are ordered
f (x+h)−f (x)
so that −∞ < x1 < x2 < . . . < xn < ∞. Writing f ′ (x) = limh→0 h , we deduce
that f ′ (x) points in the tangential direction to ∂Ω. In other words for xj < x < xj+1 ,
arg f ′ (x) is given by the direction of the line segment from vj to vj+1 . Since f ′ 6= 0 on the
simply connected region H we can define log f ′ (z) so as to be analytic on H, and hence
arg f ′ (z) is a bounded harmonic function on H which is continuous on H \ {xj }n1 . The
function π − arg(z − a), for a ∈ R, equals 0 for z < a and equals π for z > a, z ∈ R. By
Lindelöf’s Maximum Principle
n
X
′
arg f (z) = c0 + αj (π − arg(z − xj )),
j=1

and so
n
Y
′
f (z) = C (z − xj )−αj .
j=1

We have proved:
 §3: Conformal Maps to Polygonal Regions 213

Theorem 3.1 (Schwarz-Christoffel). Suppose Ω is bounded and simply connected,

and suppose ∂Ω is a polygon with vertices v1 , ..., vn whose tangent direction increases by
παj at vj . Then there exists x1 < x2 < . . . < xn and constants c1 , c2 so that
Z n
z Y
f (z) = c1 (ζ − xj )−αj dζ + c2
i j=1

is a conformal map of H onto Ω, where the integral is along any curve in H from i to z.

The difficulty in using the Schwarz-Christoffel formula is that it does not tell us how
to find the “prevertices” xj . In the formulation above −1 < αj < 1 and the tangent
Pn
direction has a total increase of 2π around ∂Ω, so that j=1 αj = 2. Thus the integral is
absolutely convergent, even if the integration takes place on R instead of H. The length of
the segment from vj to vj+1 can be expressed as:
Z xj+1 n
Y
Lj = |C| |x − xj |−αj dx.
xj j=1

So the prevertices can be written as a solution of a system of (highly non-linear) singular

integrals. For numerical applications, because the derivative is singular at some of the
prevertices, Riemann sums are too inaccurate to compute the integrals, so that Gauss-
Jacobi quadrature is needed.
The Schwarz-Christoffel Theorem can be extended to cover unbounded regions whose
boundary consisting of line segments and half lines or full lines. Allowing an angle αj to
equal π corresponds to an infinite channel. The case when the angles παj are allowed
to equal −π, corresponds to a “cut” along a line segment into a polygonal region. See
Exercises 4 and 14.
We’ll end this chapter with one of the more important open problems in complex
analysis called the Brennan Conjecture. Suppose that T is a “polygonal tree”, a con-
nected finite union of line segments and the half line [0, ∞], whose complement is a simply
connected region Ω. The conformal map ψ of the upper half plane H onto Ω can be nor-
malized so that it is asymptotic to z 2 for |z| large. Thus ψ ′ (z) is asymptotic to 2z near ∞.
Indeed, the total turning of the tangent direction following a positive orientation of ∂Ω is
214 XI. Conformal Maps and Harmonic Functions
P
−π, so that αj = −1, where παj is the change in the tangent direction at ψ(xj ). By
the Schwarz-Christoffel Theorem

2
ψ ′ (z) = Q .
(z − xj )αj

The Brennan conjecture is about the second derivative of ψ at the “tips” of the tree: Prove
that for every polygonal tree

X Y
|xk − xj |2αj ≤ 1.
k:αk =−1 j6=k

This is equivalent to proving that for any conformal map f of the disk into the plane
1/f ′ ∈ Lp (dxdy) for all p < 2. The conformal map f (z) = (1 − z)2 defined on D has
derivative whose reciprocal is not in L2 .

§4. Exercises.

1. Let f be an analytic function on |z| < 2 such that f is real-valued on a subarc of

|z| = 1. Prove that f is a constant.
2. (a) Suppose Ω is a simply connected region and suppose ∂Ω is a closed analytic curve.
Prove that if f is a Riemann map of Ω onto D then f and f −1 extend to be one-to-one
and analytic in a neighborhood of Ω, respectively D.
(b) Using the Cayley transform, show that reflection z → z across R corresponds to
“reflection” z → 1/z across ∂D.
(c) Suppose f is analytic in D and |f (z)| → 1 as z → I, where I ⊂ ∂D is an arc.
Show that f extends to be analytic in a neighborhood of D ∪ I using 1/f (1/z). The
extension will have poles at the reflection of the zeros of f .
(d) Prove that if f is analytic in a region Ω1 whose boundary contains a one-sided
analytic arc γ and if f (z) tends to an analytic arc σ as z → γ, then f extends to be
analytic in a neighborhood of Ω ∪ γ.
 §4: Exercises 215

3. Find a conformal map of H onto {x + iy : y > x2 } so that the origin is mapped to the
focus. See the outline in the text.
4. Show that the Schwarz-Christoffel formula still holds if angles παj are allowed to equal
π.
5. A conformal map of a rectangle onto the upper half plane is an example of an elliptic
function. If f maps a rectangle of width R and height 1 onto the upper half plane,
then it can be reflected across a vertical edge, to obtain a map from a rectangle twice
as long onto the slit plane. Prove that by repeated reflections we can extend the map
f to be a meromorphic function on C with the property that f (z + 2R) = f (z) and
f (z + i) = f (z). In general, an elliptic function is a meromorphic function invariant
under two linearly independent shifts. The inverse of an elliptic function, such as the
Schwarz-Christoffel integral for a rectangle map, is an example of an elliptic integral.

6. Find a conformal map of H onto the region {x + iy : x2 − y 2 /4 > 1 and x > 0}.
7. Given a continuous function f (t), 0 ≤ t ≤ π, find a harmonic function u on D+ =
D ∩ {Imz > 0} such that u is continuous on D+ , u = f on ∂D+ ∩ {Imz > 0} and
∂u
∂y = 0 on (−1, 1). Give an explicit formula.
R 2π
8. Suppose g is continuous on ∂D and 0 gdθ = 0. Find a harmonic function on D with
∂u
the property that ∂η = g on ∂D where η is the unit inward normal to ∂D. Give an
explict formula. This is called a Neumann problem. (perhaps next time do it on
the upper half plane??)
9. Given continuous functions f (t), 0 ≤ t ≤ π and g(x), −1 < x < 1 find a harmonic
function u on D+ = D ∩ {Imz > 0} such that u is continuous on D+ \ {±1}, u = f on
∂u
∂D+ ∩ {Imz > 0} and ∂y = g on (−1, 1). Give an explicit formula.
10. Suppose ϕ is analytic and one-to-one on a neighborhood of D and Ω = ϕ(D). Suppose
E is a connected subset of ∂Ω. Given continuous functions f on E and g on ∂Ω \ E,
∂u
describe how to find a harmonic function u on Ω so that u = f on E ◦ and ∂η = g on
(∂Ω \ E)◦ . This is called a mixed Dirichlet-Neumann problem.
216 XI. Conformal Maps and Harmonic Functions

11. The intersection of two disks is called a lens. Suppose zi , i = 1, . . . , n are distinct
points in C. Suppose Di+ and Di− are (open) disks whose intersection is a lens Li =
Di+ ∩ Di− with endpoints zi and zi+1 . If Di+ ∪ Di− ∩ Lj = ∅ for j = 1, . . . , i − 1
and if ∂Di+ is tangent to ∂Di−1
−
and ∂Di− is tangent to ∂Di−1
+
then we say that
{Li } is a tangential lens-chain. Prove that if {Li } is a tangential lens chain then
the geodesic algorithm constructs a conformal coformal map of the half plane onto a
simply connected region whose boundary is contained in ∪Li . This idea can be used
to reduce the number of data points zi in the geodesic algorithm by using long thin
lenses where the boundary is flat.
12. Find a conformal map of D onto the interior of an ellipse. Hint: consider the map
1
2 (z + z1 ) on the top half of an annulus.
13. Find a conformal map of D onto a regular n-gon.
14. Give a version of Schwarz-Christoffel for unbounded regions.

15. (a) Find a conformal map ϕ of the unit disk onto a rectangle of height 1 and length
R with the property that the origin is mapped from the conformal map ψ of the unit
disk to a rectangle of height 1 and length 2R. Normalize both ϕ and ψ so that the
origin is mapped to the center of the corresponding rectangle.
(b) Find an approximation to the conformal map ϕ by first mapping the disk onto
a half strip. This will map a subregion of the disk onto a rectangle of length 2n R.
Repeat the process above n times to obtain an approximation of ϕ defined on D.
(c) Find the explicit map constructed in part (b) when n = 3, where ϕ is normalized
to map the origin to the center of the rectangle of length R.
(d) Find an approximation to the inverse ϕ−1 .
Hint: Keep track of the points on the unit circle corresponding to the 4 corners of the
corresponding rectangle. When n ≥ 3, these approximations are highly accurate.
 XII

Conformal Maps to Jordan Regions

A Jordan Curve is the homeomorphic image of the unit circle. It may seem obvious,
but is by no means simple to prove, that a Jordan Curve divides the plane into exactly
two regions. We begin the chapter with some of examples of strange Jordan Curves. Then
we use complex analysis to prove a topological lemma due to Janiszewski, which is quite
useful for tackling subtle topological problems in the plane. We use this lemma many
times in Section 3 to prove the Jordan Curve Theorem. Our proof follows the exposition
in Pommerenke, Ch., Univalent Functions. In Section 4 we prove the useful and important
theorem, due to Carathéodory, that a conformal map of the unit disk onto a Jordan region
extends to be a homeomorphism of the closed disk onto the closure of the region. The
proof follows the exposition in Garnett and Marshall, Harmonic Measure.

§1. Badly Behaved Jordan Curves

Example 1.

Our first example is a Jordan Curve with positive area. First we construct a totally
disconnected compact set E of positive area, then we describe how to pass a curve through
this set.
The compact set E will be the intersection of compact sets Ej , j = 0, 1, 2, . . .. The
initial set E0 is the closed unit square E0 = [0, 1] × [0, 1]. Choose σ1 < 1 and let E1 be the
union of four squares of side length σ1 /2 located in each of the corners of E0 . For example,
the lower left square is [0, σ1 /2] × [0, σ1 /2]. The area of E1 is σ12 . Next choose σ2 < 1
and let E2 be the union of 16 squares, one in each of the corners of E1 with side length
σ1 σ2 /4 and total area (σ1 σ2 )2 . Repeating this process we obtain compact sets En ⊃ En−1
Qn Qn
consisting of 4n squares with side length j=1 (σj /2) and area j=1 σj2 . Choose σj so that
Q∞ 2
j=1 σj > 0. For instance we can take σj = 1 − 1/j . Set

∞
\
E= En .
n=0

217
218 XII. Conformal Maps to Jordan Regions

Then E is a compact set of positive area. See Figure XII.1.

The Jordan curve J is constructed in stages also. A Jordan arc is the homeomorphic
image of the unit interval [0, 1]. First we construct a Jordan arc containing the set E,
then we close it up to form a Jordan curve. Connect the four squares in E1 using 3 line
segments which leave the lower right corner of each square and connect to the upper left
corner of an adjacent square as shown in Figure XII.2. Inside each square in E1 we insert
3 line segments connecting the four squares in E2 using the same pattern. Repeat this
process for each square in each En .

We construct a mapping of a dense subset of the unit interval onto the union of these
line segments. Let I1 consist of three disjoint closed intervals contained in (0, 1) with total
length 1/2. Let I2 consist of 3 disjoint closed intervals in each of the four intervals of
[0, 1] \ J1 . Repeat this process so that In consists of 3 · 4n disjoint intervals contained in
[0, 1] \ In−1 of total length 2−n . Construct a map ϕ mapping each interval in Jn linearly
P∞ −n
onto one of the segments connecting a square in En . Since ℓ(∪∞1 Jn ) = 1 2 = 1 the
map ϕ extends to be a continuous and one-to-one map of the unit interval to a curve which
contains the set E of positive area. If we then connect the upper left corner of E0 to the
lower right corner of E0 (in C \ E0 ) we obtain a (closed) Jordan curve J with positive area.

Example 2.

Another example Jordan arc J is found by spiraling in to 0 in a counter-clockwise

direction then spiraling back out again with a Jordan arc that has finite length. See
Figure XII.3. Note that arg z is neither bounded above nor bounded below on J. Starting
with the unit circle J0 , remove a countable dense subset of subarcs of J0 with small total
length and replace each with a scaled copy of J to form a new Jordan arc J1 with the
property that J1 has finite length, in fact we can arrange that J1 has length as close to
the length of J0 as we like. Repeat this process on J1 , removing much smaller intervals to
obtain a Jordan curve J2 . We can repeat the process indefinitely to give a limit Jordan
curve J. The curve J can have finite length and there is a dense subset of points {an } ⊂ J
such that arg(z − an ) is neither bounded above nor bounded below for z ∈ J.
 §2: Janiszewski’s Lemma. 219

Example 3.

For the final example we create a Jordan arc that looks like an interval, even if we
use arbitrarily high magnification to look at it. Begin with a polygonal arc with vertices
1, a, b, −b, −a and −1, where 0 < a < 1 and 0 < Reb < a and Imb > 0. We can choose
a and b so that the slope of every segment is less than a given ε > 0. We can choose ε
less than the height of a pixel (dot) on a computer screen, so that this polygon cannot be
distinguished from a straight line on a computer. If we zoom in, or magnify, the picture on
a computer, it will always look like a straight line. Now replace each line segment with a
scaled, translated and rotated version of the same polygonal arc with the same endpoints
as the segment. Repeat this operation indefinitely. The resulting curve will have arbitrarily
large winding in both directions about a dense set of points, yet when you try to look at
it on a computer, it looks like a line segment. Moreover if you zoom in, or magnify, you
will still just see an apparent line segment. If you continue to zoom in, you will still just
see a line segment, but the direction of the segment will slowly begin to rotate. In fact as
you zoom in, the segment can rotate clockwise n times, then counterclockwise 2n times,
then 4n times clockwise, etc. It can be very difficult to tell on which “side” of J a given
point z ∈
/ J lies.

§2. Janiszewski’s Lemma

This section and the next follows the exposition in Pommerenke, Ch., Univalent Func-
tions.
For the sake of completeness we first review some topological concepts.
A polygonal path in C∗ is a path consisting of finitely many line segments or half-
lines. Suppose U is an open connected subset of the extended plane C∗ . If a ∈ U , set

Ua = {z ∈ U : there exists a polygonal path γ ⊂ U from a to z}.

The set Ua has the following properties:

(a) If z ∈ Ua then we can find a polygonal path connecting a to z which is also simple.
(b) If z, w ∈ Ua , then there is a polygonal path in Ua connecting z to w.
220 XII. Conformal Maps to Jordan Regions

(c) The set Ua is open.

(d) The set U \ Ua is open in U .
(e) There are at most countably many distinct sets Ua .
(f) If E ⊂ U is connected then E is contained in one Ua .
(g) Each component Ua of U has boundary ∂Ua contained in C∗ \ U .
(h) If F is a homeomorphism of C∗ then Ua is a component of U if and only if F (Ua ) is
a component of F (U ).
The sets Ua are called the components of U . Item (a) follows by removing “loops”.
Item (b) says that if b ∈ Ua then Ub = Ua . Items (c) and (d) follow from the fact that
any two points in a ball can be connected by a straight line segment in the ball. The
verification of these items is left as Exercise 1.

Definition 2.1. A closed set E ⊂ C∗ separates points a, b ∈

/ E if a and b belong to
distinct components of C∗ \ E.

Lemma 2.2 (Janiszewski). Suppose A1 and A2 are compact subsets of C such that
A1 ∩ A2 is connected and 0 ∈
/ A1 ∪ A2 . If A1 does not separate 0 and ∞ and if A2 does
not separate 0 and ∞ then A1 ∪ A2 does not separate 0 and ∞.

The hypothesis that A1 ∩ A2 is connected is essential for if A1 is a neighborhood of

the top half of the unit circle and if A2 is a neighborhood of the bottom half of the unit
circle, neither of which contains 0, then A1 ∪ A2 contains a neighborhood of the unit circle
and hence separates 0 and ∞.
Janiszewski’s Lemma follows from Lemma 2.3 and Lemma 2.4 below.

Lemma 2.3. If a compact set E separates 0 and ∞ then we cannot define log z to be
analytic in a neighborhood of E.

In other words, there is no function g which is analytic in a neighborhood of E

satisfying eg(z) = z, if E separates 0 and ∞

Proof. Suppose g is analytic in an open set W ⊃ E and eg(z) = z. Then 0, ∞ ∈ C∗ \ W ⊂

C∗ \ E. Since 0 and ∞ belong to distinct components of C∗ \ E, by Theorem V.2.2 we
 §2: Janiszewski’s Lemma. 221

can find a curve σ ⊂ W so that n(σ, 0) = 1. But by the chain rule, g ′ (z) = 1/z, so
R R
that 2πi = σ 1/z dz = σ g ′ (z)dz = 0 by the Fundamental Theorem of Calculus. This
contradiction proves Lemma 2.3. 

Lemma 2.4. Suppose A1 and A2 are compact sets such that A1 ∩ A2 is connected and
0, ∞ ∈
/ A1 ∪ A2 . If A1 does not separate 0 and ∞ and if A2 does not separate 0 and ∞
then we can define log z to be analytic in a neighborhood of A1 ∪ A2 .

Proof. By hypothesis, we can find simple polygonal curves σ1 and σ2 connecting 0 to ∞

with σj ∩ Aj , j = 1, 2. Then

(σ1 ∪ σ2 ) ∩ (A1 ∪ A2 ) = ∅.

By item (f) above, the connected set A1 ∩A2 is contained in one component U of C∗ \(σ1 ∪σ2 .
Since C∗ \ σj is simply connected,j = 1, 2, we can find functions fj analytic on C∗ \ σj so
that efj (z) = z on C∗ \ σj and fj′ (z) = 1/z. So (f1 − f2 )′ = 0 on C∗ \ (σ1 ∪ σ2 ), and thus
f1 − f2 is constant on each component of C∗ \ (σ1 sup σ2 ). We may add a constant to f2
so that f1 = f2 on U . Because A1 \ U and A2 \ U are disjoint compact sets, we can find
open sets Vj so that
Aj \ U ⊂ Vj ⊂ Vj ⊂ C∗ \ σ1 ,

for j = 1, 2 and V1 ∩ V2 = ∅. Set


 fj (z) for z ∈ Vj
f (z) =

f1 (z) = f2 (z) for z ∈ U
Then f is analytic on V1 ∪ V2 ∪ U , an open neighbhood of A1 ∪ A2 , and ef (z) = z. 

Janiszewski’s Lemma now follows from Lemmas 2.3 and 2.4. Moreover by item (h) we
can replace 0 and ∞ by any two points not in A1 ∪ A2 in the statement of Janiszewski’s
Lemma.

Corollary 2.5. If J is a Jordan arc in C∗ then C∗ \ J is open and connected.

Proof. Since linear fractional transformations are homeomorphisms of C∗ , it suffices to

show that J does not separate 0 and ∞. Write J = ∪n1 Jk where Jk are subarcs of J such
222 XII. Conformal Maps to Jordan Regions

that Jk ∩ Jk−1 is a single point. We may choose each Jk so small that for each k, there is a
half line from 0 to ∞ contained in C∗ \ Jk and hence no Jk separate 0 and ∞. But J1 ∩ J2
a single point, and hence connected, so that by Janiszewski’s Lemma the arc J1 ∪ J2 does
not separate 0 and ∞. The intersection of this arc with J3 again is a single point, and
hence J1 ∪ J2 ∪ J3 does not separate 0 and ∞ by Janiszewski’s Lemma. By induction J
does not separate 0 and ∞. 

§3. Jordan Curve Theorem

In this section we will prove the Jordan Curve Theorem.

Theorem 3.1 (Jordan). If J ⊂ C∗ is a (closed) Jordan curve then C∗ \ J has exactly

two components. Moreover, J is the boundary of each component.

The proof of Theorem 3.1 is divided into four Lemmas.

Lemma 3.2. If J is a (closed) Jordan curve,and if U is a component of C∗ \ J then

∂U = J.

Proof. Suppose ζ ∈ J and z0 ∈ U . Because J is the homeomorphic image of the unit

circle, given n < ∞ we can find a subarc Jn ⊂ J ∩ Dn where Dn = {z : |z − ζ| < 1/n}.
By Corollary 2.5 Jn′ = J \ Jn does not separate ζ and z0 . So there is a polygonal arc σn
from z0 to ζ such that σn ∩ Jn′ = ∅. Let zn be the first intersection of σn with ∂Dn . The
subarc αn ⊂ σn from z0 to zn does not intersect Jn′ and does not intersect Jn ⊂ Dn , and
hence does not intersect J. Since αn is connected, we have αn ⊂ U , and hence zn ∈ U .
But limn zn = ζ ∈
/ U , so ζ ∈ ∂U . This shows that ∂U ⊂ J. If ζ ∈ ∂U , then ζ does not
belong to any component of C∗ \ J since the components are open. Thus ζ ∈ J, and we’ve
shown J ⊂ ∂U. 

Next we will construct a modification σ of the curve J and prove that C∗ \ σ has
exactly two components, and then use this result to prove that C∗ \ J has exactly two
componenets.
 §3: Jordan Curve Theorem. 223

Take ζ0 ∈ C∗ \ J. Then there is a straight line segment [ζ0 , ζ1 ] with ζ1 ∈ J and

[ζ0 , ζ1 ) ∩ J = ∅. Morever we can choose another line segment [ζ0 , ζ2 ] with ζ2 ∈ J and
[ζ0 , ζ2 ) ∩ J = ∅ and [ζ0 , ζ1 ] ∩ [ζ0 , ζ2 ] = {ζ0 }. For otherwise J would be contained in a half
line from ζ1 to ∞ which is impossible. Write J = J1 ∪ J1 where Jj are Jordan arcs with
J1 ∩ J2 = {ζ1 , ζ2 }. Switching ζ1 and ζ2 if necessary,

σ = J1 ∪ [ζ1 , ζ0 ] ∪ [ζ0 , ζ2 ]

is a (closed) Jordan curve. In other words, σ is the modification of J found by replacing

J2 with the union of two intervals.

Lemma 3.3. The set C∗ \ σ has exactly two components.

Proof. If D0 is a disk centered at ζ0 with J ∩ D0 = ∅, then D0 \ ([ζ1 , ζ0 ] ∪ [ζ0 , ζ2 ])

consists of two connected open circular sectors. By item (f) in Section 2, each sector must
be contained in a component of C∗ \ σ and so by Lemma 3.2, there can be at most two
components in C∗ \ σ.
Take z1 and z2 in distinct sectors of D0 \ ([ζ1 , ζ0 ] ∪ [ζ0 , ζ2 ]). Then (σ \ D0 ) ∪ ∂D0 does
not separate z1 and z2 . Moreover σ ∩ (σ \ D0 ) ∪ ∂D0 = σ \ D0 , which is connected. So if
σ also does not separate z1 and z2 then by Janiszewski’s Lemma, E = σ ∪ (σ \ D0 ) ∪ ∂D0
does not separate z1 and z2 . But clearly ∂D0 ∪ [ζ1 , ζ0 ] ∪ [ζ0 , ζ2 ] ⊂ E does separate. This
contradiction proves that z1 and z2 are in distinct components of C∗ \ σ, proving Lemma
3.3. 

Lemma 3.4. The set C∗ \ J has at least two components.

Proof. Take ζ ∈ J1 \ {ζ1 , ζ2 } and define

α = J2 ∪ [ζ2 , ζ0 ] ∪ [ζ0 , ζ1 ].

Let Dζ be a disk centered at ζ such that Dζ ∩ α = ∅. By Lemma 3.2, ζ ∈ ∂G1 ∩ ∂G2

where G1 and G2 are the two components of the complement of σ. Take w1 ∈ G1 ∩ Dζ
and w2 ∈ G2 ∩ Dζ . The points w1 and w2 are separated by σ, but not by α. Note that
224 XII. Conformal Maps to Jordan Regions

J ∩ α = J2 is connected so if J also does not separate w1 and w2 then by Janiszewski’s

Lemma, J ∪α does not separate w1 and w2 . But this is a contradiction since σ ⊂ J ∪α. We
conclude that J must separate w1 and w2 and hence C∗ \ J has at least two components.


Lemma 3.5. The set C∗ \ J has no more than two components.

Proof. Suppose H1 and H2 are the components of C∗ \ J containing w1 and w2 from the
proof of Lemma 3.4. If C∗ \ J has another component H3 , then by Lemma 3.2 again we
can find w3 ∈ H3 ∩ Dζ . But w3 ∈
/ σ and hence w3 ∈ G1 or w3 ∈ G2 . If w3 ∈ G1 then w1
and w3 are not separated by σ and are not separated by α. But σ ∩ α = [ζ1 , ζ0 ] ∪ [ζ0 , ζ2 ] is
connected, so that by Janiszewski’s Lemma w1 and w3 are not separated by σ ∪α. But this
contradicts the assumption that J ⊂ σ ∪ α separates w1 and w3 . A similar contradiction
is obtained if w3 ∈ G2 , proving Lemma 3.5. 

Theorem 3.1 now follows from Lemma 3.4, Lemma 3.5, and Lemma 3.2. 

The proof of Theorem 3.1 yields the following useful Corollary about winding numbers.

Corollary 3.6. If J ⊂ C is a piecewise C 1 (closed) Jordan curve then n(J, z) = 0 for

z in the unbounded component of the complement of J and J can be oriented so that
n(J, z) = 1 for z in the bounded component of the complement of J.

Proof. The winding number is constant in each component of the complement of J and
zero in the unbounded component, so by the Jordan curve theorem, we need only find a
point w ∈
/ J so that n(J, w) = ±1. Let w1 , w2 ∈ Dζ be the points in distinct components
of the complement of σ as constructed in the proof of Lemma 3.4. We proved that w1 and
w2 are also in distinct components of the complement of J. Since n(J, wj ) = n(σ + α, wj ),
we have that

n(J, w1 ) − n(J, w2 ) = n(σ, w1 ) + n(α, w1 ) − n(σ, w2 ) − n(α, w2 ).

 §3: Jordan Curve Theorem. 225

But w1 , w2 ∈ Dζ and Dζ ∩ α = ∅, so that n(α, w1 ) = n(α, w2 ). Moreover wj and zj belong

to the same component Gj of the complement of σ, for j = 1, 2, and winding number is
constant in components, so that

n(J, w1 ) − n(J, w2 ) = n(σ, z1 ) − n(σ, z2 ).

If C1 is one of the two arcs in ∂D0 \ σ then we can orient C1 so that σ1 = σ ∩ D0 + C1 is

a closed curve. Because z1 , z2 ∈ D0 ,

n(σ, z1 ) − n(σ1 , z1 ) = n(σ − σ1 , z1 ) = n(σ − σ1 , z2 ) = n(σ, z2 ) − n(σ1 , z2 ),

and so
n(J, w1 ) − n(J, w2 ) = n(σ1 , z1 ) − n(σ1 , z2 ). (3.1)

Either w1 or w2 is in the unbounded component of the complement of J and hence one

of the quantities on the left side of (3.1) is zero. Likewise one of the quantities on the
right side of (3.1) is zero and moreover the other quantity on the right is either 1 or −1.
Replacing J by −J if necessary, Corollary 3.6 obtains. 

Corollary 3.7. Suppose Ω is a region bounded by finitely many pairwise disjoint piecewise
C 1 closed Jordan curves. If Ω is unbounded then we can orient ∂Ω so that n(∂Ω, z) = 0
for all z ∈ Ω and n(∂Ω, z) = 1 for all z ∈
/ Ω. If Ω is bounded, then we can orient ∂Ω so
that n(J, z) = 1 for all z ∈ Ω and n(J, z) = 0 for all z ∈
/ Ω.

Proof. Fix k. Since Ω is connected, it is contained in one component of the complement of

Jk . If Ω is contained in the bounded component, then each Ji with i 6= k is contained in the
same component, because Ji is in the closure of Ω. Moreover the unbounded component of
the complement of Jk then does not intersect any Ji , with i 6= k, and hence Jk is contained
in the unbounded component of each Ji , with i 6= k Thus if Ω is bounded, then there
is exactly one component Jk such that Ω is contained in the bounded component of the
complement of Jk . Orient Jk so that n(Jk , z) = 1 for z in the bounded component of the
complement of Jk and orient Ji , where i 6= k, so that n(Ji , z) = −1 for z in the bounded
component of the complement of Ji .
226 XII. Conformal Maps to Jordan Regions

If Ω is unbounded, then Ω and Ji must be contained in the unbounded component of

the complement of Jk for each i 6= k. For each k, orient Jk so that n(Jk , z) = 1 for each z
in the bounded component of the complement of Jk . 

§4. Carathéodory’s Theorem

This section is from Garnett and Marshall, Harmonic Measure, Cambridge Univ.
Press, 2005.
Let Ω be a simply connected domain in the extended plane C∗ . We say Ω is a Jordan
domain if Γ = ∂Ω is a Jordan curve in C∗ .

Theorem 4.1. (Carathéodory). Let ϕ be a conformal mapping from the unit disc D onto
a Jordan domain Ω. Then ϕ has a continuous extension to D, and the extension is a
one-to-one map from D onto Ω.

Because ϕ maps D onto Ω, the continuous extension (also denoted by ϕ) must map
∂D onto Γ = ∂Ω, and because ϕ is one-to-one on ∂D, ϕ(eiθ ) parameterizes the Jordan
curve Γ.

Before we prove Carathéodory’s theorem, we use it to solve the Dirichlet problem on

a Jordan domain Ω.
The Dirichlet Problem on a region Ω for a function f ∈ C(∂Ω) is to find a harmonic
function u on Ω which is continuous on Ω and equal to f on ∂Ω. More generally, if f is only
assumed to be bounded on ∂Ω and continuous on a subset E ⊂ ∂Ω, we seek a bounded
function u which is harmonic on Ω, continuous on Ω ∪ E and equal to f on E.
Fourier invented Fourier Series to solve essentially this problem on an infinite strip,
S = {z : Rez > 0 and 0 < Imz < π}. Set I = {iy : 0 < y < π}. If f (z) = 0 for z ∈ ∂S \ I.
Then the harmonic function
∞
X ∞
X
−nx
u(x, y) = an e sin ny = − an Ime−nz
n=1 n=1
P
is equal to f on ∂S \I. If we can find coefficients an so that f (iy) = an sin ny, 0 < y < π,
then u solves the Dirichlet problem on S for the function f .
 §4: Carathéodory’s Theorem. 227

Suppose Ω is a Jordan domain, a simply connected domain bounded by a Jordan

curve Γ. Let f be a bounded Borel function on Γ. If w = ϕ−1 (z), then
Z 2π
1 − |w|2 dθ
u(z) ≡ uf (z) = f ◦ ϕ(eiθ ) (4.1)
0 |eiθ − w|2 2π

is harmonic on Ω, and by Theorem 4.1 and Schwarz’s Theorem VIII.1.5,

lim u(z) = f (ζ) (4.2)

Ω∋z→ζ

whenever ϕ−1 (ζ) ∈ ∂D is a point of continuity of f ◦ ϕ.

Proof of Theorem 4.1. We may assume Ω is bounded. Fix ζ ∈ ∂D. First we show ϕ
has a continuous extension at ζ. Let 0 < δ < 1 and write

γδ = D ∩ {z : |z − ζ| = δ}.

The idea of the proof is that the image curve ϕ(γδ ) cuts off a region Uδ which shrinks to
the point ζ as δ → 0. It is not hard to show that the area of Uδ decreases to 0, but this is
not enough. We use a “length-area principle” to show that the diameter of the boundary
of Uδ , and hence the diameter of Uδ , tends to 0.
The curve ϕ(γδ ) is a Jordan arc having length
Z
L(δ) = |ϕ′ (z)|ds.
γδ

By the Cauchy-Schwarz inequality

Z
2
L (δ) ≤ πδ |ϕ′ (z)|2 ds,
γδ

so that for ρ < 1,

Z ρ Z Z
L2 (δ)
dδ ≤ π |ϕ′ (z)|2 dx dy
0 δ D∩B(ζ,ρ)
(4.5)

= πArea (ϕ(D ∩ B(ζ, ρ)) < ∞.

228 XII. Conformal Maps to Jordan Regions
αn
ζ σn
Un βn
γδn
ϕ(γδn )

Figure XII.1 Crosscuts γδn and ϕ(γδn ).

Therefore there is a sequence δn ↓ 0 such that L(δn ) → 0. When L(δn ) < ∞, the curve
ϕ(γδn ) has endpoints αn , βn , and both of these endpoints must lie on Γ = ∂Ω. Indeed, if
αn ∈ Ω, then some point near αn has two distinct preimages in D because ϕ maps D onto
Ω, and that is impossible because ϕ is one-to-one. Furthermore,

|αn − βn | ≤ L(δn ) → 0. (4.6)

Let σn be that closed subarc of Γ having endpoints αn and βn and having smaller diameter.
Then by uniform continuity (4.6) implies

diam(σn ) → 0,

because Γ is homeomorphic to the circle. By the Jordan curve theorem the curve σn ∪ϕ(γδn )
divides the plane into two (connected, open) regions, and one of these regions, say Un , is
bounded. Then Un ⊂ Ω, because C∗ \ Ω is arcwise connected. Since


diam(∂Un ) = diam σn ∪ ϕ(γδn ) → 0,

we conclude that

diam(Un ) → 0. (4.7)

Set Dn = D ∩ {z : |z − ζ| < δn }. We claim that for large n, ϕ(Dn ) = Un . If not, then by

connectedness ϕ(D \ Dn ) = Un and
 §4: Carathéodory’s Theorem. 229

diam(Un ) ≥ diam(ϕ(B(0, 1/2)) > 0,

T
which contradicts (4.7). Therefore diam(ϕ(Dn )) → 0 and ϕ(Dn ) consists of a single
point, because ϕ(Dn+1 ) ⊂ ϕ(Dn ). That means ϕ has a continuous extension to {ζ} ∪ D. It
is an exercise to show that the union over ζ of these extensions defines a continuous map
on D .

Let ϕ also denote the extension ϕ : D → Ω. Since ϕ(D) = Ω, ϕ maps D onto Ω. To

show ϕ is one-to-one, suppose ϕ(ζ1 ) = ϕ(ζ2 ) but ζ1 6= ζ2 . The argument used to show
αn ∈ Γ also shows that ϕ(∂D) = Γ, and so we can assume ζj ∈ ∂D, j = 1, 2. The Jordan
curve

 
ϕ(rζ1 ) : 0 ≤ r ≤ 1 ∪ ϕ(rζ2 ) : 0 ≤ r ≤ 1

bounds a bounded domain W ⊂ Ω, and then ϕ−1 (W ) is one of the two components of

 
D \ {rζ1 : 0 ≤ r ≤ 1} ∪ {rζ2 : 0 ≤ r ≤ 1} .

But since ϕ(∂D) ⊂ Γ,

ϕ(∂D ∩ ∂ϕ−1 (W )) ⊂ ∂W ∩ ∂Ω = {ϕ(ζ1 )}

and ϕ is constant on an arc of ∂D. It follows that ϕ is constant, either by Schwarz reflection
principle or by the Jensen formula, and this contradiction shows ϕ(ζ1 ) 6= ϕ(ζ2 ). 

One can also prove ϕ is one-to-one by repeating for ϕ−1 the proof that ϕ is continuous.
The Cauchy-Schwarz trick used to prove (4.5) is known as a length-area argument.
The length-area method is the cornerstone of the theory of extremal length.

Corollary 4.2. If h : ∂D → C is a homeomorphism then h extends to be a homeomor-

phism of C onto C.
230 XII. Conformal Maps to Jordan Regions

Proof. Suppose f and g are Riemann maps of the interior and exterior of the disk onto
the inner and outer domains of J, with g(∞) = ∞. For |ζ| = 1, set

 f (rf −1 (h(ζ))) for r ≤ 1
F (rζ) =

g(rg −1(h(ζ))) for r ≥ 1.

Corollary 4.2 fails in higher dimensions. Alexander’s Horned Sphere is a homeomor-

phic image of the ball in R3 . The boundary is the homeomorphic image of the sphere, but
the unbounded component of the complement is not simply connected in the sense that
there are curves in the complement that are not homotopic in the complement to a point.
In particular the homeomorphism cannot be extended to a homeomorphism of R3 .

§5. Exercises

A
1. Prove items (a)-(h) in section 2.
2. In the proof of Carathéodory’s theorem it was proved that the map has a continuous
extension to each ζ ∈ ∂D. Prove that the union over ζ of these extensions defines a
continuous map on D .
3. Suppose G ⊂ C∗ is a connected open set and suppose J is a (closed) Jordan curve and
let H1 and H2 be the components of the complement of J. If ∂G ⊂ H1 then either
G ⊂ H1 or H2 ⊂ G.
4. (a) If E is connected and if Ea is connected for each index a and if Ea intersects E
for all indices a, then the union of E and all Ea is connected.
(b) E connected implies its closure is connected
(c) if E is compact and connected and if Ua is a component of the complement of E
then Ua is simply connected.
(d) Suppose U is simply connected and suppose S is connected and S ∩ (C \ U ) 6= ∅.
Prove that each component of U ∩ (C \ S) is simply connected.
 §5: Exercises. 231

5. Prove that if U is simply connected then ∂U is connected. Hint: use the conformal
map of D onto U .
6. If f is continuous and closed (image of a closed set is closed) on a set X and if X is
locally connected then Y = f (X) is locally connected. Hint: fix p ∈ Y . Let Vq be
S
open the connected subsets of X containing q where f (q) = p. Then Y \ f (Vq ) =
T
f (X \ Vq ). Since f is closed, the union of f (Vq ) is open.
B

7. A compact set K is locally connected if whenever U is a relatively open subset of

K and z ∈ U ⊂ K there is a relatively open subset V of K such that V is connected
and z ∈ V ⊂ U. Let Ω be a simply connected domain such that ∂Ω contains at least
two points. Prove ∂Ω is locally connected if and only if the Riemann map ϕ : D → Ω
extends continuously to D. Hint: apply the proof of Carathéodory’s theorem to the
map of Ω onto D for one direction.
8. (a) Let Ω be a simply connected domain and let σ ⊂ Ω be a crosscut, that is, an
(open) Jordan arc in Ω having distinct endpoints in ∂Ω. Prove that Ω \ σ has two
components Ω1 and Ω2 , each simply connected, and βj = ∂Ωj \ σ is connected.
(b) Let Ω be simply connected, let ψ : Ω → D be conformal and fix z ∈ Ω and ζ ∈ ∂Ω.
Assume ζ and z can be separated by a sequence of crosscuts γn ⊂ Ω such that
length(γn ) → 0. Let Un be the component of Ω \ γn such that z ∈
/ Un . Prove that
ψ(Un ) → α ∈ ∂D.

9. Find and read a construction of Alexander’s Horned Sphere and the proof of the claim
after the proof of Corollary 4.2. Show also that the notion of simply connected in R3
is not the same as having a connected complement.
232
 XIII

The Dirichlet Problem

In this chapter we will treat the Dirichlet problem on arbitrary domains. There are
several other approaches to this material. See for example Garnett and Marshall, Harmonic
Measure. We will use the Perron process because it is the fundamental method underlying
the Uniformization Theorem which we will prove in Chapter XV.
It is not always possible to solve the Dirichlet problem. If f = 0 on ∂D and f (0) = 1,
then f ∈ C(∂Ω) where Ω = D \ {0}. But by Lindelöf’s Maximum Principle VIII.3.2, if u
is harmonic and bounded on Ω with u = 0 on ∂D then u(z) = 0 for all z ∈ Ω. Thus u
extends to be continuous at 0, but u(0) 6= f (0).

§1. Perron Process

Recall that a subharmonic function v on Ω is continuous as a map of Ω into [−∞, +∞)

and satisfies the Mean Value Inequality on all sufficiently small circles. See Definition
VIII.1.2. There are several key properties of subharmonic functions:
(1.i) A subharmonic function v on a region Ω satisfies the Maximum Principle: if there
exists z0 ∈ Ω such that v(z0 ) = supz∈Ω v(z) then v is constant on Ω.
(1.ii) If v1 , v2 are subharmonic then v1 + v2 and w = max(v1 , v2 ) are subharmonic.
(1.iii) If v is subharmonic on a region Ω and integrable on ∂D where D is a disk with
D ⊂ Ω, then the function vD which equals v on Ω \ D and equals the Poisson
integral of v on D is also subharmonic. In other words, if D = {z : |z − c| < r},
we can replace v on D by
Z 2π
1 − |(z − c)/r|2 dt
vD (z) = it 2
v(c + reit ) ,
0 |e − (z − c)/r| 2π

and still be subharmonic.

Lemma 1.1. If v is subharmonic on Ω and if f is analytic on Ω1 with f (Ω1 ) ⊂ Ω, then

v ◦ f is subharmonic on Ω1 .

233
234 XIII. The Dirichlet Problem.

Proof. The set E = {z : v = −∞} is closed and f −1 (E) is closed in Ω1 since f is open.
It suffices to prove that v ◦ f is subharmonic on Ω2 = Ω1 \ f −1 (E). Suppose c ∈ Ω2 and
suppose f ′ (c) 6= 0. Then f is one-to-one in a neighbhorhood of c. If D = {z : |z − c| < r}
is a disk with D ⊂ Ω2 and if f is one-to-one on D, let u be the Poisson integral of v ◦ f
on D. Then u ◦ f −1 is harmonic on f (D) and equal to v on ∂f (D). By the maximum
principle applied to v − u ◦ f −1 we have that v ≤ u ◦ f −1 on f (D), and hence v ◦ f ≤ u on
D. But then Z
dt
v ◦ f (c) ≤ u(c) = v ◦ f (c + reit ) .
∂D 2π
Thus the Mean Value inequality holds for sufficiently small circles centered at c. We have
shown that v ◦ f is subharmonic on Ω2 \ {f ′ = 0}.
The zeros of f ′ however are isolated. Thus if f ′ (c) = 0, and if v ◦ f is subharmonic on
0 < |z − c| < r, continuous and finite on |z − c| = r, then let u be the Poisson integral of
v ◦ f on |z − c| = r. By Lindelöf’s Maximum Principle VIII.3.2, we conclude that v ◦ f ≤ u
and so the Mean Value Inequality holds for v ◦ f on sufficiently small circles centered at c.


For example, if f is analytic then |f |, log |f | and log(1 + |f |2 ) are subharmonic.

Definition 1.2. A family F of subharmonic functions on a region Ω is called a Perron

family if it satisfies:
(i) if v1 , v2 ∈ F then max(v1 , v2 ) ∈ F , and
(ii) if v ∈ F and if D is a disk with D ⊂ Ω, then vD ∈ F , and
(iii) for each z ∈ Ω there exists v ∈ F such that v(z) > −∞.

See (1.iii) for the definition of vD .

Definition 1.3. If F is a Perron family on a region Ω then we define

uF (z) ≡ sup v(z).

v∈F

Theorem 1.4. If F is a Perron family on Ω then uF is harmonic on Ω or uF (z) = +∞

for all z ∈ Ω.
 §4: Exercises. 235

The “bootstrap technique” for climbing a cliff is to put your foot above your head,
then pull yourself up by your bootstraps or shoelaces. Repeat. The proof of Theorem 1.4
is a classic bootstrap proof.

Proof.

§2. Local Barriers

§3. Another Proof of the Riemann Mapping Theorem

§4. Exercises

A
1. Prove items (1.i)-(1.iii) at the beginning of Section 1. B
2. Does a subharmonic function on Ω satisifies the Mean-Value Inequality on each circle
in Ω that bounds a disk contained in Ω?
236
 XIV

Riemann Surfaces

§1. Analytic Continuation.

§2. Monodromy.

§3. Riemann Surface Definition.

§4. Simply Connected Covering Surface.

§5. Deck Transformations.

§6. Exercises.

237
238
 XV

The Uniformization Theorem

The Koebe uniformization theorem is a generalization of the Riemann mapping The-

orem. It says that a simply connected Riemann surface is conformally equivalent to either
the unit disk D, the plane C, or the sphere C∗ . We will give a proof that illustrates
the power of the Perron method. As is standard, the hyperbolic case (D) is proved by
constructing Green’s function. The novel part here is that the non-hyperbolic cases are
treated in a very similar manner by constructing the Dipole Green’s function.

§1. Riemann Surfaces.

A Riemann surface is a connected Hausdorff space W , together with a collection of

open subsets Uα ⊂ W and functions zα : Uα → C such that
(i) W = ∪Uα
(ii) zα is a homeomorphism of Uα onto the unit disk D, and
(iii) if Uα ∩ Uβ 6= ∅ then zβ ◦ zα−1 is analytic on zα (Uα ∩ Uβ ).
The functions zα are called coordinate functions, and the sets Uα are called coordinate
disks. We can extend our collection of coordinate functions and disks to form a base for
1
the topology on W by setting Uα,r = zα−1 (rD) and zα,r = r zα |Uα,r , for r < 1. We can
also compose each zα with a Möbius transformation of the disk onto itself so that we can
assume that for each coordinate disk Uα and each p0 ∈ Uα there is a coordinate function
zα with zα (p0 ) = 0. For the purposes of the discussion below, we will assume that our
collection {zα , Uα } includes all such maps, except that we will further assume that each
Uα has compact closure in W , by restricting our collection.
A Riemann surface W is called simply connected if every closed curve in W is homo-
topic to a point. A function f : W → C is called analytic if for every coordinate function
zα , the function f ◦ zα−1 is analytic on D. Harmonic, subharmonic, and meromorphic
functions on W are defined in a similar way. Note that by (ii) and (iii) the functions

239
240 XV. The Uniformization Theorem

zβ ◦ zα−1 are one-to-one analytic maps. This means that any theorem about plane domains
whose proof depends solely on the local behavior of functions is also a valid theorem about
Riemann surfaces. For example, the monodromy theorem (see e.g. Ahlfors[A1]) holds for
Riemann surfaces, and so does its corollary that every harmonic function on a simply con-
nected Riemann surface has the form u = Ref for some analytic function f defined on W .
The Perron process also works on a Riemann surface. A Riemann surface W is pathwise
connected since the set of points than can be connected to p0 is open for each p0 ∈ W .

§2. Green’s Function.

Suppose W is a Riemann surface. Fix p0 ∈ W and let z : U → D be a coordinate

function such that z(p0 ) = 0. Let Fp0 be the collection of subharmonic functions on W \ p0
satisfying

(a) v = 0 on W \ K, for some compact K ⊂ W with K 6= W, and

.
(b) lim sup v(p) + log |z(p)| < ∞
p→p0

Set
gW (p, p0 ) = sup{v(p) : v ∈ Fp0 }. (1)

Condition (b) does not depend on the choice of the coordinate function zα , provided
zα (p0 ) = 0. The collection Fp0 is a Perron family, so one of the following two cases
holds by Harnack’s Theorem:
Case 1: gW (p, p0 ) is harmonic in W \ {p0 }, or
Case 2: gW (p, p0 ) = +∞ for all p ∈ W \ {p0 }.
In the first case, gW (p, p0 ) is called Green’s function with pole (or logarithmic singularity)
at p0 . In the second case we say that Green’s function does not exist.

In this chapter we give an essentially self contained proof of the following result.

The Uniformization Theorem (Koebe[1907]). Suppose W is a simply connected

Riemann surface. If Green’s function exists for W , then there is a one-to-one analytic map
of W onto D. If W is compact, then there is a one-to-one analytic map of W onto C∗ . If
 §2: Green’s Function. 241

W is not compact and if Green’s function does not exist for W , then there is a one-to-one
analytic map of W onto C.

First we will prove some facts about Green’s function on Riemann surfaces, then we
will prove the Uniformization Theorem in Case 1, when Green’s function exists. Then we
give a similar proof of the Uniformization Theorem in Case 2, using the “Dipole” Green’s
function, assuming it exists. Finally we show that the Dipole Green’s function exists on
every Riemann surface.

Lemma 1. Suppose p0 ∈ W and suppose z : U → D is a coordinate function such that

z(p0 ) = 0. If gW (p, p0 ) exists then,

gW (p, p0 ) > 0 for p ∈ W \ {p0 }, and (2)

gW (p, p0 ) + log |z(p)| extends to be harmonic in U. (3)

Proof. The function 

 − log |z(p)| for p ∈ U
v0 (p) =

0 for p ∈ W \ U
is in Fp0 . Hence gW (p, p0 ) ≥ 0 and gW (p, p0 ) > 0 if p ∈ U . By the maximum principle in
W \ {p0 } (2) holds.
If v ∈ Fp0 and ε > 0, then v + (1 + ε) log |z| extends to be subharmonic in U , and
equal to −∞ at p0 . Thus

sup v + (1 + ε) log |z| = sup v ≤ sup gW < ∞.
U ∂U ∂U

Taking the supremum over v ∈ Fp0 and sending ε → 0, we obtain

gW + log |z| ≤ sup gW < ∞

∂U

in U \ {p0 }. We also have that

gW + log |z| ≥ v0 + log |z| ≥ 0 (4)

242 XV. The Uniformization Theorem

for p ∈ U \ {p0 }. Thus p0 is a removable singularity for the harmonic function gW + log |z|
and (3) holds. 

The Green’s function for the unit disk D is given by

 
 1 − az 
gD (z, a) = log .
z−a 
Indeed by the maximum principle, each candidate subharmonic function v in the Perron
family Fa is bounded by (1 + ε)gD (z, a), for ε > 0. Moreover max(gD (z, a) − ε, 0) ∈ Fa ,
when ε > 0.
The next lemma gives a large collection of Riemann surfaces for which Green’s function
exists.

Lemma 2. Suppose W0 is a Riemann surface and suppose U0 is a coordinate disk whose

closure is compact in W0 . Set W = W0 \ U0 . Then gW (p, p0 ) exists for all p, p0 ∈ W with
p 6= p0 .

Proof. Fix p0 ∈ W and let U ⊂⊂ W be a coordinate disk with coordinate function

z : U → D and z(p0 ) = 0. To prove that gW exists, we show that the family Fp0 is
bounded above. Fix r, with 0 < r < 1, and set rU = {p ∈ W : |z(p)| < r}. If v ∈ Fp0 and
ε > 0 then by the maximum principle


v(p) + (1 + ε) log |z(p)| ≤ max v(q) + (1 + ε) log |z(q)| = max v(q),
q∈∂U q∈∂U

for all p ∈ U . As before, letting ε → 0, we obtain

max v(p) + log r ≤ max v(p). (5)

p∈∂rU p∈∂U

Let ω(p) = ω(p, ∂rU, W \ rU ) be the Perron solution to the Dirichlet problem on W \ rU =
W0 \{U0 ∪rU } for the boundary data which equals 1 on ∂rU and equals 0 on ∂U0 . In other
words, we let F denote the collection of functions u which are subharmonic on W \ rU
with u = 0 on W \ K for some compact set K, depending on u, and such that

lim sup u(p) ≤ 1

p→ζ
 §2: Green’s Function. 243

for ζ ∈ ∂rU . Then

ω(p) = sup{u(p) : u ∈ F }.

By the Perron process, ω is harmonic in W \ U . Regularity for the Dirichlet problem is a

local question. Indeed we can define a local barrier at each point of the boundary of rU
and at each point of the boundary of U0 . Thus the harmonic function ω extends to be
continuous at each point of ∂U 0 and each point of ∂rU so that ω(p) = 0 for p ∈ ∂U0 and
ω(p) = 1 for p ∈ ∂rU . This implies ω is not constant and 0 < ω(p) < 1 for p ∈ W \ rU .
Returning to our function v ∈ Fp0 , by the maximum principle we have that
 
v(p) ≤ max v ω(p)
∂rU

for p ∈ W \ rU since v = 0 off a compact subset of W . So

   
max v ≤ max v max ω(p) ≤ max v (1 − δ) (6)
∂U ∂rU p∈∂U ∂rU

for some δ > 0. Adding inequalities (5) and (6) yields

1
δ max v ≤ log
∂rU r

for all v ∈ F , with δ is independent of v. This implies that Case 2 does not hold and hence
Green’s function exists. 

The next Lemma relates Green’s function on a Riemann surface to Green’s function
on its universal cover.

Lemma 3. Suppose W is a Riemann surface for which gW exists. Let W ∗ be a simply

connected universal covering surface of W and let π : W ∗ → W be the universal covering
map. Then gW ∗ exists and satisfies
X
gW (π(p∗ ), π(p∗0 )) = gW ∗ (p∗ , q ∗ ). (7)
q ∗ :π(q ∗ )=π(p∗
0)

We interpret the infinite sum on the right side of (7) to be the supremum of all sums
over finitely many q ∗ .
244 XV. The Uniformization Theorem

Proof. Suppose q1∗ , . . . , qn∗ are distinct points in W ∗ with π(qj∗ ) = p0 = π(p∗0 ). Suppose
vj ∈ Fqj∗ , the Perron family for the construction of gW ∗ (·, qj∗ ). So vj = 0 off Kj∗ , a compact
subset of W ∗ and
lim sup v(p∗ ) + log |z ◦ π(p∗ )| < ∞,
p∗ →qj∗

where z is a coordinate chart on W with z(p0 ) = 0. Recall that gW (p, p0 ) + log |z(p)|
extends to be finite and continuous at p0 , and hence

lim gW (π(p∗ ), p0 ) + log |z(π(p∗ ))|

p∗ →qj∗

exists and is finite where π(qj∗ ) = p0 . Thus for ε > 0,

X
n 
vj (p ) −(1 + ε)gW (π(p∗ ), p0 )
∗

j=1

extends to be subharmonic and equal to −∞ at qj∗ , for j = 1, . . . , n, and less than or equal
to 0 off ∪j Kj∗ . By the maximum principle, it is bounded above by 0 and letting ε → 0 and
taking the supremum over all such v we conclude that gW ∗ (p∗ , qj∗ ) exists and
n
X
gW ∗ (p∗ , qj∗ ) ≤ gW (π(p∗ ), p0 ).
j=1

Taking the supremum over all such finite sums we have

X
S(p∗ ) ≡ gW ∗ (p∗ , q ∗ ) ≤ gW (π(p∗ ), p0 ).
q ∗ ;π(q ∗ )=p0

Moreover, as a supremum of finite sums of positive harmonic functions,

S(p∗ ) + log |z(π(p∗ ))|

is harmonic in a neighborhood of each qj∗ by Harnack’s Theorem.

Now take v ∈ Fp0 , the Perron family used to construct gW (p, p0 ). Let U ∗ be a
coordinate disk containing p∗0 such that z ◦ π is a coordinate function mapping U ∗ onto D.
Then by the maximum principle

v(π(p∗ )) − (1 + ε)S(p∗ ) ≤ 0
 §2: Green’s Function. 245

for p∗ ∈ U ∗ and ε > 0. Taking the supremum over all such v and letting ε → 0, we obtain

gW (π(p∗ ), p0 ) ≤ S(p∗ )

and thus (7) holds. 

What we have proved so far works for any Riemann surface. We will now consider
simply connected Riemann surfaces and prove the Uniformization Theorem in Case 1.

Theorem 4. If W is a simply connected Riemann surface then the following are equiva-
lent:

gW (p, p0 ) exists for some p0 ∈ W (8)

gW (p, p0 ) exists for all p0 ∈ W, (9)

There is a one-to-one analytic map ϕ from W onto D. (10)

Moreover if gW exists, then

gW (p1 , p0 ) = gW (p0 , p1 ), (11)

and gW (p, p0 ) = − log |ϕ(p)|, where ϕ(p0 ) = 0.

Proof. Suppose there is a one-to-one analytic map ϕ of W onto D and let p0 ∈ W . By

composing ϕ with a Möbius transformation, we can assume that ϕ(p0 ) = 0. If v ∈ Fp0
and if ε > 0, then by (b)
v + (1 + ε) log |ϕ|

is subharmonic in W and equal to −∞ at p0 . By the maximum principle, since v = 0 off

a compact subset K ⊂ W ,
v + (1 + ε) log |ϕ| ≤ 0

on W . Taking the supremum over all such v and letting ε → 0, shows that gW (p, p0 ) < ∞
and therefore (9) holds. Clearly (9) implies (8).
246 XV. The Uniformization Theorem

Now suppose (8) holds. By (3) there is an analytic function f defined on a coordinate
disk U containing p0 so that

Ref (p) = gW (p, p0 ) + log |z(p)|

for p ∈ U . Hence the function

ϕ(p) = ze−f (p)

is analytic in U and satisfies |ϕ(p)| = e−gW (p,p0 ) and ϕ(p0 ) = 0. On any coordinate disk Uα
with p0 ∈
/ Uα , gW (p, p0 ) is the real part of an analytic function. Thus by the monodromy
theorem, there is a function ϕ, analytic on W , such that

|ϕ(p)| = e−gW (p,p0 ) < 1.

We claim that ϕ is one-to-one. If ϕ(p) = ϕ(p0 ) = 0, then clearly p = p0 . Let p1 ∈ W , with

p1 6= p0 . Then by (2), |ϕ(p1 )| < 1 and

ϕ − ϕ(p1 )
ϕ1 ≡
1 − ϕ(p1 )ϕ

is analytic on W and |ϕ1 | < 1. If v ∈ Fp1 , then by the maximum principle

v + (1 + ε) log |ϕ1 | ≤ 0.

Taking the supremum over all such v and sending ε → 0, we see that gW (p, p1 ) exists and
that
gW (p, p1 ) + log |ϕ1 | ≤ 0. (12)

Setting p = p0 in (12) gives

gW (p0 , p1 ) ≤ − log |ϕ1 (p0 )| = − log |ϕ(p1 )| = gW (p1 , p0 ).

Switching the roles of p0 and p1 gives (11). Moreover equality holds in (12) at p = p0 so
that gW (p, p1 ) = − log |ϕ1 (p)| for all p ∈ W \ {p1 }. Now if ϕ(p2 ) = ϕ(p1 ), then by the
definition ϕ1 (p2 ) = 0 and thus gW (p2 , p1 ) = ∞ and p2 = p1 . Therefore ϕ is one-o-one.
The image ϕ(W ) ⊂ D is simply connected, so if ϕ(W ) 6= D then by the Riemann Mapping
 §3: No Green’s Function. 247

Theorem we can find a one-to-one analytic map ψ of ϕ(W ) onto D with ψ(0) = 0. The
map ψ ◦ ϕ is then a one-to-one analytic map of W onto D, with ψ ◦ ϕ(p0 ) = 0, proving
(10). 

§3. No Green’s Function.

We now consider Riemann surfaces which do not have a Green’s function. As seen
above, Green’s function for the disk with pole at 0 is given by G = − log |z|. There is
no Green’s function on the sphere or the plane, but this same function G plays a similar
role. Instead of one pole, or logarithmic singularity, G has two poles on the sphere, with
opposite signs. We will call it a Dipole Green’s function. The next lemma says that a
Dipole Green’s function exists for every Riemann surface. For a simply connected Riemann
surface without Green’s function, the Dipole Green’s function will be used to construct a
conformal map to the sphere or the plane in much the same way as Green’s function was
used in Case 1.

Lemma 5. Suppose W is a Riemann surface and for j = 1, 2, suppose that zj : Uj → D

are coordinate functions with coordinate disks Uj satisfying U1 ∩ U2 = ∅, and zj (pj ) = 0.
Then there is a function G(p) ≡ G(p, p1 , p2 ), harmonic in p ∈ W \ {p1 , p2 } such that

G + log |z1 | extends to be harmonic in U1 , (13)

G − log |z2 | extends to be harmonic in U2 , (14)

and
sup |G(p)| < ∞. (15)
p∈W \{U1 ∪U2 }

Before proving Lemma 5, we will use it to prove the Uniformization Theorem in Case
2, since the proof is similar to the proof in Case 1.

Proof of the Uniformization Theorem in Case 2. By Theorem 4, we may suppose

that gW (p, p1 ) does not exist for all p, p1 ∈ W . By the monodromy theorem, as in the
248 XV. The Uniformization Theorem

proof of Theorem 4, there is a meromorphic function ϕ defined on W such that

|ϕ(p)| = e−G(p,p1 ,p2 ) .

Note that ϕ has a simple zero at p1 , a simple pole at p2 and no other zeros or poles.
Let us prove ϕ is one-to-one. Take p0 ∈ W \ {p1 , p2 }, let ϕ1 be the meromorphic
function on W such that
|ϕ1 (p)| = e−G(p,p0 ,p2 )

and consider the function

ϕ(p) − ϕ(p0 )
H(p) = .
ϕ1 (p)
Then H is analytic on W because its poles at p2 cancel and because ϕ1 has a simple zero
at p0 . By (15) and the analyticity of H, |H| is bounded on W . But if v ∈ Fp1 , the Perron
family used to construct gW (p, p1 ), and if ε > 0, then by the maximum principle
 
 H(p) − H(p1 ) 
v(p) + (1 + ε) log  ≤ 0.
2 sup |H| 
W

Since gW (p, p1 ) does not exist, sup{v(p) : v ∈ Fp1 } ≡ +∞, and therefore

H(p) ≡ H(p1 ) = −ϕ(p0 )/ϕ1 (p1 ) 6= 0, ∞.

From the definition of H, if ϕ1 (p) is finite and not zero, then ϕ(p) 6= ϕ(p0 ) since H is
a non-zero constant. If ϕ1 (p) = 0 then p = p0 from the definition of ϕ1 . Finally ϕ1 has a
pole only at p2 . But ϕ also has a pole at p2 , and only at p2 , and thus ϕ(p2 ) 6= ϕ(p0 ). We
have proved that ϕ(p) = ϕ(p0 ) only if p = p0 . Since p0 is arbitrary, this proves that ϕ is
one-to-one.
We have shown that ϕ is a one-to-one analytic map from W to a simply connected
region ϕ(W ) ⊂ C∗ . If C∗ \ ϕ(W ) contains more than one point, then by the Riemann
mapping Theorem, there is a one-to-one analytic map of ϕ(W ), and hence of W , onto D.
Since we assumed that gW does not exist, this contradicts Theorem 4. Thus C∗ \ ϕ(W )
contains at most one point, and the last two statements of Theorem 6 are now obvious. 

One consequence of the proof of the Uniformization Theorem in Case 1 is the symmetry
property of Green’s function in Corollary 6. This property will be used in the proof of
Lemma 5.
 §3: No Green’s Function. 249

Corollary 6. Suppose W is a Riemann surface for which Green’s function gW (p, q) exists,
for some p, q ∈ W , with p 6= q. Then gW (p, q) exists for all p, q ∈ W with p 6= q and

gW (p, q) = gW (q, p). (16)

Proof. By Lemma 3 and Theorem 4, we may suppose that D is a universal cover of W .

In this case gD (a, b) = − log |(a − b)/(1 − ba)|. Note that if τ is a Möbius transformation of
the disk onto the disk, then gD (a, τ (b)) = gD (τ −1 (a), b) for all a, b ∈ D with a 6= b. Let G
denote the group of deck transformations, a group of Möbius transformations τ satisfying
π ◦ τ = π and if π(q ∗ ) = π(p∗ ) then there is a τ ∈ G such that τ (p∗ ) = q ∗ . Suppose g(p, p0 )
exists for some p0 ∈ W and all p 6= p0 . Choose p∗0 ∈ D so that π(p∗0 ) = p0 . By Lemma 3
6 p∗0
for p∗ =
 ∗ 
X  p − τ (p∗0 ) 
gW (π(p ∗
), π(p∗0 )) = − log 
τ ∈G
1 − τ (p∗ )p∗ 
0

 −1 ∗ 
X  τ (p ) − p∗0 
= 
− log . (17)
−1 ∗ ∗
τ ∈G
1 − τ (p )p 0

Fix p∗ ∈ D. Each term in the sum

X  −1 ∗ 
 τ (p ) − q ∗ 
∗
S(q ) = − log 
τ ∈G
1 − τ −1 (p∗ )q ∗ 

is a positive harmonic function of q ∗ ∈ D \ τ −1 (p∗ ). Since the sum of positive harmonic

functions converges at q ∗ = p∗0 , the function S is harmonic in D \ {τ −1 (p∗ ) : τ ∈ G}, by
Harnack’s Theorem. If v ∈ Fp , the Perron family for gW (q, p) where p = π(p∗ ), then by
the maximum principle v ≤ (1 + ε)S ◦ π −1 for all ε > 0, and letting ε → 0 and taking the
supremum over all v ∈ Fp we conclude that gW (q, p) exists for all q 6= p. Equation (16)
then follows from Lemma 3 and (17). 

We now complete the proof of the Uniformization Theorem by proving Lemma 5.

Proof of Lemma 5. Suppose z0 is a coordinate function with coordinate chart U0

such that U0 ∩ Uj = ∅ for j = 1, 2. Let p0 be the point in W such that z0 (p0 ) = 0. Set
250 XV. The Uniformization Theorem

tU0 = {p ∈ W : |z0 (p)| < t} and set Wt = W \tU0 . By Lemma 2 and Theorem 4, gWt (p, p1 )
exists for all p, p1 ∈ Wt with p 6= p1 . Fix r, 0 < r < 1, and set rU1 = {p ∈ W : |z1 (p)| < r}.
By the maximum principle

gWt (p, p1 ) ≤ M1 (t) ≡ max gWt (q, p1 ), (18)

q∈∂rU1

for all p ∈ Wt \ rU1 , because the same bound holds for all candidates in the Perron family
defining gWt . By (5)
1
M1 (t) ≤ max gWt (p, p1 ) + log . (19)
p∈∂U1 r
By (18), ut (p) ≡ M1 (t) − gWt (p, p1 ) is a positive harmonic function in Wt \ rU1 and by
(19) there exists q ∈ ∂U1 with ut (q) ≤ log 1r . Riemann surfaces are pathwise connected
so that if K is a compact subset of W1 \ rU1 containing {p2 } ∪ ∂rU1 , then by Harnack’s
inequality there is a constant C < ∞ depending on K and r but not on t, so that for all
p∈K
0 ≤ ut (p) ≤ C,

and
|gWt (p, p1 ) − gWt (p2 , p1 )| = |ut (p2 ) − ut (p)| ≤ 2C.

Likewise, if K ′ is a compact subset of W1 \ {|z2 | < r} containing {p1 } ∪ ∂rU2 , there is a

constant C < ∞ so that
|gWt (p, p2 ) − gWt (p1 , p2 )| ≤ C,

for all p ∈ K ′ .
By Corollary 6, gWt (p1 , p2 ) = gWt (p2 , p1 ) and so the function

Gt (p, p1 , p2 ) ≡ gWt (p, p1 ) − gWt (p, p2 )

= (gWt (p, p1 ) − gWt (p2 , p1 )) − (gWt (p, p2 ) − gWt (p1 , p2 )

is harmonic in Wt \ {p1 , p2 } and satisfies

|Gt (p, p1 , p2 )| ≤ C,

for all p ∈ K ∩ K ′ and some finite C independent of t. We may suppose, for instance, that
K ∩ K ′ contains ∂U1 ∪ ∂U2 . If v ∈ Fp1 , the Perron family for gWt (p, p1 ), then since v = 0
 §3: No Green’s Function. 251

off a compact subset of Wt and since gWt > 0, by the maximum principle

sup v(p) − gWt (p, p2 ) ≤ max(0, sup v(p) − gWt (p, p2 ))

Wt \U1 ∂U1

≤ max(0, sup gWt (p, p1 ) − gWt (p, p2 )) ≤ C,

∂U1

and taking the supremum over all such v yields

sup G(p, p1 , p2 ) ≤ C.
p∈Wt \U1

Similarly
inf Gt (p, p1 , p2 ) = − sup −Gt (p, p1 , p2 ) ≥ −C,
p∈Wt \U2 p∈Wt \U2

and
|Gt (p, p1 , p2 )| ≤ C

for all p ∈ Wt \ {U1 ∪ U2 }. The function Gt + log |z1 | extends to be harmonic in U1 , so by

the maximum principle, we have that

sup |Gt + log |z1 || = sup |Gt + log |z1 || = sup |Gt | ≤ C.
U1 ∂U1 ∂U1

Similarly
sup |Gt − log |z2 || = sup |Gt − log |z2 || = sup |Gt | ≤ C.
U2 ∂U2 ∂U2

By normal families, there exists a sequence tn → 0 so that Gtn converges uniformly on

compact subsets of W \ {p0 , p1 , p2 } to a function G(p, p1 , p2 ) satisifying (13), (14), and
(15). The function G(p, p1 , p2 ) extends to be harmonic at p0 because it is bounded in a
punctured neighborhood of p0 . 

Since each of the spaces D, C, and C∗ is second countable, we have the following
consequence.

Corollary 7. Every simply connected Riemann surface satisfies the second axiom of
countability.
252 XV. The Uniformization Theorem

§2. Comments.

Comments:
1. It is possible to avoid the use of the Riemann Mapping Theorem (RMT) and then
RMT is a consequence of Theorem 6. In doing so, we use ideas from other proofs of that
theorem.
There are two places RMT is used. The first is at the end of the proof in Case I,
where we obtain a conformal map ϕ of W onto a possibly proper subset of D. Here is how
to prove that ϕ is onto:
If there exists a ∈ D \ ϕ(W ), then (z − a)/(1 − az) is a non-vanishing function on the
simply connected region ϕ(W ) and hence we can define an analytic square root
r
z−a
τ (z) =
1 − az

for z ∈ ϕ(W ). Set

τ (z) − τ (0)
σ(z) = .
1 − τ (0)τ (z)
Then σ ◦ ϕ is an analytic function mapping W into D. Moreover σ −1 is an analytic
function defined on all of D, mapping D into D with σ −1 (0) = 0 and it is not a rotation.
By Schwarz’s lemma |σ −1 (z)| < |z| and so

|σ ◦ ϕ(z)| > |ϕ(z)|. (20)

By the argument used to establish (9) and (11), gW + log |σ ◦ ϕ| ≤ 0. But gW (p, p0 ) =
− log |ϕ(p)|, so that log |σ ◦ ϕ| ≤ log |ϕ|, contradicting (20). 

The second place where RMT is used is at the conclusion of the proof in Case 2, where
there is no Green’s function. We proved that there is a conformal map ϕ of W onto a
subset of C∗ . If ϕ(W ) omitted at least two points in C∗ , then we used the RMT to prove
there is a conformal map onto the disk and thus contradict Theorem 4. To avoid RMT,
we can simply use a part of its proof. If there are at least two points a, b ∈ C∗ \ ϕ(W ) then

z−a
ψ(z) =
z−b
 §2: Comments. 253

is a nonvanishing analytic function on the simply connected region ϕ(W ), and hence we can
p
define an analytic square root σ(z) = ψ(z). It is not hard to check that ψ is one-to-one,
and if ψ(ϕ(W )) covers a neighborhood of z1 , z1 6= 0, then ψ(ϕ(W )) omits a neighborhood
of −z1 . We can then apply a linear fractional transformation σ so that σ(ψ(ϕ(W ))) ⊂ D.
If gD is Green’s function on D then the Perron family Fp0 for constructing Green’s function
on W is bounded by gD (f (p), f (p0 )) where f = σ ◦ ψ ◦ ϕ. Thus Green’s function exists,
contradicting our assumption.
254
 XVI

The Uniformization Theorem

The Koebe uniformization theorem is a generalization of the Riemann mapping The-

orem. It says that a simply connected Riemann surface is conformally equivalent to either
the unit disk D, the plane C, or the sphere C∗ . We will give a proof that illustrates
the power of the Perron method. As is standard, the hyperbolic case (D) is proved by
constructing Green’s function. The novel part here is that the non-hyperbolic cases are
treated in a very similar manner by constructing the Dipole Green’s function.

§1. Riemann Surfaces.

A Riemann surface is a connected Hausdorff space W , together with a collection of

open subsets Uα ⊂ W and functions zα : Uα → C such that
(i) W = ∪Uα
(ii) zα is a homeomorphism of Uα onto the unit disk D, and
(iii) if Uα ∩ Uβ 6= ∅ then zβ ◦ zα−1 is analytic on zα (Uα ∩ Uβ ).
The functions zα are called coordinate functions, and the sets Uα are called coordinate
disks. We can extend our collection of coordinate functions and disks to form a base for
1
the topology on W by setting Uα,r = zα−1 (rD) and zα,r = r zα |Uα,r , for r < 1. We can
also compose each zα with a Möbius transformation of the disk onto itself so that we can
assume that for each coordinate disk Uα and each p0 ∈ Uα there is a coordinate function
zα with zα (p0 ) = 0. For the purposes of the discussion below, we will assume that our
collection {zα , Uα } includes all such maps, except that we will further assume that each
Uα has compact closure in W , by restricting our collection.
A Riemann surface W is called simply connected if every closed curve in W is homo-
topic to a point. A function f : W → C is called analytic if for every coordinate function
zα , the function f ◦ zα−1 is analytic on D. Harmonic, subharmonic, and meromorphic
functions on W are defined in a similar way. Note that by (ii) and (iii) the functions

255
256 XV. The Uniformization Theorem

zβ ◦ zα−1 are one-to-one analytic maps. This means that any theorem about plane domains
whose proof depends solely on the local behavior of functions is also a valid theorem about
Riemann surfaces. For example, the monodromy theorem (see e.g. Ahlfors[A1]) holds for
Riemann surfaces, and so does its corollary that every harmonic function on a simply con-
nected Riemann surface has the form u = Ref for some analytic function f defined on W .
The Perron process also works on a Riemann surface. A Riemann surface W is pathwise
connected since the set of points than can be connected to p0 is open for each p0 ∈ W .

§2. Green’s Function.

Suppose W is a Riemann surface. Fix p0 ∈ W and let z : U → D be a coordinate

function such that z(p0 ) = 0. Let Fp0 be the collection of subharmonic functions on W \ p0
satisfying

(a) v = 0 on W \ K, for some compact K ⊂ W with K 6= W, and

.
(b) lim sup v(p) + log |z(p)| < ∞
p→p0

Set
gW (p, p0 ) = sup{v(p) : v ∈ Fp0 }. (1)

Condition (b) does not depend on the choice of the coordinate function zα , provided
zα (p0 ) = 0. The collection Fp0 is a Perron family, so one of the following two cases
holds by Harnack’s Theorem:
Case 1: gW (p, p0 ) is harmonic in W \ {p0 }, or
Case 2: gW (p, p0 ) = +∞ for all p ∈ W \ {p0 }.
In the first case, gW (p, p0 ) is called Green’s function with pole (or logarithmic singularity)
at p0 . In the second case we say that Green’s function does not exist.

In this chapter we give an essentially self contained proof of the following result.

The Uniformization Theorem (Koebe[1907]). Suppose W is a simply connected

Riemann surface. If Green’s function exists for W , then there is a one-to-one analytic map
of W onto D. If W is compact, then there is a one-to-one analytic map of W onto C∗ . If
 §2: Green’s Function. 257

W is not compact and if Green’s function does not exist for W , then there is a one-to-one
analytic map of W onto C.

First we will prove some facts about Green’s function on Riemann surfaces, then we
will prove the Uniformization Theorem in Case 1, when Green’s function exists. Then we
give a similar proof of the Uniformization Theorem in Case 2, using the “Dipole” Green’s
function, assuming it exists. Finally we show that the Dipole Green’s function exists on
every Riemann surface.

Lemma 1. Suppose p0 ∈ W and suppose z : U → D is a coordinate function such that

z(p0 ) = 0. If gW (p, p0 ) exists then,

gW (p, p0 ) > 0 for p ∈ W \ {p0 }, and (2)

gW (p, p0 ) + log |z(p)| extends to be harmonic in U. (3)

Proof. The function 

 − log |z(p)| for p ∈ U
v0 (p) =

0 for p ∈ W \ U
is in Fp0 . Hence gW (p, p0 ) ≥ 0 and gW (p, p0 ) > 0 if p ∈ U . By the maximum principle in
W \ {p0 } (2) holds.
If v ∈ Fp0 and ε > 0, then v + (1 + ε) log |z| extends to be subharmonic in U , and
equal to −∞ at p0 . Thus

sup v + (1 + ε) log |z| = sup v ≤ sup gW < ∞.
U ∂U ∂U

Taking the supremum over v ∈ Fp0 and sending ε → 0, we obtain

gW + log |z| ≤ sup gW < ∞

∂U

in U \ {p0 }. We also have that

gW + log |z| ≥ v0 + log |z| ≥ 0 (4)

258 XV. The Uniformization Theorem

for p ∈ U \ {p0 }. Thus p0 is a removable singularity for the harmonic function gW + log |z|
and (3) holds. 

The Green’s function for the unit disk D is given by

 
 1 − az 
gD (z, a) = log .
z−a 
Indeed by the maximum principle, each candidate subharmonic function v in the Perron
family Fa is bounded by (1 + ε)gD (z, a), for ε > 0. Moreover max(gD (z, a) − ε, 0) ∈ Fa ,
when ε > 0.
The next lemma gives a large collection of Riemann surfaces for which Green’s function
exists.

Lemma 2. Suppose W0 is a Riemann surface and suppose U0 is a coordinate disk whose

closure is compact in W0 . Set W = W0 \ U0 . Then gW (p, p0 ) exists for all p, p0 ∈ W with
p 6= p0 .

Proof. Fix p0 ∈ W and let U ⊂⊂ W be a coordinate disk with coordinate function

z : U → D and z(p0 ) = 0. To prove that gW exists, we show that the family Fp0 is
bounded above. Fix r, with 0 < r < 1, and set rU = {p ∈ W : |z(p)| < r}. If v ∈ Fp0 and
ε > 0 then by the maximum principle


v(p) + (1 + ε) log |z(p)| ≤ max v(q) + (1 + ε) log |z(q)| = max v(q),
q∈∂U q∈∂U

for all p ∈ U . As before, letting ε → 0, we obtain

max v(p) + log r ≤ max v(p). (5)

p∈∂rU p∈∂U

Let ω(p) = ω(p, ∂rU, W \ rU ) be the Perron solution to the Dirichlet problem on W \ rU =
W0 \{U0 ∪rU } for the boundary data which equals 1 on ∂rU and equals 0 on ∂U0 . In other
words, we let F denote the collection of functions u which are subharmonic on W \ rU
with u = 0 on W \ K for some compact set K, depending on u, and such that

lim sup u(p) ≤ 1

p→ζ
 §2: Green’s Function. 259

for ζ ∈ ∂rU . Then

ω(p) = sup{u(p) : u ∈ F }.

By the Perron process, ω is harmonic in W \ U . Regularity for the Dirichlet problem is a

local question. Indeed we can define a local barrier at each point of the boundary of rU
and at each point of the boundary of U0 . Thus the harmonic function ω extends to be
continuous at each point of ∂U 0 and each point of ∂rU so that ω(p) = 0 for p ∈ ∂U0 and
ω(p) = 1 for p ∈ ∂rU . This implies ω is not constant and 0 < ω(p) < 1 for p ∈ W \ rU .
Returning to our function v ∈ Fp0 , by the maximum principle we have that
 
v(p) ≤ max v ω(p)
∂rU

for p ∈ W \ rU since v = 0 off a compact subset of W . So

   
max v ≤ max v max ω(p) ≤ max v (1 − δ) (6)
∂U ∂rU p∈∂U ∂rU

for some δ > 0. Adding inequalities (5) and (6) yields

1
δ max v ≤ log
∂rU r

for all v ∈ F , with δ is independent of v. This implies that Case 2 does not hold and hence
Green’s function exists. 

The next Lemma relates Green’s function on a Riemann surface to Green’s function
on its universal cover.

Lemma 3. Suppose W is a Riemann surface for which gW exists. Let W ∗ be a simply

connected universal covering surface of W and let π : W ∗ → W be the universal covering
map. Then gW ∗ exists and satisfies
X
gW (π(p∗ ), π(p∗0 )) = gW ∗ (p∗ , q ∗ ). (7)
q ∗ :π(q ∗ )=π(p∗
0)

We interpret the infinite sum on the right side of (7) to be the supremum of all sums
over finitely many q ∗ .
260 XV. The Uniformization Theorem

Proof. Suppose q1∗ , . . . , qn∗ are distinct points in W ∗ with π(qj∗ ) = p0 = π(p∗0 ). Suppose
vj ∈ Fqj∗ , the Perron family for the construction of gW ∗ (·, qj∗ ). So vj = 0 off Kj∗ , a compact
subset of W ∗ and
lim sup v(p∗ ) + log |z ◦ π(p∗ )| < ∞,
p∗ →qj∗

where z is a coordinate chart on W with z(p0 ) = 0. Recall that gW (p, p0 ) + log |z(p)|
extends to be finite and continuous at p0 , and hence

lim gW (π(p∗ ), p0 ) + log |z(π(p∗ ))|

p∗ →qj∗

exists and is finite where π(qj∗ ) = p0 . Thus for ε > 0,

X
n 
vj (p ) −(1 + ε)gW (π(p∗ ), p0 )
∗

j=1

extends to be subharmonic and equal to −∞ at qj∗ , for j = 1, . . . , n, and less than or equal
to 0 off ∪j Kj∗ . By the maximum principle, it is bounded above by 0 and letting ε → 0 and
taking the supremum over all such v we conclude that gW ∗ (p∗ , qj∗ ) exists and
n
X
gW ∗ (p∗ , qj∗ ) ≤ gW (π(p∗ ), p0 ).
j=1

Taking the supremum over all such finite sums we have

X
S(p∗ ) ≡ gW ∗ (p∗ , q ∗ ) ≤ gW (π(p∗ ), p0 ).
q ∗ ;π(q ∗ )=p0

Moreover, as a supremum of finite sums of positive harmonic functions,

S(p∗ ) + log |z(π(p∗ ))|

is harmonic in a neighborhood of each qj∗ by Harnack’s Theorem.

Now take v ∈ Fp0 , the Perron family used to construct gW (p, p0 ). Let U ∗ be a
coordinate disk containing p∗0 such that z ◦ π is a coordinate function mapping U ∗ onto D.
Then by the maximum principle

v(π(p∗ )) − (1 + ε)S(p∗ ) ≤ 0
 §2: Green’s Function. 261

for p∗ ∈ U ∗ and ε > 0. Taking the supremum over all such v and letting ε → 0, we obtain

gW (π(p∗ ), p0 ) ≤ S(p∗ )

and thus (7) holds. 

What we have proved so far works for any Riemann surface. We will now consider
simply connected Riemann surfaces and prove the Uniformization Theorem in Case 1.

Theorem 4. If W is a simply connected Riemann surface then the following are equiva-
lent:

gW (p, p0 ) exists for some p0 ∈ W (8)

gW (p, p0 ) exists for all p0 ∈ W, (9)

There is a one-to-one analytic map ϕ from W onto D. (10)

Moreover if gW exists, then

gW (p1 , p0 ) = gW (p0 , p1 ), (11)

and gW (p, p0 ) = − log |ϕ(p)|, where ϕ(p0 ) = 0.

Proof. Suppose there is a one-to-one analytic map ϕ of W onto D and let p0 ∈ W . By

composing ϕ with a Möbius transformation, we can assume that ϕ(p0 ) = 0. If v ∈ Fp0
and if ε > 0, then by (b)
v + (1 + ε) log |ϕ|

is subharmonic in W and equal to −∞ at p0 . By the maximum principle, since v = 0 off

a compact subset K ⊂ W ,
v + (1 + ε) log |ϕ| ≤ 0

on W . Taking the supremum over all such v and letting ε → 0, shows that gW (p, p0 ) < ∞
and therefore (9) holds. Clearly (9) implies (8).
262 XV. The Uniformization Theorem

Now suppose (8) holds. By (3) there is an analytic function f defined on a coordinate
disk U containing p0 so that

Ref (p) = gW (p, p0 ) + log |z(p)|

for p ∈ U . Hence the function

ϕ(p) = ze−f (p)

is analytic in U and satisfies |ϕ(p)| = e−gW (p,p0 ) and ϕ(p0 ) = 0. On any coordinate disk Uα
with p0 ∈
/ Uα , gW (p, p0 ) is the real part of an analytic function. Thus by the monodromy
theorem, there is a function ϕ, analytic on W , such that

|ϕ(p)| = e−gW (p,p0 ) < 1.

We claim that ϕ is one-to-one. If ϕ(p) = ϕ(p0 ) = 0, then clearly p = p0 . Let p1 ∈ W , with

p1 6= p0 . Then by (2), |ϕ(p1 )| < 1 and

ϕ − ϕ(p1 )
ϕ1 ≡
1 − ϕ(p1 )ϕ

is analytic on W and |ϕ1 | < 1. If v ∈ Fp1 , then by the maximum principle

v + (1 + ε) log |ϕ1 | ≤ 0.

Taking the supremum over all such v and sending ε → 0, we see that gW (p, p1 ) exists and
that
gW (p, p1 ) + log |ϕ1 | ≤ 0. (12)

Setting p = p0 in (12) gives

gW (p0 , p1 ) ≤ − log |ϕ1 (p0 )| = − log |ϕ(p1 )| = gW (p1 , p0 ).

Switching the roles of p0 and p1 gives (11). Moreover equality holds in (12) at p = p0 so
that gW (p, p1 ) = − log |ϕ1 (p)| for all p ∈ W \ {p1 }. Now if ϕ(p2 ) = ϕ(p1 ), then by the
definition ϕ1 (p2 ) = 0 and thus gW (p2 , p1 ) = ∞ and p2 = p1 . Therefore ϕ is one-o-one.
The image ϕ(W ) ⊂ D is simply connected, so if ϕ(W ) 6= D then by the Riemann Mapping
 §3: No Green’s Function. 263

Theorem we can find a one-to-one analytic map ψ of ϕ(W ) onto D with ψ(0) = 0. The
map ψ ◦ ϕ is then a one-to-one analytic map of W onto D, with ψ ◦ ϕ(p0 ) = 0, proving
(10). 

§3. No Green’s Function.

We now consider Riemann surfaces which do not have a Green’s function. As seen
above, Green’s function for the disk with pole at 0 is given by G = − log |z|. There is
no Green’s function on the sphere or the plane, but this same function G plays a similar
role. Instead of one pole, or logarithmic singularity, G has two poles on the sphere, with
opposite signs. We will call it a Dipole Green’s function. The next lemma says that a
Dipole Green’s function exists for every Riemann surface. For a simply connected Riemann
surface without Green’s function, the Dipole Green’s function will be used to construct a
conformal map to the sphere or the plane in much the same way as Green’s function was
used in Case 1.

Lemma 5. Suppose W is a Riemann surface and for j = 1, 2, suppose that zj : Uj → D

are coordinate functions with coordinate disks Uj satisfying U1 ∩ U2 = ∅, and zj (pj ) = 0.
Then there is a function G(p) ≡ G(p, p1 , p2 ), harmonic in p ∈ W \ {p1 , p2 } such that

G + log |z1 | extends to be harmonic in U1 , (13)

G − log |z2 | extends to be harmonic in U2 , (14)

and
sup |G(p)| < ∞. (15)
p∈W \{U1 ∪U2 }

Before proving Lemma 5, we will use it to prove the Uniformization Theorem in Case
2, since the proof is similar to the proof in Case 1.

Proof of the Uniformization Theorem in Case 2. By Theorem 4, we may suppose

that gW (p, p1 ) does not exist for all p, p1 ∈ W . By the monodromy theorem, as in the
264 XV. The Uniformization Theorem

proof of Theorem 4, there is a meromorphic function ϕ defined on W such that

|ϕ(p)| = e−G(p,p1 ,p2 ) .

Note that ϕ has a simple zero at p1 , a simple pole at p2 and no other zeros or poles.
Let us prove ϕ is one-to-one. Take p0 ∈ W \ {p1 , p2 }, let ϕ1 be the meromorphic
function on W such that
|ϕ1 (p)| = e−G(p,p0 ,p2 )

and consider the function

ϕ(p) − ϕ(p0 )
H(p) = .
ϕ1 (p)
Then H is analytic on W because its poles at p2 cancel and because ϕ1 has a simple zero
at p0 . By (15) and the analyticity of H, |H| is bounded on W . But if v ∈ Fp1 , the Perron
family used to construct gW (p, p1 ), and if ε > 0, then by the maximum principle
 
 H(p) − H(p1 ) 
v(p) + (1 + ε) log  ≤ 0.
2 sup |H| 
W

Since gW (p, p1 ) does not exist, sup{v(p) : v ∈ Fp1 } ≡ +∞, and therefore

H(p) ≡ H(p1 ) = −ϕ(p0 )/ϕ1 (p1 ) 6= 0, ∞.

From the definition of H, if ϕ1 (p) is finite and not zero, then ϕ(p) 6= ϕ(p0 ) since H is
a non-zero constant. If ϕ1 (p) = 0 then p = p0 from the definition of ϕ1 . Finally ϕ1 has a
pole only at p2 . But ϕ also has a pole at p2 , and only at p2 , and thus ϕ(p2 ) 6= ϕ(p0 ). We
have proved that ϕ(p) = ϕ(p0 ) only if p = p0 . Since p0 is arbitrary, this proves that ϕ is
one-to-one.
We have shown that ϕ is a one-to-one analytic map from W to a simply connected
region ϕ(W ) ⊂ C∗ . If C∗ \ ϕ(W ) contains more than one point, then by the Riemann
mapping Theorem, there is a one-to-one analytic map of ϕ(W ), and hence of W , onto D.
Since we assumed that gW does not exist, this contradicts Theorem 4. Thus C∗ \ ϕ(W )
contains at most one point, and the last two statements of Theorem 6 are now obvious. 

One consequence of the proof of the Uniformization Theorem in Case 1 is the symmetry
property of Green’s function in Corollary 6. This property will be used in the proof of
Lemma 5.
 §3: No Green’s Function. 265

Corollary 6. Suppose W is a Riemann surface for which Green’s function gW (p, q) exists,
for some p, q ∈ W , with p 6= q. Then gW (p, q) exists for all p, q ∈ W with p 6= q and

gW (p, q) = gW (q, p). (16)

Proof. By Lemma 3 and Theorem 4, we may suppose that D is a universal cover of W .

In this case gD (a, b) = − log |(a − b)/(1 − ba)|. Note that if τ is a Möbius transformation of
the disk onto the disk, then gD (a, τ (b)) = gD (τ −1 (a), b) for all a, b ∈ D with a 6= b. Let G
denote the group of deck transformations, a group of Möbius transformations τ satisfying
π ◦ τ = π and if π(q ∗ ) = π(p∗ ) then there is a τ ∈ G such that τ (p∗ ) = q ∗ . Suppose g(p, p0 )
exists for some p0 ∈ W and all p 6= p0 . Choose p∗0 ∈ D so that π(p∗0 ) = p0 . By Lemma 3
6 p∗0
for p∗ =
 ∗ 
X  p − τ (p∗0 ) 
gW (π(p ∗
), π(p∗0 )) = − log 
τ ∈G
1 − τ (p∗ )p∗ 
0

 −1 ∗ 
X  τ (p ) − p∗0 
= 
− log . (17)
−1 ∗ ∗
τ ∈G
1 − τ (p )p 0

Fix p∗ ∈ D. Each term in the sum

X  −1 ∗ 
 τ (p ) − q ∗ 
∗
S(q ) = − log 
τ ∈G
1 − τ −1 (p∗ )q ∗ 

is a positive harmonic function of q ∗ ∈ D \ τ −1 (p∗ ). Since the sum of positive harmonic

functions converges at q ∗ = p∗0 , the function S is harmonic in D \ {τ −1 (p∗ ) : τ ∈ G}, by
Harnack’s Theorem. If v ∈ Fp , the Perron family for gW (q, p) where p = π(p∗ ), then by
the maximum principle v ≤ (1 + ε)S ◦ π −1 for all ε > 0, and letting ε → 0 and taking the
supremum over all v ∈ Fp we conclude that gW (q, p) exists for all q 6= p. Equation (16)
then follows from Lemma 3 and (17). 

We now complete the proof of the Uniformization Theorem by proving Lemma 5.

Proof of Lemma 5. Suppose z0 is a coordinate function with coordinate chart U0

such that U0 ∩ Uj = ∅ for j = 1, 2. Let p0 be the point in W such that z0 (p0 ) = 0. Set
266 XV. The Uniformization Theorem

tU0 = {p ∈ W : |z0 (p)| < t} and set Wt = W \tU0 . By Lemma 2 and Theorem 4, gWt (p, p1 )
exists for all p, p1 ∈ Wt with p 6= p1 . Fix r, 0 < r < 1, and set rU1 = {p ∈ W : |z1 (p)| < r}.
By the maximum principle

gWt (p, p1 ) ≤ M1 (t) ≡ max gWt (q, p1 ), (18)

q∈∂rU1

for all p ∈ Wt \ rU1 , because the same bound holds for all candidates in the Perron family
defining gWt . By (5)
1
M1 (t) ≤ max gWt (p, p1 ) + log . (19)
p∈∂U1 r
By (18), ut (p) ≡ M1 (t) − gWt (p, p1 ) is a positive harmonic function in Wt \ rU1 and by
(19) there exists q ∈ ∂U1 with ut (q) ≤ log 1r . Riemann surfaces are pathwise connected
so that if K is a compact subset of W1 \ rU1 containing {p2 } ∪ ∂rU1 , then by Harnack’s
inequality there is a constant C < ∞ depending on K and r but not on t, so that for all
p∈K
0 ≤ ut (p) ≤ C,

and
|gWt (p, p1 ) − gWt (p2 , p1 )| = |ut (p2 ) − ut (p)| ≤ 2C.

Likewise, if K ′ is a compact subset of W1 \ {|z2 | < r} containing {p1 } ∪ ∂rU2 , there is a

constant C < ∞ so that
|gWt (p, p2 ) − gWt (p1 , p2 )| ≤ C,

for all p ∈ K ′ .
By Corollary 6, gWt (p1 , p2 ) = gWt (p2 , p1 ) and so the function

Gt (p, p1 , p2 ) ≡ gWt (p, p1 ) − gWt (p, p2 )

= (gWt (p, p1 ) − gWt (p2 , p1 )) − (gWt (p, p2 ) − gWt (p1 , p2 )

is harmonic in Wt \ {p1 , p2 } and satisfies

|Gt (p, p1 , p2 )| ≤ C,

for all p ∈ K ∩ K ′ and some finite C independent of t. We may suppose, for instance, that
K ∩ K ′ contains ∂U1 ∪ ∂U2 . If v ∈ Fp1 , the Perron family for gWt (p, p1 ), then since v = 0
 §3: No Green’s Function. 267

off a compact subset of Wt and since gWt > 0, by the maximum principle

sup v(p) − gWt (p, p2 ) ≤ max(0, sup v(p) − gWt (p, p2 ))

Wt \U1 ∂U1

≤ max(0, sup gWt (p, p1 ) − gWt (p, p2 )) ≤ C,

∂U1

and taking the supremum over all such v yields

sup G(p, p1 , p2 ) ≤ C.
p∈Wt \U1

Similarly
inf Gt (p, p1 , p2 ) = − sup −Gt (p, p1 , p2 ) ≥ −C,
p∈Wt \U2 p∈Wt \U2

and
|Gt (p, p1 , p2 )| ≤ C

for all p ∈ Wt \ {U1 ∪ U2 }. The function Gt + log |z1 | extends to be harmonic in U1 , so by

the maximum principle, we have that

sup |Gt + log |z1 || = sup |Gt + log |z1 || = sup |Gt | ≤ C.
U1 ∂U1 ∂U1

Similarly
sup |Gt − log |z2 || = sup |Gt − log |z2 || = sup |Gt | ≤ C.
U2 ∂U2 ∂U2

By normal families, there exists a sequence tn → 0 so that Gtn converges uniformly on

compact subsets of W \ {p0 , p1 , p2 } to a function G(p, p1 , p2 ) satisifying (13), (14), and
(15). The function G(p, p1 , p2 ) extends to be harmonic at p0 because it is bounded in a
punctured neighborhood of p0 . 

Since each of the spaces D, C, and C∗ is second countable, we have the following
consequence.

Corollary 7. Every simply connected Riemann surface satisfies the second axiom of
countability.
268 XV. The Uniformization Theorem

§2. Comments.

Comments:
1. It is possible to avoid the use of the Riemann Mapping Theorem (RMT) and then
RMT is a consequence of Theorem 6. In doing so, we use ideas from other proofs of that
theorem.
There are two places RMT is used. The first is at the end of the proof in Case I,
where we obtain a conformal map ϕ of W onto a possibly proper subset of D. Here is how
to prove that ϕ is onto:
If there exists a ∈ D \ ϕ(W ), then (z − a)/(1 − az) is a non-vanishing function on the
simply connected region ϕ(W ) and hence we can define an analytic square root
r
z−a
τ (z) =
1 − az

for z ∈ ϕ(W ). Set

τ (z) − τ (0)
σ(z) = .
1 − τ (0)τ (z)
Then σ ◦ ϕ is an analytic function mapping W into D. Moreover σ −1 is an analytic
function defined on all of D, mapping D into D with σ −1 (0) = 0 and it is not a rotation.
By Schwarz’s lemma |σ −1 (z)| < |z| and so

|σ ◦ ϕ(z)| > |ϕ(z)|. (20)

By the argument used to establish (9) and (11), gW + log |σ ◦ ϕ| ≤ 0. But gW (p, p0 ) =
− log |ϕ(p)|, so that log |σ ◦ ϕ| ≤ log |ϕ|, contradicting (20). 

The second place where RMT is used is at the conclusion of the proof in Case 2, where
there is no Green’s function. We proved that there is a conformal map ϕ of W onto a
subset of C∗ . If ϕ(W ) omitted at least two points in C∗ , then we used the RMT to prove
there is a conformal map onto the disk and thus contradict Theorem 4. To avoid RMT,
we can simply use a part of its proof. If there are at least two points a, b ∈ C∗ \ ϕ(W ) then

z−a
ψ(z) =
z−b
 §2: Comments. 269

is a nonvanishing analytic function on the simply connected region ϕ(W ), and hence we can
p
define an analytic square root σ(z) = ψ(z). It is not hard to check that ψ is one-to-one,
and if ψ(ϕ(W )) covers a neighborhood of z1 , z1 6= 0, then ψ(ϕ(W )) omits a neighborhood
of −z1 . We can then apply a linear fractional transformation σ so that σ(ψ(ϕ(W ))) ⊂ D.
If gD is Green’s function on D then the Perron family Fp0 for constructing Green’s function
on W is bounded by gD (f (p), f (p0 )) where f = σ ◦ ψ ◦ ϕ. Thus Green’s function exists,
contradicting our assumption.
270 XVIII. Equivalent Definitions of Analytic

XVIII

Equivalent Definitions of Analytic

§1. Goursat’s Proof of Cauchy’s Theorem

A holomorphic function, by definition, has a complex derivative which is continuous.

The hypothesis in the next theorem is slightly weaker than holomorphic in that there is
no assumption of continuity of the derivative.

Theorem 1.1 (Cauchy’s Theorem for a Rectangle). Suppose f has a complex deriva-
tive at each point of a closed rectangle R. Then
Z
f (z)dz = 0.
∂R

Proof. (Goursat) The idea of the proof is that the conclusion holds for linear functions by
the Fundamental Theorem of Calculus, since linear functions are derivatives of quadratic
functions. The rectangle R can be split into small rectangles and approximated by a linear
function on each.
Because f has a complex derivative, it is continuous on R, so we can integrate over
R
curves. Let C = | ∂R f (z)dz|. Divide the rectangle R into four similar rectangles whose
interiors are disjoint. The integral of f around the boundary of R equals the sum of the
integrals around the boundaries of the four smaller rectangles. By the triangle inequality,
there is at least one of these four rectangles, call it R1 , with
Z 
  C
 f (z)dz  ≥ .
 4
∂R1

Note that ℓ(∂R1 ) = 21 ℓ(∂R), where ℓ(P ) denotes length of P . Repeat this process on R1 :
divide R1 into four similar rectangles whose interiors are disjoint. Find one, call it R2 , so
that Z  Z 
  1  C
  
f (z)dz  ≥  f (z)dz  ≥ 2 ,
 4 ∂R1 4
∂R2
 §4: Exercises 271

and ℓ(∂R2 ) = 2−1 ℓ(∂R1 ) = 2−2 ℓ(∂R). By induction we can find a sequence of rectangles
Rj so that Rj+1 ⊂ Rj and Z 
 
 f (z)dz  ≥ C4−n ,

∂Rn

and
ℓ(∂Rn ) = 2−n ℓ(∂R).

Since the rectangles are nested, there is a point b ∈ ∩Rn ⊂ R. Since f has a complex
derivative at b, given ε > 0 we can find a δ > 0 so that
 
 f (z) − f (b) 
 − f (b) < ε,
′
 z−b

provided |z − b| < δ. For n sufficiently large, if z ∈ Rn then

|f (z) − [f (b) + (z − b)f ′ (b)]| ≤ ε|z − b| ≤ ε 2−n ℓ(∂R).

Thus Z  Z 
   
C =  f (z)dz  ≤ 4 
n
f (z)dz 
∂R ∂Rn

Z 
n

=4  (f (z) − [f (b) + f (b)(z − b)])dz 
′
∂Rn

≤ 4n ℓ(∂Rn ) sup |f (z) − [f (b) + f ′ (b)(z − b)]|

z∈∂Rn

≤ 4n 2−n ℓ(∂R) ε2−n ℓ(∂R) = εℓ(∂R)2 .

R
Since ε > 0 was arbitrary, we must have C = 0 and ∂R f (z)dz = 0. 

Corollary 1.2. If f has a complex derivative at each point z in a region Ω then f is

holomorphic and hence analytic in Ω.

Proof. Apply Theorem 1.1 and Morera’s Theorem IV.2.9. 

§2. Weyl’s Lemma

272 XVIII. Equivalent Definitions of Analytic

§3. Conditions Equivalent to Analytic

§4. Exercises.
 §4: Exercises 273

1. Let Z
2
f (z) = ew dw
Cz

where Cz is the radial line segment from 0 to z. Prove: f is entire, non-constant, f maps
the plane onto itself, f ′ (z) is never zero, yet f is not one-to-one. (This problem is related
to the “monodromy theorem”, which we will study later this year. We can define f −1
locally but cannot extend it as an analytic function on the whole plane.)

2. Suppose f is a non-constant entire function. Prove that there are at least three solutions
z to
f (z)3 + 4f (z) + 1 = 0.