Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
Complex Analysis
Version 1.23
Matthias Beck, Gerald Marchesi, and Dennis Pixton
Department of Mathematics Department of Mathematical Sciences
San Francisco State University Binghamton University (SUNY)
San Francisco, CA 94132 Binghamton, NY 139026000
beck@math.sfsu.edu marchesi@math.binghamton.edu
dennis@math.binghamton.edu
Copyright 2002–2007 by the authors. All rights reserved. The most current version of this book is
available at the websites
http://www.math.binghamton.edu/dennis/complex.pdf
http://math.sfsu.edu/beck/complex.html.
This book may be freely reproduced and distributed, provided that it is reproduced in its entirety
from the most recent version. This book may not be altered in any way, except for changes in
format required for printing or other distribution, without the permission of the authors.
2
These are the lecture notes of a onesemester undergraduate course which we have taught several
times at Binghamton University (SUNY) and San Francisco State University. For many of our
students, complex analysis is their ﬁrst rigorous analysis (if not mathematics) class they take,
and these notes reﬂect this very much. We tried to rely on as few concepts from real analysis as
possible. In particular, series and sequences are treated “from scratch.” This also has the (maybe
disadvantageous) consequence that power series are introduced very late in the course.
We thank our students who made many suggestions for and found errors in the text. Special
thanks go to Collin Bleak, Jon Clauss, Sharma Pallekonda, and Joshua Palmatier for comments
after teaching from this book.
Contents
1 Complex Numbers 1
1.1 Deﬁnition and Algebraic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Geometric Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Elementary Topology of the Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Theorems from Calculus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2 Diﬀerentiation 13
2.1 First Steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Diﬀerentiability and Analyticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 The Cauchy–Riemann Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4 Constants and Connectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3 Examples of Functions 23
3.1 M¨obius Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Inﬁnity and the Cross Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3 Exponential and Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . . 28
3.4 The Logarithm and Complex Exponentials . . . . . . . . . . . . . . . . . . . . . . . 31
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4 Integration 37
4.1 Deﬁnition and Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Antiderivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.3 Cauchy’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.4 Cauchy’s Integral Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5 Consequences of Cauchy’s Theorem 48
5.1 Extensions of Cauchy’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.2 Taking Cauchy’s Formula to the Limit . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.3 Antiderivatives Revisited and Morera’s Theorem . . . . . . . . . . . . . . . . . . . . 53
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3
CONTENTS 4
6 Harmonic Functions 57
6.1 Deﬁnition and Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
6.2 MeanValue and Maximum/Minimum Principle . . . . . . . . . . . . . . . . . . . . . 59
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
7 Power Series 62
7.1 Sequences and Completeness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
7.2 Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.3 Sequences and Series of Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
7.4 Region of Convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
8 Taylor and Laurent Series 74
8.1 Power Series and Analytic Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
8.2 Classiﬁcation of Zeros and the Identity Principle . . . . . . . . . . . . . . . . . . . . 77
8.3 Laurent Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
9 Isolated Singularities and the Residue Theorem 85
9.1 Classiﬁcation of Singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
9.2 Residues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
9.3 Argument Principle and Rouch´e’s Theorem . . . . . . . . . . . . . . . . . . . . . . . 91
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
10 Discreet Applications of the Residue Theorem 96
10.1 Inﬁnite Sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
10.2 Binomial Coeﬃcients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
10.3 Fibonacci Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
10.4 The ‘CoinExchange Problem’ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
10.5 Dedekind sums . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Solutions to Selected Exercises 101
Index 104
Chapter 1
Complex Numbers
Die ganzen Zahlen hat der liebe Gott geschaﬀen, alles andere ist Menschenwerk.
(God created the integers, everything else is made by humans.)
Leopold Kronecker (1823–1891)
1.1 Deﬁnition and Algebraic Properties
The complex numbers can be deﬁned as pairs of real numbers,
C = {(x, y) : x, y ∈ R} ,
equipped with the addition
(x, y) + (a, b) = (x +a, y +b)
and the multiplication
(x, y) · (a, b) = (xa −yb, xb +ya) .
One reason to believe that the deﬁnitions of these binary operations are “good” is that C is an
extension of R, in the sense that the complex numbers of the form (x, 0) behave just like real
numbers; that is, (x, 0) + (y, 0) = (x + y, 0) and (x, 0) · (y, 0) = (x · y, 0). So we can think of the
real numbers being embedded in C as those complex numbers whose second coordinate is zero.
The following basic theorem states the algebraic structure that we established with our deﬁni
tions. Its proof is straightforward but nevertheless a good exercise.
Theorem 1.1. (C, +, ·) is a ﬁeld; that is:
∀ (x, y), (a, b) ∈ C : (x, y) + (a, b) ∈ C (1.1)
∀ (x, y), (a, b), (c, d) ∈ C :
_
(x, y) + (a, b)
_
+ (c, d) = (x, y) +
_
(a, b) + (c, d)
_
(1.2)
∀ (x, y), (a, b) ∈ C : (x, y) + (a, b) = (a, b) + (x, y) (1.3)
∀ (x, y) ∈ C : (x, y) + (0, 0) = (x, y) (1.4)
∀ (x, y) ∈ C : (x, y) + (−x, −y) = (0, 0) (1.5)
1
CHAPTER 1. COMPLEX NUMBERS 2
∀ (x, y), (a, b) ∈ C : (x, y) · (a, b) ∈ C (1.6)
∀ (x, y), (a, b), (c, d) ∈ C :
_
(x, y) · (a, b)
_
· (c, d) = (x, y) ·
_
(a, b) · (c, d)
_
(1.7)
∀ (x, y), (a, b) ∈ C : (x, y) · (a, b) = (a, b) · (x, y) (1.8)
∀ (x, y) ∈ C : (x, y) · (1, 0) = (x, y) (1.9)
∀ (x, y) ∈ C \ {(0, 0)} : (x, y) ·
_
x
x
2
+y
2
,
−y
x
2
+y
2
_
= (1, 0) (1.10)
Remark. What we are stating here can be compressed in the language of algebra: equations (1.1)–
(1.5) say that (C, +) is an Abelian group with unit element (0, 0), equations (1.6)–(1.10) that
(C \ {(0, 0)}, ·) is an abelian group with unit element (1, 0). (If you don’t know what these terms
mean—don’t worry, we will not have to deal with them.)
The deﬁnition of our multiplication implies the innocent looking statement
(0, 1) · (0, 1) = (−1, 0) . (1.11)
This identity together with the fact that
(a, 0) · (x, y) = (ax, ay)
allows an alternative notation for complex numbers. The latter implies that we can write
(x, y) = (x, 0) + (0, y) = (x, 0) · (1, 0) + (y, 0) · (0, 1) .
If we think—in the spirit of our remark on the embedding of R in C—of (x, 0) and (y, 0) as the
real numbers x and y, then this means that we can write any complex number (x, y) as a linear
combination of (1, 0) and (0, 1), with the real coeﬃcients x and y. (1, 0), in turn, can be thought
of as the real number 1. So if we give (0, 1) a special name, say i, then the complex number that
we used to call (x, y) can be written as x · 1 +y · i, or in short,
x +iy .
The number x is called the real part and y the imaginary part
1
of the complex number x+iy, often
denoted as Re(x +iy) = x and Im(x +iy) = y. The identity (1.11) then reads
i
2
= −1 .
We invite the reader to check that the deﬁnitions of our binary operations and Theorem 1.1 are
coherent with the usual real arithmetic rules if we think of complex numbers as given in the form
x +iy.
1.2 Geometric Properties
Although we just introduced a new way of writing complex numbers, let’s for a moment return to
the (x, y)notation. It suggests that one can think of a complex number as a twodimensional real
vector. When plotting these vectors in the plane R
2
, we will call the xaxis the real axis and the
yaxis the imaginary axis. The addition that we deﬁned for complex numbers resembles vector
addition. The analogy stops at multiplication: there is no “usual” multiplication of two vectors
CHAPTER 1. COMPLEX NUMBERS 3
DD
W
W
W
W
W
W
W
W
W
W
W
W
kk
/
/
/
/
/
/
/
/
/
/
/
/
/
WW
z
1
z
2
z
1
+z
2
Figure 1.1: Addition of complex numbers.
that gives another vector—much less so if we additionally demand our deﬁnition of the product of
two complex numbers.
Any vector in R
2
is deﬁned by its two coordinates. On the other hand, it is also determined
by its length and the angle it encloses with, say, the positive real axis; let’s deﬁne these concepts
thoroughly. The absolute value (sometimes also called the modulus) of x +iy is
r = x +iy =
_
x
2
+y
2
,
and an argument of x +iy is a number φ such that
x = r cos φ and y = r sin φ.
This means, naturally, that any complex number has many arguments; more precisely, all of them
diﬀer by a multiple of 2π.
The absolute value of the diﬀerence of two vectors has a nice geometric interpretation: it is
the distance of the (end points of the) two vectors (see Figure 1.2). It is very useful to keep this
geometric interpretation in mind when thinking about the absolute value of the diﬀerence of two
complex numbers.
DD
W
W
W
W
W
W
W
W
W
W
W
W
kk
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
j
44
z
1
z
2
z
1
−z
2
Figure 1.2: Geometry behind the “distance” between two complex numbers.
The ﬁrst hint that absolute value and argument of a complex number are useful concepts
is the fact that they allow us to give a geometric interpretation for the multiplication of two
complex numbers. Let’s say we have two complex numbers, x
1
+ iy
1
with absolute value r
1
and
argument φ
1
, and x
2
+ iy
2
with absolute value r
2
and argument φ
2
. This means, we can write
1
The name has historical reasons: people thought of complex numbers as unreal, imagined.
CHAPTER 1. COMPLEX NUMBERS 4
x
1
+iy
1
= (r
1
cos φ
1
) +i(r
1
sin φ
1
) and x
2
+iy
2
= (r
2
cos φ
2
) +i(r
2
sin φ
2
) To compute the product,
we make use of some classic trigonometric identities:
(x
1
+iy
1
)(x
2
+iy
2
) =
_
(r
1
cos φ
1
) +i(r
1
sin φ
1
)
__
(r
2
cos φ
2
) +i(r
2
sin φ
2
)
_
= (r
1
r
2
cos φ
1
cos φ
2
−r
1
r
2
sin φ
1
sin φ
2
) +i(r
1
r
2
cos φ
1
sin φ
2
+r
1
r
2
sin φ
1
cos φ
2
)
= r
1
r
2
_
(cos φ
1
cos φ
2
−sin φ
1
sin φ
2
) +i(cos φ
1
sin φ
2
+ sin φ
1
cos φ
2
)
_
= r
1
r
2
_
cos(φ
1
+φ
2
) +i sin(φ
1
+φ
2
)
_
.
So the absolute value of the product is r
1
r
2
and (one of) its argument is φ
1
+φ
2
. Geometrically, we
are multiplying the lengths of the two vectors representing our two complex numbers, and adding
their angles measured with respect to the positive xaxis.
2
FF M
M
M
M
M
M
M
M
ff
r
r
r
r
r
r
r
r
r
r
r
r
r
r
r
r
r
xx
. .
. .
.
. .
.
. .
.
. .
. .
.
.
.
. .
.
. .
.
. .
.
. .
.
. .
.
. .
. . . . . . . . .
.
. .
.
. .
.
. .
.
.
.
.
.
.
. .
.
.
.
. .
.
.
.
. .
.
.
.
. .
.
.
.
. .
.
.
.
. .
. . . . . . . . . . . . . . . .
.
.
. .
.
.
.
. .
.
.
.
. .
.
.
.
. .
.
.
.
. .
.
.
.
. .
.
.
.
. .
.
.
.
. .
.
.
.
z
1
z
2
z
1
z
2
φ
1
φ
2
φ
1
+φ
2
Figure 1.3: Multiplication of complex numbers.
In view of the above calculation, it should come as no surprise that we will have to deal with
quantities of the form cos φ + i sin φ (where φ is some real number) quite a bit. To save space,
bytes, ink, etc., (and because “Mathematics is for lazy people”
3
) we introduce a shortcut notation
and deﬁne
e
iφ
= cos φ +i sin φ.
At this point, this exponential notation is indeed purely a notation. We will later see that it has
an intimate connection to the complex exponential function. For now, we motivate this maybe
strangeseeming deﬁnition by collecting some of its properties. The reader is encouraged to prove
them.
Lemma 1.2. For any φ, φ
1
, φ
2
∈ R,
(a) e
iφ
1
e
iφ
2
= e
i(φ
1
+φ
2
)
(b) 1/e
iφ
= e
−iφ
(c) e
i(φ+2π)
= e
iφ
(d)
¸
¸
e
iφ
¸
¸
= 1
2
One should convince oneself that there is no problem with the fact that there are many possible arguments for
complex numbers, as both cosine and sine are periodic functions with period 2π.
3
Peter Hilton (Invited address, Hudson River Undergraduate Mathematics Conference 2000)
CHAPTER 1. COMPLEX NUMBERS 5
(e)
d
dφ
e
iφ
= i e
iφ
.
With this notation, the sentence “The complex number x+iy has absolute value r and argument
φ” now becomes the identity
x +iy = re
iφ
.
The lefthand side is often called the rectangular form, the righthand side the polar form of this
complex number.
From very basic geometric properties of triangles, we get the inequalities
−z ≤ Re z ≤ z and −z ≤ Imz ≤ z . (1.12)
The square of the absolute value has the nice property
x +iy
2
= x
2
+y
2
= (x +iy)(x −iy) .
This is one of many reasons to give the process of passing from x + iy to x − iy a special name:
x −iy is called the (complex) conjugate of x +iy. We denote the conjugate by
x +iy = x −iy .
Geometrically, conjugating z means reﬂecting the vector corresponding to z with respect to the
real axis. The following collects some basic properties of the conjugate. Their easy proofs are left
for the exercises.
Lemma 1.3. For any z, z
1
, z
2
∈ C,
(a) z
1
±z
2
= z
1
±z
2
(b) z
1
· z
2
= z
1
· z
2
(c)
_
z
1
z
2
_
=
z
1
z
2
(d) z = z
(e) z = z
(f) z
2
= zz
(g) Re z =
1
2
(z +z)
(h) Imz =
1
2i
(z −z)
(i) e
iφ
= e
−iφ
.
From part (f) we have a neat formula for the inverse of a nonzero complex number:
z
−1
=
1
z
=
z
z
2
.
A famous geometric inequality (which holds for vectors in R
n
) is the triangle inequality
z
1
+z
2
 ≤ z
1
 +z
2
 .
CHAPTER 1. COMPLEX NUMBERS 6
By drawing a picture in the complex plane, you should be able to come up with a geometric proof
of this inequality. To prove it algebraically, we make extensive use of Lemma 1.3:
z
1
+z
2

2
= (z
1
+z
2
) (z
1
+z
2
)
= (z
1
+z
2
) (z
1
+z
2
)
= z
1
z
1
+z
1
z
2
+z
2
z
1
+z
2
z
2
= z
1

2
+z
1
z
2
+z
1
z
2
+z
2

2
= z
1

2
+ 2 Re (z
1
z
2
) +z
2

2
.
Finally by (1.12)
z
1
+z
2

2
≤ z
1

2
+ 2 z
1
z
2
 +z
2

2
= z
1

2
+ 2 z
1
 z
2
 +z
2

2
= z
1

2
+ 2 z
1
 z
2
 +z
2

2
= (z
1
 +z
2
)
2
,
which is equivalent to our claim.
For future reference we list several variants of the triangle inequality:
Lemma 1.4. For z
1
, z
2
, · · · ∈ C, we have the following identities:
(a) The triangle inequality: ±z
1
±z
2
 ≤ z
1
 +z
2
.
(b) The reverse triangle inequality: ±z
1
±z
2
 ≥ z
1
 −z
2
.
(c) The triangle inequality for sums:
¸
¸
¸
¸
¸
n
k=1
z
k
¸
¸
¸
¸
¸
≤
n
k=1
z
k
.
The ﬁrst inequality is just a rewrite of the original triangle inequality, using the fact that
±z = z, and the last follows by induction. The reverse triangle inequality is proved in Exercise 15.
1.3 Elementary Topology of the Plane
In Section 1.2 we saw that the complex numbers C, which were initially deﬁned algebraically, can
be identiﬁed with the points in the Euclidean plane R
2
. In this section we collect some deﬁnitions
and results concerning the topology of the plane. While the deﬁnitions are essential and will be
used frequently, we will need the following theorems only at a limited number of places in the
remainder of the book; the reader who is willing to accept the topological arguments in later proofs
on faith may skip the theorems in this section.
Recall that if z, w ∈ C, then z −w is the distance between z and w as points in the plane. So
if we ﬁx a complex number a and a positive real number r then the set of z satisfying z −a = r
is the set of points at distance r from a; that is, this is the circle with center a and radius r. The
inside of this circle is called the open disk with center a and radius r, and is written D
r
(a). That
is, D
r
(a) = {z ∈ C : z −a < r}. Notice that this does not include the circle itself.
We need some terminology for talking about subsets of C.
CHAPTER 1. COMPLEX NUMBERS 7
Deﬁnition 1.1. Suppose E is any subset of C.
(a) A point a is an interior point of E if some open disk with center a lies in E.
(b) A point b is a boundary point of E if every open disk centered at b contains a point in E and
also a point that is not in E.
(c) A point c is an accumulation point of E if every open disk centered at c contains a point of E
diﬀerent from c.
(d) A point d is an isolated point of E if it lies in E and some open disk centered at d contains no
point of E other than d.
The idea is that if you don’t move too far from an interior point of E then you remain in E;
but at a boundary point you can make an arbitrarily small move and get to a point inside E and
you can also make an arbitrarily small move and get to a point outside E.
Deﬁnition 1.2. A set is open if all its points are interior points. A set is closed if it contains all
its boundary points.
Example 1.1. For R > 0 and z
0
∈ C, {z ∈ C : z −z
0
 < R} and {z ∈ C : z −z
0
 > R} are open.
{z ∈ C : z −z
0
 ≤ R} is closed.
Example 1.2. C and the empty set ∅ are open. They are also closed!
Deﬁnition 1.3. The boundary of a set E, written ∂E, is the set of all boundary points of E. The
interior of E is the set of all interior points of E. The closure of E, written E, is the set of points
in E together with all boundary points of E.
Example 1.3. If G is the open disk {z ∈ C : z −z
0
 < R} then
G = {z ∈ C : z −z
0
 ≤ R} and ∂G = {z ∈ C : z −z
0
 = R} .
That is, G is a closed disk and ∂G is a circle.
One notion that is somewhat subtle in the complex domain is the idea of connectedness. Intu
itively, a set is connected if it is “in one piece.” In the reals a set is connected if and only if it is an
interval, so there is little reason to discuss the matter. However, in the plane there is a vast variety
of connected subsets, so a deﬁnition is necessary.
Deﬁnition 1.4. Two sets X, Y ⊆ C are separated if there are disjoint open sets A and B so that
X ⊆ A and Y ⊆ B. A set W ⊆ C is connected if it is impossible to ﬁnd two separated nonempty
sets whose union is equal to W. A region is a connected open set.
The idea of separation is that the two open sets A and B ensure that X and Y cannot just
“stick together.” It is usually easy to check that a set is not connected. For example, the intervals
X = [0, 1) and Y = (1, 2] on the real axis are separated: There are inﬁnitely many choices for A and
B that work; one choice is A = D
1
(0) (the open disk with center 0 and radius 1) and B = D
1
(2)
(the open disk with center 2 and radius 1). Hence their union, which is [0, 2] \{1}, is not connected.
On the other hand, it is hard to use the deﬁnition to show that a set is connected, since we have
to rule out any possible separation.
One type of connected set that we will use frequently is a curve.
CHAPTER 1. COMPLEX NUMBERS 8
Deﬁnition 1.5. A path or curve in C is the image of a continuous function γ : [a, b] → C, where
[a, b] is a closed interval in R. The path γ is smooth if γ is diﬀerentiable.
We say that the curve is parametrized by γ. It is a customary and practical abuse of notation
to use the same letter for the curve and its parametrization. We emphasize that a curve must have
a parametrization, and that the parametrization must be deﬁned and continuous on a closed and
bounded interval [a, b].
Since we may regard C as identiﬁed with R
2
, a path can be speciﬁed by giving two continuous
realvalued functions of a real variable, x(t) and y(t), and setting γ(t) = x(t) + y(t)i. A curve is
closed if γ(a) = γ(b) and is a simple closed curve if γ(s) = γ(t) implies s = a and t = b or s = b
and t = a, that is, the curve does not cross itself.
The following seems intuitively clear, but its proof requires more preparation in topology:
Proposition 1.5. Any curve is connected.
The next theorem gives an easy way to check whether an open set is connected, and also gives
a very useful property of open connected sets.
Theorem 1.6. If W is a subset of C that has the property that any two points in W can be
connected by a curve in W then W is connected. On the other hand, if G is a connected open
subset of C then any two points of G may be connected by a curve in G; in fact, we can connect
any two points of G by a chain of horizontal and vertical segments lying in G.
A chain of segments in G means the following: there are points z
0
, z
1
, . . . , z
n
so that, for each
k, z
k
and z
k+1
are the endpoints of a horizontal or vertical segment which lies entirely in G. (It is
not hard to parametrize such a chain, so it determines is a curve.)
As an example, let G be the open disk with center 0 and radius 2. Then any two points in G can
be connected by a chain of at most 2 segments in G, so G is connected. Now let G
0
= G\ {0}; this
is the punctured disk obtained by removing the center from G. Then G is open and it is connected,
but now you may need more than two segments to connect points. For example, you need three
segments to connect −1 to 1 since you cannot go through 0.
Warning: The second part of Theorem 1.6 is not generally true if G is not open. For example,
circles are connected but there is no way to connect two distinct points of a circle by a chain of
segments which are subsets of the circle. A more extreme example, discussed in topology texts, is
the “topologist’s sine curve,” which is a connected set S ⊂ C that contains points that cannot be
connected by a curve of any sort inside S.
The reader may skip the following proof. It is included to illustrate some common techniques
in dealing with connected sets.
Proof of Theorem 1.6. Suppose, ﬁrst, that any two points of G may be connected by a path that
lies in G. If G is not connected then we can write it as a union of two nonempty separated subsets
X and Y . So there are disjoint open sets A and B so that X ⊆ A and Y ⊆ B. Since X and Y are
disjoint we can ﬁnd a ∈ X and b ∈ G. Let γ be a path in G that connects a to b. Then X
γ
= X∩γ
and Y
γ
= Y ∩ γ are disjoint and nonempty, their union is γ, and they are separated by A and B.
But this means that γ is not connected, and this contradicts Proposition 1.5.
CHAPTER 1. COMPLEX NUMBERS 9
Now suppose that G is a connected open set. Choose a point z
0
∈ G and deﬁne two sets: A is
the set of all points a so that there is a chain of segments in G connecting z
0
to a, and B is the set
of points in G that are not in A.
Suppose a is in A. Since a ∈ G there is an open disk D with center a that is contained in G.
We can connect z
0
to any point z in D by following a chain of segments from z
0
to a, and then
adding at most two segments in D that connect a to z. That is, each point of D is in A, so we
have shown that A is open.
Now suppose b is in B. Since b ∈ G there is an open disk D centered at b that lies in G. If z
0
could be connected to any point in D by a chain of segments in G then, extending this chain by at
most two more segments, we could connect z
0
to b, and this is impossible. Hence z
0
cannot connect
to any point of D by a chain of segments in G, so D ⊆ B. So we have shown that B is open.
Now G is the disjoint union of the two open sets A and B. If these are both nonempty then
they form a separation of G, which is impossible. But z
0
is in A so A is not empty, and so B must
be empty. That is, G = A, so z
0
can be connected to any point of G by a sequence of segments in
G. Since z
0
could be any point in G, this ﬁnishes the proof.
1.4 Theorems from Calculus
Here are a few theorems from real calculus that we will make use of in the course of the text.
Theorem 1.7 (ExtremeValue Theorem). Any continuous realvalued function deﬁned on a closed
and bounded subset of R
n
has a minimum value and a maximum value.
Theorem 1.8 (MeanValue Theorem). Suppose I ⊆ R is an interval, f : I → R is diﬀerentiable,
and x, x + ∆x ∈ I. Then there is 0 < a < 1 such that
f(x + ∆x) −f(x)
∆x
= f
(x +a∆x) .
Many of the most important results of analysis concern combinations of limit operations. The
most important of all calculus theorems combines diﬀerentiation and integration (in two ways):
Theorem 1.9 (Fundamental Theorem of Calculus). Suppose f : [a, b] →R is continuous. Then
(a) If F is deﬁned by F(x) =
_
x
a
f(t) dt then F is diﬀerentiable and F
(x) = f(x).
(b) If F is any antiderivative of f (that is, F
= f) then
_
b
a
f(x) dx = F(b) −F(a).
For functions of several variables we can perform diﬀerentiation operations, or integration op
erations, in any order, if we have suﬃcient continuity:
Theorem 1.10 (Equality of mixed partials). If the mixed partials
∂
2
f
∂x∂y
and
∂
2
f
∂y∂x
are deﬁned on
an open set G and are continuous at a point (x
0
, y
0
) in G then they are equal at (x
0
, y
0
).
Theorem 1.11 (Equality of iterated integrals). If f is continuous on the rectangle given by a ≤
x ≤ b and c ≤ y ≤ d then the iterated integrals
_
b
a
_
d
c
f(x, y) dy dx and
_
d
c
_
b
a
f(x, y) dxdy are equal.
Finally, we can apply diﬀerentiation and integration with respect to diﬀerent variables in either
order:
CHAPTER 1. COMPLEX NUMBERS 10
Theorem 1.12 (Leibniz’s
4
Rule). Suppose f is continuous on the rectangle R given by a ≤ x ≤ b
and c ≤ y ≤ d, and suppose the partial derivative
∂f
∂x
exists and is continuous on R. Then
d
dx
_
d
c
f(x, y) dy =
_
d
c
∂f
∂x
(x, y) dy .
Exercises
1. Find the real and imaginary parts of each of the following:
(a)
z−a
z+a
(a ∈ R).
(b)
3+5i
7i+1
.
(c)
_
−1+i
√
3
2
_
3
.
(d) i
n
for any n ∈ Z.
2. Find the absolute value and conjugate of each of the following:
(a) −2 +i.
(b) (2 +i)(4 + 3i).
(c)
3−i
√
2+3i
.
(d) (1 +i)
6
.
3. Write in polar form:
(a) 2i.
(b) 1 +i.
(c) −3 +
√
3i.
4. Write in rectangular form:
(a)
√
2 e
i3π/4
.
(b) 34 e
iπ/2
.
(c) −e
i250π
.
5. Find all solutions to the following equations:
(a) z
6
= 1.
(b) z
4
= −16.
(c) z
6
= −9.
(d) z
6
−z
3
−2 = 0.
6. Show that
4
Named after Gottfried Wilhelm Leibniz (1646–1716). For more information about Leibnitz, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Leibnitz.html.
CHAPTER 1. COMPLEX NUMBERS 11
(a) z is a real number if and only if z = z;
(b) z is either real or purely imaginary if and only if (z)
2
= z
2
.
7. Find all solutions of the equation z
2
+ 2z + (1 −i) = 0.
8. Prove Theorem 1.1.
9. Show that if z
1
z
2
= 0 then z
1
= 0 or z
2
= 0.
10. Prove Lemma 1.2.
11. Use Lemma 1.2 to derive the triple angle formulas:
(a) cos 3θ = cos
3
θ −3 cos θ sin
2
θ.
(b) sin 3θ = 3 cos
2
θ sin θ −sin
3
θ.
12. Prove Lemma 1.3.
13. Sketch the following sets in the complex plane:
(a) {z ∈ C : z −1 +i = 2} .
(b) {z ∈ C : z −1 +i ≤ 2} .
(c) {z ∈ C : Re(z + 2 −2i) = 3} .
(d) {z ∈ C : z −i +z +i = 3} .
14. Suppose p is a polynomial with real coeﬃcients. Prove that
(a) p(z) = p (z).
(b) p(z) = 0 if and only if p (z) = 0.
15. Prove the reverse triangle inequality z
1
−z
2
 ≥ z
1
 −z
2
.
16. Use the previous exercise to show that
¸
¸
¸
1
z
2
−1
¸
¸
¸ ≤
1
3
for every z on the circle z = 2e
iθ
.
17. Sketch the following sets and determine whether they are open, closed, or neither; bounded;
connected.
(a) z + 3 < 2.
(b) Imz < 1.
(c) 0 < z −1 < 2.
(d) z −1 +z + 1 = 2.
(e) z −1 +z + 1 < 3.
18. What are the boundaries of the sets in the previous exercise?
19. The set E is the set of points z in C satisfying either z is real and −2 < z < −1, or z < 1,
or z = 1 or z = 2.
CHAPTER 1. COMPLEX NUMBERS 12
(a) Sketch the set E, being careful to indicate exactly the points that are in E.
(b) Determine the interior points of E.
(c) Determine the boundary points of E.
(d) Determine the isolated points of E.
20. The set E in the previous exercise can be written in three diﬀerent ways as the union of two
disjoint nonempty separated subsets. Describe them, and in each case say brieﬂy why the
subsets are separated.
21. Let G be the annulus determined by the conditions 2 < z < 3. This is a connected open
set. Find the maximum number of horizontal and vertical segments in G needed to connect
two points of G.
22. Prove Leibniz’s Rule: Deﬁne F(x) =
_
d
c
f(x, y) dy, get an expression for F(x) − F(a) as an
iterated integral by writing f(x, y) − f(a, y) as the integral of
∂f
∂x
, interchange the order of
integrations, and then diﬀerentiate using the Fundamental Theorem of Calculus.
Chapter 2
Diﬀerentiation
Mathematical study and research are very suggestive of mountaineering. Whymper made several
eﬀorts before he climbed the Matterhorn in the 1860’s and even then it cost the life of four of
his party. Now, however, any tourist can be hauled up for a small cost, and perhaps does not
appreciate the diﬃculty of the original ascent. So in mathematics, it may be found hard to
realise the great initial diﬃculty of making a little step which now seems so natural and obvious,
and it may not be surprising if such a step has been found and lost again.
Louis Joel Mordell (1888–1972)
2.1 First Steps
A (complex) function f is a mapping from a subset G ⊆ C to C (in this situation we will write
f : G → C and call G the domain of f). This means that each element z ∈ G gets mapped to
exactly one complex number, called the image of z and usually denoted by f(z). So far there is
nothing that makes complex functions any more special than, say, functions from R
m
to R
n
. In
fact, we can construct many familiar looking functions from the standard calculus repertoire, such
as f(z) = z (the identity map), f(z) = 2z + i, f(z) = z
3
, or f(z) =
1
z
. The former three could be
deﬁned on all of C, whereas for the latter we have to exclude the origin z = 0. On the other hand,
we could construct some functions which make use of a certain representation of z, for example,
f(x, y) = x −2iy, f(x, y) = y
2
−ix, or f(r, φ) = 2re
i(φ+π)
.
Maybe the fundamental principle of analysis is that of a limit. The philosophy of the following
deﬁnition is not restricted to complex functions, but for sake of simplicity we only state it for those
functions.
Deﬁnition 2.1. Suppose f is a complex function with domain G and z
0
is an accumulation point
of G. Suppose there is a complex number w
0
such that for every > 0, we can ﬁnd δ > 0 so that
for all z ∈ G satisfying 0 < z −z
0
 < δ we have f(z) −w
0
 < . Then w
0
is the limit of f as z
approaches z
0
, in short
lim
z→z
0
f(z) = w
0
.
This deﬁnition is the same as is found in most calculus texts. The reason we require that z
0
is
an accumulation point of the domain is just that we need to be sure that there are points z of the
domain which are arbitrarily close to z
0
. Just as in the real case, the deﬁnition does not require
13
CHAPTER 2. DIFFERENTIATION 14
that z
0
is in the domain of f and, if z
0
is in the domain of f, the deﬁnition explicitly ignores the
value of f(z
0
). That is why we require 0 < z −z
0
.
Just as in the real case the limit w
0
is unique if it exists. It is often useful to investigate limits
by restricting the way the point z “approaches” z
0
. The following is a easy consequence of the
deﬁnition.
Lemma 2.1. Suppose lim
z→z
0
f(z) exists and has the value w
0
, as above. Suppose G
0
⊆ G, and
suppose z
0
is an accumulation point of G
0
. If f
0
is the restriction of f to G
0
then lim
z→z
0
f
0
(z)
exists and has the value w
0
.
The deﬁnition of limit in the complex domain has to be treated with a little more care than its
real companion; this is illustrated by the following example.
Example 2.1. lim
z→0
¯ z
z
does not exist.
To see this, we try to compute this “limit” as z →0 on the real and on the imaginary axis. In the
ﬁrst case, we can write z = x ∈ R, and hence
lim
z→0
z
z
= lim
x→0
x
x
= lim
x→0
x
x
= 1 .
In the second case, we write z = iy where y ∈ R, and then
lim
z→0
z
z
= lim
y→0
iy
iy
= lim
y→0
−iy
iy
= −1 .
So we get a diﬀerent “limit” depending on the direction from which we approach 0. Lemma 2.1
then implies that lim
z→0
¯ z
z
does not exist.
On the other hand, the following “usual” limit rules are valid for complex functions; the proofs
of these rules are everything but trivial and make for nice exercises.
Lemma 2.2. Let f and g be complex functions and c, z
0
∈ C.
(a) lim
z→z
0
f(z) +c lim
z→z
0
g(z) = lim
z→z
0
(f(z) +c g(z))
(b) lim
z→z
0
f(z) · lim
z→z
0
g(z) = lim
z→z
0
(f(z) · g(z))
(c) lim
z→z
0
f(z)/ lim
z→z
0
g(z) = lim
z→z
0
(f(z)/g(z)) .
In the last identity we have to make sure we do not divide by zero.
Because the deﬁnition of the limit is somewhat elaborate, the following fundamental deﬁnition
looks almost trivial.
Deﬁnition 2.2. Suppose f is a complex function. If z
0
is in the domain of the function and either
z
0
is an isolated point of the domain or
lim
z→z
0
f(z) = f(z
0
)
then f is continuous at z
0
. More generally, f is continuous on G ⊆ C if f is continuous at every
z ∈ G.
CHAPTER 2. DIFFERENTIATION 15
Just as in the real case, we can “take the limit inside” a continuous function:
Lemma 2.3. If f is continuous at w
0
and lim
z→z
0
g(z) = w
0
then lim
z→z
0
f(g(z)) = f(w
0
). In
other words,
lim
z→z
0
f(g(z)) = f
_
lim
z→z
0
g(z)
_
.
2.2 Diﬀerentiability and Analyticity
The fact that limits such as lim
z→0
¯ z
z
do not exist points to something special about complex
numbers which has no parallel in the reals—we can express a function in a very compact way in
one variable, yet it shows some peculiar behavior “in the limit.” We will repeatedly notice this kind
of behavior; one reason is that when trying to compute a limit of a function as, say, z →0, we have
to allow z to approach the point 0 in any way. On the real line there are only two directions to
approach 0—from the left or from the right (or some combination of those two). In the complex
plane, we have an additional dimension to play with. This means that the statement “A complex
function has a limit...” is in many senses stronger than the statement “A real function has a limit...”
This diﬀerence becomes apparent most baldly when studying derivatives.
Deﬁnition 2.3. Suppose f : G → C is a complex function and z
0
is an interior point of G. The
derivative of f at z
0
is deﬁned as
f
(z
0
) = lim
z→z
0
f(z) −f(z
0
)
z −z
0
,
provided this limit exists. In this case, f is called diﬀerentiable at z
0
. If f is diﬀerentiable for all
points in an open disk centered at z
0
then f is called analytic at z
0
. The function f is analytic on
the open set G ⊆ C if it is diﬀerentiable (and hence analytic) at every point in G. Functions which
are diﬀerentiable (and hence analytic) in the whole complex plane C are called entire.
The diﬀerence quotient limit which deﬁnes f
(z
0
) can be rewritten as
f
(z
0
) = lim
h→0
f(z
0
+h) −f(z
0
)
h
.
This equivalent deﬁnition is sometimes easier to handle. Note that h is not a real number but can
rather approach zero from anywhere in the complex plane.
The fact that the notions of diﬀerentiability and analyticity are actually diﬀerent is seen in the
following examples.
Example 2.2. The function f(z) = z
3
is entire, that is, analytic in C: For any z
0
∈ C,
lim
z→z
0
f(z) −f(z
0
)
z −z
0
= lim
z→z
0
z
3
−z
3
0
z −z
0
= lim
z→z
0
(z
2
+zz
0
+z
2
0
)(z −z
0
)
z −z
0
= lim
z→z
0
z
2
+zz
0
+z
2
0
= 3z
2
0
.
Example 2.3. The function f(z) = z
2
is diﬀerentiable at 0 and nowhere else (in particular, f is not
analytic at 0): Let’s write z = z
0
+re
iφ
. Then
z
2
−z
0
2
z −z
0
=
_
z
0
+re
iφ
_
2
−z
0
2
z
0
+re
iφ
−z
0
=
_
z
0
+re
−iφ
_
2
z
0
2
re
iφ
=
z
0
2
+ 2z
0
re
−iφ
+r
2
e
−2iφ
−z
0
2
re
iφ
=
2z
0
re
−iφ
+r
2
e
−2iφ
re
iφ
= 2z
0
e
−2iφ
+re
−3iφ
.
CHAPTER 2. DIFFERENTIATION 16
If z
0
= 0 then the limit of the righthand side as z → z
0
does not exist since r → 0 and we get
diﬀerent answers for horizontal approach (φ = 0) and for vertical approach (φ = π/2). (A more
entertaining way to see this is to use, for example, z(t) = z
0
+
1
t
e
it
, which approaches z
0
as t →∞.)
On the other hand, if z
0
= 0 then the righthand side equals re
−3iφ
= ze
−3iφ
. Hence
lim
z→0
¸
¸
¸
¸
z
2
z
¸
¸
¸
¸
= lim
z→0
¸
¸
¸ze
−3iφ
¸
¸
¸ = lim
z→0
z = 0 ,
which implies that
lim
z→0
z
2
z
= 0 .
Example 2.4. The function f(z) = z is nowhere diﬀerentiable:
lim
z→z
0
z −z
0
z −z
0
= lim
z→z
0
z −z
0
z −z
0
= lim
z→0
z
z
does not exist, as discussed earlier.
The basic properties for derivatives are similar to those we know from real calculus. In fact, one
should convince oneself that the following rules follow mostly from properties of the limit. (The
‘chain rule’ needs a little care to be worked out.)
Lemma 2.4. Suppose f and g are diﬀerentiable at z ∈ C, and that c ∈ C, n ∈ Z, and h is
diﬀerentiable at g(z).
(a)
_
f(z) +c g(z)
_
= f
(z) +c g
(z)
(b)
_
f(z) · g(z)
_
= f
(z)g(z) +f(z)g
(z)
(c)
_
f(z)/g(z)
_
=
f
(z)g(z) −f(z)g
(z)
g(z)
2
(d)
_
z
n
_
= nz
n−1
(e)
_
h(g(z))
_
= h
(g(z))g
(z) .
In the third identity we have to be aware of division by zero.
We end this section with yet another diﬀerentiation rule, that for inverse functions. As in the
real case, this rule is only deﬁned for functions which are bijections. A function f : G → H is
onetoone if for every image w ∈ H there is a unique z ∈ G such that f(z) = w. The function is
onto if every w ∈ H has a preimage z ∈ G (that is, there exists a z ∈ G such that f(z) = w). A
bijection is a function which is both onetoone and onto. If f : G →H is a bijection then g is the
inverse of f if for all z ∈ H, f(g(z)) = z.
Lemma 2.5. Suppose G and H are open sets in C, f : G → H is a bijection, g : H → G is the
inverse function of f, and z
0
∈ H. If f is diﬀerentiable at g(z
0
), f
(g(z
0
)) = 0, and g is continuous
at z
0
then g is diﬀerentiable at z
0
with
g
(z
0
) =
1
f
(g(z
0
))
.
CHAPTER 2. DIFFERENTIATION 17
Proof. The function F deﬁned by
F(z) =
_
_
_
f(w) −f(w
0
)
w −w
0
if w = w
0
,
f
(w
0
) if w = w
0
is continuous at w
0
. This appears when we calculate g
(z
0
):
lim
z→z
0
g(z) −g(z
0
)
z −z
0
= lim
z→z
0
g(z) −g(z
0
)
f(g(z)) −f(g(z
0
))
= lim
z→z
0
1
f(g(z)) −f(g(z
0
))
g(z) −g(z
0
)
= lim
z→z
0
1
F(g(z))
.
Now apply Lemma 2.3 to evaluate this last limit as
1
F(g(z
0
))
=
1
f
(g(z
0
))
.
2.3 The Cauchy–Riemann Equations
Theorem 2.6. (a) Suppose f is diﬀerentiable at z
0
= x
0
+ iy
0
. Then the partial derivatives of f
satisfy
∂f
∂x
(z
0
) = −i
∂f
∂y
(z
0
) . (2.1)
(b) Suppose f is a complex function such that the partial derivatives f
x
and f
y
exist in an open
disk centered at z
0
and are continuous at z
0
. If these partial derivatives satisfy (2.1) then f is
diﬀerentiable at z
0
.
In both cases (a) and (b), f
is given by
f
(z
0
) =
∂f
∂x
(z
0
) .
Remarks. 1. It is traditional, and often convenient, to write the function f in terms of its real and
imaginary parts. That is, we write f(z) = f(x, y) = u(x, y) +iv(x, y) where u is the real part of f
and v is the imaginary part. Then f
x
= u
x
+ iv
x
and −if
y
= −i(u
y
+ iv
y
) = v
y
− iu
y
. Using this
terminology we can rewrite the equation (2.1) equivalently as the following pair of equations:
u
x
(x
0
, y
0
) = v
y
(x
0
, y
0
)
u
y
(x
0
, y
0
) = −v
x
(x
0
, y
0
) .
(2.2)
2. The partial diﬀerential equations (2.2) are called the Cauchy–Riemann equations, named after
Augustin Louis Cauchy (1789–1857)
1
and Georg Friedrich Bernhard Riemann (1826–1866)
2
.
3. As stated, (a) and (b) are not quite converse statements. However, we will later show that if f is
analytic at z
0
= x
0
+iy
0
then u and v have continuous partials (of any order) at z
0
. That is, later
1
For more information about Cauchy, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Cauchy.html.
2
For more information about Riemann, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Riemann.html.
CHAPTER 2. DIFFERENTIATION 18
we will prove that f = u + iv is analytic in an open set G if and only if u and v have continuous
partials that satisfy (2.2) in G.
4. If u and v satisfy (2.2) and their second partials are also continuous then we obtain
u
xx
(x
0
, y
0
) = v
yx
(x
0
, y
0
) = v
xy
(x
0
, y
0
) = −u
yy
(x
0
, y
0
) ,
that is,
u
xx
(x
0
, y
0
) +u
yy
(x
0
, y
0
) = 0
and an analogous identity for v. Functions with continuous second partials satisfying this partial
diﬀerential equation are called harmonic; we will study such functions in Chapter 6. Again, as we
will see later, if f is analytic in an open set G then the partials of any order of u and v exist; hence
we will show that the real and imaginary part of a function which is analytic on an open set are
harmonic on that set.
Proof of Theorem 2.6. (a) If f is diﬀerentiable at z
0
= (x
0
, y
0
) then
f
(z
0
) = lim
∆z→0
f(z
0
+ ∆z) −f(z
0
)
∆z
.
As we saw in the last section we must get the same result if we restrict ∆z to be on the real axis
and if we restrict it to be on the imaginary axis. In the ﬁrst case we have ∆z = ∆x and
f
(z
0
) = lim
∆x→0
f(z
0
+ ∆x) −f(z
0
)
∆x
= lim
∆x→0
f(x
0
+ ∆x, y
0
) −f(x
0
, y
0
)
∆x
=
∂f
∂x
(x
0
, y
0
).
In the second case we have ∆z = i∆y and
f
(z
0
) = lim
i∆y→0
f(z
0
+i∆y) −f(z
0
)
i∆y
= lim
∆y→0
1
i
f(x
0
, y
0
+ ∆y) −f(x
0
, y
0
)
∆y
= −i
∂f
∂y
(x
0
, y
0
)
(using
1
i
= −i). Thus we have shown that f
(z
0
) = f
x
(z
0
) = −if
y
(z
0
).
(b) To prove the statement in (b), “all we need to do” is prove that f
(z
0
) = f
x
(z
0
), assuming the
Cauchy–Riemann equations and continuity of the partials. We ﬁrst rearrange a diﬀerence quotient
for f
(z
0
), writing ∆z = ∆x +i∆y:
f(z
0
+ ∆z) −f(z
0
)
∆z
=
f(z
0
+ ∆z) −f(z
0
+ ∆x) +f(z
0
+ ∆x) −f(z
0
)
∆z
=
f(z
0
+ ∆x +i∆y) −f(z
0
+ ∆x)
∆z
+
f(z
0
+ ∆x) −f(z
0
)
∆z
=
∆y
∆z
·
f(z
0
+ ∆x +i∆y) −f(z
0
+ ∆x)
∆y
+
∆x
∆z
·
f(z
0
+ ∆x) −f(z
0
)
∆x
.
Now we rearrange f
x
(z
0
):
f
x
(z
0
) =
∆z
∆z
· f
x
(z
0
) =
i∆y + ∆x
∆z
· f
x
(z
0
) =
∆y
∆z
· if
x
(z
0
) +
∆x
∆z
· f
x
(z
0
)
=
∆y
∆z
· f
y
(z
0
) +
∆x
∆z
· f
x
(z
0
) ,
CHAPTER 2. DIFFERENTIATION 19
where we used equation (2.1) in the last step to convert if
x
to i(−if
y
) = f
y
. Now we subtract our
two rearrangements and take a limit:
lim
∆z→0
f(z
0
+ ∆z) −f(z
0
)
∆z
−f
x
(z
0
)
= lim
∆z→0
_
∆y
∆z
_
f(z
0
+ ∆x +i∆y) −f(z
0
+ ∆x)
∆y
−f
y
(z
0
)
__
(2.3)
+ lim
∆z→0
_
∆x
∆z
_
f(z
0
+ ∆x) −f(z
0
)
∆x
−f
x
(z
0
)
__
.
We need to show that these limits are both 0. The fractions ∆x/∆z and ∆y/∆z are bounded by
1 in modulus so we just need to see that the limits of the expressions in parentheses are 0. The
second term in (2.3) has a limit of 0 since, by deﬁnition,
f
x
(z
0
) = lim
∆x→0
f(z
0
+ ∆x) −f(z
0
)
∆x
and taking the limit as ∆z → 0 is the same as taking the limit as ∆x → 0. We can’t do this for
the ﬁrst expression since both ∆x and ∆y are involved, and both change as ∆z →0.
For the ﬁrst term in (2.3) we apply Theorem 1.8, the real meanvalue theorem, to the real and
imaginary parts of f. This gives us real numbers a and b, with 0 < a, b < 1, so that
u(x
0
+ ∆x, y
0
+ ∆y) −u(x
0
+ ∆x, y
0
)
∆y
= u
y
(x
0
+ ∆x, y
0
+a∆y)
v(x
0
+ ∆x, y
0
+ ∆y) −v(x
0
+ ∆x, y
0
)
∆y
= v
y
(x
0
+ ∆x, y
0
+b∆y) .
Using these expressions, we have
f(z
0
+ ∆x +i∆y) −f(z
0
+ ∆x)
∆y
−f
y
(z
0
)
= u
y
(x
0
+ ∆x, y
0
+a∆y) +iv
y
(x
0
+ ∆x, y
0
+b∆y) −(u
y
(x
0
, y
0
) +iv
y
(x
0
, y
0
))
= (u
y
(x
0
+ ∆x, y
0
+a∆y) −u
y
(x
0
, y
0
)) +i (v
y
(x
0
+ ∆x, y
0
+a∆y) −v
y
(x
0
, y
0
)) .
Finally, the two diﬀerences in parentheses have zero limit as ∆z → 0 because u
y
and v
y
are
continuous at (x
0
, y
0
).
2.4 Constants and Connectivity
One of the ﬁrst applications of the meanvalue theorem in real calculus is to show that if a function
has zero derivative everywhere on an interval then it must be constant. The proof is very easy: The
meanvalue theorem for a real function says f(x + ∆x) −f(x) = f
(x +a∆x)∆x where 0 < a < 1.
If we know that f
is always zero then we know that f
(x +a∆x) = 0, so f(x + ∆x) = f(x). This
says that all values of f must be the same, so f is a constant.
However the meanvalue theorem does not have a simple analog for complex valued functions,
so we need another argument to prove that functions with derivative that are always 0 must be
CHAPTER 2. DIFFERENTIATION 20
constant. In fact, this isn’t really true. For example, if the domain of f consists of all complex
numbers with nonzero real part and
f(z) =
_
1 if Re z > 0,
−1 if Re z < 0,
then f
(z) = 0 for all z in the domain of f but f is not constant.
This may seem like a silly example, but it illustrates an important fact about complex functions.
In many cases during the course we will want to conclude that a function is constant, and in each
case we will have to allow for examples like the above. The fundamental problem is that the domain
in this example is not connected, and in fact the correct theorem is:
Theorem 2.7. If the domain of f is a region G ⊆ C and f
(z) = 0 for all z in G then f is a
constant.
Proof. First, suppose that H is a horizontal line segment in G. Consider the real part u(z) for
z ∈ H. Since H is a horizontal segment, y is constant on H, so we can consider u(z) to be just a
function of x. But u
x
(z) = Re(f
(z)) = 0 so, by the real version of the theorem, u(z) is constant
on this horizontal segment. We can argue the same way to see that the imaginary part v(z) of f(z)
is constant on H, since v
x
(z) = Im(f
(z)) = 0. Since both the real and imaginary parts of f are
constant on H, f itself is constant on H.
Next, suppose that V is a vertical segment that is contained in G, and consider the real part
u(z) for z on V . As above, we can consider u(z) to be just a function of y and, using the Cauchy–
Riemann equations, u
y
(z) = −v
x
(z) = −Im(f
(z)) = 0. Thus u is constant on V , and similarly v
is constant on V , so f is constant on V .
Now we can prove the theorem using these two facts: Fix a starting point z
0
in G and let
b = f(z
0
). Connect z
0
to a point z
1
by a horizontal segment H in G; then f is constant on H so
f(z
1
) = f(z
0
) = b. Now connect z
1
to a point z
2
by a vertical segment V in G; then f is constant
on V so f(z
2
) = f(z
1
) = b. Now connect z
2
to a point z
3
by a horizontal segment and conclude
that f(z
3
) = b. Repeating this argument we see that f(z) = b for all points that can be connected
to z
0
in this way by a ﬁnite sequence of horizontal and vertical segments. Theorem 1.6 says that
this is always possible.
There are a number of surprising applications of this theorem; see Exercises 13 and 14 for a
start.
Exercises
1. Use the deﬁnition of limit to show that lim
z→z
0
(az +b) = az
0
+b.
2. Evaluate the following limits or explain why they don’t exist.
(a) lim
z→i
iz
3
−1
z+i
.
(b) lim
z→1−i
x +i(2x +y).
3. Prove Lemma 2.2.
CHAPTER 2. DIFFERENTIATION 21
4. Prove Lemma 2.2 by using the formula for f
given in Theorem 2.6.
5. Apply the deﬁnition of the derivative to give a direct proof that f
(z) = −
1
z
2
when f(z) =
1
z
.
6. Show that if f is diﬀerentiable at z then f is continuous at z.
7. Prove Lemma 2.3.
8. Prove Lemma 2.4.
9. If u(x, y) and v(x, y) are continuous (respectively diﬀerentiable) does it follow that f(z) =
u(x, y) +iv(x, y) is continuous (resp. diﬀerentiable)? If not, provide a counterexample.
10. Where are the following functions diﬀerentiable? Where are they analytic? Determine their
derivatives at points where they are diﬀerentiable.
(a) f(z) = e
−x
e
−iy
.
(b) f(z) = 2x +ixy
2
.
(c) f(z) = x
2
+iy
2
.
(d) f(z) = e
x
e
−iy
.
(e) f(z) = cos xcosh y −i sin xsinh y.
(f) f(z) = Imz.
(g) f(z) = z
2
= x
2
+y
2
.
(h) f(z) = z Imz.
(i) f(z) =
ix+1
y
.
(j) f(z) = 4(Re z)(Imz) −i(z)
2
.
(k) f(z) = 2xy −i(x +y)
2
.
(l) f(z) = z
2
−z
2
.
11. Prove that if f(z) is given by a polynomial in z then f is entire. What can you say if f(z) is
given by a polynomial in x = Re z and y = Imz?
12. Consider the function
f(z) =
_
_
_
xy(x +iy)
x
2
+y
2
if z = 0,
0 if z = 0.
(As always, z = x + iy.) Show that f satisﬁes the Cauchy–Riemann equations at the origin
z = 0, yet f is not diﬀerentiable at the origin. Why doesn’t this contradict Theorem 2.6 (b)?
13. Prove: If f is analytic in the region G ⊆ C and always real valued, then f is constant in G.
(Hint: Use the Cauchy–Riemann equations to show that f
= 0.)
14. Prove: If f(z) and f(z) are both analytic in the region G ⊆ C then f(z) is constant in G.
15. Suppose f(z) is entire, with real and imaginary parts u(z) and v(z) satisfying u(z)v(z) = 3
for all z. Show that f is constant.
CHAPTER 2. DIFFERENTIATION 22
16. Is
x
x
2
+y
2
harmonic? What about
x
2
x
2
+y
2
?
17. The general real homogeneous quadratic function of (x, y) is
u(x, y) = ax
2
+bxy +cy
2
,
where a, b and c are real constants.
(a) Show that u is harmonic if and only if a = −c.
(b) If u is harmonic then show that it is the real part of a function of the form f(z) = Az
2
,
where A is a complex constant. Give a formula for A in terms of the constants a, b
and c.
Chapter 3
Examples of Functions
Obvious is the most dangerous word in mathematics.
E. T. Bell
3.1 M¨obius Transformations
The ﬁrst class of functions that we will discuss in some detail are built from linear polynomials.
Deﬁnition 3.1. A linear fractional transformation is a function of the form
f(z) =
az +b
cz +d
,
where a, b, c, d ∈ C. If ad −bc = 0 then f is called a M¨obius
1
transformation.
Exercise 11 of the previous chapter states that any polynomial (in z) is an entire function.
From this fact we can conclude that a linear fractional transformation f(z) =
az+b
cz+d
is analytic in
C \
_
−
d
c
_
(unless c = 0, in which case f is entire).
One property of M¨obius transformations, which is quite special for complex functions, is the
following.
Lemma 3.1. M¨obius transformations are bijections. In fact, if f(z) =
az+b
cz+d
then the inverse
function of f is given by
f
−1
(z) =
dz −b
−cz +a
.
Remark. Notice that the inverse of a M¨obius transformation is another M¨obius transformation.
Proof. Note that f : C \ {−
d
c
} →C \ {
a
c
}. Suppose f(z
1
) = f(z
2
), that is,
az
1
+b
cz
1
+d
=
az
2
+b
cz
2
+d
.
This is equivalent (unless the denominators are zero) to
(az
1
+b)(cz
2
+d) = (az
2
+b)(cz
1
+d) ,
1
Named after August Ferdinand M¨ obius (1790–1868). For more information about M¨ obius, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Mobius.html.
23
CHAPTER 3. EXAMPLES OF FUNCTIONS 24
which can be rearranged to
(ad −bc)(z
1
−z
2
) = 0 .
Since ad − bc = 0 this implies that z
1
= z
2
, which means that f is onetoone. The formula for
f
−1
: C\ {
a
c
} →C\ {−
d
c
} can be checked easily. Just like f, f
−1
is onetoone, which implies that
f is onto.
Aside from being prime examples of onetoone functions, M¨obius transformations possess fas
cinating geometric properties. En route to an example of such, we introduce some terminology.
Special cases of M¨obius transformations are translations f(z) = z +b, dilations f(z) = az, and in
versions f(z) =
1
z
. The next result says that if we understand those three special transformations,
we understand them all.
Proposition 3.2. Suppose f(z) =
az+b
cz+d
is a linear fractional transformation. If c = 0 then
f(z) =
a
d
z +
b
d
,
if c = 0 then
f(z) =
bc −ad
c
2
1
z +
d
c
+
a
c
.
In particular, every linear fractional transformation is a composition of translations, dilations, and
inversions.
Proof. Simplify.
With the last result at hand, we can tackle the promised theorem about the following geometric
property of M¨obius transformations.
Theorem 3.3. M¨obius transformations map circles and lines into circles and lines.
Proof. Translations and dilations certainly map circles and lines into circles and lines, so by the
last proposition, we only have to prove the theorem for the inversion f(z) =
1
z
.
Before going on we ﬁnd a standard form for the equation of a straight line. Starting with
ax + by = c (where z = x + iy), let α = a + bi. Then αz = ax + by + i(ay − bx) so αz + αz =
αz +αz = 2 Re(αz) = 2ax + 2by. Hence our standard equation for a line becomes
αz +αz = 2c, or Re(αz) = c. (3.1)
First case: Given a circle centered at z
0
with radius r, we can modify its deﬁning equation
z −z
0
 = r as follows:
z −z
0

2
= r
2
(z −z
0
)(z −z
0
) = r
2
zz −z
0
z −zz
0
+z
0
z
0
= r
2
z
2
−z
0
z −zz
0
+z
0

2
−r
2
= 0 .
CHAPTER 3. EXAMPLES OF FUNCTIONS 25
Now we want to transform this into an equation in terms of w, where w =
1
z
. If we solve w =
1
z
for
z we get z =
1
w
, so we make this substitution in our equation:
¸
¸
¸
¸
1
w
¸
¸
¸
¸
2
−z
0
1
w
−z
0
1
w
+z
0

2
−r
2
= 0
1 −z
0
w −z
0
w +w
2
_
z
0

2
−r
2
_
= 0 .
(To get the second line we multiply by w
2
= ww and simplify.) Now if r happens to be equal to
z
0

2
then this equation becomes 1 −z
0
w −z
0
w = 0, which is of the form (3.1) with α = z
0
, so we
have a straight line in terms of w. Otherwise z
0

2
− r
2
is nonzero so we can divide our equation
by it. We obtain
w
2
−
z
0
z
0

2
−r
2
w −
z
0
z
0

2
−r
2
w +
1
z
0

2
−r
2
= 0 .
We deﬁne
w
0
=
z
0
z
0

2
−r
2
, s
2
= w
0

2
−
1
z
0

2
−r
2
=
z
0

2
(z
0

2
−r
2
)
2
−
z
0

2
−r
2
(z
0

2
−r
2
)
2
=
r
2
(z
0

2
−r
2
)
2
.
Then we can rewrite our equation as
w
2
−w
0
w −w
0
w +w
0

2
−s
2
= 0
ww −w
0
w −ww
0
+w
0
w
0
= s
2
(w −w
0
)(w −w
0
) = s
2
w −w
0

2
= s
2
.
This is the equation of a circle in terms of w, with center w
0
and radius s.
Second case: We start with the equation of a line in the form (3.1) and rewrite it in terms of
w, as above, by substituting z =
1
w
and simplifying. We get
z
0
w +z
0
w = 2cww.
If c = 0, this describes a line in the form (3.1) in terms of w. Otherwise we can divide by 2c:
ww −
z
0
2c
w −
z
0
2c
w = 0
_
w −
z
0
2c
_
_
w −
z
0
2c
_
−
z
0

2
4c
2
= 0
¸
¸
¸
¸
w −
z
0
2c
¸
¸
¸
¸
2
=
z
0

2
4c
2
.
This is the equation of a circle with center
z
0
2c
and radius
z
0

2c
.
There is one fact about M¨obius transformations that is very helpful to understanding their
geometry. In fact, it is much more generally useful:
CHAPTER 3. EXAMPLES OF FUNCTIONS 26
Lemma 3.4. Suppose f is analytic at a with f
(a) = 0 and suppose γ
1
and γ
2
are two smooth
curves which pass through a, making an angle of θ with each other. Then f transforms γ
1
and γ
2
into smooth curves which meet at f(a), and the transformed curves make an angle of θ with each
other.
In brief, an analytic function with nonzero derivative preserves angles. Functions which preserve
angles in this way are also called conformal.
Proof. For k = 1, 2 we write γ
k
parametrically, as z
k
(t) = x
k
(t) + iy
k
(t), so that z
k
(0) = a. The
complex number z
k
(0), considered as a vector, is the tangent vector to γ
k
at the point a. Then f
transforms the curve γ
k
to the curve f(γ
k
), parameterized as f(z
k
(t)). If we diﬀerentiate f(z
k
(t))
at t = 0 and use the chain rule we see that the tangent vector to the transformed curve at the
point f(a) is f
(a)z
k
(0). Since f
(a) = 0 the transformation from z
1
(0) and z
2
(0) to f
(a)z
1
(0) and
f
(a)z
2
(0) is a dilation. A dilation is the composition of a scale change and a rotation and both of
these preserve the angles between vectors.
3.2 Inﬁnity and the Cross Ratio
Inﬁnity is not a number—this is true whether we use the complex numbers or stay in the reals.
However, for many purposes we can work with inﬁnity in the complexes much more naturally and
simply than in the reals.
In the complex sense there is only one inﬁnity, written ∞. In the real sense there is also a
“negative inﬁnity”, but −∞= ∞ in the complex sense. In order to deal correctly with inﬁnity we
have to realize that we are always talking about a limit, and complex numbers have inﬁnite limits
if they can become larger in magnitude than any preassigned limit. For completeness we repeat
the usual deﬁnitions:
Deﬁnition 3.2. Suppose G is a set of complex numbers and f is a function from G to C.
(a) lim
z→z
0
f(z) = ∞ means that for every M > 0 we can ﬁnd δ > 0 so that, for all z ∈ G satisfying
0 < z −z
0
 < δ, we have f(z) > M.
(b) lim
z→∞
f(z) = L means that for every > 0 we can ﬁnd N > 0 so that, for all z ∈ G satisfying
z > N, we have f(z) −L < .
(c) lim
z→∞
f(z) = ∞ means that for every M > 0 we can ﬁnd N > 0 so that, for all z ∈ G satisfying
z > N we have f(z) > M.
In the ﬁrst deﬁnition we require that z
0
is an accumulation point of G while in the second and third
we require that ∞ is an “extended accumulation point” of G, in the sense that for every B > 0
there is some z ∈ G with z > B.
The usual rules for working with inﬁnite limits are still valid in the complex numbers. In
fact, it is a good idea to make inﬁnity an honorary complex number so that we can more easily
manipulate inﬁnite limits. We do this by deﬁning a new set,
ˆ
C = C ∪ {∞}. In this new set we
deﬁne algebraic rules for dealing with inﬁnity based on the usual laws of limits. For example, if
lim
z→z
0
f(z) = ∞ and lim
z→z
0
g(z) = a is ﬁnite then the usual “limit of sum = sum of limits” rule gives
lim
z→z
0
(f(z) +g(z)) = ∞. This leads to the addition rule ∞+a = ∞. We summarize these rules:
CHAPTER 3. EXAMPLES OF FUNCTIONS 27
Deﬁnition 3.3. Suppose a ∈ C.
(a) ∞+a = a +∞= ∞
(b) ∞· a = a · ∞= ∞· ∞= ∞ if a = 0.
(c)
a
∞
= 0 and
a
0
= ∞ if a = 0.
If a calculation involving inﬁnity is not covered by the rules above then we must investigate the
limit more carefully. For example, it may seem strange that ∞+∞ is not deﬁned, but if we take
the limit of z +(−z) = 0 as z →∞we will get 0, but the individual limits of z and −z are both ∞.
Now we reconsider M¨obius transformations with inﬁnity in mind. For example, f(z) =
1
z
is
now deﬁned for z = 0 and z = ∞, with f(0) = ∞ and f(∞) = 0, so the proper domain for
f(z) is actually
ˆ
C. Let’s consider the other basic types of M¨obius transformations. A translation
f(z) = z + b is now deﬁned for z = ∞, with f(∞) = ∞+ b = ∞, and a dilation f(z) = az (with
a = 0) is also deﬁned for z = ∞, with f(∞) = a · ∞ = ∞. Since every M¨obius transformation
can be expressed as a composition of translations, dilations and the inversion f(z) =
1
z
we see that
every M¨obius transformation may be interpreted as a transformation of
ˆ
C onto
ˆ
C. The general
case is summarized below:
Lemma 3.5. Let f be the M¨obius transformation
f(z) =
az +b
cz +d
.
Then f is deﬁned for all z ∈
ˆ
C. If c = 0 then f(∞) = ∞, and, otherwise,
f(∞) =
a
c
and f
_
−
d
c
_
= ∞.
With this interpretation in mind we can add some insight to Theorem 3.3. Recall that f(z) =
1
z
transforms circles that pass through the origin to straight lines, but the point z = 0 must be excluded
from the circle. However, now we can put it back, so f transforms circles that pass through the
origin to straight lines plus ∞. If we remember that ∞ corresponds to being arbitrarily far away
from the origin we can visualize a line plus inﬁnity as a circle passing through ∞. If we make
this a deﬁnition then Theorem 3.3 can be expressed very simply: any M¨obius transformation of
ˆ
C
transforms circles to circles. For example, the transformation
f(z) =
z +i
z −i
transforms −i to 0, i to ∞, and 1 to i. The three points −i, i and 1 determine a circle—the unit
circle z = 1—and the three image points 0, ∞ and i also determine a circle—the imaginary axis
plus the point at inﬁnity. Hence f transforms the unit circle onto the imaginary axis plus the point
at inﬁnity.
This example relied on the idea that three distinct points in
ˆ
C determine uniquely a circle
passing through them. If the three points are on a straight line or if one of the points is ∞ then
the circle is a straight line plus ∞. Conversely, if we know where three distinct points in
ˆ
C are
transformed by a M¨obius transformation then we should be able to ﬁgure out everything about the
transformation. There is a computational device that makes this easier to see.
CHAPTER 3. EXAMPLES OF FUNCTIONS 28
Deﬁnition 3.4. If z, z
1
, z
2
, and z
3
are any four points in
ˆ
C with z
1
, z
2
, and z
3
distinct, then their
crossratio is deﬁned by
[z, z
1
, z
2
, z
3
] =
(z −z
1
)(z
2
−z
3
)
(z −z
3
)(z
2
−z
1
)
.
Here if z = z
3
, the result is inﬁnity, and if one of z, z
1
, z
2
, or z
3
is inﬁnity, then the two terms on
the right containing it are canceled.
Lemma 3.6. If f is deﬁned by f(z) = [z, z
1
, z
2
, z
3
] then f is a M¨obius transformation which
satisﬁes
f(z
1
) = 0, f(z
2
) = 1, f(z
3
) = ∞.
Moreover, if g is any M¨obius transformation which transforms z
1
, z
2
and z
3
as above then g(z) =
f(z) for all z.
Proof. Everything should be clear except the ﬁnal uniqueness statement. By Lemma 3.1 the inverse
f
−1
is a M¨obius transformation and, by Exercise 7 in this chapter, the composition h = g ◦ f
−1
is a M¨obius transformation. Notice that h(0) = g(f
−1
(0)) = g(z
1
) = 0. Similarly, h(1) = 1 and
h(∞) = ∞. If we write h(z) =
az+b
cz+d
then
0 = h(0) =
b
d
=⇒ b = 0
∞= h(∞) =
a
c
=⇒ c = 0
1 = h(1) =
a +b
c +d
=
a + 0
0 +d
=
a
d
=⇒ a = d ,
so h(z) =
az+b
cz+d
=
az+0
0+d
=
a
d
z = z. But since h(z) = z for all z we have h(f(z)) = f(z) and so
g(z) = g ◦ (f
−1
◦ f)(z) = (g ◦ f
−1
) ◦ f(z) = h(f(z)) = f(z).
So if we want to map three given points of
ˆ
C to 0, 1 and ∞ by a M¨obius transformation then
the crossratio gives us the only way to do it. What if we have three points z
1
, z
2
and z
3
and we
want to map them to three other points, w
1
, w
2
and w
3
?
Theorem 3.7. Suppose z
1
, z
2
and z
3
are distinct points in
ˆ
C and w
1
, w
2
and w
3
are distinct
points in
ˆ
C. Then there is a unique M¨obius transformation h satisfying h(z
1
) = w
1
, h(z
2
) = w
2
and h(z
3
) = w
3
.
Proof. Let h = g
−1
◦ f where f(z) = [z, z
1
, z
2
, z
3
] and g(w) = [w, w
1
, w
2
, w
3
]. Uniqueness follows
as in the proof of Lemma 3.6.
This theorem gives an explicit way to determine h from the points z
j
and w
j
but, in practice, it
is often easier to determine h directly from the conditions f(z
k
) = w
k
(by solving for a, b, c and d).
3.3 Exponential and Trigonometric Functions
To deﬁne the complex exponential function, we once more borrow concepts from calculus, namely
the real exponential function
2
and the real sine and cosine, and—in addition—ﬁnally make sense
of the notation e
it
= cos t +i sin t.
2
It is a nontrivial question how to deﬁne the real exponential function. Our preferred way to do this is through a
power series: e
x
=
P
k≥0
x
k
/k!. In light of this deﬁnition, the reader might think we should have simply deﬁned the
CHAPTER 3. EXAMPLES OF FUNCTIONS 29
Deﬁnition 3.5. The (complex) exponential function is deﬁned for z = x +iy as
exp(z) = e
x
(cos y +i sin y) = e
x
e
iy
.
This deﬁnition seems a bit arbitrary, to say the least. Its ﬁrst justiﬁcation is that all exponential
rules which we are used to from real numbers carry over to the complex case. They mainly follow
from Lemma 1.2 and are collected in the following.
Lemma 3.8. For all z, z
1
, z
2
∈ C,
(a) exp (z
1
) exp (z
2
) = exp (z
1
+z
2
)
(b)
1
exp(z)
= exp (−z)
(c) exp (z + 2πi) = exp (z)
(d) exp (z) = exp (Re z)
(e) exp(z) = 0
(f)
d
dz
exp (z) = exp (z) .
Remarks. 1. The third identity is a very special one and has no counterpart for the real exponential
function. It says that the complex exponential function is periodic with period 2πi. This has many
interesting consequences; one that may not seem too pleasant at ﬁrst sight is the fact that the
complex exponential function is not onetoone.
2. The last identity is not only remarkable, but we invite the reader to meditate on its proof. When
proving this identity through the Cauchy–Riemann equations for the exponential function, one can
get another strong reason why Deﬁnition 3.5 is reasonable. Finally, note that the last identity also
says that exp is entire.
We should make sure that the complex exponential function specializes to the real exponential
function for real arguments: if z = x ∈ R then
exp(x) = e
x
(cos 0 +i sin 0) = e
x
.
The trigonometric functions—sine, cosine, tangent, cotangent, etc.—have their complex ana
logues, however, they don’t play the same prominent role as in the real case. In fact, we can deﬁne
them as merely being special combinations of the exponential function.
Deﬁnition 3.6. The (complex) sine and cosine are deﬁned as
sin z =
1
2i
(exp(iz) −exp(−iz)) and cos z =
1
2
(exp(iz) + exp(−iz)) ,
respectively. The tangent and cotangent are deﬁned as
tan z =
sin z
cos z
= −i
exp(2iz) −1
exp(2iz) + 1
and cot z =
cos z
sin z
= i
exp(2iz) + 1
exp(2iz) −1
,
respectively.
complex exponential function through a complex power series. In fact, this is possible (and an elegant deﬁnition);
however, one of the promises of these lecture notes is to introduce complex power series as late as possible. We agree
with those readers who think that we are “cheating” at this point, as we borrow the concept of a (real) power series
to deﬁne the real exponential function.
CHAPTER 3. EXAMPLES OF FUNCTIONS 30
//
exp
−
5π
6
−
π
3
0
π
3
5π
6
−1 0 1 2
q
q
q
q
q
q
q
q
q
q
q
q
q
q
q
q
q
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
M
Figure 3.1: Image properties of the exponential function.
Note that to write tangent and cotangent in terms of the exponential function, we used the fact
that exp(z) exp(−z) = exp(0) = 1. Because exp is entire, so are sin and cos.
As with the exponential function, we should ﬁrst make sure that we’re not redeﬁning the real
sine and cosine: if z = x ∈ R then
sin x =
1
2i
(exp(ix) −exp(−ix)) =
1
2i
(cos x +i sin x −(cos(−x) +i sin(−x))) = sin x.
(The ‘sin’ on the left denotes the complex sine, the one on the right the real sine.) A similar
calculation holds for the cosine. Not too surprising, the following properties follow mostly from
Lemma 3.8.
Lemma 3.9. For all z, z
1
, z
2
∈ C,
sin(−z) = −sin z cos(−z) = cos z
sin(z + 2π) = sin z cos(z + 2π) = cos z
tan(z +π) = tan z cot(z +π) = cot z
sin(z +π/2) = cos z cos(z +π/2) = −sin z
sin (z
1
+z
2
) = sin z
1
cos z
2
+ cos z
1
sin z
2
cos (z
1
+z
2
) = cos z
1
cos z
2
−sin z
1
sin z
2
cos
2
z + sin
2
z = 1 cos
2
z −sin
2
z = cos(2z)
sin
z = cos z cos
z = −sin z .
Finally, one word of caution: unlike in the real case, the complex sine and cosine are not
bounded—consider, for example, sin(iy) as y →±∞.
We end this section with a remark on hyperbolic trig functions. The hyperbolic sine, cosine,
tangent, and cotangent are deﬁned as in the real case:
sinh z =
1
2
(exp(z) −exp(−z)) cosh z =
1
2
(exp(z) + exp(−z))
tanh z =
sinh z
cosh z
=
exp(2z) −1
exp(2z) + 1
coth z =
cosh z
sinh z
=
exp(2z) + 1
exp(2z) −1
.
CHAPTER 3. EXAMPLES OF FUNCTIONS 31
As such, they are not only yet more special combinations of the exponential function, but they
are also related with the trigonometric functions via
sinh(iz) = i sin z and cosh(iz) = cos z .
3.4 The Logarithm and Complex Exponentials
The complex logarithm is the ﬁrst function we’ll encounter that is of a somewhat tricky nature. It
is motivated as being the inverse function to the exponential function, that is, we’re looking for a
function Log such that
exp(Log z) = z = Log(exp z) .
As we will see shortly, this is too much to hope for. Let’s write, as usual, z = r e
iφ
, and suppose
that Log z = u(z) +iv(z). Then for the ﬁrst equation to hold, we need
exp(Log z) = e
u
e
iv
= r e
iφ
= z ,
that is, e
u
= r = z ⇐⇒ u = ln z (where ln denotes the real natural logarithm; in particular we
need to demand that z = 0), and e
iv
= e
iφ
⇐⇒v = φ+2πk for some k ∈ Z. A reasonable deﬁnition
of a logarithm function Log would hence be to set Log z = ln z + i Arg z where Arg z gives the
argument for the complex number z according to some convention—for example, we could agree
that the argument is always in (−π, π], or in [0, 2π), etc. The problem is that we need to stick to
this convention. On the other hand, as we saw, we could just use a diﬀerent argument convention
and get another reasonable ‘logarithm.’ Even worse, by deﬁning the multivalued map
arg z = {φ : φ is a possible argument of z}
and deﬁning the multivalued logarithm as
log z = ln z +i arg z ,
we get something that’s not a function, yet it satisﬁes
exp(log z) = z .
We invite the reader to check this thoroughly; in particular, one should note how the periodicity
of the exponential function takes care of the multivaluedness of our ‘logarithm’ log.
log is, of course, not a function, and hence we can’t even consider it to be our soughtafter
inverse of the exponential function. Let’s try to make things well deﬁned.
Deﬁnition 3.7. Any function Log : C\ {0} →C which satisﬁes exp(Log z) = z is a branch of the
logarithm. Let Arg z denote that argument of z which is in (−π, π] (the principal argument of z).
Then the principal logarithm is deﬁned as
Log z = ln z +i Arg z .
CHAPTER 3. EXAMPLES OF FUNCTIONS 32
The paragraph preceding this deﬁnition ensures that the principal logarithm is indeed a branch
of the logarithm. Even better, the evaluation of any branch of the logarithm at z can only diﬀer
from Log z by a multiple of 2πi; the reason for this is once more the periodicity of the exponential
function.
So what about the other equation Log(exp z) = z? Let’s try the principal logarithm: Suppose
z = x +iy, then
Log(exp z) = Log
_
e
x
e
iy
_
= ln
¸
¸
e
x
e
iy
¸
¸
+i Arg
_
e
x
e
iy
_
= ln e
x
+i Arg
_
e
iy
_
= x +i Arg
_
e
iy
_
.
The righthand side is equal to z = x + iy only if y ∈ (−π, π]. The same happens with any
other branch of the logarithm Log—there will always be some (in fact, many) yvalues for which
Log(exp z) = z.
To end our discussion of the logarithm on a happy note, we prove that any branch of the
logarithm has the same derivative; one just has to be cautious about where each logarithm is
analytic.
Theorem 3.10. Suppose Log is a branch of the logarithm. Then Log is diﬀerentiable wherever it
is continuous and
Log
z =
1
z
.
Proof. The idea is to apply Lemma 2.5 to exp and Log, but we need to be careful about the domains
of these functions, so that we get actual inverse functions. Suppose Log maps C \ {0} to G (this
is typically a halfopen strip; you might want to think about what it looks like if Log = Log). We
apply Lemma 2.5 with f : G →C\ {0} , f(z) = exp(z) and g : C\ {0} →G, g(z) = Log: if Log is
continuous at z then
Log
z =
1
exp
(Log z)
=
1
exp(Log z)
=
1
z
.
We ﬁnish this section by deﬁning complex exponentials. For two complex numbers a and b, the
natural deﬁnition a
b
= exp(b log a) (which is a concept borrowed from calculus) would in general
yield more than one value (Exercise 31), so it is not always useful. We turn instead to the principal
logarithm and deﬁne the principal value of a
b
as
a
b
= exp(b Log a) .
A note about e. In calculus one proves the equivalence of the real exponential function (as given,
for example, through a power series) and the function f(x) = e
x
where e is Euler’s
3
number and
can be deﬁned, for example, as e = lim
n→∞
_
1 +
1
n
_
n
. With our deﬁnition of a
b
, we can now make
a similar remark about the complex exponential function. Because e is a positive real number and
hence Arg e = 0, we obtain
e
z
= exp(z Log e) = exp (z (ln e +i Arg e)) = exp (z ln e) = exp (z) .
A word of caution: this only works out this nicely because we carefully deﬁned a
b
for complex
numbers. Diﬀerent deﬁnitions might lead to diﬀerent outcomes of e
z
versus exp z!
3
Named after Leonard Euler (1707–1783). For more information about Euler, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Euler.html.
CHAPTER 3. EXAMPLES OF FUNCTIONS 33
Exercises
1. Show that if f(z) =
az+b
cz+d
is a M¨obius transformation then f
−1
(z) =
dz−b
−cz+a
.
2. Show that the derivative of a M¨obius transformation is never zero.
3. Prove that any M¨obius transformation diﬀerent from the identity map can have at most two
ﬁxed points. (A ﬁxed point of a function f is a number z such that f(z) = z.)
4. Prove Proposition 3.2.
5. Show that the M¨obius transformation f(z) =
1+z
1−z
maps the unit circle (minus the point z = 1)
onto the imaginary axis.
6. Suppose that f is analytic on the region G and f(G) is a subset of the unit circle. Show
that f is constant. (Hint: Consider the function
1+f(z)
1−f(z)
and use Exercise 5 and a variation of
Exercise 13 in Chapter 2.)
7. Suppose A =
_
a b
c d
_
is a 2 × 2 matrix of complex numbers whose determinant ad − bc is
nonzero. Then we can deﬁne a corresponding M¨obius transformation T
A
by T
A
(z) =
az+b
cz+d
.
Show that T
A
◦T
B
= T
A·B
. (Here ◦ denotes composition and · denotes matrix multiplication.)
8. Let f(z) =
2z
z+2
. Draw two graphs, one showing the following six sets in the z plane and the
other showing their images in the w plane. Label the sets. (You should only need to calculate
the images of 0, ±2, ∞ and −1 −i; remember that M¨obius transformations preserve angles.)
(a) The xaxis, plus ∞.
(b) The yaxis, plus ∞.
(c) The line x = y, plus ∞.
(d) The circle with radius 2 centered at 0.
(e) The circle with radius 1 centered at 1.
(f) The circle with radius 1 centered at −1.
9. Find M¨obius transformations satisfying each of the following. Write your answers in standard
form, as
az+b
cz+d
.
(a) 1 →0, 2 →1, 3 →∞. (Use the crossratio.)
(b) 1 →0, 1 +i →1, 2 →∞. (Use the crossratio.)
(c) 0 →i, 1 →1, ∞→−i.
10. Let C be the circle with center 1+i and radius 1. Using the crossratio, with diﬀerent choices
of z
k
, ﬁnd two diﬀerent M¨obius transformations that transform C onto the real axis plus
inﬁnity. In each case, ﬁnd the image of the center of the circle.
11. Let C be the circle with center 0 and radius 1. Find a M¨obius transformation which transforms
C onto C and transforms 0 to
1
2
.
CHAPTER 3. EXAMPLES OF FUNCTIONS 34
12. Describe the image of the region under the transformation:
(a) The disk z < 1 under w =
iz−i
z+1
.
(b) The quadrant x > 0, y > 0 under w =
z−i
z+i
.
(c) The strip 0 < x < 1 under w =
z
z−1
.
13. The Jacobian of a transformation u = u(x, y), v = v(x, y) is the determinant of the matrix
_
∂u
∂x
∂u
∂y
∂v
∂x
∂v
∂y
_
. Show that if f = u +iv is analytic then the Jacobian equals f
(z)
2
.
14. Find the ﬁxed points in
ˆ
C of f(z) =
z
2
−1
2z+1
.
15. Find the M¨obius transformation f:
(a) f maps 0 →1, 1 →∞, ∞→0.
(b) f maps 1 →1, −1 →i, −i →−1.
(c) f maps xaxis to y = x, yaxis to y = −x, and the unit circle to itself.
16. Suppose z
1
, z
2
and z
3
are distinct points in
ˆ
C. Show that z is on the circle passing through
by z
1
, z
2
and z
3
if and only if [z, z
1
, z
2
, z
3
] is real or inﬁnite.
17. Describe the images of the following sets under the exponential function exp(z):
(a) the line segment deﬁned by z = iy, 0 ≤ y ≤ 2π.
(b) the line segment deﬁned by z = 1 +iy, 0 ≤ y ≤ 2π.
(c) the rectangle {z = x +iy ∈ C : 0 ≤ x ≤ 1, 0 ≤ y ≤ 2π}.
18. Prove Lemma 3.8.
19. Prove Lemma 3.9.
20. Let z = x +iy and show that
(a) sin z = sin xcosh y +i cos xsinh y.
(b) cos z = cos xcosh y −i sin xsinh y.
21. Let z = x +iy and show that
(a) sin z
2
= sin
2
x + sinh
2
y = cosh
2
y −cos
2
x.
(b) cos z
2
= cos
2
x + sinh
2
y = cosh
2
y −sin
2
x.
(c) If cos x = 0 then cot z
2
=
cosh
2
y−1
cosh
2
y
≤ 1.
(d) If y ≥ 1 then cot z
2
≤
sinh
2
y+1
sinh
2
y
= 1 +
1
sinh
2
y
≤ 1 +
1
sinh
2
1
≤ 2.
22. Show that tan(iz) = i tanh z.
23. Find the principal values of
CHAPTER 3. EXAMPLES OF FUNCTIONS 35
(a) log i.
(b) (−1)
i
.
(c) log(1 +i).
24. Is arg(z) = −arg(z) true for the multiplevalued argument? What about Arg(z) = −Arg(z)
for the principal branch?
25. Is there a diﬀerence between the set of all values of log
_
z
2
_
and the set of all values of 2 log z?
(Try some ﬁxed numbers for z.)
26. For each of the following functions, determine all complex numbers for which the function is
analytic. If you run into a logarithm, use the principal value (unless stated otherwise).
(a) z
2
.
(b)
sin z
z
3
+1
.
(c) Log(z −2i + 1) where Log(z) = ln z +i Arg(z) with 0 ≤ Arg(z) < 2π.
(d) exp(z).
(e) (z −3)
i
.
(f) i
z−3
.
27. Find all solutions to the following equations:
(a) Log(z) =
π
2
i.
(b) Log(z) =
3π
2
i.
(c) exp(z) = πi.
(d) sin z = cosh 4.
(e) cos z = 0.
(f) sinh z = 0.
(g) exp(iz) = exp(iz).
(h) z
1/2
= 1 +i.
28. Find the image of the annulus 1 < z < e under the principal value of the logarithm.
29. Show that a
z
 = a
Re z
if a is a positive real constant.
30. Fix c ∈ C \ {0}. Find the derivative of f(z) = z
c
.
31. Prove that exp(b log a) is singlevalued if and only if b is an integer. (Note that this means
that complex exponentials don’t clash with monomials z
n
.) What can you say if b is rational?
32. Describe the image under exp of the line with equation y = x. To do this you should ﬁnd
an equation (at least parametrically) for the image (you can start with the parametric form
x = t, y = t), plot it reasonably carefully, and explain what happens in the limits as t → ∞
and t →−∞.
33. For this problem, f(z) = z
2
.
CHAPTER 3. EXAMPLES OF FUNCTIONS 36
(a) Show that the image of a circle centered at the origin is a circle centered at the origin.
(b) Show that the image of a ray starting at the origin is a ray starting at the origin.
(c) Let T be the ﬁgure formed by the horizontal segment from 0 to 2, the circular arc from
2 to 2i, and then the vertical segment from 2i to 0. Draw T and f(T).
(d) Is the right angle at the origin in part (c) preserved? Is something wrong here?
(Hint: Use polar coordinates.)
34. As in the previous problem, let f(z) = z
2
. Let Q be the square with vertices at 0, 2, 2 + 2i
and 2i. Draw f(Q) and identify the types of image curves corresponding to the segments
from 2 to 2 + 2i and from 2 + 2i to 2i. They are not parts of either straight lines or circles.
(Hint: You can write the vertical segment parametrically as z(t) = 2 + it. Eliminate the
parameter in u +iv = f(z(t)) to get a (u, v) equation for the image curve.)
Chapter 4
Integration
Everybody knows that mathematics is about miracles, only mathematicians have a name for
them: theorems.
Roger Howe
4.1 Deﬁnition and Basic Properties
At ﬁrst sight, complex integration is not really anything diﬀerent from real integration. For a
continuous complexvalued function φ : [a, b] ⊂ R →C, we deﬁne
_
b
a
φ(t) dt =
_
b
a
Re φ(t) dt +i
_
b
a
Imφ(t) dt . (4.1)
For a function which takes complex numbers as arguments, we integrate over a curve γ (instead
of a real interval). Suppose this curve is parametrized by γ(t), a ≤ t ≤ b. If one meditates about
the substitution rule for real integrals, the following deﬁnition, which is based on (4.1) should come
as no surprise.
Deﬁnition 4.1. Suppose γ is a smooth curve parametrized by γ(t), a ≤ t ≤ b, and f is a complex
function which is continuous on γ. Then we deﬁne the integral of f on γ as
_
γ
f =
_
γ
f(z) dz =
_
b
a
f(γ(t))γ
(t) dt .
This deﬁnition can be naturally extended to piecewise smooth curves, that is, those curves γ
whose parametrization γ(t), a ≤ t ≤ b, is only piecewise diﬀerentiable, say γ(t) is diﬀerentiable on
the intervals [a, c
1
], [c
1
, c
2
], . . . , [c
n−1
, c
n
], [c
n
, b]. In this case we simply deﬁne
_
γ
f =
_
c
1
a
f(γ(t))γ
(t) dt +
_
c
2
c
1
f(γ(t))γ
(t) dt +· · · +
_
b
cn
f(γ(t))γ
(t) dt .
In what follows, we’ll usually state our results for smooth curves, bearing in mind that practically
all can be extended to piecewise smooth curves.
Example 4.1. As our ﬁrst example of the application of this deﬁnition we will compute the integral
of the function f(z) = z
2
=
_
x
2
−y
2
_
−i(2xy) over several curves from the point z = 0 to the point
z = 1 +i.
37
CHAPTER 4. INTEGRATION 38
(a) Let γ be the line segment from z = 0 to z = 1 + i. A parametrization of this curve is
γ(t) = t +it, 0 ≤ t ≤ 1. We have γ
(t) = 1 +i and f(γ(t)) = (t −it)
2
, and hence
_
γ
f =
_
1
0
(t −it)
2
(1 +i) dt = (1 +i)
_
1
0
t
2
−2it
2
−t
2
dt = −2i(1 +i)/3 =
2
3
(1 −i) .
(b) Let γ be the arc of the parabola y = x
2
from z = 0 to z = 1 + i. A parametrization of this
curve is γ(t) = t +it
2
, 0 ≤ t ≤ 1. Now we have γ
(t) = 1 + 2it and
f(γ(t)) =
_
t
2
−
_
t
2
_
2
_
−i 2t · t
2
= t
2
−t
4
−2it
3
,
whence
_
γ
f =
_
1
0
_
t
2
−t
4
−2it
3
_
(1 + 2it) dt =
_
1
0
t
2
+ 3t
4
−2it
5
dt =
1
3
+ 3
1
5
−2i
1
6
=
14
15
−
i
3
.
(c) Let γ be the union of the two line segments γ
1
from z = 0 to z = 1 and γ
2
from z = 1 to
z = 1 +i. Parameterizations are γ
1
(t) = t, 0 ≤ t ≤ 1 and γ
2
(t) = 1 +it, 0 ≤ t ≤ 1. Hence
_
γ
f =
_
γ
1
f +
_
γ
2
f =
_
1
0
t
2
· 1 dt +
_
1
0
(1 −it)
2
i dt =
1
3
+i
_
1
0
1 −2it −t
2
dt
=
1
3
+i
_
1 −2i
1
2
−
1
3
_
=
4
3
+
2
3
i .
The complex integral has some standard properties, most of which follow from their real siblings
in a straightforward way.
Proposition 4.1. Suppose γ is a smooth curve, f and g are complex functions which are continuous
on γ, and c ∈ C.
(a)
_
γ
(f +cg) =
_
γ
f + c
_
γ
g .
(b) If γ is parametrized by γ(t), a ≤ t ≤ b, deﬁne the curve −γ through −γ(t) = γ(a +b −t), a ≤
t ≤ b. Then
_
−γ
f = −
_
γ
f .
(c) If γ
1
and γ
2
are curves so that γ
2
starts where γ
1
ends then deﬁne the curve γ
1
γ
2
by following γ
1
to its end, and then continuing on γ
2
to its end. Then
_
γ
1
γ
2
f(z) dz =
_
γ
1
f(z) dz +
_
γ
2
f(z) dz .
(d)
¸
¸
¸
_
γ
f
¸
¸
¸ ≤ max
z∈γ
f(z) · length(γ) .
The curve −γ deﬁned in (b) is the curve that we obtain by traveling through γ in the opposite
direction.
In (d) the length of a smooth curve γ with parametrization γ(t), a ≤ t ≤ b, is deﬁned as
length(γ) =
_
b
a
¸
¸
γ
(t)
¸
¸
dt .
We invite the reader to use some familiar curves to see that this deﬁnition gives what one would
expect to be the length of a curve.
CHAPTER 4. INTEGRATION 39
Proof. (a) follows directly from the deﬁnition of the integral and the properties of real integrals.
(b) follows with an easy real change of variables s = a +b −t:
_
−γ
f =
_
b
a
f (γ(a +b −t)) (γ(a +b −t))
dt = −
_
b
a
f (γ(a +b −t)) γ
(a +b −t) dt
=
_
a
b
f (γ(s)) γ
(s) ds = −
_
b
a
f (γ(s)) γ
(s) ds = −
_
γ
f .
For (c) we need a suitable parameterization γ(t) for γ
1
γ
2
. If γ
1
has domain [a
1
, b
1
] and γ
2
has
domain [a
2
, b
2
] then we can use
γ(t) =
_
γ
1
(t) for a
1
≤ t ≤ b
1
,
γ
2
(t −b
1
+a
2
) for b
1
≤ t ≤ b
1
+b
2
−a
2
.
The fact that γ
1
(b
1
) = γ
2
(a
2
) is necessary to make sure that this parameterization is piecewise
smooth. Now we break the integral over γ
1
γ
2
into two pieces and apply the simple change of
variables s = t −b
1
+a
2
:
_
γ
1
γ
2
f(z) dz =
_
b
1
+b
2
−a
2
a
1
f(γ(t))γ
(t) dt =
_
b
1
a
1
f(γ(t))γ
(t) dt +
_
b
1
+b
2
−a
2
b
1
f(γ(t))γ
(t) dt
=
_
b
1
a
1
f(γ
1
(t))γ
(t) dt +
_
b
1
+b
2
−a
2
b
1
f(γ
2
(t −b
1
+a
2
))γ
2
(t −b
1
+a
2
) dt
=
_
b
1
a
1
f(γ
1
(t))γ
(t) dt +
_
b
2
a
2
f(γ
2
(s))γ
2
(s) ds
=
_
γ
1
f(z) dz +
_
γ
2
f(z) dz.
Finally, to prove (d), let φ = Arg
_
γ
f. Then
¸
¸
¸
¸
_
γ
f(z) dz
¸
¸
¸
¸
=
_
γ
f(z) dz e
−iφ
= Re
__
γ
f(z) dz e
−iφ
_
= Re
__
b
a
f(γ(t))γ
(t)e
−iφ
dt
_
=
_
b
a
Re
_
f(γ(t))e
−iφ
γ
(t)
_
dt ≤
_
b
a
¸
¸
¸f(γ(t))e
−iφ
γ
(t)
¸
¸
¸ dt =
_
b
a
f(γ(t))
¸
¸
γ
(t)
¸
¸
dt
≤ max
a≤t≤b
f(γ(t))
_
b
a
¸
¸
γ
(t)
¸
¸
dt = max
z∈γ
f(z) · length(γ) .
4.2 Antiderivatives
Just like in the real case, one easy way to compute integrals is through knowing the antiderivative
(or primitive) of the integrand f, that is, a function F such that F
= f. To be more precise, we
say that f has an antiderivative on G if there exists a function F that is analytic on G, such that
F
(z) = f(z) for all z ∈ G.
CHAPTER 4. INTEGRATION 40
Theorem 4.2. Suppose G ⊆ C is open, γ is a smooth curve in G parametrized by γ(t), a ≤ t ≤ b,
f is continuous on G, and F is a primitive of f on G. Then
_
γ
f = F (γ(b)) −F (γ(a)) .
In particular,
_
γ
f is independent of the path γ ⊂ G between γ(a) and γ(b).
Example 4.1 shows that a pathindependent integral is quite special; it also says that the
function z
2
does not have an antiderivative in, for example, the region {z ∈ C : z < 2}. (Actually,
the function z
2
does not have an antiderivative in any nonempty region—prove it!)
In the special case that γ is closed (that is, γ(a) = γ(b)), we immediately get the following nice
consequence.
Corollary 4.3. Suppose G ⊆ C is open, γ is a smooth closed curve in G, and f is continuous on
G and has an antiderivative on G. Then
_
γ
f = 0 .
Proof of Theorem 4.2. An application of the chain rule shows
d
dt
F(γ(t)) = F
(γ(t))γ
(t) ,
and then we calculate
_
γ
f =
_
b
a
f(γ(t))γ
(t) dt =
_
b
a
F
(γ(t))γ
(t) dt =
_
b
a
d
dt
F(γ(t)) dt = F (γ(b)) −F (γ(a)) ,
by Theorem 1.9 (the Fundamental Theorem of Calculus).
4.3 Cauchy’s Theorem
We now turn to the central theorem of complex analysis. It is based on the following concept.
Deﬁnition 4.2. Suppose γ
1
and γ
2
are closed curves in the open set G ⊆ C, parametrized by
γ
1
(t), 0 ≤ t ≤ 1 and γ
2
(t), 0 ≤ t ≤ 1, respectively. Then γ
1
is Ghomotopic to γ
2
, in symbols
γ
1
∼
G
γ
2
, if there is a continuous function h : [0, 1]
2
→G such that
h(t, 0) = γ
1
(t) ,
h(t, 1) = γ
2
(t) ,
h(0, s) = h(1, s) .
The function h(t, s) is called a homotopy and represents a curve for each ﬁxed s, which is
continuously transformed from γ
1
to γ
2
. The last condition simply says that each of the curves
h(t, s), 0 ≤ t ≤ 1 is closed. An example is depicted in Figure 4.1.
Here is the theorem on which most of what will follow is based.
CHAPTER 4. INTEGRATION 41
Figure 4.1: This square and the circle are (C \ {0})homotopic.
Theorem 4.4 (Cauchy’s Theorem). Suppose G ⊆ C is open, f is analytic in G, and γ
1
∼
G
γ
2
via
a homotopy with continuous second partials. Then
_
γ
1
f =
_
γ
2
f .
Remarks. 1. The condition on the smoothness of the homotopy can be omitted, however, then the
proof becomes too advanced for the scope of these notes. In all the examples and exercises that
we’ll have to deal with here, the homotopies will be ‘nice enough’ to satisfy the condition of this
theorem.
2. It is assumed that Johann Carl Friedrich Gauß (1777–1855)
1
knew a version of this theorem in
1811 but only published it in 1831. Cauchy published his version in 1825, Weierstraß
2
his in 1842.
Cauchy’s theorem is often called the Cauchy–Goursat Theorem, since Cauchy assumed that the
derivative of f was continuous, a condition which was ﬁrst removed by Goursat
3
.
An important special case is the one where a curve γ is Ghomotopic to a point, that is, a
constant curve (see Figure 4.2 for an example). In this case we simply say γ is Gcontractible, in
symbols γ ∼
G
0.
The fact that an integral over a point is zero has the following immediate consequence.
Corollary 4.5. Suppose G ⊆ C is open, f is analytic in G, and γ ∼
G
0 via a homotopy with
continuous second partials. Then
_
γ
f = 0 .
1
For more information about Gauß, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Gauss.html.
2
For more information about Karl Theodor Wilhelm Weierstraß (1815–1897), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Weierstrass.html.
3
For more information about Edouard JeanBaptiste Goursat (1858–1936), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Goursat.html.
CHAPTER 4. INTEGRATION 42
Figure 4.2: This ellipse is (C \ R)contractible.
The fact that any closed curve is Ccontractible (Exercise 11a) yields the following special case
of the previous specialcase corollary.
Corollary 4.6. If f is entire and γ is any smooth closed curve then
_
γ
f = 0 .
Proof of Theorem 4.4. Suppose h is the homotopy, and γ
s
is the curve parametrized by h(t, s), 0 ≤
t ≤ 1. Consider the integral
I(s) =
_
γs
f
as a function in s (so I(0) =
_
γ
1
f and I(1) =
_
γ
2
f). We will show that I is constant with respect
to s, and hence the statement of the theorem follows with I(0) = I(1). To prove that I is constant,
we use Theorem 1.12 (Leibniz’s rule), combined with Theorem 1.9 (the fundamental theorem of
calculus).
d
ds
I(s) =
d
ds
_
1
0
f (h(t, s))
∂h
∂t
dt =
_
1
0
∂
∂s
_
f (h(t, s))
∂h
∂t
_
dt
=
_
1
0
f
(h(t, s))
∂h
∂s
∂h
∂t
+f (h(t, s))
∂
2
h
∂t∂s
dt =
_
1
0
∂
∂t
_
f (h(t, s))
∂h
∂s
_
dt
= f (h(1, s))
∂h
∂s
(1, s) −f (h(0, s))
∂h
∂s
(0, s) = 0 .
In the last step we used the third property (according to Deﬁnition 4.2) of the homotopy h.
Note also that in the second line, we use the fact that h has continuous second partials and hence
∂
2
h
∂t∂s
=
∂
2
h
∂s∂t
.
4.4 Cauchy’s Integral Formula
Cauchy’s Theorem 4.4 yields almost immediately the following helpful result.
CHAPTER 4. INTEGRATION 43
Theorem 4.7 (Cauchy’s Integral Formula for a Circle). Let C
R
be the counterclockwise circle with
radius R centered at w and suppose f is analytic at each point of the closed disk D bounded by C
R
.
Then
f(w) =
1
2πi
_
C
R
f(z)
z −w
dz .
Proof. All circles C
r
with center w and radius r are homotopic in D \ {w}, and the function
f(z)/(z −w) is analytic in an open set containing D \ {w}. So Cauchy’s Theorem 4.4, gives
_
C
R
f(z)
z −w
dz =
_
Cr
f(z)
z −w
dz
Now by Exercise 8,
_
Cr
1
z −w
dz = 2πi ,
and we obtain with Proposition 4.1(d)
¸
¸
¸
¸
_
C
R
f(z)
z −w
dz −2πif(w)
¸
¸
¸
¸
=
¸
¸
¸
¸
_
Cr
f(z)
z −w
dz −f(w)
_
Cr
1
z −w
dz
¸
¸
¸
¸
=
¸
¸
¸
¸
_
Cr
f(z) −f(w)
z −w
dz
¸
¸
¸
¸
≤ max
z∈Cr
¸
¸
¸
¸
f(z) −f(w)
z −w
¸
¸
¸
¸
length (C
r
) = max
z∈Cr
f(z) −f(w)
r
2πr
= 2π max
z∈Cr
f(z) −f(w) .
On the righthand side, we can now take r as small as we want, and—because f is continuous
at w—this means we can make f(z) −f(w) as small as we like. Hence the lefthand side has no
choice but to be zero, which is what we claimed.
This is a useful theorem by itself, but it can be made more generally useful. For example, it
will be important to have Cauchy’s integral formula when w is anywhere inside C
R
, not just at the
center of C
R
. In fact, in many cases in which a point w is inside a simple closed curve γ we can see
a homotopy from γ to a small circle around w so that the homotopy misses w and remains in the
region where f is analytic. In that case the theorem remains true, since, by Cauchy’s theorem, the
integral of f(z)/(z −w) around γ is the same as the integral of f(z)/(z −w) around a small circle
centered at w, and Theorem 4.7 then applies to evaluate the integral. In this discussion we need to
be sure that the orientation of the curve γ and the circle match. In general, we say a simple closed
curve γ is positively oriented if it is parameterized so that the inside is on the left of γ. For a circle
this corresponds to a counterclockwise orientation.
Here’s the general form:
Theorem 4.8 (Cauchy’s Integral Formula). Suppose f is analytic on the region G, w ∈ G, and γ
is a positively oriented, simple, closed, smooth, Gcontractible curve such that w is inside γ. Then
f(w) =
1
2πi
_
γ
f(z)
z −w
dz .
CHAPTER 4. INTEGRATION 44
We have already indicated how to prove this, by combining Cauchy’s theorem and the special
case, Theorem 4.7. All we need is to ﬁnd a homotopy in G \ {w} between γ and a small circle
with center at w. In all practical cases we can see immediately how to construct such a homotopy,
but it is not at all clear how to do so in complete generality; in fact, it is not even clear how to
make sense of the “inside” of γ in general. The justiﬁcation for this is one of the ﬁrst substantial
theorems ever proved in topology. We can state it as follows:
Theorem 4.9 (Jordan Curve Theorem). If γ is a positively oriented, simple, closed curve in C
then C \ γ consists of two connected open sets, the inside and the outside of γ. If a closed disk D
centered at w lies inside γ then there is a homotopy γ
s
from γ to the positively oriented boundary
of D, and, for 0 < s < 1, γ
s
is inside γ and outside of D.
This theorem, although “intuitively obvious,” is surprisingly diﬃcult to prove. The usual state
ment of the Jordan curve theorem does not contain the homotopy information; we have borrowed
this from a companion theorem to the Jordan curve theorem which is sometimes called the “annulus
theorem.” If you want to explore this kind of mathematics you should take a course in topology.
A nice special case of Cauchy’s formula is obtained when γ is a circle centered at w, parametrized
by, say, z = w +re
it
, 0 ≤ t ≤ 2π. Theorem 4.8 gives (if the conditions are met)
f(w) =
1
2πi
_
2π
0
f
_
w +re
it
_
w +re
it
−w
ire
it
dt =
1
2π
_
2π
0
f
_
w +re
it
_
dt .
Even better, we automatically get similar formulas for the real and imaginary part of f, simply
by taking real and imaginary parts on both sides. These identities have the ﬂavor of mean values.
Let’s summarize them in the following statement, which is often called a meanvalue theorem.
Corollary 4.10. Suppose f is analytic on and inside the circle z = w +re
it
, 0 ≤ t ≤ 2π. Then
f(w) =
1
2π
_
2π
0
f
_
w +re
it
_
dt .
Furthermore, if f = u +iv,
u(w) =
1
2π
_
2π
0
u
_
w +re
it
_
dt and v(w) =
1
2π
_
2π
0
v
_
w +re
it
_
dt .
Exercises
1. Integrate the function f(z) = z over the three curves given in Example 4.1.
2. Evaluate
_
γ
1
z
dz where γ(t) = sin t +i cos t, 0 ≤ t ≤ 2π.
3. Integrate the following functions over the circle z = 2, oriented counterclockwise:
(a) z +z.
(b) z
2
−2z + 3.
(c) 1/z
4
.
(d) xy.
CHAPTER 4. INTEGRATION 45
4. Evaluate the integrals
_
γ
xdz,
_
γ
y dz,
_
γ
z dz and
_
γ
z dz along each of the following paths.
Note that you can get the second two integrals very easily after you calculate the ﬁrst two,
by writing z and z as x ±iy.
(a) γ is the line segment form 0 to 1 −i.
(b) γ is the counterclockwise circle z = 1.
(c) γ is the counterclockwise circle z −a = r. Use γ(t) = a +re
it
.
5. Evaluate
_
γ
e
3z
dz for each of the following paths:
(a) The straight line segment from 1 to i.
(b) The circle z = 3.
(c) The parabola y = x
2
from x = 0 to x = 1.
6. Evaluate
_
γ
¸
¸
z
2
¸
¸
dz where γ is the parabola with parametric equation γ(t) = t+it
2
, 0 ≤ t ≤ 1.
7. Evaluate
_
γ
z
1
2
dz where γ is the unit circle and z
1
2
is the principal branch. You can use the
parameterization γ(θ) = e
iθ
for −π ≤ θ ≤ π, and remember that the principal branch is
deﬁned by z
1
2
=
√
re
iθ/2
if z = re
iθ
for −π ≤ θ ≤ π.
8. Let γ be the circle with radius r centered at w, oriented counterclockwise. You can parame
terize this curve as z(t) = w +re
it
for 0 ≤ t ≤ 2π. Show that
_
γ
dz
z −w
= 2πi .
9. Suppose a smooth curve is parametrized by both γ(t), a ≤ t ≤ b and σ(t), c ≤ t ≤ d, and let
τ : [c, d] →[a, b] be the map which “takes γ to σ,” that is, σ = γ ◦ τ. Show that
_
d
c
f(σ(t))σ
(t) dt =
_
b
a
f(γ(t))γ
(t) dt .
(In other words, our deﬁnition of the integral
_
γ
f is independent of the parametrization of γ.)
10. Prove that ∼
G
is an equivalence relation.
11. (a) Prove that any closed curve is Ccontractible.
(b) Prove that any two closed curves are Chomotopic.
12. Show that
_
γ
z
n
dz = 0 for any closed smooth γ and any integer n = −1. [If n is negative,
assume that γ does not pass through the origin, since otherwise the integral is not deﬁned.]
13. Exercise 12 excluded n = −1 for a very good reason: Exercises 2 and 8 (with w = 0) give
counterexamples. Generalizing these, if m is any integer then ﬁnd a closed curve γ so that
_
γ
z
−1
dz = 2mπi. (Hint: Follow the counterclockwise unit circle through m complete cycles
(for m > 0). What should you do if m < 0? What if m = 0?)
CHAPTER 4. INTEGRATION 46
14. Let γ
r
be the circle centered at 2i with radius r, oriented counterclockwise. Compute
_
γr
dz
z
2
+ 1
.
(This integral depends on r.)
15. Suppose p is a polynomial and γ is a closed smooth path in C. Show that
_
γ
p = 0 .
16. Compute the real integral
_
2π
0
dθ
2 + sin θ
by writing the sine function in terms of the exponential function and making the substitution
z = e
iθ
to turn the real into a complex integral.
17. Show that F(z) =
i
2
Log(z + i) −
i
2
Log(z − i) is a primitive of
1
1+z
2
for Re(z) > 0. Is
F(z) = arctan z?
18. Prove the following integration by parts statement. Let f and g be analytic in G, and suppose
γ ⊂ G is a smooth curve from a to b. Then
_
γ
fg
= f(γ(b))g(γ(b)) −f(γ(a))g(γ(a)) −
_
γ
f
g .
19. Suppose f and g are analytic on the region G, γ is a closed, smooth, Gcontractible curve,
and f(z) = g(z) for all z ∈ γ. Prove that f(z) = g(z) for all z inside γ.
20. This exercise gives an alternative proof of Cauchy’s integral formula (Theorem 4.8), which
does not depend on Cauchy’s Theorem 4.4. Suppose f is analytic on the region G, w ∈ G,
and γ is a positively oriented, simple, closed, smooth, Gcontractible curve such that w is
inside γ.
(a) Consider the function g : [0, 1] →C, g(t) =
_
γ
f(w+t(z−w))
z−w
dz. Show that g
= 0. (Hint:
Use Theorem 1.12 (Leibniz’s rule) and then ﬁnd a primitive for
∂f
∂t
(z +t(w −z)).)
(b) Prove Theorem 4.8 by evaluating g(0) and g(1).
21. Prove Corollary 4.5 using Theorem 4.8.
22. Suppose a is a complex number and γ
0
and γ
1
are two counterclockwise circles (traversed just
once) so that a is inside both of them. Explain geometrically why γ
0
and γ
1
are homotopic
in C \ {a} .
23. Let γ
r
be the counterclockwise circle with center at 0 and radius r. Find
_
γr
dz
z−a
. You should
get diﬀerent answers for r < a and r > a. (Hint: In one case γ
r
is contractible in C \ {a}.
In the other you can combine Exercises 8 and 22.)
CHAPTER 4. INTEGRATION 47
24. Let γ
r
be the counterclockwise circle with center at 0 and radius r. Find
_
γr
dz
z
2
−2z−8
for r = 1,
r = 3 and r = 5. (Hint: Since z
2
− 2z − 8 = (z − 4)(z + 2) you can ﬁnd a partial fraction
decomposition of the form
1
z
2
−2z−8
=
A
z−4
+
B
z+2
. Now use Exercise 23.)
25. Use the Cauchy integral formula to evaluate the integral in Exercise 24 when r = 3. (Hint:
The integrand can be written in each of following ways:
1
z
2
−2z −8
=
1
(z −4)(z + 2)
=
1/(z −4)
z + 2
=
1/(z + 2)
z −4
.
Which of these forms corresponds to the Cauchy integral formula for the curve γ
3
?)
26. Compute the following integrals, where C is the boundary of the square with corners at
±4 ±4i:
(a)
_
C
e
z
z
3
dz.
(b)
_
C
e
z
(z −πi)
4
dz.
(c)
_
C
sin(2z)
(z −π)
4
dz.
(d)
_
C
e
z
cos(z)
(z −π)
3
dz.
Chapter 5
Consequences of Cauchy’s Theorem
If things are nice there is probably a good reason why they are nice: and if you do not know at
least one reason for this good fortune, then you still have work to do.
Richard Askey
5.1 Extensions of Cauchy’s Formula
We now derive formulas for f
and f
which resemble Cauchy’s formula (Theorem 4.8).
Theorem 5.1. Suppose f is analytic on the region G, w ∈ G, and γ is a positively oriented, simple,
closed, smooth, Gcontractible curve such that w is inside γ. Then
f
(w) =
1
2πi
_
γ
f(z)
(z −w)
2
dz
and
f
(w) =
1
πi
_
γ
f(z)
(z −w)
3
dz .
This innocentlooking theorem has a very powerful consequence: just from knowing that f is
analytic we know of the existence of f
, that is, f
is also analytic in G. Repeating this argument
for f
, then for f
, f
, etc., gives the following statement, which has no analog whatsoever in the
reals.
Corollary 5.2. If f is diﬀerentiable in the region G then f is inﬁnitely diﬀerentiable in G.
Proof of Theorem 5.1. The idea of the proof is very similar to the proof of Cauchy’s integral formula
(Theorem 4.8). We will study the following diﬀerence quotient, which we can rewrite as follows by
Theorem 4.8.
f(w + ∆w) −f(w)
∆w
=
1
∆w
_
1
2πi
_
γ
f(z)
z −(w + ∆w)
dz −
1
2πi
_
γ
f(z)
z −w
dz
_
=
1
2πi
_
γ
f(z)
(z −w −∆w)(z −w)
dz .
48
CHAPTER 5. CONSEQUENCES OF CAUCHY’S THEOREM 49
Hence we will have to show that the following expression gets arbitrarily small as ∆w →0:
f(w + ∆w) −f(w)
∆w
−
1
2πi
_
γ
f(z)
(z −w)
2
dz =
1
2πi
_
γ
f(z)
(z −w −∆w)(z −w)
−
f(z)
(z −w)
2
dz
= ∆w
1
2πi
_
γ
f(z)
(z −w −∆w)(z −w)
2
dz .
This can be made arbitrarily small if we can show that the integral stays bounded as ∆w → 0.
In fact, by Proposition 4.1(d), it suﬃces to show that the integrand stays bounded as ∆w → 0
(because γ and hence length(γ) are ﬁxed). Let M = max
z∈γ
f(z). Since γ is a closed set, there is
some positive δ so that the open disk of radius δ around w does not intersect γ; that is, z −w ≥ δ
for all z on γ. By the reverse triangle inequality we have for all z ∈ γ
¸
¸
¸
¸
f(z)
(z −w −∆w)(z −w)
2
¸
¸
¸
¸
≤
f(z)
(z −w −∆w)z −w
2
≤
M
(δ −∆w)δ
2
,
which certainly stays bounded as ∆w →0. The proof of the formula for f
is very similar and will
be left for the exercises (see Exercise 1).
Remarks. 1. Theorem 5.1 suggests that there are similar looking formulas for the higher derivatives
of f. This is in fact true, and theoretically one could obtain them one by one with the methods
of the proof of Theorem 5.1. However, once we start studying power series for analytic functions,
we will obtain such a result much more easily; so we save the derivation of formulas for higher
derivatives of f for later (see Corollary 8.6).
2. Theorem 5.1 can also be used to compute certain integrals. We give some examples of this
application next.
Example 5.1.
_
z=1
sin(z)
z
2
dz = 2πi
d
dz
sin(z)
¸
¸
¸
¸
z=0
= 2πi cos(0) = 2πi .
Example 5.2. To compute the integral
_
z=2
dz
z
2
(z −1)
,
we ﬁrst split up the integration path as illustrated in Figure 5.1: Introduce an additional path which
separates 0 and 1. If we integrate on these two new closed paths (γ
1
and γ
2
) counterclockwise, the
two contributions along the new path will cancel each other. The eﬀect is that we transformed an
integral, for which two singularities where inside the integration path, into a sum of two integrals,
each of which has only one singularity inside the integration path; these new integrals we know
CHAPTER 5. CONSEQUENCES OF CAUCHY’S THEOREM 50
γ
1
γ
2
0
1
Figure 5.1: Example 5.2
how to deal with.
_
z=2
dz
z
2
(z −1)
=
_
γ
1
dz
z
2
(z −1)
+
_
γ
2
dz
z
2
(z −1)
=
_
γ
1
1
z−1
z
2
dz +
_
γ
2
1
z
2
z −1
dz
= 2πi
d
dz
1
z −1
¸
¸
¸
¸
z=0
+ 2πi
1
1
2
= 2πi
_
−
1
(−1)
2
_
+ 2πi
= 0 .
Example 5.3.
_
z=1
cos(z)
z
3
dz = πi
d
2
dz
2
cos(z)
¸
¸
¸
¸
z=0
= πi (−cos(0)) = −πi .
5.2 Taking Cauchy’s Formula to the Limit
Many beautiful applications of Cauchy’s formula arise from considerations of the limiting behavior
of the formula as the curve gets arbitrarily large. We shall look at a few applications along these
lines in this section, but this will be a recurring theme throughout the rest of the book.
The ﬁrst application is understanding the roots of polynomials. As a preparation we prove
the following inequality, which is generally quite useful. It simply says that for large enough z, a
polynomial of degree d looks almost like a constant times z
d
.
Lemma 5.3. Suppose p(z) is a polynomial of degree d with leading coeﬃcient a
d
. Then there is
real number R
0
so that
1
2
a
d
 z
d
≤ p(z) ≤ 2 a
d
 z
d
for all z satisfying z ≥ R
0
.
CHAPTER 5. CONSEQUENCES OF CAUCHY’S THEOREM 51
Proof. Since p(z) has degree d its leading coeﬃcient a
d
is not zero, and we can factor out a
d
z
d
:
p(z) =
¸
¸
¸a
d
z
d
+a
d−1
z
d−1
+a
d−2
z
d−2
+· · · +a
1
z +a
0
¸
¸
¸
= a
d
 z
d
¸
¸
¸
¸
1 +
a
d−1
a
d
z
+
a
d−2
a
d
z
2
+· · · +
a
1
a
d
z
d−1
+
a
0
a
d
z
d
¸
¸
¸
¸
.
Then the sum inside the last factor has limit 1 as z →∞ so its modulus is between
1
2
and 2 for all
large enough z.
Theorem 5.4 (Fundamental theorem of algebra
1
). Every nonconstant polynomial has a root in C.
Proof
2
. Suppose (by way of contradiction) that p does not have any roots, that is, p(z) = 0 for all
z ∈ C. Then Cauchy’s formula gives us
1
p(0)
=
1
2πi
_
γ
R
1/p(z)
z
dz
where γ
R
is the circle of radius R around the origin. Notice that the value of the integral does not
depend on R, so we have
1
p(0)
= lim
R→∞
1
2πi
_
γ
R
dz
z p(z)
. (∗)
But now we can see that the limit of the integral is 0: By Lemma 5.3 we have z p(z) ≥
1
2
a
d
 z
d+1
for all large z, where d is the degree of p(z) and a
d
is the leading coeﬃcient of p(z). Hence, using
Proposition 4.1(d) and the formula for the circumference of a circle we see that the integral can be
bounded as ¸
¸
¸
¸
1
2πi
_
γ
R
dz
zp(z)
¸
¸
¸
¸
≤
1
2π
·
2
a
d
 R
d+1
· (2πR) =
2
a
d
 R
d
and this has limit 0 as R → ∞. But, plugging into (∗), we have shown that
1
p(0)
= 0, which is
impossible.
Remarks. 1. This statement implies that any polynomial p can be factored into linear terms of
the form z −a where a is a root of p, as we can apply the corollary, after getting a root a, to
p(z)
z−a
(which is again a polynomial by the division algorithm), etc. (see also Exercise 8).
2. A compact reformulation of the fundamental theorem of algebra is to say that C is algebraically
closed.
A powerful consequence of (the ﬁrst half of) Theorem 5.1 is the following.
Corollary 5.5 (Liouville’s
3
Theorem
4
). Every bounded entire function is constant.
1
The fundamental theorem of algebra was ﬁrst proved by Gauß (in his doctoral dissertation), although its statement
had been assumed to be correct long before Gauß’s times.
3
For more information about Joseph Liouville (1809–1882), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Liouville.html.
4
This theorem is for historical reasons erroneously attributed to Liouville. It was published earlier by Cauchy; in
fact, Gauß may well have known about it before Cauchy’s times.
CHAPTER 5. CONSEQUENCES OF CAUCHY’S THEOREM 52
Proof. Suppose f(z) ≤ M for all z ∈ C. Given any w ∈ C, we apply Theorem 5.1 with the circle
γ
R
of radius R centered at w. Note that we can choose any R because f is entire. Now we apply
Proposition 4.1 (d), remembering that γ
R
has circumference 2πR and z −w = R for all z on γ
R
:
¸
¸
f
(w)
¸
¸
=
¸
¸
¸
¸
1
2πi
_
γ
R
f(z)
(z −w)
2
dz
¸
¸
¸
¸
≤
1
2π
max
z∈γ
R
¸
¸
¸
¸
f(z)
(z −w)
2
¸
¸
¸
¸
· 2πR =
1
2π
max
z∈γ
R
f(z)
R
2
2πR = max
z∈γ
f(z)
R
≤
M
R
.
The righthand side can be made arbitrary small, as we are allowed to make R as large as we want.
This implies that f
= 0, and hence, by Theorem 2.7, f is constant.
As an example of the usefulness of Liouville’s theorem we give another proof of the fundamental
theorem of algebra, which is close to Gauß’s original proof:
Another proof of the fundamental theorem of algebra. Suppose (by way of contradiction) that p
does not have any roots, that is, p(z) = 0 for all z ∈ C. Then, because p is entire, the func
tion f(z) =
1
p(z)
is entire. But f →0 as z becomes large as a consequence of Lemma 5.3; that is,
f is also bounded (Exercise 7). Now apply Corollary 5.5 to deduce that f is constant. Hence p is
constant, which contradicts our assumptions.
As one more example of this theme of getting results from Cauchy’s formula by taking the limit
as a path goes to inﬁnity, we compute an improper integral.
Let σ be the counterclockwise semicircle formed by the segment S of the real axis from −R to
R, followed by the circular arc T of radius R in the upper half plane from R to −R, where R > 1.
We shall integrate the function
f(z) =
1
z
2
+ 1
=
1/(z +i)
z −i
=
g(z)
z −i
, where g(z) =
1
z +i
Since g(z) is analytic inside and on σ and i is inside σ, we can apply Cauchy’s formula:
1
2πi
_
σ
dz
z
2
+ 1
=
1
2πi
_
σ
g(z)
z −i
dz = g(i) =
1
i +i
=
1
2i
,
and so
_
S
dz
z
2
+ 1
+
_
T
dz
z
2
+ 1
=
_
σ
dz
z
2
+ 1
= 2πi ·
1
2i
= π. (∗∗)
Now this formula holds for all R > 1, so we can take the limit as R → ∞. First,
¸
¸
z
2
+ 1
¸
¸
≥
1
2
z
2
for large enough z by Lemma 5.3, so we can bound the integral over T using Proposition 4.1(d):
¸
¸
¸
¸
_
T
dz
z
2
+ 1
¸
¸
¸
¸
≤
2
R
2
· πR =
2
R
and this has limit 0 as R →∞. On the other hand, we can parameterize the integral over S using
z = t, −R ≤ t ≤ R, obtaining
_
S
dz
z
2
+ 1
=
_
R
−R
dt
1 +t
2
.
CHAPTER 5. CONSEQUENCES OF CAUCHY’S THEOREM 53
As R → ∞ this approaches an improper integral. Making these observations in the limit of the
formula (∗∗) as R →∞ now produces
_
∞
−∞
dt
t
2
+ 1
= π.
Of course this integral can be evaluated almost as easily using standard formulas from calculus.
However, just a slight modiﬁcation of this example leads to an improper integral which is far beyond
the scope of basic calculus; see Exercise 11.
5.3 Antiderivatives Revisited and Morera’s Theorem
A region G is said to be simply connected if every closed curve in G is Gcontractible. This concept
allows the following result.
Theorem 5.6. Suppose f is analytic in the simplyconnected region G. Then f has a primitive
in G.
Proof. Fix a point a ∈ G and let
F(z) =
_
γz
f
where γ
z
is any smooth curve from a to z. We should make sure that F is well deﬁned: Suppose
δ
z
is another smooth curve from a to z then γ
z
− δ
z
is closed and Gcontractible, as G is simply
connected. Hence by Corollary 4.5
0 =
_
γz−δz
f =
_
γz
f −
_
δz
f
which means we get the same integral no matter which path we take from a to z, so F is a
welldeﬁned function. It remains to show that F is a primitive of f:
F
(z) = lim
h→0
F(z +h) −F(z)
h
= lim
h→0
1
h
_
_
γ
z+h
f −
_
γz
f
_
.
Now let δ be a smooth curve in G from z to z +h. Then γ
z
+δ −γ
z+h
is a closed smooth curve
in G, and it is Gcontractible as G is simply connected. Hence again Corollary 4.5 gives us
_
γz
f +
_
δ
f −
_
γ
z+h
f = 0 ,
that is,
F
(z) = lim
h→0
1
h
_
_
γ
z+h
f −
_
γz
f
_
= lim
h→0
1
h
_
δ
f .
(One should keep in mind that δ very much depends on z and h.) If h is suﬃciently small, the
line segment l(z, z +h) between z and z +h lies in G and by Corollary 4.5 (again we use that G is
simply connected)
F
(z) = lim
h→0
1
h
_
δ
f = lim
h→0
1
h
_
l(z,z+h)
f .
CHAPTER 5. CONSEQUENCES OF CAUCHY’S THEOREM 54
γ
z+h
γ
z
a
z
z +h
δ
G
Figure 5.2: Proof of Theorem 5.6.
Now because
_
l(z,z+h)
f(z) dw = f(z)
_
l(z,z+h)
dw = f(z) h, we obtain
¸
¸
F
(z) −f(z)
¸
¸
=
¸
¸
¸
¸
¸
lim
h→0
1
h
_
l(z,z+h)
f(w) dw − lim
h→0
1
h
_
l(z,z+h)
f(z) dw
¸
¸
¸
¸
¸
= lim
h→0
¸
¸
¸
¸
¸
1
h
_
l(z,z+h)
f(w) −f(z) dw
¸
¸
¸
¸
¸
≤ lim
h→0
1
h
max
w∈l(z,z+h)
f(w) −f(z) length(l(z, z +h))
= lim
h→0
max
w∈l(z,z+h)
f(w) −f(z) = 0 .
The last equality follows from the continuity of f.
There is an interesting consequence to be drawn from this theorem. It follows from the fact
that a primitive of a function is, by deﬁnition, diﬀerentiable. This means that the primitive of
a function f obtained by Theorem 5.6 has itself a primitive, which has a primitive, which has a
primitive, which has . . . This is the same behavior which we discovered in Corollary 5.2 ‘in the other
direction.’
Another consequence comes from the proof of Theorem 5.6: we did not really need the fact
that every closed curve in G is contractible, just that every closed curve gives a zero integral for f.
This fact can be exploited to give a sort of converse statement to Corollary 4.5.
Corollary 5.7 (Morera’s
5
Theorem). Suppose f is continuous in the region G and
_
γ
f = 0
for all smooth closed paths γ ⊂ G. Then f is analytic in G.
5
For more information about Giancinto Morera (1856–1907), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Morera.html.
CHAPTER 5. CONSEQUENCES OF CAUCHY’S THEOREM 55
Proof. As in the previous proof, we ﬁx an a ∈ G and deﬁne
F(z) =
_
γz
f ,
where γ
z
is any smooth curve in G from a to z. As above, this is a welldeﬁned function because
all closed paths give a zero integral for f; and exactly as above we can show that F is a primitive
for f in G. Because F is analytic on G, Corollary 5.2 gives that f is also analytic on G.
Exercises
1. Prove the formula for f
in Theorem 5.1.
2. Integrate the following functions over the circle z = 3, oriented counterclockwise:
(a) Log(z −4i).
(b)
1
z−
1
2
.
(c)
1
z
2
−4
.
(d)
exp z
z
3
.
(e)
_
cos z
z
_
2
.
(f) i
z−3
.
(g)
sin z
(z
2
+
1
2
)
2
.
(h)
exp z
(z−w)
2
, where w is any ﬁxed complex number with w = 3.
(i)
1
(z+4)(z
2
+1)
.
3. Prove that
_
γ
z exp
_
z
2
_
dz = 0 for any closed curve γ.
4. Show that exp(sin z) has an antiderivative on C.
5. Find a (maximal size) set on which f(z) = exp
_
1
z
_
has an antiderivative. (How does this
compare with the real function f(x) = e
1/x
?)
6. Compute the following integrals; use the principal value of z
i
. (Hint: one of these integrals
is considerably easier than the other.)
(a)
_
γ
1
z
i
dz where γ
1
(t) = e
it
, −
π
2
≤ t ≤
π
2
.
(b)
_
γ
2
z
i
dz where γ
2
(t) = e
it
,
π
2
≤ t ≤
3π
2
.
7. Suppose f is continuous on C and lim
z→∞
f(z) = 0. Show that f is bounded. (Hint: From
the deﬁnition of limit at inﬁnity (with = 1) there is R > 0 so that f(z) −0 = f (z) < 1
if z > R. Is f bounded for z ≤ R?)
CHAPTER 5. CONSEQUENCES OF CAUCHY’S THEOREM 56
8. Let p be a polynomial of degree n > 0. Prove that there exist complex numbers c, z
1
, z
2
, . . . , z
k
and positive integers j
1
, . . . , j
k
such that
p(z) = c (z −z
1
)
j
1
(z −z
2
)
j
2
· · · (z −z
k
)
j
k
,
where j
1
+· · · +j
k
= n.
9. Show that a polynomial of odd degree with real coeﬃcients must have a real zero. (Hint:
Exercise 14b in Chapter 1.)
10. Suppose f is entire and there exist constants a, b such that f(z) ≤ az + b for all z ∈ C.
Prove that f is a linear polynomial (that is, of degree ≤ 1).
11. In this problem F(z) =
e
iz
z
2
+1
and R > 1. Modify the example at the end of Section 5.2:
(a) Show that
_
σ
F(z) dz =
π
e
if σ is the counterclockwise semicircle formed by the segment
S of the real axis from −R to R, followed by the circular arc T of radius R in the upper
half plane from R to −R.
(b) Show that
¸
¸
e
iz
¸
¸
≤ 1 for z in the upper half plane, and conclude that F(z) ≤
2
z
2
for z
large enough.
(c) Show that lim
R→∞
_
T
F(z) dz = 0, and hence lim
R→∞
_
S
F(z) dz =
π
e
.
(d) Conclude, by parameterizing the integral over S in terms of t and just considering the
real part, that
_
∞
−∞
cos(t)
t
2
+1
dx =
π
e
.
Chapter 6
Harmonic Functions
The shortest route between two truths in the real domain passes through the complex domain.
J. Hadamard
6.1 Deﬁnition and Basic Properties
We will now spend a chapter on certain functions deﬁned on subsets of the complex plane which
are real valued. The main motivation for studying them is that the partial diﬀerential equation
they satisfy is very common in the physical sciences.
Deﬁnition 6.1. Let G ⊆ C be a region. A function u : G →R is harmonic in G if it has continuous
second partials in G and satisﬁes the Laplace
1
equation
u
xx
+u
yy
= 0
in G.
There are (at least) two reasons why harmonic functions are part of the study of complex
analysis, and they can be found in the next two theorems.
Proposition 6.1. Suppose f = u + iv is analytic in the region G. Then u and v are harmonic
in G.
Proof. First, by Corollary 5.2, f is inﬁnitely diﬀerentiable, and hence so are u and v. In particular,
u and v have continuous second partials. By Theorem 2.6, u and v satisfy the Cauchy–Riemann
equations
u
x
= v
y
and u
y
= −v
x
in G. Hence
u
xx
+u
yy
= (u
x
)
x
+ (u
y
)
y
= (v
y
)
x
+ (−v
x
)
y
= v
yx
−v
xy
= 0
in G. Note that in the last step we used the fact that v has continuous second partials. The proof
that v satisﬁes the Laplace equation is completely analogous.
1
For more information about PierreSimon Laplace (1749–1827), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Laplace.html.
57
CHAPTER 6. HARMONIC FUNCTIONS 58
Proposition 6.1 shouts for a converse theorem. There are, however, functions which are harmonic
in a region G but not the real part (say) of an analytic function in G (Exercise 3). We do obtain a
converse of Proposition 6.1 if we restrict ourselves to simply connected regions.
Theorem 6.2. Suppose u is harmonic on the simply connected region G. Then there exists a
harmonic function v such that f = u +iv is analytic in G.
Remark. The function v is called a harmonic conjugate of u.
Proof. We will explicitly construct the analytic function f (and thus v = Imf). First, let
g = u
x
−iu
y
.
The plan is to prove that g is analytic, and then to construct an antiderivative of g, which will
be almost the function f that we’re after. To prove that g is analytic, we use Theorem 2.6: ﬁrst
because u is harmonic, Re g = u
x
and Img = −u
y
have continuous partials. Moreover, again
because u is harmonic, they satisfy the Cauchy–Riemann equations:
(Re g)
x
= u
xx
= −u
yy
= (Img)
y
and
(Re g)
y
= u
xy
= u
yx
= −(Img)
x
.
Now that we know that g is analytic in G, we can use Theorem 5.6 to obtain a primitive h of g on
G. (Note that for the application of this theorem we need the fact that G is simply connected.)
Suppose we decompose h into its real and imaginary parts as h = a + ib. Then, again using
Theorem 2.6,
g = h
= a
x
+ib
x
= a
x
−ia
y
.
(The second equation follows with the Cauchy–Riemann equations.) But the real part of g is u
x
,
so that we obtain u
x
= a
x
or u(x, y) = a(x, y) + c(y) for some function c which only depends
on y. On the other hand, comparing the imaginary parts of g and h
yields −u
y
= −a
y
or
u(x, y) = a(x, y) +c(x), and c depends only on x. Hence c has to be constant, and u = a +c. But
then
f = h −c
is a function analytic in G whose real part is u, as promised.
Remark. In hindsight, it should not be surprising that the function g which we ﬁrst constructed is
the derivative of the soughtafter function f. Namely, by Theorem 2.6 such a function f = u + iv
must satisfy
f
= u
x
+iv
x
= u
x
−iu
y
.
(The second equation follows with the Cauchy–Riemann equations.) It is also worth mentioning
that the proof shows that if u is harmonic in G then u
x
is the real part of a function analytic in G
regardless whether G is simply connected or not.
As one might imagine, the two theorems we’ve just proved allow for a powerful interplay between
harmonic and analytic functions. In that spirit, the following theorem might appear not too
surprising. It is, however, a very strong result, which one might appreciate better when looking
back at the simple deﬁnition of harmonic functions.
CHAPTER 6. HARMONIC FUNCTIONS 59
Corollary 6.3. A harmonic function is inﬁnitely diﬀerentiable.
Proof. Suppose u is harmonic in G. Fix z
0
∈ G and r > 0 such that the disk
D = {z ∈ C : z −z
0
 < r}
is contained in G. D is simply connected, so by the last theorem, there exists a function f analytic
in D such that u = Re f on D. By Corollary 5.2, f is inﬁnitely diﬀerentiable on D, and hence so
is its real part u. Because z
0
∈ D, we showed that u is inﬁnitely diﬀerentiable at z
0
, and because
z
0
was chosen arbitrarily, we proved the statement.
Remark. This is the ﬁrst in a series of proofs which uses the fact that the property of being harmonic
is a local property—it is a property at each point of a certain region. Note that we did not construct
a function f which is analytic in G but we only constructed such a function on the disk D. This f
might very well diﬀer from one disk to the next.
6.2 MeanValue and Maximum/Minimum Principle
The following identity is the harmonic analog of Cauchy’s integral formula, Theorem 4.8.
Theorem 6.4. Suppose u is harmonic in the region G, and {z ∈ C : z −w ≤ r} ⊂ G. Then
u(w) =
1
2π
_
2π
0
u
_
w +re
it
_
dt .
Proof. The disk D = {z ∈ C : z − w ≤ r} is simply connected, so by Theorem 6.2 there is a
function f analytic on D such that u = Re f on D. Now we apply Corollary 4.10 to f:
f(w) =
1
2π
_
2π
0
f
_
w +re
it
_
dt .
The statement follows by taking the real part on both sides.
Theorem 6.4 states that harmonic functions have the meanvalue property. The following result
is a fairly straightforward consequence of this property. The function u : G ⊂ C →R has a strong
relative maximum at w if there exists a disk D = {z ∈ C : z −w < R} ⊂ G such that u(z) ≤ u(w)
for all z ∈ D and u(z
0
) < u(w) for some z
0
∈ D. The deﬁnition of a strong relative minimum is
completely analogous.
Theorem 6.5. If u is harmonic in the region G, then it does not have a strong relative maximum
or minimum in G.
Proof. Assume (by way of contradiction) that w is a strong local maximum of u in G. Then there
is a disk in G centered at w containing a point z
0
with u(z
0
) < u(w). Suppose z
0
−w = r; we
apply Theorem 6.4 with this r:
u(w) =
1
2π
_
2π
0
u
_
w +re
it
_
dt .
CHAPTER 6. HARMONIC FUNCTIONS 60
w
w +re
it
1
w +re
it
0
w +re
it
2
Figure 6.1: Proof of Theorem 6.5.
Intuitively, this cannot hold, because some of the function values we’re integrating are smaller than
u(w), contradicting the meanvalue property. To make this into a thorough argument, suppose that
z
0
= w +re
it
0
. Because u(z
0
) < u(w) and u is continuous, there is a whole interval of parameters,
say t
1
≤ t ≤ t
2
(and t
0
is among those t), such that u
_
w +re
it
_
< u(w).
Now we split up the meanvalue integral:
u(w) =
1
2π
_
2π
0
u
_
w +re
it
_
dt
=
1
2π
__
t
1
0
u
_
w +re
it
_
dt +
_
t
2
t
1
u
_
w +re
it
_
dt +
_
2π
t
2
u
_
w +re
it
_
dt
_
All the integrands can be bounded by u(w), for the middle integral we get a strict inequality. Hence
u(w) <
1
2π
__
t
1
0
u(w) dt +
_
t
2
t
1
u(w) dt
_
2π
t
2
u(w) dt
_
= u(w) ,
a contradiction. The same argument works if we assume that u has a relative minimum. But in this
case there’s actually a short cut: if u has a strong relative minimum then the harmonic function
−u has a strong relative maximum, which we just showed cannot exist.
A look into the (not so distant) future. We will see in Corollary 8.12 a variation of this theorem for
a weak relative maximum w, in the sense that there exists a disk D = {z ∈ C : z −w < R} ⊂ G
such that all z ∈ D satisfy u(z) ≤ u(w). Corollary 8.12 says that if u is harmonic in the region
G, then it does not have a weak relative maximum or minimum in G. A special yet important
case of the above maximum/minimum principle is obtained when considering bounded regions.
Corollary 8.12 implies that if u is harmonic in the closure of the bounded region G then
max
z∈G
u(z) = max
z∈∂G
u(z) and min
z∈G
u(z) = min
z∈∂G
u(z) .
(Here ∂G denotes the boundary of G.) We’ll exploit this fact in the next two corollaries.
Corollary 6.6. Suppose u is harmonic in the closure of the bounded region G. If u is zero on ∂G
then u is zero in G.
CHAPTER 6. HARMONIC FUNCTIONS 61
Proof. By the remark we just made
u(z) ≤ max
z∈G
u(z) = max
z∈∂G
u(z) = max
z∈∂G
0 = 0
and
u(z) ≥ min
z∈G
u(z) = min
z∈∂G
u(z) = min
z∈∂G
0 = 0 ,
so u has to be zero in G.
Corollary 6.7. If two harmonic functions agree on the boundary of a bounded region then they
agree in the region.
Proof. Suppose u and v are harmonic in G∪∂G and they agree on ∂G. Then u−v is also harmonic
in G∪ ∂G (Exercise 2) and u −v is zero on ∂G. Now apply the previous corollary.
The last corollary states that if we know a harmonic function on the boundary of some region
then we know it inside the region. One should remark, however, that this result is of a completely
theoretical nature: it says nothing about how to extend a function given on the boundary of a
region to the full region. This problem is called the Dirichlet
2
problem and has a solution for all
simplyconnected regions. There is a fairly simple formula (involving the socalled Poisson
3
kernel )
if the region in question is a disk; for other regions one needs to ﬁnd a conformal map to the unit
disk. All of this is beyond the scope of these notes, we just remark that Corollary 6.7 says that the
solution to the Dirichlet problem is unique.
Exercises
1. Show that all partial derivatives of a harmonic function are harmonic.
2. Suppose u and v are harmonic, and c ∈ R. Prove that u +cv is also harmonic.
3. Consider u(z) = u(x, y) = ln
_
x
2
+y
2
_
.
(a) Show that u is harmonic in C \ {0}.
(b) Prove that u is not the real part of a function which is analytic in C \ {0}.
4. Let u(x, y) = e
x
sin y.
(a) Show that u is harmonic on C.
(b) Find an entire function f such that Re(f) = u.
5. Is it possible to ﬁnd a real function v so that x
3
+y
3
+iv is analytic?
2
For more information about Johann Peter Gustav Dirichlet (1805–1859), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Dirichlet.html.
3
For more information about Sim´eon Denis Poisson (1781–1840), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Poisson.html.
Chapter 7
Power Series
It is a pain to think about convergence but sometimes you really have to.
Sinai Robins
7.1 Sequences and Completeness
As in the real case (and there will be no surprises in this chapter of the nature ‘real versus complex’),
a (complex) sequence is a function from the positive (sometimes the nonnegative) integers to the
complex numbers. Its values are usually denoted by a
n
(as opposed to, say, a(n)) and we commonly
denote the sequence by (a
n
)
∞
n=1
, (a
n
)
n≥1
, or simply (a
n
). The notion of convergence of a sequence
is based on the following sibling of Deﬁnition 2.1.
Deﬁnition 7.1. Suppose (a
n
) is a sequence and a ∈ C such that for all > 0, there is an integer
N such that for all n ≥ N, we have a
n
−a < . Then the sequence (a
n
) is convergent and a is its
limit, in symbols
lim
n→∞
a
n
= a .
If no such a exists then the sequence (a
n
) is divergent.
Example 7.1. lim
n→∞
i
n
n
= 0: Given > 0, choose N > 1/. Then for any n ≥ N,
¸
¸
¸
¸
i
n
n
−0
¸
¸
¸
¸
=
¸
¸
¸
¸
i
n
n
¸
¸
¸
¸
=
i
n
n
=
1
n
≤
1
N
< .
Example 7.2. The sequence (a
n
= i
n
) diverges: Given a ∈ C, choose = 1/2. We consider two
cases: If Re a ≥ 0, then for any N, choose n ≥ N such that a
n
= −1. (This is always possible since
a
4k+2
= i
4k+2
= −1 for any k ≥ 0.) Then
a −a
n
 = a + 1 ≥ 1 >
1
2
.
If Re a < 0, then for any N, choose n ≥ N such that a
n
= 1. (This is always possible since
a
4k
= i
4k
= 1 for any k > 0.) Then
a −a
n
 = a −1 ≥ 1 >
1
2
.
62
CHAPTER 7. POWER SERIES 63
The following limit laws are the relatives of the identities stated in Lemma 2.2.
Lemma 7.1. Let (a
n
) and (b
n
) be convergent sequences and c ∈ C.
(a) lim
n→∞
a
n
+c lim
n→∞
b
n
= lim
n→∞
(a
n
+c b
n
) .
(b) lim
n→∞
a
n
· lim
n→∞
b
n
= lim
n→∞
(a
n
· b
n
) .
(c)
lim
n→∞
a
n
lim
n→∞
b
n
= lim
n→∞
_
a
n
b
n
_
.
In the quotient law we have to make sure we do not divide by zero. Moreover, if f is continuous at
a then
lim
n→∞
f(a
n
) = f(a) if lim
n→∞
a
n
= a ,
where we require that a
n
be in the domain of f.
The most important property of the real number system is that we can, in many cases, determine
that a sequence converges without knowing the value of the limit. In this sense we can use the
sequence to deﬁne a real number. In fact, all irrational numbers are actually deﬁned this way, as
limits of rational numbers. This property of the real numbers is called completeness, and it can
be formulated in many equivalent ways. We will accept the following axiom as our version of the
completeness property:
Axiom (Monotone Sequence Property). Any bounded monotone sequence converges.
Remember that a sequence is monotone if it is either nondecreasing (x
n+1
≥ x
n
) or non
increasing (x
n+1
≤ x
n
).
Example 7.3. If 0 ≤ r < 1 then lim
n→∞
r
n
= 0: First, the sequence converges because it is
decreasing and bounded below by 0. If the limit is L then, using the laws of limits, we get
L = lim
n→∞
r
n
= lim
n→∞
r
n+1
= r lim
n→∞
r
n
= rL. From L = rL we get (1 −r)L = 0, so L = 0
since 1 −r = 0
The following is a consequence of the monotone sequence property, although it is often listed
as a separate axiom:
Theorem 7.2 (Archimedean Property). If x is any real number than there is an integer N which
is greater than x.
This essentially says that there are no inﬁnities in the reals. Notice that this was already
used in Example 7.1. For a proof see Exercise 3. It is interesting to see that the Archimedean
principle underlies the construction of an inﬁnite decimal expansion for any real number, while the
monotone sequence property shows that any such inﬁnite decimal expansion actually converges to
a real number.
We close this discussion of limits with a pair of standard limits. The ﬁrst of these can be
established by calculus methods (like L’Hospital’s rule, by treating n as the variable); both of them
can be proved by more elementary considerations.
Lemma 7.3. (a) Exponentials beat polynomials: lim
n→∞
b
n
p(n) = 0 if p(n) is a polynomial of
ﬁxed degree in n and b < 1.
(b) Factorials beat exponentials: lim
n→∞
a
n
n!
= 0 if a is a constant.
CHAPTER 7. POWER SERIES 64
7.2 Series
A series is a sequence (a
n
) whose members are of the form a
n
=
n
k=1
b
k
(or a
n
=
n
k=0
b
k
); here
(b
k
) is the sequence of terms of the series. The a
n
=
n
k=1
b
k
(or a
n
=
n
k=0
b
k
) are the partial
sums of the series. If we wanted to be lazy we would for convergence of a series simply refer to
convergence of the partial sums of the series, after all, we just deﬁned series through sequences.
However, there are some convergence features which take on special appearances for series, so we
should mention them here explicitly. For starters, a series converges to the limit (or sum) a by
deﬁnition if
lim
n→∞
a
n
= lim
n→∞
n
k=1
b
k
= a .
To express this in terms of Deﬁnition 7.1, for any > 0 we have to ﬁnd an N such that for all
n ≥ N
¸
¸
¸
¸
¸
n
k=1
b
k
−a
¸
¸
¸
¸
¸
< .
In the case of a convergent series, we usually express its limit as a =
∞
k=1
b
k
or a =
k≥1
b
k
.
Example 7.4. Occasionally we can ﬁnd the limit of a sequence by manipulating the partial sums:
k≥1
1
k(k + 1)
= lim
n→∞
n
k=1
_
1
k
−
1
k + 1
_
= lim
n→∞
__
1 −
1
2
_
+
_
1
2
−
1
3
_
+
_
1
3
−
1
4
_
+· · · +
_
1
n
−
1
n + 1
__
= lim
n→∞
_
1 −
1
2
+
1
2
−
1
3
+
1
3
−
1
4
+· · · +
1
n
−
1
n + 1
_
= lim
n→∞
_
1 −
1
n + 1
_
= 1.
A series where most of the terms cancel like this is called a telescoping series.
Most of the time we need to use the completeness property to check convergence of a series,
and it is fortunate that the monotone sequence property has a very convenient translation into the
language of series of real numbers. The partial sums of a series form a nondecreasing sequence if
the terms of the series are nonnegative, and this observation immediately yields:
Lemma 7.4. If b
k
are nonnegative real numbers then
∞
k=1
b
k
converges if and only if the partial
sums are bounded.
If b
k
are nonnegative real numbers and the partial sums of the series
∞
k=1
b
k
are unbounded
then the partial sums “converge” to inﬁnity, so we can write
∞
k=1
b
k
= ∞. Using this terminology,
we can rephrase Lemma 7.4 to say:
∞
k=1
b
k
converges in the reals if and only if it is ﬁnite.
We have already used the simple fact that convergence of a sequence (a
n
) is equivalent to the
convergence of (a
n−1
), and both of these sequences have the same limit. If a
n
is the n
th
partial
sum of the series
k≥1
b
k
then a
n
= a
n−1
+b
n
. From this we conclude:
Lemma 7.5. If
k≥1
b
k
converges then lim
n→∞
b
n
= 0.
CHAPTER 7. POWER SERIES 65
A common mistake is to try to use the converse of this result, but the converse is false:
Example 7.5. The harmonic series
k≥1
1
k
diverges (even though the limit of the general term is
0): If we assume the series converges, say to L, then we have
L = 1 +
1
2
+
1
3
+
1
4
+. . .
=
_
1 +
1
3
+
1
5
+. . .
_
+
_
1
2
+
1
4
+
1
6
+. . .
_
>
_
1
2
+
1
4
+
1
6
+. . .
_
+
_
1
2
+
1
4
+
1
6
+. . .
_
=
1
2
_
1 +
1
2
+
1
3
+
1
4
+. . .
_
+
1
2
_
1 +
1
2
+
1
3
+
1
4
+. . .
_
=
1
2
L +
1
2
L = L.
Here the inequality comes from
1
k
>
1
k+1
applied to each term in the ﬁrst sum in parentheses.
But now we have L > L, which is impossible.
There is one notion of convergence that’s special to series: we say that
k≥1
c
k
converges
absolutely if
k≥1
c
k
 < ∞. Be careful: We are deﬁning the phrase “converges absolutely,” but
this deﬁnition does not say anything about convergence of the series
k≥1
c
k
; we need a proof:
Theorem 7.6. If a series converges absolutely then it converges.
Proof. First consider the case when the terms c
k
are real. Deﬁne c
+
k
to be c
k
if c
k
≥ 0, or 0 if
c
k
< 0. Then c
+
k
≥ 0 and
k≥1
c
+
k
≤
k≥1
c
k
 < ∞ so
k≥1
c
+
k
converges; let P be its limit.
Similarly, deﬁne c
−
k
to be −c
k
if c
k
≤ 0, or 0 if c
k
> 0. Then c
−
k
≥ 0 and
k≥1
c
−
k
≤
k≥1
c
k
 < ∞
so
k≥1
c
−
k
converges; let N be its limit. Since c
k
= c
+
k
− c
−
k
we see that
k≥1
c
k
converges to
P −N.
In case c
k
is complex, write c
k
= a
k
+ib
k
where a
k
and b
k
are real. Then
k≥1
a
k
 ≤
k≥1
c
k
 <
∞ and
k≥1
b
k
 ≤
k≥1
c
k
 < ∞. By what we just proved, both
k≥1
a
k
and
k≥1
b
k
converge
to real numbers, say, A and B. But then
k≥1
c
k
converges to A+iB.
Another common mistake is to try to use the converse of this result, but the converse is false:
Example 7.6. The alternating harmonic series
k≥1
(−1)
k+1
k
converges, but not absolutely: This
series does not converge absolutely, according to the previous example. To see that it does converge,
rewrite it as follows:
k≥1
(−1)
k+1
k
= 1 −
1
2
+
1
3
−
1
4
+
1
5
−
1
6
+. . .
=
_
1 −
1
2
_
+
_
1
3
−
1
4
_
+
_
1
5
−
1
6
_
+. . .
(Technically, there is a small detail to be checked here, since we are eﬀectively ignoring half the
partial sums of the original series. See Exercise 10.) The reader can verify the inequality 2k(2k−1) ≥
k(k + 1) for k > 1, so the general term satisﬁes
1
2k −1
−
1
2k
=
1
2k(2k −1)
≤
1
k(k + 1)
,
CHAPTER 7. POWER SERIES 66
so the series converges by comparison with the telescoping series of Example 7.4.
For the rest of this book we shall be concerned almost exclusively with series which converge
absolutely. Hence checking convergence of a series is usually a matter of verifying that a series
of nonnegative reals is ﬁnite. We have already used the technique of comparing a series to a
series which is known to converge; this is often called a “comparison test.” Some variants of the
comparison test will appear when we look at power series. One handy test is the following:
Lemma 7.7 (Integral Test). Suppose f is a nonincreasing, positive function deﬁned on [1, ∞).
Then
_
∞
1
f(t) dt ≤
∞
k=1
f(k) ≤ f(1) +
_
∞
1
f(t) dt
This is immediate from a picture: the integral of f(t) on the interval [k, k + 1] is bounded
between f(k) and f(k +1). Adding the pieces gives the inequalities above for the N
th
partial sum
versus the integrals from 1 to N and from 1 to N + 1, and the inequality persists in the limit.
Example 7.7.
k≥1
1
k
p
converges if p > 1 and diverges if p ≤ 1.
7.3 Sequences and Series of Functions
The fun starts when one studies sequences (f
n
) of functions f
n
. We say that such a sequence
converges at z
0
if the sequence (of complex numbers) (f
n
(z
0
)) converges. If a sequence of functions,
(f
n
), converges at all z in some subset G ⊆ C then we say that (f
n
) converges pointwise on G. So
far nothing new; but this notion of convergence does not really catch the spirit of the function as
a whole.
Deﬁnition 7.2. Suppose (f
n
) and f are functions deﬁned on G ⊆ C. If for all > 0 there is an N
such that for all z ∈ G and for all n ≥ N we have
f
n
(z) −f(z) <
then (f
n
) converges uniformly in G to f.
What’s the big deal about uniform versus pointwise convergence? It is easiest to describe
the diﬀerence with the use of quantiﬁers, namely ∀ denoting “for all” and ∃ denoting “there is.”
Pointwise convergence on G means
(∀ > 0) (∀ z ∈ G) (∃ N : n ≥ N ⇒ f
n
(z) −f(z) < ) ,
whereas uniform convergence on G translates into
(∀ > 0) (∃ N : (∀ z ∈ G) n ≥ N ⇒ f
n
(z) −f(z) < ) .
No big deal—we only exchanged two of the quantiﬁers. In the ﬁrst case, N may well depend on
z, in the second case we need to ﬁnd an N which works for all z ∈ G. And this can make all the
diﬀerence . . .
The ﬁrst example illustrating this diﬀerence says in essence that if we have a sequence of
functions (f
n
) which converges uniformly on G then for all z
0
∈ G
lim
n→∞
lim
z→z
0
f
n
(z) = lim
z→z
0
lim
n→∞
f
n
(z) .
We will need similar interchanges of limits constantly.
CHAPTER 7. POWER SERIES 67
Proposition 7.8. Suppose (f
n
) is a sequence of continuous functions on the region G converging
uniformly to f on G. Then f is continuous on G.
Proof. Let z
0
∈ G; we will prove that f is continuous at z
0
. By uniform convergence, given > 0,
there is an N such that for all z ∈ G and all n ≥ N
f
n
(z) −f(z) <
3
.
Now we make use of the continuity of the f
n
’s. This means that given (the same) > 0, there is a
δ > 0 such that whenever z −z
0
 < δ we have
f
n
(z) −f
n
(z
0
) <
3
.
All that’s left is putting those two inequalities together: by the triangle inequality
f(z) −f(z
0
) = f(z) −f
n
(z) +f
n
(z) −f
n
(z
0
) +f
n
(z
0
) −f(z
0
)
≤ f(z) −f
n
(z) +f
n
(z) −f
n
(z
0
) +f
n
(z
0
) −f(z
0
)
< ,
that is, f is continuous at z
0
.
Once we know the above result about continuity, we can ask about integration of series of
functions. The next theorem should come as no surprise, however, its consequences (which we will
only see in the next chapter) are wide ranging.
Proposition 7.9. Suppose f
n
are continuous on the smooth curve γ and converge uniformly on γ
to f. Then
lim
n→∞
_
γ
f
n
=
_
γ
f .
Proof. By Proposition 4.1(d), we can estimate
¸
¸
¸
¸
_
γ
f
n
−
_
γ
f
¸
¸
¸
¸
=
¸
¸
¸
¸
_
γ
f
n
−f
¸
¸
¸
¸
≤ max
z∈γ
f
n
(z) −f(z) length(γ) .
But f
n
→f uniformly on γ, and we can make max
z∈γ
f
n
(z) −f(z) as small as we like.
Since uniform convergence is often of critical importance, we give two practical tests: one
arguing for uniformity and the other against. They are formulated for sequences that converge to
0. If a sequence g
n
converges to a function g then we can usually apply these tests to f
n
= g −g
n
,
which does converge to 0.
Lemma 7.10. If f
n
is a sequence of functions and M
n
is a sequence of constants so that M
n
converges to 0 and f
n
(z) ≤ M
n
for all z ∈ G, then f
n
converges uniformly to 0 on G.
For example, z
n
 ≤ r
n
if z is in the closed disk
¯
D
r
(0), and r
n
→0 if r < 1, so z
n
→0 uniformly
in
¯
D
r
(0) if r < 1.
Lemma 7.11. If f
n
is a sequence of functions which converges uniformly to 0 on a set G and z
n
is any sequence in G then the sequence f
n
(z
n
) converges to 0.
CHAPTER 7. POWER SERIES 68
This is most often used to prove nonuniform convergence. For example, let f
n
(z) = z
n
and let
G be the open unit disk D
1
(0). Then z < 1 if z is in G, so z
n
→0, and so z
n
→0. However, let
z
n
= exp(−
1
n
). Then z
n
is in G but f
n
(z
n
) = e
−1
so f
n
(z
n
) does not converge to 0. Therefore z
n
does not converge uniformly to 0 on D
1
(0).
All of these notions for sequences of functions go verbatim for series of functions. Here we also
have a notion of absolute convergence (which can be combined with uniform convergence). There
is an important result about series of functions, often called the Weierstraß Mtest.
Proposition 7.12. Suppose (f
k
) are continuous on the region G, f
k
(z) ≤ M
k
for all z ∈ G, and
k≥1
M
k
converges. Then
k≥1
f
k
converges absolutely and uniformly in G.
Proof. For each ﬁxed z we have
k≥1
f
k
(z) ≤
k≥1
M
k
< ∞, so
k≥1
f
k
(z) converges; call the
limit f(z). This deﬁnes a function f on G. To see that f
n
converges uniformly to f, suppose > 0.
Since
k≥1
M
k
converges there is N so that
k>n
M
k
=
∞
k=1
M
k
−
n
k=1
M
k
<
for all n > N. Then, for any z in G, if n ≥ N then
¸
¸
¸
¸
¸
f(z) −
n
k=1
f
k
(z)
¸
¸
¸
¸
¸
=
¸
¸
¸
¸
¸
k>n
f
n
(z)
¸
¸
¸
¸
¸
≤
k>n
f
n
(z) ≤
k>n
M
k
<
and this satisﬁes the deﬁnition of uniform convergence.
7.4 Region of Convergence
For the remainder of this chapter (indeed, these lecture notes) we concentrate on some very special
series of functions.
Deﬁnition 7.3. A power series centered at z
0
is a series of functions of the form
k≥0
c
k
(z −z
0
)
k
.
The fundamental example of a power series is the geometric series, for which all c
k
= 1.
Lemma 7.13. The geometric series
k≥0
z
k
converges absolutely for z < 1 to the function
1/(1 −z). The convergence is uniform on any set of the form { z ∈ C : z ≤ r } for any r < 1.
Proof. Fix an r < 1, and let D = { z ∈ C : z ≤ r }. We will use Proposition 7.12 with f
k
(z) = z
k
and M
k
= r
k
. Hence the uniform convergence on D of the geometric series will follow if we can
show that
k≥0
r
k
converges. But this is straightforward: the partial sums of this series can be
written as
n
k=0
r
k
= 1 +r +· · · +r
n−1
+r
n
=
1 −r
n+1
1 −r
,
CHAPTER 7. POWER SERIES 69
whose limit as n → ∞ exists because r < 1. Hence, by Proposition 7.12, the geometric series
converges absolutely and uniformly on any set of the form {z ∈ C : z ≤ r} with r < 1. Since r
can be chosen arbitrarily close to 1, we have absolute convergence for z < 1. It remains to show
that for those z the limit function is 1/(1 −z), which follows by
k≥0
z
k
= lim
n→∞
n
k=0
z
k
= lim
n→∞
1 −z
n+1
1 −z
=
1
1 −z
.
By comparing a general power series to a geometric series we can give a complete description
of its region of convergence.
Theorem 7.14. Any power series
k≥0
c
k
(z − z
0
)
k
has a radius of convergence R. By this we
mean that R is a nonnegative real number, or ∞, satisfying the following.
(a) If r < R then
k≥0
c
k
(z − z
0
)
k
converges absolutely and uniformly on the closed disk
¯
D
r
(z
0
)
of radius r centered at z
0
.
(b) If z −z
0
 > R then the sequence of terms c
k
(z −z
0
)
k
is unbounded, so
k≥0
c
k
(z −z
0
)
k
does
not converge.
The open disk D
R
(z
0
) in which the power series converges absolutely is the region of convergence.
(If R = ∞ then D
R
(z
0
) is the entire complex plane, and if R = 0 then D
R
(z
0
) is the empty set.)
By way of Proposition 7.8, this theorem immediately implies the following.
Corollary 7.15. Suppose the power series
k≥0
c
k
(z −z
0
)
k
has radius of convergence R. Then
the series represents a function which is continuous on D
R
(z
0
).
While we’re at it, we might as well state what Proposition 7.9 implies for power series.
Corollary 7.16. Suppose the power series
k≥0
c
k
(z −z
0
)
k
has radius of convergence R and γ is
a smooth curve in D
R
(z
0
). Then
_
γ
k≥0
c
k
(z −z
0
)
k
dz =
k≥0
c
k
_
γ
(z −z
0
)
k
dz .
In particular, if γ is closed then
_
γ
k≥0
c
k
(z −z
0
)
k
dz = 0.
Proof of Theorem 7.14. Deﬁne C to be the set of positive real numbers for which the series
k≥0
c
k
t
k
converges, and deﬁne D to be the set of positive real numbers for which it diverges. Clearly every
positive real number is in either C or D, and these sets are disjoint. First we establish three facts
about these sets.
(∗) If r < t and t ∈ C then r ∈ C and
k≥0
c
k
(z − z
0
)
k
converges absolutely and uniformly
on
¯
D
r
(z
0
). To prove this, note that
k≥0
c
k
t
k
converges so c
k
t
k
→ 0 as k → ∞. In particular,
this sequence is bounded, so c
k
 t
k
≤ M for some constant M. Now if z ∈
¯
D
r
(z
0
) we have
¸
¸
c
k
(z −z
0
)
k
¸
¸
≤ c
k
 r
k
and
k≥0
c
k
 r
k
=
k≥0
c
k
 t
k
_
r
t
_
k
≤
k≥0
M
_
r
t
_
k
= M
k≥0
_
r
t
_
k
=
M
1 −r/t
< ∞.
CHAPTER 7. POWER SERIES 70
At the last step we recognized the geometric series, which converges since 0 ≤ r < t, and so
0 ≤ r/t < 1. This shows that r ∈ C, and uniform and absolute convergence on
¯
D
r
(z
0
) follows from
the Weierstraß Mtest.
(∗∗) If z −z
0
 = r > t and t ∈ D then r ∈ D and the sequence c
k
r
k
is unbounded, and hence
k≥0
c
k
(z − z
0
)
k
diverges. To prove this, assume that c
k
r
k
is bounded, so c
k
 r
k
≤ M for some
constant M. But now exactly the same argument as in (∗), but interchanging r and t, shows that
k≥0
c
k
t
k
converges, contradicting the assumption that t is in D.
(∗ ∗ ∗) There is an extended real number R, satisfying 0 ≤ R ≤ ∞, so that 0 < r < R implies
r ∈ C and R < r < ∞ implies r ∈ D. Notice that R = 0 works if C is empty, and R = ∞ works if
D is empty; so we assume neither is empty and we start with a
0
in C and b
0
in D. It is immediate
from (∗) or (∗∗) that a
0
< b
0
. We shall deﬁne sequences a
n
in C and b
n
in D which “zero in” on
R. First, let m
0
be the midpoint of the segment [a
0
, b
0
], so m
0
= (a
0
+b
0
)/2. If m
0
lies in C then
we deﬁne a
1
= m
0
and b
1
= b
0
; but if m
0
lies in D then we deﬁne a
1
= a
0
and b
1
= m
0
. Note that,
in either case, we have a
0
≤ a
1
< b
1
≤ b
0
, a
1
is in C, and b
1
is in D. Moreover, a
1
and b
1
are closer
together than a
0
and b
0
; in fact, b
1
−a
1
= (b
0
−a
0
)/2. We repeat this procedure to deﬁne a
2
and
b
2
within the interval [a
1
, b
1
], and so on. Summarizing, we have
a
n
≤ a
n+1
a
n
∈ C
b
n
≥ b
n+1
b
n
∈ D
a
n
< b
n
b
n
−a
n
= (b
0
−a
0
)/2
n
The sequences a
n
and b
n
are monotone and bounded (by a
0
and b
0
) so they have limits, and these
limits are the same since lim
n→∞
(b
n
−a
n
) = lim
n→∞
(b
0
−a
0
)/2
n
= 0. We deﬁne R to be this limit.
If 0 < r < R then r < a
n
for all suﬃciently large n, since a
n
converges to R, so r is in C by (∗).
On the other hand, if R < r then b
n
< r for all suﬃciently large n, so r is in D by (∗∗). Thus R
veriﬁes the statement (∗ ∗ ∗).
To prove Theorem 7.14, ﬁrst assume r < R and choose t so that r < t < R. Then t ∈ C by
(∗ ∗ ∗), so part (a) of 7.14 follows from (∗). Similarly, if r = z −z
0
 > R then choose t so that
R < t < r. Then t ∈ D by (∗ ∗ ∗), so part (b) of 7.14 follows from (∗∗).
It is worth mentioning the following corollary, which reduces the calculation of the radius of
convergence to examining the limiting behavior of the terms of the series.
Corollary 7.17. c
k
 r
k
→0 for 0 ≤ r < R but c
k
 r
k
is unbounded for r > R.
Warning: Neither Theorem 7.14 nor Corollary 7.17 says anything about convergence on the
circle z −z
0
 = R .
Exercises
1. For each of the sequences, prove convergence/divergence. If the sequence converges, ﬁnd the
limit.
(a) a
n
= e
iπn/4
.
(b)
(−1)
n
n
.
CHAPTER 7. POWER SERIES 71
(c) cos n.
(d) 2 −
in
2
2n
2
+1
.
(e) sin
_
1
n
_
.
2. Show that the limit of a convergent sequence is unique.
3. Derive the Archimedean Property from the monotone sequence property.
4. Prove:
(a) lim
n→∞
a
n
= a =⇒ lim
n→∞
a
n
 = a.
(b) lim
n→∞
a
n
= 0 ⇐⇒ lim
n→∞
a
n
 = 0.
5. Prove Lemma 7.3.
6. Prove: (c
n
) converges if and only if (Re c
n
) and (Imc
n
) converge.
7. Prove Lemma 7.1.
8. Suppose a
n
≤ b
n
≤ c
n
for all n and lim
n→∞
a
n
= L = lim
n→∞
c
n
. Prove that lim
n→∞
b
n
= L.
State and prove a similar theorem for series.
9. Find sup
_
Re
_
e
2πit
_
: t ∈ Q\ Z
_
.
10. Suppose that the terms c
n
converge to zero, and show that
∞
n=0
c
n
converges if and only if
∞
k=0
(c
2k
+ c
2k+1
) converges. Moreover, if the two series converge then they have the same
limit. Also, give an example where c
n
does not converge to 0 and one series diverges while
the other converges.
11. Prove that the series
k≥1
b
k
converges if and only if lim
n→∞
∞
k=n
b
k
= 0 .
12. (a) Show that
k≥1
1
2
k
= 1. One way to do this is to write
1
2
k
as a diﬀerence of powers of
2 so that you get a telescoping series.
(b) Show that
k≥1
k
k
2
+1
diverges. (Hint: compare the general term to
1
2k
.)
(c) Show that
k≥1
k
k
3
+1
converges. (Hint: compare the general term to
1
k
2
.)
13. Discuss the convergence of
k≥0
z
k
for z = 1.
14. Prove Lemma 7.10.
15. Prove Lemma 7.11.
16. Discuss pointwise and uniform convergence for the following sequences
(a) (nz
n
) .
(b)
_
z
n
n
_
for n > 0.
(c)
_
1
1+nz
_
, deﬁned on {z ∈ C : Re z ≥ 0}.
CHAPTER 7. POWER SERIES 72
17. Let f
n
(x) = n
2
xe
−nx
.
(a) Show that lim
n→∞
f
n
(x) = 0 for all x ≥ 0. Treat x = 0 as a special case; for x > 0 you
can use L’Hospital’s rule—but remember that n is the variable, not x.
(b) Find lim
n→∞
_
1
0
f
n
(x) dx. (Hint: the answer is not 0.)
(c) Why doesn’t your answer to part (b) violate Proposition 7.9?
18. Find a power series (and determine its radius of convergence) of the following functions.
(a)
1
1+4z
.
(b)
1
3−
z
2
.
19. (a) Suppose that the sequence c
k
is bounded and show that the radius of convergence of
k≥0
c
k
(z −z
0
)
k
is at least 1.
(b) Suppose that the sequence c
k
does not converge to 0 and show that the radius of con
vergence of
k≥0
c
k
(z −z
0
)
k
is at most 1.
20. Find the power series centered at 1 for the following functions, and compute their radius of
convergence:
(a)
1
z
.
(b) Log z.
21. Use the Weierstraß Mtest to show that each of the following series converges uniformly on
the given domain:
(a)
k≥1
z
k
k
2
on
¯
D
1
(0).
(b)
k≥0
1
z
k
on {z : z ≥ 2}.
(c)
k≥0
z
k
z
k
+ 1
on
¯
D
r
(0), where 0 ≤ r < 1.
22. Suppose L = lim
k→∞
c
k

1/k
exists. Show that
1
L
is the radius of convergence of
k≥0
c
k
(z −z
0
)
k
.
(Use the natural interpretations if L = 0 or L = ∞.)
23. Find the radius of convergence for each of the following series.
(a)
k≥0
a
k
2
z
k
, a ∈ C.
(b)
k≥0
k
n
z
k
, n ∈ Z.
(c)
k≥0
z
k!
.
CHAPTER 7. POWER SERIES 73
(d)
k≥1
(−1)
k
k
z
k(k+1)
.
(e)
k≥1
z
k
k
k
.
(f)
k≥0
cos(k)z
k
.
(g)
k≥0
4
k
(z −2)
k
.
24. Deﬁne the functions f
n
(t) =
1
n
e
−t/n
for n > 0 and 0 ≤ t < ∞.
(a) Show that the maximum of f
n
(t) is
1
n
.
(b) Show that f
n
(t) converges uniformly to 0 as n →∞.
(c) Show that
_
∞
0
f
n
(t) dt does not converge to 0 as n →∞
(d) Why doesn’t this contradict the theorem that “the integral of a uniform limit is the limit
of the integrals”?
25. Let f be analytic on the disk z < 4 and suppose f(z) ≤ 5 for all z on the circle z = 3.
Show that
¸
¸
f
(3)
(0)
¸
¸
≤
10
9
. (Hint: Use the Cauchy integral formula.) What can you say about
¸
¸
f
(3)
(1)
¸
¸
?
Chapter 8
Taylor and Laurent Series
We think in generalities, but we live in details.
A. N. Whitehead
8.1 Power Series and Analytic Functions
All of the last chapter could have been developed in greater generality, say for functions from R
n
to
R
m
. We will now, however, connect the last chapter to the theory of functions analytic on certain
regions. The cornerstone is provided by two theorems which say that any power series represents
an analytic function, and conversely, any analytic function can be represented as a power series.
The ﬁrst of them goes as follows.
Theorem 8.1. Suppose f(z) =
k≥0
c
k
(z −z
0
)
k
has radius of convergence R. Then f is analytic
in {z ∈ C : z −z
0
 < R}.
Proof. Given any closed curve γ ⊂ {z ∈ C : z −z
0
 < R}, we have by Corollary 7.16
_
γ
k≥0
c
k
(z −z
0
)
k
dz = 0 .
On the other hand, Corollary 7.15 says that f is continuous. Now apply Morera’s theorem (Corol
lary 5.7).
A special case of the last result concerns power series with inﬁnite radius of convergence: those
represent entire functions.
Now that we know that power series are analytic (i.e., diﬀerentiable) on their regions of conver
gence we can ask how to ﬁnd their derivatives. The next result says that we can simply diﬀerentiate
the series “term by term.”
Theorem 8.2. Suppose f(z) =
k≥0
c
k
(z −z
0
)
k
has radius of convergence R. Then
f
(z) =
k≥1
k c
k
(z −z
0
)
k−1
,
and the radius of convergence of this power series is also R.
74
CHAPTER 8. TAYLOR AND LAURENT SERIES 75
Proof. Let f(z) =
k≥0
c
k
(z −z
0
)
k
. Since we know that f is analytic in its region of convergence
we can use Theorem 5.1. Let γ be any simple closed curve in {z ∈ C : z −z
0
 < R}. Note that the
power series of f converges uniformly on γ, so that we are free to interchange integral and inﬁnite
sum. And then we use Theorem 5.1 again, but applied to the function (z − z
0
)
k
. Here are the
details:
f
(z) =
1
2πi
_
γ
f(w)
(w −z)
2
dw
=
1
2πi
_
γ
k≥0
c
k
(w −z
0
)
k
(w −z)
2
dw
=
k≥0
c
k
·
1
2πi
_
γ
(w −z
0
)
k
(w −z)
2
dw
=
k≥0
c
k
·
d
dw
(w −z
0
)
k
¸
¸
¸
¸
w=z
=
k≥0
k c
k
(z −z
0
)
k−1
.
The last statement of the theorem is easy to show: the radius of convergence R of f
(z) is at least
R (since we have shown that the series converges whenever z −z
0
 < R), and it cannot be larger
than R by comparison to the series for f(z), since the coeﬃcients for (z −z
0
)f
(z) are bigger than
the corresponding ones for f(z).
Naturally, the last theorem can be repeatedly applied to f
, then to f
, and so on. The various
derivatives of a power series can also be seen as ingredients of the series itself. This is the statement
of the following Taylor
1
series expansion.
Corollary 8.3. Suppose f(z) =
k≥0
c
k
(z −z
0
)
k
has a positive radius of convergence. Then
c
k
=
f
(k)
(z
0
)
k!
.
Proof. For starters, f(z
0
) = c
0
. Theorem 8.2 gives f
(z
0
) = c
1
. Applying the same theorem to f
gives
f
(z) =
k≥2
k(k −1)c
k
(z −z
0
)
k−2
and f
(z
0
) = 2c
2
. We can play the same game for f
(z
0
), f
(z
0
), etc.
Taylor’s formulas show that the coeﬃcients of any power series which converges to f on an
open disk D centered at z
0
can be determined from the the function f restricted to D. It follows
immediately that the coeﬃcients of a power series are unique:
Corollary 8.4 (Uniqueness of power series). If
k≥0
c
k
(z − z
0
)
k
and
k≥0
c
k
(z − z
0
)
k
are two
power series which both converge to the same function f(z) on an open disk centered at z
0
then
c
k
= c
k
for all k.
1
For more information about Brook Taylor (1685–1731), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Taylor.html.
CHAPTER 8. TAYLOR AND LAURENT SERIES 76
Theorem 8.5. Suppose f is a function which is analytic in D = {z ∈ C : z −z
0
 < R}. Then f
can be represented in D as a power series centered at z
0
(with a radius of convergence at least R):
f(z) =
k≥0
c
k
(z −z
0
)
k
with c
k
=
1
2πi
_
γ
f(w)
(w −z
0
)
k+1
dw.
Here γ is any positively oriented, simple, closed, smooth curve in D for which z
0
is inside γ.
Proof. Let g(z) = f(z + z
0
); so g is a function analytic in {z ∈ C : z < R}. Fix r < R, denote
the circle centered at the origin with radius r by γ
r
, and suppose that z < r. Then by Cauchy’s
integral formula (Theorem 4.8),
g(z) =
1
2πi
_
γr
g(w)
w −z
dw.
The factor 1/(w−z) in this integral can be extended into a geometric series (note that w ∈ γ
r
and
so
¸
¸
z
w
¸
¸
< 1)
1
w −z
=
1
w
1
1 −
z
w
=
1
w
k≥0
_
z
w
_
k
which converges uniformly in the variable w ∈ γ
r
(by Lemma 7.13). Hence Proposition 7.9 applies:
g(z) =
1
2πi
_
γr
g(w)
w −z
dw =
1
2πi
_
γr
g(w)
1
w
k≥0
_
z
w
_
k
dw =
k≥0
1
2πi
_
γr
g(w)
w
k+1
dwz
k
.
Now, since f(z) = g(z −z
0
), we apply an easy change of variables to obtain
f(z) =
k≥0
1
2πi
_
Γr
f(w)
(w −z
0
)
k+1
dw(z −z
0
)
k
,
where Γ
r
is a circle centered at z
0
with radius r. The only diﬀerence of this righthand side to the
statement of the theorem are the curves we’re integrating over. However, Γ
r
∼
G\{z
0
}
γ, and we can
apply Cauchy’s Theorem 4.4:
_
Γr
f(w)
(w −z
0
)
k+1
dw =
_
γ
f(w)
(w −z
0
)
k+1
dw.
If we compare the coeﬃcients of the power series obtained in Theorem 8.5 with those in Corol
lary 8.3, we arrive at the longpromised extension of Theorem 5.1 (which in itself extended Cauchy’s
integral formula, Theorem 4.8).
Corollary 8.6. Suppose f is analytic on the region G, w ∈ G, and γ is a positively oriented,
simple, closed, smooth, Gcontractible curve such that w is inside γ. Then
f
(k)
(w) =
k!
2πi
_
γ
f(z)
(z −w)
k+1
dz .
Corollary 8.6 combined with our oftenused Proposition 4.1(d) gives an inequality which is often
called Cauchy’s Estimate:
CHAPTER 8. TAYLOR AND LAURENT SERIES 77
Corollary 8.7. Suppose f is analytic in {z ∈ C : z −w < R} and f ≤ M. Then
¸
¸
¸f
(k)
(w)
¸
¸
¸ ≤
k!M
R
k
.
Proof. Let γ be a circle centered at w with radius r < R. Then Corollary 8.6 applies, and we can
estimate using Proposition 4.1(d):
¸
¸
¸f
(k)
(w)
¸
¸
¸ =
¸
¸
¸
¸
k!
2πi
_
γ
f(z)
(z −w)
k+1
dz
¸
¸
¸
¸
≤
k!
2π
max
z∈γ
¸
¸
¸
¸
f(z)
(z −w)
k+1
¸
¸
¸
¸
length(γ) ≤
k!
2π
M
r
k+1
2πr =
k!M
r
k
.
The statement now follows since r can be chosen arbitrarily close to R.
8.2 Classiﬁcation of Zeros and the Identity Principle
Basic algebra shows that if a polynomial p(z) of positive degree d has a a zero at a (in other words,
if p(a) = 0) then p(z) has z −a as a factor. That is, p(z) = (z −a)q(z) where q(z) is a polynomial
of degree d − 1. We can then ask whether q(z) itself has a zero at a and, if so, we can factor out
another factor of z −a. continuing in this way we see that we can factor p(z) as p(z) = (z −a)
m
g(z)
where m is a positive integer, not bigger than d, and g(z) is a polynomial which does not have a
zero at a. The integer m is called the multiplicity of the zero a of p(z).
Almost exactly the same thing happens for analytic functions:
Theorem 8.8 (Classiﬁcation of Zeros). Suppose f is an analytic function deﬁned on an open set
G and suppose f has a zero at a point a in G. Then there are exactly two possibilities:
(a) Either: f is identically zero on some open disk D centered at a (that is, f(z) = 0 for all z in
D);
(b) or: there is a positive integer m and an analytic function g, deﬁned on G, satisfying f(z) =
(z −a)
m
g(z) for all z in G, with g(a) = 0
The integer m in the second case is uniquely determined by f and a and is called the multiplicity
of the zero at a.
Proof. We have a power series expansion for f(z) in some disk D
r
(a) of radius r around a, so
f(z) =
k≥0
c
k
(z − a)
k
, and c
0
= f(0) is zero since a is a zero of f. There are now exactly two
possibilities:
(a) Either c
k
= 0 for all k;
(b) or there is some positive integer m so that c
k
= 0 for all k < m but c
m
= 0.
The ﬁrst case clearly gives us f(z) = 0 for all z in D = D
r
(a). So now consider the second case.
Notice that
f(z) = c
m
(z −a)
m
+c
m+1
(z −a)
m+1
+· · · = (z −a)
m
(c
m
+c
m+1
(z −a) +· · · )
= (z −a)
m
k≥0
c
k+m
(z −a)
k
.
CHAPTER 8. TAYLOR AND LAURENT SERIES 78
Then we can deﬁne a function g on G by
g(z) =
_
¸
_
¸
_
k≥0
c
k+m
(z −a)
k
if z −a < r
f(z)
(z −a)
m
if z ∈ G\ {a}
According to our calculations above, the two deﬁnitions give the same value when both are appli
cable. The function g is analytic at a by the ﬁrst deﬁnition; and g is analytic at other points of G
by the second deﬁnition. Finally, g(a) = c
m
= 0.
Clearly m is unique, since it is deﬁned in terms of the power series expansion of f at a, which
is unique.
The proof of this last theorem immediately yields the following.
Corollary 8.9. Suppose f is an analytic function deﬁned on an open set G and suppose f has a
zero at a ∈ G. Then the multiplicity of a equals m if and only if we can write f as a power series
of the form f(z) =
k≥m
c
k
(z −a)
k
with c
m
= 0.
To start using the intimate connection of analytic functions and power series, we apply Theo
rem 8.8 to obtain the following result, which is sometimes also called the uniqueness theorem.
Theorem 8.10 (Identity Principle). Suppose f and g are analytic in the region G and f(z
k
) = g(z
k
)
at a sequence that converges to a ∈ G with z
k
= a for all k. Then f(z) = g(z) for all z in G.
Proof. We start by deﬁning h = f −g. Then h is analytic on G, h(z
n
) = 0, and we will be ﬁnished
if we can deduce that h is identically zero on G. Now notice the following: If b is in G then exactly
one of the following occurs:
(a) Either there is an open disk D centered at b so that h(z) = 0 for all z in D;
(b) or there is an open disk D centered at b so that h(z) = 0 for all z in D \ {b}.
To see this, suppose that h(b) = 0. Then, by continuity, there is an open disk D centered at b so
that h(z) = 0 for all z ∈ D, so b satisﬁes the second condition. If h(b) = 0 then, by the classiﬁcation
of zeros, either h(z) = 0 for all z in some open disk D centered at b, so b satisﬁes the ﬁrst condition;
or h(z) = (z −b)
m
φ(z) for all z in G, where φ is analytic and φ(b) = 0. Then, since φ is continuous,
there is an open disk D centered at b so that φ(z) = 0 for all z in D. Then h(z) = (z −b)
m
φ(z) = 0
for all z in D except z = b, so b satisﬁes the second condition.
Now deﬁne two sets X, Y ⊆ G, so that b ∈ X if b satisﬁes the ﬁrst condition above, and b ∈ Y
if b satisﬁes the second condition. If b ∈ X and D is an open disk centered at b as in the ﬁrst
condition then it is clear that D ⊆ X. If b ∈ Y and D is an open disk centered at b as in the second
condition then D ⊆ Y , since if z ∈ D \ {b} then h(z) = 0, and we saw that this means z satisﬁes
the second condition.
Finally, we check that our original point a lies in X. To see this, suppose a ∈ Y , and let D be
an open disk centered at a so that h(z) = 0 for all z in D except z = b. But, since the sequence z
k
converges to a, there is some k so that z
k
is in D, so h(z
k
) = 0. Since z
k
= a, this is a contradiction.
Now we ﬁnish the proof using the deﬁnition of connectedness. X and Y are disjoint open sets
whose union is G, so one of them must be empty. Since a is in X, we must have Y = ∅ and X = G.
But X = G implies that every z in G satisﬁes the ﬁrst condition above, so h(z) = 0.
CHAPTER 8. TAYLOR AND LAURENT SERIES 79
Using the identity principle, we can prove yet another important property of analytic functions.
Theorem 8.11 (MaximumModulus Theorem). Suppose f is analytic and not constant in the
region G. Then f does not attain a weak relative maximum in G.
There are many reformulations of this theorem, such as: If G is a bounded region and f is
analytic in the closure of G, then the maximum of f is attained on the boundary of G.
Proof. Suppose there is a point a in G and an open disk D
0
centered at a so that f(z) ≤ f(a)
for all z in D
0
. If f(a) = 0 then f(z) = 0 for all z in D
0
, so f is identically zero, by the identity
principle. So we assume f(a) = 0. In this case we can deﬁne an analytic function g(z) = f(z)/f(a),
and we have the condition g(z) ≤ g(a) = 1 for all z in D
0
. Since g(a) = 1 we can ﬁnd, using
continuity, a smaller open disk D centered at a so that g(z) has positive real part for all z in D. Thus
the function h = Log ◦g is deﬁned and analytic on D, and we have h(a) = Log(g(a)) = Log(1) = 0
and Re h(z) = Re Log(g(z)) = ln(g(z)) ≤ ln(1) = 0.
We now refer to Exercise 27, which shows that h must be identically zero in D. Hence g(z) =
e
h(z)
must be equal to e
0
= 1 for all z in D, and so f(z) = f(a)g(z) must have the constant value
f(a) for all z in D. Hence, by the identity principle, f(z) has the constant value f(a) for all z
in G.
Theorem 8.11 can be used to give a proof of the analogous theorem for harmonic functions,
Theorem 6.5, in the process strengthening that theorem to cover weak maxima and weak minima.
Corollary 8.12. If u is harmonic in the region G, then it does not have a weak relative maximum
or minimum in G.
Since the last corollary also covers minima of harmonic functions, we should not be too surprised
to ﬁnd the following result whose proof we leave for the exercises.
Corollary 8.13 (MinimumModulus Theorem). Suppose f is analytic and not constant in the
region G. Then f does not attain a weak relative minimum at a in G unless f(a) = 0.
8.3 Laurent Series
Theorem 8.5 gives a powerful way of describing analytic functions. It is, however, not as general
as it could be. It is natural, for example, to think about representing exp
_
1
z
_
as
exp
_
1
z
_
=
k≥0
1
k!
_
1
z
_
k
=
k≥0
1
k!
z
−k
,
a “power series” with negative exponents. To make sense of expressions like the above, we introduce
the concept of a double series
k∈Z
a
k
=
k≥0
a
k
+
k≥1
a
−k
.
Here a
k
∈ C are terms indexed by the integers. A double series converges if both its deﬁning series
do. Absolute and uniform convergence are deﬁned analogously. Equipped with this, we can now
state the following central deﬁnition.
CHAPTER 8. TAYLOR AND LAURENT SERIES 80
Deﬁnition 8.1. A Laurent
2
series centered at z
0
is a double series of the form
k∈Z
c
k
(z −z
0
)
k
.
Example 8.1. The series which started this section is the Laurent series of exp
_
1
z
_
centered at 0.
Example 8.2. Any power series is a Laurent series (with c
k
= 0 for k < 0).
We should pause for a minute and ask for which z such a Laurent series can possibly converge.
By deﬁnition
k∈Z
c
k
(z −z
0
)
k
=
k≥0
c
k
(z −z
0
)
k
+
k≥1
c
−k
(z −z
0
)
−k
.
The ﬁrst of the series on the righthand side is a power series with some radius of convergence
R
2
, that is, it converges in {z ∈ C : z −z
0
 < R
2
}. The second we can view as a “power series
in
1
z−z
0
,” it will converge for
1
z−z
0
<
1
R
1
for some R
1
, that is, in {z ∈ C : z −z
0
 > R
1
}. For the
convergence of our Laurent series, we need to combine those two notions, whence the Laurent series
converges on the annulus {z ∈ C : R
1
< z −z
0
 < R
2
} (if R
1
< R
2
). Even better, Theorem 7.14
implies that the convergence is uniform on a set of the form {z ∈ C : r
1
≤ z −z
0
 ≤ r
2
} for any
R
1
< r
1
< r
2
< R
2
. Theorem 8.1 says that the Laurent series represents a function which is
analytic on {z ∈ C : R
1
< z −z
0
 < R
2
}. The fact that we can conversely represent any function
analytic in such an annulus by a Laurent series is the substance of the next theorem.
Theorem 8.14. Suppose f is a function which is analytic in A = {z ∈ C : R
1
< z −z
0
 < R
2
}.
Then f can be represented in A as a Laurent series centered at z
0
:
f(z) =
k∈Z
c
k
(z −z
0
)
k
with c
k
=
1
2πi
_
γ
f(w)
(w −z
0
)
k+1
dw.
Here γ is any circle in A centered at z
0
.
Remark. Naturally, by Cauchy’s Theorem 4.4 we can replace the circle in the formula for the
Laurent series by any closed, smooth path that is Ahomotopic to the circle.
Proof. Let g(z) = f(z +z
0
); so g is a function analytic in {z ∈ C : R
1
< z < R
2
}. Fix R
1
< r
1
<
z < r
2
< R
2
, and let γ
1
and γ
2
be positively oriented circles centered at 0 with radii r
1
and
r
2
, respectively. By introducing an “extra piece” (see Figure 8.1), we can apply Cauchy’s integral
formula (Theorem 4.8) to the path γ
2
−γ
1
:
g(z) =
1
2πi
_
γ
2
−γ
1
g(w)
w −z
dw =
1
2πi
_
γ
2
g(w)
w −z
dw −
1
2πi
_
γ
1
g(w)
w −z
dw. (8.1)
For the integral over γ
2
we play exactly the same game as in Theorem 8.5. The factor 1/(w−z) in
this integral can be expanded into a geometric series (note that w ∈ γ
2
and so
¸
¸
z
w
¸
¸
< 1)
1
w −z
=
1
w
1
1 −
z
w
=
1
w
k≥0
_
z
w
_
k
,
2
For more information about Pierre Alphonse Laurent (1813–1854), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Laurent Pierre.html.
CHAPTER 8. TAYLOR AND LAURENT SERIES 81
γ
1
γ
2
Figure 8.1: Proof of Theorem 8.14.
which converges uniformly in the variable w ∈ γ
2
(by Lemma 7.13). Hence Proposition 7.9 applies:
_
γ
2
g(w)
w −z
dw =
_
γ
2
g(w)
1
w
k≥0
_
z
w
_
k
dw =
k≥0
_
γ
2
g(w)
w
k+1
dwz
k
.
The integral over γ
1
is computed in a similar fashion; now we expand the factor 1/(w−z) into the
following geometric series (note that w ∈ γ
1
and so
¸
¸
w
z
¸
¸
< 1)
1
w −z
= −
1
z
1
1 −
w
z
= −
1
z
k≥0
_
w
z
_
k
,
which converges uniformly in the variable w ∈ γ
1
(by Lemma 7.13). Again Proposition 7.9 applies:
_
γ
1
g(w)
w −z
dw = −
_
γ
1
g(w)
1
z
k≥0
_
w
z
_
k
dw = −
k≥0
_
γ
1
g(w)w
k
dwz
−k−1
= −
k≤−1
_
γ
1
g(w)
w
k+1
dwz
k
.
Putting everything back into (8.1) gives
g(z) =
1
2πi
k≥0
_
γ
2
g(w)
w
k+1
dwz
k
+
k≤−1
_
γ
1
g(w)
w
k+1
dwz
k
.
We can now change both integration paths to a circle γ centered at 0 with a radius between R
1
and R
2
(by Cauchy’s Theorem 4.4), which ﬁnally gives
g(z) =
1
2πi
k∈Z
_
γ
g(w)
w
k+1
dwz
k
.
The statement follows now with f(z) = g(z −z
0
) and an easy change of variables.
CHAPTER 8. TAYLOR AND LAURENT SERIES 82
We ﬁnish this chapter with a consequence of the above theorem: because the coeﬃcients of a
Laurent series are given by integrals, we immediately obtain the following:
Corollary 8.15. The coeﬃcients of a Laurent series are unique.
This result seems a bit artiﬁcial; what it says is simply the following: if we expand a function
(that is analytic in some annulus) into a Laurent series, there is only one possible outcome.
Exercises
1. For each of the following series, determine where the series converges absolutely/uniformly:
(a)
k≥2
k(k −1) z
k−2
.
(b)
k≥0
1
(2k + 1)!
z
2k+1
.
(c)
k≥0
_
1
z −3
_
k
.
2. What functions are represented by the series in the previous exercise?
3. Find the power series centered at 1 for exp z.
4. Prove Lemma 3.8 using the power series of exp z centered at 0.
5. By integrating a series for
1
1+z
2
term by term, ﬁnd a power series for arctan(z). What is its
radius of convergence?
6. Find the terms through third order and the radius of convergence of the power series for each
following functions, centered at z
0
. Do not ﬁnd the general form for the coeﬃcients.
(a) f(z) =
1
1+z
2
, z
0
= 1.
(b) f(z) =
1
e
z
+1
, z
0
= 0.
(c) f(z) =
√
1 +z, z
0
= 0 (use the principal branch).
(d) f(z) = e
z
2
, z
0
= i.
7. Prove the following generalization of Theorem 8.1: Suppose f
n
are analytic on the region
G and converge uniformly to f on G. Then f is analytic in G. (This result is called the
Weierstraß convergence theorem.)
8. Use the previous exercise and Corollary 8.6 to prove the following: Suppose f
n
are analytic
on the region G and converge uniformly to f on G. Then for any k ∈ N, the k
th
derivatives
f
(k)
n
converge (pointwise) to f
(k)
.
9. Prove the minimummodulus theorem (Corollary 8.13).
CHAPTER 8. TAYLOR AND LAURENT SERIES 83
10. Find the maximum and minimum of f(z) on the unit disc {z ∈ C : z ≤ 1}, where
f(z) = z
2
−2.
11. Give another proof of the fundamental theorem of algebra (Theorem 5.4), using the mini
mummodulus theorem (Corollary 8.13). (Hint: Use Lemma 5.3 to show that a polynomial
does not achieve its minimum modulus on a large circle; then use the minimummodulus
theorem to deduce that the polynomial has a zero.)
12. Find a Laurent series for
1
(z−1)(z+1)
centered at z = 1 and specify the region in which it
converges.
13. Find a Laurent series for
1
z(z−2)
2
centered at z = 2 and specify the region in which it converges.
14. Find a Laurent series for
z−2
z+1
centered at z = −1 and specify the region in which it converges.
15. Find the ﬁrst ﬁve terms in the Laurent series for
1
sin z
centered at z = 0.
16. Find the ﬁrst 4 nonzero terms in the power series expansion of tan z centered at the origin.
What is the radius of convergence?
17. (a) Find the power series representation for e
az
centered at 0, where a is any constant.
(b) Show that e
z
cos(z) =
1
2
_
e
(1+i)z
+e
(1−i)z
_
.
(c) Find the power series expansion for e
z
cos(z) centered at 0.
18. Show that
z−1
z−2
=
k≥0
1
(z−1)
k
for z −1 > 1.
19. Prove: If f is entire and Im(f) is constant on the unit disc {z ∈ C : z ≤ 1} then f is
constant.
20. (a) Find the Laurent series for
cos z
z
2
centered at z = 0.
(b) Prove that
f(z) =
_
cos z−1
z
2
if z = 0,
−
1
2
if z = 0
is entire.
21. Suppose that f(z) has a zero of multiplicity m at a. Explain why
1
f(z)
has a pole of order m
at a.
22. Suppose that f(z
0
) = 0 and f
(z
0
) = 0. Show that f has a zero of multiplicity 1 at z
0
.
23. Find the multiplicities of the zeros:
(a) f(z) = e
z
−1, z
0
= 2kπi, where k is any integer.
(b) f(z) = sin(z) −tan(z), z
0
= 0.
(c) f(z) = cos(z) −1 +
1
2
sin
2
(z), z
0
= 0.
24. Find the zeros of the following, and determine their multiplicities:
(a) (1 +z
2
)
4
.
CHAPTER 8. TAYLOR AND LAURENT SERIES 84
(b) sin
2
z.
(c) 1 +e
z
.
(d) z
3
cos z.
25. Find the poles of the following, and determine their orders:
(a) (z
2
+ 1)
−3
(z −1)
−4
.
(b) z cot(z).
(c) z
−5
sin(z).
(d)
1
1−e
z
.
(e)
z
1−e
z
.
26. Suppose that f(z) has exactly one zero, at a, inside the circle γ, and that it has multiplicity 1.
Show that a =
1
2πi
_
γ
zf
(z)
f(z)
dz.
27. Suppose f is analytic and not identically zero on an open disk D centered at a, and suppose
f(a) = 0. Follow the following outline to show that Re f(z) > 0 for some z in D.
(a) Why can you write f(z) = (z −a)
m
g(z) where m > 0, g is analytic, and g(a) = 0?
(b) Write g(a) in polar form as g(a) = c e
iα
and deﬁne G(z) = e
−iα
g(z). Why is Re G(a) > 0?
(c) Why is there a positive constant δ so that Re G(z) > 0 for all z in the open disk D
δ
(a)?
(d) Write z = a +re
iθ
for 0 < r < δ. Show that f(z) = r
m
e
imθ
e
iα
G(z).
(e) Find a value of θ so that f(z) has positive real part.
28. Suppose c
n
 ≥ 2
n
for all n. What can you say about the radius of convergence of
k≥0
c
k
z
k
?
29. Suppose the radius of convergence of
k≥0
c
k
z
k
is R. What is the radius of convergence of
each of the following?
(a)
k≥0
k
2
c
k
z
k
.
(b)
k≥0
c
k
z
2k
.
(c)
k≥0
c
k
z
k+5
.
(d)
k≥0
3
k
c
k
z
k
.
(e)
k≥0
c
2
k
z
k
.
Chapter 9
Isolated Singularities and the Residue
Theorem
1/r
2
has a nasty singularity at r = 0, but it did not bother Newton—the moon is far enough.
Edward Witten
9.1 Classiﬁcation of Singularities
What is the diﬀerence between the functions
sin z
z
,
1
z
4
, and exp
_
1
z
_
? All of them are not deﬁned at
0, but the singularities are of a very diﬀerent nature. For complex functions there are three types
of singularities, which are classiﬁed as follows.
Deﬁnition 9.1. If f is analytic in the punctured disk {z ∈ C : 0 < z −z
0
 < R} for some R > 0
but not at z = z
0
then z
0
is an isolated singularity of f. The singularity z
0
is called
(a) removable if there is a function g analytic in {z ∈ C : z −z
0
 < R} such that f = g in
{z ∈ C : 0 < z −z
0
 < R},
(b) a pole if lim
z→z
0
f(z) = ∞,
(c) essential if z
0
is neither removable nor a pole.
Example 9.1. The function
sin z
z
has a removable singularity at 0, as for z = 0
sin z
z
=
1
z
k≥0
(−1)
k
(2k + 1)!
z
2k+1
=
k≥0
(−1)
k
(2k + 1)!
z
2k
.
and the power series on the righthand side represents an entire function (you may meditate on the
fact why it has to be entire).
Example 9.2. The function
1
z
4
has a pole at 0, as
lim
z→0
¸
¸
¸
¸
1
z
4
¸
¸
¸
¸
= ∞.
85
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 86
Example 9.3. The function exp
_
1
z
_
does not have a removable singularity (consider, for example,
lim
x→0
+ exp
_
1
x
_
= ∞). On the other hand, exp
_
1
z
_
approaches 0 as z approaches 0 from the
negative real axis. Hence lim
z→0
¸
¸
exp
_
1
z
_¸
¸
= ∞, that is, exp
_
1
z
_
has an essential singularity at 0.
To get a feel for the diﬀerent types of singularities, we start with the following results.
Proposition 9.1. Suppose z
0
is a isolated singularity of f. Then
(a) z
0
is removable if and only if lim
z→z
0
(z −z
0
) f(z) = 0;
(b) if z
0
is a pole then lim
z→z
0
(z −z
0
)
n+1
f(z) = 0 for some positive integer n.
Remark. The smallest possible n in (b) is the order of the pole. We will see in the proof that “near
the pole z
0
” we can write f(z) as
h(z)
(z−z
0
)
n
for some function h which is analytic (and not zero) at
z
0
. This is very similar to the game we played with zeros in Chapter 8: f has a zero of order (or
multiplicity) m at z
0
if we can write f(z) = (z − z
0
)
m
h(z), where h is analytic and not zero at
z
0
. We will make use of the notions of zeros and poles of certain orders quite extensively in this
chapter.
Proof. (a) Suppose z
0
is removable, and g is analytic on D
R
(z
0
), the open disk with radius R
centered at z
0
such that f = g for z = z
0
. Then we can make use of the fact that g is continuous
at z
0
:
lim
z→z
0
(z −z
0
) f(z) = lim
z→z
0
(z −z
0
) g(z) = g(z
0
) lim
z→z
0
(z −z
0
) = 0 .
Conversely, suppose that lim
z→z
0
(z −z
0
) f(z) = 0, and f is analytic on the punctured disk
ˆ
D
R
(z
0
) = D
R
(z
0
) \ {z
0
}. Then deﬁne
g(z) =
_
(z −z
0
)
2
f(z) if z = z
0
,
0 if z = z
0
.
Clearly g is analytic for z = z
0
, and it is also diﬀerentiable at z
0
, since we can calculate
g
(z
0
) = lim
z→z
0
g(z) −g(z
0
)
z −z
0
= lim
z→z
0
(z −z
0
)
2
f(z)
z −z
0
= lim
z→z
0
(z −z
0
)f(z) = 0
So g is analytic in D
R
(z
0
) with g(z
0
) = 0 and g
(z
0
) = 0, so it has a power series expansion
g(z) =
k≥0
c
k
(z −z
0
)
k
with c
0
= c
1
= 0. Hence we can factor (z −z
0
)
2
from the series, so
g(z) = (z −z
0
)
2
k≥0
c
k+2
(z −z
0
)
k
= (z −z
0
)
2
f(z).
Hence, for z = z
0
, f(z) =
k≥0
c
k+2
(z−z
0
)
k
, and this series deﬁnes an analytic function in D
R
(z
0
).
(b) Suppose that z
0
is a pole of f. Then there is some R > 0 so that f(z) > 1 in the punctured
disk
ˆ
D
R
(z
0
), and
lim
z→z
0
1
f(z)
= 0 .
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 87
So, if we deﬁne g(z) by
g(z) =
_
1
f(z)
if z ∈
ˆ
D
R
(z
0
),
0 if z = z
0
,
then g is analytic in D
R
(z
0
) (by part (a)). By the classiﬁcation of zeros, g(z) = (z −z
0
)
n
φ(z) where
φ is analytic in D
R
(z
0
) and φ(z
0
) = 0. In fact, φ(z) = 0 for all z in D
R
(z
0
) since g(z) = 0 for
z ∈
ˆ
D
R
(z
0
). Hence h =
1
φ
is an analytic function in D
R
(z
0
) and
f(z) =
1
g(z)
=
1
(z −z
0
)
n
φ(z)
=
h(z)
(z −z
0
)
n
.
But then, since h is continuous at z
0
,
lim
z→z
0
(z −z
0
)
n+1
f(z) = lim
z→z
0
(z −z
0
)h(z) = h(z
0
) lim
z→z
0
(z −z
0
) = 0 .
The reader might have noticed that the previous proposition did not include any result on
essential singularities. Not only does the next theorem make up for this but it also nicely illustrates
the strangeness of essential singularities. To appreciate the following result, we suggest meditating
about its statement for a couple of minutes over a good cup of coﬀee.
Theorem 9.2 (Casorati
1
Weierstraß). If z
0
is an essential singularity of f and D = {z ∈ C : 0 <
z − z
0
 < R} for some R > 0, then any w ∈ C is arbitrarily close to a point in f(D), that is, for
any w ∈ C and any > 0 there exists z ∈ D such that w −f(z) < .
Remarks. 1. In the language of topology, the CasoratiWeierstraß theorem says that the image of
any punctured disc centered at an essential singularity is dense in C.
2. There is a much stronger theorem, which is beyond the scope of this book, and which implies
the CasoratiWeierstraß theorem. It is due to Charles Emile Picard (1856–1941)
2
and says that the
image of any punctured disc centered at an essential singularity misses at most one point of C. (It
is worth meditating about coming up with examples of functions which do not miss any point in C
and functions which miss exactly one point. Try it!)
Proof. Suppose (by way of contradiction) that there is a w ∈ C and an > 0 such that for all z in
the punctured disc D (centered at z
0
)
w −f(z) ≥ .
Then the function g(z) =
1
(f(z)−w)
stays bounded as z →z
0
, and so
lim
z→z
0
(z −z
0
)g(z) = lim
z→z
0
z −z
0
f(z) −w
= 0 .
(The previous proposition tells us that g has a removable singularity at z
0
.) Hence
lim
z→z
0
¸
¸
¸
¸
f(z) −w
z −z
0
¸
¸
¸
¸
= ∞.
But this implies that f has a pole or a removable singularity at z
0
, which is a contradiction.
1
For more information about Felice Casorati (1835–1890), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Casorati.html.
2
For more information about Picard, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Picard Emile.html.
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 88
Deﬁnition 9.1 is not always handy. The following classiﬁes singularities according to their
Laurent series.
Proposition 9.3. Suppose z
0
is an isolated singularity of f with Laurent series
f(z) =
k∈Z
c
k
(z −z
0
)
k
(valid in {z ∈ C : 0 < z −z
0
 < R} for some R > 0). Then
(a) z
0
is removable if and only if there are no negative exponents (that is, the Laurent series is a
power series);
(b) z
0
is a pole if and only if there are ﬁnitely many negative exponents;
(c) z
0
is essential if and only if there are inﬁnitely many negative exponents.
Proof. (a) Suppose z
0
is removable, and g is analytic on {z ∈ C : z −z
0
 < R} such that f = g in
{z ∈ C : 0 < z −z
0
 < R}. Then the Laurent series of g in this region is a power series, and by
Corollary 8.15 (uniqueness theorem for Laurent series) it has to coincide with the Laurent series of
f. Conversely, if the Laurent series of f at z
0
has only nonnegative powers, we can use it to deﬁne
a function which is analytic at z
0
.
(b) Suppose z
0
is a pole of order n. Then by Proposition 9.1, the function (z −z
0
)
n
f(z) has a
removable singularity at z
0
. By part (a), we can hence expand
(z −z
0
)
n
f(z) =
k≥0
c
k
(z −z
0
)
k
,
that is,
f(z) =
k≥0
c
k
(z −z
0
)
k−n
=
k≥−n
c
k
(z −z
0
)
k
.
Conversely, suppose that
f(z) =
k≥−n
c
k
(z −z
0
)
k
= (z −z
0
)
−n
k≥−n
c
k
(z −z
0
)
k+n
= (z −z
0
)
−n
k≥0
c
k−n
(z −z
0
)
k
,
where c
−n
= 0. Deﬁne
g(z) =
k≥0
c
k−n
(z −z
0
)
k
.
Then since g(z
0
) = c
−n
= 0,
lim
z→z
0
f(z) = lim
z→z
0
¸
¸
¸
¸
g(z)
(z −z
0
)
n
¸
¸
¸
¸
= ∞.
(c) This follows by deﬁnition: an essential singularity is neither removable nor a pole.
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 89
9.2 Residues
Suppose z
0
is an isolated singularity of f, γ is a positively oriented, simple, closed, smooth path
around z
0
, which lies in the domain of the Laurent series of f at z
0
. Then—essentially by Proposition
7.9—we can integrate term by term:
_
γ
f =
_
γ
k∈Z
c
k
(z −z
0
)
k
dz =
k∈Z
c
k
_
γ
(z −z
0
)
k
dz .
The integrals inside the summation are easy: for nonnegative powers k the integral
_
γ
(z − z
0
)
k
is
0 (because (z − z
0
)
k
is entire), and the same holds for k ≤ −2 (because (z − z
0
)
k
has a primitive
on C \ {z
0
}). Finally, for k = −1 we can use Exercise 8 of Chapter 4. Because all the other terms
give a zero integral, c
−1
is the only term of the series which survives:
_
γ
f =
k∈Z
c
k
_
γ
(z −z
0
)
k
dz = 2πi c
−1
.
(One might also notice that Theorem 8.14 gives the same identity.) Reason enough to give the
c
−1
coeﬃcient of a Laurent series a special name.
Deﬁnition 9.2. Suppose z
0
is an isolated singularity of f with Laurent series
k∈Z
c
k
(z − z
0
)
k
.
Then c
−1
is the residue of f at z
0
, denoted by Res
z=z
0
(f(z)) or Res(f(z), z = z
0
).
The following theorem generalizes the discussion at the beginning of this section.
γ
z
1
z
2
z
3
z
4
z
5
z
6
Figure 9.1: Proof of Theorem 9.4.
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 90
Theorem 9.4 (Residue Theorem). Suppose f is analytic in the region G, except for isolated
singularities, and γ is a positively oriented, simple, closed, smooth, Gcontractible curve. Then
_
γ
f = 2πi
k
Res
z=z
k
(f(z)) ,
where the sum is taken over all singularities z
k
inside γ.
Proof. Draw two circles around each isolated singularity inside γ, one with positive, and one with
negative orientation, as pictured in Figure 9.1. Each of these pairs cancel each other when we
integrate over them. Now connect the circles with negative orientation with γ. This gives a curve
which is contractible in the region of analyticity of f. But this means that we can replace γ by the
positively oriented circles; now all we need to do is described at the beginning of this section.
Computing integrals is as easy (or hard!) as computing residues. The following two lemmas
start the range of tricks one can use when computing residues.
Lemma 9.5. Suppose f and g are analytic in a region containing z
0
, which is a simple zero of g,
and f(z
0
) = 0. Then
f
g
has a simple pole at z
0
and
Res
z=z
0
_
f(z)
g(z)
_
=
f(z
0
)
g
(z
0
)
.
Proof. The functions f and g have power series centered at z
0
; the one for g has by assumption no
constant term:
g(z) =
k≥1
c
k
(z −z
0
)
k
= (z −z
0
)
k≥1
c
k
(z −z
0
)
k−1
.
The series on the right represents an analytic function, call it h; note that h(z
0
) = c
1
= 0. Hence
f(z)
g(z)
=
f(z)
(z −z
0
)h(z)
,
and the function
f
h
is analytic at z
0
. Even more, the residue of
f
g
equals the constant term of the
power series of
f
h
(that’s how we get the (−1)st term of
f
g
). But this constant term is computed,
as always, by
f(z
0
)
h(z
0
)
. But h(z
0
), in turn, is the constant term of h or the second term of g, which by
Taylor’s formula (Corollary 8.3) equals g
(z
0
).
Lemma 9.6. Suppose z
0
is a pole of f of order n. Then
Res
z=z
0
(f(z)) =
1
(n −1)!
lim
z→z
0
d
n−1
dz
n−1
_
(z −z
0
)
n
f(z)
_
.
Proof. We know by Proposition 9.3 that the Laurent series at z
0
looks like
f(z) =
k≥−n
c
k
(z −z
0
)
k
.
But then
(z −z
0
)
n
f(z) =
k≥−n
c
k
(z −z
0
)
k+n
represents a power series, and we can use Taylor’s formula (Corollary 8.3) to compute c
−1
.
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 91
9.3 Argument Principle and Rouch´e’s Theorem
There are many situations where we want to restrict ourselves to functions which are analytic in
some region except possibly for poles. Such functions are called meromorphic. In this section, we
will study these functions, especially with respect to their zeros and poles, which—as the reader
might have guessed already—can be thought of as siblings.
Suppose we have a diﬀerentiable function f. Diﬀerentiating Log f (where Log is a branch of the
logarithm) gives
f
f
, which is one good reason why this quotient is called the logarithmic derivative
of f. It has some remarkable properties, one of which we would like to discuss here.
Let’s say we have two functions f and g analytic in some region. Then the logarithmic derivative
of their product behaves very nicely:
(fg)
fg
=
f
g +fg
fg
=
f
f
+
g
g
.
We can apply this fact to the following situation: Suppose that f is analytic on the region G, and
f has the (ﬁnitely many) zeros z
1
, . . . , z
j
of order n
1
, . . . , n
j
, respectively. Then we can express f
as
f(z) = (z −z
1
)
n
1
· · · (z −z
j
)
n
j
g(z) ,
where g is also analytic in G and never zero. Let’s compute the logarithmic derivative of f and
play the same remarkable cancellation game as above:
f
(z)
f(z)
=
n
1
(z −z
1
)
n
1
−1
(z −z
2
)
n
2
· · · (z −z
j
)
n
j
g(z) +· · · + (z −z
1
)
n
1
· · · (z −z
j
)
n
j
g
(z)
(z −z
1
)
n
1
· · · (z −z
j
)
n
j
g(z)
=
n
1
z −z
1
+
n
2
z −z
2
+. . .
n
j
z −z
j
+
g
(z)
g(z)
.
Something similar happens to the poles of f. We invite the reader to prove that if p
1
, . . . , p
k
are
all the poles of f in G with order m
1
, . . . , m
k
, respectively, then the logarithmic derivative of f can
be expressed as
f
(z)
f(z)
= −
m
1
z −p
1
−
m
2
z −p
2
−· · · −
m
k
z −p
k
+
g
(z)
g(z)
, (9.1)
where g is a function without poles in G. Naturally, we can combine the expressions we got for
zeros and poles, which is the starting point of the following theorem.
Theorem 9.7 (Argument Principle). Suppose f is meromorphic in the region G and γ is a posi
tively oriented, simple, closed, smooth, Gcontractible curve, which does not pass through any zero or
pole of f. Denote by Z(f, γ) the number of zeros of f inside γ—counted according to multiplicity—
and by P(f, γ) the number of poles of f inside γ, again counted according to multiplicity. Then
1
2πi
_
γ
f
f
= Z(f, γ) −P(f, γ) .
Proof. Suppose the zeros of f inside γ are z
1
, . . . , z
j
of order n
1
, . . . , n
j
, respectively, and the poles
inside γ are p
1
, . . . , p
k
with order m
1
, . . . , m
k
, respectively. (You may meditate about the fact why
there can only be ﬁnitely many zeros and poles inside γ.) In fact, we may shrink G, if necessary, so
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 92
that these are the only zeros and poles in G. Our discussion before the statement of the theorem
yielded that the logarithmic derivative of f can be expressed as
f
(z)
f(z)
=
n
1
z −z
1
+· · · +
n
j
z −z
j
−
m
1
z −p
1
−· · · −
m
k
z −p
k
+
g
(z)
g(z)
,
where g is a function which is analytic in G (in particular, without poles) and never zero. Thanks
to Exercise 8 of Chapter 4, the integral is easy:
_
γ
f
f
= n
1
_
γ
dz
z −z
1
+ · · · + n
j
_
γ
dz
z −z
j
− m
1
_
γ
dz
z −p
1
− · · · − m
k
_
γ
dz
z −p
k
+
_
γ
g
g
= 2πi (n
1
+· · · +n
j
−m
1
−· · · −m
k
) +
_
γ
g
g
.
Finally,
g
g
is analytic in G (recall that g is never zero in G), so that Corollary 4.5 (to Cauchy’s
Theorem 4.4) gives that
_
γ
g
g
= 0 .
As a nice application of the argument principle, we present a famous theorem due to Eugene
Rouch´e (1832–1910)
3
.
Theorem 9.8 (Rouch´e’s Theorem). Suppose f and g are analytic in a region G, and γ is a
positively oriented, simple, closed, smooth, Gcontractible curve such that for all z ∈ γ, f(z) >
g(z). Then
Z(f +g, γ) = Z(f, γ) .
This theorem is of surprising practicality. It allows us to locate the zeros of a function fairly
precisely. As an illustration, we prove:
Example 9.4. All the roots of the polynomial p(z) = z
5
+z
4
+z
3
+z
2
+z + 1 have absolute value
less than two.
4
To see this, let f(z) = z
5
and g(z) = z
4
+ z
3
+ z
2
+ z + 1, and let γ denote the
circle centered at the origin with radius 2. Then for z ∈ γ
g(z) ≤ z
4
+z
3
+z
2
+z + 1 = 16 + 8 + 4 + 2 + 1 = 31 < 32 = z
5
= f(z) .
So g and f satisfy the condition of the Theorem 9.8. But f has just a root of order 5 at the origin,
whence
Z(p, γ) = Z(f +g, γ) = Z(f, γ) = 5 .
3
For more information about Rouch´e, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Rouche.html.
4
The fundamental theorem of algebra (Theorem 5.4) asserts that p has ﬁve roots in C. What’s special about the
statement of Example 9.4 is that they all have absolute value < 2. Note also that there is no general formula for
computing roots of a polynomial of degree 5. (Although for this p it’s not hard to ﬁnd one root—and therefore all of
them.)
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 93
Proof of Theorem 9.8. By our analysis in the beginning of this section and by the argument prin
ciple (Theorem 9.7)
Z(f +g, γ) =
1
2πi
_
γ
(f +g)
f +g
=
1
2πi
_
γ
_
f
_
1 +
g
f
__
f
_
1 +
g
f
_ =
1
2πi
_
γ
_
_
_
f
f
+
_
1 +
g
f
_
1 +
g
f
_
_
_
= Z(f, γ) +
1
2πi
_
γ
_
1 +
g
f
_
1 +
g
f
.
We are assuming that
¸
¸
¸
g
f
¸
¸
¸ < 1 on γ, which means that the function 1 +
g
f
evaluated on γ stays
away from the nonpositive real axis. But then Log
_
1 +
g
f
_
is a well deﬁned analytic function on γ.
Its derivative is
_
1 +
g
f
_
1 +
g
f
, which implies by Corollary 4.3 that
1
2πi
_
γ
_
1 +
g
f
_
1 +
g
f
= 0 .
Exercises
1. Prove (9.1).
2. Show that if f has an essential singularity at z
0
then
1
f
also has an essential singularity at z
0
.
3. Suppose f is a nonconstant entire function. Prove that any complex number is arbitrarily
close to a number in f(C). (Hint: If f is not a polynomial, use Theorem 9.2 for f
_
1
z
_
.)
4. Suppose f is meromorphic in the region G, g is analytic in G, and γ is a positively oriented,
simple, closed, Gcontractible curve, which does not pass through any zero or pole of f.
Denote the zeros and poles of f inside γ by z
1
, . . . , z
j
and p
1
, . . . , p
k
, respectively, counted
according to multiplicity. Prove that
1
2πi
_
γ
g
f
f
=
j
m=1
g(z
m
) −
k
n=1
g(p
n
) .
5. Find the number of zeros of
(a) 3 exp z −z in {z ∈ C : z ≤ 1} ;
(b)
1
3
exp z −z in {z ∈ C : z ≤ 1} ;
(c) z
4
−5z + 1 in {z ∈ C : 1 ≤ z ≤ 2} .
6. Give another proof of the fundamental theorem of algebra (Theorem 5.4), using Rouch´e’s
Theorem 9.8. (Hint: If p(z) = a
n
z
n
+ a
n−1
z
n−1
+ · · · + a
1
z + 1, let f(z) = a
n
z
n
and
g(z) = a
n−1
z
n−1
+ a
n−2
z
n−2
+ · · · + a
1
z + 1, and choose as γ a circle which is large enough
to make the condition of Rouch´e’s theorem work. You might want to ﬁrst apply Lemma 5.3
to g(z).)
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 94
7. (a) Find a Laurent series for
1
(z
2
−4)(z−2)
centered at z = 2 and specify the region in which
it converges.
(b) Compute
_
γ
dz
(z
2
−4)(z−2)
, where γ is the positively oriented circle centered at 2 of radius 1.
8. Evaluate the following integrals for γ(t) = 3 e
it
, 0 ≤ t ≤ 2π.
(a)
_
γ
cot z dz
(b)
_
γ
z
3
cos
_
3
z
_
dz
(c)
_
γ
dz
(z + 4)(z
2
+ 1)
(d)
_
γ
z
2
exp
_
1
z
_
dz
(e)
_
γ
exp z
sinh z
dz
(f)
_
γ
i
z+4
(z
2
+ 16)
2
dz
9. (a) Find the power series of exp z centered at z = −1.
(b) Find
_
γ
exp z
(z+1)
34
dz, where γ is the circle z + 2 = 2, positively oriented.
10. Suppose f has a simple pole (i.e., a pole of order 1) at z
0
and g is analytic at z
0
. Prove that
Res
z=z
0
_
f(z)g(z)
_
= g(z
0
) · Res
z=z
0
_
f(z)
_
.
11. Find the residue of each function at 0:
(a) z
−3
cos(z).
(b) csc(z).
(c)
z
2
+ 4z + 5
z
2
+z
.
(d) e
1−
1
z
.
(e)
e
4z
−1
sin
2
z
.
12. Use residues to evaluate the following:
(a)
_
γ
dz
z
4
+ 4
, where γ is the circle z + 1 −i = 1.
(b)
_
γ
dz
z(z
2
+z −2)
, where γ is the circle z −i = 2.
(c)
_
γ
e
z
dz
z
3
+z
, where γ is the circle z = 2.
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 95
(d)
_
γ
dz
z
2
sin z
, where γ is the circle z = 1.
13. Suppose f has an isolated singularity at z
0
.
(a) Show that f
also has an isolated singularity at z
0
.
(b) Find Res
z=z
0
(f
).
14. Given R > 0, let γ
R
be the half circle deﬁned by γ
R
(t) = Re
it
, 0 ≤ t ≤ π, and Γ
R
be the
closed curve composed of γ
R
and the line segment [−R, R].
(a) Compute
_
Γ
R
dz
(1+z
2
)
2
.
(b) Prove that lim
R→∞
_
γ
R
dz
(1+z
2
)
2
= 0 .
(c) Combine (a) and (b) to evaluate the real integral
_
∞
−∞
dx
(1+x
2
)
2
.
15. Suppose f is entire, and a, b ∈ C with a, b < R. Let γ be the circle centered at 0 with
radius R. Evaluate
_
γ
f(z)
(z −a)(z −b)
dz ,
and use this to give an alternate proof of Liouville’s Theorem 5.5. (Hint: Show that if f is
bounded then the above integral goes to zero as R increases.)
Chapter 10
Discreet Applications of the Residue
Theorem
All means (even continuous) sanctify the discrete end.
Doron Zeilberger
On the surface, this chapter is just a collection of exercises. They are more involved than any of
the ones we’ve given so far at the end of each chapter, which is one reason why we lead the reader
through each of the following ones step by step. On the other hand, these sections should really
be thought of as a continuation of the lecture notes, just in a diﬀerent format. All of the following
‘problems’ are of a discrete mathematical nature, and we invite the reader to solve them using
continuous methods—namely, complex integration. It might be that there is no other result which
so intimately combines discrete and continuous mathematics as does the Residue Theorem 9.4.
10.1 Inﬁnite Sums
In this exercise, we evaluate—as an example—the sums
k≥1
1
k
2
and
k≥1
(−1)
k
k
2
. We hope the
idea how to compute such sums in general will become clear.
1. Consider the function f(z) =
π cot(πz)
z
2
. Compute the residues at all the singularities of f.
2. Let N be a positive integer and γ
N
be the rectangular curve from N+1/2−iN to N+1/2+iN
to −N −1/2 +iN to −N −1/2 −iN back to N + 1/2 −iN.
(a) Show that for all z ∈ γ
N
,  cot(πz) < 2. (Use Exercise 21 in Chapter 3.)
(b) Show that lim
N→∞
_
γ
N
f = 0.
3. Use the Residue Theorem 9.4 to arrive at an identity for
k∈Z\{0}
1
k
2
.
4. Evaluate
k≥1
1
k
2
.
5. Repeat the exercise with the function f(z) =
π
z
2
sin(πz)
to arrive at an evaluation of
k≥1
(−1)
k
k
2
.
96
CHAPTER 10. DISCREET APPLICATIONS OF THE RESIDUE THEOREM 97
(Hint: To bound this function, you may use the fact that 1/ sin
2
z = 1 + cot
2
z.)
6. Evaluate
k≥1
1
k
4
and
k≥1
(−1)
k
k
4
.
10.2 Binomial Coeﬃcients
The binomial coeﬃcient
_
n
k
_
is a natural candidate for being explored analytically, as the binomial
theorem
1
tells us that
_
n
k
_
is the coeﬃcient of z
k
in (1 +z)
n
. As an example, we outline a proof of
the identity (for −1/4 < x < 1/4)
k≥0
_
2k
k
_
x
k
=
1
√
1 −4x
.
1. Convince yourself that
_
2k
k
_
=
1
2πi
_
γ
(1 +w)
2k
w
k+1
dw,
where γ is any simple closed curve such that 0 is inside γ.
2. Suppose x < 1/4. Find a simple closed curve γ surrounding the origin such that
k≥0
_
(1 +w)
2
w
x
_
k
converges uniformly on γ (as a function in w). Evaluate this sum.
3. Convince yourself that
k≥0
_
2k
k
_
x
k
=
1
2πi
k≥0
_
γ
(1 +w)
2k
w
k
x
k
dw
w
,
use 2. to interchange summation and integral, and use the Residue Theorem 9.4 to evaluate
the integral.
10.3 Fibonacci Numbers
The Fibonacci
2
numbers are a sequence of integers deﬁned recursively as:
f
0
= 1,
f
1
= 1,
f
n
= f
n−1
+f
n−2
for n ≥ 2.
Let F(z) =
k≥0
f
n
z
n
.
1
The binomial theorem says that for x, y ∈ C and n ∈ N, (x + y)
n
=
P
n
k=0
`
n
k
´
x
k
y
n−k
.
2
For more information about Leonardo Pisano Fibonacci (1170–1250), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Fibonacci.html.
CHAPTER 10. DISCREET APPLICATIONS OF THE RESIDUE THEOREM 98
1. Show that F has a positive radius of convergence.
2. Show that the recurrence relation among the f
n
implies that F(z) =
1
1−z−z
2
. (Hint: Write
down the power series of zF(z) and z
2
F(z) and rearrange both so that you can easily add.)
3. Verify that Res
z=0
_
1
z
n+1
(1−z−z
2
)
_
= f
n
.
4. Use the Residue Theorem 9.4 to derive an identity for f
n
. (Hint: Integrate
1
z
n+1
(1−z−z
2
)
around a circle with center 0 and radius R, and show that this integral vanishes as R →∞.)
5. Generalize to other recurrence relations.
10.4 The ‘CoinExchange Problem’
In this exercise, we will solve and extend a classical problem of Ferdinand Georg Frobenius (1849–
1917)
3
. Suppose a and b are relatively prime
4
positive integers, and t is a positive integer. Consider
the function
f(z) =
1
(1 −z
a
) (1 −z
b
) z
t+1
.
1. Compute the residues at all nonzero poles of f.
2. Verify that Res
z=0
(f) = N(t), where
N(t) = #{(m, n) ∈ Z : m, n ≥ 0, ma +nb = t} .
3. Use the Residue Theorem 9.4 to derive an identity for N(t). (Hint: Integrate f around a
circle with center 0 and radius R, and show that this integral vanishes as R →∞.)
4. Use the following three steps to simplify this identity to
N(t) =
t
ab
−
_
b
−1
t
a
_
−
_
a
−1
t
b
_
+ 1 .
Here, {x} denotes the fractional part
5
of x, and a
−1
a ≡ 1 (mod b)
6
, and b
−1
b ≡ 1 (mod a).
(a) Verify that for b = 1,
N(t) = #{(m, n) ∈ Z : m, n ≥ 0, ma +n = t} = #{m ∈ Z : m ≥ 0, ma ≤ t}
= #
__
0,
t
a
_
∩ Z
_
=
t
a
−
_
t
a
_
+ 1 .
3
For more information about Frobenius, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Frobenius.html.
4
this means that the integers don’t have any common factor
5
The fractional part of a real number x is, loosely speaking, the “part after the decimal point.” More thoroughly,
the greatest integer function of x, denoted by x, is the greatest integer not exceeding x. The fractional part is then
{x} = x − x.
6
This means that a
−1
is an integer such that a
−1
a = 1 + kb for some k ∈ Z.
CHAPTER 10. DISCREET APPLICATIONS OF THE RESIDUE THEOREM 99
(b) Use this together with the identity found in 3. to obtain
1
a
a−1
k=1
1
(1 −e
2πik/a
)e
2πikt/a
= −
_
t
a
_
+
1
2
−
1
2a
.
(c) Verify that
a−1
k=1
1
(1 −e
2πikb/a
)e
2πikt/a
=
a−1
k=1
1
(1 −e
2πik/a
)e
2πikb
−1
t/a
.
5. Prove that N(ab −a −b) = 0, and N(t) > 0 for all t > ab −a −b.
6. More generally, prove that, if k is a nonnegative integer, N ((k + 1)ab −a −b) = k, and
N(t) > k for all t > (k + 1)ab −a −b.
Historical remark. Given relatively prime positive integers a
1
, . . . , a
n
, let’s call an integer t repre
sentable if there exist nonnegative integers m
1
, . . . , m
n
such that
t =
n
j=1
m
j
a
j
.
In the late 19th century, Frobenius raised the problem of ﬁnding the largest integer which is
not representable. We call this largest integer the Frobenius number g(a
1
, . . . , a
n
). It is well
known (probably at least since the 1880’s, when James Joseph Sylvester (1814–1897)
7
studied the
Frobenius problem) that g(a
1
, a
2
) = a
1
a
2
− a
1
− a
2
. We veriﬁed this result in 5. For n > 2, there
is no known closed formula for g(a
1
, . . . , a
n
). The formula in 4. is due to Popoviciu. The notion of
an integer being representable k times and the respective formula obtained in 6. can only be found
in the most recent literature.
10.5 Dedekind sums
This exercise outlines yet another nontraditional application of the Residue Theorem 9.4. Given
two positive, relatively prime integers a and b, let
f(z) = cot(πaz) cot(πbz) cot(πz) .
1. Choose an > 0 such that the rectangular path γ
R
from 1 − −iR to 1 − +iR to − +iR
to − −iR back to 1 − −iR does not pass through any of the poles of f.
(a) Compute the residues for the poles of f inside γ
R
.
Hint: use the periodicity of the cotangent and the fact that
cot z =
1
z
−
1
3
z + higherorder terms .
7
For more information about Sylvester, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Sylvester.html.
CHAPTER 10. DISCREET APPLICATIONS OF THE RESIDUE THEOREM 100
(b) Prove that lim
R→∞
_
γ
R
f = −2i and deduce that for any R > 0
_
γ
R
f = −2i .
2. Deﬁne
s(a, b) =
1
4b
b−1
k=1
cot
_
πka
b
_
cot
_
πk
b
_
. (10.1)
Use the Residue Theorem 9.4 to show that
s(a, b) +s(b, a) = −
1
4
+
1
12
_
a
b
+
1
ab
+
b
a
_
. (10.2)
3. Can you generalize (10.1) and (10.2)?
Historical remark. The sum (10.1) is called a Dedekind
8
sum. It ﬁrst appeared in the study of the
Dedekind ηfunction
η(z) = exp
_
πiz
12
_
k≥1
(1 −exp(2πikz))
in the 1870’s and has since intrigued mathematicians from such diﬀerent areas as topology, number
theory, and discrete geometry. The reciprocity law (10.2) is the most important and famous identity
of the Dedekind sum. The proof that is outlined here is due to Hans Rademacher (1892–1969)
9
.
8
For more information about Julius Wilhelm Richard Dedekind (1831–1916), see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Dedekind.html.
9
For more information about Rademacher, see
http://wwwgroups.dcs.stand.ac.uk/∼history/Biographies/Rademacher.html.
Solutions to Selected Exercises
Chapter 1
1. (b)
19
25
−
8
25
i
(c) 1
(d) 1 if n = 4k, k ∈ Z; i if n = 1 + 4k, k ∈ Z; −1 if n = 2 + 4k, k ∈ Z; −i if n = 3 + 4k, k ∈ Z.
2. (a)
√
5, −2 −i
(b) 5
√
5, 5 −10i
(c)
_
10
11
,
3
11
(
√
2 −1) +
i
11
(
√
2 + 9)
(d) 8, 8i
3. (a) 2e
i
π
2
(b)
√
2e
i
π
4
(c) 2
√
3e
i
5π
6
4. (a) −1 +i
(b) 34i
(c) −1
5. (a) z = e
i
π
3
k
, k = 0, 1, . . . , 5
(b) z = 2e
i
π
4
+
π
2
k
, k = 0, 1, 2, 3
7. z = e
i
π
4
−1 and z = e
i
5π
4
−1
Chapter 2
2. (a) 0
(b) 1 +i
10. (a) diﬀerentiable and analytic in C with derivative −e
−x
e
−iy
(b) nowhere diﬀerentiable or analytic
(c) diﬀerentiable on {x +iy ∈ C : x = y} with derivative 2x, nowhere analytic
(d) nowhere diﬀerentiable or analytic
(e) diﬀerentiable and analytic in C with derivative −sin xcosh y −i cos xsinh y
(f) diﬀerentiable at 0 with derivative 0, nowhere analytic
(g) diﬀerentiable at 0 with derivative 0, nowhere analytic
(h) diﬀerentiable only at i with derivative i, nowhere analytic
(i) diﬀerentiable and analytic in C with derivative −2iz
Chapter 3
26. (a) diﬀerentiable at 0, nowhere analytic
101
SOLUTIONS TO SELECTED EXERCISES 102
(b) diﬀerentiable and analytic on C \
_
−1, e
i
π
3
, e
−i
π
3
_
(c) diﬀerentiable and analytic on C \ {x +iy ∈ C : x ≥ −1, y = 2}
(d) nowhere diﬀerentiable or analytic
(e) diﬀerentiable and analytic on C \ {x +iy ∈ C : x ≤ 3, y = 0}
(f) diﬀerentiable and analytic in C (i.e. entire)
27. (a) z = i
(b) There is no solution.
(c) z = ln π +i
_
π
2
+ 2πk
_
, k ∈ Z
(d) z =
π
2
+ 2πk ±4i, k ∈ Z
(e) z =
π
2
+πk, k ∈ Z
(f) z = πk, k ∈ Z
(g) z = 2i
30. f
(z) = c z
c−1
Chapter 4
2. −2πi
3. (a) 8πi
(b) 0
(c) 0
(d) 0
14. 0
16.
2π
√
3
23 0 for r < a; 2πi for r > a
24 0 for r = 1; −
πi
3
for r = 3; 0 for r = 5
Chapter 5
2. (a) 0
(b) 2πi
(c) 0
(d) πi
(e) 0
(f) 0
5. Any simply connected set which does not contain the origin, for example, C \ (−∞, 0].
Chapter 7
1. (a) divergent
(b) convergent (limit 0)
(c) divergent
(d) convergent (limit 2 −
i
2
)
(e) convergent (limit 0)
16. (a) pointwise convergent for z < 1, uniform for z ≤ R for any ﬁxed R < 1
(b) pointwise and uniformly convergent for z ≤ 1
(c) pointwise convergent for all z ∈ H := {Re z ≥ 0}, uniform on H∩{z ≥ R} for any ﬁxed R > 0.
SOLUTIONS TO SELECTED EXERCISES 103
18. (a)
k≥0
(−4)
k
z
k
(b)
k≥0
1
3·6
k
z
k
20. (a)
k≥0
(−1)
k
(z −1)
k
(b)
k≥1
(−1)
k−1
k
(z −1)
k
23. (a) ∞ if a < 1, 1 if a = 1, and 0 if a > 1.
(b) 1
(c) 1 (careful reasoning!)
(d) 1 (careful reasoning!)
Chapter 8
1. (a) {z ∈ C : z < 1}, {z ∈ C : z ≤ r} for any r < 1
(b) C, {z ∈ C : z ≤ r} for any r
(c) {z ∈ C : z −3 > 1}, {z ∈ C : r ≤ z −3 ≤ R} for any 1 < r ≤ R
3.
k≥0
e
k!
(z −1)
k
10. The maximum is 3 (attained at z = ±i), and the minimum is 1 (attained at z = ±1).
12. One Laurent series is
k≥0
(−2)
k
(z −1)
−k−2
, converging for z −1 > 2.
13. One Laurent series is
k≥0
(−2)
k
(z −2)
−k−3
, converging for z −2 > 2.
14. One Laurent series is −3(z + 1)
−1
+ 1, converging for z = −1.
15.
1
sin z
= z
−1
+
1
6
z +
7
360
z
3
+. . .
20. (a)
k≥0
(−1)
k
(2k)!
z
2k−2
Chapter 9
5. (a) 0
(b) 1
(c) 4
7. (a) One Laurent series is
k≥−2
(−1)
k
4
k+3
(z −2)
k
, converging for 0 < z −2 < 4.
(b) −
πi
8
8. (a) 2πi
(b)
27πi
4
(c) −
2πi
17
(d)
πi
3
(e) 2πi
(f) 0
9. (a)
k≥0
1
e k!
(z + 1)
k
(b)
2πi
e 33!
14. (c)
π
2
Index
absolute convergence, 65
absolute value, 3
addition, 1
analytic, 15
antiderivative, 39, 53
Arg, 31
arg, 31
argument, 3
axis
imaginary, 2
real, 2
bijection, 16
binomial coeﬃcient, 97
boundary, 60
branch of the logarithm, 31
CasoratiWeierstraß theorem, 87
Cauchy’s estimate, 76
Cauchy’s integral formula, 42, 46
extensions of, 48, 76
Cauchy’s theorem, 40
Cauchy–Riemann equations, 17
chain rule, 16
closed
algebraically, 51
curve, 8
closed set, 7
coﬀee, 87
conjugate, 5
connected, 7
continuous, 14
contractible, 41
convergent
sequence, 62
series, 64
cosine, 29
cotangent, 29
curve, 8
Dedekind sum, 99
dense, 87
derivative, 15
diﬀerence quotient, 15
diﬀerentiable, 15
diﬀerentiation rule, 16
dilation, 24
Dirichlet problem, 61
distance
of numbers, 3, 6
divergent, 62
domain, 13
double series, 79
e, 32
embedding of R in C, 2
entire, 15, 51, 56
essential singularity, 85
exponential function, 29
exponential rules, 29
Fibonacci numbers, 97
ﬁeld, 1
Frobenius problem, 98
function, 13
fundamental theorem
of algebra, 51, 83, 92, 93
of calculus, 9
geometric series, 68
group, 2
abelian, 2
harmonic, 18, 57
harmonic conjugate, 58
homotopic, 40
homotopy, 40
104
INDEX 105
hyperbolic trig functions, 30
i, 2
identity map, 13
image, 13
imaginary part, 2
integral, 37
integration by parts, 46
inverse function, 16
inversion, 24
isolated singularity, 85
Laplace equation, 57
Laurent series, 80
Leibniz’s rule, 10
length, 38
limit
of a function, 13, 15
of a sequence, 62
of a series, 64
linear fractional transformation, 23
Log, 31
log, 31
logarithm, 31
logarithmic derivative, 91
max/min property for harmonic functions, 59,
79
maximum
strong relative, 59
weak relative, 60, 79
maximummodulus theorem, 79
meanvalue theorem
for analytic functions, 44
for harmonic functions, 59
for real functions, 9
meromorphic, 91
minimum
strong relative, 59
weak relative, 79
minimummodulus theorem, 82
M¨obius transformation, 23
modulus, 3
Morera’s theorem, 54
multiplication, 1
obvious, 13, 23
onetoone, 16
onto, 16
open set, 7
order
of a pole, 86
parametrization, 37
path, 8
path independent, 40
periodic, 29
Picard’s theorem, 87
piecewise smooth, 37
pointwise convergence, 66
polar form, 5
pole, 85
polynomial, 11, 21, 51, 56
power series, 68
diﬀerentiation of, 74
integration of, 69
primitive, 39, 53
principal argument, 31
principal logarithm, 31
principal value of a
b
, 32
real part, 2
rectangular form, 5
region, 7
region of convergence, 69
removable singularity, 85
residue, 89
residue theorem, 90
reverse triangle inequality, 11
Rouch´e’s theorem, 92
separated, 7
sequence, 62
series, 64
simple closed curve, 8
simply connected, 53
sine, 29
singularity, 85
smooth, 8
tangent, 29
Taylor series expansion, 75
INDEX 106
topology, 6
translation, 24
triangle inequality, 5
trigonometric functions, 29
trivial, 14
uniform convergence, 66
uniqueness theorem, 78
Weierstraß Mtest, 68
Weierstraß convergence theorem, 82
2
These are the lecture notes of a onesemester undergraduate course which we have taught several times at Binghamton University (SUNY) and San Francisco State University. For many of our students, complex analysis is their ﬁrst rigorous analysis (if not mathematics) class they take, and these notes reﬂect this very much. We tried to rely on as few concepts from real analysis as possible. In particular, series and sequences are treated “from scratch.” This also has the (maybe disadvantageous) consequence that power series are introduced very late in the course. We thank our students who made many suggestions for and found errors in the text. Special thanks go to Collin Bleak, Jon Clauss, Sharma Pallekonda, and Joshua Palmatier for comments after teaching from this book.
Contents
1 Complex Numbers 1.1 Deﬁnition and Algebraic Properties 1.2 Geometric Properties . . . . . . . . 1.3 Elementary Topology of the Plane 1.4 Theorems from Calculus . . . . . . Exercises . . . . . . . . . . . . . . . . . 2 Diﬀerentiation 2.1 First Steps . . . . . . . . . . . . 2.2 Diﬀerentiability and Analyticity . 2.3 The Cauchy–Riemann Equations 2.4 Constants and Connectivity . . . Exercises . . . . . . . . . . . . . . . . 1 1 2 6 9 10 13 13 15 17 19 20 23 23 26 28 31 33 37 37 39 40 42 44 48 48 50 53 55
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
3 Examples of Functions 3.1 M¨bius Transformations . . . . . . . . . . . o 3.2 Inﬁnity and the Cross Ratio . . . . . . . . . 3.3 Exponential and Trigonometric Functions . 3.4 The Logarithm and Complex Exponentials Exercises . . . . . . . . . . . . . . . . . . . . . . 4 Integration 4.1 Deﬁnition and Basic Properties 4.2 Antiderivatives . . . . . . . . . 4.3 Cauchy’s Theorem . . . . . . . 4.4 Cauchy’s Integral Formula . . . Exercises . . . . . . . . . . . . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
5 Consequences of Cauchy’s Theorem 5.1 Extensions of Cauchy’s Formula . . . . 5.2 Taking Cauchy’s Formula to the Limit 5.3 Antiderivatives Revisited and Morera’s Exercises . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . Theorem . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
3
CONTENTS 6 Harmonic Functions 6.1 Deﬁnition and Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2 MeanValue and Maximum/Minimum Principle . . . . . . . . . . . . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Power Series 7.1 Sequences and Completeness . . . 7.2 Series . . . . . . . . . . . . . . . . 7.3 Sequences and Series of Functions 7.4 Region of Convergence . . . . . . . Exercises . . . . . . . . . . . . . . . . .
4 57 57 59 61 62 62 64 66 68 70 74 74 77 79 82 85 85 89 91 93 96 96 97 97 98 99 101 104
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
8 Taylor and Laurent Series 8.1 Power Series and Analytic Functions . 8.2 Classiﬁcation of Zeros and the Identity 8.3 Laurent Series . . . . . . . . . . . . . . Exercises . . . . . . . . . . . . . . . . . . .
. . . . . . Principle . . . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
9 Isolated Singularities and the Residue Theorem 9.1 Classiﬁcation of Singularities . . . . . . . . . . . 9.2 Residues . . . . . . . . . . . . . . . . . . . . . . . 9.3 Argument Principle and Rouch´’s Theorem . . . e Exercises . . . . . . . . . . . . . . . . . . . . . . . . . 10 Discreet Applications of the Residue 10.1 Inﬁnite Sums . . . . . . . . . . . . . 10.2 Binomial Coeﬃcients . . . . . . . . . 10.3 Fibonacci Numbers . . . . . . . . . . 10.4 The ‘CoinExchange Problem’ . . . . 10.5 Dedekind sums . . . . . . . . . . . . Solutions to Selected Exercises Index Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
Chapter 1
Complex Numbers
Die ganzen Zahlen hat der liebe Gott geschaﬀen, alles andere ist Menschenwerk. (God created the integers, everything else is made by humans.) Leopold Kronecker (1823–1891)
1.1
Deﬁnition and Algebraic Properties
The complex numbers can be deﬁned as pairs of real numbers, C = {(x, y) : x, y ∈ R} , equipped with the addition (x, y) + (a, b) = (x + a, y + b) and the multiplication (x, y) · (a, b) = (xa − yb, xb + ya) . One reason to believe that the deﬁnitions of these binary operations are “good” is that C is an extension of R, in the sense that the complex numbers of the form (x, 0) behave just like real numbers; that is, (x, 0) + (y, 0) = (x + y, 0) and (x, 0) · (y, 0) = (x · y, 0). So we can think of the real numbers being embedded in C as those complex numbers whose second coordinate is zero. The following basic theorem states the algebraic structure that we established with our deﬁnitions. Its proof is straightforward but nevertheless a good exercise. Theorem 1.1. (C, +, ·) is a ﬁeld; that is: ∀ (x, y), (a, b) ∈ C : (x, y) + (a, b) ∈ C ∀ (x, y), (a, b), (c, d) ∈ C : (x, y) + (a, b) + (c, d) = (x, y) + (a, b) + (c, d) ∀ (x, y), (a, b) ∈ C : (x, y) + (a, b) = (a, b) + (x, y) ∀ (x, y) ∈ C : (x, y) + (0, 0) = (x, y) ∀ (x, y) ∈ C : (x, y) + (−x, −y) = (0, 0) (1.1) (1.2) (1.3) (1.4) (1.5)
1
CHAPTER 1. COMPLEX NUMBERS ∀ (x, y), (a, b) ∈ C : (x, y) · (a, b) ∈ C ∀ (x, y), (a, b), (c, d) ∈ C : (x, y) · (a, b) · (c, d) = (x, y) · (a, b) · (c, d) ∀ (x, y), (a, b) ∈ C : (x, y) · (a, b) = (a, b) · (x, y) ∀ (x, y) ∈ C : (x, y) · (1, 0) = (x, y) ∀ (x, y) ∈ C \ {(0, 0)} : (x, y) ·
x , x2 +y 2 −y x2 +y 2
2 (1.6) (1.7) (1.8) (1.9) = (1, 0) (1.10)
Remark. What we are stating here can be compressed in the language of algebra: equations (1.1)– (1.5) say that (C, +) is an Abelian group with unit element (0, 0), equations (1.6)–(1.10) that (C \ {(0, 0)}, ·) is an abelian group with unit element (1, 0). (If you don’t know what these terms mean—don’t worry, we will not have to deal with them.) The deﬁnition of our multiplication implies the innocent looking statement (0, 1) · (0, 1) = (−1, 0) . This identity together with the fact that (a, 0) · (x, y) = (ax, ay) allows an alternative notation for complex numbers. The latter implies that we can write (x, y) = (x, 0) + (0, y) = (x, 0) · (1, 0) + (y, 0) · (0, 1) . If we think—in the spirit of our remark on the embedding of R in C—of (x, 0) and (y, 0) as the real numbers x and y, then this means that we can write any complex number (x, y) as a linear combination of (1, 0) and (0, 1), with the real coeﬃcients x and y. (1, 0), in turn, can be thought of as the real number 1. So if we give (0, 1) a special name, say i, then the complex number that we used to call (x, y) can be written as x · 1 + y · i, or in short, x + iy . The number x is called the real part and y the imaginary part1 of the complex number x + iy, often denoted as Re(x + iy) = x and Im(x + iy) = y. The identity (1.11) then reads i2 = −1 . We invite the reader to check that the deﬁnitions of our binary operations and Theorem 1.1 are coherent with the usual real arithmetic rules if we think of complex numbers as given in the form x + iy. (1.11)
1.2
Geometric Properties
Although we just introduced a new way of writing complex numbers, let’s for a moment return to the (x, y)notation. It suggests that one can think of a complex number as a twodimensional real vector. When plotting these vectors in the plane R2 , we will call the xaxis the real axis and the yaxis the imaginary axis. The addition that we deﬁned for complex numbers resembles vector addition. The analogy stops at multiplication: there is no “usual” multiplication of two vectors
CHAPTER 1. This means. The ﬁrst hint that absolute value and argument of a complex number are useful concepts is the fact that they allow us to give a geometric interpretation for the multiplication of two complex numbers. that any complex number has many arguments. imagined. it is also determined by its length and the angle it encloses with. The absolute value of the diﬀerence of two vectors has a nice geometric interpretation: it is the distance of the (end points of the) two vectors (see Figure 1. we can write 1 The name has historical reasons: people thought of complex numbers as unreal. let’s deﬁne these concepts thoroughly. COMPLEX NUMBERS z1 +/ z2 W 3 // z1 // D // // / z2 kWWWWW WWWWW /// WW Figure 1. It is very useful to keep this geometric interpretation in mind when thinking about the absolute value of the diﬀerence of two complex numbers.1: Addition of complex numbers.2). Let’s say we have two complex numbers. and x2 + iy2 with absolute value r2 and argument φ2 . Any vector in R2 is deﬁned by its two coordinates. more precisely. x2 + y 2 .2: Geometry behind the “distance” between two complex numbers. that gives another vector—much less so if we additionally demand our deﬁnition of the product of two complex numbers. On the other hand. jj jjjj jjjj z2 jjjWW kWWW WWWWW WW z1 j z1 − z2 jjjjj 4D Figure 1. all of them diﬀer by a multiple of 2π. say. This means. naturally. The absolute value (sometimes also called the modulus) of x + iy is r = x + iy = and an argument of x + iy is a number φ such that x = r cos φ and y = r sin φ . the positive real axis. . x1 + iy1 with absolute value r1 and argument φ1 .
.. .. ... . . ..MMM ..CHAPTER 1. .... .. ... . ....... . ......... F . φ1 + φ2..2 .... we are multiplying the lengths of the two vectors representing our two complex numbers.. and adding their angles measured with respect to the positive xaxis.... COMPLEX NUMBERS 4 x1 + iy1 = (r1 cos φ1 ) + i(r1 sin φ1 ) and x2 + iy2 = (r2 cos φ2 ) + i(r2 sin φ2 ) To compute the product.. .. . . . .. .. So the absolute value of the product is r1 r2 and (one of) its argument is φ1 + φ2 .. we make use of some classic trigonometric identities: (x1 + iy1 )(x2 + iy2 ) = (r1 cos φ1 ) + i(r1 sin φ1 ) (r2 cos φ2 ) + i(r2 sin φ2 ) = (r1 r2 cos φ1 cos φ2 − r1 r2 sin φ1 sin φ2 ) + i(r1 r2 cos φ1 sin φ2 + r1 r2 sin φ1 cos φ2 ) = r1 r2 (cos φ1 cos φ2 − sin φ1 sin φ2 ) + i(cos φ1 sin φ2 + sin φ1 cos φ2 ) = r1 r2 cos(φ1 + φ2 ) + i sin(φ1 + φ2 ) . Geometrically.. . φ2 . z1 .. ... z . . .. 2 fM. .
. φ . .. .. MMM .
.
. . M. . . 1 .. . .
.
z1 z2 xrrr Figure 1. . . ink.2. Hudson River Undergraduate Mathematics Conference 2000) 2 . We will later see that it has an intimate connection to the complex exponential function. this exponential notation is indeed purely a notation. φ1 . as both cosine and sine are periodic functions with period 2π. (a) eiφ1 eiφ2 = ei(φ1 +φ2 ) (b) 1/eiφ = e−iφ (c) ei(φ+2π) = eiφ (d) eiφ = 1 One should convince oneself that there is no problem with the fact that there are many possible arguments for complex numbers. . To save space. φ2 ∈ R. . . rr . The reader is encouraged to prove them. rrrrr . Lemma 1. we motivate this maybe strangeseeming deﬁnition by collecting some of its properties.. bytes. (and because “Mathematics is for lazy people”3 ) we introduce a shortcut notation and deﬁne eiφ = cos φ + i sin φ . rr . . For any φ.. r . In view of the above calculation. it should come as no surprise that we will have to deal with quantities of the form cos φ + i sin φ (where φ is some real number) quite a bit. r ..rrr . 3 Peter Hilton (Invited address. For now. etc. .3: Multiplication of complex numbers. At this point.
. Their easy proofs are left for the exercises. Geometrically. we get the inequalities −z ≤ Re z ≤ z and − z ≤ Im z ≤ z . the righthand side the polar form of this complex number. the sentence “The complex number x+iy has absolute value r and argument φ” now becomes the identity x + iy = reiφ . From very basic geometric properties of triangles. (1.12) The square of the absolute value has the nice property x + iy2 = x2 + y 2 = (x + iy)(x − iy) . COMPLEX NUMBERS (e) d dφ 5 eiφ = i eiφ . conjugating z means reﬂecting the vector corresponding to z with respect to the real axis. z1 . Lemma 1. The following collects some basic properties of the conjugate. The lefthand side is often called the rectangular form. With this notation. We denote the conjugate by x + iy = x − iy .CHAPTER 1. This is one of many reasons to give the process of passing from x + iy to x − iy a special name: x − iy is called the (complex) conjugate of x + iy. (a) z1 ± z2 = z1 ± z2 (b) z1 · z2 = z1 · z2 (c) z1 z2 = z1 z2 (d) z = z (e) z = z (f) z2 = zz (g) Re z = (h) Im z = 1 2 (z + z) (z − z) 1 2i (i) eiφ = e−iφ .3. For any z. z2 ∈ C. From part (f) we have a neat formula for the inverse of a nonzero complex number: z −1 = z 1 = 2. z z A famous geometric inequality (which holds for vectors in Rn ) is the triangle inequality z1 + z2  ≤ z1  + z2  .
4. and is written Dr (a). 1. So if we ﬁx a complex number a and a positive real number r then the set of z satisfying z − a = r is the set of points at distance r from a. . n n (c) The triangle inequality for sums: k=1 zk ≤ k=1 zk . The reverse triangle inequality is proved in Exercise 15. which were initially deﬁned algebraically. The ﬁrst inequality is just a rewrite of the original triangle inequality. w ∈ C. Recall that if z. z2 . We need some terminology for talking about subsets of C. we have the following identities: (a) The triangle inequality: ±z1 ± z2  ≤ z1  + z2 . can be identiﬁed with the points in the Euclidean plane R2 . In this section we collect some deﬁnitions and results concerning the topology of the plane. For z1 . using the fact that ±z = z. we make extensive use of Lemma 1. For future reference we list several variants of the triangle inequality: Lemma 1. COMPLEX NUMBERS 6 By drawing a picture in the complex plane. Notice that this does not include the circle itself.3 Elementary Topology of the Plane In Section 1. · · · ∈ C. this is the circle with center a and radius r. we will need the following theorems only at a limited number of places in the remainder of the book.2 we saw that the complex numbers C. then z − w is the distance between z and w as points in the plane. Dr (a) = {z ∈ C : z − a < r}. Finally by (1.CHAPTER 1. and the last follows by induction. To prove it algebraically. which is equivalent to our claim. The inside of this circle is called the open disk with center a and radius r.12) z1 + z2 2 ≤ z1 2 + 2 z1 z2  + z2 2 = z1 2 + 2 z1  z2  + z2 2 = z1 2 + 2 z1  z2  + z2 2 = (z1  + z2 )2 . you should be able to come up with a geometric proof of this inequality. the reader who is willing to accept the topological arguments in later proofs on faith may skip the theorems in this section. (b) The reverse triangle inequality: ±z1 ± z2  ≥ z1  − z2 .3: z1 + z2 2 = (z1 + z2 ) (z1 + z2 ) = (z1 + z2 ) (z1 + z2 ) = z 1 z1 + z1 z 2 + z 2 z1 + z2 z 2 = z1 2 + z1 z2 + z1 z2 + z2 2 = z1 2 + 2 Re (z1 z2 ) + z2 2 . While the deﬁnitions are essential and will be used frequently. that is. That is.
For example. so there is little reason to discuss the matter. is not connected. since we have to rule out any possible separation. They are also closed! Deﬁnition 1. Example 1.” In the reals a set is connected if and only if it is an interval. 7 (b) A point b is a boundary point of E if every open disk centered at b contains a point in E and also a point that is not in E. {z ∈ C : z − z0  < R} and {z ∈ C : z − z0  > R} are open. 2] \ {1}.4. Example 1. If G is the open disk {z ∈ C : z − z0  < R} then G = {z ∈ C : z − z0  ≤ R} That is. A set is open if all its points are interior points. However. it is hard to use the deﬁnition to show that a set is connected. G is a closed disk and ∂G is a circle. The interior of E is the set of all interior points of E. The boundary of a set E.1. Suppose E is any subset of C. Y ⊆ C are separated if there are disjoint open sets A and B so that X ⊆ A and Y ⊆ B. Intuitively. COMPLEX NUMBERS Deﬁnition 1. and ∂G = {z ∈ C : z − z0  = R} .3. so a deﬁnition is necessary. 1) and Y = (1. The idea is that if you don’t move too far from an interior point of E then you remain in E.1. (a) A point a is an interior point of E if some open disk with center a lies in E. C and the empty set ∅ are open. .2. the intervals X = [0. Deﬁnition 1. one choice is A = D1 (0) (the open disk with center 0 and radius 1) and B = D1 (2) (the open disk with center 2 and radius 1). {z ∈ C : z − z0  ≤ R} is closed. (d) A point d is an isolated point of E if it lies in E and some open disk centered at d contains no point of E other than d. One notion that is somewhat subtle in the complex domain is the idea of connectedness. is the set of all boundary points of E. a set is connected if it is “in one piece. but at a boundary point you can make an arbitrarily small move and get to a point inside E and you can also make an arbitrarily small move and get to a point outside E. Hence their union. A set is closed if it contains all its boundary points. On the other hand. One type of connected set that we will use frequently is a curve. Two sets X. in the plane there is a vast variety of connected subsets.” It is usually easy to check that a set is not connected. which is [0. Example 1. Deﬁnition 1. A set W ⊆ C is connected if it is impossible to ﬁnd two separated nonempty sets whose union is equal to W . written E. The closure of E.CHAPTER 1. A region is a connected open set. (c) A point c is an accumulation point of E if every open disk centered at c contains a point of E diﬀerent from c.2. 2] on the real axis are separated: There are inﬁnitely many choices for A and B that work. The idea of separation is that the two open sets A and B ensure that X and Y cannot just “stick together. is the set of points in E together with all boundary points of E. written ∂E.3. For R > 0 and z0 ∈ C.
. b]. b] is a closed interval in R. The following seems intuitively clear. If G is not connected then we can write it as a union of two nonempty separated subsets X and Y .5.” which is a connected set S ⊂ C that contains points that cannot be connected by a curve of any sort inside S.6 is not generally true if G is not open. their union is γ. you need three segments to connect −1 to 1 since you cannot go through 0. For example. if G is a connected open subset of C then any two points of G may be connected by a curve in G. and that the parametrization must be deﬁned and continuous on a closed and bounded interval [a. z1 . But this means that γ is not connected. Then any two points in G can be connected by a chain of at most 2 segments in G. we can connect any two points of G by a chain of horizontal and vertical segments lying in G. that is. this is the punctured disk obtained by removing the center from G.5.6. A more extreme example. Then Xγ = X ∩ γ and Yγ = Y ∩ γ are disjoint and nonempty. x(t) and y(t). It is included to illustrate some common techniques in dealing with connected sets.5. The reader may skip the following proof. For example. Theorem 1. ﬁrst. A curve is closed if γ(a) = γ(b) and is a simple closed curve if γ(s) = γ(t) implies s = a and t = b or s = b and t = a. . Now let G0 = G \ {0}. for each k. Proof of Theorem 1. We emphasize that a curve must have a parametrization. zn so that. We say that the curve is parametrized by γ. Any curve is connected. . . Suppose.6. The next theorem gives an easy way to check whether an open set is connected. so it determines is a curve. and setting γ(t) = x(t) + y(t)i. that any two points of G may be connected by a path that lies in G. It is a customary and practical abuse of notation to use the same letter for the curve and its parametrization. circles are connected but there is no way to connect two distinct points of a circle by a chain of segments which are subsets of the circle. On the other hand. A path or curve in C is the image of a continuous function γ : [a. zk and zk+1 are the endpoints of a horizontal or vertical segment which lies entirely in G. If W is a subset of C that has the property that any two points in W can be connected by a curve in W then W is connected. the curve does not cross itself. Let γ be a path in G that connects a to b. COMPLEX NUMBERS 8 Deﬁnition 1.CHAPTER 1. and they are separated by A and B. is the “topologist’s sine curve. a path can be speciﬁed by giving two continuous realvalued functions of a real variable. but now you may need more than two segments to connect points. A chain of segments in G means the following: there are points z0 . The path γ is smooth if γ is diﬀerentiable. so G is connected.) As an example. in fact. Then G is open and it is connected. but its proof requires more preparation in topology: Proposition 1. where [a. So there are disjoint open sets A and B so that X ⊆ A and Y ⊆ B. Warning: The second part of Theorem 1. Since X and Y are disjoint we can ﬁnd a ∈ X and b ∈ G. and also gives a very useful property of open connected sets. (It is not hard to parametrize such a chain. Since we may regard C as identiﬁed with R2 . and this contradicts Proposition 1. let G be the open disk with center 0 and radius 2. . discussed in topology texts. b] → C.
b a f (x) dx (b) If F is any antiderivative of f (that is. Now G is the disjoint union of the two open sets A and B. For functions of several variables we can perform diﬀerentiation operations. Hence z0 cannot connect to any point of D by a chain of segments in G. That is. so z0 can be connected to any point of G by a sequence of segments in G. We can connect z0 to any point z in D by following a chain of segments from z0 to a.4 Theorems from Calculus Here are a few theorems from real calculus that we will make use of in the course of the text. Then (a) If F is deﬁned by F (x) = x a f (t) dt then F is diﬀerentiable and F (x) = f (x). Theorem 1. Then there is 0 < a < 1 such that f (x + ∆x) − f (x) = f (x + a∆x) . 2 2 Theorem 1. y) dx dy are equal. Since a ∈ G there is an open disk D with center a that is contained in G. Since b ∈ G there is an open disk D centered at b that lies in G. Suppose I ⊆ R is an interval. and then adding at most two segments in D that connect a to z.8 (MeanValue Theorem). If f is continuous on the rectangle given by a ≤ b d d b x ≤ b and c ≤ y ≤ d then the iterated integrals a c f (x. The most important of all calculus theorems combines diﬀerentiation and integration (in two ways): Theorem 1. y0 ). this ﬁnishes the proof.7 (ExtremeValue Theorem). and so B must be empty. Theorem 1. Suppose a is in A.9 (Fundamental Theorem of Calculus). and this is impossible. ∆x Many of the most important results of analysis concern combinations of limit operations. F = f ) then = F (b) − F (a). which is impossible. Choose a point z0 ∈ G and deﬁne two sets: A is the set of all points a so that there is a chain of segments in G connecting z0 to a. x + ∆x ∈ I. If the mixed partials ∂x∂y and ∂y∂x are deﬁned on an open set G and are continuous at a point (x0 . so we have shown that A is open. So we have shown that B is open. If these are both nonempty then they form a separation of G.11 (Equality of iterated integrals). y0 ) in G then they are equal at (x0 . y) dy dx and c a f (x. and B is the set of points in G that are not in A.10 (Equality of mixed partials). f : I → R is diﬀerentiable. extending this chain by at most two more segments. b] → R is continuous. and x. But z0 is in A so A is not empty. we could connect z0 to b. If z0 could be connected to any point in D by a chain of segments in G then. we can apply diﬀerentiation and integration with respect to diﬀerent variables in either order: . each point of D is in A. or integration operations. Now suppose b is in B.CHAPTER 1. so D ⊆ B. in any order. 1. Finally. Since z0 could be any point in G. G = A. Suppose f : [a. That is. COMPLEX NUMBERS 9 Now suppose that G is a connected open set. if we have suﬃcient continuity: ∂ f ∂ f Theorem 1. Any continuous realvalued function deﬁned on a closed and bounded subset of Rn has a minimum value and a maximum value.
and suppose the partial derivative ∂f exists and is continuous on R.CHAPTER 1. (c) −3 + √ 3i. y) dy = c c ∂f (x.dcs. (c) z 6 = −9. Then ∂x d dx d d f (x.ac. (c) √3−i . ∂x Exercises 1. Write in polar form: (a) 2i. . (b) 34 eiπ/2 . Write in rectangular form: √ (a) 2 ei3π/4 . For more information about Leibnitz.html. Find all solutions to the following equations: (a) z 6 = 1. 2. 6. Show that 4 Named after Gottfried Wilhelm Leibniz (1646–1716). Find the real and imaginary parts of each of the following: (a) (b) (c) z−a z+a 3+5i 7i+1 . 5. Find the absolute value and conjugate of each of the following: (a) −2 + i. (c) −ei250π . y) dy . 3. (b) z 4 = −16. 4. (d) z 6 − z 3 − 2 = 0. 2+3i (d) (1 + i)6 . (b) (2 + i)(4 + 3i).uk/∼history/Biographies/Leibnitz.stand. COMPLEX NUMBERS 10 Theorem 1. 2 (d) in for any n ∈ Z. (b) 1 + i. see http://wwwgroups. (a ∈ R).12 (Leibniz’s4 Rule). Suppose f is continuous on the rectangle R given by a ≤ x ≤ b and c ≤ y ≤ d. √ 3 −1+i 3 .
Prove the reverse triangle inequality z1 − z2  ≥ z1  − z2 . 13. 18. (c) 0 < z − 1 < 2. 15. COMPLEX NUMBERS (a) z is a real number if and only if z = z. 14. (b) sin 3θ = 3 cos2 θ sin θ − sin3 θ. Sketch the following sets in the complex plane: (a) {z ∈ C : z − 1 + i = 2} . 16. 11.1. (e) z − 1 + z + 1 < 3. Sketch the following sets and determine whether they are open. 17. or neither.2 to derive the triple angle formulas: (a) cos 3θ = cos3 θ − 3 cos θ sin2 θ. Show that if z1 z2 = 0 then z1 = 0 or z2 = 0. (a) z + 3 < 2. (b) z is either real or purely imaginary if and only if (z)2 = z 2 . Find all solutions of the equation z 2 + 2z + (1 − i) = 0. Prove Lemma 1. (d) {z ∈ C : z − i + z + i = 3} . closed. 7. Prove that (a) p(z) = p (z). Suppose p is a polynomial with real coeﬃcients.CHAPTER 1. or z = 1 or z = 2. (c) {z ∈ C : Re(z + 2 − 2i) = 3} . (d) z − 1 + z + 1 = 2. 12. (b) Im z < 1. Use Lemma 1. (b) p(z) = 0 if and only if p (z) = 0. or z < 1. 9. bounded.3. The set E is the set of points z in C satisfying either z is real and −2 < z < −1. connected. Prove Theorem 1. What are the boundaries of the sets in the previous exercise? 19. 8. Prove Lemma 1. . Use the previous exercise to show that 1 z 2 −1 11 ≤ 1 3 for every z on the circle z = 2eiθ . (b) {z ∈ C : z − 1 + i ≤ 2} . 10.2.
12 20. and then diﬀerentiate using the Fundamental Theorem of Calculus. COMPLEX NUMBERS (a) Sketch the set E. d . Find the maximum number of horizontal and vertical segments in G needed to connect two points of G. Describe them.CHAPTER 1. (c) Determine the boundary points of E. and in each case say brieﬂy why the subsets are separated. y) − f (a. interchange the order of ∂x integrations. 21. (d) Determine the isolated points of E. Prove Leibniz’s Rule: Deﬁne F (x) = c f (x. being careful to indicate exactly the points that are in E. (b) Determine the interior points of E. y) dy. This is a connected open set. The set E in the previous exercise can be written in three diﬀerent ways as the union of two disjoint nonempty separated subsets. y) as the integral of ∂f . get an expression for F (x) − F (a) as an iterated integral by writing f (x. 22. Let G be the annulus determined by the conditions 2 < z < 3.
Maybe the fundamental principle of analysis is that of a limit.1 First Steps A (complex) function f is a mapping from a subset G ⊆ C to C (in this situation we will write f : G → C and call G the domain of f ). it may be found hard to realise the great initial diﬃculty of making a little step which now seems so natural and obvious. functions from Rm to Rn . we can construct many familiar looking functions from the standard calculus repertoire. but for sake of simplicity we only state it for those functions. or f (r. whereas for the latter we have to exclude the origin z = 0. The former three could be deﬁned on all of C. f (x. On the other hand. and it may not be surprising if such a step has been found and lost again. such 1 as f (z) = z (the identity map). Suppose f is a complex function with domain G and z0 is an accumulation point of G. and perhaps does not appreciate the diﬃculty of the original ascent. called the image of z and usually denoted by f (z). however. Then w0 is the limit of f as z approaches z0 .Chapter 2 Diﬀerentiation Mathematical study and research are very suggestive of mountaineering. Just as in the real case. or f (z) = z .1. we could construct some functions which make use of a certain representation of z. z→z0 This deﬁnition is the same as is found in most calculus texts. Deﬁnition 2. the deﬁnition does not require 13 . f (z) = 2z + i. Suppose there is a complex number w0 such that for every > 0. y) = y 2 − ix. In fact. y) = x − 2iy. The reason we require that z0 is an accumulation point of the domain is just that we need to be sure that there are points z of the domain which are arbitrarily close to z0 . any tourist can be hauled up for a small cost. for example. Louis Joel Mordell (1888–1972) 2. This means that each element z ∈ G gets mapped to exactly one complex number. in short lim f (z) = w0 . Whymper made several eﬀorts before he climbed the Matterhorn in the 1860’s and even then it cost the life of four of his party. φ) = 2rei(φ+π) . Now. say. So in mathematics. we can ﬁnd δ > 0 so that for all z ∈ G satisfying 0 < z − z0  < δ we have f (z) − w0  < . f (z) = z 3 . So far there is nothing that makes complex functions any more special than. f (x. The philosophy of the following deﬁnition is not restricted to complex functions.
f is continuous on G ⊆ C if f is continuous at every z ∈ G. More generally. and suppose z0 is an accumulation point of G0 .CHAPTER 2. That is why we require 0 < z − z0 .2. and then z iy −iy = lim = lim = −1 . Suppose G0 ⊆ G. and hence x z x = lim = lim = 1 .1 ¯ then implies that limz→0 z does not exist. Lemma 2. we write z = iy where y ∈ R. Let f and g be complex functions and c. we try to compute this “limit” as z → 0 on the real and on the imaginary axis. Just as in the real case the limit w0 is unique if it exists. If z0 is in the domain of the function and either z0 is an isolated point of the domain or z→z0 lim f (z) = f (z0 ) then f is continuous at z0 .2. the following “usual” limit rules are valid for complex functions. If f0 is the restriction of f to G0 then limz→z0 f0 (z) exists and has the value w0 . Because the deﬁnition of the limit is somewhat elaborate. z→z0 z→z0 z→z0 In the last identity we have to make sure we do not divide by zero. In the ﬁrst case. Suppose f is a complex function. DIFFERENTIATION 14 that z0 is in the domain of f and. lim does not exist.1. x→0 x x→0 x z→0 z lim In the second case. z→0 z y→0 iy y→0 iy lim So we get a diﬀerent “limit” depending on the direction from which we approach 0. the following fundamental deﬁnition looks almost trivial. Suppose limz→z0 f (z) exists and has the value w0 . this is illustrated by the following example. The following is a easy consequence of the deﬁnition. Deﬁnition 2. if z0 is in the domain of f .1. (a) lim f (z) + c lim g(z) = lim (f (z) + c g(z)) z→z0 z→z0 z→z0 (b) lim f (z) · lim g(z) = lim (f (z) · g(z)) z→z0 z→z0 z→z0 (c) lim f (z)/ lim g(z) = lim (f (z)/g(z)) . the proofs of these rules are everything but trivial and make for nice exercises. The deﬁnition of limit in the complex domain has to be treated with a little more care than its real companion. z ¯ Example 2. the deﬁnition explicitly ignores the value of f (z0 ). Lemma 2. It is often useful to investigate limits by restricting the way the point z “approaches” z0 . z→0 z To see this. we can write z = x ∈ R. . as above. z0 ∈ C. Lemma 2. z On the other hand.
” This diﬀerence becomes apparent most baldly when studying derivatives.3. The fact that the notions of diﬀerentiability and analyticity are actually diﬀerent is seen in the following examples. Example 2. The function f (z) = z 3 is entire. we have an additional dimension to play with.. Functions which are diﬀerentiable (and hence analytic) in the whole complex plane C are called entire. In other words. reiφ 2 . one reason is that when trying to compute a limit of a function as.. Then 2 z0 + reiφ − z0 2 z0 + re−iφ z0 2 z 2 − z0 2 z0 2 + 2z0 re−iφ + r2 e−2iφ − z0 2 = = = z − z0 z0 + reiφ − z0 reiφ reiφ 2z0 re−iφ + r2 e−2iφ = = 2z0 e−2iφ + re−3iφ . If f is continuous at w0 and limz→z0 g(z) = w0 then limz→z0 f (g(z)) = f (w0 ). f (z0 ) = lim z→z0 lim 3 2 f (z) − f (z0 ) z 3 − z0 (z 2 + zz0 + z0 )(z − z0 ) 2 2 = lim = lim = lim z 2 + zz0 + z0 = 3z0 . This means that the statement “A complex function has a limit.3. say. f (z0 ) = lim z→z0 The diﬀerence quotient limit which deﬁnes f (z0 ) can be rewritten as f (z0 + h) − f (z0 ) .2 Diﬀerentiability and Analyticity ¯ The fact that limits such as limz→0 z do not exist points to something special about complex z numbers which has no parallel in the reals—we can express a function in a very compact way in one variable. Note that h is not a real number but can rather approach zero from anywhere in the complex plane. z→z0 lim f (g(z)) = f z→z0 lim g(z) . Deﬁnition 2.. The function f is analytic on the open set G ⊆ C if it is diﬀerentiable (and hence analytic) at every point in G.CHAPTER 2. The derivative of f at z0 is deﬁned as f (z) − f (z0 ) . z → 0. we can “take the limit inside” a continuous function: 15 Lemma 2. we have to allow z to approach the point 0 in any way. f is called diﬀerentiable at z0 .” is in many senses stronger than the statement “A real function has a limit. Suppose f : G → C is a complex function and z0 is an interior point of G. z − z0 provided this limit exists. analytic in C: For any z0 ∈ C.3.” We will repeatedly notice this kind of behavior. 2. h→0 h This equivalent deﬁnition is sometimes easier to handle. On the real line there are only two directions to approach 0—from the left or from the right (or some combination of those two). that is.2. yet it shows some peculiar behavior “in the limit. If f is diﬀerentiable for all points in an open disk centered at z0 then f is called analytic at z0 . In the complex plane. In this case. DIFFERENTIATION Just as in the real case. f is not analytic at 0): Let’s write z = z0 + reiφ . z→z0 z − z0 z→z0 z→z0 z − z0 z − z0 Example 2. The function f (z) = z 2 is diﬀerentiable at 0 and nowhere else (in particular..
4. Hence lim z2 = lim ze−3iφ = lim z = 0 . The basic properties for derivatives are similar to those we know from real calculus. In the third identity we have to be aware of division by zero. f : G → H is a bijection. that for inverse functions. g : H → G is the inverse function of f . if z0 = 0 then the righthand side equals re−3iφ = ze−3iφ . (a) f (z) + c g(z) = f (z) + c g (z) (b) f (z) · g(z) = f (z)g(z) + f (z)g (z) (c) f (z)/g(z) = (d) z n = nz n−1 f (z)g(z) − f (z)g (z) g(z)2 (e) h(g(z)) = h (g(z))g (z) .5. and z0 ∈ H. If f is diﬀerentiable at g(z0 ). A function f : G → H is onetoone if for every image w ∈ H there is a unique z ∈ G such that f (z) = w. there exists a z ∈ G such that f (z) = w). as discussed earlier. this rule is only deﬁned for functions which are bijections. In fact. z→0 z→0 z z→0 which implies that z2 = 0.4. for example. f (g(z0 )) . A bijection is a function which is both onetoone and onto. and g is continuous at z0 then g is diﬀerentiable at z0 with g (z0 ) = 1 . one should convince oneself that the following rules follow mostly from properties of the limit. z→0 z Example 2. f (g(z0 )) = 0.) t On the other hand. Suppose f and g are diﬀerentiable at z ∈ C. The function f (z) = z is nowhere diﬀerentiable: lim lim z − z0 z − z0 z = lim = lim z→z0 z − z0 z→0 z z − z0 z→z0 does not exist. z(t) = z0 + 1 eit . The function is onto if every w ∈ H has a preimage z ∈ G (that is. n ∈ Z. Lemma 2. and that c ∈ C.CHAPTER 2.) Lemma 2. Suppose G and H are open sets in C. As in the real case. We end this section with yet another diﬀerentiation rule. If f : G → H is a bijection then g is the inverse of f if for all z ∈ H. and h is diﬀerentiable at g(z). (A more entertaining way to see this is to use. f (g(z)) = z. which approaches z0 as t → ∞. DIFFERENTIATION 16 If z0 = 0 then the limit of the righthand side as z → z0 does not exist since r → 0 and we get diﬀerent answers for horizontal approach (φ = 0) and for vertical approach (φ = π/2). (The ‘chain rule’ needs a little care to be worked out.
CHAPTER 2. we will later show that if f is analytic at z0 = x0 + iy0 then u and v have continuous partials (of any order) at z0 .html.uk/∼history/Biographies/Riemann.3 to evaluate this last limit as 1 1 = .1) then f is diﬀerentiable at z0 . DIFFERENTIATION Proof.dcs.dcs.1) equivalently as the following pair of equations: ux (x0 . y0 ) = vy (x0 . Then fx = ux + ivx and −ify = −i(uy + ivy ) = vy − iuy . (a) and (b) are not quite converse statements. That is. and often convenient.6. y0 ) = −vx (x0 . we write f (z) = f (x. Using this terminology we can rewrite the equation (2.stand. y0 ) uy (x0 .html. to write the function f in terms of its real and imaginary parts. If these partial derivatives satisfy (2. It is traditional. y) + iv(x. 2 For more information about Riemann. This appears when we calculate g (z0 ): z→z0 lim g(z) − g(z0 ) g(z) − g(z0 ) 1 1 = lim = lim = lim . As stated.2) 2. y) = u(x. The partial diﬀerential equations (2. However. The function F deﬁned by f (w) − f (w0 ) w − w0 F (z) = f (w0 ) if w = w0 . z→z0 f (g(z)) − f (g(z0 )) z→z0 f (g(z)) − f (g(z0 )) z→z0 F (g(z)) z − z0 g(z) − g(z0 ) Now apply Lemma 2. (2. named after Augustin Louis Cauchy (1789–1857)1 and Georg Friedrich Bernhard Riemann (1826–1866)2 . y) where u is the real part of f and v is the imaginary part.ac. see http://wwwgroups.1) ∂x ∂y (b) Suppose f is a complex function such that the partial derivatives fx and fy exist in an open disk centered at z0 and are continuous at z0 .3 The Cauchy–Riemann Equations Theorem 2. ∂x Remarks.ac. later For more information about Cauchy. 3. F (g(z0 )) f (g(z0 )) 2. y0 ) .uk/∼history/Biographies/Cauchy. Then the partial derivatives of f satisfy ∂f ∂f (z0 ) = −i (z0 ) . (a) Suppose f is diﬀerentiable at z0 = x0 + iy0 . if w = w0 17 is continuous at w0 . In both cases (a) and (b). That is.2) are called the Cauchy–Riemann equations. see http://wwwgroups. 1 . (2. 1. f is given by f (z0 ) = ∂f (z0 ) .stand.
y0 ) . ∆z ∆z→0 As we saw in the last section we must get the same result if we restrict ∆z to be on the real axis and if we restrict it to be on the imaginary axis. DIFFERENTIATION 18 we will prove that f = u + iv is analytic in an open set G if and only if u and v have continuous partials that satisfy (2. y0 ). y0 + ∆y) − f (x0 . y0 ) ∂f = lim = (x0 . y0 ) − f (x0 . Thus we have shown that f (z0 ) = fx (z0 ) = −ify (z0 ). In the ﬁrst case we have ∆z = ∆x and f (z0 ) = lim f (z0 + ∆x) − f (z0 ) f (x0 + ∆x. uxx (x0 . “all we need to do” is prove that f (z0 ) = fx (z0 ). y0 ) ∆y→0 i i∆y ∆y ∂y i∆y→0 (using 1 = −i). writing ∆z = ∆x + i∆y: f (z0 + ∆z) − f (z0 ) f (z0 + ∆z) − f (z0 + ∆x) + f (z0 + ∆x) − f (z0 ) = ∆z ∆z f (z0 + ∆x + i∆y) − f (z0 + ∆x) f (z0 + ∆x) − f (z0 ) = + ∆z ∆z ∆y f (z0 + ∆x + i∆y) − f (z0 + ∆x) ∆x f (z0 + ∆x) − f (z0 ) = · + · .2) in G. y0 ) then f (z0 ) = lim f (z0 + ∆z) − f (z0 ) . Functions with continuous second partials satisfying this partial diﬀerential equation are called harmonic. (a) If f is diﬀerentiable at z0 = (x0 . We ﬁrst rearrange a diﬀerence quotient for f (z0 ). y0 ) = vxy (x0 . as we will see later. if f is analytic in an open set G then the partials of any order of u and v exist.6. If u and v satisfy (2. y0 ) ∂f = lim = −i (x0 . that is. we will study such functions in Chapter 6. ∆x→0 ∆x→0 ∆x ∆x ∂x In the second case we have ∆z = i∆y and f (z0 ) = lim f (z0 + i∆y) − f (z0 ) 1 f (x0 . y0 ) = 0 and an analogous identity for v. ∆z ∆z . 4. hence we will show that the real and imaginary part of a function which is analytic on an open set are harmonic on that set. y0 ) = −uyy (x0 . ∆z ∆y ∆z ∆x Now we rearrange fx (z0 ): fx (z0 ) = i∆y + ∆x ∆y ∆x ∆z · fx (z0 ) = · fx (z0 ) = · ifx (z0 ) + · fx (z0 ) ∆z ∆z ∆z ∆z ∆y ∆x = · fy (z0 ) + · fx (z0 ) .2) and their second partials are also continuous then we obtain uxx (x0 . i (b) To prove the statement in (b). Proof of Theorem 2. y0 ) + uyy (x0 . Again. y0 ) = vyx (x0 .CHAPTER 2. assuming the Cauchy–Riemann equations and continuity of the partials.
4 Constants and Connectivity One of the ﬁrst applications of the meanvalue theorem in real calculus is to show that if a function has zero derivative everywhere on an interval then it must be constant. Now we subtract our two rearrangements and take a limit: f (z0 + ∆z) − f (z0 ) − fx (z0 ) ∆z→0 ∆z ∆y f (z0 + ∆x + i∆y) − f (z0 + ∆x) = lim − fy (z0 ) ∆z→0 ∆z ∆y ∆x f (z0 + ∆x) − f (z0 ) − fx (z0 ) . For the ﬁrst term in (2. y0 + a∆y) ∆y v(x0 + ∆x.3) We need to show that these limits are both 0. The fractions ∆x/∆z and ∆y/∆z are bounded by 1 in modulus so we just need to see that the limits of the expressions in parentheses are 0. we have f (z0 + ∆x + i∆y) − f (z0 + ∆x) − fy (z0 ) ∆y = uy (x0 + ∆x. so f is a constant. 2. The proof is very easy: The meanvalue theorem for a real function says f (x + ∆x) − f (x) = f (x + a∆x)∆x where 0 < a < 1. and both change as ∆z → 0. the two diﬀerences in parentheses have zero limit as ∆z → 0 because uy and vy are continuous at (x0 .3) we apply Theorem 1. ∆y Using these expressions. y0 ). y0 )) . to the real and imaginary parts of f . + lim ∆z→0 ∆z ∆x lim (2. b < 1. If we know that f is always zero then we know that f (x + a∆x) = 0. so we need another argument to prove that functions with derivative that are always 0 must be .1) in the last step to convert ifx to i(−ify ) = fy .CHAPTER 2. However the meanvalue theorem does not have a simple analog for complex valued functions. y0 + a∆y) + ivy (x0 + ∆x. y0 + ∆y) − u(x0 + ∆x. y0 )) = (uy (x0 + ∆x. y0 + ∆y) − v(x0 + ∆x.3) has a limit of 0 since. y0 + b∆y) − (uy (x0 . The second term in (2. with 0 < a.8. by deﬁnition. y0 + a∆y) − vy (x0 . This says that all values of f must be the same. y0 + b∆y) . This gives us real numbers a and b. so f (x + ∆x) = f (x). Finally. fx (z0 ) = lim f (z0 + ∆x) − f (z0 ) ∆x ∆x→0 and taking the limit as ∆z → 0 is the same as taking the limit as ∆x → 0. so that u(x0 + ∆x. DIFFERENTIATION 19 where we used equation (2. y0 ) = vy (x0 + ∆x. We can’t do this for the ﬁrst expression since both ∆x and ∆y are involved. the real meanvalue theorem. y0 ) = uy (x0 + ∆x. y0 )) + i (vy (x0 + ∆x. y0 ) + ivy (x0 . y0 + a∆y) − uy (x0 .
2. but it illustrates an important fact about complex functions. z+i z→i 3 (b) z→1−i lim x + i(2x + y). then f (z) = 0 for all z in the domain of f but f is not constant. Now we can prove the theorem using these two facts: Fix a starting point z0 in G and let b = f (z0 ). For example. There are a number of surprising applications of this theorem. and consider the real part u(z) for z on V . u(z) is constant on this horizontal segment. Use the deﬁnition of limit to show that limz→z0 (az + b) = az0 + b. Now connect z1 to a point z2 by a vertical segment V in G. Evaluate the following limits or explain why they don’t exist. and in each case we will have to allow for examples like the above. then f is constant on V so f (z2 ) = f (z1 ) = b. Proof. suppose that V is a vertical segment that is contained in G. Since both the real and imaginary parts of f are constant on H. DIFFERENTIATION 20 constant. this isn’t really true.6 says that this is always possible. suppose that H is a horizontal line segment in G. Next. then f is constant on H so f (z1 ) = f (z0 ) = b. If the domain of f is a region G ⊆ C and f (z) = 0 for all z in G then f is a constant. if Re z < 0.7.2. f itself is constant on H. (a) lim iz −1 . Now connect z2 to a point z3 by a horizontal segment and conclude that f (z3 ) = b. since vx (z) = Im(f (z)) = 0. We can argue the same way to see that the imaginary part v(z) of f (z) is constant on H. . uy (z) = −vx (z) = − Im(f (z)) = 0. if the domain of f consists of all complex numbers with nonzero real part and f (z) = 1 −1 if Re z > 0. Thus u is constant on V . by the real version of the theorem. But ux (z) = Re(f (z)) = 0 so. As above. and similarly v is constant on V . see Exercises 13 and 14 for a start. In many cases during the course we will want to conclude that a function is constant. This may seem like a silly example.CHAPTER 2. so we can consider u(z) to be just a function of x. Since H is a horizontal segment. Exercises 1. Repeating this argument we see that f (z) = b for all points that can be connected to z0 in this way by a ﬁnite sequence of horizontal and vertical segments. 3. and in fact the correct theorem is: Theorem 2. we can consider u(z) to be just a function of y and. Consider the real part u(z) for z ∈ H. Prove Lemma 2. The fundamental problem is that the domain in this example is not connected. Connect z0 to a point z1 by a horizontal segment H in G. so f is constant on V . In fact. using the Cauchy– Riemann equations. First. Theorem 1. y is constant on H.
Show that f is constant. provide a counterexample. then f is constant in G.6 (b)? 13. 11. Prove Lemma 2. Suppose f (z) is entire. Prove Lemma 2. Apply the deﬁnition of the derivative to give a direct proof that f (z) = − z12 when f (z) = z . . y) is continuous (resp. yet f is not diﬀerentiable at the origin. 15. if z = 0. DIFFERENTIATION 4. (g) f (z) = z2 = x2 + y 2 . (j) f (z) = 4(Re z)(Im z) − i(z)2 . (Hint: Use the Cauchy–Riemann equations to show that f = 0. 21 1 5. (l) f (z) = z 2 − z 2 . (b) f (z) = 2x + ixy 2 . Prove: If f is analytic in the region G ⊆ C and always real valued. 10. Where are the following functions diﬀerentiable? Where are they analytic? Determine their derivatives at points where they are diﬀerentiable. (e) f (z) = cos x cosh y − i sin x sinh y. Show that if f is diﬀerentiable at z then f is continuous at z. 8. y) + iv(x. (k) f (z) = 2xy − i(x + y)2 . (As always. Prove Lemma 2. z = x + iy. (d) f (z) = ex e−iy .4. with real and imaginary parts u(z) and v(z) satisfying u(z)v(z) = 3 for all z.) 14.6. 7.2 by using the formula for f given in Theorem 2. (a) f (z) = e−x e−iy . 9.3. (f) f (z) = Im z. Prove: If f (z) and f (z) are both analytic in the region G ⊆ C then f (z) is constant in G. If u(x. What can you say if f (z) is given by a polynomial in x = Re z and y = Im z? 12. diﬀerentiable)? If not.) Show that f satisﬁes the Cauchy–Riemann equations at the origin z = 0. y) and v(x. (i) f (z) = ix+1 y . Prove that if f (z) is given by a polynomial in z then f is entire.CHAPTER 2. 6. y) are continuous (respectively diﬀerentiable) does it follow that f (z) = u(x. Consider the function xy(x + iy) f (z) = x2 + y 2 0 if z = 0. (h) f (z) = z Im z. Why doesn’t this contradict Theorem 2. (c) f (z) = x2 + iy 2 .
. Give a formula for A in terms of the constants a. The general real homogeneous quadratic function of (x. (a) Show that u is harmonic if and only if a = −c.CHAPTER 2. Is x x2 +y 2 22 x2 ? x2 +y 2 harmonic? What about 17. y) is u(x. b and c. DIFFERENTIATION 16. b and c are real constants. (b) If u is harmonic then show that it is the real part of a function of the form f (z) = Az 2 . where A is a complex constant. y) = ax2 + bxy + cy 2 . where a.
see o o http://wwwgroups. d ∈ C. o Exercise 11 of the previous chapter states that any polynomial (in z) is an entire function. Suppose f (z1 ) = f (z2 ). in which case f is entire). T. For more information about M¨bius. A linear fractional transformation is a function of the form f (z) = az + b .1 M¨bius Transformations o The ﬁrst class of functions that we will discuss in some detail are built from linear polynomials. o o Proof. M¨bius transformations are bijections. 1 Named after August Ferdinand M¨bius (1790–1868).uk/∼history/Biographies/Mobius. b. In fact. that is. which is quite special for complex functions. if f (z) = az+b then the inverse o cz+d function of f is given by dz − b f −1 (z) = . From this fact we can conclude that a linear fractional transformation f (z) = az+b is analytic in cz+d C \ − d (unless c = 0.stand.1.ac. Deﬁnition 3.dcs. c c az1 + b az2 + b = . −cz + a Remark.1.Chapter 3 Examples of Functions Obvious is the most dangerous word in mathematics. c One property of M¨bius transformations. If ad − bc = 0 then f is called a M¨bius1 transformation. Notice that the inverse of a M¨bius transformation is another M¨bius transformation. cz + d where a. c. Bell 3. 23 . Note that f : C \ {− d } → C \ { a }. Lemma 3.html. E. is the o following. cz1 + d cz2 + d This is equivalent (unless the denominators are zero) to (az1 + b)(cz2 + d) = (az2 + b)(cz1 + d) .
Proof. let α = a + bi. The formula for f −1 : C \ { a } → C \ {− d } can be checked easily. o Theorem 3. Hence our standard equation for a line becomes αz + αz = 2c. we can tackle the promised theorem about the following geometric property of M¨bius transformations. and ino 1 versions f (z) = z . Proposition 3. Aside from being prime examples of onetoone functions. which implies that c c f is onto. M¨bius transformations map circles and lines into circles and lines. or Re(αz) = c. we can modify its deﬁning equation z − z0  = r as follows: z − z0 2 = r2 (z − z0 )(z − z0 ) = r2 zz − z0 z − zz0 + z0 z0 = r2 z2 − z0 z − zz0 + z0 2 − r2 = 0 .2. EXAMPLES OF FUNCTIONS which can be rearranged to (ad − bc)(z1 − z2 ) = 0 . we only have to prove the theorem for the inversion f (z) = z . Starting with ax + by = c (where z = x + iy). Simplify. M¨bius transformations possess faso cinating geometric properties. which means that f is onetoone.3. . f −1 is onetoone.1) First case: Given a circle centered at z0 with radius r. o Proof. Special cases of M¨bius transformations are translations f (z) = z + b. Before going on we ﬁnd a standard form for the equation of a straight line. dilations. Suppose f (z) = az+b cz+d is a linear fractional transformation. c if c = 0 then f (z) = bc − ad 1 c2 z + d c + In particular. En route to an example of such. The next result says that if we understand those three special transformations. Translations and dilations certainly map circles and lines into circles and lines. and inversions. Just like f . With the last result at hand. Then αz = ax + by + i(ay − bx) so αz + αz = αz + αz = 2 Re(αz) = 2ax + 2by. (3. we understand them all. dilations f (z) = az. so by the 1 last proposition. If c = 0 then f (z) = a b z+ . every linear fractional transformation is a composition of translations. we introduce some terminology. d d a . 24 Since ad − bc = 0 this implies that z1 = z2 .CHAPTER 3.
z0 2 .) Now if r happens to be equal to z0 2 then this equation becomes 1 − z0 w − z0 w = 0.1) in terms of w. Otherwise we can divide by 2c: ww − w− z0 2c z0 z0 w− w=0 2c 2c z0 z0 2 w− =0 − 2c 4c2 w− This is the equation of a circle with center z0 2c z0 2c 2 = z0  2c .1) with α = z0 .1) and rewrite it in terms of 1 w. This is the equation of a circle in terms of w. If c = 0. We get z0 w + z0 w = 2cww . If we solve w = 1 z we get z = w . it is much more generally useful: . Second case: We start with the equation of a line in the form (3. We obtain z0 z0 1 w− = 0. 2 z0  − r2 1 z0 2 z0 2 − r2 r2 s = w0  − = − = . so we have a straight line in terms of w. z0 2 − r2 (z0 2 − r2 )2 (z0 2 − r2 )2 (z0 2 − r2 )2 2 2 Then we can rewrite our equation as w2 − w0 w − w0 w + w0 2 − s2 = 0 ww − w0 w − ww0 + w0 w0 = s2 (w − w0 )(w − w0 ) = s2 w − w0 2 = s2 .CHAPTER 3. w2 − w+ 2 2 2 2 2 z0  − r z0  − r z0  − r2 We deﬁne z0 w0 = . which is of the form (3. 4c2 and radius There is one fact about M¨bius transformations that is very helpful to understanding their o geometry. by substituting z = w and simplifying. this describes a line in the form (3. where w = z . EXAMPLES OF FUNCTIONS 1 Now we want to transform this into an equation in terms of w. so we make this substitution in our equation: 1 z 25 for 1 w 2 − z0 1 1 − z0 + z0 2 − r2 = 0 w w 1 − z0 w − z0 w + w2 z0 2 − r2 = 0 . (To get the second line we multiply by w2 = ww and simplify. as above. In fact. with center w0 and radius s. Otherwise z0 2 − r2 is nonzero so we can divide our equation by it.
parameterized as f (zk (t)). In this new set we deﬁne algebraic rules for dealing with inﬁnity based on the usual laws of limits. is the tangent vector to γk at the point a. but −∞ = ∞ in the complex sense.2 Inﬁnity and the Cross Ratio Inﬁnity is not a number—this is true whether we use the complex numbers or stay in the reals.4. If we diﬀerentiate f (zk (t)) at t = 0 and use the chain rule we see that the tangent vector to the transformed curve at the point f (a) is f (a)zk (0). Suppose f is analytic at a with f (a) = 0 and suppose γ1 and γ2 are two smooth curves which pass through a. Suppose G is a set of complex numbers and f is a function from G to C. In the ﬁrst deﬁnition we require that z0 is an accumulation point of G while in the second and third we require that ∞ is an “extended accumulation point” of G. and the transformed curves make an angle of θ with each other. For example. For k = 1. Then f transforms γ1 and γ2 into smooth curves which meet at f (a). In the complex sense there is only one inﬁnity. C = C ∪ {∞}. we have f (z) − L < . (a) lim f (z) = ∞ means that for every M > 0 we can ﬁnd δ > 0 so that. as zk (t) = xk (t) + iyk (t). We summarize these rules: . The complex number zk (0). Since f (a) = 0 the transformation from z1 (0) and z2 (0) to f (a)z1 (0) and f (a)z2 (0) is a dilation. This leads to the addition rule ∞ + a = ∞. and complex numbers have inﬁnite limits if they can become larger in magnitude than any preassigned limit. However. for all z ∈ G satisfying z > N . (c) lim f (z) = ∞ means that for every M > 0 we can ﬁnd N > 0 so that. for all z ∈ G satisfying z→∞ z > N we have f (z) > M .CHAPTER 3. For completeness we repeat the usual deﬁnitions: Deﬁnition 3. for many purposes we can work with inﬁnity in the complexes much more naturally and simply than in the reals. in the sense that for every B > 0 there is some z ∈ G with z > B. We do this by deﬁning a new set. so that zk (0) = a.2. EXAMPLES OF FUNCTIONS 26 Lemma 3. for all z ∈ G satisfying z→z0 0 < z − z0  < δ. In order to deal correctly with inﬁnity we have to realize that we are always talking about a limit. if lim f (z) = ∞ and lim g(z) = a is ﬁnite then the usual “limit of sum = sum of limits” rule gives z→z0 z→z0 z→z0 lim (f (z) + g(z)) = ∞. Then f transforms the curve γk to the curve f (γk ). an analytic function with nonzero derivative preserves angles. Functions which preserve angles in this way are also called conformal. written ∞. In fact. In the real sense there is also a “negative inﬁnity”. making an angle of θ with each other. 2 we write γk parametrically. considered as a vector. Proof. The usual rules for working with inﬁnite limits are still valid in the complex numbers. (b) lim f (z) = L means that for every z→∞ > 0 we can ﬁnd N > 0 so that. 3. it is a good idea to make inﬁnity an honorary complex number so that we can more easily ˆ manipulate inﬁnite limits. In brief. A dilation is the composition of a scale change and a rotation and both of these preserve the angles between vectors. we have f (z) > M .
Recall that f (z) = z transforms circles that pass through the origin to straight lines. and a dilation f (z) = az (with a = 0) is also deﬁned for z = ∞. For example. The general o case is summarized below: (c) Lemma 3. dilations and the inversion f (z) = z we see that ˆ ˆ every M¨bius transformation may be interpreted as a transformation of C onto C.5. i to ∞. 27 a a = 0 and = ∞ if a = 0. if we know where three distinct points in C are transformed by a M¨bius transformation then we should be able to ﬁgure out everything about the o transformation. There is a computational device that makes this easier to see. but the individual limits of z and −z are both ∞. 1 With this interpretation in mind we can add some insight to Theorem 3. 1 Now we reconsider M¨bius transformations with inﬁnity in mind. . EXAMPLES OF FUNCTIONS Deﬁnition 3. For example. If we remember that ∞ corresponds to being arbitrarily far away from the origin we can visualize a line plus inﬁnity as a circle passing through ∞. ∞ 0 If a calculation involving inﬁnity is not covered by the rules above then we must investigate the limit more carefully. but if we take the limit of z + (−z) = 0 as z → ∞ we will get 0. ∞ and i also determine a circle—the imaginary axis plus the point at inﬁnity. with f (0) = ∞ and f (∞) = 0. otherwise. Suppose a ∈ C. If the three points are on a straight line or if one of the points is ∞ then ˆ the circle is a straight line plus ∞. For example. so the proper domain for ˆ f (z) is actually C. f (∞) = a c and f − d c = ∞. A translation o f (z) = z + b is now deﬁned for z = ∞. If c = 0 then f (∞) = ∞.CHAPTER 3. but the point z = 0 must be excluded from the circle. with f (∞) = a · ∞ = ∞. now we can put it back. Let f be the M¨bius transformation o f (z) = az + b .3.3 can be expressed very simply: any M¨bius transformation of C o transforms circles to circles. Conversely. cz + d ˆ Then f is deﬁned for all z ∈ C. If we make ˆ this a deﬁnition then Theorem 3. Let’s consider the other basic types of M¨bius transformations. Hence f transforms the unit circle onto the imaginary axis plus the point at inﬁnity. However. and 1 to i. with f (∞) = ∞ + b = ∞. i and 1 determine a circle—the unit circle z = 1—and the three image points 0. the transformation f (z) = z+i z−i transforms −i to 0. and.3. it may seem strange that ∞ + ∞ is not deﬁned. ˆ This example relied on the idea that three distinct points in C determine uniquely a circle passing through them. (a) ∞ + a = a + ∞ = ∞ (b) ∞ · a = a · ∞ = ∞ · ∞ = ∞ if a = 0. so f transforms circles that pass through the origin to straight lines plus ∞. Since every M¨bius transformation o 1 can be expressed as a composition of translations. f (z) = z is o now deﬁned for z = 0 and z = ∞. The three points −i.
z1 . Everything should be clear except the ﬁnal uniqueness statement. h(1) = 1 and o h(∞) = ∞. c+d 0+d d 0 = h(0) = so h(z) = az+b = az+0 = a z = z. the reader might think we should have simply deﬁned the . it is often easier to determine h directly from the conditions f (zk ) = wk (by solving for a. z2 . z2 and z3 as above then g(z) = o f (z) for all z.3 Exponential and Trigonometric Functions To deﬁne the complex exponential function. What if we have three points z1 . Let h = g −1 ◦ f where f (z) = [z. o and h(z3 ) = w3 . w1 . Suppose z1 . z1 . the composition h = g ◦ f −1 o is a M¨bius transformation. Moreover. and—in addition—ﬁnally make sense of the notation eit = cos t + i sin t. z2 . w1 . z2 and z3 and we want to map them to three other points. or z3 is inﬁnity. Lemma 3. Our preferred way to do this is through a P power series: ex = k≥0 xk /k!. and z3 are any four points in C with z1 . 1 and ∞ by a M¨bius transformation then o the crossratio gives us the only way to do it.7.4. then their crossratio is deﬁned by (z − z1 )(z2 − z3 ) [z. then the two terms on the right containing it are canceled. But since h(z) = z for all z we have h(f (z)) = f (z) and so cz+d 0+d d g(z) = g ◦ (f −1 ◦ f )(z) = (g ◦ f −1 ) ◦ f (z) = h(f (z)) = f (z). z3 ] then f is a M¨bius transformation which o satisﬁes f (z1 ) = 0. z1 . z1 . z3 ] = . Similarly. If f is deﬁned by f (z) = [z. z2 .6. if g is any M¨bius transformation which transforms z1 . Notice that h(0) = g(f −1 (0)) = g(z1 ) = 0.1 the inverse f −1 is a M¨bius transformation and. z3 ] and g(w) = [w. c and d). w3 ]. w2 . z2 . we once more borrow concepts from calculus. 3. Proof.CHAPTER 3. In light of this deﬁnition. the result is inﬁnity. z2 . b.6. This theorem gives an explicit way to determine h from the points zj and wj but. By Lemma 3. f (z2 ) = 1. w2 and w3 are distinct ˆ Then there is a unique M¨bius transformation h satisfying h(z1 ) = w1 . namely the real exponential function2 and the real sine and cosine. EXAMPLES OF FUNCTIONS 28 ˆ Deﬁnition 3. z2 . by Exercise 7 in this chapter. z2 and z3 are distinct points in C and w1 . w2 and w3 ? ˆ Theorem 3. Proof. z1 . If we write h(z) = az+b then cz+d b =⇒ b = 0 d a ∞ = h(∞) = =⇒ c = 0 c a+b a+0 a 1 = h(1) = = = =⇒ a = d . If z. 2 It is a nontrivial question how to deﬁne the real exponential function. f (z3 ) = ∞ . (z − z3 )(z2 − z1 ) Here if z = z3 . and if one of z. h(z2 ) = w2 points in C. and z3 distinct. Uniqueness follows as in the proof of Lemma 3. ˆ So if we want to map three given points of C to 0. in practice.
5 is reasonable. z2 ∈ C. Lemma 3. one of the promises of these lecture notes is to introduce complex power series as late as possible. cosine. Its ﬁrst justiﬁcation is that all exponential rules which we are used to from real numbers carry over to the complex case. one that may not seem too pleasant at ﬁrst sight is the fact that the complex exponential function is not onetoone.8. In fact. EXAMPLES OF FUNCTIONS Deﬁnition 3. The third identity is a very special one and has no counterpart for the real exponential function.6. cotangent.CHAPTER 3. We agree with those readers who think that we are “cheating” at this point. In fact. The tangent and cotangent are deﬁned as sin z = tan z = respectively. For all z. 29 This deﬁnition seems a bit arbitrary. as we borrow the concept of a (real) power series to deﬁne the real exponential function. z1 . etc. but we invite the reader to meditate on its proof. sin z exp(2iz) − 1 = −i cos z exp(2iz) + 1 and cot z = cos z exp(2iz) + 1 =i . 2i 2 respectively. complex exponential function through a complex power series. The trigonometric functions—sine. note that the last identity also says that exp is entire. we can deﬁne them as merely being special combinations of the exponential function. one can get another strong reason why Deﬁnition 3. It says that the complex exponential function is periodic with period 2πi. The (complex) sine and cosine are deﬁned as 1 1 (exp(iz) − exp(−iz)) and cos z = (exp(iz) + exp(−iz)) . this is possible (and an elegant deﬁnition).2 and are collected in the following. The (complex) exponential function is deﬁned for z = x + iy as exp(z) = ex (cos y + i sin y) = ex eiy . Remarks. (a) exp (z1 ) exp (z2 ) = exp (z1 + z2 ) (b) 1 exp(z) = exp (−z) (c) exp (z + 2πi) = exp (z) (d) exp (z) = exp (Re z) (e) exp(z) = 0 (f) d dz exp (z) = exp (z) . they don’t play the same prominent role as in the real case. however.5. however. 1. 2. When proving this identity through the Cauchy–Riemann equations for the exponential function. Finally. Deﬁnition 3.—have their complex analogues. sin z exp(2iz) − 1 . The last identity is not only remarkable. tangent. They mainly follow from Lemma 1. This has many interesting consequences. We should make sure that the complex exponential function specializes to the real exponential function for real arguments: if z = x ∈ R then exp(x) = ex (cos 0 + i sin 0) = ex . to say the least.
EXAMPLES OF FUNCTIONS MM MM MM MM MM MM MM MM M q1 qq 11 qq 11 q qq 11 qq q 11 qq qq 11 qq 11 11 11 / 30 5π 6 π 3 exp 0 −π 3 − 5π 6 −1 0 1 2 Figure 3. For all z. Note that to write tangent and cotangent in terms of the exponential function. the one on the right the real sine. 2i 2i (The ‘sin’ on the left denotes the complex sine. the complex sine and cosine are not bounded—consider. the following properties follow mostly from Lemma 3. we used the fact that exp(z) exp(−z) = exp(0) = 1. sinh z exp(2z) − 1 cosh z = . Finally. and cotangent are deﬁned as in the real case: 1 (exp(z) − exp(−z)) 2 sinh z exp(2z) − 1 tanh z = = cosh z exp(2z) + 1 sinh z = 1 (exp(z) + exp(−z)) 2 cosh z exp(2z) + 1 coth z = = . for example. z2 ∈ C. so are sin and cos.8.) A similar calculation holds for the cosine. Because exp is entire. The hyperbolic sine.9. cosine.1: Image properties of the exponential function. As with the exponential function. tangent. Not too surprising. We end this section with a remark on hyperbolic trig functions. Lemma 3. sin(iy) as y → ±∞. we should ﬁrst make sure that we’re not redeﬁning the real sine and cosine: if z = x ∈ R then sin x = 1 1 (exp(ix) − exp(−ix)) = (cos x + i sin x − (cos(−x) + i sin(−x))) = sin x . sin(−z) = − sin z sin(z + 2π) = sin z tan(z + π) = tan z sin(z + π/2) = cos z sin (z1 + z2 ) = sin z1 cos z2 + cos z1 sin z2 cos z + sin z = 1 sin z = cos z 2 2 cos(−z) = cos z cos(z + 2π) = cos z cot(z + π) = cot z cos(z + π/2) = − sin z cos (z1 + z2 ) = cos z1 cos z2 − sin z1 sin z2 cos2 z − sin2 z = cos(2z) cos z = − sin z . one word of caution: unlike in the real case.CHAPTER 3. z1 .
and eiv = eiφ ⇐⇒ v = φ+2πk for some k ∈ Z. eu = r = z ⇐⇒ u = ln z (where ln denotes the real natural logarithm. not a function. π] (the principal argument of z). and hence we can’t even consider it to be our soughtafter inverse of the exponential function. On the other hand. The problem is that we need to stick to this convention. A reasonable deﬁnition of a logarithm function Log would hence be to set Log z = ln z + i Arg z where Arg z gives the argument for the complex number z according to some convention—for example.4 The Logarithm and Complex Exponentials The complex logarithm is the ﬁrst function we’ll encounter that is of a somewhat tricky nature. we could just use a diﬀerent argument convention and get another reasonable ‘logarithm.7. as we saw. We invite the reader to check this thoroughly. we get something that’s not a function. they are not only yet more special combinations of the exponential function.CHAPTER 3. log is. π]. or in [0.’ Even worse. . we could agree that the argument is always in (−π. in particular we need to demand that z = 0). that is. as usual. this is too much to hope for. Then for the ﬁrst equation to hold. and suppose that Log z = u(z) + iv(z). of course. 2π). in particular. we need exp(Log z) = eu eiv = r eiφ = z . by deﬁning the multivalued map arg z = {φ : φ is a possible argument of z} and deﬁning the multivalued logarithm as log z = ln z + i arg z . yet it satisﬁes exp(log z) = z . z = r eiφ . one should note how the periodicity of the exponential function takes care of the multivaluedness of our ‘logarithm’ log. Let Arg z denote that argument of z which is in (−π. but they are also related with the trigonometric functions via sinh(iz) = i sin z and cosh(iz) = cos z . that is. etc. Then the principal logarithm is deﬁned as Log z = ln z + i Arg z . It is motivated as being the inverse function to the exponential function. we’re looking for a function Log such that exp(Log z) = z = Log(exp z) . Any function Log : C \ {0} → C which satisﬁes exp(Log z) = z is a branch of the logarithm. As we will see shortly. Deﬁnition 3. Let’s write. EXAMPLES OF FUNCTIONS 31 As such. 3. Let’s try to make things well deﬁned.
The righthand side is equal to z = x + iy only if y ∈ (−π. π].html. We apply Lemma 2. Suppose Log maps C \ {0} to G (this is typically a halfopen strip.dcs. The idea is to apply Lemma 2. the reason for this is once more the periodicity of the exponential function. the natural deﬁnition ab = exp(b log a) (which is a concept borrowed from calculus) would in general yield more than one value (Exercise 31). then Log(exp z) = Log ex eiy = ln ex eiy + i Arg ex eiy = ln ex + i Arg eiy = x + i Arg eiy .ac. The same happens with any other branch of the logarithm Log—there will always be some (in fact. but we need to be careful about the domains of these functions. For two complex numbers a and b. see http://wwwgroups. f (z) = exp(z) and g : C \ {0} → G.5 to exp and Log. for example. Even better. Theorem 3. through a power series) and the function f (x) = ex where e is Euler’s3 number and 1 n can be deﬁned. In calculus one proves the equivalence of the real exponential function (as given. With our deﬁnition of ab . Suppose Log is a branch of the logarithm. for example. we obtain ez = exp(z Log e) = exp (z (ln e + i Arg e)) = exp (z ln e) = exp (z) . z Proof.stand. g(z) = Log: if Log is continuous at z then 1 1 1 Log z = = = . you might want to think about what it looks like if Log = Log). we can now make a similar remark about the complex exponential function.10. one just has to be cautious about where each logarithm is analytic. EXAMPLES OF FUNCTIONS 32 The paragraph preceding this deﬁnition ensures that the principal logarithm is indeed a branch of the logarithm. So what about the other equation Log(exp z) = z? Let’s try the principal logarithm: Suppose z = x + iy. the evaluation of any branch of the logarithm at z can only diﬀer from Log z by a multiple of 2πi. A word of caution: this only works out this nicely because we carefully deﬁned ab for complex numbers. we prove that any branch of the logarithm has the same derivative.uk/∼history/Biographies/Euler. For more information about Euler.CHAPTER 3.5 with f : G → C \ {0} . Because e is a positive real number and hence Arg e = 0. To end our discussion of the logarithm on a happy note. many) yvalues for which Log(exp z) = z. We turn instead to the principal logarithm and deﬁne the principal value of ab as ab = exp(b Log a) . exp (Log z) exp(Log z) z We ﬁnish this section by deﬁning complex exponentials. so that we get actual inverse functions. so it is not always useful. Then Log is diﬀerentiable wherever it is continuous and 1 Log z = . 3 . Diﬀerent deﬁnitions might lead to diﬀerent outcomes of ez versus exp z! Named after Leonard Euler (1707–1783). as e = limn→∞ 1 + n . A note about e.
ﬁnd the image of the center of the circle. remember that M¨bius transformations preserve angles. (e) The circle with radius 1 centered at 1. Suppose that f is analytic on the region G and f (G) is a subset of the unit circle. Let C be the circle with center 0 and radius 1. Draw two graphs. 1+z 1−z maps the unit circle (minus the point z = 1) 6. Using the crossratio. Label the sets. ±2. plus ∞. Let C be the circle with center 1 + i and radius 1. Find M¨bius transformations satisfying each of the following. Let f (z) = z+2 .) 2z 8. 1 → 1. as az+b . cz+d (a) 1 → 0. Prove Proposition 3. one showing the following six sets in the z plane and the other showing their images in the w plane. Suppose A = a b is a 2 × 2 matrix of complex numbers whose determinant ad − bc is c d nonzero. ∞ → −i. with diﬀerent choices of zk .) o (a) The xaxis.) 7. In each case. 11. Write your answers in standard o form. plus ∞. (Here ◦ denotes composition and · denotes matrix multiplication. 2 → 1. 9. . 10. Show that f is constant.2. ∞ and −1 − i.) (b) 1 → 0. 1 + i → 1. Then we can deﬁne a corresponding M¨bius transformation TA by TA (z) = az+b . o 3. (d) The circle with radius 2 centered at 0. plus ∞. 3 → ∞. Show that the derivative of a M¨bius transformation is never zero. o cz+d Show that TA ◦TB = TA·B . (Use the crossratio.) (c) 0 → i. (f) The circle with radius 1 centered at −1. Prove that any M¨bius transformation diﬀerent from the identity map can have at most two o ﬁxed points. (You should only need to calculate the images of 0. (Hint: Consider the function 1+f (z) and use Exercise 5 and a variation of 1−f (z) Exercise 13 in Chapter 2. (Use the crossratio. Find a M¨bius transformation which transforms o 1 C onto C and transforms 0 to 2 . Show that the M¨bius transformation f (z) = o onto the imaginary axis. Show that if f (z) = az+b cz+d is a M¨bius transformation then f −1 (z) = o dz−b −cz+a . 5.CHAPTER 3.) 4. 2. (A ﬁxed point of a function f is a number z such that f (z) = z. 2 → ∞. EXAMPLES OF FUNCTIONS 33 Exercises 1. ﬁnd two diﬀerent M¨bius transformations that transform C onto the real axis plus o inﬁnity. (b) The yaxis. (c) The line x = y.
Prove Lemma 3. y > 0 under w = (c) The strip 0 < x < 1 under w = ∂u ∂x ∂v ∂x ∂u ∂y ∂v ∂y z z−1 .8. 13. y). 18. −i → −1. yaxis to y = −x. Let z = x + iy and show that (a) sin z2 = sin2 x + sinh2 y = cosh2 y − cos2 x. Find the M¨bius transformation f : o (a) f maps 0 → 1. 0 ≤ y ≤ 2π. Show that if f = u + iv is analytic then the Jacobian equals f (z)2 . and the unit circle to itself. −1 → i. 21. z3 ] is real or inﬁnite. 1 → ∞. z1 . z 2 −1 2z+1 . ∞ → 0. 34 (b) The quadrant x > 0. Let z = x + iy and show that (a) sin z = sin x cosh y + i cos x sinh y. 23. Describe the image of the region under the transformation: (a) The disk z < 1 under w = iz−i z+1 . 17. ˆ 16. Find the ﬁxed points in C of f (z) = 15. (b) the line segment deﬁned by z = 1 + iy. Show that tan(iz) = i tanh z. (c) the rectangle {z = x + iy ∈ C : 0 ≤ x ≤ 1. z2 and z3 are distinct points in C. (c) f maps xaxis to y = x.9. 0 ≤ y ≤ 2π}. 0 ≤ y ≤ 2π. (c) If cos x = 0 then cot z2 = (d) If y ≥ 1 then cot z2 ≤ 22.CHAPTER 3. v = v(x. Describe the images of the following sets under the exponential function exp(z): (a) the line segment deﬁned by z = iy. y) is the determinant of the matrix . Show that z is on the circle passing through by z1 . (b) f maps 1 → 1. 20. EXAMPLES OF FUNCTIONS 12. 1 sinh2 y sinh2 y+1 sinh2 y =1+ ≤1+ 1 sinh2 1 ≤ 2. . z2 . (b) cos z2 = cos2 x + sinh2 y = cosh2 y − sin2 x. ˆ 14. 19. z−i z+i . Prove Lemma 3. z2 and z3 if and only if [z. Find the principal values of cosh2 y−1 cosh2 y ≤ 1. The Jacobian of a transformation u = u(x. Suppose z1 . (b) cos z = cos x cosh y − i sin x sinh y.
(f) sinh z = 0. 31. (h) z 1/2 = 1 + i. 35 24. Prove that exp(b log a) is singlevalued if and only if b is an integer. Find all solutions to the following equations: (a) Log(z) = π i. use the principal value (unless stated otherwise).) What can you say if b is rational? 32. plot it reasonably carefully. EXAMPLES OF FUNCTIONS (a) log i. Describe the image under exp of the line with equation y = x. Is there a diﬀerence between the set of all values of log z 2 and the set of all values of 2 log z? (Try some ﬁxed numbers for z. 2 (b) Log(z) = 3π 2 i. (Note that this means that complex exponentials don’t clash with monomials z n . f (z) = z 2 . (b) sin z . 29. (e) cos z = 0. (c) exp(z) = πi. (b) (−1)i . Fix c ∈ C \ {0}. 27. Find the image of the annulus 1 < z < e under the principal value of the logarithm. Show that az  = aRe z if a is a positive real constant. (a) z 2 . (f) iz−3 . determine all complex numbers for which the function is analytic. (d) exp(z). . If you run into a logarithm. 33. y = t).CHAPTER 3. 28. For this problem. 30. (e) (z − 3)i . z 3 +1 (c) Log(z − 2i + 1) where Log(z) = ln z + i Arg(z) with 0 ≤ Arg(z) < 2π.) 26. and explain what happens in the limits as t → ∞ and t → −∞. (c) log(1 + i). Is arg(z) = − arg(z) true for the multiplevalued argument? What about Arg(z) = − Arg(z) for the principal branch? 25. Find the derivative of f (z) = z c . For each of the following functions. To do this you should ﬁnd an equation (at least parametrically) for the image (you can start with the parametric form x = t. (d) sin z = cosh 4. (g) exp(iz) = exp(iz).
the circular arc from 2 to 2i. Draw f (Q) and identify the types of image curves corresponding to the segments from 2 to 2 + 2i and from 2 + 2i to 2i.) . and then the vertical segment from 2i to 0.) 34. Let Q be the square with vertices at 0. They are not parts of either straight lines or circles.CHAPTER 3. 2. (b) Show that the image of a ray starting at the origin is a ray starting at the origin. (c) Let T be the ﬁgure formed by the horizontal segment from 0 to 2. (d) Is the right angle at the origin in part (c) preserved? Is something wrong here? (Hint: Use polar coordinates. v) equation for the image curve. As in the previous problem. EXAMPLES OF FUNCTIONS 36 (a) Show that the image of a circle centered at the origin is a circle centered at the origin. Eliminate the parameter in u + iv = f (z(t)) to get a (u. let f (z) = z 2 . Draw T and f (T ). 2 + 2i and 2i. (Hint: You can write the vertical segment parametrically as z(t) = 2 + it.
1 Deﬁnition and Basic Properties At ﬁrst sight. In this case we simply deﬁne c1 c2 b f= γ a f (γ(t))γ (t) dt + c1 f (γ(t))γ (t) dt + · · · + cn f (γ(t))γ (t) dt . we deﬁne b b b φ(t) dt = a a Re φ(t) dt + i a Im φ(t) dt .Chapter 4 Integration Everybody knows that mathematics is about miracles. is only piecewise diﬀerentiable. only mathematicians have a name for them: theorems. a ≤ t ≤ b.1) For a function which takes complex numbers as arguments. cn ]. we integrate over a curve γ (instead of a real interval). the following deﬁnition. b] ⊂ R → C. which is based on (4. Suppose γ is a smooth curve parametrized by γ(t). a ≤ t ≤ b. Then we deﬁne the integral of f on γ as b f= γ γ f (z) dz = a f (γ(t))γ (t) dt . As our ﬁrst example of the application of this deﬁnition we will compute the integral of the function f (z) = z 2 = x2 − y 2 − i(2xy) over several curves from the point z = 0 to the point z = 1 + i. a ≤ t ≤ b. c2 ]. In what follows. complex integration is not really anything diﬀerent from real integration. (4.1. c1 ]. Deﬁnition 4. [cn . we’ll usually state our results for smooth curves. .1) should come as no surprise. . . Suppose this curve is parametrized by γ(t). . say γ(t) is diﬀerentiable on the intervals [a. Roger Howe 4. those curves γ whose parametrization γ(t). bearing in mind that practically all can be extended to piecewise smooth curves. and f is a complex function which is continuous on γ. [cn−1 . b]. [c1 . Example 4. For a continuous complexvalued function φ : [a. 37 . that is. If one meditates about the substitution rule for real integrals. This deﬁnition can be naturally extended to piecewise smooth curves.1.
3 (b) Let γ be the arc of the parabola y = x2 from z = 0 to z = 1 + i. A parametrization of this curve is γ(t) = t + it2 . In (d) the length of a smooth curve γ with parametrization γ(t). most of which follow from their real siblings in a straightforward way. Then −γ f = − γ f .CHAPTER 4. f= γ 0 t2 − t4 − 2it3 (1 + 2it) dt = 0 t2 + 3t4 − 2it5 dt = 1 1 14 i 1 + 3 − 2i = − . Hence 1 1 f= γ γ1 f+ γ2 f= 0 t2 · 1 dt + 0 (1 − it)2 i dt = 1 +i 3 1 1 − 2it − t2 dt 0 = 1 1 1 + i 1 − 2i − 3 2 3 = 4 2 + i. The curve −γ deﬁned in (b) is the curve that we obtain by traveling through γ in the opposite direction. Parameterizations are γ1 (t) = t. is deﬁned as b length(γ) = a γ (t) dt . 3 5 6 15 3 (c) Let γ be the union of the two line segments γ1 from z = 0 to z = 1 and γ2 from z = 1 to z = 1 + i. Then γ1 γ2 f (z) dz = γ1 f (z) dz + γ2 f (z) dz . deﬁne the curve −γ through −γ(t) = γ(a + b − t). (b) If γ is parametrized by γ(t).1. Proposition 4. a ≤ t ≤ b. (a) γ (f + cg) = γ f +c γ g. (d) γ f ≤ maxz∈γ f (z) · length(γ) . 0 ≤ t ≤ 1 and γ2 (t) = 1 + it. 0 ≤ t ≤ 1. and then continuing on γ2 to its end. 0 ≤ t ≤ 1. We have γ (t) = 1 + i and f (γ(t)) = (t − it)2 . a ≤ t ≤ b. 0 ≤ t ≤ 1. Now we have γ (t) = 1 + 2it and f (γ(t)) = t2 − t2 whence 1 1 2 − i 2t · t2 = t2 − t4 − 2it3 . We invite the reader to use some familiar curves to see that this deﬁnition gives what one would expect to be the length of a curve. 3 3 The complex integral has some standard properties. Suppose γ is a smooth curve. and c ∈ C. (c) If γ1 and γ2 are curves so that γ2 starts where γ1 ends then deﬁne the curve γ1 γ2 by following γ1 to its end. and hence 1 f= γ 0 (t − it)2 (1 + i) dt = (1 + i) 0 1 2 t2 − 2it2 − t2 dt = −2i(1 + i)/3 = (1 − i) . f and g are complex functions which are continuous on γ. . INTEGRATION 38 (a) Let γ be the line segment from z = 0 to z = 1 + i. A parametrization of this curve is γ(t) = t + it. a ≤ t ≤ b.
CHAPTER 4. (b) follows with an easy real change of variables s = a + b − t: b b f= −γ a a f (γ(a + b − t)) (γ(a + b − t)) dt = − a b f (γ(a + b − t)) γ (a + b − t) dt f. such that F (z) = f (z) for all z ∈ G. If γ1 has domain [a1 . The fact that γ1 (b1 ) = γ2 (a2 ) is necessary to make sure that this parameterization is piecewise smooth. one easy way to compute integrals is through knowing the antiderivative (or primitive) of the integrand f . Then f (z) dz e−iφ γ b a b b Finally. for b1 ≤ t ≤ b1 + b2 − a2 . f . z∈γ 4. . let φ = Arg f (z) dz = γ γ b f (z) dz e−iφ = Re = Re a f (γ(t))γ (t)e−iφ dt b = a Re f (γ(t))e−iφ γ (t) dt ≤ f (γ(t))e−iφ γ (t) dt = a f (γ(t)) γ (t) dt ≤ max f (γ(t)) a≤t≤b a γ (t) dt = max f (z) · length(γ) . that is. (a) follows directly from the deﬁnition of the integral and the properties of real integrals. we say that f has an antiderivative on G if there exists a function F that is analytic on G. a function F such that F = f . γ = b f (γ(s)) γ (s) ds = − a f (γ(s)) γ (s) ds = − For (c) we need a suitable parameterization γ(t) for γ1 γ2 . INTEGRATION 39 Proof. b2 ] then we can use γ(t) = γ1 (t) γ2 (t − b1 + a2 ) for a1 ≤ t ≤ b1 . to prove (d). Now we break the integral over γ1 γ2 into two pieces and apply the simple change of variables s = t − b1 + a2 : b1 +b2 −a2 b1 b1 +b2 −a2 f (z) dz = γ1 γ2 a1 b1 f (γ(t))γ (t) dt = f (γ1 (t))γ (t) dt + a1 b1 b1 b2 f (γ(t))γ (t) dt + b1 f (γ(t))γ (t) dt a1 b1 +b2 −a2 = = a1 f (γ2 (t − b1 + a2 ))γ2 (t − b1 + a2 ) dt f (γ2 (s))γ2 (s) ds a2 f (γ1 (t))γ (t) dt + f (z) dz + γ1 γ2 γ = f (z) dz.2 Antiderivatives Just like in the real case. b1 ] and γ2 has domain [a2 . To be more precise.
s) = h(1.9 (the Fundamental Theorem of Calculus).CHAPTER 4. γ is a smooth closed curve in G. γ Proof of Theorem 4. 0 ≤ t ≤ 1. 0 ≤ t ≤ 1 is closed. Then f = F (γ(b)) − F (γ(a)) .1. the region {z ∈ C : z < 2}.2. The function h(t. dt and then we calculate b b b f= γ a f (γ(t))γ (t) dt = a F (γ(t))γ (t) dt = a d F (γ(t)) dt = F (γ(b)) − F (γ(a)) . h(0. which is continuously transformed from γ1 to γ2 . a ≤ t ≤ b. the function z 2 does not have an antiderivative in any nonempty region—prove it!) In the special case that γ is closed (that is. and F is a primitive of f on G. γ(a) = γ(b)). It is based on the following concept. s). Suppose G ⊆ C is open. parametrized by γ1 (t). if there is a continuous function h : [0. γ f is independent of the path γ ⊂ G between γ(a) and γ(b).2. The last condition simply says that each of the curves h(t. h(t. 0 ≤ t ≤ 1 and γ2 (t). . γ is a smooth curve in G parametrized by γ(t). we immediately get the following nice consequence. it also says that the function z 2 does not have an antiderivative in. f is continuous on G. 0) = γ1 (t) . Example 4. (Actually. respectively. Suppose γ1 and γ2 are closed curves in the open set G ⊆ C. 1) = γ2 (t) . γ In particular.3 Cauchy’s Theorem We now turn to the central theorem of complex analysis.2. 1]2 → G such that h(t. Then f = 0. Then γ1 is Ghomotopic to γ2 . An application of the chain rule shows d F (γ(t)) = F (γ(t))γ (t) . and f is continuous on G and has an antiderivative on G. Corollary 4. dt by Theorem 1. for example.1 shows that a pathindependent integral is quite special. INTEGRATION 40 Theorem 4. 4. s) is called a homotopy and represents a curve for each ﬁxed s. s) . Deﬁnition 4. Here is the theorem on which most of what will follow is based.3. in symbols γ1 ∼G γ2 . Suppose G ⊆ C is open. An example is depicted in Figure 4.
the homotopies will be ‘nice enough’ to satisfy the condition of this theorem. f is analytic in G. and γ1 ∼G γ2 via a homotopy with continuous second partials.ac. INTEGRATION 41 Figure 4. Corollary 4. then the proof becomes too advanced for the scope of these notes. Suppose G ⊆ C is open. see http://wwwgroups. 1 .2 for an example). however.5.uk/∼history/Biographies/Gauss.ac.stand. Suppose G ⊆ C is open. 3 For more information about Edouard JeanBaptiste Goursat (1858–1936). Then f = 0.uk/∼history/Biographies/Goursat. since Cauchy assumed that the derivative of f was continuous.4 (Cauchy’s Theorem).html. In this case we simply say γ is Gcontractible. 2.dcs.stand. f is analytic in G.CHAPTER 4.ac. Cauchy’s theorem is often called the Cauchy–Goursat Theorem.stand. Theorem 4. 2 For more information about Karl Theodor Wilhelm Weierstraß (1815–1897). a constant curve (see Figure 4. γ For more information about Gauß. 1. The fact that an integral over a point is zero has the following immediate consequence. An important special case is the one where a curve γ is Ghomotopic to a point. and γ ∼G 0 via a homotopy with continuous second partials. The condition on the smoothness of the homotopy can be omitted.dcs. Then f= γ1 γ2 f.dcs. see http://wwwgroups. Weierstraß2 his in 1842. In all the examples and exercises that we’ll have to deal with here. Remarks.1: This square and the circle are (C \ {0})homotopic.uk/∼history/Biographies/Weierstrass. Cauchy published his version in 1825. in symbols γ ∼G 0.html.html. see http://wwwgroups. that is. a condition which was ﬁrst removed by Goursat3 . It is assumed that Johann Carl Friedrich Gauß (1777–1855)1 knew a version of this theorem in 1811 but only published it in 1831.
6. We will show that I is constant with respect to s. The fact that any closed curve is Ccontractible (Exercise 11a) yields the following special case of the previous specialcase corollary. 0 ≤ t ≤ 1. combined with Theorem 1. s) − f (h(0. s)) dt ∂t ∂t 0 0 ∂s 1 1 ∂2h ∂ ∂h ∂h ∂h + f (h(t. ∂s ∂s 1 f (h(t. s)) dt In the last step we used the third property (according to Deﬁnition 4. If f is entire and γ is any smooth closed curve then f = 0. s)) dt = f (h(t. ∂t∂s ∂s∂t 4. s)) = f (h(t.9 (the fundamental theorem of calculus).2: This ellipse is (C \ R)contractible.4 yields almost immediately the following helpful result. INTEGRATION 42 Figure 4. s) = 0 . . Corollary 4. s). and hence the statement of the theorem follows with I(0) = I(1). Consider the integral I(s) = γs f as a function in s (so I(0) = γ1 f and I(1) = γ2 f ). To prove that I is constant. we use the fact that h has continuous second partials and hence ∂2h ∂2h = .4 Cauchy’s Integral Formula Cauchy’s Theorem 4. s)) (0. Note also that in the second line. γ Proof of Theorem 4. Suppose h is the homotopy.4. s)) (1.2) of the homotopy h. s)) ∂s ∂t ∂t∂s ∂t ∂s 0 0 ∂h ∂h = f (h(1.12 (Leibniz’s rule).CHAPTER 4. we use Theorem 1. and γs is the curve parametrized by h(t. d d I(s) = ds ds 1 ∂h ∂ ∂h dt = f (h(t.
simple. All circles Cr with center w and radius r are homotopic in D \ {w}. z∈Cr CR On the righthand side.1(d) f (z) dz − 2πif (w) = z−w 1 f (z) f (z) − f (w) dz − f (w) dz = dz z−w Cr z − w Cr z − w Cr f (z) − f (w) f (z) − f (w) length (Cr ) = max 2πr ≤ max z∈Cr z∈Cr z−w r = 2π max f (z) − f (w) . For example. smooth. In general.CHAPTER 4. z−w γ . and the function f (z)/(z − w) is analytic in an open set containing D \ {w}. and—because f is continuous at w—this means we can make f (z) − f (w) as small as we like. closed. 2πi CR z − w Proof.4. and Theorem 4. gives f (z) dz = z−w f (z) dz z−w CR Cr Now by Exercise 8. This is a useful theorem by itself. in many cases in which a point w is inside a simple closed curve γ we can see a homotopy from γ to a small circle around w so that the homotopy misses w and remains in the region where f is analytic. In that case the theorem remains true. z−w and we obtain with Proposition 4. Then 1 f (z) f (w) = dz . In this discussion we need to be sure that the orientation of the curve γ and the circle match. we say a simple closed curve γ is positively oriented if it is parameterized so that the inside is on the left of γ.7 (Cauchy’s Integral Formula for a Circle).8 (Cauchy’s Integral Formula). Let CR be the counterclockwise circle with radius R centered at w and suppose f is analytic at each point of the closed disk D bounded by CR . Cr 1 dz = 2πi . Then f (w) = 1 2πi f (z) dz . w ∈ G. Here’s the general form: Theorem 4. INTEGRATION 43 Theorem 4. we can now take r as small as we want. and γ is a positively oriented. So Cauchy’s Theorem 4.7 then applies to evaluate the integral. Suppose f is analytic on the region G. For a circle this corresponds to a counterclockwise orientation. it will be important to have Cauchy’s integral formula when w is anywhere inside CR . Gcontractible curve such that w is inside γ. since. In fact. Hence the lefthand side has no choice but to be zero. not just at the center of CR . which is what we claimed. the integral of f (z)/(z − w) around γ is the same as the integral of f (z)/(z − w) around a small circle centered at w. but it can be made more generally useful. by Cauchy’s theorem.
All we need is to ﬁnd a homotopy in G \ {w} between γ and a small circle with center at w. Suppose f is analytic on and inside the circle z = w + reit . if f = u + iv. we have borrowed this from a companion theorem to the Jordan curve theorem which is sometimes called the “annulus theorem. it is not even clear how to make sense of the “inside” of γ in general.10. parametrized by. If γ is a positively oriented. 2. 3.1. Evaluate 1 γ z dz where γ(t) = sin t + i cos t. Integrate the function f (z) = z over the three curves given in Example 4.CHAPTER 4. 0 ≤ t ≤ 2π. for 0 < s < 1. γs is inside γ and outside of D. in fact. simple. 0 ≤ t ≤ 2π. Then f (w) = Furthermore. (c) 1/z 4 .” If you want to explore this kind of mathematics you should take a course in topology.” is surprisingly diﬃcult to prove. (b) z 2 − 2z + 3. we automatically get similar formulas for the real and imaginary part of f . These identities have the ﬂavor of mean values. simply by taking real and imaginary parts on both sides. The justiﬁcation for this is one of the ﬁrst substantial theorems ever proved in topology. This theorem.9 (Jordan Curve Theorem). Theorem 4. which is often called a meanvalue theorem. u(w) = 1 2π 2π 1 2π 2π f w + reit dt . 0 Exercises 1. Let’s summarize them in the following statement. . but it is not at all clear how to do so in complete generality. 0 ≤ t ≤ 2π. We can state it as follows: Theorem 4. by combining Cauchy’s theorem and the special case. Integrate the following functions over the circle z = 2. closed curve in C then C \ γ consists of two connected open sets. oriented counterclockwise: (a) z + z. and. The usual statement of the Jordan curve theorem does not contain the homotopy information. If a closed disk D centered at w lies inside γ then there is a homotopy γs from γ to the positively oriented boundary of D. say. (d) xy. z = w + reit . Corollary 4. although “intuitively obvious. In all practical cases we can see immediately how to construct such a homotopy. the inside and the outside of γ. 0 Even better. 0 u w + reit dt 0 and v(w) = 1 2π 2π v w + reit dt .8 gives (if the conditions are met) f (w) = 1 2πi 2π 0 f w + reit 1 ireit dt = it − w w + re 2π 2π f w + reit dt . Theorem 4. A nice special case of Cauchy’s formula is obtained when γ is a circle centered at w. INTEGRATION 44 We have already indicated how to prove this.7.
8. b] be the map which “takes γ to σ. 6. since otherwise the integral is not deﬁned. f is independent of the parametrization of γ. [If n is negative. Let γ be the circle with radius r centered at w. γ y dz. z−w γ 9. Exercise 12 excluded n = −1 for a very good reason: Exercises 2 and 8 (with w = 0) give counterexamples. σ = γ ◦ τ . Evaluate γ e3z dz for each of the following paths: (a) The straight line segment from 1 to i. Suppose a smooth curve is parametrized by both γ(t). You can parameterize this curve as z(t) = w + reit for 0 ≤ t ≤ 2π. γ z dz and γ z dz along each of the following paths. INTEGRATION 45 4. Show that d b f (σ(t))σ (t) dt = c a γ f (γ(t))γ (t) dt . c ≤ t ≤ d. d] → [a. Note that you can get the second two integrals very easily after you calculate the ﬁrst two.) (In other words. (Hint: Follow the counterclockwise unit circle through m complete cycles γz (for m > 0). Evaluate the integrals γ x dz.CHAPTER 4. Show that dz = 2πi . 1 1 7. (a) Prove that any closed curve is Ccontractible. a ≤ t ≤ b and σ(t). Generalizing these. (b) The circle z = 3. by writing z and z as x ± iy. Show that γ z n dz = 0 for any closed smooth γ and any integer n = −1. our deﬁnition of the integral 10. 0 ≤ t ≤ 1. 12. and remember that the principal branch is √ 1 deﬁned by z 2 = reiθ/2 if z = reiθ for −π ≤ θ ≤ π.” that is. What should you do if m < 0? What if m = 0?) . if m is any integer then ﬁnd a closed curve γ so that −1 dz = 2mπi. (c) The parabola y = x2 from x = 0 to x = 1. 11. Evaluate γ z 2 dz where γ is the parabola with parametric equation γ(t) = t+it2 . 5. (b) Prove that any two closed curves are Chomotopic. assume that γ does not pass through the origin. Evaluate γ z 2 dz where γ is the unit circle and z 2 is the principal branch. and let τ : [c. You can use the parameterization γ(θ) = eiθ for −π ≤ θ ≤ π. Prove that ∼G is an equivalence relation. (c) γ is the counterclockwise circle z − a = r.] 13. Use γ(t) = a + reit . (b) γ is the counterclockwise circle z = 1. (a) γ is the line segment form 0 to 1 − i. oriented counterclockwise.
You should get diﬀerent answers for r < a and r > a. Prove the following integration by parts statement. Then f g = f (γ(b))g(γ(b)) − f (γ(a))g(γ(a)) − γ γ f g. (Hint: In one case γr is contractible in C \ {a}. 21. and γ is a positively oriented. Compute the real integral 2π 0 dθ 2 + sin θ by writing the sine function in terms of the exponential function and making the substitution z = eiθ to turn the real into a complex integral.5 using Theorem 4.8 by evaluating g(0) and g(1). Prove that f (z) = g(z) for all z inside γ. This exercise gives an alternative proof of Cauchy’s integral formula (Theorem 4. Suppose f and g are analytic on the region G. (Hint: ∂f ∂t Use Theorem 1. Show that p = 0. Let f and g be analytic in G. 20. INTEGRATION 14. (a) Consider the function g : [0. Gcontractible curve. Prove Corollary 4.) 15.) .4. z2 + 1 46 γr (This integral depends on r. Suppose a is a complex number and γ0 and γ1 are two counterclockwise circles (traversed just once) so that a is inside both of them. simple. (z + t(w − z)). which does not depend on Cauchy’s Theorem 4.CHAPTER 4. and f (z) = g(z) for all z ∈ γ. Suppose p is a polynomial and γ is a closed smooth path in C. g(t) = γ f (w+t(z−w)) z−w dz.8). and suppose γ ⊂ G is a smooth curve from a to b. γ is a closed. Is 18. γ 16. closed. Show that F (z) = F (z) = arctan z? i 2 Log(z + i) − i 2 Log(z − i) is a primitive of 1 1+z 2 for Re(z) > 0. Gcontractible curve such that w is inside γ. smooth. Let γr be the counterclockwise circle with center at 0 and radius r. Let γr be the circle centered at 2i with radius r. w ∈ G. Compute dz . Show that g = 0. smooth. Suppose f is analytic on the region G. oriented counterclockwise. 1] → C. 17. dz 23. 19.8. Find γr z−a .12 (Leibniz’s rule) and then ﬁnd a primitive for (b) Prove Theorem 4. In the other you can combine Exercises 8 and 22. Explain geometrically why γ0 and γ1 are homotopic in C \ {a} .) 22.
Let γr be the counterclockwise circle with center at 0 and radius r. r = 3 and r = 5. where C is the boundary of the square with corners at ±4 ± 4i: (a) C (b) C (c) C (d) C ez dz. (Hint: Since z 2 − 2z − 8 = (z − 4)(z + 2) you can ﬁnd a partial fraction 1 A B decomposition of the form z 2 −2z−8 = z−4 + z+2 . Use the Cauchy integral formula to evaluate the integral in Exercise 24 when r = 3. Now use Exercise 23. (z − πi)4 sin(2z) dz.) 25. − 2z − 8 (z − 4)(z + 2) z+2 z−4 Which of these forms corresponds to the Cauchy integral formula for the curve γ3 ?) 26. (Hint: The integrand can be written in each of following ways: z2 1 1 1/(z − 4) 1/(z + 2) = = = . z3 ez dz. Compute the following integrals.CHAPTER 4. Find γr z 2 −2z−8 for r = 1. INTEGRATION 47 dz 24. (z − π)4 ez cos(z) dz. (z − π)3 .
Theorem 5.1. Richard Askey 5.8). (z − w)3 This innocentlooking theorem has a very powerful consequence: just from knowing that f is analytic we know of the existence of f . gives the following statement. Then f (w) = and f (w) = 1 2πi 1 πi f (z) dz (z − w)2 γ γ f (z) dz . etc. w ∈ G. which has no analog whatsoever in the reals. which we can rewrite as follows by Theorem 4. We will study the following diﬀerence quotient.1. f is also analytic in G. closed. f . If f is diﬀerentiable in the region G then f is inﬁnitely diﬀerentiable in G. Repeating this argument for f . 1 f (w + ∆w) − f (w) = ∆w ∆w 1 = 2πi 1 2πi γ f (z) 1 dz − z − (w + ∆w) 2πi γ f (z) dz .1 Extensions of Cauchy’s Formula We now derive formulas for f and f which resemble Cauchy’s formula (Theorem 4.8. then for f . Proof of Theorem 5.. Suppose f is analytic on the region G.2.8). simple. and γ is a positively oriented. then you still have work to do. (z − w − ∆w)(z − w) γ f (z) dz z−w 48 .Chapter 5 Consequences of Cauchy’s Theorem If things are nice there is probably a good reason why they are nice: and if you do not know at least one reason for this good fortune. smooth. Gcontractible curve such that w is inside γ. The idea of the proof is very similar to the proof of Cauchy’s integral formula (Theorem 4. that is. Corollary 5.
6). Theorem 5. and theoretically one could obtain them one by one with the methods of the proof of Theorem 5. By the reverse triangle inequality we have for all z ∈ γ f (z) M f (z) ≤ . these new integrals we know . each of which has only one singularity inside the integration path. The eﬀect is that we transformed an integral.1: Introduce an additional path which separates 0 and 1. We give some examples of this application next.1.2. the two contributions along the new path will cancel each other. Example 5. z − w ≥ δ for all z on γ.1 can also be used to compute certain integrals. z=1 sin(z) d dz = 2πi sin(z) 2 z dz = 2πi cos(0) = 2πi . Remarks. − 1) z=2 we ﬁrst split up the integration path as illustrated in Figure 5. This is in fact true. The proof of the formula for f is very similar and will be left for the exercises (see Exercise 1). However. it suﬃces to show that the integrand stays bounded as ∆w → 0 (because γ and hence length(γ) are ﬁxed). CONSEQUENCES OF CAUCHY’S THEOREM Hence we will have to show that the following expression gets arbitrarily small as ∆w → 0: f (w + ∆w) − f (w) 1 − ∆w 2πi f (z) 1 dz = (z − w)2 2πi f (z) f (z) dz − (z − w − ∆w)(z − w) (z − w)2 γ f (z) 1 dz . once we start studying power series for analytic functions. To compute the integral z 2 (z dz . Since γ is a closed set. Theorem 5. by Proposition 4. z=0 Example 5. there is some positive δ so that the open disk of radius δ around w does not intersect γ. 2. we will obtain such a result much more easily. so we save the derivation of formulas for higher derivatives of f for later (see Corollary 8.1(d). In fact.1. that is. If we integrate on these two new closed paths (γ1 and γ2 ) counterclockwise.1 suggests that there are similar looking formulas for the higher derivatives of f . for which two singularities where inside the integration path. into a sum of two integrals. Let M = maxz∈γ f (z). = ∆w 2πi γ (z − w − ∆w)(z − w)2 49 γ This can be made arbitrarily small if we can show that the integral stays bounded as ∆w → 0.CHAPTER 5. ≤ 2 2 (z − w − ∆w)(z − w) (z − w − ∆w)z − w (δ − ∆w)δ 2 which certainly stays bounded as ∆w → 0. 1.
CHAPTER 5. As a preparation we prove the following inequality. which is generally quite useful. but this will be a recurring theme throughout the rest of the book. The ﬁrst application is understanding the roots of polynomials.2 Taking Cauchy’s Formula to the Limit Many beautiful applications of Cauchy’s formula arise from considerations of the limiting behavior of the formula as the curve gets arbitrarily large. We shall look at a few applications along these lines in this section. . Suppose p(z) is a polynomial of degree d with leading coeﬃcient ad . z=0 5. Lemma 5. a polynomial of degree d looks almost like a constant times z d . CONSEQUENCES OF CAUCHY’S THEOREM 50 γ2 1 0 γ1 Figure 5.2 how to deal with. = 2πi Example 5. z=1 cos(z) d2 dz = πi 2 cos(z) z3 dz = πi (− cos(0)) = −πi . z 2 (z dz = − 1) = γ1 z=2 γ1 z 2 (z 1 z−1 z2 dz + − 1) dz + γ2 γ2 z 2 (z dz − 1) 1 z2 z−1 dz d 1 1 + 2πi 2 dz z − 1 z=0 1 1 = 2πi − + 2πi (−1)2 = 0. Then there is real number R0 so that 1 ad  zd ≤ p(z) ≤ 2 ad  zd 2 for all z satisfying z ≥ R0 .1: Example 5.3. It simply says that for large enough z.3.
etc. 1. Notice that the value of the integral does not depend on R. 4 This theorem is for historical reasons erroneously attributed to Liouville. 1 .ac.1 is the following. Gauß may well have known about it before Cauchy’s times. Corollary 5.stand. and we can factor out ad z d : p(z) = ad z d + ad−1 z d−1 + ad−2 z d−2 + · · · + a1 z + a0 = ad  zd 1 + ad−1 ad−2 a1 a0 + ··· + + . The fundamental theorem of algebra was ﬁrst proved by Gauß (in his doctoral dissertation).5 (Liouville’s3 Theorem4 ). CONSEQUENCES OF CAUCHY’S THEOREM Proof. after getting a root a.dcs. so we have 1 dz 1 = lim . although its statement had been assumed to be correct long before Gauß’s times. in fact. p(z) = 0 for all z ∈ C.uk/∼history/Biographies/Liouville. This statement implies that any polynomial p can be factored into linear terms of p(z) the form z − a where a is a root of p. Hence. Every bounded entire function is constant. It was published earlier by Cauchy. where d is the degree of p(z) and ad is the leading coeﬃcient of p(z).1(d) and the formula for the circumference of a circle we see that the integral can be bounded as dz 1 1 2 2 ≤ · · (2πR) = 2πi γR zp(z) 2π ad  Rd+1 ad  Rd and this has limit 0 as R → ∞. But.html. as we can apply the corollary. Proof 2 . (see also Exercise 8). we have shown that impossible. (∗) p(0) R→∞ 2πi γR z p(z) But now we can see that the limit of the integral is 0: By Lemma 5. A compact reformulation of the fundamental theorem of algebra is to say that C is algebraically closed. Since p(z) has degree d its leading coeﬃcient ad is not zero.CHAPTER 5. that is.3 we have z p(z) ≥ 1 ad  zd+1 2 for all large z. Then Cauchy’s formula gives us 1 1 = p(0) 2πi 1/p(z) dz z γR where γR is the circle of radius R around the origin. 3 For more information about Joseph Liouville (1809–1882).4 (Fundamental theorem of algebra1 ). to z−a (which is again a polynomial by the division algorithm). Suppose (by way of contradiction) that p does not have any roots. Every nonconstant polynomial has a root in C. and 2 for all Theorem 5. + 2 d−1 ad z ad z ad z ad z d 1 2 51 Then the sum inside the last factor has limit 1 as z → ∞ so its modulus is between large enough z. plugging into (∗). 2. see http://wwwgroups. using Proposition 4. which is Remarks. A powerful consequence of (the ﬁrst half of) Theorem 5. 1 p(0) = 0.
This implies that f = 0. CONSEQUENCES OF CAUCHY’S THEOREM 52 Proof. Hence p is constant. Note that we can choose any R because f is entire. where g(z) = z2 + 1 z−i z−i z+i Since g(z) is analytic inside and on σ and i is inside σ. Suppose (by way of contradiction) that p does not have any roots. we can parameterize the integral over S using z = t. R f (z) f (z) 1 1 f (z) f (z) dz ≤ 2πR = max max · 2πR = max z∈γ (z − w)2 2π z∈γR (z − w)2 2π z∈γR R2 R γR The righthand side can be made arbitrary small. Then. because p is entire.3. so we can bound the integral over T using Proposition 4. as we are allowed to make R as large as we want. that is.1 with the circle γR of radius R centered at w. by Theorem 2. the func1 tion f (z) = p(z) is entire. Let σ be the counterclockwise semicircle formed by the segment S of the real axis from −R to R. so we can take the limit as R → ∞. which is close to Gauß’s original proof: Another proof of the fundamental theorem of algebra. that is. But f → 0 as z becomes large as a consequence of Lemma 5. First. On the other hand.1(d): dz 2 2 ≤ 2 · πR = z2 + 1 R R T and this has limit 0 as R → ∞. We shall integrate the function f (z) = 1 1/(z + i) g(z) 1 = = . Suppose f (z) ≤ M for all z ∈ C. Now apply Corollary 5. As an example of the usefulness of Liouville’s theorem we give another proof of the fundamental theorem of algebra. remembering that γR has circumference 2πR and z − w = R for all z on γR : f (w) = 1 2πi M ≤ .7. where R > 1. we can apply Cauchy’s formula: 1 2πi and so S σ z2 dz 1 = +1 2πi σ g(z) 1 1 dz = g(i) = = .1 (d). −R ≤ t ≤ R. we apply Theorem 5. Now we apply Proposition 4. Given any w ∈ C. obtaining R dt dz = . and hence. z 2 + 1 ≥ 1 z2 2 for large enough z by Lemma 5.3. As one more example of this theme of getting results from Cauchy’s formula by taking the limit as a path goes to inﬁnity. which contradicts our assumptions. f is also bounded (Exercise 7). f is constant. p(z) = 0 for all z ∈ C. z2 + 1 2i dz + z2 + 1 T dz = z2 + 1 (∗∗) σ Now this formula holds for all R > 1.CHAPTER 5.5 to deduce that f is constant. followed by the circular arc T of radius R in the upper half plane from R to −R. we compute an improper integral. 2+1 2 S z −R 1 + t . z−i i+i 2i dz 1 = 2πi · = π.
6. Then γz + δ − γz+h is a closed smooth curve in G. just a slight modiﬁcation of this example leads to an improper integral which is far beyond the scope of basic calculus. h→0 h δ h→0 h l(z. t2 + 1 Of course this integral can be evaluated almost as easily using standard formulas from calculus. see Exercise 11. Suppose f is analytic in the simplyconnected region G. and it is Gcontractible as G is simply connected. However. 5. It remains to show that F is a primitive of f : F (z) = lim F (z + h) − F (z) 1 = lim h→0 h→0 h h f− γz+h γz f . This concept allows the following result. Hence by Corollary 4.) If h is suﬃciently small.5 0= γz −δz f= γz f− δz f which means we get the same integral no matter which path we take from a to z. z + h) between z and z + h lies in G and by Corollary 4.z+h) .3 Antiderivatives Revisited and Morera’s Theorem A region G is said to be simply connected if every closed curve in G is Gcontractible. Then f has a primitive in G. Theorem 5. as G is simply connected. δ (One should keep in mind that δ very much depends on z and h. We should make sure that F is well deﬁned: Suppose δz is another smooth curve from a to z then γz − δz is closed and Gcontractible.CHAPTER 5.5 (again we use that G is simply connected) 1 1 F (z) = lim f = lim f.5 gives us f+ γz δ f− γz+h f = 0. Making these observations in the limit of the formula (∗∗) as R → ∞ now produces ∞ −∞ dt = π. Fix a point a ∈ G and let F (z) = γz f where γz is any smooth curve from a to z. Hence again Corollary 4. Proof. Now let δ be a smooth curve in G from z to z + h. CONSEQUENCES OF CAUCHY’S THEOREM 53 As R → ∞ this approaches an improper integral. the line segment l(z. so F is a welldeﬁned function. that is. F (z) = lim 1 h→0 h f− γz+h γz f = lim 1 h→0 h f.
Now because f (z) dw = f (z) l(z. which has a primitive.6 has itself a primitive.z+h) ≤ lim = lim h→0 1 max f (w) − f (z) length(l(z.uk/∼history/Biographies/Morera. which has a primitive. This means that the primitive of a function f obtained by Theorem 5.6: we did not really need the fact that every closed curve in G is contractible. h→0 w∈l(z. which has .z+h) F (z) − f (z) = lim = lim h→0 1 h 1 h f (w) dw − lim l(z. This fact can be exploited to give a sort of converse statement to Corollary 4.z+h) h→0 f (w) − f (z) dw l(z. CONSEQUENCES OF CAUCHY’S THEOREM 54 a γz z γz+h G δ z+h Figure 5. For more information about Giancinto Morera (1856–1907). There is an interesting consequence to be drawn from this theorem.z+h) max f (w) − f (z) = 0 .5.CHAPTER 5.z+h) h→0 1 h f (z) dw l(z. z + h)) h w∈l(z.7 (Morera’s5 Theorem). Then f is analytic in G.dcs.z+h) dw = f (z) h.6.2: Proof of Theorem 5.2 ‘in the other direction.stand. see http://wwwgroups.ac. It follows from the fact that a primitive of a function is. . Corollary 5. . we obtain l(z.z+h) The last equality follows from the continuity of f .html. Suppose f is continuous in the region G and f =0 γ for all smooth closed paths γ ⊂ G. This is the same behavior which we discovered in Corollary 5. diﬀerentiable. by deﬁnition. just that every closed curve gives a zero integral for f .’ Another consequence comes from the proof of Theorem 5. 5 .
z 2 −4 exp z . π 2 (b) γ2 ≤t≤ 3π 2 . 4.2 gives that f is also analytic on G. 5. use the principal value of z i . Suppose f is continuous on C and limz→∞ f (z) = 0. Compute the following integrals. Prove that z exp z 2 dz = 0 for any closed curve γ. 2. Exercises 1. − π ≤ t ≤ π .CHAPTER 5.) (a) γ1 z i dz where γ1 (t) = eit . (b) (c) (d) (e) (f) (g) (h) (i) 1 z− 1 2 . 1 (z 2 + 2 )2 exp z . where γz is any smooth curve in G from a to z. z iz−3 .1. Is f bounded for z ≤ R?) . Because F is analytic on G. 1 . (z+4)(z 2 +1) γ w is any ﬁxed complex number with w = 3. (Hint: From the deﬁnition of limit at inﬁnity (with = 1) there is R > 0 so that f (z) − 0 = f  (z) < 1 if z > R. Corollary 5. CONSEQUENCES OF CAUCHY’S THEOREM Proof. z3 cos z 2 . Integrate the following functions over the circle z = 3. we ﬁx an a ∈ G and deﬁne F (z) = γz 55 f. 3. As in the previous proof. 2 2 z i dz where γ2 (t) = eit . Show that f is bounded. and exactly as above we can show that F is a primitive for f in G. oriented counterclockwise: (a) Log(z − 4i). sin z . As above. (How does this 6. (Hint: one of these integrals is considerably easier than the other. Find a (maximal size) set on which f (z) = exp compare with the real function f (x) = e1/x ?) 1 z has an antiderivative. Prove the formula for f in Theorem 5. Show that exp(sin z) has an antiderivative on C. 7. where (z−w)2 1 . this is a welldeﬁned function because all closed paths give a zero integral for f .
and conclude that F (z) ≤ large enough. Show that a polynomial of odd degree with real coeﬃcients must have a real zero. 11. followed by the circular arc T of radius R in the upper half plane from R to −R. z2 . b such that f (z) ≤ az + b for all z ∈ C.2: (a) Show that σ F (z) dz = π if σ is the counterclockwise semicircle formed by the segment e S of the real axis from −R to R. . by parameterizing the integral over S in terms of t and just considering the ∞ real part. 9. jk such that p(z) = c (z − z1 )j1 (z − z2 )j2 · · · (z − zk )jk . z1 . that −∞ cos(t) dx = π .CHAPTER 5. Let p be a polynomial of degree n > 0. Prove that f is a linear polynomial (that is. . CONSEQUENCES OF CAUCHY’S THEOREM 56 8. (c) Show that limR→∞ T 2 z2 for z F (z) dz = 0. zk and positive integers j1 . (Hint: Exercise 14b in Chapter 1. and hence limR→∞ S F (z) dz = π . . Suppose f is entire and there exist constants a. . . Modify the example at the end of Section 5. . of degree ≤ 1).) 10. . In this problem F (z) = eiz z 2 +1 and R > 1. . e (d) Conclude. e t2 +1 . (b) Show that eiz ≤ 1 for z in the upper half plane. Prove that there exist complex numbers c. where j1 + · · · + jk = n.
and hence so are u and v. 57 . Hadamard 6. and they can be found in the next two theorems. There are (at least) two reasons why harmonic functions are part of the study of complex analysis. by Corollary 5. By Theorem 2. Proposition 6.2. The proof that v satisﬁes the Laplace equation is completely analogous. In particular. Let G ⊆ C be a region. Suppose f = u + iv is analytic in the region G.uk/∼history/Biographies/Laplace. A function u : G → R is harmonic in G if it has continuous second partials in G and satisﬁes the Laplace1 equation uxx + uyy = 0 in G.1.6. see http://wwwgroups. 1 For more information about PierreSimon Laplace (1749–1827). Then u and v are harmonic in G. Hence uxx + uyy = (ux )x + (uy )y = (vy )x + (−vx )y = vyx − vxy = 0 in G.dcs.html.stand. Note that in the last step we used the fact that v has continuous second partials.1 Deﬁnition and Basic Properties We will now spend a chapter on certain functions deﬁned on subsets of the complex plane which are real valued.1. The main motivation for studying them is that the partial diﬀerential equation they satisfy is very common in the physical sciences. Proof. First. Deﬁnition 6.ac. u and v satisfy the Cauchy–Riemann equations ux = vy and uy = −vx in G. J. u and v have continuous second partials. f is inﬁnitely diﬀerentiable.Chapter 6 Harmonic Functions The shortest route between two truths in the real domain passes through the complex domain.
comparing the imaginary parts of g and h yields −uy = −ay or u(x. however. y) = a(x. which one might appreciate better when looking back at the simple deﬁnition of harmonic functions. y) = a(x. which will be almost the function f that we’re after. the two theorems we’ve just proved allow for a powerful interplay between harmonic and analytic functions. In that spirit. The function v is called a harmonic conjugate of u. To prove that g is analytic. (The second equation follows with the Cauchy–Riemann equations.1 shouts for a converse theorem. and then to construct an antiderivative of g. .CHAPTER 6. But then f =h−c is a function analytic in G whose real part is u. Suppose u is harmonic on the simply connected region G. we use Theorem 2. The plan is to prove that g is analytic. (Note that for the application of this theorem we need the fact that G is simply connected. functions which are harmonic in a region G but not the real part (say) of an analytic function in G (Exercise 3). however. the following theorem might appear not too surprising. As one might imagine.) Suppose we decompose h into its real and imaginary parts as h = a + ib.6 such a function f = u + iv must satisfy f = ux + ivx = ux − iuy . Proof. by Theorem 2. (The second equation follows with the Cauchy–Riemann equations. On the other hand.1 if we restrict ourselves to simply connected regions. a very strong result. again because u is harmonic.) It is also worth mentioning that the proof shows that if u is harmonic in G then ux is the real part of a function analytic in G regardless whether G is simply connected or not.6: ﬁrst because u is harmonic. We will explicitly construct the analytic function f (and thus v = Im f ). Hence c has to be constant.) But the real part of g is ux . y) + c(y) for some function c which only depends on y. and c depends only on x. and u = a + c. Now that we know that g is analytic in G. Then there exists a harmonic function v such that f = u + iv is analytic in G.6 to obtain a primitive h of g on G. Then. Moreover. we can use Theorem 5. it should not be surprising that the function g which we ﬁrst constructed is the derivative of the soughtafter function f .6. In hindsight. let g = ux − iuy . First. There are.2. Remark. g = h = ax + ibx = ax − iay . Re g = ux and Im g = −uy have continuous partials. again using Theorem 2. Namely. y) + c(x). Theorem 6. they satisfy the Cauchy–Riemann equations: (Re g)x = uxx = −uyy = (Im g)y and (Re g)y = uxy = uyx = − (Im g)x . as promised. It is. HARMONIC FUNCTIONS 58 Proposition 6. Remark. We do obtain a converse of Proposition 6. so that we obtain ux = ax or u(x.
D is simply connected. Suppose u is harmonic in G. Then there is a disk in G centered at w containing a point z0 with u(z0 ) < u(w). Theorem 6. there exists a function f analytic in D such that u = Re f on D.3. and because z0 was chosen arbitrarily. Assume (by way of contradiction) that w is a strong local maximum of u in G. Theorem 4. Now we apply Corollary 4. If u is harmonic in the region G. The function u : G ⊂ C → R has a strong relative maximum at w if there exists a disk D = {z ∈ C : z − w < R} ⊂ G such that u(z) ≤ u(w) for all z ∈ D and u(z0 ) < u(w) for some z0 ∈ D. 0 Proof. This f might very well diﬀer from one disk to the next.10 to f : f (w) = 1 2π 2π f w + reit dt . Proof. 6. Suppose z0 − w = r. By Corollary 5. Fix z0 ∈ G and r > 0 such that the disk D = {z ∈ C : z − z0  < r} 59 is contained in G.4. then it does not have a strong relative maximum or minimum in G.8. Remark. so by Theorem 6. The disk D = {z ∈ C : z − w ≤ r} is simply connected.CHAPTER 6. we proved the statement. Theorem 6. This is the ﬁrst in a series of proofs which uses the fact that the property of being harmonic is a local property—it is a property at each point of a certain region. HARMONIC FUNCTIONS Corollary 6. Because z0 ∈ D. and {z ∈ C : z − w ≤ r} ⊂ G. A harmonic function is inﬁnitely diﬀerentiable.4 with this r: u(w) = 1 2π 2π u w + reit dt . Note that we did not construct a function f which is analytic in G but we only constructed such a function on the disk D.2.4 states that harmonic functions have the meanvalue property.2 MeanValue and Maximum/Minimum Principle The following identity is the harmonic analog of Cauchy’s integral formula. 0 The statement follows by taking the real part on both sides.2 there is a function f analytic on D such that u = Re f on D. The following result is a fairly straightforward consequence of this property. Suppose u is harmonic in the region G. we apply Theorem 6. so by the last theorem. we showed that u is inﬁnitely diﬀerentiable at z0 . Then u(w) = 1 2π 2π u w + reit dt . Proof. 0 . The deﬁnition of a strong relative minimum is completely analogous. Theorem 6.5. f is inﬁnitely diﬀerentiable on D. and hence so is its real part u.
for the middle integral we get a strict inequality. Corollary 6. To make this into a thorough argument. Hence u(w) < 1 2π t1 t2 2π u(w) dt + 0 t1 u(w) dt t2 u(w) dt = u(w) . Corollary 8. Suppose u is harmonic in the closure of the bounded region G. We will see in Corollary 8. z∈G z∈∂G (Here ∂G denotes the boundary of G. Corollary 8. . there is a whole interval of parameters. then it does not have a weak relative maximum or minimum in G. in the sense that there exists a disk D = {z ∈ C : z − w < R} ⊂ G such that all z ∈ D satisfy u(z) ≤ u(w).12 a variation of this theorem for a weak relative maximum w. If u is zero on ∂G then u is zero in G. a contradiction.) We’ll exploit this fact in the next two corollaries.12 says that if u is harmonic in the region G.5. contradicting the meanvalue property. But in this case there’s actually a short cut: if u has a strong relative minimum then the harmonic function −u has a strong relative maximum.12 implies that if u is harmonic in the closure of the bounded region G then max u(z) = max u(z) z∈G z∈∂G and min u(z) = min u(z) . Now we split up the meanvalue integral: u(w) = 1 2π 1 = 2π 2π u w + reit dt 0 t1 0 u w + reit dt + t2 t1 2π u w + reit dt + t2 u w + reit dt All the integrands can be bounded by u(w). A look into the (not so distant) future. say t1 ≤ t ≤ t2 (and t0 is among those t). A special yet important case of the above maximum/minimum principle is obtained when considering bounded regions. Intuitively. suppose that z0 = w + reit0 . such that u w + reit < u(w). this cannot hold.1: Proof of Theorem 6.CHAPTER 6. The same argument works if we assume that u has a relative minimum. Because u(z0 ) < u(w) and u is continuous. because some of the function values we’re integrating are smaller than u(w). which we just showed cannot exist. HARMONIC FUNCTIONS 60 w w + reit1 w + reit0 w + reit2 Figure 6.6.
Suppose u and v are harmonic. One should remark.dcs.CHAPTER 6.7 says that the solution to the Dirichlet problem is unique. (a) Show that u is harmonic in C \ {0}. 3 For more information about Sim´on Denis Poisson (1781–1840). we just remark that Corollary 6. Now apply the previous corollary. Consider u(z) = u(x. z∈G z∈∂G z∈∂G so u has to be zero in G. (b) Find an entire function f such that Re(f ) = u. y) = ln x2 + y 2 . see http://wwwgroups. y) = ex sin y. Show that all partial derivatives of a harmonic function are harmonic.stand. see e http://wwwgroups.uk/∼history/Biographies/Dirichlet. Is it possible to ﬁnd a real function v so that x3 + y 3 + iv is analytic? For more information about Johann Peter Gustav Dirichlet (1805–1859). Prove that u + cv is also harmonic. 2.html.ac. Let u(x.uk/∼history/Biographies/Poisson. that this result is of a completely theoretical nature: it says nothing about how to extend a function given on the boundary of a region to the full region. for other regions one needs to ﬁnd a conformal map to the unit disk.dcs. If two harmonic functions agree on the boundary of a bounded region then they agree in the region. Exercises 1.7. 4.html. 3. All of this is beyond the scope of these notes. HARMONIC FUNCTIONS Proof. There is a fairly simple formula (involving the socalled Poisson3 kernel ) if the region in question is a disk. Proof. By the remark we just made u(z) ≤ max u(z) = max u(z) = max 0 = 0 z∈G z∈∂G z∈∂G 61 and u(z) ≥ min u(z) = min u(z) = min 0 = 0 . The last corollary states that if we know a harmonic function on the boundary of some region then we know it inside the region. and c ∈ R. 2 . This problem is called the Dirichlet2 problem and has a solution for all simplyconnected regions. Suppose u and v are harmonic in G ∪ ∂G and they agree on ∂G. Corollary 6. (b) Prove that u is not the real part of a function which is analytic in C \ {0}.ac.stand. Then u − v is also harmonic in G ∪ ∂G (Exercise 2) and u − v is zero on ∂G. (a) Show that u is harmonic on C. however. 5.
Then for any n ≥ N . in in in 1 1 −0 = = = ≤ < . a(n)) and we commonly denote the sequence by (an )∞ . The sequence (an = in ) diverges: Given a ∈ C.Chapter 7 Power Series It is a pain to think about convergence but sometimes you really have to. n→∞ If no such a exists then the sequence (an ) is divergent. (an )n≥1 . Then the sequence (an ) is convergent and a is its limit. (This is always possible since a4k+2 = i4k+2 = −1 for any k ≥ 0. then for any N . we have an − a < . choose n ≥ N such that an = 1. choose n ≥ N such that an = −1. in symbols lim an = a .2. n n n n N Example 7. choose N > 1/ . Example 7. (This is always possible since a4k = i4k = 1 for any k > 0. a (complex) sequence is a function from the positive (sometimes the nonnegative) integers to the complex numbers. choose = 1/2. Sinai Robins 7. Its values are usually denoted by an (as opposed to.1 Sequences and Completeness As in the real case (and there will be no surprises in this chapter of the nature ‘real versus complex’). say. Suppose (an ) is a sequence and a ∈ C such that for all > 0.1. We consider two cases: If Re a ≥ 0. then for any N . 2 . 2 If Re a < 0.1. Deﬁnition 7. or simply (an ).1.) Then a − an  = a − 1 ≥ 1 > 62 1 . The notion of convergence of a sequence n=1 is based on the following sibling of Deﬁnition 2. there is an integer N such that for all n ≥ N . limn→∞ in n = 0: Given > 0.) Then a − an  = a + 1 ≥ 1 > 1 .
Let (an ) and (bn ) be convergent sequences and c ∈ C. This essentially says that there are no inﬁnities in the reals. n→∞ n→∞ n→∞ 63 (b) lim an · lim bn = lim (an · bn ) . the sequence converges because it is decreasing and bounded below by 0. We will accept the following axiom as our version of the completeness property: Axiom (Monotone Sequence Property). in many cases. If the limit is L then. . n→∞ n→∞ where we require that an be in the domain of f . if f is continuous at a then lim f (an ) = f (a) if lim an = a . In the quotient law we have to make sure we do not divide by zero. all irrational numbers are actually deﬁned this way.1. both of them can be proved by more elementary considerations. by treating n as the variable). This property of the real numbers is called completeness. while the monotone sequence property shows that any such inﬁnite decimal expansion actually converges to a real number. For a proof see Exercise 3. we get L = limn→∞ rn = limn→∞ rn+1 = r limn→∞ rn = rL. Remember that a sequence is monotone if it is either nondecreasing (xn+1 ≥ xn ) or nonincreasing (xn+1 ≤ xn ). The most important property of the real number system is that we can. (a) lim an + c lim bn = lim (an + c bn ) .1. Any bounded monotone sequence converges.3. If x is any real number than there is an integer N which is greater than x. Lemma 7.2.CHAPTER 7. The ﬁrst of these can be established by calculus methods (like L’Hospital’s rule. It is interesting to see that the Archimedean principle underlies the construction of an inﬁnite decimal expansion for any real number.2 (Archimedean Property). (b) Factorials beat exponentials: limn→∞ an n! = 0 if a is a constant. so L = 0 since 1 − r = 0 The following is a consequence of the monotone sequence property. POWER SERIES The following limit laws are the relatives of the identities stated in Lemma 2. Example 7. In fact.3. using the laws of limits. Moreover. In this sense we can use the sequence to deﬁne a real number. and it can be formulated in many equivalent ways. (a) Exponentials beat polynomials: limn→∞ bn p(n) = 0 if p(n) is a polynomial of ﬁxed degree in n and b < 1. n→∞ n→∞ n→∞ (c) limn→∞ an = lim n→∞ limn→∞ bn an bn . From L = rL we get (1 − r)L = 0. If 0 ≤ r < 1 then limn→∞ rn = 0: First. Lemma 7. We close this discussion of limits with a pair of standard limits. although it is often listed as a separate axiom: Theorem 7. Notice that this was already used in Example 7. as limits of rational numbers. determine that a sequence converges without knowing the value of the limit.
. we usually express its limit as a = ∞ k=1 bk or a = k≥1 bk . k=1 To express this in terms of Deﬁnition 7.4. The partial sums of a series form a nondecreasing sequence if the terms of the series are nonnegative. For starters. and it is fortunate that the monotone sequence property has a very convenient translation into the language of series of real numbers.4 to say: ∞ bk converges in the reals if and only if it is ﬁnite. POWER SERIES 64 7. and this observation immediately yields: Lemma 7. If we wanted to be lazy we would for convergence of a series simply refer to convergence of the partial sums of the series. k=1 We have already used the simple fact that convergence of a sequence (an ) is equivalent to the convergence of (an−1 ). k=1 we can rephrase Lemma 7. k=1 In the case of a convergent series.CHAPTER 7.1. a series converges to the limit (or sum) a by deﬁnition if n n→∞ lim an = lim n→∞ bk = a . However. n+1 1− A series where most of the terms cancel like this is called a telescoping series.4. If an is the nth partial sum of the series k≥1 bk then an = an−1 + bn . ∞ k=1 bk converges if and only if the partial If bk are nonnegative real numbers and the partial sums of the series ∞ bk are unbounded k=1 then the partial sums “converge” to inﬁnity. here k=0 k=1 (bk ) is the sequence of terms of the series. If k≥1 bk converges then limn→∞ bn = 0. Occasionally we can ﬁnd the limit of a sequence by manipulating the partial sums: 1 = lim k(k + 1) n→∞ = lim = lim n→∞ n k=1 k≥1 1 1 − k k+1 1 2 + n→∞ = lim n→∞ 1 1 1 1 1 1 − + − + ··· + − 2 3 3 4 n n+1 1 1 1 1 1 1 1 1 − + − + − + ··· + − 2 2 3 3 4 n n+1 1 1− = 1. there are some convergence features which take on special appearances for series. so we can write ∞ bk = ∞. for any n≥N n > 0 we have to ﬁnd an N such that for all bk − a < . Most of the time we need to use the completeness property to check convergence of a series.5. If bk are nonnegative real numbers then sums are bounded.2 Series A series is a sequence (an ) whose members are of the form an = n bk (or an = n bk ). after all. and both of these sequences have the same limit. Using this terminology. From this we conclude: Lemma 7. so we should mention them here explicitly. Example 7. The an = n bk (or an = n bk ) are the partial k=0 k=1 sums of the series. we just deﬁned series through sequences.
2 3 4 1 1 1 1 1 1 + + + . say. 2 4 6 2 4 6 1 1 1 1 1 1 1 1 1 + + + + .CHAPTER 7. but the converse is false: converges. The harmonic series k≥1 k diverges (even though the limit of the general term is 0): If we assume the series converges. let N be its limit. there is a small detail to be checked here.6. let P be its limit. then we have L=1+ = > = = 1 1 1 + + + . 2k − 1 2k 2k(2k − 1) k(k + 1) .” but this deﬁnition does not say anything about convergence of the series k≥1 ck . write ck = ak +ibk where ak and bk are real. The alternating harmonic series k≥1 (−1) k series does not converge absolutely. deﬁne c− to be −ck if ck ≤ 0. but the converse is false: 65 1 Example 7.. POWER SERIES A common mistake is to try to use the converse of this result. 2 2 1 1 Here the inequality comes from k > k+1 applied to each term in the ﬁrst sum in parentheses. k k k Similarly.. Then c− ≥ 0 and k≥1 c− ≤ k≥1 ck  < ∞ k k k so k≥1 c− converges... Then k≥1 ak  ≤ k≥1 ck  < ∞ and k≥1 bk  ≤ k≥1 ck  < ∞.. Deﬁne c+ to be ck if ck ≥ 0. or 0 if ck > 0. 3 5 2 4 6 1 1 1 1 1 1 + + + . There is one notion of convergence that’s special to series: we say that k≥1 ck converges absolutely if k≥1 ck  < ∞. Since ck = c+ − c− we see that k≥1 ck converges to k k k P − N. k 2 3 4 5 6 = 1− 1 2 + 1 1 − 3 4 + 1 1 − 5 6 + . In case ck is complex. Another common mistake is to try to use the converse of this result.. or 0 if k ck < 0. rewrite it as follows: (−1)k+1 1 1 1 1 1 = 1 − + − + − + . 2 2 3 4 2 2 3 4 1 1 L + L = L. + + + + ... k+1 k≥1 (Technically. + 1 + + + + . See Exercise 10. But now we have L > L. Then c+ ≥ 0 and k≥1 c+ ≤ k≥1 ck  < ∞ so k≥1 c+ converges.) The reader can verify the inequality 2k(2k−1) ≥ k(k + 1) for k > 1. A and B. Proof... To see that it does converge.. according to the previous example. + + + + . so the general term satisﬁes 1 1 1 1 − = ≤ . we need a proof: Theorem 7. First consider the case when the terms ck are real. both k≥1 ak and k≥1 bk converge to real numbers. since we are eﬀectively ignoring half the partial sums of the original series.6.. But then k≥1 ck converges to A + iB... say to L.... Be careful: We are deﬁning the phrase “converges absolutely. If a series converges absolutely then it converges. which is impossible. By what we just proved. but not absolutely: This Example 7..5.
Adding the pieces gives the inequalities above for the N th partial sum versus the integrals from 1 to N and from 1 to N + 1. And this can make all the diﬀerence . namely ∀ denoting “for all” and ∃ denoting “there is. Then ∞ ∞ ∞ f (t) dt ≤ 1 k=1 f (k) ≤ f (1) + 1 f (t) dt This is immediate from a picture: the integral of f (t) on the interval [k. whereas uniform convergence on G translates into (∀ > 0) (∃ N : (∀ z ∈ G) n ≥ N ⇒ fn (z) − f (z) < ) . (fn ).CHAPTER 7. k + 1] is bounded between f (k) and f (k + 1). but this notion of convergence does not really catch the spirit of the function as a whole. 7. Deﬁnition 7.2. POWER SERIES 66 so the series converges by comparison with the telescoping series of Example 7. positive function deﬁned on [1. In the ﬁrst case. Suppose (fn ) and f are functions deﬁned on G ⊆ C. in the second case we need to ﬁnd an N which works for all z ∈ G. If a sequence of functions.3 Sequences and Series of Functions The fun starts when one studies sequences (fn ) of functions fn . We say that such a sequence converges at z0 if the sequence (of complex numbers) (fn (z0 )) converges.7.” Pointwise convergence on G means (∀ > 0) (∀ z ∈ G) (∃ N : n ≥ N ⇒ fn (z) − f (z) < ) . ∞). For the rest of this book we shall be concerned almost exclusively with series which converge absolutely. We have already used the technique of comparing a series to a series which is known to converge. . .7 (Integral Test). 1 Example 7. this is often called a “comparison test. .4. k≥1 kp converges if p > 1 and diverges if p ≤ 1.” Some variants of the comparison test will appear when we look at power series. The ﬁrst example illustrating this diﬀerence says in essence that if we have a sequence of functions (fn ) which converges uniformly on G then for all z0 ∈ G n→∞ z→z0 lim lim fn (z) = lim lim fn (z) . Suppose f is a nonincreasing. z→z0 n→∞ We will need similar interchanges of limits constantly. If for all > 0 there is an N such that for all z ∈ G and for all n ≥ N we have fn (z) − f (z) < then (fn ) converges uniformly in G to f . What’s the big deal about uniform versus pointwise convergence? It is easiest to describe the diﬀerence with the use of quantiﬁers. and the inequality persists in the limit. converges at all z in some subset G ⊆ C then we say that (fn ) converges pointwise on G. N may well depend on z. No big deal—we only exchanged two of the quantiﬁers. Hence checking convergence of a series is usually a matter of verifying that a series of nonnegative reals is ﬁnite. One handy test is the following: Lemma 7. So far nothing new.
This means that given (the same) δ > 0 such that whenever z − z0  < δ we have fn (z) − fn (z0 ) < 3 . Lemma 7. Then f is continuous on G. we can ask about integration of series of functions. They are formulated for sequences that converge to 0. POWER SERIES 67 Proposition 7. we give two practical tests: one arguing for uniformity and the other against. . Then n→∞ γ lim fn = γ f. we will prove that f is continuous at z0 . in D Lemma 7. Once we know the above result about continuity.1(d). given there is an N such that for all z ∈ G and all n ≥ N fn (z) − f (z) < 3 . Proposition 7.9. so z n → 0 uniformly ¯ r (0) if r < 1. and rn → 0 if r < 1.CHAPTER 7. Suppose fn are continuous on the smooth curve γ and converge uniformly on γ to f . and we can make maxz∈γ fn (z) − f (z) as small as we like. f is continuous at z0 . All that’s left is putting those two inequalities together: by the triangle inequality f (z) − f (z0 ) = f (z) − fn (z) + fn (z) − fn (z0 ) + fn (z0 ) − f (z0 ) ≤ f (z) − fn (z) + fn (z) − fn (z0 ) + fn (z0 ) − f (z0 ) < . Since uniform convergence is often of critical importance.8.10. > 0. ¯ For example. there is a > 0. its consequences (which we will only see in the next chapter) are wide ranging. If fn is a sequence of functions and Mn is a sequence of constants so that Mn converges to 0 and fn (z) ≤ Mn for all z ∈ G. that is. By Proposition 4. The next theorem should come as no surprise. If fn is a sequence of functions which converges uniformly to 0 on a set G and zn is any sequence in G then the sequence fn (zn ) converges to 0. Suppose (fn ) is a sequence of continuous functions on the region G converging uniformly to f on G. z∈γ But fn → f uniformly on γ. z n  ≤ rn if z is in the closed disk Dr (0). Now we make use of the continuity of the fn ’s. then fn converges uniformly to 0 on G. Proof. If a sequence gn converges to a function g then we can usually apply these tests to fn = g − gn . Proof.11. By uniform convergence. however. Let z0 ∈ G. which does converge to 0. we can estimate fn − γ γ f = γ fn − f ≤ max fn (z) − f (z) length(γ) .
these lecture notes) we concentrate on some very special series of functions. Hence the uniform convergence on D of the geometric series will follow if we can show that k≥0 rk converges. Suppose (fk ) are continuous on the region G.12. Then z < 1 if z is in G. k≥0 The fundamental example of a power series is the geometric series. The geometric series k≥0 z converges absolutely for z < 1 to the function 1/(1 − z). Then. For each ﬁxed z we have k≥1 fk (z) ≤ k≥1 Mk < ∞.13. let 1 zn = exp(− n ). suppose > 0. so zn → 0. Deﬁnition 7. Since k≥1 Mk converges there is N so that ∞ n Mk = k>n k=1 Mk − k=1 Mk < for all n > N . Proof.12 with fk (z) = z k and Mk = rk . For example. However. A power series centered at z0 is a series of functions of the form ck (z − z0 )k . for any z in G. fk (z) ≤ Mk for all z ∈ G. and k≥1 Mk converges.CHAPTER 7. and so z n → 0. To see that fn converges uniformly to f . Then k≥1 fk converges absolutely and uniformly in G. Then zn is in G but fn (zn ) = e−1 so fn (zn ) does not converge to 0. All of these notions for sequences of functions go verbatim for series of functions. Proposition 7. We will use Proposition 7.4 Region of Convergence For the remainder of this chapter (indeed. Therefore z n does not converge uniformly to 0 on D1 (0). This deﬁnes a function f on G. k Lemma 7. But this is straightforward: the partial sums of this series can be written as n 1 − rn+1 . often called the Weierstraß M test. There is an important result about series of functions. rk = 1 + r + · · · + rn−1 + rn = 1−r k=0 . and let D = { z ∈ C : z ≤ r }. Proof. The convergence is uniform on any set of the form { z ∈ C : z ≤ r } for any r < 1. POWER SERIES 68 This is most often used to prove nonuniform convergence.3. 7. Fix an r < 1. call the limit f (z). so k≥1 fk (z) converges. for which all ck = 1. Here we also have a notion of absolute convergence (which can be combined with uniform convergence). if n ≥ N then n f (z) − k=1 fk (z) = k>n fn (z) ≤ k>n fn (z) ≤ k>n Mk < and this satisﬁes the deﬁnition of uniform convergence. let fn (z) = z n and let G be the open unit disk D1 (0).
which follows by n z k = lim k≥0 n→∞ z k = lim k=0 1 − z n+1 1 = . Now if z ∈ Dr (z0 ) we have k ≤ c  r k and ck (z − z0 ) k ck  rk = k≥0 k≥0 ck  tk r t k ≤ k≥0 M r t k =M k≥0 r t k = M < ∞. Theorem 7. Deﬁne C to be the set of positive real numbers for which the series k≥0 ck tk converges.8. or ∞. (If R = ∞ then DR (z0 ) is the entire complex plane. By this we mean that R is a nonnegative real number. Corollary 7. so ck  tk ≤ M for some constant M . Suppose the power series a smooth curve in DR (z0 ). k≥0 ck (z − z0 )k does The open disk DR (z0 ) in which the power series converges absolutely is the region of convergence. satisfying the following. 1 − r/t .12.15. Hence.16. Any power series k≥0 ck (z − z0 )k has a radius of convergence R. POWER SERIES 69 whose limit as n → ∞ exists because r < 1. (∗) If r < t and t ∈ C then r ∈ C and k≥0 ck (z − z0 )k converges absolutely and uniformly ¯ on Dr (z0 ). Then k≥0 ck (z − z0 )k has radius of convergence R and γ is ck (z − z0 )k dz = γ k≥0 k≥0 k≥0 ck ck γ (z − z0 )k dz . the geometric series converges absolutely and uniformly on any set of the form {z ∈ C : z ≤ r} with r < 1. note that k≥0 ck tk converges so ck tk → 0 as k → ∞. While we’re at it. n→∞ 1 − z 1−z By comparing a general power series to a geometric series we can give a complete description of its region of convergence. ¯ (a) If r < R then k≥0 ck (z − z0 )k converges absolutely and uniformly on the closed disk Dr (z0 ) of radius r centered at z0 . this theorem immediately implies the following. Proof of Theorem 7. so not converge. Clearly every positive real number is in either C or D. we might as well state what Proposition 7. In particular. and these sets are disjoint. Then the series represents a function which is continuous on DR (z0 ). if γ is closed then γ (z − z0 )k dz = 0. Corollary 7. ¯ this sequence is bounded. In particular. (b) If z − z0  > R then the sequence of terms ck (z − z0 )k is unbounded. and deﬁne D to be the set of positive real numbers for which it diverges.9 implies for power series.) By way of Proposition 7. To prove this.CHAPTER 7. First we establish three facts about these sets. we have absolute convergence for z < 1. It remains to show that for those z the limit function is 1/(1 − z). by Proposition 7.14. Since r can be chosen arbitrarily close to 1. Suppose the power series k≥0 ck (z − z0 )k has radius of convergence R.14. and if R = 0 then DR (z0 ) is the empty set.
in fact. ﬁnd the limit. It is immediate from (∗) or (∗∗) that a0 < b0 . Corollary 7. an ∈ C bn ∈ D Exercises 1. so m0 = (a0 + b0 )/2. a1 and b1 are closer together than a0 and b0 . shows that k k≥0 ck t converges. If m0 lies in C then we deﬁne a1 = m0 and b1 = b0 .14. contradicting the assumption that t is in D. prove convergence/divergence. . POWER SERIES 70 At the last step we recognized the geometric series. so we assume neither is empty and we start with a0 in C and b0 in D.CHAPTER 7. assume that ck r is bounded. Thus R veriﬁes the statement (∗ ∗ ∗). which converges since 0 ≤ r < t. and b1 is in D. (a) an = eiπn/4 . and so ¯ 0 ≤ r/t < 1. Then t ∈ C by (∗ ∗ ∗). so ck  r ≤ M for some constant M . (∗∗) If z − z0  = r > t and t ∈ D then r ∈ D and the sequence ck rk is unbounded.17. b1 − a1 = (b0 − a0 )/2. It is worth mentioning the following corollary. if r = z − z0  > R then choose t so that R < t < r. which reduces the calculation of the radius of convergence to examining the limiting behavior of the terms of the series. First. On the other hand. but if m0 lies in D then we deﬁne a1 = a0 and b1 = m0 . since an converges to R. and hence k k k k≥0 ck (z − z0 ) diverges.17 says anything about convergence on the circle z − z0  = R . Warning: Neither Theorem 7. in either case. and so on. Then t ∈ D by (∗ ∗ ∗). Similarly.14 nor Corollary 7. we have a0 ≤ a1 < b1 ≤ b0 . satisfying 0 ≤ R ≤ ∞. but interchanging r and t. Moreover. (∗ ∗ ∗) There is an extended real number R.14 follows from (∗∗). so r is in C by (∗). if R < r then bn < r for all suﬃciently large n. Summarizing. If 0 < r < R then r < an for all suﬃciently large n. If the sequence converges. so part (a) of 7. Note that. b0 ]. we have an ≤ an+1 bn ≥ bn+1 an < bn bn − an = (b0 − a0 )/2n The sequences an and bn are monotone and bounded (by a0 and b0 ) so they have limits. Notice that R = 0 works if C is empty. so part (b) of 7. (b) (−1)n n . so r is in D by (∗∗).14 follows from (∗). But now exactly the same argument as in (∗). We deﬁne R to be this limit. This shows that r ∈ C. so that 0 < r < R implies r ∈ C and R < r < ∞ implies r ∈ D. For each of the sequences. ﬁrst assume r < R and choose t so that r < t < R. We repeat this procedure to deﬁne a2 and b2 within the interval [a1 . and R = ∞ works if D is empty. and these limits are the same since limn→∞ (bn − an ) = limn→∞ (b0 − a0 )/2n = 0. b1 ]. let m0 be the midpoint of the segment [a0 . We shall deﬁne sequences an in C and bn in D which “zero in” on R. To prove Theorem 7. a1 is in C. To prove this. ck  rk → 0 for 0 ≤ r < R but ck  rk is unbounded for r > R. and uniform and absolute convergence on Dr (z0 ) follows from the Weierstraß M test.
9. Prove Lemma 7. Discuss pointwise and uniform convergence for the following sequences (a) (nz n ) . Prove Lemma 7. 11. 10. Prove Lemma 7.) to k12 . deﬁned on {z ∈ C : Re z ≥ 0}. 71 2.CHAPTER 7. . State and prove a similar theorem for series.1. and show that ∞ cn converges if and only if n=0 ∞ k=0 (c2k + c2k+1 ) converges. Prove that limn→∞ bn = L. 16.) diverges. give an example where cn does not converge to 0 and one series diverges while the other converges. 3.10.3. 8. 15. for z = 1. 12. lim an  = 0. Also. (b) Show that (c) Show that k k≥1 k2 +1 k k≥1 k3 +1 as a diﬀerence of powers of 1 2k . (Hint: compare the general term to converges. Suppose an ≤ bn ≤ cn for all n and limn→∞ an = L = limn→∞ cn . (a) Show that k≥1 21 = 1. Find sup Re e2πit : t ∈ Q \ Z . (b) (c) zn n for n > 0. if the two series converge then they have the same limit. Prove: (cn ) converges if and only if (Re cn ) and (Im cn ) converge. One way to do this is to write k 2 so that you get a telescoping series. Prove: (a) lim an = a n→∞ =⇒ ⇐⇒ n→∞ lim an  = a. 2n2 +1 1 n . Discuss the convergence of 14. Show that the limit of a convergent sequence is unique. POWER SERIES (c) cos n. (Hint: compare the general term k≥0 z k 13. (d) 2 − (e) sin in2 . Prove that the series k≥1 bk converges if and only if limn→∞ 1 2k ∞ k=n bk = 0. Prove Lemma 7. Derive the Archimedean Property from the monotone sequence property. 6. 1 1+nz . Moreover. 7. 4. (b) lim an = 0 n→∞ n→∞ 5.11. Suppose that the terms cn converge to zero.
Show that L is the radius of convergence of (Use the natural interpretations if L = 0 or L = ∞.) (z − z0 )k . a ∈ C. (b) Log z.CHAPTER 7. where 0 ≤ r < 1. (b) Find limn→∞ 1 0 fn (x) dx. 21. k n z k . 72 (a) Show that limn→∞ fn (x) = 0 for all x ≥ 0. and compute their radius of convergence: (a) 1 z. (a) Suppose that the sequence ck is bounded and show that the radius of convergence of k k≥0 ck (z − z0 ) is at least 1. k≥0 2 (b) (c) k≥0 z k! .9? 18. POWER SERIES 17. Find the power series centered at 1 for the following functions. Find a power series (and determine its radius of convergence) of the following functions. . Use the Weierstraß M test to show that each of the following series converges uniformly on the given domain: (a) k≥1 zk ¯ on D1 (0). zk + 1 k≥0 ck (b) k≥0 (c) k≥0 1 22. k2 1 on {z : z ≥ 2}. Suppose L = limk→∞ ck 1/k exists. Treat x = 0 as a special case. (Hint: the answer is not 0. for x > 0 you can use L’Hospital’s rule—but remember that n is the variable. 23. 1 3− z . Let fn (x) = n2 xe−nx . zk zk ¯ on Dr (0). Find the radius of convergence for each of the following series. (a) (b) 1 1+4z . not x. (a) k≥0 ak z k . 20. n ∈ Z.) (c) Why doesn’t your answer to part (b) violate Proposition 7. (b) Suppose that the sequence ck does not converge to 0 and show that the radius of convergence of k≥0 ck (z − z0 )k is at most 1. 2 19.
does not converge to 0 as n → ∞ (d) Why doesn’t this contradict the theorem that “the integral of a uniform limit is the limit of the integrals”? 25. Show that f (3) (0) ≤ 10 . kk cos(k)z k . POWER SERIES (d) k≥1 73 (−1)k k(k+1) z . (Hint: Use the Cauchy integral formula.) What can you say about 9 f (3) (1) ? . (e) k≥1 (f) k≥0 (g) k≥0 4k (z − 2)k . Deﬁne the functions fn (t) = n e−t/n for n > 0 and 0 ≤ t < ∞. (b) Show that fn (t) converges uniformly to 0 as n → ∞.CHAPTER 7. (a) Show that the maximum of fn (t) is (c) Show that ∞ 0 fn (t) dt 1 n. k zk . Let f be analytic on the disk z < 4 and suppose f (z) ≤ 5 for all z on the circle z = 3. 1 24.
and conversely. diﬀerentiable) on their regions of convergence we can ask how to ﬁnd their derivatives. Given any closed curve γ ⊂ {z ∈ C : z − z0  < R}. Suppose f (z) = in {z ∈ C : z − z0  < R}. Then f is analytic Proof.” Theorem 8.2. Now apply Morera’s theorem (Corollary 5.Chapter 8 Taylor and Laurent Series We think in generalities.e.. say for functions from Rn to Rm . Suppose f (z) = k≥0 ck (z − z0 )k has radius of convergence R. A special case of the last result concerns power series with inﬁnite radius of convergence: those represent entire functions. N. γ k≥0 On the other hand. Corollary 7. Now that we know that power series are analytic (i. The ﬁrst of them goes as follows.15 says that f is continuous.1. Whitehead 8. Then k ck (z − z0 )k−1 . k≥1 f (z) = and the radius of convergence of this power series is also R. any analytic function can be represented as a power series.1 Power Series and Analytic Functions All of the last chapter could have been developed in greater generality. The cornerstone is provided by two theorems which say that any power series represents an analytic function.16 ck (z − z0 )k dz = 0 . we have by Corollary 7. but we live in details. 74 . Theorem 8.7). connect the last chapter to the theory of functions analytic on certain regions. however. The next result says that we can simply diﬀerentiate the series “term by term. k≥0 ck (z − z0 )k has radius of convergence R. A. We will now.
Then ck = f (k) (z0 ) . Note that the power series of f converges uniformly on γ. and so on. If k≥0 ck (z − z0 )k and k≥0 ck (z − z0 )k are two power series which both converge to the same function f (z) on an open disk centered at z0 then ck = ck for all k. Applying the same theorem to f gives f (z) = k(k − 1)ck (z − z0 )k−2 k≥2 and f (z0 ) = 2c2 .html. This is the statement of the following Taylor1 series expansion. Taylor’s formulas show that the coeﬃcients of any power series which converges to f on an open disk D centered at z0 can be determined from the the function f restricted to D.1. The various derivatives of a power series can also be seen as ingredients of the series itself.2 gives f (z0 ) = c1 . Suppose f (z) = k≥0 ck (z − z0 )k has a positive radius of convergence. f (z0 ) = c0 . We can play the same game for f (z0 ). Theorem 8.stand.uk/∼history/Biographies/Taylor. since the coeﬃcients for (z − z0 )f (z) are bigger than the corresponding ones for f (z).1 again. so that we are free to interchange integral and inﬁnite sum. Here are the details: f (z) = = = k≥0 1 2πi 1 2πi γ f (w) dw (w − z)2 k≥0 ck (w γ − z0 )k dw (w − z)2 γ ck · ck · k≥0 1 2πi (w − z0 )k dw (w − z)2 = = k≥0 d (w − z0 )k dw w=z k ck (z − z0 )k−1 . Since we know that f is analytic in its region of convergence we can use Theorem 5. Let f (z) = k≥0 ck (z − z0 )k . TAYLOR AND LAURENT SERIES 75 Proof. Let γ be any simple closed curve in {z ∈ C : z − z0  < R}. f (z0 ). the last theorem can be repeatedly applied to f . then to f .CHAPTER 8.3. but applied to the function (z − z0 )k . etc.ac.dcs. . The last statement of the theorem is easy to show: the radius of convergence R of f (z) is at least R (since we have shown that the series converges whenever z − z0  < R). Corollary 8. see http://wwwgroups. 1 For more information about Brook Taylor (1685–1731).4 (Uniqueness of power series). It follows immediately that the coeﬃcients of a power series are unique: Corollary 8. For starters. and it cannot be larger than R by comparison to the series for f (z). Naturally. And then we use Theorem 5. k! Proof.
Γr ∼G\{z0 } γ. Let g(z) = f (z + z0 ). closed.CHAPTER 8. Then f (k) (w) = k! 2πi f (z) dz .5 with those in Corollary 8.1 (which in itself extended Cauchy’s integral formula. Proof. However. simple. (w − z0 )k+1 Here γ is any positively oriented.6 combined with our oftenused Proposition 4. and suppose that z < r.8).5. (w − z0 )k+1 where Γr is a circle centered at z0 with radius r.3.4: f (w) dw = (w − z0 )k+1 f (w) dw . (w − z0 )k+1 Γr γ If we compare the coeﬃcients of the power series obtained in Theorem 8. and γ is a positively oriented. closed. Hence Proposition 7. Suppose f is a function which is analytic in D = {z ∈ C : z − z0  < R}.1(d) gives an inequality which is often called Cauchy’s Estimate: . so g is a function analytic in {z ∈ C : z < R}.6. we apply an easy change of variables to obtain f (z) = k≥0 1 2πi Γr f (w) dw (z − z0 )k . Theorem 4. wk+1 Now.8). since f (z) = g(z − z0 ). TAYLOR AND LAURENT SERIES 76 Theorem 8. Then f can be represented in D as a power series centered at z0 (with a radius of convergence at least R): f (z) = k≥0 ck (z − z0 )k with ck = 1 2πi γ f (w) dw . smooth.9 applies: g(z) = 1 2πi g(w) 1 dw = w−z 2πi g(w) γr γr 1 w k≥0 z w k dw = k≥0 1 2πi γr g(w) dw z k . Then by Cauchy’s integral formula (Theorem 4. w−z γr The factor 1/(w − z) in this integral can be extended into a geometric series (note that w ∈ γr and z so w < 1) 1 1 1 1 z k = z = w−z w 1− w w w k≥0 which converges uniformly in the variable w ∈ γr (by Lemma 7. denote the circle centered at the origin with radius r by γr . (z − w)k+1 γ Corollary 8. w ∈ G.13). smooth curve in D for which z0 is inside γ. Suppose f is analytic on the region G. we arrive at the longpromised extension of Theorem 5. and we can apply Cauchy’s Theorem 4. simple. g(z) = 1 2πi g(w) dw . Gcontractible curve such that w is inside γ. Corollary 8. Fix r < R. The only diﬀerence of this righthand side to the statement of the theorem are the curves we’re integrating over.
we can factor out another factor of z − a. deﬁned on G. We can then ask whether q(z) itself has a zero at a and. TAYLOR AND LAURENT SERIES Corollary 8. not bigger than d. and g(z) is a polynomial which does not have a zero at a. Almost exactly the same thing happens for analytic functions: Theorem 8. Proof. There are now exactly two possibilities: (a) Either ck = 0 for all k. and we can estimate using Proposition 4. if so. Then there are exactly two possibilities: (a) Either: f is identically zero on some open disk D centered at a (that is. . with g(a) = 0 The integer m in the second case is uniquely determined by f and a and is called the multiplicity of the zero at a. continuing in this way we see that we can factor p(z) as p(z) = (z − a)m g(z) where m is a positive integer.CHAPTER 8. f (z) = 0 for all z in D). Then Corollary 8. 8. p(z) = (z − a)q(z) where q(z) is a polynomial of degree d − 1. satisfying f (z) = (z − a)m g(z) for all z in G. We have a power series expansion for f (z) in some disk Dr (a) of radius r around a. and c0 = f (0) is zero since a is a zero of f .1(d): f (k) (w) = k! 2πi f (z) k! k! M k!M f (z) dz ≤ length(γ) ≤ 2πr = k . That is. The integer m is called the multiplicity of the zero a of p(z). (b) or there is some positive integer m so that ck = 0 for all k < m but cm = 0. Then f (k) (w) ≤ k!M . so f (z) = k≥0 ck (z − a)k . if p(a) = 0) then p(z) has z − a as a factor. Notice that f (z) = cm (z − a)m + cm+1 (z − a)m+1 + · · · = (z − a)m (cm + cm+1 (z − a) + · · · ) = (z − a)m k≥0 ck+m (z − a)k .7. The ﬁrst case clearly gives us f (z) = 0 for all z in D = Dr (a). Let γ be a circle centered at w with radius r < R. Rk 77 Proof.2 Classiﬁcation of Zeros and the Identity Principle Basic algebra shows that if a polynomial p(z) of positive degree d has a a zero at a (in other words. Suppose f is an analytic function deﬁned on an open set G and suppose f has a zero at a point a in G. So now consider the second case. (b) or: there is a positive integer m and an analytic function g.6 applies. max k+1 k+1 k+1 2π z∈γ (z − w) 2π r (z − w) r γ The statement now follows since r can be chosen arbitrarily close to R. Suppose f is analytic in {z ∈ C : z − w < R} and f  ≤ M .8 (Classiﬁcation of Zeros).
Then the multiplicity of a equals m if and only if we can write f as a power series of the form f (z) = k≥m ck (z − a)k with cm = 0. Then f (z) = g(z) for all z in G. (b) or there is an open disk D centered at b so that h(z) = 0 for all z in D \ {b}. or h(z) = (z − b)m φ(z) for all z in G. If h(b) = 0 then. where φ is analytic and φ(b) = 0. the two deﬁnitions give the same value when both are applicable. and g is analytic at other points of G by the second deﬁnition. suppose that h(b) = 0. this is a contradiction. so h(zk ) = 0. To see this. there is an open disk D centered at b so that h(z) = 0 for all z ∈ D. Then h is analytic on G. and b ∈ Y if b satisﬁes the second condition.CHAPTER 8. Finally. so b satisﬁes the ﬁrst condition. g(a) = cm = 0. . by continuity. there is an open disk D centered at b so that φ(z) = 0 for all z in D. If b ∈ Y and D is an open disk centered at b as in the second condition then D ⊆ Y . Suppose f and g are analytic in the region G and f (zk ) = g(zk ) at a sequence that converges to a ∈ G with zk = a for all k. Finally. which is sometimes also called the uniqueness theorem. Suppose f is an analytic function deﬁned on an open set G and suppose f has a zero at a ∈ G. since the sequence zk converges to a. Then. Now notice the following: If b is in G then exactly one of the following occurs: (a) Either there is an open disk D centered at b so that h(z) = 0 for all z in D. If b ∈ X and D is an open disk centered at b as in the ﬁrst condition then it is clear that D ⊆ X. Then h(z) = (z − b)m φ(z) = 0 for all z in D except z = b. Proof. The proof of this last theorem immediately yields the following. To start using the intimate connection of analytic functions and power series. there is some k so that zk is in D. Theorem 8.10 (Identity Principle). we must have Y = ∅ and X = G. by the classiﬁcation of zeros. Y ⊆ G. Now we ﬁnish the proof using the deﬁnition of connectedness. Since zk = a. We start by deﬁning h = f − g. which is unique. we apply Theorem 8. we check that our original point a lies in X. and we saw that this means z satisﬁes the second condition. Corollary 8. But X = G implies that every z in G satisﬁes the ﬁrst condition above. Then. so b satisﬁes the second condition. Clearly m is unique. since it is deﬁned in terms of the power series expansion of f at a. Since a is in X. so that b ∈ X if b satisﬁes the ﬁrst condition above. Now deﬁne two sets X. TAYLOR AND LAURENT SERIES Then we can deﬁne a function g on G by k≥0 ck+m (z − a)k g(z) = f (z) (z − a)m 78 if z − a < r if z ∈ G \ {a} According to our calculations above. so one of them must be empty. since φ is continuous. so h(z) = 0.9. But.8 to obtain the following result. and we will be ﬁnished if we can deduce that h is identically zero on G. X and Y are disjoint open sets whose union is G. h(zn ) = 0. To see this. since if z ∈ D \ {b} then h(z) = 0. The function g is analytic at a by the ﬁrst deﬁnition. suppose a ∈ Y . and let D be an open disk centered at a so that h(z) = 0 for all z in D except z = b. either h(z) = 0 for all z in some open disk D centered at b. so b satisﬁes the second condition.
It is natural.CHAPTER 8. Hence g(z) = eh(z) must be equal to e0 = 1 for all z in D. and we have the condition g(z) ≤ g(a) = 1 for all z in D0 .11 (MaximumModulus Theorem). not as general 1 as it could be. Suppose f is analytic and not constant in the region G. To make sense of expressions like the above. for example. Hence. Then f  does not attain a weak relative maximum in G. It is.5. . by the identity principle. we introduce the concept of a double series ak = ak + a−k . Corollary 8. Thus the function h = Log ◦g is deﬁned and analytic on D.13 (MinimumModulus Theorem). then it does not have a weak relative maximum or minimum in G. using continuity. Suppose there is a point a in G and an open disk D0 centered at a so that f (z) ≤ f (a) for all z in D0 . Suppose f is analytic and not constant in the region G. and we have h(a) = Log(g(a)) = Log(1) = 0 and Re h(z) = Re Log(g(z)) = ln(g(z)) ≤ ln(1) = 0. so f is identically zero. such as: If G is a bounded region and f is analytic in the closure of G. Since g(a) = 1 we can ﬁnd. to think about representing exp z as exp 1 z = k≥0 1 k! 1 z k = k≥0 1 −k z . Corollary 8. Absolute and uniform convergence are deﬁned analogously. we can now state the following central deﬁnition.11 can be used to give a proof of the analogous theorem for harmonic functions. in the process strengthening that theorem to cover weak maxima and weak minima. 8. We now refer to Exercise 27. Theorem 8. Theorem 8. we should not be too surprised to ﬁnd the following result whose proof we leave for the exercises. If f (a) = 0 then f (z) = 0 for all z in D0 . Theorem 6. we can prove yet another important property of analytic functions. So we assume f (a) = 0. Then f  does not attain a weak relative minimum at a in G unless f (a) = 0. f (z) has the constant value f (a) for all z in G. and so f (z) = f (a)g(z) must have the constant value f (a) for all z in D. TAYLOR AND LAURENT SERIES 79 Using the identity principle. which shows that h must be identically zero in D. In this case we can deﬁne an analytic function g(z) = f (z)/f (a). Equipped with this. If u is harmonic in the region G. Since the last corollary also covers minima of harmonic functions. however.3 Laurent Series Theorem 8. A double series converges if both its deﬁning series do. k! a “power series” with negative exponents. a smaller open disk D centered at a so that g(z) has positive real part for all z in D. by the identity principle. then the maximum of f  is attained on the boundary of G. Proof. k∈Z k≥0 k≥1 Here ak ∈ C are terms indexed by the integers.12.5 gives a powerful way of describing analytic functions. There are many reformulations of this theorem.
ac.html. smooth path that is Ahomotopic to the circle.stand. Let g(z) = f (z + z0 ). TAYLOR AND LAURENT SERIES Deﬁnition 8. k∈Z k≥0 k≥1 The ﬁrst of the series on the righthand side is a power series with some radius of convergence R2 . whence the Laurent series converges on the annulus {z ∈ C : R1 < z − z0  < R2 } (if R1 < R2 ).” it will converge for z−z0 < R1 for some R1 . We should pause for a minute and ask for which z such a Laurent series can possibly converge. Naturally.uk/∼history/Biographies/Laurent Pierre. respectively. 2 For more information about Pierre Alphonse Laurent (1813–1854). so g is a function analytic in {z ∈ C : R1 < z < R2 }.1) γ2 −γ1 γ2 γ1 For the integral over γ2 we play exactly the same game as in Theorem 8. The series which started this section is the Laurent series of exp Example 8. Remark. it converges in {z ∈ C : z − z0  < R2 }. centered at 0. (w − z0 )k+1 Here γ is any circle in A centered at z0 .1. by Cauchy’s Theorem 4. .5. Proof. see http://wwwgroups. Theorem 8.1. The factor 1/(w − z) in z this integral can be expanded into a geometric series (note that w ∈ γ2 and so w < 1) 1 1 1 = w−z w 1− z w = 1 w k≥0 z w k .CHAPTER 8. and let γ1 and γ2 be positively oriented circles centered at 0 with radii r1 and r2 . Fix R1 < r1 < z < r2 < R2 . we can apply Cauchy’s integral formula (Theorem 4. Theorem 8. Even better. 1 z k∈Z ck 80 (z − z0 )k . A Laurent2 series centered at z0 is a double series of the form Example 8. we need to combine those two notions.dcs. By introducing an “extra piece” (see Figure 8. Suppose f is a function which is analytic in A = {z ∈ C : R1 < z − z0  < R2 }. By deﬁnition ck (z − z0 )k = ck (z − z0 )k + c−k (z − z0 )−k . in {z ∈ C : z − z0  > R1 }. that is.1).1 says that the Laurent series represents a function which is analytic on {z ∈ C : R1 < z − z0  < R2 }. The second we can view as a “power series 1 1 1 in z−z0 .4 we can replace the circle in the formula for the Laurent series by any closed.14 implies that the convergence is uniform on a set of the form {z ∈ C : r1 ≤ z − z0  ≤ r2 } for any R1 < r1 < r2 < R2 . The fact that we can conversely represent any function analytic in such an annulus by a Laurent series is the substance of the next theorem. Theorem 7. w−z (8.14. that is.2. Any power series is a Laurent series (with ck = 0 for k < 0).8) to the path γ2 − γ1 : g(z) = 1 2πi g(w) 1 dw = w−z 2πi g(w) 1 dw − w−z 2πi g(w) dw . Then f can be represented in A as a Laurent series centered at z0 : f (z) = k∈Z ck (z − z0 )k with ck = 1 2πi γ f (w) dw . For the convergence of our Laurent series.
wk+1 k∈Z γ The statement follows now with f (z) = g(z − z0 ) and an easy change of variables. TAYLOR AND LAURENT SERIES 81 γ2 γ1 Figure 8. wk+1 Putting everything back into (8.1) gives g(z) = 1 2πi g(w) dw z k + wk+1 g(w) dw z k . .1: Proof of Theorem 8. now we expand the factor 1/(w − z) into the following geometric series (note that w ∈ γ1 and so w < 1) z 1 1 1 =− w−z z 1− w z =− 1 z k≥0 w z k . wk+1 The integral over γ1 is computed in a similar fashion.14. which converges uniformly in the variable w ∈ γ1 (by Lemma 7.9 applies: g(w) dw = − w−z g(w) γ1 γ1 1 z k≥0 w z k dw = − k≥0 γ1 g(w)wk dw z −k−1 = − k≤−1 γ1 g(w) dw z k . which ﬁnally gives g(z) = 1 2πi g(w) dw z k .CHAPTER 8.13). Hence Proposition 7.4).9 applies: g(w) dw = w−z g(w) γ2 γ2 1 w k≥0 z w k dw = k≥0 γ2 g(w) dw z k . wk+1 k≥0 γ2 k≤−1 γ1 We can now change both integration paths to a circle γ centered at 0 with a radius between R1 and R2 (by Cauchy’s Theorem 4.13). which converges uniformly in the variable w ∈ γ2 (by Lemma 7. Again Proposition 7.
Prove the following generalization of Theorem 8. The coeﬃcients of a Laurent series are unique.8 using the power series of exp z centered at 0. 7. ﬁnd a power series for arctan(z). What functions are represented by the series in the previous exercise? 3. Do not ﬁnd the general form for the coeﬃcients. . Then f is analytic in G.) 8. Exercises 1. centered at z0 . Then for any k ∈ N. z0 = 0 (use the principal branch). (a) f (z) = (b) f (z) = (c) f (z) = 1 .CHAPTER 8. Prove Lemma 3. 2. (2k + 1)! 1 z−3 k (b) k≥0 (c) k≥0 . Find the terms through third order and the radius of convergence of the power series for each following functions. determine where the series converges absolutely/uniformly: (a) k≥2 k(k − 1) z k−2 . we immediately obtain the following: Corollary 8. z0 = 0. What is its 6. By integrating a series for radius of convergence? 1 1+z 2 term by term. there is only one possible outcome. what it says is simply the following: if we expand a function (that is analytic in some annulus) into a Laurent series. This result seems a bit artiﬁcial.15. the k th derivatives (k) fn converge (pointwise) to f (k) . z0 = 1. z0 = i. √ 1 + z.6 to prove the following: Suppose fn are analytic on the region G and converge uniformly to f on G. 4. (This result is called the Weierstraß convergence theorem. 1 z 2k+1 . 9. Prove the minimummodulus theorem (Corollary 8. 1+z 2 1 ez +1 . 5. 2 (d) f (z) = ez .1: Suppose fn are analytic on the region G and converge uniformly to f on G.13). For each of the following series. Use the previous exercise and Corollary 8. TAYLOR AND LAURENT SERIES 82 We ﬁnish this chapter with a consequence of the above theorem: because the coeﬃcients of a Laurent series are given by integrals. Find the power series centered at 1 for exp z.
13. z0 = 0. 1 sin z 15. Find the ﬁrst 4 nonzero terms in the power series expansion of tan z centered at the origin. (c) f (z) = cos(z) − 1 + 1 sin2 (z). centered at z = −1 and specify the region in which it converges.3 to show that a polynomial does not achieve its minimum modulus on a large circle. (b) f (z) = sin(z) − tan(z). Suppose that f (z) has a zero of multiplicity m at a. 19. Prove: If f is entire and Im(f ) is constant on the unit disc {z ∈ C : z ≤ 1} then f is constant. 2 24. 18. . where f (z) = z 2 − 2. (a) Find the Laurent series for (b) Prove that f (z) = is entire. 1 f (z) cos z−1 z2 1 −2 cos z z2 centered at z = 0. 23. Suppose that f (z0 ) = 0 and f (z0 ) = 0. where k is any integer. TAYLOR AND LAURENT SERIES 83 10. Find the zeros of the following. where a is any constant. What is the radius of convergence? 17. Find a Laurent series for 1 (z−1)(z+1) centered at z = 1 and specify the region in which it 1 z(z−2)2 z−2 z+1 centered at z = 2 and specify the region in which it converges. and determine their multiplicities: (a) (1 + z 2 )4 . 16.4). Find the maximum and minimum of f (z) on the unit disc {z ∈ C : z ≤ 1}. (b) Show that ez cos(z) = 1 2 e(1+i)z + e(1−i)z . Find the ﬁrst ﬁve terms in the Laurent series for centered at z = 0. 11. then use the minimummodulus theorem to deduce that the polynomial has a zero.) 12.13). Find a Laurent series for converges. Find the multiplicities of the zeros: (a) f (z) = ez − 1. Find a Laurent series for 14. (Hint: Use Lemma 5. if z = 0 has a pole of order m 22. Explain why at a. (c) Find the power series expansion for ez cos(z) centered at 0. z0 = 2kπi. Show that f has a zero of multiplicity 1 at z0 . Show that z−1 z−2 = 1 k≥0 (z−1)k for z − 1 > 1. z0 = 0. 21. Give another proof of the fundamental theorem of algebra (Theorem 5. 20. if z = 0.CHAPTER 8. (a) Find the power series representation for eaz centered at 0. using the minimummodulus theorem (Corollary 8.
g is analytic. (z) 1 Show that a = 2πi γ zf(z) dz. (c) z −5 sin(z). (a) Why can you write f (z) = (z − a)m g(z) where m > 0. inside the circle γ.CHAPTER 8. and determine their orders: (a) (z 2 + 1)−3 (z − 1)−4 . (d) (e) 1 1−ez . Find the poles of the following. f 27. Suppose cn  ≥ 2n for all n. Show that f (z) = rm eimθ eiα G(z). What is the radius of convergence of k 2 ck z k . k≥0 (d) (e) k≥0 c2 z k . z 1−ez . and that it has multiplicity 1. and suppose f (a) = 0. at a. What can you say about the radius of convergence of 29. k . (d) z 3 cos z. Suppose the radius of convergence of each of the following? (a) k≥0 k≥0 ck z k k≥0 ck z k? is R. 28. Suppose f is analytic and not identically zero on an open disk D centered at a. 84 26. 3k ck z k . Why is Re G(a) > 0? (c) Why is there a positive constant δ so that Re G(z) > 0 for all z in the open disk Dδ (a)? (d) Write z = a + reiθ for 0 < r < δ. (e) Find a value of θ so that f (z) has positive real part. Suppose that f (z) has exactly one zero. and g(a) = 0? (b) Write g(a) in polar form as g(a) = c eiα and deﬁne G(z) = e−iα g(z). 25. (c) 1 + ez . TAYLOR AND LAURENT SERIES (b) sin2 z. (b) z cot(z). Follow the following outline to show that Re f (z) > 0 for some z in D. k≥0 (b) (c) k≥0 ck z k+5 . ck z 2k .
1 Classiﬁcation of Singularities 1 What is the diﬀerence between the functions sin z . The singularity z0 is called (a) removable if there is a function g analytic in {z ∈ C : z − z0  < R} such that f = g in {z ∈ C : 0 < z − z0  < R}. but the singularities are of a very diﬀerent nature.2. Example 9.Chapter 9 Isolated Singularities and the Residue Theorem 1/r2 has a nasty singularity at r = 0. If f is analytic in the punctured disk {z ∈ C : 0 < z − z0  < R} for some R > 0 but not at z = z0 then z0 is an isolated singularity of f . The function 1 z4 has a pole at 0. z→z0 (c) essential if z0 is neither removable nor a pole. z4 z→0 85 . as lim 1 = ∞. (2k + 1)! sin z 1 = z z k≥0 k≥0 and the power series on the righthand side represents an entire function (you may meditate on the fact why it has to be entire). (b) a pole if lim f (z) = ∞. as for z = 0 (−1)k z 2k+1 = (2k + 1)! (−1)k z 2k .1. Deﬁnition 9.1. z14 . and exp z ? All of them are not deﬁned at z 0. Edward Witten 9. For complex functions there are three types of singularities. The function sin z z has a removable singularity at 0. but it did not bother Newton—the moon is far enough. Example 9. which are classiﬁed as follows.
suppose that lim (z − z0 ) f (z) = 0. so g(z) = (z − z0 )2 k≥0 ck+2 (z − z0 )k = (z − z0 )2 f (z). Then (a) z0 is removable if and only if lim (z − z0 ) f (z) = 0. Then there is some R > 0 so that f (z) > 1 in the punctured ˆ disk DR (z0 ). where h is analytic and not zero at z0 . exp z has an essential singularity at 0. the open disk with radius R centered at z0 such that f = g for z = z0 . and g is analytic on DR (z0 ). and f is analytic on the punctured disk z→z0 ˆ DR (z0 ) = DR (z0 ) \ {z0 }. lim z→z0 f (z) . and it is also diﬀerentiable at z0 . Hence we can factor (z − z0 )2 from the series. so it has a power series expansion g(z) = k≥0 ck (z − z0 )k with c0 = c1 = 0. (a) Suppose z0 is removable. we start with the following results. Clearly g is analytic for z = z0 . for example. (b) Suppose that z0 is a pole of f . ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 86 1 Example 9. Then deﬁne g(z) = (z − z0 )2 f (z) 0 if z = z0 . and this series deﬁnes an analytic function in DR (z0 ). To get a feel for the diﬀerent types of singularities. Proof. exp z approaches 0 as z approaches 0 from the 1 1 negative real axis.3. that is. Proposition 9. This is very similar to the game we played with zeros in Chapter 8: f has a zero of order (or multiplicity) m at z0 if we can write f (z) = (z − z0 )m h(z). Suppose z0 is a isolated singularity of f . The smallest possible n in (b) is the order of the pole. Then we can make use of the fact that g is continuous at z0 : lim (z − z0 ) f (z) = lim (z − z0 ) g(z) = g(z0 ) lim (z − z0 ) = 0 . z→z0 Remark. and 1 = 0. We will see in the proof that “near h(z) the pole z0 ” we can write f (z) as (z−z0 )n for some function h which is analytic (and not zero) at z0 .1.CHAPTER 9. Hence. for z = z0 . since we can calculate g (z0 ) = lim g(z) − g(z0 ) (z − z0 )2 f (z) = lim = lim (z − z0 )f (z) = 0 z→z0 z→z0 z − z0 z − z0 z→z0 So g is analytic in DR (z0 ) with g(z0 ) = 0 and g (z0 ) = 0. The function exp z does not have a removable singularity (consider. 1 1 limx→0+ exp x = ∞). On the other hand. z→z0 (b) if z0 is a pole then lim (z − z0 )n+1 f (z) = 0 for some positive integer n. z→z0 z→z0 z→z0 Conversely. We will make use of the notions of zeros and poles of certain orders quite extensively in this chapter. Hence limz→0 exp z = ∞. if z = z0 . f (z) = k≥0 ck+2 (z −z0 )k .
uk/∼history/Biographies/Picard Emile.) Hence z→z0 lim f (z) − w = ∞. if we deﬁne g(z) by g(z) = 1 f (z) 87 ˆ if z ∈ DR (z0 ).stand. Suppose (by way of contradiction) that there is a w ∈ C and an the punctured disc D (centered at z0 ) w − f (z) ≥ . Remarks. In fact.ac. for any w ∈ C and any > 0 there exists z ∈ D such that w − f (z) < . Theorem 9. (It is worth meditating about coming up with examples of functions which do not miss any point in C and functions which miss exactly one point.dcs. see http://wwwgroups. and which implies the CasoratiWeierstraß theorem. if z = z0 . By the classiﬁcation of zeros. which is beyond the scope of this book. If z0 is an essential singularity of f and D = {z ∈ C : 0 < z − z0  < R} for some R > 0. There is a much stronger theorem. z − z0 But this implies that f has a pole or a removable singularity at z0 . the CasoratiWeierstraß theorem says that the image of any punctured disc centered at an essential singularity is dense in C.CHAPTER 9. Try it!) Proof. 0 then g is analytic in DR (z0 ) (by part (a)). ISOLATED SINGULARITIES AND THE RESIDUE THEOREM So.html. z→z0 lim (z − z0 )n+1 f (z) = lim (z − z0 )h(z) = h(z0 ) lim (z − z0 ) = 0 . z→z0 z→z0 The reader might have noticed that the previous proposition did not include any result on essential singularities. In the language of topology.2 (Casorati1 Weierstraß).ac. we suggest meditating about its statement for a couple of minutes over a good cup of coﬀee. then any w ∈ C is arbitrarily close to a point in f (D). Then the function g(z) = 1 (f (z)−w) > 0 such that for all z in stays bounded as z → z0 . Hence h = φ is an analytic function in DR (z0 ) and f (z) = 1 h(z) 1 = . For more information about Felice Casorati (1835–1890). see http://wwwgroups. It is due to Charles Emile Picard (1856–1941)2 and says that the image of any punctured disc centered at an essential singularity misses at most one point of C. that is. which is a contradiction. and so z→z0 z→z0 lim (z − z0 )g(z) = lim z − z0 = 0. To appreciate the following result. 2.uk/∼history/Biographies/Casorati.html. Not only does the next theorem make up for this but it also nicely illustrates the strangeness of essential singularities.stand. since h is continuous at z0 .dcs. = n φ(z) g(z) (z − z0 ) (z − z0 )n But then. 2 For more information about Picard. g(z) = (z − z0 )n φ(z) where φ is analytic in DR (z0 ) and φ(z0 ) = 0. 1. f (z) − w (The previous proposition tells us that g has a removable singularity at z0 . 1 . φ(z) = 0 for all z in DR (z0 ) since g(z) = 0 for 1 ˆ z ∈ DR (z0 ).
Then the Laurent series of g in this region is a power series. f (z) = k≥0 ck (z − z0 )k−n = k≥−n ck (z − z0 )k . the Laurent series is a power series). and by Corollary 8. (a) Suppose z0 is removable. The following classiﬁes singularities according to their Laurent series. Deﬁne g(z) = k≥0 ck−n (z − z0 )k . we can use it to deﬁne a function which is analytic at z0 . suppose that f (z) = k≥−n ck (z − z0 )k = (z − z0 )−n k≥−n ck (z − z0 )k+n = (z − z0 )−n k≥0 ck−n (z − z0 )k . Conversely. where c−n = 0. and g is analytic on {z ∈ C : z − z0  < R} such that f = g in {z ∈ C : 0 < z − z0  < R}. Then by Proposition 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 88 Deﬁnition 9. that is. Proof. if the Laurent series of f at z0 has only nonnegative powers. . lim f (z) = lim g(z) = ∞.15 (uniqueness theorem for Laurent series) it has to coincide with the Laurent series of f . (b) z0 is a pole if and only if there are ﬁnitely many negative exponents. Proposition 9. (b) Suppose z0 is a pole of order n.3. By part (a). Then (a) z0 is removable if and only if there are no negative exponents (that is. Conversely. (c) z0 is essential if and only if there are inﬁnitely many negative exponents.CHAPTER 9.1. the function (z − z0 )n f (z) has a removable singularity at z0 . we can hence expand (z − z0 )n f (z) = k≥0 ck (z − z0 )k . Suppose z0 is an isolated singularity of f with Laurent series f (z) = k∈Z ck (z − z0 )k (valid in {z ∈ C : 0 < z − z0  < R} for some R > 0).1 is not always handy. (z − z0 )n z→z0 z→z0 (c) This follows by deﬁnition: an essential singularity is neither removable nor a pole. Then since g(z0 ) = c−n = 0.
. for k = −1 we can use Exercise 8 of Chapter 4. z = z0 ).4. (One might also notice that Theorem 8.1: Proof of Theorem 9. Deﬁnition 9. Finally.2. simple. Suppose z0 is an isolated singularity of f with Laurent series Then c−1 is the residue of f at z0 .9—we can integrate term by term: f= γ γ k∈Z ck (z − z0 )k dz = k∈Z ck γ (z − z0 )k dz . smooth path around z0 .) Reason enough to give the c−1 coeﬃcient of a Laurent series a special name. k∈Z ck (z − z0 ) k . The following theorem generalizes the discussion at the beginning of this section.2 Residues Suppose z0 is an isolated singularity of f . Then—essentially by Proposition 7. which lies in the domain of the Laurent series of f at z0 . closed. denoted by Resz=z0 (f (z)) or Res(f (z). z6 γ z2 z4 z1 z3 z5 Figure 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 89 9. and the same holds for k ≤ −2 (because (z − z0 )k has a primitive on C \ {z0 }).CHAPTER 9. The integrals inside the summation are easy: for nonnegative powers k the integral γ (z − z0 )k is 0 (because (z − z0 )k is entire). c−1 is the only term of the series which survives: f= γ k∈Z ck γ (z − z0 )k dz = 2πi c−1 . γ is a positively oriented.14 gives the same identity. Because all the other terms give a zero integral.
Hence f (z) f (z) = . by f (z0 ) . Computing integrals is as easy (or hard!) as computing residues. except for isolated singularities. and one with negative orientation. (n − 1)! z→z0 dz n−1 Proof. Now connect the circles with negative orientation with γ.4 (Residue Theorem). g (z0 ) Proof. We know by Proposition 9. turn. is the constant term of h or the second term of g.6. and f (z0 ) = 0. and we can use Taylor’s formula (Corollary 8.CHAPTER 9. Gcontractible curve. Suppose f and g are analytic in a region containing z0 . This gives a curve which is contractible in the region of analyticity of f . the residue of get the (−1)st term of f g ). Proof.3) to compute c−1 .3 that the Laurent series at z0 looks like f (z) = k≥−n ck (z − z0 )k . as pictured in Figure 9. and γ is a positively oriented. The functions f and g have power series centered at z0 . where the sum is taken over all singularities zk inside γ. But this means that we can replace γ by the positively oriented circles. . one with positive. Even more. Then Resz=z0 (f (z)) = 1 dn−1 lim (z − z0 )n f (z) .3) equals g (z0 ). Then f has a simple pole at z0 and g Resz=z0 f (z) g(z) = f (z0 ) . Then f = 2πi γ k Resz=zk (f (z)) . Each of these pairs cancel each other when we integrate over them. call it h. Lemma 9. Suppose f is analytic in the region G. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 90 Theorem 9. which by Taylor’s formula (Corollary 8. The following two lemmas start the range of tricks one can use when computing residues. simple. now all we need to do is described at the beginning of this section. Lemma 9. the one for g has by assumption no constant term: g(z) = ck (z − z0 )k = (z − z0 ) ck (z − z0 )k−1 . closed. g(z) (z − z0 )h(z) and the function f h is analytic at z0 . k≥1 k≥1 The series on the right represents an analytic function. smooth. But h(z0 ).1. Draw two circles around each isolated singularity inside γ. Suppose z0 is a pole of f of order n. f g equals the constant term of the power series of f (that’s how we h 0) as always. note that h(z0 ) = c1 = 0. which is a simple zero of g. in h(z But this constant term is computed.5. But then (z − z0 )n f (z) = k≥−n ck (z − z0 )k+n represents a power series.
nj . respectively. fg fg f g We can apply this fact to the following situation: Suppose that f is analytic on the region G. pk with order m1 . we may shrink G. + . . zj of order n1 . one of which we would like to discuss here. . Such functions are called meromorphic. . (9.1) f (z) z − p1 z − p2 z − pk g(z) where g is a function without poles in G. which is the starting point of the following theorem. Suppose the zeros of f inside γ are z1 . γ) the number of zeros of f inside γ—counted according to multiplicity— and by P (f. . . .. Denote by Z(f. respectively. we will study these functions.) In fact. . zj of order n1 . again counted according to multiplicity. Then 1 2πi f = Z(f. . Naturally. . where g is also analytic in G and never zero. . In this section. γ) . z − z1 z − z2 z − zj g(z) Something similar happens to the poles of f . . . . respectively. . . then the logarithmic derivative of f can be expressed as f (z) m1 m2 mk g (z) =− − − ··· − + . which—as the reader might have guessed already—can be thought of as siblings. γ) − P (f. so . especially with respect to their zeros and poles. . Let’s say we have two functions f and g analytic in some region. we can combine the expressions we got for zeros and poles. . and f has the (ﬁnitely many) zeros z1 . Theorem 9. . Diﬀerentiating Log f (where Log is a branch of the logarithm) gives f . which does not pass through any zero or pole of f . γ) the number of poles of f inside γ. which is one good reason why this quotient is called the logarithmic derivative f of f . It has some remarkable properties. . Suppose we have a diﬀerentiable function f . and the poles inside γ are p1 .3 Argument Principle and Rouch´’s Theorem e There are many situations where we want to restrict ourselves to functions which are analytic in some region except possibly for poles. Let’s compute the logarithmic derivative of f and play the same remarkable cancellation game as above: n1 (z − z1 )n1 −1 (z − z2 )n2 · · · (z − zj )nj g(z) + · · · + (z − z1 )n1 · · · (z − zj )nj g (z) f (z) = f (z) (z − z1 )n1 · · · (z − zj )nj g(z) nj n1 n2 g (z) = + + . pk are all the poles of f in G with order m1 . . . respectively. simple. . . Then we can express f as f (z) = (z − z1 )n1 · · · (z − zj )nj g(z) . nj . smooth. . .. . .CHAPTER 9. . We invite the reader to prove that if p1 . closed. f γ Proof. . if necessary. . . (You may meditate about the fact why there can only be ﬁnitely many zeros and poles inside γ. mk . mk . ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 91 9. Then the logarithmic derivative of their product behaves very nicely: (f g) f g + fg f g = = + .7 (Argument Principle). Suppose f is meromorphic in the region G and γ is a positively oriented. . Gcontractible curve.
But f has just a root of order 5 at the origin. Thanks to Exercise 8 of Chapter 4. Note also that there is no general formula for computing roots of a polynomial of degree 5. γ g As a nice application of the argument principle.4. So g and f satisfy the condition of the Theorem 9.stand. e Theorem 9.4) gives that g = 0. so that Corollary 4. Gcontractible curve such that for all z ∈ γ. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 92 that these are the only zeros and poles in G. see e http://wwwgroups. It allows us to locate the zeros of a function fairly precisely. f (z) > g(z). γ) . smooth. What’s special about the statement of Example 9.dcs. γ) = Z(f + g. we prove: Example 9. let f (z) = z 5 and g(z) = z 4 + z 3 + z 2 + z + 1.uk/∼history/Biographies/Rouche. whence Z(p. γ) = Z(f. Then for z ∈ γ g(z) ≤ z4 + z3 + z2 + z + 1 = 16 + 8 + 4 + 2 + 1 = 31 < 32 = z5 = f (z) . g Finally. closed. For more information about Rouch´. we present a famous theorem due to Eugene Rouch´ (1832–1910)3 . (Although for this p it’s not hard to ﬁnd one root—and therefore all of them.4 To see this.ac. and let γ denote the circle centered at the origin with radius 2. γ) = 5 .CHAPTER 9.8. As an illustration. g is analytic in G (recall that g is never zero in G).5 (to Cauchy’s g Theorem 4.4) asserts that p has ﬁve roots in C. simple. without poles) and never zero. 4 The fundamental theorem of algebra (Theorem 5.4 is that they all have absolute value < 2.8 (Rouch´’s Theorem).) 3 . Then Z(f + g. Our discussion before the statement of the theorem yielded that the logarithmic derivative of f can be expressed as nj f (z) m1 mk n1 g (z) + ··· + − − ··· − = + . f (z) z − z1 z − zj z − p1 z − pk g(z) where g is a function which is analytic in G (in particular. the integral is easy: f = n1 f dz + · · · + nj z − z1 dz − m1 z − zj γ γ γ γ γ dz − · · · − mk z − p1 γ dz + z − pk γ g g = 2πi (n1 + · · · + nj − m1 − · · · − mk ) + g . This theorem is of surprising practicality. Suppose f and g are analytic in a region G. γ) = Z(f.html. and γ is a e positively oriented. All the roots of the polynomial p(z) = z 5 + z 4 + z 3 + z 2 + z + 1 have absolute value less than two.
pk .3 that 1+ γ g f g +f 1 2πi 1 = 0. closed. 6. 3. let f (z) = an z n and g(z) = an−1 z n−1 + an−2 z n−2 + · · · + a1 z + 1. use Theorem 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 93 Proof of Theorem 9. . Suppose f is meromorphic in the region G. γ) = + g 2πi γ f + g 2πi γ f 1 + g 2πi γ f 1+ f f 1 = Z(f. (Hint: If p(z) = an z n + an−1 z n−1 + · · · + a1 z + 1. Find the number of zeros of (a) 3 exp z − z in {z ∈ C : z ≤ 1} . Exercises 1. simple. Give another proof of the fundamental theorem of algebra (Theorem 5. .) 4. Prove (9. using Rouch´’s e Theorem 9. . and γ is a positively oriented. . . respectively.2 for f z . You might want to ﬁrst apply Lemma 5. 3 (c) z 4 − 5z + 1 in {z ∈ C : 1 ≤ z ≤ 2} .8. . γ) + 2πi We are assuming that g f 1+ γ 1 g f g +f . which does not pass through any zero or pole of f . Denote the zeros and poles of f inside γ by z1 .4). zj and p1 . . Show that if f has an essential singularity at z0 then 1 f also has an essential singularity at z0 .3 e to g(z). 2. g f < 1 on γ. By our analysis in the beginning of this section and by the argument principle (Theorem 9.8. . Prove that any complex number is arbitrarily 1 close to a number in f (C).) g γ f = f j k g(zm ) − m=1 n=1 g(pn ) . (b) 1 exp z − z in {z ∈ C : z ≤ 1} .1). counted according to multiplicity. g is analytic in G. and choose as γ a circle which is large enough to make the condition of Rouch´’s theorem work.CHAPTER 9. .7) g g f 1+ f 1+ f (f + g) 1 1 1 f = = Z(f + g. Suppose f is a nonconstant entire function. Prove that 1 2πi 5. which means that the function 1 + g f evaluated on γ stays away from the nonpositive real axis. Gcontractible curve. But then Log 1 + 1+ Its derivative is 1 g f g +f is a well deﬁned analytic function on γ. . which implies by Corollary 4. (Hint: If f is not a polynomial.
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM 7. (a) Find a Laurent series for it converges. (b) Compute
dz γ (z 2 −4)(z−2) , 1 (z 2 −4)(z−2)
94
centered at z = 2 and specify the region in which
where γ is the positively oriented circle centered at 2 of radius 1.
8. Evaluate the following integrals for γ(t) = 3 eit , 0 ≤ t ≤ 2π. (a)
γ
cot z dz z 3 cos
γ 3 z
(b) (c)
γ
dz
dz (z + 4)(z 2 + 1) z 2 exp
1 z
(d)
γ
dz
(e)
γ
exp z dz sinh z iz+4 dz (z 2 + 16)2
exp z γ (z+1)34
(f)
γ
9. (a) Find the power series of exp z centered at z = −1. (b) Find dz, where γ is the circle z + 2 = 2, positively oriented.
10. Suppose f has a simple pole (i.e., a pole of order 1) at z0 and g is analytic at z0 . Prove that Resz=z0 f (z)g(z) = g(z0 ) · Resz=z0 f (z) . 11. Find the residue of each function at 0: (a) z −3 cos(z). (b) csc(z). (c) z 2 + 4z + 5 . z2 + z
1
(d) e1− z . (e) e4z − 1 . sin2 z
12. Use residues to evaluate the following: (a)
γ
z4
dz , where γ is the circle z + 1 − i = 1. +4 dz , where γ is the circle z − i = 2. + z − 2)
(b)
γ
z(z 2
(c)
γ
ez dz , where γ is the circle z = 2. z3 + z
CHAPTER 9. ISOLATED SINGULARITIES AND THE RESIDUE THEOREM (d)
γ
95
dz z 2 sin z
, where γ is the circle z = 1.
13. Suppose f has an isolated singularity at z0 . (a) Show that f also has an isolated singularity at z0 . (b) Find Resz=z0 (f ). 14. Given R > 0, let γR be the half circle deﬁned by γR (t) = Reit , 0 ≤ t ≤ π, and ΓR be the closed curve composed of γR and the line segment [−R, R]. (a) Compute
dz ΓR (1+z 2 )2
.
dz γR (1+z 2 )2
(b) Prove that limR→∞
=0.
∞ dx −∞ (1+x2 )2
(c) Combine (a) and (b) to evaluate the real integral
.
15. Suppose f is entire, and a, b ∈ C with a, b < R. Let γ be the circle centered at 0 with radius R. Evaluate f (z) dz , γ (z − a)(z − b) and use this to give an alternate proof of Liouville’s Theorem 5.5. (Hint: Show that if f is bounded then the above integral goes to zero as R increases.)
Chapter 10
Discreet Applications of the Residue Theorem
All means (even continuous) sanctify the discrete end. Doron Zeilberger
On the surface, this chapter is just a collection of exercises. They are more involved than any of the ones we’ve given so far at the end of each chapter, which is one reason why we lead the reader through each of the following ones step by step. On the other hand, these sections should really be thought of as a continuation of the lecture notes, just in a diﬀerent format. All of the following ‘problems’ are of a discrete mathematical nature, and we invite the reader to solve them using continuous methods—namely, complex integration. It might be that there is no other result which so intimately combines discrete and continuous mathematics as does the Residue Theorem 9.4.
10.1
Inﬁnite Sums
1 k2
In this exercise, we evaluate—as an example—the sums k≥1 idea how to compute such sums in general will become clear. 1. Consider the function f (z) =
and
k≥1
(−1)k . k2
We hope the
π cot(πz) . Compute the residues at all the singularities of f . z2
2. Let N be a positive integer and γN be the rectangular curve from N +1/2−iN to N +1/2+iN to −N − 1/2 + iN to −N − 1/2 − iN back to N + 1/2 − iN . (a) Show that for all z ∈ γN ,  cot(πz) < 2. (Use Exercise 21 in Chapter 3.) (b) Show that limN →∞ γN f = 0. 3. Use the Residue Theorem 9.4 to arrive at an identity for 4. Evaluate
1 k≥1 k2 . π z 2 sin(πz) 1 k∈Z\{0} k2 .
5. Repeat the exercise with the function f (z) =
to arrive at an evaluation of
k≥1
(−1)k . k2
96
y ∈ C and n ∈ N. f1 = 1. (x + y)n = n . 10. k4 10. fn = fn−1 + fn−2 Let F (z) = 1 for n ≥ 2. As an example. Convince yourself that 2k k 1 x = k 2πi (1 + w)2k k dw x . k≥0 fn z n.2 Binomial Coeﬃcients The binomial coeﬃcient n is a natural candidate for being explored analytically. k 1 − 4x k≥0 1. 2.) 6. w wk k≥0 k≥0 γ use 2. we outline a proof of k the identity (for −1/4 < x < 1/4) 2k k 1 x =√ . you may use the fact that 1/ sin2 z = 1 + cot2 z.uk/∼history/Biographies/Fibonacci. Convince yourself that 2k k = 1 2πi γ (1 + w)2k dw . wk+1 where γ is any simple closed curve such that 0 is inside γ. Evaluate 1 k≥1 k4 97 and k≥1 (−1)k . Evaluate this sum.ac.html. DISCREET APPLICATIONS OF THE RESIDUE THEOREM (Hint: To bound this function. k=0 k x y 2 For more information about Leonardo Pisano Fibonacci (1170–1250).CHAPTER 10.3 Fibonacci Numbers The Fibonacci2 numbers are a sequence of integers deﬁned recursively as: f0 = 1.stand. .dcs. to interchange summation and integral. `n´ k n−k P The binomial theorem says that for x. as the binomial k theorem1 tells us that n is the coeﬃcient of z k in (1 + z)n .4 to evaluate the integral. 3. and use the Residue Theorem 9. see http://wwwgroups. Find a simple closed curve γ surrounding the origin such that (1 + w)2 x w k k≥0 converges uniformly on γ (as a function in w). Suppose x < 1/4.
The fractional part is then {x} = x − x . is the greatest integer not exceeding x.ac. f (z) = a ) (1 − z b ) z t+1 (1 − z 1. 3. ma + n = t} = # {m ∈ Z : m ≥ 0.uk/∼history/Biographies/Frobenius.) 4. 6 This means that a−1 is an integer such that a−1 a = 1 + kb for some k ∈ Z. n ≥ 0. 98 1 2. Compute the residues at all nonzero poles of f . n) ∈ Z : m. n ≥ 0. ma + nb = t} . For more information about Frobenius.) 3. n) ∈ Z : m. ma ≤ t} =# 3 0. 1 4. 4 this means that the integers don’t have any common factor 5 The fractional part of a real number x is.html. 2. Suppose a and b are relatively prime4 positive integers. Verify that Resz=0 (f ) = N (t). see http://wwwgroups. Show that the recurrence relation among the fn implies that F (z) = 1−z−z 2 . Use the following three steps to simplify this identity to N (t) = t − ab b−1 t a − a−1 t b + 1.dcs. and show that this integral vanishes as R → ∞. where N (t) = # {(m. Consider the function 1 .) 5. Here. and b−1 b ≡ 1 (mod a). {x} denotes the fractional part5 of x.” More thoroughly. the “part after the decimal point. we will solve and extend a classical problem of Ferdinand Georg Frobenius (1849– 1917)3 . (Hint: Integrate z n+1 (1−z−z 2 ) around a circle with center 0 and radius R. and a−1 a ≡ 1 (mod b)6 . DISCREET APPLICATIONS OF THE RESIDUE THEOREM 1.4 The ‘CoinExchange Problem’ In this exercise. t ∩Z a = t − a t a + 1. Show that F has a positive radius of convergence.CHAPTER 10. Use the Residue Theorem 9. loosely speaking.4 to derive an identity for N (t). N (t) = # {(m. denoted by x . 10.4 to derive an identity for fn . (a) Verify that for b = 1. (Hint: Write down the power series of zF (z) and z 2 F (z) and rearrange both so that you can easily add. Verify that Resz=0 1 z n+1 (1−z−z 2 ) = fn . the greatest integer function of x. and t is a positive integer. . Use the Residue Theorem 9. Generalize to other recurrence relations. and show that this integral vanishes as R → ∞. (Hint: Integrate f around a circle with center 0 and radius R.stand.
Given relatively prime positive integers a1 . More generally. 5. The formula in 4. In the late 19th century.4. . let’s call an integer t representable if there exist nonnegative integers m1 .stand. . and N (t) > k for all t > (k + 1)ab − a − b. For n > 2. . . a2 ) = a1 a2 − a1 − a2 .dcs. Given two positive. . . mn such that n t= j=1 mj aj .5 Dedekind sums This exercise outlines yet another nontraditional application of the Residue Theorem 9. see http://wwwgroups. We veriﬁed this result in 5. . It is well known (probably at least since the 1880’s. . . an ).uk/∼history/Biographies/Sylvester. an . DISCREET APPLICATIONS OF THE RESIDUE THEOREM (b) Use this together with the identity found in 3. 6. Historical remark. is due to Popoviciu. let f (z) = cot(πaz) cot(πbz) cot(πz) . an ).ac.html. (a) Compute the residues for the poles of f inside γR . Choose an > 0 such that the rectangular path γR from 1 − − iR to 1 − + iR to − + iR to − − iR back to 1 − − iR does not pass through any of the poles of f . relatively prime integers a and b. when James Joseph Sylvester (1814–1897)7 studied the Frobenius problem) that g(a1 . Hint: use the periodicity of the cotangent and the fact that cot z = 7 1 1 − z + higherorder terms . can only be found in the most recent literature. . . The notion of an integer being representable k times and the respective formula obtained in 6. . 10. and N (t) > 0 for all t > ab − a − b. to obtain 1 a (c) Verify that a−1 k=1 a−1 k=1 99 1 (1 − e2πik/a )e2πikt/a =− t a + 1 1 − . . Prove that N (ab − a − b) = 0. 2 2a 1 (1 − e2πikb/a )e2πikt/a a−1 = k=1 1 (1 − e2πik/a )e2πikb−1 t/a . We call this largest integer the Frobenius number g(a1 . prove that. . Frobenius raised the problem of ﬁnding the largest integer which is not representable. 1. . there is no known closed formula for g(a1 . z 3 For more information about Sylvester.CHAPTER 10. . . N ((k + 1)ab − a − b) = k. if k is a nonnegative integer.
dcs.4 to show that 1 1 s(a.stand.html. Deﬁne 1 s(a. (10. DISCREET APPLICATIONS OF THE RESIDUE THEOREM (b) Prove that limR→∞ γR 100 f = −2i and deduce that for any R > 0 f = −2i .1) and (10. It ﬁrst appeared in the study of the Dedekind ηfunction η(z) = exp πiz (1 − exp(2πikz)) 12 k≥1 a 1 b + + b ab a .stand. see http://wwwgroups. The reciprocity law (10. γR 2. 8 . The sum (10. b) = 4b b−1 cot k=1 πka b cot πk b . The proof that is outlined here is due to Hans Rademacher (1892–1969)9 . For more information about Julius Wilhelm Richard Dedekind (1831–1916).2)? Historical remark.2) in the 1870’s and has since intrigued mathematicians from such diﬀerent areas as topology. and discrete geometry.uk/∼history/Biographies/Rademacher.ac. (10. a) = − + 4 12 3. Can you generalize (10.uk/∼history/Biographies/Dedekind.1) is called a Dedekind8 sum.2) is the most important and famous identity of the Dedekind sum.1) Use the Residue Theorem 9. 9 For more information about Rademacher. number theory. see http://wwwgroups.CHAPTER 10.ac.dcs.html. b) + s(b.
k ∈ Z. −i if n = 3 + 4k. 5 π π (b) z = 2ei 4 + 2 k . 3 7. k ∈ Z. 8i π 3. (a) diﬀerentiable and analytic in C with derivative −e−x e−iy (b) nowhere diﬀerentiable or analytic (c) diﬀerentiable on {x + iy ∈ C : x = y} with derivative 2x. i if n = 1 + 4k. 5 − 10i √ √ i 3 (c) 10 . √ 2. (b) 19 − 25 i 25 (c) 1 (d) 1 if n = 4k. k ∈ Z. (a) 2ei 2 √ iπ (b) 2e 4 √ 5π (c) 2 3ei 6 4. 1. . nowhere analytic 101 . (a) diﬀerentiable at 0. 2. 11 ( 2 − 1) + 11 ( 2 + 9) 11 (d) 8. k = 0. k ∈ Z. . nowhere analytic (i) diﬀerentiable and analytic in C with derivative −2iz Chapter 3 26. (a) 0 (b) 1 + i 10. nowhere analytic (d) nowhere diﬀerentiable or analytic (e) diﬀerentiable and analytic in C with derivative − sin x cosh y − i cos x sinh y (f) diﬀerentiable at 0 with derivative 0. k = 0. (a) −1 + i (b) 34i (c) −1 π 5. (a) z = ei 3 k . (a) 5. 1. nowhere analytic (h) diﬀerentiable only at i with derivative i. −1 if n = 2 + 4k. .Solutions to Selected Exercises Chapter 1 8 1. nowhere analytic (g) diﬀerentiable at 0 with derivative 0. −2 − i √ (b) 5 5. z = ei 4 − 1 and z = ei π 5π 4 −1 Chapter 2 2. .
y = 0} (f) diﬀerentiable and analytic in C (i. (a) 0 (b) 2πi (c) 0 (d) πi (e) 0 (f) 0 5. (a) z = i (b) There is no solution. entire) 27. 2πi for r > a 24 0 for r = 1. 0]. uniform on H ∩ {z ≥ R} for any ﬁxed R > 0. (a) 8πi (b) 0 (c) 0 (d) 0 14. 0 2π 16.SOLUTIONS TO SELECTED EXERCISES (b) diﬀerentiable and analytic on C \ −1. C \ (−∞. (a) pointwise convergent for z < 1. π π 102 Chapter 7 1. ei 3 . k ∈ Z (g) z = 2i 30. for example. − πi for r = 3. . uniform for z ≤ R for any ﬁxed R < 1 (b) pointwise and uniformly convergent for z ≤ 1 (c) pointwise convergent for all z ∈ H := {Re z ≥ 0}. k ∈ Z 2 (f) z = πk. e−i 3 (c) diﬀerentiable and analytic on C \ {x + iy ∈ C : x ≥ −1. y = 2} (d) nowhere diﬀerentiable or analytic (e) diﬀerentiable and analytic on C \ {x + iy ∈ C : x ≤ 3. k ∈ Z 2 (d) z = π + 2πk ± 4i.e. √3 23 0 for r < a. (c) z = ln π + i π + 2πk . −2πi 3. k ∈ Z 2 (e) z = π + πk. Any simply connected set which does not contain the origin. 0 for r = 5 3 Chapter 5 2. (a) divergent (b) convergent (limit 0) (c) divergent i (d) convergent (limit 2 − 2 ) (e) convergent (limit 0) 16. f (z) = c z c−1 Chapter 4 2.
(a) k≥0 e1 (z + 1)k k! (b) e2πi 33! 14. One Laurent series is 1 1 7 15. 20. (a) 0 (b) 1 (c) 4 (−1)k (2k)! k −k−3 . sin z = z −1 + 6 z + 360 z 3 + . {z ∈ C : r ≤ z − 3 ≤ R} for any 1 < r ≤ R e k 3. z 2k−2 7. converging for z k≥0 (−2) (z − 2) −3(z + 1)−1 + 1. (a) {z ∈ C : z < 1}. 12. (c) π 2 (−1)k k≥−2 4k+3 (z − 2)k . . 13. The maximum is 3 (attained at z = ±i). (a) 2πi (b) 27πi 4 (c) − 2πi 17 (d) πi 3 (e) 2πi (f) 0 9. . converging for 0 < z − 2 < 4. (a) (b) 23.SOLUTIONS TO SELECTED EXERCISES 18. . (a) ∞ if a < 1. (a) k≥0 (−4)k z k 1 (b) k≥0 3·6k z k 20. 1 if a = 1. {z ∈ C : z ≤ r} for any r < 1 (b) C. (a) One Laurent series is (b) − πi 8 8. converging for z − 1 > 2. (a) k≥0 Chapter 9 5. (b) 1 (c) 1 (careful reasoning!) (d) 1 (careful reasoning!) Chapter 8 1. {z ∈ C : z ≤ r} for any r (c) {z ∈ C : z − 3 > 1}. k k k≥0 (−1) (z − 1) (−1)k−1 (z − 1)k k≥1 k 103 − 2 > 2. and the minimum is 1 (attained at z = ±1). One Laurent series is 14. One Laurent series is k≥0 (−2)k (z − 1)−k−2 . converging for z = −1. k≥0 k! (z − 1) 10. and 0 if a > 1.
40 Cauchy–Riemann equations. 93 of calculus. 32 embedding of R in C. 2 harmonic. 2 real. 15 antiderivative. 46 extensions of. 15 diﬀerentiation rule. 31 argument. 1 Frobenius problem. 9 geometric series. 7 coﬀee. 13 fundamental theorem of algebra. 79 e. 39. 62 series. 97 boundary. 41 convergent sequence. 87 derivative. 2 abelian. 16 dilation. 61 distance of numbers. 5 connected. 58 homotopic. 62 domain. 87 conjugate. 1 analytic. 85 exponential function. 18. 3 addition. 48. 51. 92. 64 cosine. 65 absolute value. 6 divergent. 3. 8 closed set. 76 Cauchy’s integral formula.Index absolute convergence. 13 double series. 14 contractible. 56 essential singularity. 97 ﬁeld. 15 diﬀerence quotient. 3 axis imaginary. 98 function. 15. 29 cotangent. 2 entire. 57 harmonic conjugate. 31 CasoratiWeierstraß theorem. 16 binomial coeﬃcient. 24 Dirichlet problem. 51 curve. 15 diﬀerentiable. 60 branch of the logarithm. 42. 40 homotopy. 29 curve. 31 arg. 2 bijection. 17 chain rule. 76 Cauchy’s theorem. 29 Fibonacci numbers. 40 104 . 87 Cauchy’s estimate. 99 dense. 51. 53 Arg. 83. 29 exponential rules. 7 continuous. 68 group. 8 Dedekind sum. 16 closed algebraically.
51. 23 o smooth. 8 strong relative.INDEX hyperbolic trig functions. 2 identity map. 80 Leibniz’s rule. 8 modulus. 79 sine. 75 . 7 maximum region of convergence. 87 piecewise smooth. 31 log. 37 path. 37 integration by parts. 69 strong relative. 24 isolated singularity. 59. 39. 59 separated. 31 principal logarithm. 23 Log. 59 removable singularity. 5 pole. 60. 56 power series. 90 meanvalue theorem reverse triangle inequality. 64 minimum simple closed curve. 7 for real functions. 53 principal argument. 62 meromorphic. 79 residue theorem. 64 linear fractional transformation. 15 of a sequence. 13 image. rectangular form. 31 logarithmic derivative. 8 path independent. 85 weak relative. 16 open set. 53 weak relative. 16 inversion. 91 series. 44 Rouch´’s theorem. 66 polar form. 92 e for harmonic functions. 74 integration of. 85 polynomial. 38 limit of a function. 89 maximummodulus theorem. 11. 86 parametrization. 40 periodic. 13. 91 obvious. 30 i. 85 Laplace equation. 62 of a series. 23 onetoone. 13. 29 Picard’s theorem. 82 singularity. 10 length. 79 residue. 46 inverse function. 54 tangent. 85 M¨bius transformation. 37 pointwise convergence. 57 Laurent series. 59 simply connected. 16 onto. 2 integral. 69 primitive. 31 logarithm. 9 sequence. 5 79 region. 7 order of a pole. 3 Morera’s theorem. 13 imaginary part. 68 diﬀerentiation of. 2 max/min property for harmonic functions. 29 minimummodulus theorem. 21. 1 Taylor series expansion. 29 multiplication. 11 for analytic functions. 32 105 real part. 31 principal value of ab .
6 translation. 5 trigonometric functions. 14 uniform convergence. 66 uniqueness theorem. 82 106 . 78 Weierstraß M test. 68 Weierstraß convergence theorem.INDEX topology. 29 trivial. 24 triangle inequality.
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.