Complex analysis

Lecture notes, Fall 2024

(Version: December 23, 2024)

Pu-Zhao Kow
D EPARTMENT OF M ATHEMATICAL S CIENCES , NATIONAL C HENGCHI U NIVERSITY
Email address: [email protected]
 Preface

The main theme of this lecture note is to explain

• the equivalence of analyticity (complex differentiability), Cauchy-Riemann equation, and
the power series representation;
• Morera theorems (Section 4.7.1); and
• the Cauchy residual theorem (Theorem 5.3.6).
This lecture note is prepared for the course Complex Analysis during Fall Semester 2024 (113-1),
which gives an introduction to complex numbers and functions, mainly based on [BN10], but not
following the order. The e-book is available in https://www.lib.nccu.edu.tw (NCCU library).
The lecture note may updated during the course. Previously, this lecture note also used during the
Fall Semester 2023 (112-1).

Title. Complex Analysis (Fall Semester 2024, 3 credits)

Lectures. Wednesday (10:10–12:00) and Friday (10:10–11:00). Begins at September 9, 2024 and
ends at January 10, 2025.

Language. Chinese and English. Materials will be prepared in English.

Instructor. Pu-Zhao Kow (Email: [email protected])

Office hour. Wednesday (12:10–13:00)

TA class. Friday (11:10–12:00)

Completion. Homework Assignments 60%, Midterm Exam 20% (hope that at least Chapter 1–
Chapter 4 are covered), Final Exam 20%

Acknowledgments. I would like to give special thanks to students who pointed out my mistakes
in this note.

Exam rules.
(1) Textbook or other materials are not allowed to be used during exams.
(2) All electronic devices (including calculator, smartphone, pad, computer, ...) are prohibited
during exams.
(3) One also not allowed to bring your own extra paper. TA will provide answer sheets.
(4) Before go to washroom, one must inform us before do so.
i
 PREFACE ii

(5) If you violate one of the above rule, we will immednate terminate your writing and the
marks of the exam/quiz will be 0.
(6) One must show student card or national identity card or national health insurance card or
passport or resident certificate (driving license not accepted) for verification. Before the
exam begins, TA should reminds all of you to bring it. If one fails to show it during exam,
we consider this as a cheating and the marks of the quiz will be 0.
 Contents

Chapter 1. The complex numbers 1

1.1. Definition of complex plane C 1
1.2. Topological aspects of C 5

Chapter 2. Differentiation 14
2.1. Complex derivative and Cauchy-Riemann equation 14
2.2. Power series 19
2.3. Exponential, sine and cosine functions 25

Chapter 3. Integration 28
3.1. The fundamental theorem of line integral 28
3.2. Cauchy closed curve theorem in rectangle 32
3.3. Cauchy closed curve theorem in simply connected open sets 37

Chapter 4. Properties of Analytic functions 41

4.1. Cauchy integral formula for entire functions 41
4.2. Power series (with R = ∞) and entire function 44
4.3. Liouville theorem and the fundamental theorem of algebra 46
4.4. The roots of ±1 48
4.5. Cauchy integral formula in a ball 49
4.6. Power series (with R < ∞) and analytic function 50
4.7. Morera Theorems 53

Chapter 5. Laurent series and the Cauchy residual theorem 56

5.1. Riemann’s principle of removable singularities 56
5.2. Laurent expansions 58
5.3. Winding numbers and the Cauchy residue theorem 64
5.4. Some applications in combinatorics : Egorychev method 69

Chapter 6. Some special analytic functions 76

6.1. The analytic function log z 76
6.2. Infinite products and Weierstrass product theorem 79
6.3. The Gamma function: an extension of factorial function 86
iii
 CONTENTS iv

Bibliography 91
 CHAPTER 1

The complex numbers

1.1. Definition of complex plane C

We shall introduce the complex plane using a rather simple (and direct) way. Given a number
√
x ∈ R≥0 , it is well-known that the square root x of x is well-defined, which satisfies
√ √ √
(1.1.1) ( x)2 = x · x = x for all x ≥ 0.

This arises a natural question: It is possible to extend (1.1.1) for all x ∈ R? Or, we shall ask: How
√
to define i ≡ −1? Clearly, we should expect that

(1.1.2) i2 = −1.

We will answer this question in Section 4.4.

Formally, we expect the linearity

(1.1.3a) (a + ib) + (c + id) = (a + c) + i(b + d) for all a, b, c, d ∈ R.

By using the formal identity (1.1.2), we also formally computed that

(a + ib) · (c + id) = ac + ibc + iad + i2 bd

(1.1.3b) = (ac − bd) + i(ad + bc) for all a, b, c, d ∈ R.

At this point, we not yet define the element i, therefore the identities (1.1.3a)–(1.1.3b) are not yet
well-defined. However, we can rephrase (1.1.3a)–(1.1.3b) without involving the formal element i
(which is not yet well-defined).

D EFINITION 1.1.1. We define the set C := R × R ≡ {(x, y) : x, y ∈ R}. We define the binomial
operations “+” and “·” on C by

(1.1.4a) C + C → C, (a, b) + (c, d) := (a + c, b + d),

(1.1.4b) C · C → C, (a, b) · (c, d) := (ac − bd, ad + bc).

P ROPOSITION 1.1.2. (C, +, ·) is a field with additive identity (0, 0) and multiplicative identity
(1, 0).

R EMARK 1.1.3. The main point here is to define what is the meaning of “divide an element by
another element”. Here the multiplication is sometimes called the complex multiplication, not the
1
 1.1. DEFINITION OF COMPLEX PLANE C 2

inner product of Rn . While reading research articles, remember to make sure the definition of the
multiplication (for example, the · in the CGO solution means inner product [Sal08]).

P ROOF OF P ROPOSITION 1.1.2. Verify (C, +) forms a commutative group with additivity
identity (0, 0).
Additive associativity. ((a1 , b1 ) + (a2 , b2 )) + (a3 , b3 ) = (a1 , b1 ) + ((a2 , b2 ) + (a3 , b3 ))
Additive identity. (a, b) + (0, 0) = (0, 0) + (a, b) = (a, b)
Additive inverse element. One can easily verify that the additive inverse of (a, b) is (−a, −b):

(a, b) + (−a, −b) = (−a, −b) + (a, b) = (0, 0).

In other words, −(a, b) = (−a, −b).

Additive commutative. (a, b) + (c, d) = (c, d) + (a, b)
We now verify some properties of the multiplication operator.
Multiplicative associativity. One can directly verify that

((a1 , b1 ) · (a2 , b2 )) · (a3 , b3 )

= (a1 a2 − b1 b2 , a1 b2 + a2 b1 ) · (a3 , b3 )

1 2 3 4
= ( a1 a2 a3 − a3 b1 b2 − a1 b2 b3 − a2 b1 b3 , a1 a2 b3 − b1 b2 b3 + a1 a3 b2 + a2 a3 b1 ),
5 6 7 8
and

(a1 , b1 ) · ((a2 , b2 ) · (a3 , b3 ))

= (a1 , b1 ) · (a2 a3 − b2 b3 , a2 b3 + a3 b2 )

1 3 4 2
= ( a1 a2 a3 − a1 b2 b3 − a2 b1 b3 − a3 b1 b2 , a1 a2 b3 + a1 a3 b2 + a2 a3 b1 − b1 b2 b3 ),
5 7 8 6
therefore ((a1 , b1 ) · (a2 , b2 )) · (a3 , b3 ) = (a1 , b1 ) · ((a2 , b2 ) · (a3 , b3 )).
Multiplicative identity. (a, b) · (1, 0) = (1, 0) · (a, b) = (a, b)
Multiplicative inverse element. For each (a, b) ̸= (0, 0), we define
 
−1 a −b
(1.1.5) (a, b) := , .
a2 + b2 a2 + b2
We see that
 
−1 a −b a −b
(a, b) · (a, b) = a· 2 2
−b· 2 2
,b· 2 2
+a· 2 = (1, 0)
a +b a +b a +b a + b2
as well as (a, b)−1 · (a, b) = (1, 0).
Multiplicative commutative. (a, b) · (c, d) = (c, d) · (a, b)
 1.1. DEFINITION OF COMPLEX PLANE C 3

The above four axioms imply that (C \ {(0, 0)}, ·) forms a commutative group. We have one
more axiom to verify:
Distributive laws. This properties describe how the additive operator and multiplicative operator
act together. We compute that

(a1 , b1 ) · ((a2 , b2 ) + (a3 , b3 ))

= (a1 , b1 ) · (a2 + a3 .b2 + b3 )
= (a1 a2 + a1 a3 − b1 b2 − b1 b3 , a1 b2 + a1 b3 + b1 a2 + b1 a3 )
= (a1 a2 − b1 b2 , a1 b2 + b1 a2 ) + (a1 a3 − b1 b3 , a1 b3 + b1 a3 )
= (a1 , b1 ) · (a2 , b2 ) + (a1 , b1 ) · (a3 , b3 )

and the multiplicative commutative also gives us that

((a2 , b2 ) + (a3 , b3 )) · (a1 , b1 ) = (a2 , b2 ) · (a1 , b1 ) + (a3 , b3 ) · (a1 , b1 ).

We conclude our proposition. □

The additive identity (0, 0) is unique: If (0′ , 0′′ ) is also an additive identity, then

(0, 0) = (0′ , 0′′ ) + (0, 0) = (0, 0) + (0′ , 0′′ ) = (0′ , 0′′ ).

Similar argument also shows that the multiplicative identity (1, 0) is unique. In the context of
abstract algebra, we sometimes called (0, 0) the zero, and called (1, 0) the one. We now define the
“mysterious” element i rigorously.

D EFINITION 1.1.4. We define i := (0, 1) ∈ C, and we call it the imaginary unit.

Obviously,

(1.1.6) (a, 0) · (x, y) = (ax, ay),

and therefore in particular,

(a, 0) · (b, 0) = (ab, 0).
In addition, we have
(a, 0) + (b, 0) = (a + b, 0).
Therefore, the mapping
ι : R → {(a, 0) : a ∈ R} , a 7→ (a, 0)
is a field isomorphism. Therefore, we somehow abuse the notation by simply writing

R ≡ {(a, 0) : a ∈ R} , 1 ≡ (1, 0).

Since
(x, y) = (x, 0) + (0, y) = (x, 0) · (1, 0) + (y, 0) · (0, 1),
 1.1. DEFINITION OF COMPLEX PLANE C 4

then we see that:

L EMMA 1.1.5. Each complex number z = (x, y) can be written uniquely in the form z = x + iy.
The map
R × R → C, (x, y) 7→ x + yi
is a bijection.

D EFINITION 1.1.6. If z = x + iy, then we write Re z := x (the real part of z) and Im z := y

(the imaginary part of z). Note that Im z is a real number. We also define the conjugate z of z by
z = x − iy.

It is useful to observe that

z + z = 2x, z − z = 2iy,
therefore,
1 1 i−1
(1.1.7) Re z = (z + z), Im z = (z − z) ≡ (z − z),
2 2i 2
where i−1 is given by (1.1.5). From (1.1.3b), it is also useful to see that

(a + ib) · (c + id)
= (ac − bd) + i(ad + bc)
= (ac − bd) − i(ad + bc)
= (ac − (−b)(−d)) + i(a(−d) + (−b)c)
= (a − ib) · (c − id)
= (a + ib) · (c + id),

that is,

(1.1.8) zw = z · w for all z, w ∈ C.

Therefore, one also can write (1.1.6) as

a(x + iy) = ax + iay.

One can directly verify that

(±i)2 = ±i · ±i = (0, ±1) · (0, ±1) = (−1, 0) ≡ −1.

This somehow suggests (1.1.2). At this moment, we first keep this question in mind, we will come
back to answer this later.
 1.2. TOPOLOGICAL ASPECTS OF C 5

1.2. Topological aspects of C

We now discuss the topological aspects of the complex plane, in other words, we want to discuss
how the open sets in C looks like and define the continuous functions on C. Here we also refer to
the monograph [Mun00] for general abstract theory.

1.2.1. Sequences in C. In this section, we shall see that there are many facts in calculus also
holds true for complex numbers.

D EFINITION 1.2.1. The absolute value (or modulus) |z| of z, is defined by

√ q
|z| := zz ≡ (Re z)2 + (Im z)2 ≡ ∥(Re z, Im z)∥R2 ,

which is just simply the Euclidean norm in R2 .

It is not difficult to see the absolute homogeneity (i.e. |rz| = |r||z| for all r ∈ R) and positive
definiteness of | · | (i.e. |z| ≥ 0 and the equality holds if and only if z = 0). To verify that | · | is a
norm, we only need to verify the following:

L EMMA 1.2.2 (Triangle inequality, subadditivity). |z1 ± z2 | ≤ |z1 | + |z2 | for all z1 , z2 ∈ C.

P ROOF. We now define the inner product on R2 ∼

= C by

⟨z1 , z2 ⟩ := (Re z1 )(Re z2 ) + (Im z1 )(Im z2 ) for all z1 , z2 ∈ C.

One sees that ⟨z, z⟩ = (Re z)2 + (Im z)2 = |z|2 . We also see that

|z1 ± z2 |2 = ⟨z1 ± z2 , z1 ± z2 ⟩ = |z1 |2 ± 2⟨z1 , z2 ⟩ + |z2 |2

(|z1 | + |z2 |)2 = |z1 |2 + 2|z1 ||z2 | + |z2 |2 ,

therefore it is suffice to show the following Cauchy Schwartz inequality

±⟨z1 , z2 ⟩ ≤ |z1 ||z2 | equivalently, |⟨z1 , z2 ⟩| ≤ |z1 ||z2 |.

In fact, we only need to prove the above inequality for the case both z1 ̸= 0 and z2 ̸= 0. In this case,
by writing w1 := |zz11 | and w2 := |zz22 | , we only need to prove

(1.2.1) ±⟨w1 , w2 ⟩ ≤ 1.

Since |w1 | = |w2 | = 1, then

0 ≤ |w1 ± w2 |2 = ⟨w1 ± w2 , w1 ± w2 ⟩ = 2 ± 2⟨w1 , w2 ⟩,

which concludes (1.2.1). □

R EMARK 1.2.3. We recall that (C, +, ·) forms a field, where · represents the complex
multiplication. As a comparison, (R2 , +, ⟨·, ·⟩) forms a ring, but not a field. Roughly speaking,
we cannot define quotient for inner product, but we can define quotient for complex multiplication.
 1.2. TOPOLOGICAL ASPECTS OF C 6

D EFINITION 1.2.4. The sequence {zn }n∈N converges to z in C if the sequence of real numbers
|zn − z| converges to 0. Precisely, given any ε > 0, there exists N > 0 such that |z − zn | < ε for all
n ≥ N.

E XERCISE 1.2.5. Show that

max{|Re z|, |Im z|} ≤ |z| ≤ |Re z| + |Im z|.

From this, one can easily see that zn → z if and only if Re zn → Re z and Im zn → Im z.

We also can rephrase the above definition in a more geometric terms:

(1.2.2) Given any ε > 0, there exists N > 0 such that zn ∈ Bε (z) for all n ≥ N,

where Br (z) is the ball in R2 with radius r and centered at z. In the context of complex analysis,
some authors refer Br (z) the disk.
While taking limit, we always need to check whether it exists or not, which is very inconvenient.
For future convenience, here we recall a simple but nice concept, called the limit supremum
and limit infimum. This should be already taught in calculus course. Here we follow [Rud76,
Definition 3.16]. Given any sequence {an } ⊂ R, we define

lim sup an := lim sup am ≡ inf sup am ,

n→∞ n→∞ m≥n n∈N m≥n

lim inf an := lim inf am ≡ sup inf am .

n→∞ n→∞ m≥n n∈N m≥n

Unlike limit, both limit supremum and limit infimum always exist (because supm≥n am and
infm≥n am are monotone), but may takes “values” ±∞ ∈ / R (but only make sense for R). It is
clear that
lim inf an ≤ lim sup an
n→∞ n→∞
lim sup an ≤ lim sup bn , lim inf an ≤ lim inf bn if an ≤ bn for all n ≥ N for some N > 0.
n→∞ n→∞ n→∞ n→∞
In addition, for a real-valued sequence {an } ⊂ R, one has

(1.2.3) lim an = a∞ ∈ R ⇐⇒ lim sup an = lim inf an = a∞ ∈ R.

n→∞ n→n n→n

However, one has to be careful that, we only have subadditivity (resp. superaddivity) property for
limit supremum (resp. limit infimum):

 lim sup(an + bn ) ≤ lim sup an + lim sup bn
(1.2.4) n→∞ n→∞ n→∞ for {an }, {bn } ⊂ R,
 lim inf(an + bn ) ≥ lim inf an + lim inf bn
n→∞ n→∞ n→∞
 1.2. TOPOLOGICAL ASPECTS OF C 7

holds whenever the right hand side is not ∞ − ∞ or −∞ + ∞. For the case when limn→∞ bn exists
and finite, by writing an = (an + bn ) + (−bn ), using (1.2.4) we obtain

 lim sup an ≤ lim sup(an + bn ) − lim bn
n→∞ n→∞ n→∞
 lim inf an ≥ lim inf(an + bn ) − lim bn
n→∞ n→∞ n→∞
which implies 
 lim sup an + lim bn ≤ lim sup(an + bn ),
n→∞ n→∞ n→∞
 lim inf an + lim bn ≥ lim inf(an + bn ).
n→∞ n→∞ n→∞
Combining this with (1.2.4), we reach

 lim sup(an + bn ) = lim sup an + lim bn
n→∞ n→∞ n→∞
(1.2.5) when lim bn exists and finite.
 lim inf(an + bn ) = lim inf an + lim bn n→∞
n→∞ n→∞ n→∞

If {an } is bounded and limn→∞ bn exists which converges to some b ≥ 0, by writing an bn =

an b + an (bn − b) and using (1.2.5), one sees that
  
(∵b≥0) 
 lim sup(an bn ) = lim sup(an b) ≡
 lim sup an lim bn ,
n→∞ n→∞ n→∞ n→∞
(1.2.6)
 lim inf(a b ) = lim inf(a b) (∵b≥0)
   
≡

n n n lim inf an lim bn .
n→∞ n→∞ n→∞ n→∞

If we choose trivial sequence bn = b ≥ 0 for all n, then we reach

(1.2.7) lim sup(ban ) = b lim sup an for b ≥ 0.

n→∞ n→∞

However, one should be aware that when b ≥ 0, we have

lim sup(ban ) = − lim inf(|b|an ) = −|b| lim inf an = b lim inf an for b ≤ 0.
n→∞ n→∞ n→∞ n→∞

E XERCISE 1.2.6. Compute lim supn→∞ (an bn ) and lim infn→∞ (an bn ) when {an } is bounded and
limn→∞ bn exists which converges to some b ≤ 0.

If both {an } and {bn } are non-negative, one also has

   
 lim sup(an bn ) ≤ lim sup an

lim sup bn
(1.2.8) n→∞  n→∞   n→∞  for non-negative {an }, {bn }
 lim inf(an bn ) ≥ lim inf an lim inf bn

n→∞ n→∞ n→∞
holds whenever the right hand side is not 0 · ∞ or ∞ · 0. From (1.2.3) we have the following:

L EMMA 1.2.7. zn → z ∈ C if and only if lim supn→∞ |zn − z| = 0.

This simple observation can simplify the proofs. We can always take limit supremum in the
proof, which may simplify the proof in the future. One only need to be careful about (1.2.4).
 1.2. TOPOLOGICAL ASPECTS OF C 8

D EFINITION 1.2.8. The sequence {zn }n∈N is called a Cauchy sequence in C if, given any ε > 0,
there exists N > 0 such that |zn − zm | < ε for all n, m ≥ N.

L EMMA 1.2.9. The complex field (C, | · |) is complete, that is, the sequence {zn } converges if
and only if {zn } is a Cauchy sequence.

P ROOF. We first assume that the sequence {zn } converges to some limit z. By using the triangle
inequality in Lemma 1.2.2, one has

|zn − zm | ≤ |zn − z| + |z − zm |,

which immediately shows that {zn } is Cauchy. Conversely, suppose that {zn } is a Cauchy sequence.
From Definition 1.2.1 it is easy to see that

|Re zn − Re zm | = |Re (zn − zm )| ≤ |zn − zm |,

|Im zn − Im zm | = |Im (zn − zm )| ≤ |zn − zm |,

so that both {Re zn } and {Im zn } form Cauchy sequence in R, therefore there exist a, b ∈ R such
that
lim Re zn = a, lim Im zn = b.
n→∞ n→∞
We define z := a + bi, and from Exercise 1.2.5 and (1.2.4), one sees that

lim sup |zn − z| ≤ lim sup (|Re (zn − z)| + |Im (zn − z)|)
n→∞ n→∞
≤ lim sup |Re (zn − z)| + lim sup |Im (zn − z)|
n→∞ n→∞
(1.2.9) = lim sup |Re zn − a| + lim sup |Im zn − b| = 0,
n→∞ n→∞
which conclude our lemma. □

D EFINITION 1.2.10. We now given a sequence {zk }k∈N ⊂ C, and we define its partial sum
n
sn := ∑ zk .
k=1

An infinite series ∑∞
k=1 zk is said to converge in C if sn converges in C.

The following basic properties can be proved using same ideas as in calculus:
(1) If ∑∞ ∞ ∞
k=1 zk and ∑k=1 wk are converge in C, then ∑k=1 (zk ± wk ) are converge in C.
(2) If ∑∞k=1 zk is converges in C, then zk → 0 ∈ C.
(3) If ∑∞ ∞
k=1 |zk | converges in R, then ∑k=1 zk converges in C (this can be easily proved using
triangle inequality in Lemma 1.2.2).

D EFINITION 1.2.11. If ∑∞ ∞
k=1 |zk | converges in R, then we say that ∑k=1 zk converges in C
absolutely. Otherwise, we call the convergence is conditionally.
 1.2. TOPOLOGICAL ASPECTS OF C 9

1.2.2. Open sets in complex plane C.

D EFINITION 1.2.12. Let Ω be a set in C. We say that Ω is open in C if, given any z ∈ Ω, there
exists a ε > 0 such that Bε (z) ⊂ Ω.

This means that the open sets in C is exactly same as in R2 . Therefore we can borrow a lot of
topological terminology from R2 :
(1) An open set Ω contained z sometimes called the neighborhood of z.
(2) A set A is topological closed in C if its complement A∁ := C \ A is open in C. In this case,
A is closed in C if and only any Cauchy sequence {zn } ⊂ A converges to a limit z ∈ A.
(3) The boundary ∂ S of a set S is defined as: x ∈ ∂ S if and only if Bε (x) ∩ S ̸= 0/ and Bε (x) ∩
S∁ ̸= 0/ for all ε > 0.
(4) The closure S of a set S is defined by S := S ∪ ∂ S.
(5) Sometimes we called the boundary ∂ BR (z) of a ball BR (z) the circle.
(6) A set S is bounded if S ⊂ BR ≡ BR (0) for some R > 0.

D EFINITION 1.2.13. A set S is called compact in C if the following holds:

Oα for some collection of open sets {Oα }α∈Λ

[
S⊂
α∈Λ

Oα for some Λ′ ⊂ Λ which is finite.

[
=⇒ S ⊂
α∈Λ′

In fact, we have the Heine-Borel theorem: S is compact in C if and only if S is topological closed
and bounded. Using Bolzano-Weierstrass theorem, we also see that S is compact in C if and only
if any sequence in S must have a subsequence which is converges in S.

D EFINITION 1.2.14. Let S be any set in C. A subset S0 ⊂ S is said to be relative open in S if

there exists an open set Ω ⊂ C such that S0 = S ∩ Ω. Similarly, a subset S1 ⊂ S is said to be relative
topological closed in S if there exists a topological closed set F ⊂ C such that S1 = S ∩ F. A set S
is said to be connected if the following holds:

if S0 ⊂ S is both relative open and relative topological closed in S

(1.2.10) then either S0 = 0/ or S0 = S.

R EMARK 1.2.15 (Relative open sets in open sets). If S is an open set (resp. topological closed
set) in C and S0 ⊂ S, then S0 is open (resp. topological closed) in C if and only if S0 is relative open
(resp. relative tolopogical closed) in S. This can be easily see by the trivial set inclusion S0 = S ∩ S0 .
It is make sense to say that a set S is said to be disconnected if (1.2.10) does not hold. This
means that there exists 0/ ̸= S0 ⊊ S

there exists 0/ ̸= S0 ⊊ S such that

S0 is both relative open and relative topological closed in S.
 1.2. TOPOLOGICAL ASPECTS OF C 10

In this case, if we define S1 := S \ S0 , it is easy to see that 0/ ̸= S1 ⊊ S0 is also both relative open
and relative topological closed in S. Therefore one see that S0 and S1 are both disjoint (open)
components of S.

D EFINITION 1.2.16. We denote [z1 , z2 ] the line segment with endpoints z1 and z2 . A polygonal
line is a finite union of line segments of the form [z0 , z1 ] ∪ [z1 , z2 ] ∪ · · · ∪ [zn−1 , zn ].

L EMMA 1.2.17. Let Ω be an open set in C. Then Ω is connected if and only if for any a, b ∈ Ω
there exists a polygonal line in Ω connecting a and b.

R EMARK 1.2.18. Sometimes we also called an open connected set a region or domain.

P ROOF OF L EMMA 1.2.17. “⇒” Let a ∈ Ω and let

n o
A := x ∈ Ω there exists a polygonal line connecting a and x .

It is clear that a ∈ A, which shows that A ̸= 0.

/
Given any x ∈ A ⊂ Ω, since Ω is open, then there exists ε = ε(x) > 0 such that Bε (x) ⊂ Ω.
Clearly any point in Bε (x) can be connected to x by using a straight line, then any point in Bε (x)
can be connected to a by a polygonal line. In other words, Bε (x) ⊂ A. By arbitrariness of x ∈ A, we
conclude that A is open in C, and hence also relative open in Ω.
Similar argument shows that Ω \ A is also relative open in Ω. This shows that A is relative
topological closed in Ω. Since A ̸= 0,
/ then A = Ω.
“⇐” Let 0/ ̸= A ⊂ Ω be a set such that it is both relative open and relative topological closed in
Ω. Suppose the contrary, that A ̸= Ω, i.e. Ω \ A ̸= 0./ Choose a ∈ A and b ∈ Ω \ A. By assumption,
one can find a polygonal line connecting a and b, says [z0 , z1 ] ∪ [z1 , z2 ] ∪ · · · ∪ [zn−1 , zn ] with z0 = a
and zn = b. We define a continuous function f on [0, n] by

f (t) = z j + (t − j)(z j+1 − z j ) when t ∈ [ j, j + 1] for j = 0, 1, · · · , n − 1.

We now define the sets (called the preimage, this is just a notation, does not mean f is invertible)
n o n o
f −1 (A) := x ∈ Ω f (x) ∈ A , f −1 (Ω \ A) := x ∈ Ω f (x) ∈ Ω \ A .

Since both A and Ω \ A are open (in C if and only if relative to Ω), then both f −1 (A) and f −1 (S \ A)
are relative open in [0, n]. This is a special case of a general topological fact, but here we give a
simple argument to show that both f −1 (A) and f −1 (S \ A) are relative open in [0, n]. We only show
f −1 (A) is relative open in [0, n], since the same thing can be done for f −1 (S \ A). Let x0 ∈ f −1 (A),
i.e. f (x0 ) ∈ A.
Case 1: x0 ̸= 0 and x0 ̸= n. Since A is open, there exists ε > 0 such that Bε ( f (x0 )) ⊂ A. By
continuity of f at x0 , there exists δ > 0 such that
⇐⇒ y∈B (x0 ) ⇐⇒ f (y)∈Bε ( f (x0 ))
z }| δ { z }| {
|y − x0 | < δ =⇒ | f (y) − f (x0 )| < ε .
 1.2. TOPOLOGICAL ASPECTS OF C 11

Hence we see that

⇐⇒ y∈B (x0 ) ⇐⇒ y∈ f −1 (A)
z }| δ { z }| {
|y − x0 | < δ =⇒ f (y) ⊂ A
this meas that Bδ (x0 ) ⊂ f −1 (A).
Case 2: x0 = 0 (similar treatment for x0 = n). In this case, the continuity of f at x0 = 0 means
there exists δ > 0 (without loss of generality, we may choose δ < n) such that
⇐⇒ y∈B (x0 )∩[0,n]
δ ⇐⇒ f (y)∈Bε ( f (x0 ))
z }| { z }| {
0 ≤ y ≡ y − x0 < δ =⇒ | f (y) − f (x0 )| < ε .

Hence we see that

⇐⇒ y∈B (x0 )∩[0,n]
δ ⇐⇒ y∈ f −1 (A)
z }| { z }| {
0 ≤ y ≡ y − x0 < δ =⇒ f (y) ⊂ A .
This means that Bδ (x0 ) ∩ [0, n] ⊂ f −1 (A).
Combining these two cases, we now conclude that

given any x ∈ f −1 (A), there exists δ = δ (x) > 0 such that Bδ (x) ∩ [0, 1] ⊂ f −1 (A).

This means that f −1 (A) is relative open in [0, n], because

open in C
z [ }| {
−1
f (A) = [0, 1] ∩ Bδ (x) (x) .
x∈ f −1 (A)

Similar arguments also show that f −1 (S \ A) is relative open in [0, n], and hence f −1 (A) is relative
topological closed in [0, n]. Since the interval [0, n] is connected and f −1 (A) ̸= 0,
/ then f −1 (A) =
[0, n] and hence f −1 (S \ A) = 0,
/ which is a contradiction. This means that the assumption A ̸= Ω in
the contradiction argument does not hold. Hence we conclude that A = Ω. □
R EMARK 1.2.19. The above exhibits a standard argument when dealing with open connected
set:
(1) First show that the target set A (i.e. the set of the property which we wish to show) is
nonempty.
(2) Show that A is relative open.
(3) Show that Ω \ A is relative open.
To show an open set is connected, one of course can try to construct a continuous path
In my opinion, even though Lemma 1.2.17 gives a quite easy understanding, but Mathematically
sometimes this characterization is not convenient to manipulate. Personally I prefer the definition
(1.2.10): Even though it is abstract, but this is quite convenient to manipulate in Mathematical
proof.
L EMMA 1.2.20. zn → z if and only if: Given any open set Ω ∋ z, there exists N > 0 such that
zn ∈ Ω for all n ≥ N.
 1.2. TOPOLOGICAL ASPECTS OF C 12

P ROOF. We first suppose that zn → z. Given any open set Ω ∋ z, by definition there exists ε > 0
such that
Bε (z) ⊂ Ω.
By using (1.2.2), there exists N > 0 such that zn ∈ Bε (z) ⊂ Ω for all n ≥ N, which complete our
proof. The converse is trivial by choosing Ω = Bε (z) for arbitrary ε > 0. □

1.2.3. Continuous functions on C.

D EFINITION 1.2.21. Let z ∈ C and let Ω be an open neighborhood of z. We say that function
f : Ω → C is continuous at z if

zn → z ∈ C =⇒ f (zn ) → f (z) ∈ C.

Alternatively, given any ε > 0, there exists δ > 0, which depends on z, such that

(1.2.11) | f (z) − f (y)| < ε for all |z − y| < δ .

In other words, f (y) ∈ Bε ( f (z)) for all y ∈ Bδ (x). We say that f is continuous on Ω, denoted by
f ∈ C(Ω), if f is continuous at all point z ∈ Ω.
R EMARK 1.2.22. If one can find δ in (1.2.11) which is independent of z ∈ Ω, then one call such
function is uniformly continuous. In this case, it is also convenient to write (1.2.11) as

sup | f (z) − f (y)| < ε.

z,y∈Ω,|z−y|<δ

This notation emphasized that δ is independent of both y and z.

If we split f into its real and imaginary parts

f (z) = u(x, y) + iv(x, y) for z = x + iy ∈ Ω

it is clear that f is continuous at z = x + yi if and only if both u and v continuous at (x, y).
D EFINITION 1.2.23. We say that f ∈ Cm if and only if both u, v ∈ Cm , i.e. have continuous
partial derivatives of the mth order.
D EFINITION 1.2.24. A sequence of functions { fn } is said to be converge to f uniformly in Ω,
if for each ε > 0 there is an N > 0, which independent of z ∈ Ω, such that

(1.2.12) n ≥ N =⇒ | fn (z) − f (z)| < ε for all z ∈ Ω.

We now define the sup-norm on Ω by

∥g∥L∞ (Ω) := sup |g(z)| for all g ∈ C(Ω)

z∈Ω

By using this notations, we see that is equivalent to

(1.2.13) ∥ fn − f ∥L∞ (Ω) ≡ sup | fn (z) − f (z)| < ε for all n ≥ N.

z∈Ω
 1.2. TOPOLOGICAL ASPECTS OF C 13

L EMMA 1.2.25. Let Ω be an open set in C. Then fn converges to f uniformly in Ω if and only
if lim supn→∞ ∥ fn − f ∥L∞ (Ω) = 0.

Therefore, we also can say that fn → f in L∞ (Ω)-sense. Sometimes we also refer f the uniform
limit of f . It is well-known (see e.g. [Rud76]) that the uniform limit of real-valued continuous
function is continuous. By using the triangle inequality of ∥ · ∥L∞ (Ω) , which can be easily proved
using Lemma 1.2.2, one can easily obtain the following lemma.

L EMMA 1.2.26. Let Ω be an open set in C and let { fn } ⊂ C(Ω). If fn converges to f uniformly
in Ω, then f ∈ C(Ω).

C OROLLARY 1.2.27 (Weierstrass M-test). Let Ω be an open set in C and let { fn } ⊂ C(Ω). If
∥ fk ∥L∞ (Ω) ≤ Mk and ∑∞ ∞
k=1 Mk converges in R, then ∑k=1 f k (z) converges to a continuous function
uniformly in Ω.

P ROOF. It is easy to see that f (z) = ∑∞

k=1 f k (z) pointwisely. Moreover, we see that
n ∞ ∞
lim sup f − ∑ fk = lim sup ∑ fk ≤ lim sup ∑ Mk = 0,
n→∞ k=1 n→∞ k=n+1 n→∞ k=n+1
L∞ (Ω) L∞ (Ω)

which concludes our corollary. □

R EMARK 1.2.28 (A general trick). Here is a suggested standard procedure of proving uniform
convergence: First prove pointwise convergence to make sure the existence of limit function
(candidate), and then verify the convergence is uniform. This procedure is based on the fact that
the uniform limit is necessarily also a pointwise limit.
 CHAPTER 2

Differentiation

2.1. Complex derivative and Cauchy-Riemann equation

Inspired by calculus, it is not surprising to introduce the following definition.

D EFINITION 2.1.1. A complex-valued function f , defined in a neighborhood of z, is said to be

(complex) differentiable at z if
f (z + h) − f (z)
lim exists.
C∋h→0 h
In this case, the limit is denoted by f ′ (z) or ∂z f (z) or ∂∂z f (z) or dz
d
f (z). Let Ω be an open set in C.
A function f : Ω → C which is differentiable at every point in Ω is also called (complex) analytic
or holomorphic in Ω. A function f : C → C which is differentiable at every point in C is also called
entire.

R EMARK 2.1.2. It is important to note that in the above definition, h is not necessarily real.

R EMARK 2.1.3. Let Ω be an open set in C. Some authors call a function f : Ω → C is called
analytic at a point a ∈ Ω if there exists an open neighborhood U ⊂ Ω of a such that f is analytic in
U. Personally, I would prefer to say

(2.1.1) such function f is analytic near a ∈ Ω (rather than "at").

Throughout this course, we shall use the terminology (2.1.1) to avoid confusion.

E XERCISE 2.1.4. Show that the function f (z) = zz is differentiable at z = 0, but not analytic
near z = 0.

This exercise reminds us to be carefully while stating the terms “at” and “near”.

L EMMA 2.1.5. If f and g are both differentiable at z, then so are h1 = f + g and h2 = f g. If

g ̸= 0 near z, then h3 = f /g also differentiable at z. In the respective cases,

h′1 (z) = f ′ (z) + g′ (z),

h′2 (z) = f ′ (z)g(z) + f (z)g′ (z),
f ′ (z)g(z) − f (z)g′ (z)
h′3 (z) = .
g2 (z)

14
 2.1. COMPLEX DERIVATIVE AND CAUCHY-RIEMANN EQUATION 15

E XAMPLE 2.1.6. If P(z) = α0 + α1 z + · · · + αN zN for some complex numbers α0 , · · · , αN , then

P is differentiable at all points z and P′ (z) = α1 + 2α2 z + · · · + NαN zN−1 .

E XERCISE 2.1.7. Prove Lemma 2.1.5 and verify Example 2.1.6.

L EMMA 2.1.8. If f = u + iv is differentiable at z = x + iy, then the partial derivatives ∂x f and

∂y f of f both exist, and they satisfy the Cauchy-Riemann equation ∂y f = i∂x f .

P ROOF. The existence of ∂x f and ∂y f can be easily seen from the identities
f (z + h) − f (z) f (x + h, y) − f (x, y)
lim = lim = ∂x f (x, y),
R∋h→0 h R∋h→0 h
f (z + ih) − f (z) f (x, y + h) − f (x, y) 1
lim = lim = ∂y f (x, y).
R∋h→0 ih R∋h→0 ih i
Since f is differentiable at z, then the above identities must be identical, which conclude our lemma.
□

The converse of the above lemma does not hold true: There exist functions which are not
differentiable at a point despite the fact that the partial derivatives exist and satisfy the Cauchy-
Riemann equations here.

E XAMPLE 2.1.9. We consider


 xy(x+iy) , z ̸= 0,
x2 +y2
f (z) = f (x, y) =
0 , z = 0.

Since f = 0 on both axes x = 0 and y = 0, so that ∂x f (0, 0) = ∂y f (0, 0) = 0 (and hence satisfies the
Cauchy-Riemann equation). However, for each α ∈ R, one sees that
f (z) − f (0) x(αx)(x + iαx) α
lim = lim 2 2
= .
z→0,y=αx z z=x+iαx→0 x + (αx) 1 + α2
This shows that ∂z f (0, 0) does not exist. Suppose the contrary, that ∂z f (0, 0) exists, then
f (z) − f (0) f (z) − f (0) α
∂z f (0, 0) = lim = lim = for all α ∈ R,
C∋z→0 z z→0,y=αx z 1 + α2
which is a contradiction since ∂z f (0, 0) is independent of α.

However, it is worth to mention and proof that the equivalence holds when f is sufficiently
regular:

T HEOREM 2.1.10. Suppose that f ∈ C1 in a neighborhood of z = x + iy (sometimes we simply

say f ∈ C1 near z), that is, ∂x f and ∂y f are continuous in a neighborhood of z. We have the
following equivalence:

f satisfies the Cauchy-Riemann equation ∂y f = i∂x f at z ⇐⇒ f is differentiable at z.

 2.1. COMPLEX DERIVATIVE AND CAUCHY-RIEMANN EQUATION 16

R EMARK 2.1.11. If we write f = u + iv, the Cauchy-Riemann equation can be written as

∂x u = ∂y v, ∂y u = −∂x v.

From this, we see that

∆u := ∂x2 u + ∂y2 u = ∂x ∂y v − ∂y ∂x v = 0,
∆v := ∂x2 v + ∂y2 v = −∂x ∂y u + ∂y ∂x u = 0,

in other words, both u and v are harmonic.

P ROOF OF T HEOREM 2.1.10. The implication “⇐” was proved by Lemma 2.1.8. We only
need to prove the implication “⇒”.
We write h = h1 + ih2 . By using mean value theorem (for real functions of a real variable), one
sees that
Re f (z + h) − Re f (z) Re f (x + h1 , y + h2 ) − Re f (z)
=
h h1 + ih2
Re f (x + h1 , y + h2 ) − Re f (x + h1 , y) Re f (x + h1 , y) − Re f (x, y)
= +
h1 + ih2 h1 + ih2
h2 h1
= ∂y Re f (x + h1 , y + η2 ) + ∂x Re f (x + η1 , y)
h1 + ih2 h1 + ih2
ih2 1 h1
= ∂y Re f (x + h1 , y + η2 ) + ∂x Re f (x + η1 , y)
h1 + ih2 i h1 + ih2
for some |η1 | ≤ |h1 | and |η2 | ≤ |h2 |. Using the exactly same arguments, one also see that
Im f (z + h) − Im f (z)
h
ih2 1 h1
= ∂y Im f (x + h1 , y + η4 ) + ∂x Im f (x + η3 , y)
h1 + ih2 i h1 + ih2
for some |η3 | ≤ |h1 | and |η4 | ≤ |h2 |. Therefore, one has

f (z + h) − f (z)
− ∂x f (z)
h  
ih2 1
= ∂y (Re f (x + h1 , y + η2 ) + iIm f (x + h1 , y + η4 )) − ∂x f (x, y)
h1 + ih2 i
h1
+ (∂x (Re f (x + η1 , y) + iIm f (x + η3 , y)) − ∂x f (x, y)) .
h1 + ih2
 2.1. COMPLEX DERIVATIVE AND CAUCHY-RIEMANN EQUATION 17

By using the Cauchy-Riemann equation, we can write the above equation as

f (z + h) − f (z)
− ∂x f (z)
h  
ih2 1 1
= ∂y (Re f (x + h1 , y + η2 ) + iIm f (x + h1 , y + η4 )) − ∂y f (x, y)
h1 + ih2 i i
h1
+ (∂x (Re f (x + η1 , y) + iIm f (x + η3 , y)) − ∂x f (x, y))
h1 + ih2
h2
= (∂y (Re f (x + h1 , y + η2 ) + iIm f (x + h1 , y + η4 )) − ∂y f (x, y))
h1 + ih2
h1
+ (∂x (Re f (x + η1 , y) + iIm f (x + η3 , y)) − ∂x f (x, y))
h1 + ih2
h2
= (∂y Re f (x + h1 , y + η2 ) − ∂y Re f (x, y))
h1 + ih2
ih2
+ (∂y Im f (x + h1 , y + η4 ) − ∂y Im f (x, y))
h1 + ih2
h1
+ (∂x Re f (x + η1 , y) − ∂x Re f (x, y))
h1 + ih2
ih2
+ (∂x Im f (x + η3 , y) − ∂x Im f (x, y)) .
h1 + ih2
Hence
f (z + h) − f (z)
lim sup − ∂x f (z)
C∋h→0 h
≤ lim sup ∂y Re f (x + h1 , y + η2 ) − ∂y Re f (x, y)
C∋h→0
+ lim sup ∂y Im f (x + h1 , y + η4 − ∂y Im f (x, y)
C∋h→0
+ lim sup |∂x Re f (x + η1 , y) − ∂x Re f (x, y)|
C∋h→0
+ lim sup |∂x Im f (x + η3 , y) − ∂x Im f (x, y)|
C∋h→0
= 0,

which complete our proof with ∂z f = ∂x f . □

R EMARK 2.1.12. Suppose that all assumptions in Theorem 2.1.10 hold near z ∈ C. Let f be
a complex-valued function which is analytic at z. By using the Cauchy-Riemann equation and
∂z f = ∂x f , one see that
1
(2.1.2a) ∂z f = (∂x f − i∂y f ).
2
 2.1. COMPLEX DERIVATIVE AND CAUCHY-RIEMANN EQUATION 18

We now define the operator ∂z on C1 function by

1
(2.1.2b) ∂z f := (∂x f + i∂y f ).
2
By introducing this notation, one sees that Theorem 2.1.10 can be restated as (see Exercise 2.1.4)
f is analytic near z ⇐⇒ ∂z f = 0 near z
(2.1.3)
(assuming that all assumptions in Theorem 2.1.10 hold)
In particular, the operators (2.1.2a) and (2.1.2b) are called the Wirtinger operators.

R EMARK 2.1.13. Wirtinger operators can be defined in terms of weak derivatives (even
distributional derivatives), and it interesting to mention that the quasiconformal mapping is related
to the Beltrami equation:
∂z f = µ∂z f with ∥µ∥L∞ ≤ c < 1.
When µ ≡ 0, this reduces to (2.1.3) (Note: We called a mapping is conformal if it is holomorphic
and injective, therefore the term “quasiconformal” make sense). For more details about the
quasiconformal mapping and Beltrami equation, one can refer to the monograph [AIM09].

Warning: If ∂z f ̸= 0 (i.e. does not satisfy Cauchy-Riemann equation), the function ∂z f in

(2.1.2a) is not equivalent to the one we introduced in Definition 2.1.1.

Warning: Even though (2.1.3) suggests that analytic function must not contained z, to show
a function is analytic or not, we still have to verify the definition carefully, see Exercise 2.1.4.

Warning: Always remember to check the assumptions in Theorem 2.1.10.

E XAMPLE 2.1.14. Any complex-valued polynomial P takes the form P = ∑N

n=0 Qn for some
N ∈ Z≥0 with
n
Qn (z) = Qn (x, y) = ∑ Cn,k xn−k yk
k=0
= Cn,0 x n
+Cn,1 xn−1 y +Cn,2 xn−2 y2 + · · · +Cn,n yn
 2.2. POWER SERIES 19

for some CN,m ∈ C. One computes that

n−1 n
2∂z Qn (z) = ∑ Cn,k (n − k)xn−k−1yk + i ∑ Cn,k kxn−k yk−1
k=0 k=1
n−1 n−1
n−k−1 k
= ∑ Cn,k (n − k)x y + i ∑ Cn,(k̃+1) (k̃ + 1)xn−k̃−1 yk̃
k=0 k̃=0
n−1 n−1
= ∑ Cn,k (n − k)xn−k−1yk + i ∑ Cn,(k+1)(k + 1)xn−k−1yk
k=0 k=0
n−1
Cn,k (n − k) + iCn,(k+1) (k + 1) xn−k−1 yk .


= ∑
k=0

If P satisfies the Cauchy-Riemann equation (that is, P is analytic), then

Cn,k (n − k) + iCn,(k+1) (k + 1) = 0 for all n = 0, 1, · · · , N and k = 0, · · · , n − 1.

By using induction, one also can verify that

!
n
(2.1.4) Cn,k = ik Cn,0 for all n = 0, 1, · · · , N and k = 0, · · · , n − 1.
k
Substitute (2.1.4) into P, one reaches
!
N n N N
n
(2.1.5) P(z) = ∑ Cn,0 ∑ xn−k (iy)k = ∑ Cn,0(x + iy)n = ∑ Cn,0zn.
n=0 k=0 k n=0 n=0

Combining with Example 2.1.6, we know that a polynomial P enjoys the following property:

(2.1.6) P is analytic ⇐⇒ P takes the form (2.1.5).

Therefore, if a polynomial takes the form (2.1.5) (or in Example 2.1.6), we called it an analytic
polynomial.
It is interesting to compare the following exercise with Remark 2.1.11.
E XERCISE 2.1.15. If f (z) = u + iv is a complex function such that u and v are both harmonic
in an open set Ω, is f (z) necessarily analytic in Ω?

An application. In one of my research paper [KLSS24], we use complex polynomial to construct

some explicit examples of domain which is non-scattering with respect to some acoustic wave
(which satisfies time-harmonic wave equation).

2.2. Power series

Example 2.1.14 immediately suggests a wider class of direct functions of z, those given by
“infinite polynomials” in z:
 2.2. POWER SERIES 20

D EFINITION 2.2.1. A power series in z is an infinite series (in the sense of Definition 1.2.10)
of the form ∑∞ k
k=0 Ck z .

We now prove some properties which are similar to the power series on R (see e.g. [Rud76]).

T HEOREM 2.2.2. Given a sequence {Ck } ⊂ C.

1
(a) If lim sup |Ck | k = 0, then ∑ Ck zk converges absolutely for all z ∈ C. In addition, for each
k→∞
r > 0, ∑ Ck zk converges uniformly1 in z ∈ Br .
1
(b) If lim sup |Ck | k = +∞, then ∑ Ck zk converges for z = 0 only.
k→∞
1
(c) If 0 < lim sup |Ck | k < +∞, then ∑ Ck zk converges absolutely for |z| < R and diverges for
k→∞
|z| > R, where
 −1
1
(2.2.1) R = lim sup |Ck | k .
k→∞

In addition, for each 0 < ε < R, ∑ Ck zk converges uniformly2 in z ∈ BR−ε .

R EMARK 2.2.3 (Inconclusive on BR ). For (a) and (b), we simply say the radius of convergence
R = ∞ and R = 0 respectively. The uniform convergence only holds true for BR−ε , but not for BR .
If the uniform convergence is on BR , then the sequence converges on |z| = R, however this is not
true, see Exercises 2.2.6, 2.2.7 and 2.2.8.

R EMARK 2.2.4 (Structure of power series). If z ∈ C satisfies |z| > R, then by (1.2.7) we have
1 1
1 < R−1 |z| = lim sup |Ck | k |z| = lim sup |Ck zk | k .
k→∞ k→∞

This shows that the sequence {Ck zk }k∈N does not converge to 0. Otherwise, suppose the contrary
that {Ck zk }k∈N converge to 0, then it must be bounded, says |Ck zk | ≤ L for all k. Hence we see that
1 1
lim sup |Ck zk | k ≤ lim sup L k = 1,
k→∞ k→∞

which is a contradiction. Since {Ck zk }k∈N does not converge to 0, thus ∑ Ck zk diverges. In view of
Theorem 2.2.2, it is make sense to call such constant R is called the radius of convergence of the
power series ∑ Ck zk .

R EMARK 2.2.5. By using previous remark, it is important to notice that, if ∑ Ck zk converges at

z0 , then it also converges in B|z0 | , i.e. the ball with radius |z0 | (not include the boundary, which is
inconclusive). Similarly, if ∑ Ck zk diverges at z0 , the it is also diverges in C \ B|z0 | .
1The rate of convergence depends on r. We do not know whether the series converges uniformly on the whole C or
not.
2The rate of convergence depends on ε. We do not know whether the series converges uniformly on the whole B or
R
not.
 2.2. POWER SERIES 21

P ROOF OF ( A ). For each r̃ > 0, there exists N > 0, which depends on r̃, such that
1 1 1
|Ck | k ≤ for all k ≥ N =⇒ |Ck |r̃k ≤ k for all k ≥ N.
2r̃ 2
• For each z ∈ C, by choosing r̃ = |z|, we see that
∞
1
lim sup ∑ |Ck zk | = lim sup ∑ |Ck |r̃k ≤ lim sup ∑ k
= 0,
n→∞ k=n n→∞ k≥n n→∞ k≥n 2

which concludes that the series converges absolutely at each z ∈ C.

• On the other hand, for each r > 0, one can choose r̃ = r to see that
∞
1
lim sup sup ∑ Ck zk ≤ lim sup ∑ |Ck |r̃k ≤ lim sup ∑ k
= 0,
n→∞ z∈Br k=n n→∞ k≥n n→∞ k≥n 2

which concludes that the series converges uniformly in Br .

□

P ROOF OF ( B ). For any z ̸= 0, there exists a sequence {kn } ⊂ N with kn → +∞ such that
1 1
|Ckn | kn ≥ for all n =⇒ |Ckn zkn | = |Ckn ||z|kn ≥ 1 for all n,
|z|
which shows that ∑ Ck zk does not converges for all z ̸= 0 (Note: If ∑ Ck zk converges, then it is
necessarily that Ck zk → 0, which will led a contradiction). □

P ROOF OF ( C ). We first consider the case when |z| > R. There exists δ > 0 such that |z| =
R + δ , and there exists a sequence {kn } ⊂ N with kn → +∞ such that
1 1
|Ckn | kn ≥ for all n =⇒ |Ckn zkn | = |Ckn ||z|kn ≥ 1 for all n,
R+δ
so that ∑ Ck zk does not converges. We now fix any 0 < r̃ < R, and we write 2δ = R − r̃ > 0. By
using the definition of (2.2.1), one see that there exists N > 0, which depends on r̃, such that
R − 2δ k
 
1 1 k
|Ck | k ≤ for all k ≥ N =⇒ |Ck |r̃ ≤ for all k ≥ N.
R−δ R−δ
• For each z ∈ BR , by choosing r̃ = |z|, we see that
∞ ∞ ∞  k
k k R − 2δ
lim sup ∑ |Ck z | = lim sup ∑ |Ck |r̃ ≤ lim sup ∑ = 0,
n→∞ k=n n→∞ k=n n→∞ k=n R−δ
which concludes that the series converges absolutely at each z ∈ BR .
• On the other hand, for each 0 < ε < R, we choose r̃ = R − ε to see that
∞ 
R − 2δ k
∞ ∞ 
k k
lim sup sup ∑ Ck z ≤ lim sup ∑ |Ck |r̃ ≤ lim sup ∑ = 0,
n→∞ z∈BR−ε k=n n→∞ k=n n→∞ k=n R−δ
which concludes that the series converges uniformly in BR−ε .
 2.2. POWER SERIES 22

□
When the radius of convergence R ∈ (0, ∞), there is no guarantee for the convergence or
divergence at z ∈ ∂ BR (however, this is related to Fourier series, see Remark 2.3.7 below). This
demonstrates by the following exercises.

E XERCISE 2.2.6. Show that the radius of convergence of ∑∞ n

n=1 nz is R = 1, and the series also
diverges for |z| = 1.

E XERCISE 2.2.7. Show that the radius of convergence of ∑∞ −2 n

n=1 n z is R = 1, and the series
also converges for |z| = 1.

E XERCISE 2.2.8. Show that the radius of convergence of ∑∞ −1 n

n=1 n z is R = 1. In addition,
show that the series converges for all z ∈ ∂ B1 \ {1} but diverges at z = 1.

We now show that power series, like polynomials, are differentiable functions of z (in the sense
of Definition 2.1.1).

T HEOREM 2.2.9. Suppose that the series f (z) = ∑∞ n

n=0 Cn z has the radius of convergence 0 <
R ≤ +∞ given in (2.2.1) (see Theorem (2.2.2)), then f ′ (z) exists (in the sense of Definition 2.1.1)
and equal to
∞ ∞ ∞
(2.2.2) g(z) := ∑ nCnzn−1 ≡ ∑ nCnzn−1 ≡ ∑ C̃mzm with C̃m := (m + 1)Cm+1
n=0 n=1 m=0

in BR , and g also has the radius of convergence R, which is same as f . As an immediate

consequence, power series are infinitely differentiable (in the sense of Definition 2.1.1) within their
domain of convergence.

P ROOF. We divide the proof into several steps.

Step 1: Radius of convergence. By using (1.2.6) one sees that
1 1 1 n 1 1
lim sup |C̃m | m = lim sup |nCn | n−1 = lim (n n ) n−1 lim sup |Cn | n−1 = lim sup |Cn | n−1 .
m→∞ n→∞ n→∞ n→∞ n→∞

There exists a subsequence {Cnk } such that

1 1 nk 1
1 ·n
lim sup |Cn | n−1 = lim |Cnk | nk −1 = lim |Cnk | nk k −1 = lim |Cnk | nk
n→∞ k→∞ k→∞ k→∞
1 1
≤ lim sup |Cm | = lim sup |Cn | .
m n
k→∞ m≥nk n→∞

Conversely, we also can find another subsequence {Cnℓ } such that

1 1 nℓ −1 1
1 ·
lim sup |Cn | n = lim |Cnℓ | nℓ = lim |Cnℓ | nℓ −1 nℓ
= lim |Cnℓ | nℓ −1
n→∞ ℓ→∞ ℓ→∞ ℓ→∞
1 1
≤ lim sup |Cm | m−1 = lim sup |Cn | n−1 .
ℓ→∞ m≥nℓ n→∞
 2.2. POWER SERIES 23

Combining the above three equations, we reach

1 1
lim sup |C̃m | m = lim sup |Cn | n ,
m→∞ n→∞
hence we conclude that g also has the radius of convergence R, which is same as f .

Step 2: Show that f ′ exists and it equal to g. We now further divide our discussions in subcases.

Step 2a: When R = ∞. Given any h ∈ C \ {0} with |h| < 1. The absolute convergence allows us to
rearrange the sum, hence
f (z + h) − f (z) ∞
(z + h)n − zn ∞ ∞
− g(z) = ∑ Cn − ∑ nCn zn−1 = ∑ Cn bn ,
h n=0 h n=0 n=0

where
! !
(z + h)n − zn 1 n n k n−k n
bn = − nzn−1 = ∑ h z − z − nzn−1 (binomial theorem)
h h k=0 k
! ! !
1 n n k n−k n
n k−1 n−k n
n
= ∑ h z − nzn−1 = ∑ h z − nzn−1 = ∑ hk−1 zn−k .
h k=1 k k=1 k k=2 k

Then
! !
n n
n k−1 n−k n
|bn | ≤ ∑ |h| |z| ≤ |h| ∑ |z|n−k
k=2 k k=2 k
!
n
n
≤ |h| ∑ |z|n−k = |h|(|z| + 1)n (again binomial theorem),
k=0 k
and hence
< +∞ because R=∞
z }| {
∞ ∞
f (z + h) − f (z)
− g(z) ≤ ∑ |Cn ||bn | ≤ |h| ∑ |Cn |(|z| + 1)n .
h n=0 n=0
Taking h → 0 (in the sense of limit supremum), we conclude f ′ exists and f ′ (z) = g(z) for all z ∈ C.

Step 2b: When 0 < R < ∞. Given any |z| < R, and write |z| = R − 2δ for some δ > 0. We now
let h ∈ C \ {0} with |h| < min{δ , 1}. Then |z + h| ≤ |z| + |h| ≤ R − 2δ + δ = R − δ < R, and as in
above, we can write
!
∞ ∞
f (z + h) − f (z) n k−1 n−k
− g(z) = ∑ Cn bn , bn = ∑ h z .
h n=0 k=2 k
 2.2. POWER SERIES 24

If z = 0, then bn = hn−1 and the proof follows easily (left as exercise). We now consider the case
when z ̸= 0. For each 2 ≤ k ≤ n we see that
! ! !
n n−k+1 n n − k + 1 n − (k − 1) + 1 n
= = ·
k k k−1 k k−1 k−2
! ! !
n − 2 + 1 n − (2 − 1) + 1 n n−1 n n
≤ · = ·n ≤ n2
2 2−1 k−2 2 k−2 k−2
n−k+1 n−(k−1)+1
since both k and k−1 are monotone decreasing on k. We now have
! !
n 2 |h| n
n n n
|bn | ≤ n2 ∑ |h|k−1 |z|n−k = 2 ∑ k−2
|h|k−2 |z|n−(k−2)
k=2 k − 2 |z| k=2
! !
n2 |h| n−2 n n 2 |h| n n
= ∑ j |h| j |z|n− j ≤ |z|2 ∑ j |h| j |z|n− j
|z|2 j=2 j=2

n2 |h|
= (|z| + |h|)n (binomial theorem)
|z|2
n2 |h| n |h|  2 n

≤ (R − δ ) = 2 (R − δ )n n ,
|z|2 |z|
then we now reach
√
n n→1 as n→∞
< +∞ since
z }| {
∞
f (z + h) − f (z) |h| ∞  2 n

− g(z) ≤ ∑ |Cn ||bn | ≤ 2 ∑ |Cn | (R − δ )n n .
h n=0 |z| n=0
Taking h → 0 (in the sense of limsup), we conclude f ′ exists in BR and f ′ (z) = g(z) for all z ∈
BR . □

E XERCISE 2.2.10. Show that if f (z) = ∑∞ n

n=0 Cn z has a nonzero radius of convergence, then

f (n) (0)
Cn = for all n = 0, 1, 2, · · · ,
n!
where f (n) is the nth derivative of f (in the sense of Definition 2.1.1). Show that for each n =
0, 1, 2, · · · that
(n + 2)!
f (n) (z) = n!Cn + (n + 1)!Cn+1 z + Cn+2 z2 + · · ·
2!
for all z in the domain of convergence.

T HEOREM 2.2.11 (Uniqueness of power series). Suppose that the power series f (z) =
n
∑∞
n=0 Cn z has a nonzero radius of convergence. If there exists a sequence {zk } in the domain
of convergence such that

zk → 0 ∈ C, zk ̸= 0 and f (zk ) = 0 for all k = 1, 2, 3, · · ·

 2.3. EXPONENTIAL, SINE AND COSINE FUNCTIONS 25

then f ≡ 0.

R EMARK 2.2.12. If a power series equals to zero at all the points of a set with an accumulation
point at the origin, the power series is identically zero in the domain of convergence. As an
immediate consequence, if ∑ an zn and ∑ bn zn converge and agree on a set of points with an
accumulation point at the origin, then an = bn for all n.

P ROOF. We want to show Cn = 0 for all n = 0, 1, 2, · · · by using strong mathematical induction.

• By continuity of f at the origin, we see that

C0 = f (0) = lim f (z) = lim f (zk ) = 0.

z→∞ k→∞
• We now assume the induction hypothesis that C j = 0 for all j = 0, 1, 2, · · · , n − 1. The
induction hypothesis guarantees that the function
f (z)
g(z) = = Cn +Cn+1 z +Cn+2 z2 + · · · ,
zn
is continuous at the origin by defining g(0) := Cn . Since
f (zk )
0= = g(zk ) for all k = 1, 2, 3, · · ·
znk
then we conclude our result by taking k → ∞ (so that zk → 0).
We conclude the theorem by strong mathematical induction. □

2.3. Exponential, sine and cosine functions

We define the exponential function

(2.3.1) ez := ex (cos θ + i sin θ ) for all z = x + iθ ∈ C.

It is easy to see that |ez | = ex and ez ̸= 0 for all z = x + iy ∈ C.

E XERCISE 2.3.1. Prove that ez1 +z2 = ez1 ez2 for all z1 , z2 ∈ C.

Euler’s formula is just a special case of (2.3.1):

(2.3.2) eiθ = cos θ + i sin θ for all θ ∈ R.

E XERCISE 2.3.2 (Euler, De Moivre). For each n ∈ N, show that (cos θ + i sin θ )n = cos(nθ ) +
i sin(nθ ) for all θ ∈ R.

It is useful to see that

(2.3.3) z = |z|eiθ for all z ∈ C

for some θ ∈ [0, 2π), which is just simply the polar coordinate in R2 .
 2.3. EXPONENTIAL, SINE AND COSINE FUNCTIONS 26

E XERCISE 2.3.3. Show that ez is entire (Definition 2.1.1) by verifying the Cauchy-Riemann
equation.

In fact, ez has a unique power series representation (see Theorem 4.2.1 below) with radius of
convergence R = +∞.

E XERCISE 2.3.4. Prove that

∞
zn
ez = ∑ for all z ∈ C.
n=0 n!
By using Exercise 2.3.3 and Remark 2.1.12, one can easily see that

∂z ez = ∂x (ex (cos y + i sin y)) = ex (cos y + i sin y) = ez for all z = x + iy ∈ C.

From (2.3.2), we see that

1 iθ
sin θ = (e − e−iθ ) for all θ ∈ R,
2i
1 iθ
cos θ = (e + e−iθ ) for all θ ∈ R.
2
Therefore it is natural to define the entire functions
1
(2.3.4a) sin z := (eiz − e−iz ) for all z ∈ C,
2i
1 iz
(2.3.4b) cos z := (e + e−iz ) for all z ∈ C.
2
We remind the readers that cos z and sin z are not bounded in modulus by 1, since
1 −θ
sin(iθ ) = (e − eθ ) for all θ ∈ R,
2i
1
cos(iθ ) = (e−θ + eθ ) for all θ ∈ R.
2
E XERCISE 2.3.5. Show that sin z and cos z are entire (Definition 2.1.1) by verifying the Cauchy-
Riemann equation. Verify the identities

sin 2z = 2 sin z cos z, sin2 z + cos2 z = 1, (sin z)′ = cos z.

Compute (cos z)′ .

In fact, each sin z and cos z has a unique power series representation (see Theorem 4.2.1 below)
with radius of convergence R = +∞.

E XERCISE 2.3.6. Show that

z3 z5
+ − + · · · for all z ∈ C.
sin z = z −
3! 5!
Compute the power series representation of cos z as well.

Finally, we end this chapter by the following remark.

 2.3. EXPONENTIAL, SINE AND COSINE FUNCTIONS 27

R EMARK 2.3.7. Let ∑n cn zn be a power series with radius of convergence of 0 < R < ∞, as
described in Theorem 2.2.2. We do not know whether the power series converges on z ∈ ∂ BR or
not, see Exercise 2.2.6, Exercise 2.2.7 and Exercise 2.2.8. For each z ∈ ∂ BR , we can write z = Reiθ ,
and plut this form into the power series to obtain

(2.3.5) ∑ c̃neinθ with c̃n = cn Rn .

n
The series (2.3.5) is indeed a special case of Fourier series of period 2π, see e.g. my previous
lecture note [Kow22] for further details.
 CHAPTER 3

Integration

In previous chapter, we are focusing in (complex) differentiability of complex-valued functions.

We now discuss its counterpart: the integral.

3.1. The fundamental theorem of line integral

Before we consider the function with complex domain, we first deal with the functions defined
on interval in R.

D EFINITION 3.1.1. Let φ ≡ Re φ + iIm φ : [a, b] ⊂ R → C, which is continuous on [a, b], that
is Re φ , Im φ ∈ C([a, b]). The integral of φ is defined by
Z b Z b Z b
φ (t) dt := Re φ (t) dt + i Im φ (t) dt,
a a a
Rb Rb
where a Re φ (t) dt andIm φ (t) dt are just the usual Riemann integral.
a
h i
D EFINITION 3.1.2. Let C = z(t) = x(t) + iy(t) a ≤ t ≤ b be an (oriented) continuous
curve in C. If x and y are both differentiable at some t ∈ (a, b), then we set

ż(t) := x′ (t) + iy′ (t) for such t.

We call the curve C is piecewise-C1 if both x, y ∈ C([a, b]) and x, y ∈ C1 on each subinterval
int [a, x1 ], int [x1 , x2 ], ..., int [xn−1 , b] of some partition of [a, b]. If in addition that ż(t) ̸= 0 for
all but finitely many t ∈ (a, b) (i.e. there are at most finitely many t0 such that x′ (t0 ) = y′ (t0 ) = 0),
then we called it a parametrizable continuous piecewise-C1 curve.

R EMARK 3.1.3. Usually we refer an oriented curve C is smooth when both x, y ∈ C∞ (a, b).
Here we will not follow the terminology in [BN10].

Finally, we define the important concept of a line integral. This concept also introduced in the
vector calculus, see e.g. [GM12].
h i
D EFINITION 3.1.4 (Line integral). Let C = z(t) a ≤ t ≤ b be a parametrizable continuous
piecewise-C1 curve and suppose the complex-valued function f is continuous on C (up to
endpoints). The (line) integral of f along C is defined by
Z Z Z b Z b
f≡ f (z) dz := f (z)|z=z(t) ż(t) dt = f (z(t))ż(t) dt,
C C a a
28
 3.1. THE FUNDAMENTAL THEOREM OF LINE INTEGRAL 29

where the integrand (i.e. the quantity being integrated) is the complex multiplication of f (z(t)) and
ż(t).

It is clear that the integral depends on the curve C , more precisely, the integral
h depends oni
parametrization z (hence depends on its orientation). Therefore we denote C = z(t) a ≤ t ≤ b
n o
rather than z(t) a ≤ t ≤ b to emphasize the orientation of the curve. However, it is possible
R
to perturb the integral curve without changing the values of the line integral C f .
h i h i
L EMMA 3.1.5. Let C1 = z(t) a ≤ t ≤ b and C2 = w(t) c ≤ t ≤ d be two
parametrizable continuous piecewise-C1 curves in C. If there exists an injective C1 mapping
λ : [c, d] → [a, b] such that

(3.1.1) λ (c) = a, λ (d) = b, w(t) = z(λ (t)) for all t ∈ [c, d],
R R
then C1 f= C2 f.

E XERCISE 3.1.6. Prove Lemma 3.1.5.

h i h i
E XERCISE 3.1.7. Let C1 = z(t) a ≤ t ≤ b and C2 = w(t) c ≤ t ≤ d be two
parametrizable continuous piecewise-C1 curves in C. We define the relation ∼ by

(3.1.2) C1 ∼ C2 ⇐⇒ there exists λ ∈ C1 ([c, d]) satisfies (3.1.1).

Show that ∼ is an equivalence relation, i.e. show that:

(1) Reflexivity. C ∼ C for any parametrizable continuous piecewise-C1 curve C in C.
(2) Symmetry. C1 ∼ C2 ⇐⇒ C2 ∼ C1 for all parametrizable continuous piecewise-C1 curves
C1 , C2 in C.
(3) Transitivity. Let C1 , C2 , C3 be parametrizable continuous piecewise-C1 curves in C. If
C1 ∼ C2 and C2 ∼ C3 , then C1 ∼ C3 .
Therefore, we can rephrase Lemma 3.1.5 as: If C1 and C2 are parametrizable continuous piecewise-
C1 curves in C which are equivalent in the sense of (3.1.2), then C1 f = C2 f .
R R

h i
L EMMA 3.1.8. Let C = z(t) a ≤ t ≤ b be a parametrizable continuous piecewise-C1
curve in C. If we define h i
C := z(b + a − t) a ≤ t ≤ b ,
rev
R R
then C rev f =− C f.

One should notice that, C and C rev are identical as sets, but reverse oriented.

E XERCISE 3.1.9. Prove Lemma 3.1.8.

The following lemma exhibit a basic property of line integral.

 3.1. THE FUNDAMENTAL THEOREM OF LINE INTEGRAL 30

L EMMA 3.1.10. Let C be a parametrizable continuous piecewise-C1 curve, then the mapping
R
f 7→ C f is C-linear, that is,
R R R
(1) C ( f + g) = C f + C g for all f , g ∈ C(C ),
R R
(2) C α f = α C f for all f ∈ C(C ) and α ∈ C.
Here C(C ) denotes the collection of continuous functions on C (up to endpoints).

E XERCISE 3.1.11. Prove Lemma 3.1.10.

L EMMA 3.1.12. If the complex-valued function G ∈ C([a, b]), then

Z b Z b
G(t) dt ≤ |G(t)| dt.
a a

The LHS of the above integral is defined in the sense of Definition 3.1.1, while the RHS is the usual
Riemann integral.
Rb
P ROOF. We first write a G(t) dt in terms of polar coordinate, that is,
Z b Z b
G(t) dt = G(t) dt eiθ
a a
R
for some θ ∈ [0, 2π). By linearity of we reachC,
Z b Z b
def b
Z   Z b  
−iθ −iθ
G(t) dt = e G(t) dt ≡ Re e G(t) dt + i Im e−iθ G(t) dt.
a a a a

Taking real part of the above equation, we reach

Z b Z b  
G(t) dt = Re e−iθ G(t) dt.
a a

Since |Re e−iθ G(t) | ≤ |e−iθ G(t)| = |G(t)|, by the monotonicity of the usual Riemann integral,

we reach Z b Z   Z b b
G(t) dt = Re e−iθ G(t) dt ≤ |G(t)| dt,
a a a
which is our desired result. □
h i
D EFINITION 3.1.13. Let C = z(t) a ≤ t ≤ b be a parametrizable continuous piecewise-C1
curve. Its length H 1 (C ) is defined by
Z b Z bq
H (C ) :=1
|ż(t)| dt = (x′ (t))2 + (y′ (t))2 dt.
a a

At the first glance, the definition of the arc length H 1 (C ) depends on the parameterization
z(t). In fact, H 1 (C ) is equal to the 1-dimensional Hausdorff measure of the curve C [BBI01,
Theorem 2.6.2], and this implies that the quantity H 1 (C ) actually does not depend on the choice
of parameterization z(t).
 3.1. THE FUNDAMENTAL THEOREM OF LINE INTEGRAL 31

L EMMA 3.1.14. Let C be a parametrizable continuous piecewise-C1 curve with length

H 1 (C ), then
Z
f ≤ ∥ f ∥L∞ (C ) H 1 (C ) for all f ∈ C(C ).
C

P ROOF. By Lemma 3.1.12 we see that

Z Z b Z b Z b
f = f (z(t))ż(t) dt ≤ | f (z(t))||ż(t)| dt ≤ ∥ f ∥L∞ (C ) |ż(t)| dt.
C a a a

We combine the above two equations and conclude the lemma. □

L EMMA 3.1.15. Suppose { fn } is a sequence of continuous functions and fn → f uniformly on

the parametrizable continuous piecewise-C1 curve C . Then
Z Z
f = lim fn .
C n→∞ C
R
P ROOF. By (1.2.7), linearity of C and Lemma 3.1.14, one easily sees that
Z Z Z
lim sup f (z) dz − fn (z) dz = lim sup ( f (z) − fn (z)) dz
n→∞ C C n→∞ C

≤ H 1 (C ) lim sup ∥ f − fn ∥L∞ (C ) = 0,

n→∞
which conclude our lemma. □

We now prove one of the main result of this section, which also can be regard as a generalization
of fundamental theorem of calculus (integral operator as an inverse operator of differentiation
operator).
h i
T HEOREM 3.1.16 (Fundamental theorem of line integral). Let C = z(t) a ≤ t ≤ b be a
parametrizable continuous piecewise-C1 curve. If f ∈ C1 (C ) is (complex) differentiable on C ,
then Z
f ′ = f (z(b)) − f (z(a)).
C

R EMARK 3.1.17. The C1 assumption on f is to ensure that f ′ ∈ C(C ) so that C f ′ is well-

R

defined according to Definition 3.1.4. If we directly use Lemma 2.1.8, the complex differentiability
of f only guarantees the existence of partial derivatives. We will later prove in Corollary 4.6.2 that
the differentiability of f near a point will implies that it is C∞ near such point, therefore the C1
assumption in Theorem 3.1.16 can be easily verified in practical.

P ROOF OF T HEOREM 3.1.16. By assumptions, we have ż(t) ̸= 0 for all but finitely many a <
t < b. For such t, we can find δt > 0 so that z(t + h) − z(t) ̸= 0 and a < t + h < b for all |h| < δt .
We see that
f (z(t + h)) − f (z(t)) f (z(t + h)) − f (z(t)) z(t + h) − z(t)
= · for all 0 < |h| < δt ,
h z(t + h) − z(t) h
 3.2. CAUCHY CLOSED CURVE THEOREM IN RECTANGLE 32

which gives
d f (z(t + h)) − f (z(t))
( f (z(t))) = lim
dt R∋h→0 h
f (z(t + h)) − f (z(t)) z(t + h) − z(t)
= lim · lim
R∋h→0 z(t + h) − z(t) R∋h→0 h
f (w) − f (z(t)) z(t + h) − z(t)
= lim · lim
C∋w→z(t) w − z(t) R∋h→0 h
= f ′ (z) z=z(t)
ż(t) (complex multiplication).

Hence by the definition of line integral, one sees that

Z b Z b
d
Z
′ def ′
f (z) dz ≡ f (z) z=z(t)
ż(t) dt = ( f (z(t))) dt = f (z(b)) − f (z(a)),
C a a dt
where the last equality is just simply the fundamental theorem of calculus [Rud76, Theorem 6.21].
□

3.2. Cauchy closed curve theorem in rectangle

We begin our discussions by the following definition.

h i
D EFINITION 3.2.1. A parametrizable continuous curve C = z(t) a ≤ t ≤ b
piecewise-C1
is closed if z(a) = z(b). If, in addition, z(t1 ) ̸= z(t2 ) for all t1 < t2 with (t1 ,t2 ) ̸= (a, b), then we call
such closed curve simple.

R EMARK 3.2.2. The curve with shape “∞” is closed but not simple.

Here not to be confused with the terminology “topological closed”. For example, a straight line
with finite length is topological closed, but not closed in the sense of Definition 3.2.1. For later
convenience, we again clarify the following notion (despite we already introduced before):

D EFINITION 3.2.3. Let K be a topological closed set in C. We say that f is analytic near K
if there exists an open neighborhood Ω of K (i.e. an open set Ω such that K ⊂ Ω) such that f is
analytic in Ω. If K = {z} is a one point set, then we say that f is analytic near z. In particular, one
sees that f is analytic near z if and only if there exists ε > 0 such that f is analytic in the ball Bε (z).

The main theme of this section is to prove the following result, which somehow can be view as
a generalization of Exercise 3.1.7, see also [GM12, Theorems 6.6.2 and 6.6.3] for analogous result
on vector fields on Rn .

T HEOREM 3.2.4 (Cauchy closed curve theorem in rectangle). Let C be a parametrizable

continuous piecewise-C1 closed curve. If f is analytic near a topological closed rectangle R such
that C ⊂ R, then C f = 0.
R
 3.2. CAUCHY CLOSED CURVE THEOREM IN RECTANGLE 33

R EMARK 3.2.5. The above theorem holds true for any parametrization of the curve C . The
main point here is f has no singularity in the area enclosed by the curve C . If f has some singularity
inside it, then the above theorem does not hold. We will discuss such cases later in Chapter 5. We
also also prove a fairly general version of the Cauchy closed curve theorem later in Section 3.3.

L EMMA 3.2.6. Let C be a parametrizable continuous piecewise-C1 closed curve. If f (z) =

R
α + β z for some α, β ∈ C (that is, a linear function), then C f = 0.

PhROOF. If we definei F(z) := αz + 21 β z2 , by using Exercise 2.1.6, one has F ′ = f . By writing

C = z(t) a ≤ t ≤ b and using the fundamental theorem of line integral (Theorem 3.1.16), one
sees that Z Z
f = F ′ = F(z(b)) − F(z(a)) = 0,
C C
which immediately conclude our lemma. □

E XERCISE 3.2.7. Let {K (k) } be a sequence of nonempty compact sets in C ∼ = R2 such that
K (1) ⊃ K (2) ⊃ K (3) ⊃ · · · . Show that k∈N K (k) ̸= 0.
/ [Hint: consider the complement of K (k) .]
T

We now prove the following technical lemma.

L EMMA 3.2.8 (Rectangle lemma). Let Γ be the boundary of a topological closed rectangle R.
If f is analytic near R, then Γ f = 0.
R

P ROOF. Without loss of generality, we may choose a parametrization of Γ in counterclockwise

orientation, since the reverse orientation will gives a minus sign (Lemma 3.1.8), which does not
affect our lemma at all.
We split the topological closed rectangle R into 4 congruent subrectangles, by bisecting each
of the sides. We let Γ1 , Γ2 , Γ3 , Γ4 denote the boundaries (counterclockwise orientation) of the four
topological closed subrectangles (also in counterclockwise order) as in the following figure:

F IGURE 3.2.1. Splitting a rectangle into 4 congruent subrectangles

Since the integrals along the interior lines appear in the opposite directions and thus cancel
(Lemma 3.1.8), hence we see that
Z 4 Z
f=∑ f.
Γ i=1 Γi
 3.2. CAUCHY CLOSED CURVE THEOREM IN RECTANGLE 34

From this, one sees that

1
Z Z
f ≥ f for some i = 1, 2, 3, 4.
Γi 4 Γ

We denote Γ(0) = Γ and Γ(1) = Γi for i which satisfies the above inequality. Let R (1) be
the topological closed rectangle enclosed by the closed curve Γ(1) . We now show (by using
mathematical induction) that one can obtain a sequence of topological closed rectangles
1
Z Z
(3.2.1) R (1)
⊃R (2)
⊃R (3)
⊃ ··· with f ≥ k f ,
Γ(k) 4 Γ

where Γ(k) is the boundary of the R (k) .

We already show (3.2.1) when k = 1. We now assume the induction hypothesis that (3.2.1)
(ℓ) (ℓ) (ℓ) (ℓ)
holds for k = ℓ. We now splitting the topological closed rectangle R (ℓ) into R1 , R2 , R3 , R4
(ℓ) (ℓ)
as in Figure 3.2.1, where Γi are boundary of Ri . We again see that
Z 4 Z
f=∑ (ℓ)
f.
Γ(ℓ) i=1 Γi
From this, one sees that
1 1
Z Z Z
f ≥ f ≥ f for some i = 1, 2, 3, 4.
Γi
(ℓ)
4 Γ(ℓ) 4ℓ+1 Γ

(k)
We now denote choose Γ(ℓ+1) := Γi for i satisfies the above inequality and R (ℓ+1) be the
topological closed rectangle enclosed by the closed curve Γ(ℓ+1) . We now complete the proof
of (3.2.1) by induction.
By using Exercise 3.2.7, we know that k∈N R (k) ̸= 0. / We now fix one z0 ∈ k∈N R (k) . One
T T

sees that
f (z) − f (z0 ) f (z) − f (z0 ) − f ′ (z0 )(z − z0 )
lim = f ′ (z0 ) ⇐⇒ lim = 0.
z→z0 z − z0 z→z0 z − z0
For later convenience, we denote
f (z) − f (z0 ) − f ′ (z0 )(z − z0 )
oz :=
z − z0
so that
f (z) = f (z0 ) + f ′ (z0 )(z − z0 ) + oz · (z − z0 ), lim oz = 0.
z→z0
By using Lemma 3.2.6, we see that
Z Z
f= oz · (z − z0 ) dz.
Γ(n) Γ(n)
 3.2. CAUCHY CLOSED CURVE THEOREM IN RECTANGLE 35

Let s be the length of the largest side of the original boundary Γ (so that H 1 (Γ) ≤ 4s and |z − z0 | ≤
√
2s), then √
1 4s 2s
H (Γ ) = n H (Γ) ≤ n , sup |z − z0 | ≤ n .
1 (n) 1
2 2 z∈Γ(n) 2
By the definition of oz , given any ε > 0, there exists N such that
√
2s
|oz | ≤ ε for all |z − z0 | ≤ N ,
2
which shows that
sup |oz | ≤ ε for all n ≥ N.
z∈Γ(n)
By using Lemma 3.1.14 and (3.2.1), by fixing any n ≥ N, we see that
Z Z Z √
f ≤4 n
f =4 n
oz · (z − z0 ) dz ≤ ε4 2s2 .
Γ Γ(n) Γ(n)

We see that the first and last terms of the above are independent of N. By arbitrariness of ε, we
conclude our lemma. □

We now prove an important theorem, which is analogue to the fundamental theorem of calculus
(antiderivative).

T HEOREM 3.2.9 (Fundamental theorem of antiderivative in rectangle). If f is analytic near a

topological closed rectangle R, then there exists a function F which is analytic and F ′ = f near
R. Such analytic function F is called the (complex) antiderivative of f . Combining this with the
fundamental theorem of line integral (Theorem 3.1.16), we have
Z Z
(3.2.2) f= F ′ = F(z(b)) − F(z(a))
C C
h i
for any parametrizable continuous piecewise-C1 curve C = z(t) a ≤ t ≤ b ⊂ R.

R EMARK 3.2.10. We will later show a fairly general version of the above theorem in
Theorem 3.3.10 later.

P ROOF OF T HEOREM 3.2.9. Without loss of generality, we may assume 0 ∈ R. We define

Z z Z
(3.2.3) F(z) := f (ζ ) dζ ≡ f (ζ ) dζ ,
0 C1

where C1 denotes the oriented curve consists of the straight lines from 0 to Re z and then from Re z
to z. For each h ∈ C, we also denote
Z z+h Z
f (ζ ) dζ ≡ f (ζ ) dζ ,
z C2
 3.2. CAUCHY CLOSED CURVE THEOREM IN RECTANGLE 36

where C2 denotes the oriented curve consists of the straight lines from z to z + Re h and then from
z + Re h to z + h. By the definition (3.2.3), we have
Z z+h Z
F(z + h) = f (ζ ) dζ = f (ζ ) dζ ,
0 C3

where C3 denotes the oriented curve consists of the straight lines from 0 to Re (z + h) and then
from Re (z + h) to z + h. In particular, one sees that
Z z+h
F(z) + f (ζ ) dζ = F(z + h),
z
see the following figure:

F IGURE 3.2.2. The sketch of the curves C1 , C2 and C3

Since Z z+h
F(z + h) − F(z) = f (ζ ) dζ
z
and
Z z+h from z to z+Re h from z+Re h to z+h
1 1z }| { 1z }| {
1 dz = ((z + Re h) − z) + ((z + h) − (z + Re h)) = 1,
h z h h
then
F(z + h) − F(z) 1 z+h
Z
− f (z) = ( f (ζ ) − f (z)) dζ .
h h z
Since H 1 (C2 ) = |Re h| + |Im h| ≤ 2|h|, finally by Lemma 3.1.14 and the (uniform) continuity of
f , we have
F(z + h) − F(z) 1
lim sup − f (z) ≤ lim sup ∥ f − f (z)∥L∞ (C2 ) H 1 (C2 )
C∋h→0 h C∋h→0 |h|
≤ 2 lim sup ∥ f − f (z)∥L∞ (C2 ) = 0
C∋h→0

which conclude our theorem. □

 3.3. CAUCHY CLOSED CURVE THEOREM IN SIMPLY CONNECTED OPEN SETS 37

With this fundamental theorem at hand, we finally now ready to prove the main result of this
section, that is, the Cauchy closed curve theorem in rectangle.
h i
P ROOF OF T HEOREM 3.2.4. Write C = z(t) a ≤ t ≤ b ⊂ R with z(a) = z(b). Since f is
analytic near R, by the fundamental theorem of antiderivative in rectangle (Theorem 3.2.9), there
exists a function F which is analytic and F ′ = f near R such that
Z Z
f= F ′ = F(z(b)) − F(z(a)) = 0,
C C
which immediately conclude the theorem. □

3.3. Cauchy closed curve theorem in simply connected open sets

In this section will prove a version of Cauchy closed curve theorem, which generalized
Theorem 3.2.4. The main theme of this section is to remove the analyticity assumption on
rectangles. Let A and B are sets, then we denote the distance between them by

dist (A, B) = inf |a − b|.

a∈A,b∈B

If A is a one point set {z0 }, we simply denote dist (z0 , B).

D EFINITION 3.3.1. Let Ω be an open set. If Ω is connected and its complement is “connected
to ∞ by a continuous curve within ε-neighborhood of C \ Ω” in the following sense: if for any
z0 ∈
/ Ω and ε > 0, there is a continuous curve γ = [γ(t) : 0 ≤ t < ∞] such that

dist (γ(t), C \ Ω) < ε for all t ≥ 0, γ(0) = z0 , lim |γ(t)| = ∞,

t→∞

then we call such set Ω is simply connected open set in C.

n o
E XAMPLE 3.3.2. The annulus z ∈ C 1 < |z| < 3 is not simply connected, because its
complement cannot be “connected to ∞ by a continuous curve within ε-neighborhood of C \ Ω”.
n o
E XAMPLE 3.3.3. The infinite strip S = z ∈ C −1 < Im z < 1 is connected. Note that in
this case, C \ S is not connected.

E XERCISE 3.3.4. A set S is called star-like if there exists a point α ∈ S such that the line
segment connecting α and z is contained in S for all z ∈ S. Show that a star-like region is simply
connected.

E XAMPLE 3.3.5. The complement of the connected domain

n o n o
x + iy ∈ C 0 < x ≤ 1, y = sin 1x ∪ iy ∈ C −1 < y < ∞

is simply connected.

D EFINITION 3.3.6. Let Γ be a polygonal path (Definition 1.2.16) consists of horizontal lines
and vertical lines, i.e. either parallel to real axis or parallel to v the imaginary axis. The y0 -level is
 3.3. CAUCHY CLOSED CURVE THEOREM IN SIMPLY CONNECTED OPEN SETS 38

the set n o
Γy0 := x + iy0 x ∈ R ∩ Γ.

If Γy1 , · · · , Γyn , for some y1 > y2 > · · · > yn , are all levels of Γ, then we say that n ∈ N is the
number of levels. We also say Γy1 the top level of Γ. We also say that Γy j+1 is the next level of Γy j .

E XERCISE 3.3.7. Let K be a compact set in C and let F be a topological closed set in C. If
K ∩ F = 0,
/ show that dist (K, F) > 0. On the other hand, construct topological closed sets F1 , F2 in
C such that F1 ∩ F2 = 0/ but dist (F1 , F2 ) = 0.

L EMMA 3.3.8. Let Γ be a simple closed polygonal path (Definition 3.2.1) consists of horizontal

lines and vertical lines, such that it contained in a simply connected open set Ω. Let Γy1 , · · · , Γyn ,
for some y1 > y2 > · · · > yn , be all levels of Γ. Let X1 be the topological closed set in R such that
n o
Γy0 = x + iy1 x ∈ X1 .
n o
Then the set R := z = x + iy x ∈ X1 , y2 ≤ y ≤ y1 is contained in Ω.

S KETCH OF PROOF. Note that R is a finite union of disjoint topological closed rectangles. In
addition, by using Exercise 3.3.7, we also see that δ := dist (Γ, C \ Ω) > 0. Let z0 ∈ R and let γ be
any continuous curve which “connecting z0 to ∞” in the sense of γ = [γ(t) : 0 ≤ t < ∞] with

γ(0) = z0 , lim |γ(t)| = ∞.

t→∞

In fact, we have γ ∩ Γ ̸= 0,
/ this is just simply the fact that, a connected line from R (inside the region
bound by Γ) to outside the region bound by Γ, must pass through the boundary. One can refer to
[BN10, Chapter 8] for those technical details.
We now want to show z0 ∈ Ω. Suppose the contrary, that z0 ∈ / Ω. Since Ω is simply connected,
there exists a continuous curve γ0 “connected to ∞ by a continuous curve within δ2 -neighborhood
of C \ Ω”, that is,
δ
dist (γ0 (t), C \ Ω) < for all t ≥ 0, γ0 (0) = z0 , lim |γ0 (t)| = ∞.
2 t→∞

The previous paragraph says that γ0 ∩ Γ ̸= 0,/ then there exists t0 ≥ 0 such that γ0 (t0 ) ∈ Γ. From this,
we have
dist (γ0 (t0 ), C \ Ω) ≥ dist (Γ, C \ Ω) = δ ,
which is a contradiction. □

We now generalize the rectangle lemma (Lemma 3.2.8).

L EMMA 3.3.9. Let f be an analytic function on a simply connected open set Ω, and let Γ be a
simple closed polygonal path consists of horizontal lines and vertical lines, which contained in Ω.
R
Then Γ f = 0.
 3.3. CAUCHY CLOSED CURVE THEOREM IN SIMPLY CONNECTED OPEN SETS 39

S KETCH OF PROOF. We will prove the result by induction on the number of levels. If Γ has
only two levels, then Γ is simply the boundary of a closed rectangle, and this case can be concluded
by the rectangle lemma (Lemma 3.2.8). The induction step can be done as in the following diagram:

F IGURE 3.3.1. Induction hypothesis for Γ with j-levels (red), and induction step
(blue)

Each induction step are done by the rectangle lemma (Lemma 3.2.8). □

From this, we can obtain Fundamental theorem of antiderivative in simply connected domain.

T HEOREM 3.3.10 (Fundamental theorem of antiderivative in simply connected domain). If f

is an analytic function on a simply connected open set Ω, then there exists a function F which is
analytic and F ′ = f in Ω. Similarly, such analytic function F is called the (complex) antiderivative
of f . Combining this with the fundamental theorem of line integral (Theorem 3.1.16), we have
Z Z
(3.3.1) f= F ′ = F(z(b)) − F(z(a))
C C
h i
for any parametrizable continuous piecewise-C1 curve C = z(t) a ≤ t ≤ b ⊂ R.

S KETCH OF PROOF. Choose z0 ∈ Ω and define

Z z
F(z) = f (ζ ) dζ ,
z0

where the path of integration is the simple polygonal path consists of horizontal lines and vertical
lines, which contained in Ω. This is well-defined by the rectangle lemma (Lemma 3.3.9). Then the
rest of proof can be done as in Theorem 3.2.9, which we leave it as an exercise. □

Finally, we state (without proof) the Cauchy closed curve theorem which we needed, which can
be proved using Theorem 3.3.10 following the arguments in Theorem 3.2.4. We leave the proof as
an exercise.

T HEOREM 3.3.11 (Cauchy closed curve theorem in simply connected open set). Let f be an
analytic function on a simply connected open set Ω. For each parametrizable continuous piecewise-
C1 closed curve C which contained in Ω, then C f = 0.
R
 3.3. CAUCHY CLOSED CURVE THEOREM IN SIMPLY CONNECTED OPEN SETS 40

E XERCISE 3.3.12. Let Ω be a simply connected open set in C ∼ = R2 , and let u ∈ C2 (Ω) be a
given real-valued harmonic function (i.e. ∂x2 u + ∂y2 u = 0 in Ω). Show that there exists a real-valued
function v ∈ C2 (Ω) such that F = u + iv is analytic in Ω. [Hint: First define f := ∂x u − i∂y u and the
consider its antiderivative F as in Theorem 3.3.10. Show that ℜF = u.]
 CHAPTER 4

Properties of Analytic functions

Now we have obtained some fundamental tools connecting the differentiation and integration.
We now ready to further study the analytic functions. We first consider the simplest case: the entire
functions, which is analytic in the whole C.

4.1. Cauchy integral formula for entire functions

We now try to study the situation stated in Remark 3.2.5. In order to deal with this case, for
each point a ∈ C and an entire function f , we define the auxiliary function

 f (z)− f (a) , z ̸= a,
z−a
(4.1.1) g(z) =
 f ′ (a) , z = a.
It is clear that g is continuous. One of the main theme of this section is to prove that g is entire. We
first prove the following technical lemma, which sometimes also referred as “rectangle theorem”.
L EMMA 4.1.1. Let f be an entire function and let g be the auxiliary function given in (4.1.1).
If Γ is the boundary of a topological closed rectangle R, then Γ g = 0.
R

/ R, then clearly g is analytic near R, and the lemma immediately follows from
P ROOF. If a ∈
Cauchy closed curve theorem (Theorem 3.3.11).
For the case when a ∈ Γ = ∂ R, by using Cauchy closed curve theorem (Theorem 3.3.11) one
sees that Z Z
g= g,
Γ Γ1
where Γ1 ∋ a is the boundary of the square with side length ε, as showed in the following figure):

F IGURE 4.1.1. The sketch of the curves Γ1 and Γ

By using Lemma 3.1.14, it is easy to see that

Z Z
g = g ≤ 4∥g∥L∞ (Γ1 ) ε ≤ 4∥g∥L∞ (R) ε.
Γ Γ1
41
 4.1. CAUCHY INTEGRAL FORMULA FOR ENTIRE FUNCTIONS 42
R
By arbitrariness of ε > 0, we conclude that Γ g = 0.
For the case when a ∈ int (R), by using Cauchy closed curve theorem (Theorem 3.3.11), one
sees that Z Z
g= g,
Γ Γ2
where Γ2 is the boundary of the square (which containing a in its interior) with side length ε, as
showed in the following figure:

F IGURE 4.1.2. The sketch of the curves Γ2 and Γ

As in previous case, by using Lemma 3.1.14, it is easy to see that

Z Z
g = g ≤ 4∥g∥L∞ (Γ2 ) ε ≤ 4∥g∥L∞ (R) ε.
Γ Γ2
R
By arbitrariness of ε > 0, we conclude that Γ g = 0. □

The following exercise can be done using similar arguments as in the Fundamental theorem
of antiderivative in rectangle (Theorem 3.2.9) and the Cauchy closed theorem in rectangle
(Theorem 3.2.4):

E XERCISE 4.1.2. Let a ∈ C and let f be an entire function. Show that there exists an entire
function G such that G′ = g, where g is the auxiliary function given in (4.1.1). In addition, one
also has C g = 0 for all parametrizable continuous piecewise-C1 closed curve C . [Hint: g is
R

continuous.]

R EMARK 4.1.3. Even though we have Lemma 4.1.2, we still don’t know whether g is entire or
not. At this point, we do not know yet whether the (complex) derivative of entire function is also
entire or not.

We now prove the following lemma, which is related to Remark 3.2.5.

 4.1. CAUCHY INTEGRAL FORMULA FOR ENTIRE FUNCTIONS 43

L EMMAh 4.1.4. If Cρ (z0 ) is the iboundary of Bρ (z0 ) in counterclockwise orientation, that is,
Cρ (z0 ) = Reiθ + z0 0 ≤ θ ≤ 2π , then

1
Z
dz = 2πi for all a ∈ Bρ (z0 ).
Cρ (z0 ) z − a

P ROOF. We first consider the case when a = z0 . In this case, from the definition of line integral
(Definition 3.1.4), we see that
iReiθ
Z 2π
1
Z
dz = dθ = 2πi.
Cρ (z0 ) z − z0 0 Reiθ
By using the fundamental theorem of line integral (Theorem 3.1.16)
 
1 1
Z Z
2
dz = − ∂z dz = 0.
Cρ (z0 ) (z − z0 ) Cρ (z0 ) z − z0
Inductively, we also see that
 
1 1 1
Z Z
(4.1.2) k+1
dz = − ∂z dz = 0 for all k = 1, 2, · · · .
Cρ (z0 ) (z − z0 ) k Cρ (z0 ) (z − z0 )k
We now prove Lemma 4.1.4 for a ∈ Bρ (z0 ). We write
1 1 1 1
= = · for all z ∈ Cρ (z0 ).
z − a (z − z0 ) − (a − z0 ) z − z0 1 − a−z
z−z
0
0

Since
a − z0 |a − z0 |
(4.1.3) = <1 for all z ∈ Cρ (z0 ),
z − z0 ρ
1
then the fact that = 1 + w + w2 + · · · for w ∈ C with |w| < 1 (geometric sequence), then
1−w
!
a − z0 2 (a − z0 )k
  ∞
1 1 a − z0
(4.1.4) = · 1+ + +··· = ∑ k+1
for all z ∈ Cρ (z0 ).
z − a z − z0 z − z0 z − z0 k=0 (z − z0 )

Again by (4.1.3), we have

n
(a − z0 )k 1 ∞
(a − z0 )k
lim sup ∑ (z − z0)k+1 − z − a = lim sup ∑ k+1
n→∞ k=0 L∞ (Cρ (z0 ))
n→∞ k=n+1 (z − z0 ) L∞ (Cρ (z0 ))

(a − z0 )k ∞ 
|a − z0 | k
∞ 
1
≤ lim sup ∑ = lim sup ∑ = 0,
n→∞ k=n+1 (z − z0 )k+1 L∞ (Cρ (z0 )) ρ n→∞ k=n+1 ρ
that is, the convergence in (4.1.4) is uniform. Therefore, from (4.1.2) we obtain
∞
1 1 1
Z Z Z
dz = dz + ∑ (a − z0 )k k+1
dz = 2πi,
Cρ (z0 ) z − a Cρ (z0 ) z − z0 k=1 Cρ (z0 ) (z − z0 )

which conclude our lemma. □

 4.2. POWER SERIES (WITH R = ∞) AND ENTIRE FUNCTION 44

Warning: In general the infinite sum and integral are not commute. The uniform
convergence is a sufficient condition that guarantees that this idea work.
E XERCISE 4.1.5. Prove (4.1.2) by direct evaluation in the definition of line integral
(Definition 3.1.4).
We now ready to state and proof the main theorem of this section.
T HEOREM 4.1.6h (Cauchy integral formula
i for entire functions). Let f be an entire function, let
a ∈ C and let C = Re iθ 0 ≤ θ ≤ 2π with R > |a|. Then

1 f (z)
Z
f (a) = dz.
2πi C z − a
P ROOF. By Exercise 4.1.2 and Lemma 4.1.4, one has
f (z) − f (a) f (z) 1 f (z)
Z Z Z Z
0= dz = dz − f (a) dz = dz − 2πi f (a),
C z−a C z−a C z−a C z−a
which conclude our theorem. □

4.2. Power series (with R = ∞) and entire function

In Chapter 2 we have showed that each power series represents an analytic function inside its
domain of convergence. In real analysis, it is known that there exists a C∞ function such that its
Taylor expansion does not converges to it. For example, we consider the function
 1
e− x2 , x > 0,
f (x) =
0 , x ≤ 0,

which is in C∞ (R) but f (n) (0) = 0 for all n ∈ N (so that its Taylor expansion at 0 vanishes
identically, therefore does not converge to f ). In other words, the differentiability (existence of
partial derivatives) does not guarantee the convergence of Taylor sequence. However, the complex
differentiation has the following surprising properties, which is the main result of this section:
T HEOREM 4.2.1. f is entire if and only if it has a power series representation
(centered at each a ∈ C with radius of convergence = ∞). In this case, for each a ∈ C, the complex
derivatives { f (k) (a)}∞
k=1 exist and satisfies
∞
f (k) (a)
(4.2.1) f (z) = ∑ (z − a)k for all z ∈ C.
k=0 k!
R EMARK 4.2.2. The above theorem means that { f (k) (a)}∞
k=1 exist for all a ∈ C, that is, f is
infinitely complex differentiable.
T HEOREM 4.2.1. If f has a power series representation at a ∈ C with radius of convergence =
∞, i.e. there exist Ck ∈ C such that f (z) = ∑∞ k
k=0 Ck (z − a) for all z ∈ C. By applying Theorem 2.2.9
g(z) = f (z + a) = ∑∞ k
k=0 Ck z , we know that g is entire, and so is f .
 4.2. POWER SERIES (WITH R = ∞) AND ENTIRE FUNCTION 45

Conversely, we now suppose that f is entire. Given any a ∈ C, we define the entire function
g(z) := f (z + a) for all z ∈ C. If we can show that
g(k) (0) k
∞
(4.2.2) g(z) = ∑ z for all z ∈ C,
k=0 k!
g (0)(k) f (a) (k)
then f (z) = g(z − a) = ∑∞ k ∞ k
k=0 k! (z − a) = ∑k=0 k! (z − a) , which conclude (4.2.1).
It is remain to prove (4.2.2). Given any z ∈ C, one can choose R > 0 such that |z| < R. By using
the Cauchy integral formula for entire function, one has
1 g(w)
Z
g(y) = dw for all y ∈ BR .
2πi ∂ BR w − y

By using the geometric sequence (which used in the proof of Lemma 4.1.4), one sees that
1 1 1 y y2 ∞
yk
= = + + + · · · = ∑ k+1
w − y w(1 − wy ) w w2 w3 k=0 w

which is uniformly converge, so that

∞ Z 
1 g(w)
g(y) = ∑ k+1
dw yk for all y ∈ BR .
k=0 2πi ∂ BR w

Then by Exercise 2.2.10, one reaches

g(k) (0)
Z 
1 g(w)
dw = for all k = 0, 1, 2, · · ·
2πi ∂ BR wk+1 k!
and hence
g(k) (0) k
∞
g(y) = ∑ k! y for all y ∈ BR .
k=0
Since z ∈ BR , then
∞
g(k) (0) k
g(z) = ∑ k! z .
k=0
Since the above procedure holds true for all z ∈ C, hence we conclude (4.2.2). □

E XERCISE 4.2.3 (Higher order

h Cauchy integral formula
i for entire functions). Let f be an entire
function, let a ∈ C and let C = Re iθ 0 ≤ θ ≤ 2π with R > |a|. Show that

k! f (z)
Z
(k)
f (a) = dz for all k = 0, 1, 2, · · ·
2πi C (z − a)k+1

P ROPOSITION 4.2.4. If f is entire, then the auxiliary function g given in (4.1.1) is also entire.

P ROOF. We can write (4.2.1) as

∞
f (k) (a)
f (z) − f (a) = ∑ (z − a)k for all z ∈ C,
k=1 k!
 4.3. LIOUVILLE THEOREM AND THE FUNDAMENTAL THEOREM OF ALGEBRA 46

where we choose a ∈ C be the number as in (4.1.1). Dividing the above equation by (z − a), we
reach
f (z) − f (a) ∞
f (k) (a) ∞
f (m+1) (a)
g(z) ≡ =∑ (z − a)k−1 = ∑ (z − a)m for all z ̸= a.
z−a k=1 k! m=0 (m + 1)!
Since g is continuous on C, and the right hand side of the above inequality is entire (hence
continuous), thus the above identity also holds for all z ∈ C, which completes the proof. □
E XERCISE 4.2.5. Suppose that f is entire with zeros a1 , a2 , · · · aN , that is, f (ak ) = 0 for k =
1, 2, · · · , N, and we define
f (z)
g(z) := for all z ∈ C \ {a1 , a2 , · · · , aN }.
(z − a1 )(z − a2 ) · · · (z − aN )
Show that if limz→ak g(z) exists for all k = 1, 2, · · · , N, then the extension g̃ of g defined by

g(z) , z ∈ C \ {a1 , a2 , · · · , aN },
g̃(z) :=
 lim g(z) , z = ak for k = 1, 2, · · · , N,
z→ak

is also entire.

4.3. Liouville theorem and the fundamental theorem of algebra

By using the Cauchy integral formula for entire functions, we also can obtain some powerful
tools, which are well-known.
T HEOREM 4.3.1 (Liouville theorem). A bounded entire function is constant.
P ROOF. Let a and b represent any two complex numbers and let C be any positively oriented
(i.e. counter clockwise oriented) circle centered at 0 and with radius R > max{|a|, |b|}. By using
the Cauchy integral formula for entire functions (Theorem 4.1.6), we see that
1 f (z) 1 f (z) 1 f (z)(b − a)
Z Z Z
f (b) − f (a) = dz − dz = dz.
2πi C z−b 2πi C z−a 2πi C (z − a)(z − b)
Since the arc length H 1 (C ) of C is 2πR, then
1 ∥ f ∥L∞ (C ) |b − a| ∥ f ∥L∞ (C) |b − a|
| f (b) − f (a)| ≤ H 1 (C ) ≤ R.
2π (R − |a|)(R − |b|) (R − |a|)(R − |b|)
Taking R → ∞ (in the sense of limit supremum), we conclude f (a) = f (b). Since a, b are arbitrary,
then we conclude our theorem. □
T HEOREM 4.3.2 (Extended Liouville theorem). Let A > 0, B > 0 and k ∈ Z≥0 . If the entire
function f satisfies

(4.3.1) | f (z)| ≤ A + B|z|k for all z ∈ C,

then f is an analytic polynomial of degree at most k.

 4.3. LIOUVILLE THEOREM AND THE FUNDAMENTAL THEOREM OF ALGEBRA 47

P ROOF. We prove the above result by induction on k. The statement for k = 0 is just simply
Theorem 4.3.1.
It is suffice to prove the result for k = ℓ + 1 if Theorem 4.3.2 holds true for some k = ℓ ≥ 0. Let
g be the auxiliary function given in 4.1.1 and choosing a = 0. From Proposition 4.2.4 we know that
such g is entire. We also see that
| f (z) − f (0)| | f (z)| + | f (0)| 2A + B|z|ℓ+1
|g(z)| = ≤ ≤ ≤ 2A + B|z|ℓ for all |z| ≥ 1,
|z| |z| |z|
and thus
|g(z)| ≤ ∥g∥L∞ (B1 ) + 2A + B|z|ℓ .
By using the induction hypothesis that Theorem 4.3.2 holds true for k = ℓ ≥ 0, we know that g is
an analytic polynomial of degree at most ℓ. Since

f (z) = zg(z) + f (0) for all z ̸= 0,

by analyticity of both f and g, in particular the above identity also holds true for all z ∈ C. Therefore
f is analytic polynomial of degree at most ℓ + 1. This conclude Theorem 4.3.2 by induction. □
3
E XERCISE 4.3.3. Suppose f is entire and | f (z)| ≤ A + B|z| 2 for all z ∈ C. Show that f is linear
polynomial.

E XERCISE 4.3.4. Suppose f is entire and | f ′ (z)| ≤ |z| for all z ∈ C. Show that f (z) = a + bz2
with |b| ≤ 21 .

L EMMA 4.3.5. Let P(z) be a analytic polynomial which is not identical to a constant function.
Then there exists z0 ∈ C such that P(z0 ) = 0.

P ROOF. Suppose the contrary that such z0 ∈ C does not exist, that is, P(z) ̸= 0 for all z ∈ C.
1
Then by Lemma 2.1.5 one sees that f (z) := P(z) is an entire function. Since P is non-constant, then
we can write
N
P(z) = ∑ c jz j
j=0
for some N ∈ N with cN ̸= 0 and cn = 0 for all n > N. Then we see that
!
N−1
lim inf |P(z)| ≥ lim inf |cN ||z|N − ∑ |c j ||z| j = ∞,
z→∞ z→∞
j=0

which shows that

lim | f (z)| = 0.
z→∞
Therefore f is a bounded entire function, which is a constant by Liouville theorem (Theorem 4.3.1),
this shows that P must identical to a constant function, which is a contradiction. □

We finally end this section by proving an important theorem in the field theory.
 4.4. THE ROOTS OF ±1 48

T HEOREM 4.3.6 (Fundamental theorem of algebra). Let P(z) be an analytic polynomial which
is not identical to a constant function, then there exists A, α1 , · · · , αN ∈ C such that P(z) = A(z −
α1 ) · · · (z − αN ) for all z ∈ C. In other words, the complex field C is algebraically complete.

P ROOF. Write P(z) = ∑Nj=0 c j z j for some N ∈ N with cN ̸= 0. Similar in the proof of the
extended Liouville theorem (Theorem 4.3.2), we see that the auxiliary function g given in 4.1.1 and
choosing a = α satisfies
|g(z)| ≤ A + B|z|N−1 ,
and hence by the extended Liouville theorem (Theorem 4.3.2), g must be an analytic polynomial.
Again, similar in the proof of the extended Liouville theorem (Theorem 4.3.2), we have

P(z) = g(z)(z − α) for all z ∈ C,

this shows that g must be a polynomial of degree N − 1. Repeating the above arguments on g, we
conclude our theorem. □

4.4. The roots of ±1

We now include some materials from [FB09]. In the very beginning of this course, we
√
asked a question regarding how to define −1. By using the fundamental theorem of algebra
(Theorem 4.3.6), we now know that the equation z2 + 1 = 0 has exactly two solutions in C, and
they are ±i. As a corollary, we note that

the equation z2 + 1 = 0 has no roots in R.

Therefore, the polynomial P(z) = z2 + 1 is irreducible in R[z]. For convenience, we usually write
√ √
−1 := i, but one should be aware that −1 is not well-defined as a function in general. In
√
complex analysis, we call −i is another branch of −1.
√
It is well-known that the n-root of 1 is well-defined in R, which is given by n 1 = 1. However,
in complex field, we have the following interesting observation (one also asks similar questions in
finite field):

T HEOREM 4.4.1. For each n ∈ N, there are exactly n different solutions {ζ j }nj=1 (or roots) of
zn − 1 = 0, and they have the formula
2π j 2π j
(4.4.1) ζ j = cos + i sin for j = 0, 1, 2, · · · , n − 1.
n n
We called (4.4.1) the nth roots of unity. We also called zn − 1 the cyclotomic equation, since (4.4.1)
is exactly the vertex of regular n-gon in C

P ROOF. By using Exercise 2.3.2, one can directly verify that (4.4.1) are n different roots of
zn − 1 = 0. By using the fundamental theorem of algebra (Theorem 4.3.6), they are exactly all the
n different solutions. □
 4.5. CAUCHY INTEGRAL FORMULA IN A BALL 49

E XERCISE 4.4.2 (n-roots of −1). For each integer n ≥ 2, determine all roots of the equation
zn + 1 = 0.

4.5. Cauchy integral formula in a ball

We have proved the Cauchy integral formula for entire functions in Section 4.1. By carefully
inspecting the arguments, in fact we can obtain a local version. Here we will exhibit the details.
Let f be an analytic function in a ball Br (z0 ). By using the fundamental theorem of antideriva-
tive in rectangle (see Theorem 3.2.9 and (3.2.3)), one sees that the function
Z z Z
F(z) = f (ζ ) dζ ≡ f (ζ ) dζ is analytic and satisfies F ′ = f on Br (z0 ),
z0 C

where C denotes the oriented curve consists of the straight lines from z0 to z0 + Re (z − z0 ) and then
from z0 + Re (z − z0 ) to z. It is important to notice that one can find a topological closed rectangle
consists of z0 and z which is contained in Br (z0 ).
We consider the auxiliary function g similar to (4.1.1): If f is analytic in Br (z0 ) and a ∈ Br (z0 ),
then we define the function

 f (z)− f (a) , z ∈ B (z ) \ {a},
z−a r 0
(4.5.1) g(z) =
 f ′ (a) , z = a,

which is continuous on Br (z0 ). At this moment, we don’t know whether g is analytic in D yet.
However, by continuity of g and following the same arguments as in Exercise 4.1.2, one can show
that

(4.5.2) there exists an analytic function G with G′ = g on Br (z0 ).

In addition, one also has

Z
(4.5.3) g = 0 for all parametrizable continuous piecewise-C1 closed curve C ⊂ Br (z0 ).
C
We now can easily proof the local version of Cauchy integral formula.

T HEOREM 4.5.1 (Cauchy integral formula in a ball). Suppose that f is analytic in Br (z0 ) and
let a ∈ Br (z0 ). For each 0 < ρ < r with a ∈ Bρ (z0 ), one has
1 f (ω)
Z
f (a) = dω,
2πi Cρ (z0 ) ω − a
h i
where Cρ (z0 ) is the closed curve Cρ (z0 ) = z0 + ρeiθ 0 ≤ θ ≤ 2π , that is, Cρ (z0 ) = ∂ Bρ (z0 )
with counterclockwise oriented.
 4.6. POWER SERIES (WITH R < ∞) AND ANALYTIC FUNCTION 50

P ROOF. Let g be the auxiliary function given in (4.5.1). By using (4.5.3) and Lemma 4.1.4,
one has
f (ω) − f (a) f (ω) f (a) f (ω)
Z Z Z Z
0= dω = dω − dω = dω − 2πi f (a),
Cρ (z0 ) ω −a Cρ (z0 ) ω − a Cρ (z0 ) ω − a Cρ (z0 ) ω − a

which conclude our theorem. □

E XERCISE 4.5.2. Let Ω be an open set, let f be an analytic function on Ω and let a ∈ Ω. Show
that f (a) is equal to the mean value of f takes around the boundary of any disc centered at a
contained in D, that is,
1 2π
Z
f (a) = f (a + reiθ ) dθ
2π 0
whenever ∂ Br (a) ⊂ D.

R EMARK 4.5.3. As we see in Remark 2.1.11, an analytic function always a harmonic function.
In fact, the mean value theorem also holds true for harmonic function, see [GT01]. This even holds
true for Helmholtz operator ∆ + k2 , see e.g. my work [KLSS24, Appendix].

4.6. Power series (with R < ∞) and analytic function

In Chapter 2 we have showed that each power series represents an analytic function inside its
domain of convergence. We denote R be its radius of convergence. In Section 4.2 we have showed
the converve of this theorem for the case when R = ∞. We now turn to the question about the case
when R < ∞.

T HEOREM 4.6.1. If f is analytic in BR (z0 ), then there exist constants Ck such that
∞
f (z) = ∑ Ck (z − z0)k for all z ∈ BR (z0 ).
k=0

P ROOF. For each 0 < ρ < R, by using the Cauchy integral formula in a ball (Theorem 4.5.1)
with a = z, we have (Theorem 4.5.1)
1 f (ω)
Z
f (z) = dω for all z ∈ Bρ (z0 ).
2πi Cρ (z0 ) ω − z

Recall (4.1.4) and changing the notation z → ω and a → z:

!
z − z0 2 (z − z0 )k
  ∞
1 1 z − z0
= · 1+ + +··· = ∑ k+1
for all ω ∈ Cρ (z0 ),
ω − z ω − z0 ω − z0 ω − z0 k=0 (ω − z0 )

which converges uniformly on Cρ (z0 ). Combining the above two equations, we reach
Z 
1 ∞ f (ω)
f (z) = ∑ Cρ (z0) (ω − z0)k+1 dω (z − z0)k .
2πi k=0
 4.6. POWER SERIES (WITH R < ∞) AND ANALYTIC FUNCTION 51

Arguing as in Theorem 4.2.1 (which involving Exercise 2.2.10), we again have

1 f (ω) f (k) (z0 )
Z
k+1
dω = ,
2πi Cρ (z0 ) (ω − z0 ) k!
and thus
f (k) (z0 )
f (z) = (z − z0 )k for all z ∈ Bρ (z0 ).
k!
Since 0 < ρ < R is arbitrary, then we conclude our theorem. □

From Theorem 4.6.1, we immediately conclude the following corollary.

C OROLLARY 4.6.2 (Local power series representation). Let Ω be an open set in C. Then f is
analytic in Ω if and only if it has a local power series near each point in Ω, i.e. for each z0 ∈ Ω we
can write f as
∞
f (z) = ∑ Ck (z − z0)k
k=0
for all z ∈ BR (z0 ), where R = sup r. In this case, the complex derivatives { f (k) (z0 )}∞
k=1 exist
Br (z0 )⊂Ω
and satisfies
∞
f (k) (z0 )
(4.6.1) f (z) = ∑ (z − z0 )k
k=0 k!
for all z ∈ BR (z0 ), where R = sup r.
Br (z0 )⊂Ω

R EMARK 4.6.3. One sees that Theorem 4.2.1 is just a special case Ω = C of Corollary 4.6.2.
One should aware that the power series (4.6.1) in general not holds for all z ∈ D, i.e. not global!
See Remark 2.2.5. This is the reason why we called (4.6.1) the local power series.

P ROPOSITION 4.6.4. If f is analytic near a, then so is the auxiliary function g given in (4.5.1).

P ROOF. By using Corollary 4.6.2, we see that

∞
f (k) (a)
f (z) − f (a) = ∑ (z − a)k for all z near a.
k=1 k!
and thus
f (z) − f (a) ∞
f (k) (a) ∞
f (ℓ+1) (a)
g(z) = =∑ (z − a)k−1 = ∑ (z − a)ℓ for all z ̸= a near a.
z−a k=1 k! ℓ=0 (ℓ + 1)!
By continuity of g, we see that the above identity also holds true for z = a, which conclude our
proposition. □

T HEOREM 4.6.5 (Unique continuation property). Let f be an analytic function on an open

connected set Ω. If there exists a nonempty open set D ⊂ Ω such that f |D = 0, then f ≡ 0 in Ω.
 4.6. POWER SERIES (WITH R < ∞) AND ANALYTIC FUNCTION 52

R EMARK 4.6.6. By using the Carleman estimate, this property can be extended to large class
of solution of elliptic equations and systems (recall that analytic function also harmonic, see also
Remark 2.1.11). A related problem is called the Landis conjecture, which can be referred as the
unique continuation property from infinity.

P ROOF OF T HEOREM 4.6.5. We will prove this using a standard argument for open connected
set in Remark 1.2.19. We define
( )
there exists a sequence {zn } ⊂ Ω \ {z0 } such that
A := z0 ∈ Ω .
zn → z0 and f (zn ) = 0 for all n ∈ N
We first show that A is open (in C iff relative to Ω, since Ω is open, see Remark 1.2.15). Let z0 ∈ A.
By Corollary 4.6.2, one can represent f using a local power series near z0 , that is, there exists ε > 0
such that f (z) = ∑k Ck (z − z0 )k for all z ∈ Bε (z0 ). Then by the uniqueness theorem of power series
(Theorem 2.2.11) we see that f = 0 in Bε (z0 ), and hence Bε (z0 ) ⊂ A. By arbitrariness of z0 ∈ A,
we conclude that A is open. On the other hand, we want to show that Ω \ A is open as well. For
each z0 ∈ Ω \ A, by using a contradiction argument, one can show that there exists δ > 0 such that1

f (w) ̸= 0 for all w ∈ Bδ (z0 ) \ {z0 }.

By continuity of f , one sees that Bδ (z0 ) \ {z0 } ⊂ Ω \ A. Since z0 ∈ Ω \ A, we now reach that
Bδ (z0 ) ⊂ Ω \ A. By arbitrariness of z0 ∈ Ω \ A, this shows that Ω \ A is open. Since both A and
Ω \ A are open, by connectness of Ω, we see that either Ω = 0/ or Ω = A. Since A ⊃ D ̸= 0, / we
finally conclude that Ω = A, which conclude our theorem. □

C OROLLARY 4.6.7 (Uniqueness theorem). Let f be an analytic function on an open connected

set Ω. If there exists a sequence {zn } ⊂ Ω \ {z0 } such that zn → z0 ∈ Ω and f (zn ) = 0 for all n ∈ N,
then f ≡ 0 in Ω.

P ROOF. By using Corollary 4.6.2, one can represent f using a local power series near z0 . By
using the uniqueness theorem of power series (Theorem 2.2.11), one sees that there exists r > 0
such that f |Br (z0 ) = 0. Hence our result immediately follows from the unique continuation property
of analytic function (Theorem 4.6.5). □

E XAMPLE 4.6.8. We consider f (z) = sin z, which is entire (i.e. analytic in Ω = C). One sees
that f has at least infinitely many zeros: f (nπ) = 0 for all n ∈ Z. These zeros does not converge
in C. In fact, by using Corollary 4.6.7, the set of zeros of f does not have a limit point. Therefore,
given any bounded set, it contains at most finitely many zeros of f .

E XAMPLE 4.6.9. We consider f (z) = sin( 1z ), which is analytic in Ω = C \ {0}. One sees that
1
f has infinitely many zeros: f ( nπ ) = 0 for all n ∈ Z, and these zeros converge at 0. This illustrate
the analyticity assumption in Corollary 4.6.7 is essential.
1However, this does not exclude the possibility for f (z ) = 0.
0
 4.7. MORERA THEOREMS 53

T HEOREM 4.6.10. If f is entire and if | f (z)| → ∞ as |z| → ∞, then f is a polynomial.

P ROOF. By hypothesis, there exists R > 0 such that | f (z)| > 1 for all |z| > R. This shows
that f cannot have any zeros outside BR (0), and hence there at most finitely many zeros in BR (0).
If not, by using Bolzano-Weierstrass theorem, there exists a sequence {zn } ⊂ BR (0) converges to
z ∈ BR (0) with f (zn ) = 0. Hence the uniqueness theorem in Corollary 4.6.7 (with Ω = C) implies
that f ≡ 0 throughout C, which is a contradiction.
We now denote α1 , · · · , αN ∈ BR (0) be the zeros of f (it is possible that αi = α j for some i ̸= j).
By using Exercise 4.2.5, we see that the function
f (z)
g(z) :=
(z − α1 )(z − α2 ) · · · (z − αN )
is entire and also g(z) ̸= 0 for all z ∈ C. Hence we see that
1 (z − α1 )(z − α2 ) · · · (z − αN )
h(z) := =
g(z) f (z)
is also entire. Since | f (z)| → ∞ as z → ∞, then |h(z)| ≤ A + |z|N . By using the extended Liouville
1
theorem (Theorem 4.3.2), we see that h is a polynomial. But however h(z) = g(z) ̸= 0 for all z ∈ C,
then by fundamental theorem of algebra (Theorem 4.3.6), we conclude that h is a constant function,
says h(z) = k for some constant k ̸= 0. By the definition of h, we see that
1
f (z) = (z − α1 )(z − α2 ) · · · (z − αN ),
k
which conclude our theorem. □

4.7. Morera Theorems

The key result in our discussion of analytic functions so far has been the Cauchy closed curve
theorem (Theorem 3.3.11). In fact, the partial converse holds true as below (for future convenience,
we will refer all theorems in this section the “Morera theorems”):

T HEOREM 4.7.1 (Morera). Let f be a continuous function in an open set Ω. If

Z
f (z) dz = 0
Γ
for all Γ the boundary of topological closed rectangle in Ω, each segment is either horizontal (i.e.
parallel to real axis) or vertical (i.e. parallel to imaginary axis), then f is analytic in Ω.

R EMARK 4.7.2. In view of the Cauchy integral formula (Theorem 4.5.1), one sees that the
continuity of f is a necessary hypothesis.

E XERCISE 4.7.3. Prove Theorem 4.7.1 by modifying the arguments in the fundamental theorem
of antiderivative in rectangle (Theorem 3.2.9).

Morera’s theorem is often used to establish the analyticity of functions given in integral form.
 4.7. MORERA THEOREMS 54

E XERCISE 4.7.4. Using Morera’s theorem and Fubini’s theorem (carefully check the sufficient
R ezt
conditions for Fubini Theorem!) to show that the function f (z) = 0∞ t+1 dt is analytic in the left
n o
half plane z ∈ C Re (z) < 0 .

T HEOREM 4.7.5 (Morera’s uniform convergence theorem). Suppose { fn } represents a

sequence of analytic functions on an open set Ω satisfies

lim ∥ fn − f ∥L∞ (K) = 0 for all compact set K ⊂ Ω,

n→∞

then f is analytic in Ω.

P ROOF. Given any z ∈ Ω, there exists r > 0 such that Br (z) ⊂ Ω. We choose the compact set
K = B 2r (z). Hence we have
lim ∥ fn − f ∥L∞ (B r (z)) = 0.
n→∞ 2
This shows that f is continuous on K. Furthermore, for each Γ the boundary of any topological
closed rectangle in K, the uniform convergence of fn to f (on Γ) guarantees that
Z Z
f = lim fn = 0,
Γ n→∞ Γ

where the second identity is just simplyby the Cauchy closed curve theorem (Theorem 3.3.11). By
Morera’s theorem, we conclude that f is analytic in B 2r (z). By arbitrariness of z ∈ Ω, we conclude
the theorem. □
iθ
n E XERCISE 4.7.6. o Show that g(z) = z0 + e z with θ = arg(z1 − z0 ), maps the real axis
z ∈ C Im z = 0 onto the line L through z0 and z1 . Here arg w is defined (modulo 2π) as
that number θ for which
Im w Re w
sin θ = , cos θ = .
|w| |w|
Clearly, g defines an entire function.

T HEOREM 4.7.7 (Morera’s continuity theorem). Let Ω be an open set and let L be a straight
line in C. If f is continuous in Ω and analytic in Ω \ L, then f is analytic in Ω.

P ROOF. By using Exercise 4.7.6, it is suffice to show the theorem when L is the real axis. Let
z0 ∈ L ∩ Ω, and let r > 0 be such that Br (z0 ) ⊂ Ω. Let Γ the boundary of any topological closed
rectangle in Br (z0 ) which are parallel to the real and imaginary axes.
Case 1: L does not meet the topological closed rectangle enclosed by Γ. In this case, f is
R
analytic near the topological closed rectangle and thus Γ f = 0 by Cauchy closed curve theorem
(Theorem 3.3.11).
Case 2: the bottom side of Γ touches L. Let ε > 0 sufficiently small and let Γε be the rectangle
composed of the sides of Γ with bottom side shifted up by ε. By the continuity of f , we see that
Z Z
f = lim f = 0,
Γ ε→0 Γε
 4.7. MORERA THEOREMS 55

where the second identity follows by the Cauchy closed curve theorem (Theorem 3.3.11).
Case 3: the top side of Γ coincides with L. We can treat this case similar as previous case.
Case 4: The line L pass through the interior of the rectangle enclosed by Γ. In this case, we
can divide the rectangle into two rectangle by L. Let Γ1 and Γ2 are boundary of these two rectangles.
R R R R R
By using Case 2 and Case 3, we see that Γ1 f = 0 and Γ2 f = 0, and hence Γ f = Γ1 f + Γ2 f = 0.
Putting these 4 cases together, we conclude that f is analytic in Br (z0 ). By arbitrariness of
z0 ∈ L, we conclude our theorem. □
 CHAPTER 5

Laurent series and the Cauchy residual theorem

5.1. Riemann’s principle of removable singularities

In Remark 3.2.5, we posting the question about what we get if we integral over a simple
closed curve which surrounding some singularity. We have encounter some singularities in the
Cauchy integral formula (Theorem 4.5.1). Before studying the singularities, let us first classify the
singularities. Then we can at least partially answer this question for some class of singularities (so
that make this course easier).

D EFINITION 5.1.1. We call the set BR (z0 ) \ {z0 } the punctured ball centered at z0 with radius
R (or called the deleted neighborhood). A function f is said to have an isolated singularity at z0 if
f is analytic in a punctured ball centered at z0 and f is not complex differentiable (in the sense of
Definition 2.1.1) at z0 .

R EMARK 5.1.2. By using Theorem 4.7.7, we see that z0 is an isolated singularity if and only if
f discontinuous at z0 .

D EFINITION 5.1.3. Suppose f has an isolated singularity at z0 .

(1) If there exists a function g, analytic near z0 , such that f (z) = g(z) in a punctured ball
centered at z0 , we say that f has a removable singularity at z0 .
(2) If there exist functions A and B, both analytic near z0 with A(z0 ) ̸= 0 and B(z0 ) = 0, such
A(z)
that f (z) = B(z) in a punctured ball centered at z0 , then we say that f has a pole at z0 .
(3) If f has neither a removable singularity nor a pole at z0 , we say f has an essential
singularity at z0 .

In next section, we will fully characterize (necessary and sufficient condition) in next section
(Theorem 5.2.6) in terms of Laurent series. In plain words, removable singularity is the one we
can basically ignored, while essential singularity is the one that too difficult to handle within this
chapter. The pole is the one we want to discuss in this chapter. In this section, we first study some
sufficient conditions.

L EMMA 5.1.4 (Riemann’s principle of removable singularities). If f is analytic in a punctured

ball centered at z0 and that limz→z0 (z − z0 ) f (z) = 0, then f has at most a removable singularity at
z0 , i.e. there exists a function A, analytic near z0 , such that A = f in a punctured ball centered at
z0 .
56
 5.1. RIEMANN’S PRINCIPLE OF REMOVABLE SINGULARITIES 57

P ROOF OF L EMMA 5.1.4. If f is continuous at z0 , then by Theorem 4.7.7 we know that f is

analytic near z0 , and we have nothing to proof. If f is discontinuous at z0 , then z0 is an isolated
singularity of f . It is easy to see that the function

(z − z ) f (z) , z ̸= z ,
0 0
h(z) =
0 , z = z0 ,

is continuous at z0 . By using Theorem 4.7.7, we see that h is analytic near z0 . Since h(z0 ) = 0, then
h(z)
the function A(z) = z−z 0
is analytic near z0 (see Exercise 4.2.5). Since A = f in a punctured ball
centered at z0 , then we conclude our lemma. □
R EMARK 5.1.5. If f is analytic and bounded in a punctured ball centered at z0 , then clearly
limz→z0 (z − z0 ) f (z) = 0, and thus the above lemma follows that f has (at most) a removable
singularity at z0 .
R EMARK 5.1.6 (Riemann’s principle of removable singularities). If f is analytic in a punctured
ball centered at z0 and there exists k ∈ Z≥0 such that
analytic in a punctured
ball centered at z0
z }| {
k+1 k
(5.1.1) lim (z − z0 ) f (z) ≡ lim (z − z0 ) (z − z0 ) f (z) = 0,
z→z0 z→z0

by using the above lemma, we immediately see that there exists an analytic function A, analytic
near z0 , such that

(5.1.2) A(z) = (z − z0 )k f (z) in a punctured ball centered at z0 .

If k = 0, this implies that z0 is a removable singularity; if k > 0, this implies that z0 is a pole of f .
D EFINITION 5.1.7. Let f as in (5.1.2). If k = 0, then we called such z0 the pole of order 0 (can
be either removable singularity or f is analytic near z0 ). If k > 0 and A(z0 ) ̸= 0, then we say that
the pole z0 has order k.
R EMARK 5.1.8. By using a mathematical induction, one can easily see that (5.1.1) implies that
the pole has order at most k. Therefore one also can refer the removable singularity as the pole of
order 0. This remark generalizes Exercise 4.2.5.
E XAMPLE 5.1.9. Suppose that f has an isolated singularity at x0 = 0 (says) and there exists
C0
C0 > 0 such that it satisfies | f (z)| ≤ |z|α in a punctured ball centered at 0 for some α > 0 with

α∈ / Z. Let ⌈α⌉ be the smallest integer that ≥ α, and let ⌊α⌋ be the largest integer that ≤ α. One
sees that
lim sup |z⌈α⌉ f (z)| = lim sup |z|⌈α⌉ | f (z)| ≤ lim sup C0 |z|⌈α⌉−α = 0.
z→0 z→0 z→0
Then by Remark 5.1.6, one has

z⌊α⌋ f (z) = A(z) in a punctured ball centered at 0

 5.2. LAURENT EXPANSIONS 58

for some analytic function A. Hence it is not possible to find C1 > 0 and ⌊α⌋ < β ≤ α such that
C1
| f (z)| ≥ |z|β in a punctured ball centered at 0 (otherwise one can easily obtain a contradiction).

If f has an essential singularity at z0 , then one sees that

if lim (z − z0 )k+1 f (z) exists for some k ∈ Z≥0 , then lim (z − z0 )k+1 f (z) ̸= 0,
z→z0 z→z0

otherwise we can immediately obtain a contradiction from Remark 5.1.6. In this case, it is not
difficult see that limz→z0 | f (z)| = ∞. But, however, we do not know whether limz→z0 (z − z0 )k+1 f (z)
exists or not. We now closing this section by the following theorem.

T HEOREM
n 5.1.10. If f has an essential
o singularity at z0 , then for each R > 0 the set f (BR (z0 ) \
{z0 }) := f (z) z ∈ BR (z0 ) \ {z0 } is dense in C.

P ROOF. Suppose the contrary, that there exists a ball Bδ (w0 ) in C such that

Bδ (w0 ) ∩ f (BR (z0 ) \ {z0 }) = 0.

/

This means that | f (z) − w0 | ≥ δ for all z ∈ BR (z0 ) \ {z0 }, therefore

1 1
≤ for all z ∈ BR (z0 ) \ {z0 }.
f (z) − w0 δ
By using Remark 5.1.5, it follows that there exists a function A, which is analytic near z0 , such that
1 1
= A(z) ⇐⇒ f (z) = w0 +
f (z) − w0 A(z)
in a punctured ball centered at z0 . This implies that f has either a pole at z0 (if A(z0 ) = 0) or a
removable singularity at z0 (if A(z0 ) ̸= 0), which is a singularity. □

5.2. Laurent expansions

We now introduce a powerful tool to help us to study the isolated singularities.

D EFINITION 5.2.1. Let {µk }k∈Z be a sequence in C. We say that ∑k∈Z µk = L for some L ∈ C
−1
if both ∑∞ ∞
k=0 µk and ∑k=−∞ µk ≡ ∑k=1 µ−k converge and satisfies
∞ −1
∑ µk + ∑ µk = L.
k=0 k=−∞
We first show that the Laurent expansion make senses:

L EMMA 5.2.2. The Laurent expansion f (z) = ∑k∈Z ak zk is converge in the domain
n o
(5.2.1) AR1 ,R2 = z ∈ C R1 < |z| < R2

where
 −1
1 1
(5.2.2) R1 = lim sup |a−k | ,k R2 = lim sup |ak | k .
k→+∞ k→+∞
 5.2. LAURENT EXPANSIONS 59

If 0 ≤ R1 < R2 ≤ +∞, then f is analytic in the annulus AR1 ,R2 .

P ROOF. By using Theorem 2.2.2, one sees that

∞
f1 (z) = ∑ ak zk converges and it is an analytic function on BR2 .
k=0

If R2 = +∞, we interpret BR2 as the whole complex plane C. On the other hand, we also see that
∞  k −1
1 1 1 1
f2 (z) := ∑ a−k ≡ ∑ ak zk converges for those z ∈ C with = < .
k=1 z k=−∞ |z| z R1
In particular,
f2 converges and it is an analytic function on C \ BR1 .
Hence we conclude the theorem with f = f1 + f2 . □

The following theorem shows that the Laurent series will be a very powerful tool to study the
singularities.

T HEOREM 5.2.3. If f is analytic in the annulus AR1 ,R2 (5.2.1) with 0 ≤ R1 < R2 ≤ +∞, then f
has a Laurent expansion f (z) = ∑k∈Z ak zk in AR1 ,R2 .

P ROOF. Let C1 and C2 represent circles centered at 0 of radii r1 and r2 respectively, with
R1 < r1 < r2 < R2 , with counterclockwise orientation. We fix z ∈ Br2 \ Br1 and see that
f (w) − f (z)
g(w) =
w−z
is analytic at w ∈ AR1 ,R2 , and by Cauchy closed curve theorem (Theorem 3.3.11), we see that
Z
g(w) dw = 0,
C2 ∪C1rev

where C1rev is given by Lemma 3.1.8. One has to be careful that the annulus is not simply connected
(Example 3.3.2). However, this problem can be overcomed by splitting the annulus as showed in
the following diagram:
 5.2. LAURENT EXPANSIONS 60

F IGURE 5.2.1. Splitting the contour C2 ∪ C1rev into two closed curves

Combining the above two equations, we reach

f (w) 1
Z Z
dw = f (z) dw
C2 ∪C1rev w − z C2 ∪C1rev w − z
= 2πi =0
 
zZ }| { Z z }| {
 1 1 
= f (z)   C w−z dw − dw  = 2πi f (z) for all z ∈ Br \ Br ,
2 1
2 C1 w − z 

where the first term is due to Cauchy integral formula (Theorem 4.5.1) and the second term is
simply by the Cauchy closed curve theorem (Theorem 3.3.11). Hence we reach
f (w) f (w)
Z Z
2πi f (z) = dw − dw for all z ∈ Br2 \ Br1 .
C2 w − z C1 w − z

Since |w| > |z| for all w ∈ C2 , then recall the geometric sequence (see e.g. the proof of
Theorem 4.6.1)
1 1 1 z z2 ∞
zk
= = + + + · · · = ∑ k+1 for all w ∈ C2 ,
w − z w(1 − wz ) w w2 w3 k=0 w

which converges uniformly on C2 . Since |w| < |z| for all w ∈ C1 , similarly we have the geometric
sequence
1 −1 1 w w2 ∞
wk
= = − − 2 − 3 − · · · = − ∑ k+1 for all w ∈ C1 ,
w−z z−w z z z k=0 z
 5.2. LAURENT EXPANSIONS 61

which converges uniformly on C1 . Combining the above three equations, we reach

∞   ∞  
1 f (w) 1
Z Z
f (z) = ∑ k+1
k
dw z + ∑ f (w)w dw z−k−1
k
k=0 2πi C2 w k=0 2πi C1

= ak with k≥0 k = a with k<0

z }| { z }| {
∞  −1 
1 f (w) 1 f (w)
Z Z
= ∑ k+1
dw zk + ∑ k+1
dw zk
k=0 2πi C2 w k=−∞ 2πi C1 w

for all z ∈ Br2 \ Br1 . Since wf (w)

k+1 is analytic on the annulus Ω, by using Cauchy closed curve theorem

(Theorem 3.3.11) and the technique sketched by Figure 5.2.1, one sees that for each k ∈ Z that
1 f (w)
Z
(5.2.3) ak = dw
2πi C wk+1
for all counterclockwise circle C centered at 0, hence each ak is actually independent of r1 and r2 .
Hence we conclude our theorem. □
We now state and proof the following representation theorem.
n o
T HEOREM 5.2.4. If f is analytic in the annulus AR1 ,R2 (z0 ) = z ∈ C R1 < |z − z0 | < R2
with 0 ≤ R1 < R2 ≤ ∞, then f has a unique representation
1 f (z)
Z
k
(5.2.4) f (z) = ∑ ak (z − z0) , ak =
2πi CR (z0 ) (z − z0 )
k+1
dz
k∈Z

for any counterclockwise circle CR (z0 ) centered at z0 with radius R provided R1 < R < R2 .

P ROOF. It is easy to see that we only need to prove the proposition for z0 = 0. Since f (z) =
∑k∈Z ak zk converges in the annulus AR1 ,R2 , then it converges uniformly along C , and thus
f (z)
Z Z
(5.2.5) dz = ∑ ak zk−n−1 dz for any n ∈ Z.
C zn+1 k∈Z C

By using the Cauchy integral formula (Theorem 4.5.1), one has

Z
zm dz = 0 for all m ∈ Z≥0 .
C
By using the Cauchy integral formula (Theorem 4.5.1), we have
Z
z−1 dz = 2πi.
C
By using the fundamental theorem of line integral (Theorem 3.1.16), one also see that
Z
z−m dz = 0 for all m ∈ Z≥2 .
C
For future convenience, we record the above three equations as in below:

Z 2πi , m = −1,
(5.2.6) zm dz =
C 0 , m ∈ Z \ {−1}.
 5.2. LAURENT EXPANSIONS 62

Combining (5.2.5) and (5.2.6), we reach

f (z)
Z Z
n+1
dz = an z−1 dz = 2πian for all n ∈ Z,
C z C
which conclude our proposition. □

We now consider the case when z0 is an isolated singularity. If R1 = 0 and R2 < ∞, then
AR1 ,R2 = BR2 (z0 ) \ {z0 }, i.e. the punctured ball we consider in the previous section. Let f be an
analytic function on BR (z0 ) \ {z0 }. By Theorem 5.2.4, f has a unique Laurent series representation

(5.2.7) f (z) = ∑ ak (z − z0)k for all z ∈ BR (z0 ) \ {z0 }.

k∈Z

D EFINITION 5.2.5. We called ∑k≥0 ak (z − z0 )k the analytic part of f , while ∑k<0 ak (z − z0 )k

the principal part of f .

Since the analytic part of f does nothing with the singularity, we are now interested in the
principal part of f . From (5.2.7) we now able to give a full characterization for isolated singularities
in terms of Laurent series:

T HEOREM 5.2.6. Let f be an analytic function on a punctured ball centered at z0 . By

Theorem 5.2.4, f has a unique Laurent series representation (5.2.7). Then either one of the
following must holds:

(i) If f has a pole at z0 of order 0 (i.e. removable singularity or f is analytic near z0 ), then
C−k = 0 for all k ∈ N.
(ii) If f has a pole at z0 of order n ∈ N , thenC−n ̸=  0 and C−k = 0 for all k > n. In other
words, the principal part of f is simply P z−z0 for some polynomial P with degree n.
1

(iii) If f has an essential singularity at z0 , then C−k ̸= 0 for infinitely many k ∈ N.

R EMARK . Let g be an analytic function in an open set Ω. Suppose that z0 ∈ Ω is a zero of g,

then we consider its power series around z0 (Theorem 4.6.1):
∞
g(z) = ∑ Ck (z − z0)k .
k=0

From g(z0 ) = 0, one has C0 = 0. If g is nontrivial, then there exists k0 ∈ N such that Ck0 ̸= 0 and
Ck = 0 for all 0 ≤ k < k0 , and we write
!
∞ ∞
g(z) = ∑ Ck (z − z0 )k = (z − z0 )k0 ∑ Cℓ+k0 (z − z0)ℓ ,
k=k0 ℓ=0

which means that the zero z0 must have finite order. This also implies that each pole must have
finite order, therefore all isolated singularities are actually classified by Theorem 5.2.6.
 5.2. LAURENT EXPANSIONS 63

P ROOF OF ( I ). By definition, there exists a function A, analytic near z0 , such that f (z) = A(z)
in a punctured ball centered at z0 . Then by Theorem 5.2.4, the Laurent series of f must equal to
the power series of A. □
P ROOF OF ( II ). By definition, one writes
A(z)
f (z) = in a punctured ball centered at z0 ,
(z − z0 )n
where A is analytic near z0 . Using the local power series representation (Theorem 4.6.1), we write
A(z) = ∑∞ k
k=0 ak (z − z0 ) and we see that
∞ ∞
f (z) = ∑ ak (z − z0)k−n = ∑ an+ j (z − z0 ) j
k=0 j=−n

in a punctured ball centered at z0 . Finally, by Theorem 5.2.4, the above equation representations
the unique Laurent series of f , which conclude our theorem. □
P ROOF OF ( III ). Suppose the contrary, there exists n ∈ N such that C−k = 0 for all k > n.
Riemann’s principle of removable singularities (Remark 5.1.6) shows that z0 is pole, which is a
contradiction. □
Finally, we closed this section by exhibit an application of the representation formula of Laurent
series – together with Liouville theorem and fundamental theorem of algebra – in abstract algebra
(field theory).

T HEOREM 5.2.7 (Partial fraction decomposition of rational functions). Any proper rational
P(z)
function Q(z) , where P and Q are polynomials with deg P < deg Q, can be expanded as a sum of
1
polynomials in z−zk , where {z1 , z2 , · · · , zn } are the set of distinct zeros of Q.

S KETCH OF PROOF. By using fundamental theorem of algebra (Theorem 4.3.6), we can write
P(z)
Q(z) = A(z − z1 )k1 (z − z2 )k2 · · · (z − zn )kn for some n ≤ deg Q. This shows that Q(z) has a pole of
order at most k j at z j .
 
P(z)
(1) Using Theorem 5.2.6, the principal part of A0 (z) := Q(z) near z1 takes the form P1 z−z 1
1
 
polynomial P1 . Clearly, P1 z−z1 is analytic in C \ {z1 }. We now define A1 (z) :=
1
 
P(z)
Q(z) − P 1
1 z−z1 .
 
(2) Using Theorem 5.2.6, the principal part of A1 (z) near z2 , takes the form P2 z−z2 1
 
polynomial P2 . Clearly, P2 z−z 1
2
is analytic in C \ {z2 }. We now define A2 (z) :=
   
P(z)
Q(z) − P1 z−z1 − P2 z−z2 .
1 1

By repeting the above steps (can be rigorously written down using mathematical induction), one
sees that    
P(z) 1 1
− P1 − · · · − Pn
Q(z) z − z1 z − zn
 5.3. WINDING NUMBERS AND THE CAUCHY RESIDUE THEOREM 64

is an entire function. Since deg P < deg Q, by taking |z| → ∞, we see that actually above entire
function is bounded. Therefore the Liouville theorem (Theorem 4.3.1) implies that there exists a
constant C ∈ C such that
   
P(z) 1 1
− P1 − · · · − Pn ≡ C for all z ∈ C,
Q(z) z − z1 z − zn
which conclude our theorem1. □

5.3. Winding numbers and the Cauchy residue theorem

Let f be an analytic function on a punctured ball centered at z0 . By using Theorem 5.2.4, one
can write
1 f (z)
Z
k
(5.3.1) f (z) = ∑ ak (z − z0 ) , ak = dz
k∈Z 2πi C (z − z0 )k+1
for any counterclockwise circle C centered at z0 (within the analyticity region of f ). From (5.3.1),
we reach Z
f = 2πia−1 .
C
This suggests the coefficient a−1 is of special significance in this context.

D EFINITION 5.3.1. The coefficient a−1 is called the residue of f at z0 , and we denote
Res ( f ; z0 ) := a−1 .

P ROPOSITION 5.3.2 (Evaluation of residues via complex differentiation). If f has a pole of

order at most k ∈ N at z0 , then
1  
Res ( f ; z0 ) = ∂zk−1 (z − z0 )k f (z) .
(k − 1)! z→z0

R EMARK . Intuitively, we want to remove the pole of f by multiplying (z − z0 )k . The “price”

of doing so is some complex differentiations.

P ROOF. By Theorem 5.2.6, one can write

f (z) = a−k (z − z0 )−k + · · · + a−1 (z − z0 )−1 + a0 + a1 (z − z0 ) + · · · .

Then we see that

(z − z0 )k f (z) = a−k + · · · + a−1 (z − z0 )k−1 + a0 (z − z0 )k + a1 (z − z0 )k+1 + · · · ,

and hence  
∂zk−1 (z − z0 )k f (z) = (k − 1)!a−1 + a0 k!(z − z0 ) + · · · .
Evaluate z = z0 in the above equation, we conclude our proposition. □
1In fact, since deg P < deg Q, by taking |z| → ∞, we see that indeed C = 0.
 5.3. WINDING NUMBERS AND THE CAUCHY RESIDUE THEOREM 65

R EMARK 5.3.3. In most cases of higher-order poles, as with essential singularities, the most
convenient way to determine the residue is directly from the Laurent expansion.
R
To evaluate γ f when γ is a general closed curve (and when f may have isolated singularities),
we introduce the following concept.

D EFINITION 5.3.4. Suppose that γ is a parametrizable continuous piecewise-C1 closed curve

and that a ∈
/ γ. Then the number
1 1
Z
wind (γ, a) = dz
2πi γ z−a
is called the winding number of γ around a.

If γ = C be the counterclockwise circle C , then by Cauchy closed curve theorem

(Theorem 3.3.11) we see that

1 if a is inside the circle,
wind (γ, a) =
0 if a is outside the circle.

If γ circles the point a k-times via the parametrization γ = z0 + reiθ : 0 ≤ θ ≤ 2kπ , then
 

Z 2π
1
wind (γ, a) = i dθ = k,
2πi 0
which suggests the terminology “winding number”. We now need to prove this idea make senses
for general closed curve.
For each fixed parametrizable continuous piecewise-C1 closed curve γ, it is important to observe
that
the mapping a 7→ wind (γ, a), also can be denoted by wind (γ, ·),
is continuous as long as a ∈
/ γ.

P ROPOSITION 5.3.5. For any parametrizable continuous piecewise-C1 closed curve γ and a ∈ /
γ, the winding number wind (γ, a) is an integer. In addition, the mapping wind (γ, ·) is locally
constant (i.e. it is constant in the connected open components of C \ γ).
h i
P ROOF. Write γ = z(t) 0 ≤ t ≤ 1 , and set
Z s
ż(t)
F(s) = dt for 0 ≤ s ≤ 1,
0 z(t) − a
where ż denotes the differentiation of z with respect to t (see Definition 3.1.2). By fundamental
theorem of calculus on R, one sees that
ż(s)
Ḟ(s) = for all 0 < s < 1,
z(s) − a
 5.3. WINDING NUMBERS AND THE CAUCHY RESIDUE THEOREM 66

and thus (by the technique of integral factor, should be taughted in ODE course)
d  −F(s)

(z(s) − a)e = 0 for all 0 < s < 1.
ds
Since the open interval (0, 1) is connected, then

(z(s) − a)e−F(s) ≡ C for all 0 ≤ s ≤ 1

for some constant C ∈ C. Note: the equation also holds for endpoints s = 0 and s = 1, because F
and z are continuous on [0, 1]. Therefore, we have

(z(s) − a)e−F(s) = z(0) − a for all 0 ≤ s ≤ 1.

Since a ∈
/ γ, then z(0) − a ̸= 0, and then we have
z(s) − a
eF(s) = for all 0 ≤ s ≤ 1.
z(0) − a
Since γ is a closed curve, then z(1) = z(0), and then
z(1) − a
eF(1) = = 1.
z(0) − a
This implies that
F(1) = 2πik for some integer k ∈ Z,
1
and hence we conclude that wind (γ, a) = 2πi F(1) = k. □

Here we exhibit some graphical examples from Wikipedia:

 5.3. WINDING NUMBERS AND THE CAUCHY RESIDUE THEOREM 67

( A ) wind = −2 ( B ) wind = −1 ( C ) wind = 0

( D ) wind = 1 ( E ) wind = 2 ( F ) wind = 3

F IGURE 5.3.1. Winding numbers (By Jim.belk - Own work, Public Domain)

F IGURE 5.3.2. wind (γ, a) = 2 (By Jim.belk - Own work, Public Domain)

We finally able to prove the following theorem.

 5.3. WINDING NUMBERS AND THE CAUCHY RESIDUE THEOREM 68

T HEOREM 5.3.6 (Cauchy residue theorem). Suppose f is analytic in a simply connected open
set Ω except for isolated singularities at z1 , z2 , · · · , zm ∈ Ω. Let γ be a parametrizable continuous
piecewise-C1 closed curve in Ω, which not intersecting any of the singularities. Then
Z m
f = 2πi ∑ wind (γ, zk ) Res ( f ; zk ).
γ k=1

R EMARK (Cauchy closed curve theorem). For those f which is analytic in a simply connected
open set Ω, one has Res ( f ; z) = 0 for all z ∈ Ω, which can be easily see from Definition 5.3.1.
R
Therefore one has γ f = 0. Therefore the Cauchy closed curve theorem (Theorem 3.3.11) is a
special case of Cauchy residue theorem above.

g(z)
R EMARK (Cauchy integral formula). By considering f (z) = z−a with analytic function g and
a ∈ Ω, one has Res ( f ; a) = g(a), which can be easily see from Definition 5.3.1. If we choose C be
a parametrizable continuous piecewise-C1 closed curve in Ω, which not intersecting a and is simple
(i.e. wind (C , a) = 1), one sees that
g(z)
Z
dz = 2πi Res ( f ; a) = 2πi g(a).
C z−a
Therefore the Cauchy integral formula (Theorem 4.5.1) is a special case of Cauchy residue theorem
above.

P ROOF OF T HEOREM 5.3.6. Similar to Theorem 5.2.7, if we subtract the principal parts
   
1 1
P1 , · · · , Pm
z − z1 z − zm
from f , one sees that the difference
m  
1
g(z) = f (z) − ∑ Pk
k=1 z − zk
is analytic on D. Hence the Cauchy closed curve theorem (Theorem 3.3.11) implies that
m Z  
1
Z Z
(5.3.2) 0 = g = f − ∑ Pk .
γ γ k=1 γ z − zk
By the definition of principal part (Definition 5.2.5) and the definition of residual (Definition 5.3.1),
one sees that  
1 Res ( f , zk ) a−2 a−3
Pk = + 2
+ +··· ,
z − zk z − zk (z − zk ) (z − zk )3
and the above sequence converges uniformly on γ. By using the fundamental theorem of line
integral (Theorem 3.1.16), it is easy to see that
1
Z
dz = 0 for all k = 2, 3, 4, · · · ,
γ (z − zk )k
 5.4. SOME APPLICATIONS IN COMBINATORICS : EGORYCHEV METHOD 69

because γ is a closed curve. Hence we see that

 
1 1
Z Z
Pk = Res ( f , zk ) dz = 2πiwind (γ, zk ) Res ( f ; zk ).
γ z − zk γ z − zk

Plugging the above equation into (5.3.2), we conclude our theorem. □

5.4. Some applications in combinatorics : Egorychev method

The connection between binomial coefficients and contour integration is an immediate corollary
of the Residue theorem (Theorem 5.3.6). These techniques sometimes also referred as the
Egorychev method, which is a collection of techniques introduced by Georgy Egorychev for finding
identities among sums of binomial coefficients, Stirling numbers, Bernoulli numbers, Harmonic
numbers, Catalan numbers and other combinatorial numbers [Ego84].

T HEOREM 5.4.1 (First binomial coefficient integral). For each n ∈ N and k = 0, 1, · · · , n, one
has
!
n 1 (1 + z)n
Z
(5.4.1) = dz
k 2πi C zk+1

for all simple closed (parametrizable continuous piecewise-C1 ) curve C surrounding the origin.

P ROOF. For each k = 0, 1, · · · , n, by choosng

!
n
(1 + z)n n j−k−1
f (z) = k+1 = ∑ z
z j=0 j
in the Residue theorem (Theorem 5.3.6), one sees that
(1 + z)n
Z
dz = 2πi Res ( f ; 0) (Theorem 5.3.6)
C zk+1
!
n
= 2πi (Definition 5.3.1)
k
where we interpret n(n − 1) · · · (n − k + 1) = 1 when k = 0, which conclude the following theorem.
□
 5.4. SOME APPLICATIONS IN COMBINATORICS : EGORYCHEV METHOD 70

E XAMPLE 5.4.2. Let C be any simple closed (parametrizable continuous piecewise-C1 ) curve
surrounding the origin. By using (5.4.1), it is easy to see that
! !
n−1 n−1
+
k−1 k
1 (1 + z)n−1 1 (1 + z)n−1
Z Z
= dz + dz
2πi C zk 2πi C zk+1
1 (1 + z)n−1 z + (1 + z)n−1
Z
= dz
2πi C zk+1
!
1 (1 + z)n n
Z
= dz = ,
2πi C zk+1 k
which is the well-known Pascal triangle.

E XAMPLE 5.4.3 (A special case of Chu-Vandermonde identity). Let C be any simple closed
(parametrizable continuous piecewise-C1 ) curve surrounding the origin. By using binomial
theorem, one sees that
! ! ! !
n n n n
n n n n k−ℓ
(1 + z)n (1 + z−1 )n = ∑ zk · ∑ z−ℓ = ∑ ∑ z .
k=0 k ℓ=0 ℓ k=0 ℓ=0 k ℓ
(1+z)n (1+z−1 )n
By choosing f (z) = z , one sees that
(1 + z)n (1 + z−1 )n
Z
dz = 2πi Res ( f ; 0) (Theorem 5.3.6)
C z
!2
n
n
= 2πi ∑ (Definition 5.3.1).
k=0 k

On the other hand, we compute that

!2
n
n 1 (1 + z)n (1 + z−1 )n
Z
∑ k = 2πi C z
dz
k=0
1 (1 + z)n (z + 1)n
Z
= dz
2πi C zn+1
!
1 (1 + z)2n 2n
Z
= dz = ,
2πi C zn+1 n
where the last equality is given by (5.4.1).

E XAMPLE 5.4.4. We now want to prove the binomial identity:

! ! ! ! !
n
n n + k k n n + j
∑ (−1)k k k j
= (−1)n
j n
.
k=0
 5.4. SOME APPLICATIONS IN COMBINATORICS : EGORYCHEV METHOD 71

By using the first binomial coefficient integral (Theorem 5.4.1), one has
! !
n+k 1 (1 + z)n+k k 1 (1 + w)k
Z Z
= dz, = dw
k 2πi Cr zk+1 j 2πi Cs w j+1
for some r > 0 and s > 0, where Cρ is the circle with radius ρ centered at 0, which is
counterclockwise oriented. This yields
! ! !
n
n n + k k
∑ (−1)k k k j
k=0
binomial theorem
z ! }| {
n k
1 (1 + z)n 1 1 n (1 + z)(1 + w)
Z Z
= j+1 ∑ (−1)k dw dz
2πi Cr z 2πi Cs w k=0 k z
(1 + z)n 1 (1 + z)(1 + w) n
 
1 1
Z Z
= 1− dw dz
2πi Cr z 2πi Cs w j+1 z
1 (1 + z)n 1 1
Z Z
= n+1 j+1
(z − (1 + z)(1 + w))n dw dz
2πi Cr z 2πi Cs w
binomial theorem
(−1)n (1 + z)n 1 1 z
Z Z }| {
n
= (1 + w(1 + z)) dw dz
2πi Cr zn+1 2πi Cs w
j+1
!
(−1)n (1 + z)n 1 1 n n
Z Z
=
2πi Cr
∑ q wq(1 + z)q dw dz
zn+1 2πi Cs w j+1 q=0
!
n
(−1)n (1 + z)n 1 n
Z Z
(5.4.2) = n+1 ∑ wq− j−1 (1 + z)q dw dz.
2πi Cr z 2πi Cs q=0 q
We now define !
n
n
f (w) := ∑ wq− j−1 (1 + z)q .
q=0 q
!
n
By Definition 5.3.1, it is easy to see that Res ( f ; 0) = (1 + z) j , and thus by using the Residue
j
theorem (Theorem 5.3.6), one sees that
! !
n
1 n n
Z
∑ q wq− j−1(1 + z)q dw = j (1 + z) j .
2πi Cs q=0
 5.4. SOME APPLICATIONS IN COMBINATORICS : EGORYCHEV METHOD 72

Plugging the above equation into (5.4.2), we reach

! ! !
n
k n n+k k
∑ (−1) k k j
k=0
!
(−1)n (1 + z)n n
Z
= (1 + z) j dz
2πi Cr zn+1 j
!
n (−1)n (1 + z)n+ j
Z
= dz
j 2πi Cr zn+1
! !
n n+ j
= (−1)n ,
j n
where the last identity follows from the first binomial coefficient integral (Theorem 5.4.1).

T HEOREM 5.4.5 (Second binomial coefficient integral). For each n ∈ N and k = 0, 1, · · · , n, one
has
!
n 1 1
Z
(5.4.3) = dz
k 2πi Cρ (1 − z) zn−k+1
k+1

for all 0 < ρ < 1, where Cρ is the circle with radius ρ centered at 0, which is counterclockwise
oriented.

1
R EMARK . The reason we restrict 0 < ρ < 1 is to make sure that (1−z)k+1
is well-defined (as a
uniformly converge geometric sequence).

P ROOF. For each k = 0, 1, · · · , n, and let

1
f (z) = .
(1 − z)k+1 zn−k+1
 5.4. SOME APPLICATIONS IN COMBINATORICS : EGORYCHEV METHOD 73

Since f has pole of order n − k + 1 at z0 = 0, by using Proposition 5.3.2, one has

1 n−k

n−k+1

Res ( f ; 0) = ∂ z f (z)
(n − k)! z z→0
1  
= ∂zn−k (1 − z)−k−1
(n − k)! z→0
1 n−k−1

−k−2

= (k + 1) ∂z (1 − z)
(n − k)! z→0
1  
= (k + 1)(k + 2) ∂zn−k−2 (1 − z)−k−3
(n − k)! z→0
..
.
1
= (k + 1)(k + 2) · · · n
(n − k)!
!
n
= .
k
Therefore, by using the Residue theorem (Theorem 5.3.6), we immediately conclude (5.4.3). □
E XERCISE 5.4.6. Prove Theorem 5.4.1 by using Residue theorem (Theorem 5.3.6) and
evaluation the residues via complex differentiation (Proposition 5.3.2).
T HEOREM 5.4.7 (Exponential integral). For each n ∈ N and k = 0, 1, · · · , one has
k! enz
Z
k
n = dz
2πi C zk+1
for all simple closed (parametrizable continuous piecewise-C1 ) curve C surrounding the origin.
E XERCISE 5.4.8. Prove Theorem 5.4.7 by using the Residue theorem (Theorem 5.3.6) [Hint:
nz
Consider the function zek+1 ] .
T HEOREM 5.4.9. For each k ∈ Z and n ∈ Z, one has
1 1 1
Z
χ{(n,k)∈Z×Z : n≥k} (n, k) = dz
2πi Cρ zn−k+1 1−z
for all 0 < ρ < 1, where Cρ is the circle with radius ρ centered at 0, which is counterclockwise
oriented. Here χA is the indicator function defined by

1 , x ∈ A,
χA (x) =
0 , x ∈
/ A.
1
R EMARK . The reason we restrict 0 < ρ < 1 is to make sure that 1−z is well-defined (as a
uniformly converge geometric sequence).
R EMARK (Iverson bracket). In many cases, we simplyfied the notations by simply writing
{n ≥ k} = {(n, k) ∈ Z × Z : n ≥ k}. The Iverson bracket J·K, given by Jx ∈ AK := χA (x). One
 5.4. SOME APPLICATIONS IN COMBINATORICS : EGORYCHEV METHOD 74

note that the Kronecker delta can be expressed as δi j = J{i = j}K ≡ Ji = jK. By slightly abusing
notations, sometimes we write Theorem 5.4.9 as
1 1 1
Z
Jn ≥ kK ≡ χ{n≥k} = dz.
2πi Cρ zn−k+1 1−z

P ROOF OF T HEOREM 5.4.9. We consider the function

1 1
f (z) = n−k+1 .
z 1−z
When n + 1 ≤ k (iff n < k), then f (z) is analytic in B1 , so Res ( f ; 0) = 0. Otherwise when n + 1 > k
(iff n ≥ k), then f (z) has a pole of order n − k + 1 at z0 = 0, and hence by Proposition 5.3.2 we see
that
1 n−k

n−k+1

Res ( f ; 0) = ∂ z f (z)
(n − k)! z z→0
1
∂zn−k (1 − z)−1


=
(n − k)! z→0
1
∂zn−k−1 (1 − z)−2


=
(n − k)! z→0
1
2 ∂zn−k−2 (1 − z)−3


=
(n − k)! z→0
1
2 · 3 ∂zn−k−3 (1 − z)−4


=
(n − k)! z→0
..
.
1
= 2 · 3 · · · · · (n − k) = 1.
(n − k)!
Therefore, by using the Residue theorem (Theorem 5.3.6), we immediately conclude our theorem.
□
( )
n
The Stirling set number (also known as the Stirling number of second kind) is the number
k
of ways of partitioning a set of n elements into k nonempty sets, which is given by (https://
dlmf.nist.gov/26.8) ( ) !
n 1 k k
= ∑ (−1)k− j jn .
k k! j=0 j

T HEOREM 5.4.10. For each n ∈ N and k = 1, · · · , n, one has

( )
n n! 1 (ez − 1)k
Z
= dz
k k! 2πi C zn+1

for all simple closed (parametrizable continuous piecewise-C1 ) curve C surrounding the origin.
 5.4. SOME APPLICATIONS IN COMBINATORICS : EGORYCHEV METHOD 75

P ROOF. It is easy to see that the function

!
k
(ez − 1)k 1 k− j k jz
f (z) = n+1
= n+1 ∑ (−1) e
z z j=0 j
has a pole of order at most n + 1 at z0 = 0, and hence by Proposition 5.3.2 we see that
1
Res ( f ; 0) = ∂zn zn+1 f (z) z→0


n! !
1 k k
(−1)k− j ∂zn e jz z→0


= ∑
n! j=0 j
!
k
1 k n
= ∑ (−1)k− j j
n! j=0 j
( )
k! n
= .
n! k
Therefore, by using the Residue theorem (Theorem 5.3.6), we immediately conclude our theorem.
□
 CHAPTER 6

Some special analytic functions

6.1. The analytic function log z

In real analysis, the (natural) logarithmic function log x for x > 0 is defined by the inverse
function of the exponential function ex . The main difficulty to extend this to complex number is the
function ez is not injective.

D EFINITION 6.1.1. We say that f is an analytic branch of log z in a domain D if f is analytic

in D and e f (z) = z.

R EMARK 6.1.2. If f is an analytic branch of log z, then all other branches are g(z) = f (z)+2πki
for k ∈ Z.

For each x > 0, it is well-known that

d 1
log x = .
dx x
If we fix any x0 > 0, then the fundamental theorem of calculus implies
Z x
1
log x = dy + log x0 .
x0 y
This suggests us to define the complex logarithmic as in the following:

T HEOREM 6.1.3. Suppose that D is simply connected and that 0 ∈ / D (this condition is quite
natural since log 0 is not well-defined). Choose z0 ∈ D, fix a value of log z0 ∈ C such that elog z0 = z0
and set Z z
1
f (z) := dζ + log z0 .
z0 ζ
Then f is an analytic branch of log z in D, satisfying f ′ (z) = 1
z for all z ∈ D.
Rz 1 1
R EMARK 6.1.4. Here z0 ζ dζ means the integral along any paths from z0 to z. Since ζ is
Rz 1
analytic in D, by using the Cauchy residual theorem (Theorem 5.3.6), one sees that z0 ζ dζ is
indeed independent of the chosen path.

P ROOF OF T HEOREM 6.1.3. It is easy to see that f is analytic in D with f ′ (z) = 1z . The
remaining task is to show e f (z) = z. We define

g(z) = ze− f (z) .

76
 6.1. THE ANALYTIC FUNCTION log z 77

Since g′ (z) = e− f (z) − z f ′ (z)e− f (z) = 0 in D and D is simply connected, by using the fundamental
theorem of line integral (Theorem 3.1.16) one sees that g is a constant function and

g(z) = g(z0 ) = z0 e− log z0 = 1,

hence we conclude e f (z) = z. □

In a typical situation (unless stated), we choose D = C \ R≤0 and z0 = 1:
D EFINITION 6.1.5. The function Log z := 1z ζ1 dζ for all z ∈ C \ R≤0 , where R≤0 =
R

{x ∈ R : x ≤ 0}, which defined in the sense of Remark 6.1.4, is called the (standard) principal
branch of log z.
It is easy to see that
1
(Log z)′ = and − π < Im (Log z) < π.
z
One can use Remark 6.1.2 to construct all other branches

(6.1.1) Log z + 2πki for all k ∈ Z,

which also corresponding to D = C \ R≤0 and z0 = 1 as well. We also can define the logarithms to
other bases by
Log z
Logw z := .
Log w
Recall that exp(Log z) = z for all z ∈ C \ R≤0 , that is, Log is the right-inverse of exp (with respect
to the composition operator of functions).
Q UESTION 6.1.6. How about Log (exp w) for w ∈ C satisfies exp w ∈ C \ R≤0 ?
The above question can be easily answered by the following theorem gives an equivalent
definition of Log z:
T HEOREM 6.1.7 (Equivalent definition of principal branch of log z). For each z ∈ C \ R≤0 , one
can write z = Reiθ for some R > 0 and −π < θ < π. Then

Log z = log R + iθ ≡ log |z| + iθ .

R EMARK 6.1.8 (Left inverse of exponential). For each w ∈ C, one sees that

exp w = eRe w+iIm w = eRe w eiIm w = eRe w (cos(Im w) + i sin(Im w)).

If − π2 < Im w < π2 , then Re (exp w) = eRe w cos(Im w) > 0. Therefore, at least when − π2 < Im w <
π
2 , one can choose z = exp w in Theorem 6.1.7 to see that

Log (exp w) = log(eRe w ) + iIm w = Re w + iIm w = w.

This shows that the principal branch of complex logarithmic is the left inverse of complex
exponential in suitable domain, it is valid when − π2 < Im w < π2 , but not all w ∈ C. It is clearly
 6.1. THE ANALYTIC FUNCTION log z 78

that this is not true for other branch (6.1.1). This also explains why we only consider right inverse
in Definition 6.1.1, and we usually consider the principal branch of complex logarithmic (in many
literature, we always consider this principal branch unless stated).

P ROOF OF T HEOREM 6.1.7. We see that

Z z Z |z| Z z Z Reiθ
1 1 1 1
Log z = dζ = dζ + dζ = log R + dζ .
1 ζ 1 ζ |z| ζ R ζ
h i
We now choose the curve C = Reit 0 ≤ t ≤ θ , and by the definition of line integral we see
that Z Reiθ
1 1 1
Z Z θ Z θ
it
dζ = dζ = it
Rie dt = i 1 dt = iθ ,
R ζ C ζ 0 Re 0
which conclude the theorem. □

E XERCISE 6.1.9. Given any z1 , z2 ∈ C \ R≤0 , show that there exists n ∈ {−1, 0, 1} such that
Log (z1 z2 ) = Log z1 + Log z2 + i2πn.

E XAMPLE 6.1.10. We now can define the roots of complex number by using Log z. For
example,
√
 
1
the principal branch of z := exp Log z for all z ∈ C \ R≤0 .
2
√
Note that different branches of log z may yield different branches of z. Unlike log z, there are
√
only two different branches of z. This follows from the fact that the equation w2 = z has exactly
two different solutions for any z ̸= 0, which is a consequence of fundamental theorem of algebra
(Theorem 4.3.6).
√
E XERCISE 6.1.11. Find all the two branches of i.

E XAMPLE 6.1.12. The same technique may be used to define arbitrary powers of any nonzero
complex number. For example, the principal branch of ii is defined by exp(iLog i). By using
Theorem 6.1.7, one sees that
π iπ
Log i = log 1 + i = ,
2 2
iπ
then i = exp(i 2 ) = exp(− 2 ). It is interesting to note that ii is a real number.
i π

E XERCISE 6.1.13. Determine all the other branches of ii .

E XERCISE 6.1.14. Compute Log (1 + i).

E XERCISE 6.1.15. Show that

∞
zn
Log (1 + z) = − ∑ (−1)n for all z ∈ B1 .
n=1 n
We end this section by giving an example which has interesting branches which are different to
the principal branch.
 6.2. INFINITE PRODUCTS AND WEIERSTRASS PRODUCT THEOREM 79

E XAMPLE (Lambert W -function). The Lambert W -function W (z) is the complex-valued

solution of the equation
WeW = z.
On the z-interval [0, ∞) there is one real solution, and it is nonnegative and increasing. On the z-
interval (−e−1 , 0), there are two real solutions, one increasing and the other decreasing. We call the
increasing solution for which W (x) ≥ W (−e−1 ) = −1 the principal branch and denote it by W0 (x),
and the decreasing solution can be identified as W±1 (x ∓ i0), see Figure 6.1.1. Here x ∓ i0 means
the (formal) limit x ∓ iy as y → 0+ . Rather than elaborate all details here, we refer to DLMF:4.13
for more details about this function.

F IGURE 6.1.1. Branches W0 (x), W± (x ∓ i0) of the Lambert W -function (Credit:

https://dlmf.nist.gov/4.13.F1.mag)

6.2. Infinite products and Weierstrass product theorem

Similar to the infinite sum (power series), we also can consider the infinite product by using a
similar manner:

D EFINITION 6.2.1. Let {uk }∞ k=1 be a sequence of nonzero complex numbers. The infinite
product Πk∈N uk ≡ Πk=1 uk is said to converge to a nonzero limit if the sequence of partial products
∞

PN := ΠN
k=1 uk = u1 u2 · · · uN

converges to a nonzero limit (in C, in the sense of Definition 1.2.4) as N → ∞.

R EMARK 6.2.2. In this case, it is easy to see that PN = uN PN−1 . The infinite product converges
means PN → P for some P ∈ C \ {0}, and thus
PN limN→∞ PN P
lim uN = lim = = = 1.
N→∞ N→∞ PN−1 limN→∞ PN−1 P
 6.2. INFINITE PRODUCTS AND WEIERSTRASS PRODUCT THEOREM 80

Obviously, Π∞ ∞
k=1 uk converges to a nonzero limit if and only if Πk=N0 uk converges to a nonzero limit
for any fixed N0 ∈ N.

D EFINITION 6.2.3. If PN → 0, we say the infinite product diverges to zero. If there are finitely
many terms uk are equal to zero and Πk∈N,uk ̸=0 uk converges (in C), then we say the product
Πk∈N uk ≡ Π∞
k=1 uk converges to zero.

We now give an example to explain why we introduce the term “diverges to zero”.

E XAMPLE 6.2.4. Fix any N0 ∈ N, we see that the partial sum of the series ∏∞
k=N0 (1 − 1/k) is
given by
N   N
1 k−1
PN := ∏ 1 − = ∏
k=N0 k k=N0 k
N0 − 1  N0  N0+1 N
 −1
= · · · · · · ·
N0 
 N0+ 1 
 N0+ 2

 N
N0 − 1
= .
N
According to Definition 6.2.1, one has
∞   N  
1 1
(6.2.1) ∏ 1 − k := N→∞ lim ∏ 1 −
k
= 0.
k=N0 k=N0

However, we see that 1 − 1k → 1 as k → ∞. Fix a large N0 , and we formally see that

≈1 ≈1
z }| { 
z }| {
∞   
1 1 1 (?)
∏ 1 − k = 1 − N0 1 − N0 + 1 · · · ≈ 1 ̸= 0.
k=N0

Due to this inconsistency, therefore we call (6.2.1) that the series ∏∞

k=N0 (1 − 1/k) is diverges to
zero.
 
1
E XERCISE 6.2.5. Prove that ∏∞ k=2 1 − k2
converges to a nonzero limit.

E XERCISE 6.2.6. Let {ak }∞

k=1 be a sequence of positive real numbers. Show that
N
a1 + a2 + · · · + aN ≤ ∏ (1 + ak ) ≤ ea1 +a2 +···+aN for all N ∈ N.
k=1

By using this, show that ∏∞ ∞

k=1 (1 + ak ) converges to a nonzero limit if and only if ∑k=1 ak converges.

However, the following exercise demonstrates the necessity of the positivity of such {ak }∞
k=1 :
(−1) k
E XERCISE 6.2.7. Let ak := √ for all k = 2, 3, 4, · · · . Show that ∑∞
k=2 ak converges, but
k
∞
∏k=2 (1 + ak ) diverges to zero.

For general (complex) case, we still have the following result.

 6.2. INFINITE PRODUCTS AND WEIERSTRASS PRODUCT THEOREM 81

T HEOREM 6.2.8. Let 1 + zk ∈ C \ R≤0 for all k ∈ N.

(a) If ∑∞ ∞
k=1 Log (1 + zk ) converges, then ∏k=1 (1 + zk ) converges to a nonzero limit
exp (∑∞k=1 Log (1 + zk )).
(b) If ∏k=1 (1 + zk ) converges to a nonzero limit, then ∑∞
∞
k=1 Log (1 + zk ) converges.

R EMARK . The tricky part in (b) is when the limit of ∏∞ k=1 (1 + zk ) is in {z ∈ C : Re z ≤ 0},
therefore the limit cannot express in terms of the standard logarithmic branches (6.1.1). Therefore
∗ ∗
the limit of ∑∞ ∞
k=1 Log (1 + zk ) is actually log (∏k=1 (1 + zk )), where log is some branch of the
logarithm given in Theorem 6.1.3 with some suitable domain D, which may differ with the standard
choice C \ R≤0 .

R EMARK . Both results (a) and (b) can be extended for all zk ̸= −1 with different branches
log(k) (corresponding to different domains D(k) ) of complex logarithmic for each k.

P ROOF OF (a). Let SN = ∑N N SN

k=1 Log (1 + zk ) and PN = ∏k=1 (1 + zk ) = e . The condition
S
∑∞k=1 Log (1 + zk ) converges (to some S ∈ C) means SN → S, and hence PN → e ̸= 0, which
conclude (a). □
P ROOF OF (b). The condition ∏∞k=1 (1 + zk ) converges to some nonzero limit P ∈ C means
PN → P. As explained in the remark, one can find some branch of the complex logarithm log∗∗
(given in Theorem 6.1.3 with some suitable domain D) such that

log∗∗ PN → log∗∗ P in C as N → ∞.

By using Theorem 6.1.3 and (6.1.1), for each k ∈ N, one can find nk ∈ Z such that
N
∑ (Log (1 + zk ) + 2πink ) = log∗∗ PN ,
k=1
and thus
N
∑ (Log (1 + zk ) + 2πink ) → log∗∗ P as N → ∞.
k=1
It is easy to verify that (this can be showed by, e.g. a contradiction argument)

Log (1 + zk ) + 2πink → 0 as k → ∞.

By using Remark 6.2.2, we have zk → 0, and thus the above limit implies nk → 0 as k → ∞. Since
nk ∈ Z, thus nk = 0 for all k ≥ N0 for some N0 ∈ N. Then one sees that
∞ N0 ∞
∑ Log (1 + zk ) + 2πi ∑ nk = ∑ (Log (1 + zk ) + 2πink ) → log∗∗ P as N → ∞,
k=1 k=1 k=1

which proves (b) with the branch log∗ = log∗∗ −2πi ∑N0
k=1 nk . □

C OROLLARY 6.2.9. If ∑∞ ∞ ∞
k=1 zk converges absolutely, that is, ∑k=1 |zk | < ∞, then ∏k=1 (1 + zk )
converges.
 6.2. INFINITE PRODUCTS AND WEIERSTRASS PRODUCT THEOREM 82

1
P ROOF. Since ∑∞ k=1 |zk | < ∞, then one can find N0 ∈ N such that |zk | < 2 for all k ≥ N0 . Hence
by Exercise 6.1.15, one has
z2k z3k
 
1 1
|Log (1 + zk )| = zk − + − + · · · ≤ |zk | 1 + + + · · · ≤ 2|zk | for all k ≥ N0 .
2 3 2 4
Hence ∞ ∞
∑ |Log (1 + zk )| ≤ 2 ∑ |zk | < ∞
k=N0 k=N0
and our result follows from Theorem 6.2.8(a). □

D EFINITION 6.2.10. We say that the product Π∞ ∞

k=1 (1 + zk ) is absolutly convergent if Πk=1 (1 +
|zk |) < ∞.
∞
L EMMA 6.2.11. If Π∞
k=1 (1 + zk ) is absolutly convergent, then ∏k=1 (1 + zk ) converges.

∞
P ROOF. Since Π∞k=1 (1 + |zk |) < ∞, by Exercise 6.2.6, we have ∑k=1 |zk | < ∞,. Hence we
conclude our lemma by Corollary 6.2.9. □

We wish to consider analytic functions defined by infinite products, i.e. functions of the form
∞
(6.2.2) f (z) = ∏ (1 + uk (z)).
k=1
By using Morera’s uniform convergence theorem (Theorem 4.7.5), if each uk are analytic on an
open set D and the partial products converges to their limit function uniformly on each compact set
K in D, then one sees that f is analytic on D.

E XERCISE 6.2.12. Let K be a compact set in C, and we consider a continuous function g : K →

C. Show that the set g(K) := {g(z) : z ∈ K} is compact in C.

Based on this observation, one can prove the following theorem.

T HEOREM 6.2.13. Suppose that for each k = 1, 2, · · · that uk is analytic in an open set D, and
that ∑∞k=1 |uk (z)| converges uniformly on all compact set in D. Then the product (6.2.2) converges
uniformly on on all compact set in D, and it defines an analytic function in D.

R EMARK . The uniform convergence of ∑∞

k=1 uk (z) does not imply the uniform convergence of
∞
∑k=1 |uk (z)|.

P ROOF. Let K be any compact set in D. Since ∑∞ k=1 |uk (z)| converges uniformly on K, then
1
there exists N0 ∈ N such that ∥uk ∥L∞ (K) ≤ 2 , hence 1 + uk ̸= 0 for all k ≥ N0 . Given any ε > 0, one
can choose integer N1 ≥ N0 such that
∞
∑ |uk (z)| ≤ ε.
k=N1
 6.2. INFINITE PRODUCTS AND WEIERSTRASS PRODUCT THEOREM 83

Hence by Exercise 6.1.15, one has

(uk (z))2 (uk (z))3
|Log (1 + uk (z))| = uk (z) − + −+···
2 3
 
1 1
≤ |uk (z)| 1 + + + · · · ≤ 2|uk (z)| for all k ≥ N1 ,
2 4
and thus ∞ ∞
∑ |Log (1 + uk (z))| ≤ 2 ∑ |uk (z)| ≤ 2ε for all z ∈ K.
k=N1 k=N1
Hence we know that ∑∞
k=1 Log (1 + uk (z)) converges uniformly on K to a limit function g(z). Since
g is continuous, by Exercise 6.2.12 it follows that g(K) := {g(z) : z ∈ K} is bounded. Finally, since
the exponential function is uniformly continuous in any bounded domain, then
!
N
PN (z) := exp ∑ Log (1 + uk (z))
k=1

converges uniformly to its limit function f (z) = eg(z) . Hence we conclude our theorem by the above
observation involving Morera’s uniform convergence theorem (Theorem 4.7.5). □

E XERCISE 6.2.14. Show that ∏∞ k

k=1 (1 + z ) converges uniformly on any compact subset of B1
(therefore it defines an analytic function on B1 ).

1
E XERCISE 6.2.15. Show that ∏∞ k=1 (1 + kz ) converges uniformly on any compact subset of the
half-space {z ∈ C : Re (z) > 1} (therefore it defines an analytic function on {z ∈ C : Re (z) > 1}).

T HEOREM 6.2.16 (Weierstrass product theorem). Suppose {λk }k∈N ⊂ C which |λk | → ∞ as
k → ∞. Then there exists an entire function f such that

f (λk ) = 0 for all k ∈ N f (z) ̸= 0 for all z ∈

/ {λk }k∈N .

(see (6.2.4) for the precise formula for such f )

R EMARK . According to the uniqueness theorem (Corollary 4.6.7), a nontrivial entire function
cannot have an accumulation point of zeros. This means that, if f (λk ) = 0 for those {λk }k∈N ⊂ C
converges in C, then f ≡ 0 in the whole complex plane C. Therefore the assumption |λk | → ∞
as k → ∞ seems necessary. It would seem natural to write f (z) = ∏∞ k=1 (z − λk ). However, since
|λk | → ∞, the terms of the product would not approach 1, even pointwisely, for each z ∈ C. The
product would diverge.

R EMARK . An entire function may be zero at all the points of a sequence which diverges to ∞,
see Example 4.6.8 for sin z. Weierstrass product theorem (Theorem 6.2.16) shows that this example
is in no way exceptional.
 6.2. INFINITE PRODUCTS AND WEIERSTRASS PRODUCT THEOREM 84

P ROOF OF T HEOREM 6.2.16. We first consider the case when λk ̸= 0 for all k = 2, 3, · · · , and
set !
z z2 zk
Ek (z) := exp + +···+ k .
λk 2λk2 kλk
Given any M > 0, and let |z| < M. Since |λk | → ∞, one can find N0 ∈ N such that |λk | ≥ 2M for all
k ≥ N0 . By using Exercise 6.1.15, we see that
    
z z
Log 1− Ek (z) = Log 1 − + Log (Ek (z))
λk λk
z2 zk
 
z z
= Log 1 − + + 2 +···+ k
λk λk 2λk kλk
z j
∞  
1
=− ∑ −
j=k+1 j λj
1
which is valid since | λz | ≤ 2 for all k ≥ N0 . Hence
k

zj
   ∞
z
Log 1− Ek (z) ≤ ∑ j
λk j=k+1 jλk
k k ∞ ℓ
z ∞
z j−k z 1 z
= ∑ j−k
= ∑ ℓ + k λk
λk j=k+1 jλk λk ℓ=1
k ∞ k
z 1 z 1
≤ ∑ 2ℓ = λ k ≤ .
λk ℓ=1 2k
This shows that the sum ∞   
z
∑ Log 1 − λk Ek (z)
k=N0
is uniform converges in all compact set in BM . By taking the exponential in each partial sum, one
also can verify that the product
∞  
z
(6.2.3) g(z) := ∏ 1 − Ek (z)
k=2 λk
is also uniform converges in all compact set in BM . By arbitrariness of M, in fact (6.2.3) defines an
entire function, satisfying

g(λk ) = 0 for all k ∈ N / {λk }∞

g(z) ̸= 0 for all z ∈ k=2 .

Finally, if we seek an entire function with zeros λ1 = 0 at the origin as well, we only need to set
∞  
p p z
(6.2.4) f (z) = z g(z) = z ∏ 1 − Ek (z)
k=2 λk
so that λ1 = 0 is the zero of f with multiplicity p. □
 6.2. INFINITE PRODUCTS AND WEIERSTRASS PRODUCT THEOREM 85

E XAMPLE 6.2.17. By using (6.2.4), it is easy to see that an entire function with zeros at all the
points λk = log k for all k ∈ N is given by
∞ 
z2 zk
  
z z
f (z) = z ∏ 1 − exp + 2
+···+ k
.
k=2 log k log k 2(log k) k(log k)

E XERCISE 6.2.18. Show that

∞
 z −z
f (z) := ∏ 1 + e k
k=1 k
is an entire function with a single zero at every negative integer λk = −k. In fact, this function is
related to Gamma function (will be introduced later in Section 6.3) by the formula
e−γz
f (z) =
Γ(z)z
where γ is the Euler constant, see (6.3.5). [Hint: Modifying the ideas in the proof of
Theorem 6.2.16.]

E XAMPLE 6.2.19. By using Exercise 6.2.18, it is easy to see that

!  !
∞ 
z −z ∞ 
z z ∞ 
z 2
(6.2.5) f (z) = z ∏ 1 +
k
e k ∏ 1 − j e j = z ∏ 1 − k2
k=1 j=1 k=1
 
z2
is an entire function with a single zero at every integer, since the partial ∏N
k=1 1 − k2 is the
product of the partial sum
M1  M2  
z −z z z
∏ 1 + k e and ∏ 1 − j e j
k

k=1 j=1

with the  case N = M1 = M2 . As an exercise, here we give a direct justification of

 special
∞ z2
z ∏k=1 1 − k2 without refering Exercise 6.2.18 but only modifying the proof of Theorem 6.2.16:
Given any M > 1, and let |z| < M. By using Exercise 6.1.15, we see that
2  j
z2 (− kz2 ) j 1 z2
  ∞ ∞ 3
Log 1 − 2 = − ∑ (−1) j
=−∑ 2
for all k ≥ ⌈2 2 M 3 ⌉.
k j=1 j j=1 j k

Hence we see that

j
z2 z2
  ∞ ∞ ∞
1 1 1 1
Log 1 − 2 ≤ ∑ ≤ ∑ 4 ≤ 4 ∑ 2j = 4 for all |z| < M.
k j=1 k2 j
j=1 2 k 3 j
k 3 j=1 k3
This shows that the sum
z2
∞  
∑3 Log 1 − 2
k
k=⌈2 2 M 3 ⌉
 6.3. THE GAMMA FUNCTION: AN EXTENSION OF FACTORIAL FUNCTION 86

is uniform converges in all compact set in BM . By taking the exponential in each partial sum, one
also can verify that the product
∞ 
z2

(6.2.6) g(z) := ∏ 1 − 2
k=1 k
is also uniform converges in all compact set in BM . By arbitrariness of M, in fact such g defines an
entire function, satisfying

g(k) = 0 for all k ∈ Z \ {0} g(z) ̸= 0 for all z ∈

/ Z \ {0}.

Finally, we conclude (6.2.5) is our desired analytic function since f (z) = zg(z).
We have the following fact:
T HEOREM 6.2.20. For each z ∈ C, we have
sin πz
= f (z),
π
where f is the function given in (6.2.5).
We shall skip the proof of the above theorem, since it is too technical. Here we refer to [BN10,
Proposition 17.8] for a proof. As an immediate consequence, we have:
C OROLLARY 6.2.21. All zeros of sin z are real (in other words, there is no zeros other than in
Example 4.6.8).
Moreover, we also have the following representation for complex cosine (here we state without
proof, see [BN10, Exercise 9 in Chapter 17]).
T HEOREM 6.2.22. For each z ∈ C, we have
∞ 
4z2

cos πz = ∏ 1 − .
k=0 (2k + 1)2

6.3. The Gamma function: an extension of factorial function

We begin this section by the following lemma.

L EMMA 6.3.1. Let D be an open set in C and let ∞ ≤ a < b ≤ +∞. Suppose that for a.e.
t ∈ (a, b) the mapping ϕ(·,t) is an analytic function on D. If
Z b
(6.3.1) |ϕ(z,t)| dt < +∞ for all z ∈ D or ϕ(z,t) ≥ 0 for a.e. (z,t) ∈ D × (a, b),
a
then Z b
f (z) = ϕ(z,t) dt
a
is analytic in D with complex derivative
Z b
(6.3.2) f ′ (z) = ∂z ϕ(z,t) dt.
a
 6.3. THE GAMMA FUNCTION: AN EXTENSION OF FACTORIAL FUNCTION 87

P ROOF. Let Γ the boundary of topological closed rectangle in D, each segment is either
horizontal (i.e. parallel to real axis) or vertical (i.e. parallel to imaginary axis). The assumption
(6.3.1) allows us to use Fubini’s theorem (for Lebesgue integral) to see that
Z Z Z b Z b Z 
f (z) dz = ϕ(z,t) dt dz = ϕ(z,t) dz dt.
Γ Γ a a Γ
R
Since ϕ is analytic in z, by Cauchy’s residual theorem (Theorem 5.3.6) one sees that Γ ϕ(z,t) dz =
R
0, and thus Γ f (z) dz = 0. By arbitrariness of Γ ⊂ D, we conclude f is analytic on D by Morera’s
theorem (Theorem 4.7.1). Since f is analytic, then f ′ (z) = ∂x f , whenever z = x + iy. Therefore
(6.3.2) immediately follows from the Leibniz integral rule (this step only requires the continuity of
∂x ϕ), here we omit the details. □

We consider the integral

Z ∞
In = e−t t n dt for n = 0, 1, 2, · · · ,
0
which can be interpret as improper Riemann integral. In this case, this is same as the Lebesgue
integral.

E XERCISE 6.3.2. By interpreting the above as improper Riemann integral, show that I0 = 1 and
In = nIn−1 for all n ∈ N. From this, one sees that In = n! = n(n − 1)(n − 2) · · · · · 2 · 1.

For any z ∈ C and t > 0, we define t z−1 := e(z−1) logt . One sees that |t z−1 | = |e(z−1) logt | =
e(Re (z−1)) logt = t Re (z−1) for all t > 0. Hence one sees that the gamma function
Z ∞
Γ(z) = e−t t z−1 dt
0
is uniformly convergent in the right half-plane {z ∈ C : Re z > 0}. Hence by Lemma 6.3.1, one
sees that Γ is analytic in the right half-plane {z ∈ C : Re z > 0} with complex derivatives or order
k: Z ∞
(k)
Γ (z) = t z−1 (logt)k e−t dt for all z ∈ C with Re z > 0.
0
Using the same arguments in Exercise 6.3.2, it is easy to show that

Γ(z + 1) = zΓ(z) for all z ∈ C with Re z > 0.

We can extend Γ for −1 < Re z < 0 by the formula

Γ(z + 1)
Γ(z) := for all z ∈ C with − 1 < Re z < 0.
z
It is easy to see that Γ is continuous at each z ∈ C\{0} with Re z > −1, and hence by using Morera’s
continuity theorem (Theorem 4.7.7), Γis also analytic there. Continuing in the same manner, we
 6.3. THE GAMMA FUNCTION: AN EXTENSION OF FACTORIAL FUNCTION 88

can define
Γ(z + 2)
Γ(z) := for all z ∈ C with − 2 < Re z < −1,
z(z + 1)
Γ(z + 3)
Γ(z) := for all z ∈ C with − 3 < Re z < −2,
z(z + 1)(z + 2)
Γ(z + k + 1)
(6.3.3) Γ(z) := for all z ∈ C with − k − 1 < Re z < −k,
z(z + 1) · · · (z + k)
and applying Morera’s continuity theorem (Theorem 4.7.7), we see that:

T HEOREM 6.3.3. Γ defines an analytic function on C \ {0, −1, −2, · · · }, with Res (Γ; −k) =
(−1)k
limz→−k (z + k)Γ(z) = k! .

P ROOF. By using Proposition 5.3.2, one can easily compute

Γ(1) (−1)k
Res (Γ; −k) = lim (z + k)Γ(z) = = ,
z→−k (−k)(−k + 1) · · · (−1) k!
which concludes our result. This also means that {0, −1, −2, · · · } are all poles or order 1. □

From now on, we will only sketch the ideas (since this part is quite technical), see [BN10,
Chapter 18] for more details. By using the fact that limn→∞ (1 − nt )n = e−t , one can show
Z n
z−1
 t n
Γ(z) = lim t 1− dt
n→∞ 0 n
1 n z−1
Z
= lim n t (n − t)n dt whenever Re z > 0,
n→∞ n 0
see [BN10, Exercise 7 in Chapter 18]. By using integration by parts, we have
1 n n z
Z
Γ(z) = lim n · t (n − t)n−1 dt
n→∞ n z 0
Z n
1 n(n − 1) · · · 1
= lim n t z+n−1 dt
n→∞ n z(z + 1) · · · (z + n − 1) 0
nz 1 2 n
= lim · · ····· .
n→∞ z z + 1 z + 2 z+n
Thus we reach the Gauss’ product representation for Gamma function:
1  z  z
= lim zn−z (1 + z) 1 + ··· 1+
Γ(z) n→∞ 2 n
n 
z
(6.3.4) = lim zn−z ∏ 1 + ,
n→∞
k=1 k
see also [FB09, Proposition IV.1.10].

R EMARK . This immediately shows that Γ has no zeros.

Here we exhibit a real analysis result in [BN10, Lemma 18.8]:

 6.3. THE GAMMA FUNCTION: AN EXTENSION OF FACTORIAL FUNCTION 89

L EMMA 6.3.4. If sn = 1 + 21 + · · · + 1n − log n, then limn→∞ sn exists. This limit is called the
Euler constant, usually denoted as γ, which is approximately 0.577 · · · .

We write (6.3.4) as
!
n
1 z(1+ 21 +···+ 1n −log n)
 z −z
= lim e z∏ 1+ e k for all z ∈ C.
Γ(z) n→∞ k=1 k
By using Theorem 6.2.20 and Lemma 6.3.4, we have [FB09, Lemma IV.1.9]
∞ 
1 γz z −z
(6.3.5) = ze ∏ 1 + e k for all z ∈ C,
Γ(z) k=1 k
see Exercise 6.2.18. By using the extension formula (6.3.3), in fact
∞ 
z2

1 2
= −z ∏ 1 − 2 for all z ∈ C,
Γ(z)Γ(−z) k=1 k
this somehow formally replace z by −z (but in fact not so obvious). Therefore, we reach
−π
Γ(z)Γ(−z) = for all z ∈ C \ Z,
z sin πz
that is (see also [FB09, Proposition IV.1.11]):

π
T HEOREM 6.3.5 (Completion Formula). Γ(z)Γ(1 − z) = sin πz for all z ∈ C \ Z.

As an immediate consequence, we have

√
Γ(1/2) = π.
1√ √
Applying the identity Γ(z + 1) = zΓ(z), we also have Γ(3/2) = 2 π, Γ(5/2) = 3 π/4, and so on.

E XERCISE 6.3.6. Show that

√ n−1
   
1 1
Γ n+ = π ∏ k+ for all n = 0, 1, 2, · · · .
2 k=0 2

We now restrict Γ(z) for z > 0. In fact, log ◦Γ is convex on (0, ∞), see [Rud76, Theorem 8.18].
It is a rather surprising fact (discovered by Borh and Mollerup) that: If f is a positive function
on (0, ∞) such that f (x + 1) = x f (x), f (1) = 1 and log ◦ f is convex, then f = Γ on (0, ∞), see
[Rud76, Theorem 8.19]. See also [FB09, Proposition IV.1.3] for a characterization of the complex
Γ-function. We finally end this section by exhibit a version of Stirling’s formula, which can be
found in [FB09, Proposition IV.1.14].

T HEOREM 6.3.7 (General Stirling’s formula). Let H be the function

∞     
1 1
H(z) = ∑ z+n+ Log 1 + −1 .
n=0 2 z+n
 6.3. THE GAMMA FUNCTION: AN EXTENSION OF FACTORIAL FUNCTION 90

Then for all z ∈ C \ R≤0 one has

√ 1
Γ(z) = 2πzz− 2 e−z eH(z) .

In any angular domain Wδ = z = |z|eiθ : −π + δ ≤ θ ≤ π − δ with 0 < δ ≤ π, we have H(z) → 0



as z → ∞. In addition, we have
1
0 < H(x) < for all x > 0.
12x
R EMARK . The above theorem is a generalization of the classical Stirling formula
√  n n ϕ(n)
n! = 2πn e 12n with 0 < ϕ(n) < 1,
e
which is established by J. Stirling at 1730.
 Bibliography

[AIM09] K. Astala, T. Iwaniec, and G Martin. Elliptic partial differential equations and quasiconformal mappings in
the plane, volume 48 of Princeton Mathematical Series. Princeton University Press, Princeton, NJ, 2009.
MR2472875, Zbl:1182.30001, doi:10.1515/9781400830114.
[BN10] J. Bak and D. Newman. Complex Analysis. Undergraduate Texts in Mathematics. Springer, New York, third
edition, 2010. MR2675489, Zbl:1205.30001, doi:10.1007/978-1-4419-7288-0.
[BBI01] D. Burago, Y. Burago, and S. Ivanov. A course in metric geometry, volume 33 of Graduate Studies
in Mathematics. American Mathematical Society, Providence, RI, 2001. MR1835418, Zbl:0981.51016,
doi:10.1090/gsm/033.
[Ego84] G. P. Egorychev. Integral representation and the computation of combinatorial sums, volume 59 of
Transl. Math. Monogr. American Mathematical Society, Providence, RI, 1984. Translated from the
Russian by H. H. McFadden. Translation edited by Lev J. Leifman. MR0736151, Zbl:0524.05001,
doi:10.1090/mmono/059.
[FB09] E. Freitag and R. Busam. Complex Analysis. Universitext. Springer-Verlag, Berlin, second edition, 2009.
MR2513384, Zbl:1167.30001, doi:10.1007/978-3-540-93983-2.
[GM12] A. Galbis and M. Maestre. Vector analysis versus vector calculus. Universitext. Springer, New York, 2012.
MR2895926, Zbl:1246.26001, doi:10.1007/978-1-4614-2200-6.
[GT01] D. Gilbarg and N. S. Trudinger. Elliptic partial differential equations of second order (reprint of the 1998
edition), volume 224 of Classics in Mathematics. Springer-Verlag Berlin Heidelberg, 2001. MR1814364,
Zbl:1042.35002, doi:10.1007/978-3-642-61798-0.
[Kow22] P.-Z. Kow. Fourier analysis and distribution theory. Lecture notes. University of Jyväskylä, 2022.
https://puzhaokow1993.github.io/homepage.
[KLSS24] P.-Z. Kow, S. Larson, M. Salo, and H. Shahgholian. Quadrature domains for the Helmholtz equation
with applications to non-scattering phenomena. Potential Anal., 60(1):387–424, 2024. MR4696043,
Zbl:1535.35044, doi:10.1017/s11118-022-10054-5. The results in the appendix are well-known, and the
proofs can found at arXiv:2204.13934.
[Mun00] J. R. Munkres. Topology. Prentice Hall, Inc., Upper Saddle River, NJ, second edition, 2000. MR3728284,
Zbl:0951.54001.
[Rud76] W. Rudin. Principles of mathematical analysis. International Series in Pure and Applied Mathematics.
McGraw-Hill Book Co., New York-Auckland-Düsseldorf, thid edition, 1976. MR0385023,
Zbl:0346.26002.
[Sal08] M. Salo. Calderón problem. Lecture notes. University of Helsinki, 2008. http://users.jyu.fi/∼salomi.

91