Complex Analysis Notes

University of
Cambridge
Mathematics Tripos
Part IB
Complex Analysis
erger
ger
Lent, 2018
erger
ergergergerge
Lectures by Liguori editore City:
G. P. Paternain
Year: 2001 Edition: Secon
er Language: Italian Pages: 248
Liguori editore City:
Year: 2001 Edition: ISBN:
Secon9788820768331 ID: 2505023
Notes by
Language: Italian Time
Pages:added:
Size: ID:
248 2020-04-20 12:59:23 Time modified: 2020-04-20 16
1 Mb (1285337)v
ISBN: 9788820768331 2505023
Qiangru Kuang
Time added: 2020-04-20 12:59:23 Time modified: 2020-04-20 16:00:01
Size: 1 Mb (1285337)
erg
erg Language: Italian Pages: 248
ISBN: 9788820768331 ID: 2505023
gregerg g Time added: 2020-04-20 12:59:23
erg Time modified: 2020-04-20 16:00:01
Size: 1 Mb (1285337) Liguori editore Ci
Liguori editore City: Year: 2001 Edition: Secon
Year: 2001 Edition: Secon Language: Italian Pages: 248
Language: Italian Pages: 248 ISBN: 9788820768331 ID: 2505023
ISBN: 9788820768331 ID: 2505023 Time added: 2020-04-20 12:59:23
Time modified:
Time added: 2020-04-20 12:59:23 Time modified: 2020-04-20 16:00:012020-04-20 16:00:01
Size: 1 Mb (1285337)erggggggggggggggggggggggggggg Size: 1 Mb (1285337)
Language: Italian Pages: 248
ISBN: 9788820768331 ID: 2505023
Size: 1 Mb (1285337)
ergergerg
regreg
Contents
Contents
ISBN: 9788820768331 ID: 2505023
1 Basic notions Time added: 2020-04-20 12:59:23 Time modified: 2020-04-20 2 16:00:01
1.1 Complex differentiationSize: 1 .Mb. .(1285337)
. . . . . Liguori
. . . .editore
. . . City:
. . . . . . . . 2
1.2 Power series . . . . Language:
. . . . . .Italian
. . .Pages:
. . . .248. . . . . . . . . . . . . 4
1.3 Conformal maps . ISBN: . . . 9788820768331
. . . . . . . . .ID:. 2505023
. . . . . . . . . . . . . 8
2 Complex Integration Size: I 1 Mb (1285337) Liguori editore City: 10
2.1 Integration along curves . . .Edition:
Year: 2001 . . . .Secon
. . . . . . . . . . . . . . . . 10
2.2 Cauchy’s Theorem, Language:
weak version Italian.Pages:
. . . .248. . . . . . . . . . . . . 15
ISBN: 9788820768331 ID: 2505023
2.3 Cauchy Integral Formula, weak version . . . . . . . . . . . . . . . 17
2.4 Application of Cauchy Size: 1 Mb (1285337) Liguori .editore
Integration Formula . . . City:
. . . . . . . . 18
2.5 Uniform limits of holomorphic functions
Year: 2001 Edition: Secon. . . . . . . . . . . . . . 20
2.6 Zeros
gerg of holomorphic functions
Language: . .Pages:
Italian . . . .248. . . . . . . . . . . . . 22
2.7 Analytic continuation ISBN:. 9788820768331
. . . . . . . . .ID:. 2505023
. . . . . . . . . . . . . 22
3 Complex Integration rgergerII 1 Mb (1285337) Liguori editore City:
Size:
egergreg 24
3.1 Winding number . Language:
. . . . . .Italian
. . .Pages:
. . . .248. . . . . . . . . . . . . 24
3.2 General form of Cauchy’s theorem . . .ID:. 2505023
ISBN: 9788820768331 . . . . . . . . . . . . . 26
4 Laurent expansion, Singularities
Size: 1 Mb (1285337)and the Residue
Liguori theorem
editore City: 29
4.1 Laurent expansion . .City:
Liguori editore
Year: . 2001
. . .Edition:
. . . .Secon
. . . . . . . . . . . . . . . . 29
4.2 Isolated gerg
singularities . . .Secon
Year: 2001 Edition:
Language: . .Italian
. . .Pages:
. . . .248. . . . . . . . . . . . . 30
Language: Italian
ISBN: Pages: 248
9788820768331 ID: 2505023
4.3 Application and techniques
ISBN: 9788820768331 of
Time added:ID:
integration . . . . Time
2505023 12:59:23
2020-04-20
. . .modified:
. . . . .2020-04-20
. 34 16:00:01
Time added:Size:
2020-04-20 12:59:23 Time modified: 2020-04-20 16:00:01
1 Mb (1285337)
5 The Argument principle,
Size: Local
1 Mb (1285337) degree,
Liguori Open
editore City: mapping theorem
& Rouché’s theorem
Year: 2001 Edition:
op[po[po[po[ Secon 38
Index ISBN: 9788820768331 ID: 2505023 41
Size: 1 Mb (1285337)
1
1 Basic notions
1 Basic notions
Some preliminary notations/definitions:
Notation.
• 𝐷(𝑎, 𝑟) is the open disc of radius 𝑟 > 0 and centred at 𝑎 ∈ C.
• 𝑈 ⊆ C is open if for any 𝑎 ∈ 𝑈, there exists 𝜀 > 0 such that 𝐷(𝑎, 𝜀) ⊆ 𝑈.
• A curve is a continuous map from a closed interval 𝜑 ∶ [𝑎, 𝑏] → C. It is
continuously differentiable, i.e. 𝐶 1 , if 𝜑′ exists and is continuous on [𝑎, 𝑏].
• An open set 𝑈 ⊆ C is path-connected if for every 𝑧, 𝑤 ∈ 𝑈 there exists a
curve 𝜑 ∶ [0, 1] → 𝑈 with endpoints 𝑧, 𝑤.
Definition (Domain). A domain is an non-empty path-connected open

subset of C.
1.1 Complex differentiation

The goal of this course is to study functions 𝑓 ∶ 𝑈 → C where 𝑈 ⊆ C is open or
is a domain. Given such an 𝑓, we may write
𝑓(𝑥 + 𝑖𝑦) = 𝑢(𝑥, 𝑦) + 𝑖𝑣(𝑥, 𝑦)
where 𝑢, 𝑣 ∶ 𝑈 → R are the real and imaginary part of 𝑓. Here we use (𝑥, 𝑦) ∈ R2
to denote the coordinates of 𝑎 + 𝑏𝑖 ∈ C.
Definition (Differentiable, holomorphic).
1. 𝑓 ∶ 𝑈 → C is differentiable at 𝑤 ∈ 𝑈 if the limit
𝑓(𝑧) − 𝑓(𝑤)
𝑓 ′ (𝑤) = lim
𝑧→𝑤 𝑧−𝑤
exists. 𝑓 ′ (𝑤) is called the derivative of 𝑓 at 𝑤.
2. 𝑓 is holomorphic at 𝑤 if there exists 𝜀 > 0 such that 𝑓 is differentiable

at all points of 𝐷(𝑤, 𝜀). 𝑓 is holomorphic on 𝑈 if it is differentiable at
all 𝑤 ∈ 𝑈. Equivalent, 𝑓 is holomorphic at at all 𝑤 ∈ 𝑈.
Remark.
1. There is an alternative term analytic. In actuality, it is the same as
holomorphic for complex functions. However, it comes with a flavour
associated with the Taylor expansion and sometimes defined in terms of
such. Later in the course we will prove that the definitions are equivalent.
2. Complex differentiation follows the same rules as real differentiation. For
example, sums of differentiable functions are differentiable, product, quo-
tient, chain rules etc also hold.
2
1 Basic notions
Definition (Entire). An entire function is a holomorphic function 𝑓 ∶ C → C.
Example.
1. Polynomials are entire functions.
2. If 𝑝(𝑧) and 𝑞(𝑧) are polynomials with 𝑞(𝑧) not identically zero, then 𝑝
𝑞 is
holomorphic on C \ {zeros of 𝑞}.
In IB Analysis II we studied function from R𝑛 to R𝑚 and their differentiability.
Indeed a complex function C → C can be view as a function R2 → R2 so how
does complex differentiability relates to differentiablitiy in R2 ? It turns out that
in addition to satisfy the differentiability on R2 , the function has to satisfy a
particular partial differential equation.
Recall that 𝑢 is differentiable at (𝑐, 𝑑) ∈ 𝑈 if there exists (𝜆, 𝜇) ∈ R2 such
that
𝑢(𝑥, 𝑦) − 𝑢(𝑐, 𝑑) − (𝜆(𝑥 − 𝑐) + 𝜇(𝑦 − 𝑑))
→0
√(𝑥 − 𝑐)2 + (𝑦 − 𝑑)2
as (𝑥, 𝑦) → (𝑐, 𝑑). In this case 𝒟𝑢(𝑐, 𝑑) = (𝜆, 𝜇) is the derivative of 𝑢 at (𝑐, 𝑑).
If this holds then 𝜆 = 𝑢𝑥 (𝑐, 𝑑) and 𝜇 = 𝑢𝑦 (𝑐, 𝑑), the partial derivatives of 𝑢 at
(𝑐, 𝑑).
Theorem 1.1 (Cauchy-Riemann equations). 𝑓 ∶ 𝑈 → C is differentiable at

𝑤 = 𝑐 + 𝑖𝑑 ∈ 𝑈 if and only if the functions 𝑢 and 𝑣 are differentiable at (𝑐, 𝑑)
and
𝑢𝑥 (𝑐, 𝑑) = 𝑣𝑦 (𝑐, 𝑑)
𝑢𝑦 (𝑐, 𝑑) = −𝑣𝑥 (𝑐, 𝑑)
in which case
𝑓 ′ (𝑤) = 𝑢𝑥 (𝑐, 𝑑) + 𝑖𝑣𝑥 (𝑐, 𝑑).
Proof. From the definition, 𝑓 will be differentiable at 𝑤 with derivative 𝑓 ′ (𝑤) =

𝑝 + 𝑖𝑞 if and only if
𝑓(𝑧) − 𝑓(𝑤) − 𝑓 ′ (𝑤)(𝑧 − 𝑤)
lim =0
𝑧→𝑤 |𝑧 − 𝑤|
or equivalently, splitting into real and imaginary parts,
𝑢(𝑥, 𝑦) − 𝑢(𝑐, 𝑑) − (𝑝(𝑥 − 𝑐) − 𝑞(𝑦 − 𝑑))
lim =0
(𝑥,𝑦)→(𝑐,𝑑) √(𝑥 − 𝑐)2 + (𝑦 − 𝑑)2
𝑣(𝑥, 𝑦) − 𝑣(𝑐, 𝑑) − (𝑞(𝑥 − 𝑐) + 𝑝(𝑦 − 𝑑))
lim =0
(𝑥,𝑦)→(𝑐,𝑑) √(𝑥 − 𝑐)2 + (𝑦 − 𝑑)2
since 𝑓 ′ (𝑤)(𝑧 − 𝑤) = (𝑝(𝑥 − 𝑐) − 𝑞(𝑦 − 𝑑)) + 𝑖(𝑞(𝑥 − 𝑐) + 𝑝(𝑦 − 𝑑)). So 𝑓 is
differentiable at 𝑤 with derivative 𝑓 ′ (𝑤) = 𝑝 + 𝑖𝑞 if and only if 𝑢 and 𝑣 are
differentiable at (𝑐, 𝑑) with
𝒟𝑢(𝑐, 𝑑) = (𝑝, −𝑞)
𝒟𝑣(𝑐, 𝑑) = (𝑞, 𝑝)
hence the result.
3
1 Basic notions
Example. Let 𝑓(𝑧) = 𝑧. Then 𝑓(𝑥 + 𝑖𝑦) = 𝑥 − 𝑖𝑦, 𝑢(𝑥, 𝑦) = 𝑥, 𝑣(𝑥, 𝑦) = −𝑦.
We have
𝑢𝑥 = 1 ≠ −1 = 𝑣𝑦
so 𝑓 is not differentiable anywhere in C.
Remark.
1. We could have discovered Cauchy-Riemann as follows: let 𝑧 = 𝑤 + ℎ where

ℎ ∈ R. Then
𝑓(𝑤 + ℎ) − 𝑓(𝑤)
𝑓 ′ (𝑤) = lim = 𝑢𝑥 (𝑐, 𝑑) + 𝑖𝑣𝑥 (𝑐, 𝑑).
ℎ→0 ℎ
Let 𝑧 = 𝑤 + 𝑖ℎ, we get
𝑓(𝑤 + 𝑖ℎ) − 𝑓(𝑤)

𝑓 ′ (𝑤) = lim = 𝑣𝑦 (𝑐, 𝑑) − 𝑖𝑢𝑦 (𝑐, 𝑑).
ℎ→0 ℎ
As 𝑓 is assumed to be differentiable, these two must agree.
2. Later on we’ll see that if 𝑓 is holomorphic, so is 𝑓 ′ . This will imply right
away that all partial derivatives of 𝑢 and 𝑣 exist and are continuous, i.e.
𝐶 ∞ . Thus
𝑢𝑥𝑥 = 𝑣𝑦𝑥
𝑢𝑦𝑦 = −𝑣𝑥𝑦
By symmetry of second derivatives, we get that 𝑢 satisfies the Laplace

equation
𝑢𝑥𝑥 + 𝑢𝑦𝑦 = 0.
Hence 𝑓 is holomorphic implies that the real and imaginary parts are
harmonic functions.
Corollary 1.2. Let 𝑓 = 𝑢 + 𝑖𝑣 ∶ 𝑈 → C. Suppose the functions 𝑢 and 𝑣 have

continuous partial derivatives everywhere in 𝑈 and satisfy Cauchy-Riemann
equations. Then 𝑓 is holomorphic in 𝑈.
Proof. From IB Analysis II, 𝑢 and 𝑣 are differentiable. The result follows from
the previous theorem.
Corollary 1.3. Let 𝑓 ∶ 𝐷 → C be holomorphic on a domain 𝐷 and suppose

𝑓 ′ (𝑧) = 0 for all 𝑧 ∈ 𝐷. Then 𝑓 is constant on 𝐷.
Proof. Follows from the analogous result for differentiable functions on a path-
connected subset of R2 (Mean Value Inequality from IB Analysis II).
1.2 Power series

Recall that
4
1 Basic notions
Theorem 1.4 (Radius of convergence). Let 𝑐𝑛 be a sequence of complex

numbers. Then there exists a unique 𝑅 ∈ [0, ∞], the radius of convergence
of the series, such that
∞
∑ 𝑐𝑛 (𝑧 − 𝑎)𝑛 , 𝑧, 𝑎 ∈ C
𝑛=0
converges absolutely if |𝑧 − 𝑎| < 𝑅 and diverges if |𝑧 − 𝑎| > 𝑅. If 0 < 𝑟 < 𝑅

then the series converges uniformly on {|𝑧−𝑎| ≤ 𝑟}. The radius of convergence
is given by
𝑅 = sup{𝑟 ≥ 0 ∶ |𝑐𝑛 |𝑟𝑛 → 0}.
Theorem 1.5. Let 𝑓(𝑧) = ∑∞ 𝑐 (𝑧 − 𝑎)𝑛 be a complex power series with

𝑛=0 𝑛
radius of convergence 𝑅 > 0. Then
1. 𝑓 is holomorphic on 𝐷(𝑎, 𝑅),
2. its derivative is given by the series ∑𝑛=1 𝑛𝑐𝑛 (𝑧 − 𝑎)𝑛−1 , which also has
∞
radius of convergence 𝑅,
3. 𝑓 has derivatives of all orders on 𝐷(𝑎, 𝑅) and 𝑓 (𝑛) (𝑎) = 𝑛!𝑐𝑛 ,
4. if 𝑓 vanishes identically on some disc 𝐷(𝑎, 𝜀) then 𝑐𝑛 = 0 for all 𝑛.
Proof. wlog assume 𝑎 = 0. Claim that ∑∞

𝑛=1
𝑛𝑐𝑛 𝑧𝑛−1 has radius of convergence
𝑅.
Proof. |𝑛𝑐𝑛 | ≥ |𝑐𝑛 | so its radius of convergence 𝑅′ is at most 𝑅.

Suppose 0 ≤ 𝑟 < 𝑅 and pick 𝜌 ∈ (𝑟, 𝑅). Then ∑ |𝑐𝑛 |𝜌𝑛 converges and
𝑛|𝑐𝑛 |𝑟𝑛−1 𝑛 𝑟 𝑛
𝑛
= ( ) →0
|𝑐𝑛 |𝜌 𝑟 ⏟
𝜌
<1
so by comparison test ∑ 𝑛𝑐𝑛 𝑟𝑛−1 converges and hence 𝑅′ = 𝑅.

To show the derivative is of the desired form we shall do something clever.
Consider the continous function
𝑛−1 𝑧𝑛 −𝑤𝑛
if 𝑧 ≠ 𝑤
ℎ𝑛 (𝑧, 𝑤) = ∑ 𝑧 𝑗 𝑤𝑛−1−𝑗 = { 𝑧−𝑤
𝑗=0 𝑛𝑤 𝑛−1
if 𝑧 = 𝑤
for 𝑛 ≥ 1. Consider the series

∞
∑ 𝑐𝑛 ℎ𝑛 (𝑧, 𝑤). (∗)
𝑛=1
Claim that for every 𝑟 < 𝑅, (∗) converges uniformly on the set {(𝑧, 𝑤) ∶ |𝑧|, |𝑤| ≤
𝑟}.
Proof. |𝑐𝑛 ℎ𝑛 (𝑧, 𝑤)| ≤ |𝑐𝑛 |𝑛𝑟𝑛−1 = 𝑀𝑛 . Since we know ∑ 𝑀𝑛 < ∞, by Weier-
strass M-test (∗) converges uniformly.
5
1 Basic notions
Thus it converges to a continuous function 𝑔(𝑧, 𝑤). By definition
if 𝑧 ≠ 𝑤
∞ 𝑛 𝑛 𝑓(𝑧)−𝑓(𝑤)
∑𝑛=1 𝑐𝑛 𝑧 𝑧−𝑤
−𝑤
= 𝑧−𝑤
𝑔(𝑧, 𝑤) = { ∞
∑𝑛=1 𝑐𝑛 𝑛𝑤 𝑛−1
if 𝑧 = 𝑤
As 𝑔 is continuous, fixing 𝑤 and letting 𝑧 → 𝑤 we get

∞
𝑓(𝑧) − 𝑓(𝑤)
lim = ∑ 𝑛𝑐𝑛 𝑤𝑛−1 .
𝑧→𝑤 𝑧−𝑤 𝑛=1
So 𝑓 ′ (𝑤) exists and equals to 𝑔(𝑤, 𝑤) as desired. This proves (1) and (2). (3)
follows by induction on 𝑛. Finally if 𝑓 vanishes identically on a disc about 𝑎
then 𝑓 (𝑛) = 0 for all 𝑛 and by (3) 𝑐𝑛 = 0 for all 𝑛.
Proposition 1.6 (Weierstrass M-test). Let 𝑓𝑛 be a sequence of functions

such that |𝑓𝑛 (𝑥)| ≤ 𝑀𝑛 for all 𝑥 ∈ 𝐴. If ∑ 𝑀𝑛 < ∞ then ∑ 𝑓𝑛 (𝑥) converges
uniformly on 𝐴.
Proof. Exercise.
Definition (Exponential). The complex exponential function is defined as

∞
𝑧𝑛
𝑒𝑧 = exp(𝑧) = ∑ .
𝑛=0
𝑛!
Proposition 1.7.
1. 𝑒𝑧 is an entire function and (𝑒𝑧 )′ = 𝑒𝑧 .

2. For all 𝑧, 𝑤 ∈ C, 𝑒𝑧+𝑤 = 𝑒𝑧 𝑒𝑤 and 𝑒𝑧 ≠ 0.
3. If 𝑧 = 𝑥 + 𝑖𝑦, 𝑒𝑧 = 𝑒𝑥 (cos 𝑦 + 𝑖 sin 𝑦).
4. exp(𝑧) = 1 if and only if 𝑧 ∈ 2𝜋𝑖Z.

5. If 𝑤 ∈ C, there exists 𝑧 ∈ C such that 𝑒𝑧 = 𝑤 if and only if 𝑤 ≠ 0.
Proof. Similar to the proof in IA Analysis I.

We also define
𝑒𝑖𝑧 + 𝑒−𝑖𝑧
cos 𝑧 =
2
𝑒𝑖𝑧 − 𝑒−𝑖𝑧
sin 𝑧 =
2𝑖
𝑒𝑧 + 𝑒−𝑧
cosh 𝑧 =
2
𝑒𝑧 − 𝑒−𝑧
sinh 𝑧 =
2
6
1 Basic notions
Definition (Logarithm). If 𝑧 ∈ C, we say that 𝑤 ∈ C is a logarithm of 𝑧 if

exp(𝑤) = 𝑧.
From above we know 𝑧 has a logarithm if and only if 𝑧 ≠ 0. However, unlike

in the real case, 𝑧 would have an infinite number of logarithms, all different by
an integer multiplier of 2𝜋.
Definition (Branch of logarithm). Let 𝑈 ⊆ C \ {0} be open. We say that a

continuous function 𝜆 ∶ 𝑈 → C is a branch of logarithm if 𝑒𝜆(𝑧) = 𝑧.
Remark. Any branch of logarithm is in fact automatically holomorphic:

𝜆(𝑧) − 𝜆(𝑤) 𝜆(𝑧) − 𝜆(𝑤)
= 𝜆(𝑧)
𝑧−𝑤 𝑒 − 𝑒𝜆(𝑤)
Let 𝜆(𝑤) = 𝑘, 𝜆(𝑧) = 𝑘 + ℎ. Due to continuity of 𝜆, 𝜆(𝑧) − 𝜆(𝑤) = ℎ → 0 so
ℎ
=
𝑒𝑘+ℎ − 𝑒𝑘
1
→
𝑒𝑘
1
=
𝑤
as 𝑧 → 𝑤. Thus 𝜆 is holomorphic and 𝜆′ (𝑧) = 1𝑧 .
One useful choice of branch is
Definition (Principal branch of logarithm). Let 𝑈 = C \ {𝑥 ∈ R ∶ 𝑥 ≤ 0}.

The principal branch of logarithm is the function
log ∶ 𝑈 → C
𝑧 ↦ log |𝑧| + 𝑖 arg(𝑧)
where arg(𝑧) takes the unique argument of 𝑧 in the interval (−𝜋, 𝜋).
Check that this is indeed a branch of logarithm: the projection onto 𝑆 1

C \ {0} → 𝑆 1 , 𝑧 ↦ 𝑧/|𝑧| is continuous and maps 𝑈 to 𝑆 1 \ {−1}. (−𝜋, 𝜋) →
𝑆 1 \ {−1}, 𝜃 ↦ 𝑒𝑖𝜃 is a homeomorphism so has a continuous inverse. Thus
𝑧 ↦ arg(𝑧) is continuous. Also
exp(log 𝑧) = 𝑒𝑖 log |𝑧| (cos arg(𝑧) + 𝑖 sin arg(𝑧)) = 𝑧.
Proposition 1.8.
1. log 𝑧 is holomorphic on 𝑈 with derivative 1𝑧 .
2. If |𝑧| < 1 then
∞
(−1)𝑛−1 𝑧𝑛
log(1 + 𝑧) = ∑ .
𝑛=1
𝑛
Proof.
7
1 Basic notions
1. See remarks above.

2. Note that
𝑑 log(1 + 𝑧) 1
=
𝑑𝑧 1+𝑧
𝑑 ∞ (−1)𝑛−1 𝑧𝑛 ∞
1
∑ = ∑(−1)𝑛−1 𝑧 𝑛−1 =
𝑑𝑧 𝑛=1 𝑛 𝑛=1
1 + 𝑧
Thus their difference is constant. Now set 𝑧 = 0.
Remark. There is no way to extend log 𝑧 to a holomorphic (or even just

continuous) function on C \ {0} since
lim log 𝑒𝑖𝜃 = −𝑖𝜋

𝜃→−𝜋
lim log 𝑒𝑖𝜃 = 𝑖𝜋
𝜃→𝜋
Later on we will see that there is no branch of logarithm on C \ {0}.

We can also define fractional/complex powers by the formula
𝑧 𝛼 = exp(𝛼 log 𝑧)
for 𝑧 ∈ 𝑈. See example sheet.
1.3 Conformal maps

Let 𝑓 ∶ 𝑈 → C be a holomorphic function on an open set 𝑈 and suppose 𝑤 ∈ 𝑈
and 𝑓 ′ (𝑤) ≠ 0. In this section we study the property of 𝑓 in neighbourhoods of
𝑤. Take two 𝐶 1 -curves 𝛾𝑖 ∶ [0, 1] → 𝑈, 𝑖 = 1, 2 such that 𝛾𝑖 (0) = 𝑤, 𝛾𝑖′ (0) ≠ 0.
Define the angle between 𝛾1 , 𝛾2 to be
angle(𝛾1 , 𝛾2 ) = arg(𝛾1′ (0)) − arg(𝛾2′ (0)).
Here arg(𝑧) = 𝜃 ∈ [0, 2𝜋) for 𝑧 = 𝑟𝑒𝑖𝜃 . Consider the images 𝛿𝑖 (𝑡) = 𝑓(𝛾𝑖 (𝑡)). The
new angle is
angle(𝛿1 , 𝛿2 ) = arg((𝑓 ∘ 𝛾1 )′ (0)) − arg((𝑓 ∘ 𝛾2 )′ (0)).
Now recall that

𝑧
arg(𝑧) − arg(𝑤) = arg( )
𝑤
so
(𝑓 ∘ 𝛾1 )′ (0) 𝑓 ′ (𝑤)𝛾1′ (0) 𝛾1′ (0)
= =
(𝑓 ∘ 𝛾2 )′ (0) 𝑓 ′ (𝑤)𝛾2′ (0) 𝛾2′ (0)
since 𝑓 ′ (𝑤) ≠ 0. Thus 𝑓 preserves angles at 𝑤. We say 𝑓 is conformal at 𝑤 in
this case.
8
1 Basic notions
Definition (Conformal equivalence). If 𝑓 ∶ 𝐷 → C is holomorphic on a

domain 𝐷 with 𝑓 ′ ≠ 0 everywhere and 𝑓 injective, we say 𝑓 is a conformal
equivalence.
Example.
1. Möbius maps 𝑓(𝑧) = 𝑎𝑧+𝑏𝑐𝑧+𝑑 where 𝑎𝑑 − 𝑏𝑐 ≠ 0. It is a bijection from C ∪ {∞}
to itself. It maps circles/lines to circles/lines.
2. 𝑧 ↦ 𝑧 𝑛 , {𝑧 ∈ C \ {0} ∶ 0 < arg 𝑧 < 𝜋/𝑛} → {𝑧 ∈ C ∶ Im 𝑧 > 0}. The

inverse, 𝑧 1/𝑛 using the principal branch, is also a conformal map.
3. exp maps conformally the strip {𝑧 ∈ C ∶ Im(𝑧) ∈ (−𝜋, 𝜋)} to C \ {𝑧 ≤ 0},
with inverse log 𝑧 with principal branch cut.
We will just state an important theorem here which will not be proven in
this course:
Theorem 1.9 (Riemann mapping theorem). Let 𝐷 ⊆ C be any domain

bounded by a simple closed curve. Then there exists a conformal equivalence
𝐷 → 𝐷(0, 1) between 𝐷 and 𝐷(0, 1).
More generally, this holds for any simply connected domain which is not
all C.
9
2 Complex Integration I
2.1 Integration along curves
Let 𝑓 ∶ [𝑎, 𝑏] → C be a continuous map, then its integral is simply the sum of
the integral of the real and imaginary part:
𝑏 𝑏 𝑏
∫ 𝑓(𝑡)𝑑𝑡 = ∫ Re 𝑓(𝑡)𝑑𝑡 + 𝑖 ∫ Im 𝑓(𝑡)𝑑𝑡.
𝑎 𝑎 𝑎
Proposition 2.1.
𝑏
∣∫ 𝑓(𝑡)𝑑𝑡∣ ≤ (𝑏 − 𝑎) sup |𝑓(𝑡)|
𝑎 𝑡∈[𝑎,𝑏]
with equality if and only if 𝑓 is constant.
Proof. Let 𝜃 = arg ∫ 𝑓(𝑡)𝑑(𝑡) and let 𝑀 = sup𝑡∈[𝑎,𝑏] |𝑓(𝑡)|. Clearly

𝑏
𝑎
𝑏 𝑏
∫ 𝑓(𝑡)𝑑𝑡 = ∣∫ 𝑓(𝑡)𝑑𝑡∣ 𝑒𝑖𝜃
𝑎 𝑎
so
𝑏 𝑏
∣∫ 𝑓(𝑡)𝑑𝑡∣ = 𝑒−𝑖𝜃 ∫ 𝑓(𝑡)𝑑𝑡
𝑎 𝑎
𝑏
= ∫ 𝑒−𝑖𝜃 𝑓(𝑡)𝑑𝑡
𝑎
𝑏
= ∫ Re(𝑒−𝑖𝜃 𝑓(𝑡))𝑑𝑡
𝑎
𝑏
≤ ∫ |𝑓(𝑡)|𝑑𝑡
𝑎
≤ 𝑀 (𝑏 − 𝑎)
If the equality holds then |𝑓(𝑡)| = 𝑀 is constant by the last inequality, and it
follows from the second last inequality that arg 𝑓(𝑡) = 𝜃 so 𝑓 is constant.
If 𝛾 ∶ [𝑎, 𝑏] → C is a 𝐶 1 curve, we can define its length as follow. Write

𝛾(𝑡) = 𝑥(𝑡) + 𝑖𝑦(𝑡) and define the length to be
𝑏 𝑏
∫ |𝛾 ′ (𝑡)|𝑑𝑡 = ∫ √𝑥′ (𝑡) + 𝑦′ (𝑡)𝑑𝑡.
𝑎 𝑎
Definition (Simple curve). We say 𝛾 is simple if 𝛾(𝑡1 ) ≠ 𝛾(𝑡2 ) unless 𝑡1 = 𝑡2

or {𝑡1 , 𝑡2 } = {𝑎, 𝑏}.
10
Definition (Integration along a curve). Suppose 𝑓 ∶ 𝑈 → C is continuous

with 𝑈 ⊆ C an open subset. Let 𝛾 ∶ [𝑎, 𝑏] → 𝑈 be a 𝐶 1 curve. The integral
of 𝑓 along 𝛾 is
𝑏
∫ 𝑓(𝑧)𝑑𝑧 = ∫ 𝑓(𝛾(𝑡))𝛾 ′ (𝑡)𝑑𝑡.
𝛾 𝑎
Some basic properties:

1. linearity:
∫(𝑐1 𝑓1 (𝑧) + 𝑐2 𝑓2 (𝑧))𝑑𝑧 = 𝑐1 ∫ 𝑓1 (𝑧)𝑑𝑧 + 𝑐2 ∫ 𝑓2 (𝑧)𝑑𝑧.

𝛾 𝛾 𝛾
2. additivity: if 𝑎 < 𝑎′ < 𝑏,
∫ 𝑓(𝑧)𝑑𝑧 = ∫ 𝑓(𝑧)𝑑𝑧 + ∫ 𝑓(𝑧)𝑑𝑧.

𝛾 𝛾|[𝑎,𝑎′ ] 𝛾|[𝑎′ ,𝑏]
3. inverse path: define the inverse path of 𝛾 to be
−𝛾 ∶ [−𝑏, −𝑎] → 𝑈
𝑡 ↦ 𝛾(−𝑡)
then
∫ 𝑓(𝑧)𝑑𝑧 = − ∫ 𝑓(𝑧)𝑑𝑧.
−𝛾 𝛾
4. independence of reparameterisation: if 𝜙 ∶ [𝑎′ , 𝑏′ ] → [𝑎, 𝑏] is 𝐶 1 and

𝜙(𝑎′ ) = 𝑎, 𝜙(𝑏′ ) = 𝑏, then let 𝛿 = 𝛾 ∘ 𝜙 ∶ [𝑎′ , 𝑏′ ] → 𝑈, we have
∫ 𝑓(𝑧)𝑑𝑧 = ∫𝑓(𝑧)𝑑𝑧.
𝛾 𝛿
All except the last one is trivial. A quick check

𝑏′
∫𝑓(𝑧)𝑑𝑧 = ∫ 𝑓(𝛿(𝑡))𝛿 ′ (𝑡)𝑑𝑡
𝛿 𝑎′
𝑏′
= ∫ 𝑓(𝛾(𝜙(𝑡)))𝛾 ′ (𝜙(𝑡))𝜙′ (𝑡)𝑑𝑡
𝑎′
𝑏
= ∫ 𝑓(𝛾(𝑠))𝛾 ′ (𝑠)𝑑𝑠
𝑎
= ∫ 𝑓(𝑧)𝑑𝑧
𝛾
In reality, we may encounter curves that are not 𝐶 1 , for example along the
sides of a square. However, they are made of pieces of 𝐶 1 curves. In a sense that
will be address by the definition below, this is the worst case we will encounter
in this course.
11
Definition (Piecewise 𝐶 1 ). Let 𝛾 ∶ [𝑎, 𝑏] → C be a continuous curve. Sup-

pose we have
𝑎 = 𝑎 0 < 𝑎 1 < ⋯ < 𝑎𝑛 = 𝑏
such that 𝛾|[𝑎𝑖−1 ,𝑎𝑖 ] is 𝐶 1 for 1 ≤ 𝑖 ≤ 𝑛. Then we say that 𝛾 is piecewise
differentiable or piecewise 𝐶 1 .
Then we define 𝑛
∫ 𝑓(𝑧)𝑑𝑧 = ∑ ∫ 𝑓(𝑧)𝑑𝑧.
𝛾 𝑖=1 𝛾𝑖
Example.
1. 𝑓(𝑧) = 𝑧 𝑛 where 𝑛 ∈ Z. Form now on define the punctured complex plane

to be C∗ = C \ {0}. Let 𝑈 = C∗ . Let
𝛾 ∶ [0, 2𝜋] → 𝑈
𝑡 ↦ 𝑒𝑖𝑡
which is a circle around the origin. Then

2𝜋 2𝜋
2𝜋𝑖 if 𝑛 = −1
∫ 𝑓(𝑧)𝑑𝑧 = ∫ 𝑒𝑖𝑛𝑡 𝑖𝑒𝑖𝑡 𝑑𝑡 = 𝑖 ∫ 𝑒𝑖(𝑛+1)𝑡 𝑑𝑡 = {
𝛾 0 0 0 otherwise
In the case 𝑛 = −1 𝑓 is not defined at 0, which is encircled by the closed

path.
2. 𝑓(𝑧) = 𝑧 2 . 𝛾 = 𝛾1 + 𝛾2 where
𝛾1 ∶ [−𝑅, 𝑅] → C
𝑡↦𝑡
𝛾2 ∶ [0, 1] → C
𝑡 ↦ 𝑅𝑒𝑖𝜋𝑡
Then
∫ 𝑓(𝑧)𝑑𝑧
𝛾
= ∫ 𝑓(𝑧)𝑑𝑧 + ∫ 𝑓(𝑧)𝑑𝑧
𝛾1 𝛾2
𝑅 1
= ∫ 𝑡2 𝑑𝑡 + ∫ 𝑅2 𝑒2𝜋𝑖𝑡 𝑖𝜋𝑅𝑒𝑖𝜋𝑡 𝑑𝑡
−𝑅 0
2𝑅3 2𝑅3
= −
3 3
=0
In general, for a holomorphic function, if the closed path does not include a
pole of the function then the integral along the curve is 0.
12
Proposition 2.2. For any continuous 𝑓 ∶ 𝑈 → C and any curve 𝛾 ∶ [𝑎, 𝑏] →

𝑈,
∣∫ 𝑓(𝑧)𝑑𝑧∣ ≤ ℓ(𝛾) ⋅ sup |𝑓|

𝛾 𝛾
where sup𝛾 |𝑓| = sup𝑡∈[𝑎,𝑏] |𝑓(𝛾(𝑡))|.
Proof. wlog assume 𝛾 is 𝐶 1 ,

𝑏
∣∫ 𝑓(𝑧)𝑑𝑧∣ = ∣∫ 𝑓(𝛾(𝑡))𝛾 ′ (𝑡)𝑑𝑡∣
𝛾 𝑎
𝑏
≤ ∫ |𝑓(𝛾(𝑡))||𝛾 ′ (𝑡)|𝑑𝑡
𝑎
𝑏
≤ sup |𝑓| ∫ |𝛾 ′ (𝑡)|𝑑𝑡
𝛾 𝑎
= sup |𝑓| ⋅ ℓ(𝑓)
𝛾
Theorem 2.3 (Fundamental Theorem of Calculus). Suppose 𝑓 ∶ 𝑈 → C is

continuous and suppose there exists 𝐹 (𝑧) such that for all 𝑧 ∈ 𝑈, 𝐹 ′ (𝑧) = 𝑓(𝑧).
Then for any curve 𝛾 ∶ [𝑎, 𝑏] → 𝑈, we have
∫ 𝑓(𝑧)𝑑𝑧 = 𝐹 (𝛾(𝑏)) − 𝐹 (𝛾(𝑎)).

𝛾
Note. 𝐹 is called an anti-derivative of 𝑓 on 𝑈.

Proof.
𝑏
∫ 𝑓(𝑧)𝑑𝑧 = ∫ 𝑓(𝛾(𝑡))𝛾 ′ (𝑡)𝑑𝑡
𝛾 𝑎
𝑏
= ∫ (𝐹 ∘ 𝛾)′ (𝑡)𝑑𝑡
𝑎
= 𝐹 (𝛾(𝑏)) − 𝐹 (𝛾(𝑎))
Corollary 2.4. If 𝛾 is closed (𝛾(𝑏) = 𝛾(𝑎)) and 𝑓 is continuous and has an

anti-derivative 𝐹 on 𝑈, then
∫ 𝑓(𝑧)𝑑𝑧 = 0.
𝛾
Example. Consider 𝑓(𝑧) = 𝑧 𝑛 where 𝑛 ∈ Z defined on C∗ . If 𝑛 ≠ −1,

𝑑 𝑧𝑛+1
𝑓(𝑧) = ( )
𝑑𝑧 𝑛 + 1
13
so the integral along any closed curve is 0.

However, if 𝑛 = −1 we do not have an anti-derivative on all C∗ and we
computed last time
𝑑𝑧
∫ = 2𝜋𝑖.
𝛾
𝑧
This is because we can’t define log as an anti-derivative of 1

𝑧 on all C∗ .
Now we prove a converse to the corollary:
Proposition 2.5. Let 𝑈 ⊆ C be a domain. If 𝑓 ∶ 𝑈 → C is continuous and

∫ 𝑓(𝑧)𝑑𝑧 = 0 for all closed curves 𝛾 on 𝑈, then 𝑓 has an anti-derivative 𝐹
𝛾
on 𝑈, i.e. there exists 𝐹 such that 𝐹 ′ = 𝑓.
Proof. Given that 𝑈 is path-connected, the immediate choice of 𝐹 coming to

mind is
𝐹 (𝑤) = ∫ 𝑓(𝑧)𝑑𝑧
𝛾𝑤
where 𝛾𝑤 ∶ [0, 1] → 𝑈 is a path from a fixed point 𝑎0 to 𝑤. Note that 𝐹 is

independent of the choice of 𝛾𝑤 precisely because of the hypothesis.
Now claim that 𝐹 is holomorphic and 𝐹 ′ = 𝑓. Since 𝑈 is open, there exists
𝑟 > 0 such that 𝐷(𝑤, 𝑟) ⊆ 𝑈. Let ℎ be such that |ℎ| < 𝑟 and let 𝛿ℎ be the radial
path from 𝑤 to 𝑤 + ℎ. Note that 𝛿𝑛 ⊆ 𝐷(𝑤, 𝑟). Let 𝛾 = 𝛾𝑤 ⋅ 𝛿ℎ ⋅ (−𝛾𝑤+ℎ ) where
⋅ denotes path concatenation. Then
0 = ∫ 𝑓(𝑧)𝑑𝑧 = 𝐹 (𝑤) + ∫ 𝑓(𝑧)𝑑𝑧 − 𝐹 (𝑤 + ℎ).

𝛾 𝛿ℎ
Now
𝐹 (𝑤 + ℎ) − 𝐹 (𝑤) 1
∣ − 𝑓(𝑤)∣ = ∣∫ (𝑓(𝑧) − 𝑓(𝑤))𝑑𝑧∣
ℎ |ℎ| 𝛿
ℎ
ℓ(𝛿ℎ )
≤ sup |𝑓(𝑧) − 𝑓(𝑤)|
|ℎ| 𝑧∈𝛿ℎ
→1⋅0
as ℎ → 0 since 𝑓 is continuous.
This proposition, or the proof thereof contains an idea that can be applied
to more general circumstances. We first define
Definition (Star-shaped domain). A domain 𝑈 is star-shaped (or called

a star domain) if there exists 𝑝 ∈ 𝑈 such that for all 𝑎 ∈ 𝑈, the straight
segment from 𝑎 to 𝑝 is contained in 𝑈.
We thus have
disc ⟹ convex ⟹ star-shape ⟹ domain.
14
Corollary 2.6. If 𝑈 is star-shaped, 𝑓 ∶ 𝑈 → C is continuous and ∫ 𝑓(𝑧)𝑑𝑧 =

𝛾
0 for all triangles 𝛾 ⊆ 𝑈 then 𝑓 has an anti-derivative on 𝑈.
A triangle is, as one might infer, the region bounded by a closed curve made
of three straight line segments.
Proof. Let 𝑎0 = 𝑎 and 𝛿𝑤 and 𝛿𝑤+ℎ be straight line segments in the previous
proof.
2.2 Cauchy’s Theorem, weak version
Theorem 2.7 (Cauchy’s Theorem for triangles). Let 𝑈 be a domain and 𝑇

be a triangle contained in 𝑈. If 𝑓 ∶ 𝑈 → C is holomorphic, then
∫ 𝑓(𝑧)𝑑𝑧 = 0.
𝜕𝑇
Proof. Let 𝑦 = | ∫ 𝑓(𝑧)𝑑𝑧|, ℓ = ℓ(𝜕𝑇 ). Let 𝑇 = 𝑇 0 = 𝑇1 ∪ ⋯ ∪ 𝑇4 be a

𝜕𝑇
subdivision into 4 equal triangles all with the same orientation. Then
4
∫ 𝑓(𝑧)𝑑𝑧 = ∑ ∫ 𝑓(𝑧)𝑑𝑧
𝜕𝑇 𝑖=1 𝜕𝑇𝑖
since internal lines cancel in pair. Thus there exists 𝑖 such that
𝑦
∣∫ 𝑓(𝑧)𝑑𝑧∣ ≥ .
𝜕𝑇𝑖
4
Let 𝑇 1 = 𝑇𝑖 for this 𝑖 and repeat. We produce a sequence
𝑇 0, 𝑇 1, …
such that
𝑦
∣∫ 𝑓(𝑧)𝑑𝑧∣ ≥
𝜕𝑇 𝑖
4𝑖
and
ℓ
ℓ(𝜕𝑇 𝑖 ) = .
2𝑖
Now as 𝑇 is compact, the 𝑇 𝑖 are closed nested sets, they have non-empty
intersection, i.e. there exists
∞
𝑧0 ∈ ⋂ 𝑇 𝑖 .
𝑖=0
As 𝑓 is differentiable at 𝑧0 , given 𝜀 > 0, there exists 𝛿 > 0 such that |𝑤 − 𝑧0 | < 𝛿

implies
|𝑓(𝑤) − 𝑓(𝑧0 ) − (𝑤 − 𝑧0 )𝑓 ′ (𝑧0 )| < 𝜀|𝑤 − 𝑧0 |.
15
Pick 𝑛 such that 𝑇 𝑛 ⊆ 𝐷(𝑧0 , 𝛿). Then

𝑦
≤ ∣∫ 𝑓(𝑧)𝑑𝑧∣
4𝑛 𝜕𝑇 𝑛
= ∣∫ (𝑓(𝑧) − 𝑓(𝑧 ′
0 ) − (𝑧 − 𝑧0 )𝑓 (𝑧0 ))𝑑𝑧∣
⏟⏟⏟⏟⏟⏟⏟⏟⏟
𝜕𝑇 𝑛 has anti-derivative so =0
≤ ℓ(𝜕𝑇 𝑛 )𝜀 sup |𝑧 − 𝑧0 |
𝑧∈𝜕𝑇 𝑛
𝜀ℓ2
≤ 𝑛.
4
Thus 𝑦 ≤ ℓ2 𝜀. Since 𝜀 is arbitrary, 𝑦 = 0.
Corollary 2.8 (Convex Cauchy). Let 𝑓 be a holomorphic function on a

star-domain 𝑈. Then
∫ 𝑓(𝑧)𝑑𝑧 = 0
𝛾
for all closed curves in 𝑈.

Proof. By the previous theorem, ∫ 𝑓(𝑧)𝑑𝑧 = 0 for all triangles 𝑇 ⊆ 𝑈. By
𝜕𝑇
Corollary 2.6, 𝑓 has an anti-derivative 𝐹 on 𝑈. Then the Fundamental Theorem
of Calculus says
∫ 𝑓(𝑧)𝑑𝑧 = 0
𝛾
for all closed paths 𝛾 in 𝑈.
Actually, we can prove the same result even if 𝑓 is not assumed to be
holomorphic on finitely many points:
Proposition 2.9. Let 𝑈 be a star-domain and 𝑆 ⊆ 𝑈 be a finite set. Let

𝑓 ∶ 𝑈 → C be continuous on 𝑈 and holomorphic on 𝑈 \𝑆. Then the conclusion
of the convex Cauchy theorem still holds.
Proof. It suffices to prove that the integral along a triangle is zero for all triangles
in 𝑈. Subdivide a triangle 𝑇 into 4𝑛 equal triangles as before. Let 𝑀 = sup𝑇 |𝑓|.
Let 𝑇 ′ ⊆ 𝑇 be a subtriangle. If 𝑇 ′ ∩ 𝑆 = ∅ then the integral is zero by Cauchy’s
Theorem for triangles. Otherwise, for any 𝑇 ′ ,
∣∫ 𝑓(𝑧)𝑑𝑧∣ ≤ ℓ(𝜕𝑇 ′ )𝑀 2−𝑛

𝜕𝑇 ′
so
∣∫ 𝑓(𝑧)𝑑𝑧∣ = ∣∑ ∫ 𝑓(𝑧)𝑑𝑧∣
𝜕𝑇 𝑇′ 𝜕𝑇 ′
≤ ∑ ∣∫ 𝑓(𝑧)𝑑𝑧∣
𝑇′ 𝜕𝑇 ′
≤ 6|𝑆|ℓ𝑀 2−𝑛
where the last line is because a point can belong to at most 6 triangles. Let
𝑛 → ∞ gives the result.
16
In fact, we will later prove that 𝑓 is also holomorphic on 𝑆.
2.3 Cauchy Integral Formula, weak version
Theorem 2.10 (Cauchy integral formula for a disc). Let 𝐷 = 𝐷(𝑎, 𝑟) be a

disc and 𝑓 ∶ 𝐷 → C holomorphic. For every 𝑤 ∈ 𝐷 and 𝜌 with |𝑤−𝑎| < 𝜌 < 𝑟,
1 𝑓(𝑧)
𝑓(𝑤) = ∫ 𝑑𝑧.
2𝜋𝑖 |𝑧−𝑎|=𝜌 𝑧 − 𝑤
This is a surprising result as the value of a holomorphic function inside a

disc is completely determined by its value on the boundary.
Proof. Consider the function
𝑓(𝑧)−𝑓(𝑤)
𝑧−𝑤 if 𝑧 ≠ 𝑤
𝑔(𝑧) = { ′
𝑓 (𝑤) if 𝑧 = 𝑤
Then 𝑔 is continuous on 𝐷 and holomorphic on 𝐷\{𝑤}. Thus by the previous

result
∫ 𝑔(𝑧)𝑑𝑧 = 0.
|𝑧−𝑎|=𝜌
It would be easy if the path is a circle around 𝑤 but unfortuntely it isn’t. Instead,
𝑓(𝑧) 𝑓(𝑤)
∫ 𝑑𝑧 = ∫ 𝑑𝑧 (1)
|𝑧−𝑎|=𝜌
𝑧−𝑤 |𝑧−𝑎|=𝜌
𝑧 −𝑤
∞
𝑓(𝑤)(𝑤 − 𝑎)𝑛
= ∑∫ 𝑑𝑧 (2)
𝑛=0 |𝑧−𝑎|=𝜌
(𝑧 − 𝑎)𝑛+1
𝑓(𝑤)
=∫ 𝑑𝑧 (3)
|𝑧−𝑎|=𝜌
𝑧−𝑎
= 2𝜋𝑖𝑓(𝑤) (4)
where (3) is justified by using the example of 𝑧 𝑛 before and (2) is justified by
the geometric series
∞
1 1 (𝑤 − 𝑎)𝑛
= 𝑤−𝑎 =∑
𝑧−𝑤 (𝑧 − 𝑎) (1 − 𝑧−𝑎 ) 𝑛=0
(𝑧 − 𝑎)𝑛+1
which converge uniformly as a function of 𝑧 on |𝑧 −𝑎| = 𝜌, as well as the following

easy lemma:
Lemma 2.11. If 𝑓𝑛 ∶ 𝑈 → C and 𝑓 ∶ 𝑈 → C are continuous and 𝛾 ∶ [𝑎, 𝑏] →

𝑈 is a curve such that 𝑓𝑛 → 𝑓 uniformly on 𝛾([𝑎, 𝑏]), then
∫ 𝑓𝑛 (𝑧)𝑑𝑧 → ∫ 𝑓(𝑧)𝑑𝑧.
𝛾 𝛾
17
Proof. Let 𝑀𝑛 = sup𝛾 |𝑓𝑛 − 𝑓|. By uniform convergence lim𝑛→∞ 𝑀𝑛 = 0. Hence
∣∫ 𝑓𝑛 (𝑧)𝑑𝑧 − ∫ 𝑓(𝑧)𝑑𝑧∣ ≤ 𝑀𝑛 ℓ(𝛾) → 0

𝛾 𝛾
as 𝑛 → ∞.
Corollary 2.12 (Mean-value property). If 𝑓 ∶ 𝐷(𝑤, 𝑅) → C is holomorphic,

then for every 0 < 𝑟 < 𝑅,
1
𝑓(𝑤) = ∫ 𝑓(𝑤 + 𝑟𝑒2𝜋𝑖𝑡 )𝑑𝑡.
0
Proof. Take 𝑤 = 𝑎 in the previous theorem and parameterise the circle of

integration as 𝛾(𝑡) = 𝑤 + 𝑟𝑒2𝜋𝑖𝑡 , 𝑡 ∈ [0, 1].
2.4 Application of Cauchy Integration Formula
Theorem 2.13 (Liouville). Every bounded entire function is constant.
Proof. Suppose 𝑓 ∶ C → C is entire such that |𝑓| ≤ 𝑀 and let 𝑤 ∈ C. Then if

𝑅 > |𝑤|,
1 1 1
|𝑓(𝑤) − 𝑓(0)| = ∣∫ 𝑓(𝑧) ( − ) 𝑑𝑧∣
2𝜋 |𝑧|=𝑅 𝑧−𝑤 𝑧
1 𝑤𝑓(𝑧)
= ∣∫ 𝑑𝑧∣
2𝜋 |𝑧|=𝑅 𝑧(𝑧 − 𝑤)
1 2𝜋𝑅|𝑤|𝑀
≤
2𝜋 𝑅(𝑅 − |𝑤|)
→0
as 𝑅 → ∞.
Theorem 2.14 (Fundamental Theorem of Algebra). Every non-constant

polynomial with complex coefficients has a complex root.
Proof. Let 𝑝(𝑧) = 𝑧 𝑛 + 𝑐𝑛−1 𝑧 𝑛−1 + ⋯ + 𝑐0 be a polynomial of degree 𝑛 > 0. Then

|𝑝(𝑧)| → ∞ as 𝑧 → ∞, so there exists 𝑅 > 0 such that |𝑝(𝑧)| > 1 for all |𝑧| > 𝑅.
Now consider 𝑓(𝑧) = 𝑝(𝑧) 1
. If 𝑝 does not have a zero then 𝑓 is an entire
function. Moreover, since 𝑓 is bounded on {𝑧 ∶ |𝑧| ≤ 𝑅} by continuity and
|𝑓(𝑧)| < 1 if |𝑧| > 𝑅. By Liouville 𝑓 is constant. Absurd.
18
Theorem 2.15 (Local maximum principle). Let 𝑓 ∶ 𝐷(𝑎, 𝑟) → C be holo-

morphic. If for every 𝑧 ∈ 𝐷(𝑎, 𝑟), |𝑓(𝑧)| ≤ |𝑓(𝑎)| then 𝑓 is constant.
Thus on a open domain 𝑈, if the function 𝑓 extends analytically to the

boundary 𝜕𝑈 then the maximum can only be achieved on the boundary 𝑈, unless
𝑓 is constant.
Proof. By Mean-value property we have for 0 < 𝜌 < 𝑟,
1
|𝑓(𝑎)| = ∣∫ 𝑓(𝑎 + 𝜌𝑒2𝜋𝑖𝑡 )𝑑𝑡∣ ≤ sup |𝑓(𝑧)| ≤ |𝑓(𝑎)|
0 |𝑧−𝑎|=𝜌
Hence equality holds, and by Proposition 2.1 |𝑓(𝑧)| = |𝑓(𝑎)| for all 𝑧 on |𝑧−𝑎| = 𝜌.
Since 𝜌 is arbitrary, |𝑓(𝑧)| = |𝑓(𝑎)| for all 𝑧 ∈ 𝐷(𝑎, 𝑟). Thus 𝑓 is constant (see
example sheet 1).
Theorem 2.16 (Taylor expansion). Let 𝑓 ∶ 𝐷(𝑎, 𝑟) → C be holomorphic.

Then 𝑓 has a convergent power series on 𝐷(𝑎, 𝑟):
∞
𝑓(𝑧) = ∑ 𝑐𝑛 (𝑧 − 𝑎)𝑛
𝑛=0
where
𝑓 (𝑛) (𝑎) 1 𝑓(𝑧)
𝑐𝑛 = = ∫ 𝑑𝑧
𝑛! 2𝜋𝑖 |𝑧−𝑎|=𝜌 (𝑧 − 𝑎)𝑛+1
for any 0 < 𝜌 < 𝑟.
An immediate corollary is that holomorphic functions are analytic. The

converse is trivial so the two terms are semantically synonymous, although the
choice of terminology may enunciate your standing point: the term “holomorphic”
is preferred in pure mathematics, while “analytic” is used more often in applied
mathematics.
Proof. If |𝑤 − 𝑎| < 𝜌 < 𝑟 then by Cauchy integral formula for a disc
1 𝑓(𝑧)
𝑓(𝑤) = ∫ 𝑑𝑧
2𝜋𝑖 |𝑧−𝑎|=𝜌 𝑧 − 𝑤
∞
1 (𝑤 − 𝑎)𝑛
= ∫ ∑ 𝑑𝑧
2𝜋𝑖 |𝑧−𝑎|=𝜌 𝑛=0 (𝑧 − 𝑎)𝑛+1
∞
1 𝑓(𝑧)
=∑( ∫ 𝑑𝑧)(𝑤 − 𝑎)𝑛
2𝜋𝑖
𝑛=0 ⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟⏟
|𝑧−𝑎|=𝜌
(𝑧 − 𝑎)𝑛+1
𝑐𝑛
Corollary 2.17. If 𝑓 is holomorphic in 𝑈 then its derivatives of all orders

exist and are holomorphic.
19
Proof. Taylor expansion plus theory on power series.

The next theorem is a converse to Cauchy:
Corollary 2.18 (Morera). Let 𝑈 be a domain and 𝑓 ∶ 𝑈 → C continuous.

If ∫ 𝑓(𝑧)𝑑𝑧 = 0 for every closed paths in 𝑈, then 𝑓 is holomorphic in 𝑈.
𝛾
Proof. By Proposition 2.5, 𝑓 has an anti-derivative 𝐹. 𝐹 is holomorphic and

𝐹 ′ = 𝑓 so 𝑓 is holomorphic on 𝑈.
This gives us something we promised a while ago in Proposition 2.9: we see
that 𝑓 ends up being holomorphic also on 𝑆.
Proposition 2.19 (Cauchy integral formula for derivatives). Use the same
notation as in Taylor expansion, we have
𝑛! 𝑓(𝑧)
𝑓 (𝑛) (𝑤) = ∫ 𝑑𝑧.
2𝜋𝑖 |𝑧−𝑎|=𝜌 (𝑧 − 𝑤)𝑛+1
Proof.
𝑑 𝑓(𝑧) 𝑓 ′ (𝑧) 𝑓(𝑧)
( )= −
𝑑𝑧 𝑧 − 𝑤 𝑧 − 𝑤 (𝑧 − 𝑤)2
Now integrate over |𝑧 − 𝑎| = 𝜌, since anti-derivative of 𝑑 𝑓(𝑧)
𝑑𝑧 𝑧−𝑤 exists,
𝑑 𝑓(𝑧) 𝑓 ′ (𝑧) 𝑓(𝑧)

0=∫ ( ) 𝑑𝑧 = ∫ 𝑑𝑧 − ∫ 𝑑𝑧
|𝑧−𝑎|=𝜌
𝑑𝑧 𝑧 − 𝑤 |𝑧−𝑎|=𝜌
𝑧−𝑤 |𝑧−𝑎|=𝜌
(𝑧 − 𝑤)2
Now apply Cauchy integral formula for a disc to the first term. Inductively, we
derive
𝑛! 𝑓(𝑧)
𝑓 (𝑛) (𝑤) = ∫ 𝑑𝑧.
2𝜋𝑖 |𝑧−𝑎|=𝜌 (𝑧 − 𝑤)𝑛+1
Note the subtlety between this and Taylor’s theorem: Taylor’s theorem gives
us the coefficient of the power series expansion around 𝑧 (which holds trivially
at 𝑧) while Cauchy integral formula tells us the value of derivatives of any order
at a point in the disc.
2.5 Uniform limits of holomorphic functions

We want to construct new holomorphic functions from old ones. Recall in IB
Analysis II uniform limit theorem tells us that the uniform limit of continuous
functions is continuous. We are going to adopt the same idea except one catch:
continuity is a local property so if we can produce a cover on each element
of which a given sequence of functions converges uniformly, then the resulting
function by patching everything together is continuous on the entire space. This
is a weaker requirement than uniform convergence and indeed sometimes this is
the best that we can do, illustrated by the example below.
20
Definition (Locally uniform convergence). Let 𝑈 be an open domain and

𝑓𝑛 ∶ 𝑈 → C be a sequence of functions. We say that {𝑓𝑛 } is locally uniformly
convergent in 𝑈 if for any 𝑎 ∈ 𝑈 there exists 𝑟 > 0 such that 𝐷(𝑎, 𝑟) ⊆ 𝑈 on
which {𝑓𝑛 } is uniformly convergent.
Example. Let 𝑈 = 𝐷(0, 1) and 𝑓𝑛 (𝑧) = 1−𝑧 1

𝑛 . Note that 𝑓𝑛 → 1 pointwise as
𝑛 → ∞. It is not uniformly convergent on 𝐷(0, 1) but it is locally uniformly

convergent. In fact it is uniformly convergent on any |𝑧| ≤ 𝜌 < 1.
Proposition 2.20. A sequence of functions 𝑓𝑛 ∶ 𝑈 → C is locally uniformly

convergent if and only if it converges uniformly on all compact subsets of 𝑈.
Proof. If 𝑓𝑛 → 𝑓 uniformly on all compact subsets, then given 𝑎 ∈ 𝑈 and 𝑟 > 0

such that 𝐷(𝑎, 𝑟) ⊆ 𝑈, then 𝑓𝑛 → 𝑓 uniformly on 𝐷(𝑎, 𝑟), hence 𝑓𝑛 → 𝑓 locally
uniformly.
Conversely, if 𝑓𝑛 → 𝑓 locally uniformly on 𝑈, let 𝐾 ⊆ 𝑈 be a compact set.
For each 𝑎 ∈ 𝐾 there exists 𝐷(𝑎, 𝑟) ⊆ 𝑈 on which 𝑓𝑛 → 𝑓 uniformly. As 𝐾 is
compact, there exists a finite set 𝑆 ⊆ 𝑈 such that 𝐾 ⊆ ⋃𝑎∈𝑆 𝐷(𝑎, 𝑟). Hence
𝑓𝑛 → 𝑓 converges uniformly on 𝐾.
Theorem 2.21. Let {𝑓𝑛 } be a sequence of holomorphic functions on 𝑈 which

is locally uniformly convergent. Then the limit function 𝑓 is holomorphic
and the sequence {𝑓𝑛′ } converges locally uniformly to 𝑓 ′ on 𝑈.
This is a generalisation of the corresponding statements about power series.

Proof. The key here is to use the following characterisation of holomorphicity:
a function is holomorphic on a domain if and only if its integral along any
closed path vanishes (if given by Morera, only if given by Cauchy’s Theorem for
triangles).
Let 𝐷 = 𝐷(𝑎, 𝑟) ⊆ 𝑈 be any disc. Then by Cauchy’s Theorem for triangles,
for any closed curve 𝛾 in 𝐷, ∫ 𝑓𝑛 (𝑧)𝑑𝑧 = 0. Now 𝑓𝑛 → 𝑓 uniformly on 𝛾 since
𝛾
𝛾([𝑎, 𝑏]) is compact. Since 𝑓 is continuous,
0 = ∫ 𝑓𝑛 (𝑧)𝑑𝑧 → ∫ 𝑓(𝑧)𝑑𝑧 = 0
𝛾 𝛾
so by Morera 𝑓 is holomorphic on 𝐷.
Next, by Cauchy integral formula for derivatives for any 𝑤 ∈ 𝐷(𝑎, 𝑟/2),
1 𝑓(𝑧) − 𝑓𝑛 (𝑧) 2𝜋𝑟 sup|𝑧−𝑎|=𝑟 |𝑓(𝑧) − 𝑓𝑛 (𝑧)|

|𝑓 ′ (𝑤) − 𝑓𝑛′ (𝑤)| = ∣∫ 2
∣≤ .
2𝜋 |𝑧−𝑎|=𝑟 (𝑧 − 𝑤) 2𝜋 𝑟2 /4
Since 𝑓𝑛 → 𝑓 uniformly on |𝑧 − 𝑎| = 𝑟 we see that 𝑓𝑛′ → 𝑓 ′ uniformly on

𝐷(𝑎, 𝑟/2). Done.
Example (Riemann zeta function). The Riemann zeta function is the map
∞
1
𝜁(𝑠) = ∑ 𝑠
.
𝑛=1
𝑛
21
To find where the function is defined, i.e. the series is convergent,

|𝑛𝑠 | = |𝑒𝑠 log 𝑛 | = 𝑛Re(𝑠)
so if Re(𝑠) > 1 the series is absolutely convergent. Moreover, if Re(𝑠) ≥ 𝜎 > 1,
by comparison test with ∑ 𝑛1𝜎 we have uniform convergence on {𝑠 ∶ Re(𝑠) ≥ 𝜎}.
Thus by the previous theorem 𝜁(𝑠) defines a holomorphic function on Re(𝑠) > 1.
2.6 Zeros of holomorphic functions

Consider 𝑓 ∶ 𝐷(𝑤, 𝑅) → C holomorphic and write 𝑓 as a power series
∞
𝑓(𝑧) = ∑ 𝑐𝑛 (𝑧 − 𝑤)𝑛 .
𝑛=0
If 𝑓 is not identically zero then not all 𝑐𝑛 ’s are zero. Let

𝑚 = min{𝑛 ∈ Z, 𝑛 ≥ 0 ∶ 𝑐𝑛 ≠ 0}.
Then
𝑓(𝑧) = (𝑧 − 𝑤)𝑚 𝑔(𝑧)
where 𝑔(𝑧) = ∑𝑛=𝑚 𝑐𝑛 (𝑧 − 𝑤)𝑛−𝑚 is holomorphic on 𝐷(𝑤, 𝑅) and 𝑔(𝑤) ≠ 0. If
∞
𝑚 > 0 we say that 𝑓 has a zero of order 𝑚 at 𝑧 = 𝑤. Clearly 𝑚 is the least 𝑛

such that 𝑓 (𝑛) (𝑤) ≠ 0.
Theorem 2.22 (Principle of isolated zeros). Let 𝑓 ∶ 𝐷(𝑤, 𝑅) → C be

holomorphic and not identically zero. Then there exists 0 < 𝑟 ≤ 𝑅 such that
𝑓(𝑧) ≠ 0 for 0 < |𝑧 − 𝑤| < 𝑟.
Proof. Suppose 𝑓(𝑤) ≠ 0 then by continuity there exists 𝑟 > 0 such that 𝑓(𝑧) ≠ 0
for 𝑧 ∈ 𝐷(𝑤, 𝑟).
Otherwise, 𝑓 has a zero of some order 𝑚 > 0 at 𝑧 = 𝑤. Hence 𝑓(𝑧) =
(𝑧 − 𝑤)𝑚 𝑔(𝑧) where 𝑔 is holomorphic and 𝑔(𝑤) ≠ 0. Hence there exist 𝑟 > 0
such that 𝑔 is non-zero on 𝐷(𝑤, 𝑟) and then 𝑓(𝑧) ≠ 0 for 0 < |𝑧 − 𝑤| < 𝑟.
2.7 Analytic continuation

The section discusses some results coming from the interplay of holomorphic
functions and general topology.
Theorem 2.23 (Uniqueness of analytic continuation). Let 𝐷′ ⊆ 𝐷 be

domains and 𝑓 ∶ 𝐷′ → C analytic. Then there is at most one analytic
function 𝑔 ∶ 𝐷 → C such that 𝑔(𝑧) = 𝑓(𝑧) for 𝑧 ∈ 𝐷′ .
Proof. This is really an exercise in connectedness. Let 𝑔1 , 𝑔2 ∶ 𝐷 → C be analytic
continuations of 𝑓. Then ℎ = 𝑔1 − 𝑔2 ∶ 𝐷 → C is analytic and ℎ(𝑧) = 0 on 𝐷′ .
We’ll prove that ℎ = 0 on 𝐷. Let
𝐷0 = {𝑤 ∈ 𝐷 ∶ ℎ is identically zero on some open disc 𝐷(𝑤, 𝑟)}
𝐷1 = {𝑤 ∈ 𝐷 ∶ ℎ(𝑛) (𝑤) ≠ 0 for some 𝑛 ≥ 0}
We saw last time that 𝐷 = 𝐷0 ∪ 𝐷1 , 𝐷0 ∩ 𝐷1 = ∅. Moreover, 𝐷0 and 𝐷1 are
open. But 𝐷0 ⊇ 𝐷′ so is non-zero. By (path-)connectedness of 𝐷, 𝐷0 = 𝐷.
22
Corollary 2.24 (Identity principle). Let 𝑓, 𝑔 ∶ 𝐷 → C be analytic functions

on a domain. If
𝑆 = {𝑧 ∈ 𝐷 ∶ 𝑓(𝑧) = 𝑔(𝑧)}
contains a non-isolated point then 𝑓 = 𝑔 on 𝐷.
Proof. Let 𝑤 be such a non-isolated point. 𝑓 − 𝑔 is holomorphic in 𝐷 and

vanishes on 𝑆 so it has a non-isolated zero. Thus by Theorem 2.22 𝑓 − 𝑔 vanishes
on an open disc with centre 𝑤 and by the previous theorem 𝑓 = 𝑔 on 𝐷.
Remark. Given 𝑓 ∶ 𝐷′ → C analytic and 𝐷 ⊇ 𝐷′ , extending 𝑓 could be a tough
problem if not impossible.
Example.
1. 𝑓(𝑧) = ∑𝑛=0 𝑧𝑛 defines an analytic function on 𝐷(0, 1) and extends to

∞
1−𝑧 on C \ {1}.
1
2. ∑𝑛=1 𝑧 𝑛! defines an analytic function on 𝐷(0, 1) but cannot be extended

∞
analytically to a domain larger than 𝐷(0, 1). |𝑧| = 1 is called the natural
boundary. See example sheet 2.
3. 𝜁(𝑠) = ∑∞ 1
𝑛=1 𝑛𝑠
defines an analytic function in {𝑠 ∶ Re 𝑠 > 1} and has an
analytic continuation to C \ {1} but it takes effort to prove!
Corollary 2.25 (Global maximum principle). Let 𝑈 ⊆ C be a bounded

domain. Let 𝑈 be the closure of 𝑈 and 𝑓 ∶ 𝑈 → C continuous and holomorphic
in 𝑈. Then |𝑓| attains a maximum on 𝑈 \ 𝑈.
Proof. Recall that if 𝑈 is bounded then 𝑈 is bounded and closed so |𝑓| attains a
maximum on 𝑈. Suppose the maximum is achieved at 𝑎 ∈ 𝑈. Then Theorem 2.15
says that 𝑓 is constant on 𝐷(𝑎, 𝑟) for some 𝑟 > 0. Thus 𝑓 is constant on 𝑈 and
hence on 𝑈.
23
3 Complex Integration II
In this section we are going to prove the general form of Cauchy’s theorem, by
formalising “number of loops” of a path around a point (which is as an aside,
easy to do for 𝐶 1 curves but difficult in the continuous case), and subsequently
observe the interplay between holomorphic functions and the global topology.
3.1 Winding number

Suppose 𝛾 ∶ [𝑎, 𝑏] → C is a closed curve (which is assumed to be 𝐶 1 in this
course) and 𝑤 ∈ C which is not in 𝛾([𝑎, 𝑏]). We would like to make sense of the
intuitive notion “number of times 𝛾 winds around 𝑤”.
Definition (Choice of argument). A continuous choice of argument on 𝛾 is

a continuous map 𝜃 ∶ [𝑎, 𝑏] → R such that 𝛾(𝑡) = 𝑤 + 𝑟(𝑡)𝑒𝑖𝜃(𝑡) for 𝑡 ∈ [𝑎, 𝑏],
where 𝑟(𝑡) = |𝛾(𝑡) − 𝑤| > 0.
Definition (Winding number/index). We define
𝜃(𝑏) − 𝜃(𝑎)
𝐼(𝛾, 𝑤) =
2𝜋
to be the index of 𝛾 with respect to 𝑤 or the winding number of 𝛾 with
respect to 𝑤.
We first show it is well-defined. If 𝜃 and 𝜃′ both satisfy the equation in the

definition above and are both continuous, then their difference is a continuous
function with values in 2𝜋Z, so it is constant. Hence if such 𝜃 exists 𝐼(𝛾, 𝑤) is
well-defined.
Next we show that a choice of argument does exist:
Lemma 3.1. Let 𝛾 ∶ [𝑎, 𝑏] → C \ {𝑤} be a closed curve. Then there exists
𝜃 ∶ [𝑎, 𝑏] → R continuous (in fact piecewise 𝐶 1 ) such that
𝛾(𝑡) = 𝑤 + 𝑟(𝑡)𝑒𝑖𝜃(𝑡)
for all 𝑡 ∈ [𝑎, 𝑏] where 𝑟(𝑡) = |𝛾(𝑡) − 𝑤|.
Proof. Wlog assume 𝛾 is 𝐶 1 . Define

𝑡
𝛾 ′ (𝑠)
ℎ(𝑡) = ∫ 𝑑𝑠.
𝑎
𝛾(𝑠) − 𝑤
′
From IA Analysis I ℎ′ (𝑡) = 𝛾 (𝑡)
𝛾(𝑡)−𝑤 so
𝑑
((𝛾(𝑡) − 𝑤)𝑒−ℎ(𝑡) ) = 𝛾 ′ (𝑡)𝑒−ℎ(𝑡) + (𝛾(𝑡) − 𝑤)𝑒−ℎ(𝑡) (−ℎ′ (𝑡)) = 0
𝑑𝑡
so 𝛾(𝑡) = 𝑤 + (𝛾(𝑎) − 𝑤)𝑒ℎ(𝑡) so if we set 𝜃(𝑡) = arg(𝛾(𝑎) − 𝑤) + Im ℎ(𝑡) then it
is 𝐶 1 .
24
Remark. A continuous 𝜃 exists assuming only 𝛾 is continuous (but it is a bit

harder).
Lemma 3.2. Let 𝛾 ∶ [𝑎, 𝑏] → C \ {𝑤} be a closed curve. Then
1 𝑑𝑧
𝐼(𝛾, 𝑤) = ∫ .
2𝜋𝑖 𝛾 𝑧 − 𝑤
Proof. Write 𝛾(𝑡) = 𝑤 + 𝑟(𝑡)𝑒𝑖𝜃(𝑡) . Then

𝑏
𝑑𝑧 𝛾 ′ (𝑡)
∫ =∫ 𝑑𝑡
𝛾
𝑧−𝑤 𝑎
𝛾(𝑡) − 𝑤
𝑏
𝑟′
=∫ ( + 𝑖𝜃′ )𝑑𝑡
𝑎
𝑟
𝑏
= log 𝑟(𝑡) + 𝑖𝜃(𝑡)∣
𝑎
= 𝑖(𝜃(𝑏) − 𝜃(𝑎))
= 2𝜋𝑖𝐼(𝛾, 𝑤)
Lemma 3.3.
1. 𝐼(𝛾, 𝑤) is constant on each path-component of C \ 𝛾([𝑎, 𝑏]).
2. If 𝑤 is in the unique unbounded component of C\𝛾([𝑎, 𝑏]) then 𝐼(𝛾, 𝑤) =
0.
Proof. We know 𝐼(𝛾, ⋅) ∶ C \ 𝛾([𝑎, 𝑏]) → Z takes value in Z which is a discrete
space so we would like to show it is continuous, which then implies that 𝐼(𝛾, ⋅)
is constant on each path-connected component.
Take 𝐷(𝑤, 𝑟) such that 𝐷(𝑤, 𝑟) ⊆ C \ 𝛾([𝑎, 𝑏]). Then for all ℎ ∈ 𝐷(𝑤, 𝑟),
1 1 1
|𝐼(𝛾, 𝑤 + ℎ) − 𝐼(𝛾, 𝑤)| = ∣∫ ( − ) 𝑑𝑧∣
2𝜋 𝛾 𝑧 − 𝑤 − ℎ 𝑧 − 𝑤
1 ℎ
= ∣∫ 𝑑𝑧∣
2𝜋 𝛾 (𝑧 − 𝑤 − ℎ)(𝑧 − 𝑤)
|ℎ| 𝑑𝑧
= ∣∫ ∣
2𝜋 𝛾 (𝑧 − 𝑤 − ℎ)(𝑧 − 𝑤)
|ℎ|
≤ ℓ𝑀
2𝜋
→0
as ℎ → 0 where 𝑀 = sup |𝑧−𝑤−ℎ||𝑧−𝑤|

1
exists by continuity on a compact set.
For the second part, let
𝑤 ∈ {𝑧 ∈ C ∶ |𝑧| > 2 max |𝑢|}.

𝑢∈𝛾([𝑎,𝑏])
25
Then
1 𝑑𝑧 1 ℓ
|𝐼(𝛾, 𝑤)| = ∣∫ ∣≤ →0
2𝜋 𝛾 𝑧 − 𝑤 2𝜋 |𝑤|/2
as |𝑤| → ∞.
3.2 General form of Cauchy’s theorem

The winding number 𝐼(𝛾, 𝑤) can be seen as a function taking two arguments.
We have shown that 𝐼 is continuous with resepct to 𝑤. What if we perturb 𝛾
a little bit? Hopefully the winding number should also be continuous in this
case. It turns out to be true, and an immediate corollary of which is that we can
generalise Cauchy’s theorem to a much larger class of closed curves. But before
that, we have to define what it means for one curve to change continuously to
another.
Definition (Homotopy). Let 𝜙, 𝜓 ∶ [𝑎, 𝑏] → 𝑈 be two piecewise 𝐶 1 closed

paths. A homotopy from 𝜙 to 𝜓 is a map 𝐹 ∶ [0, 1] × [𝑎, 𝑏] → 𝑈 such that
• 𝐹 is continuous,
• 𝐹 |{0}×[𝑎,𝑏] = 𝜙, 𝐹 |{1}×[𝑎,𝑏] = 𝜓,
• for all 𝑠 ∈ [0, 1], 𝐹𝑠 (𝑡) = 𝐹 (𝑠, 𝑡) ∶ [𝑎, 𝑏] → 𝑈 is a closed piecewise 𝐶 1

curve.
Equipped with this definition, we can now make rigorous the notion of simple-
connectedness, without resorting to some sloppy definition using “absence of a
hole”.
Definition (Simply connected). A domain 𝑈 is simply connected if every

piecewise 𝐶 1 closed path is homotopic to a constant path.
Example. A star-domain is simply-connected. Suppose 𝑝 ∈ 𝑈 is such that for

all 𝑤 ∈ 𝑈 the straight segment from 𝑝 to 𝑤 lies in 𝑈. Given a path 𝛾,
𝐹 (𝑠, 𝑡) = 𝑠𝛾(𝑡) + (1 − 𝑠)𝑝
is a homotopy.
Definition (Elementary deformation). Let 𝜙, 𝜓 ∶ [0, 1] → 𝑈 be closed

piecewise 𝐶 1 curves. We say that 𝜓 is an elementary deformation of 𝜙
if there exists 0 = 𝑥0 < 𝑥1 < ⋯ < 𝑥𝑛 = 1 and convex open sets 𝐶1 , … , 𝐶𝑛 in
𝑈 such that for all 𝑡 ∈ [𝑥𝑖−1 , 𝑥𝑖 ], 𝜙(𝑡), 𝜓(𝑡) ∈ 𝐶𝑖 .
Next up is a key technical proposition:
Proposition 3.4. If 𝜙 and 𝜓 are homotopic closed paths in the domain

𝑈, then there are 𝜙 = 𝜙0 , 𝜙1 , … , 𝜙𝑁 = 𝜓 such that 𝜙𝑡+1 is an elementary
deformation of 𝜙𝑡 .
26
This says that given two homotopic paths, we can find finitely many inter-
mediate paths, each of which is an elementary deformation of the previous one.
Elementary deformation localises paths to convex open sets on which convex
Cauchy’s theorem applies, ergo allowing us to deduce the general Cauchy’s
theorem.
Proof. This is an exercise in uniform continuity. Let 𝐹 be the homotopy between
𝜙 and 𝜓. Im 𝐹 is a compact set and C \ 𝑈 is closed so
dist(Im 𝐹 , C \ 𝑈 ) = 𝜀 > 0.
For this 𝜀, 𝐷(𝐹 (𝑠, 𝑡), 𝜀) ⊆ 𝑈 for all (𝑠, 𝑡). Since 𝐹 is uniformly continuous, there
exists 𝛿 > 0 such that
‖(𝑠′ , 𝑡′ ) − (𝑠, 𝑡)‖ < 𝛿 ⟹ |𝐹 (𝑠′ , 𝑡′ ) − 𝐹 (𝑠, 𝑡)| < 𝜀. (∗)
Pick 𝑛 such that 1+(𝑏−𝑎)

𝑛 < 𝛿 and let 𝑥𝑗 = 𝑎 + 𝑏−𝑎
𝑛 𝑗 for 0 ≤ 𝑗 ≤ 𝑛. Let
𝜙𝑖 = 𝐹 |{ 𝑖 }×[𝑎,𝑏]
𝑛
and
𝑖
𝐶𝑖𝑗 = 𝐷(𝐹 ( , 𝑥𝑗 ), 𝜀) ⊆ 𝑈 .
𝑛
By (∗) we check that if 𝑠 ∈ [ 𝑖−1
𝑛 𝑛, 𝑖
], 𝑡 ∈ [𝑥𝑗−1 , 𝑥𝑗 ] then 𝐹 (𝑠, 𝑡) ⊆ 𝐶𝑖𝑗 . Then 𝜙𝑖 is
an elementary deformation of 𝜙𝑖−1 .
Theorem 3.5 (Homotopy form of Cauchy’s theorem). Let 𝑓 ∶ 𝑈 → C be

holomorphic on a domain 𝑈. If 𝜙 and 𝜓 are homotopic closed paths, then
∫ 𝑓(𝑧)𝑑𝑧 = ∫ 𝑓(𝑧)𝑑𝑧.
𝜙 𝜓
In particular if 𝜙 is homotopic to a constant then
∫ 𝑓(𝑧)𝑑𝑧 = 0.
𝜙
Proof. Suppose first 𝜓 is an elementary deformation of 𝜙. Let 𝜙𝑖 = 𝜙|[𝑥𝑖−1 ,𝑥𝑖 ] , 𝜓𝑖 =

𝜓|[𝑥𝑖−1 ,𝑥𝑖 ] ⊆ 𝐶𝑖 where 𝐶𝑖 ⊆ 𝑈 is convex. Let 𝛾𝑖 be the straight segment connecting
𝜙(𝑥𝑖 ) to 𝜓(𝑥𝑖 ), which is contained in 𝐶𝑖 by convexity. Now the concatenation
𝜙𝑖 ⋅ 𝛾𝑖 ⋅ (−𝜓𝑖 ) ⋅ (−𝛾𝑖−1 )
lies in 𝐶𝑖 and by convex Cauchy, the integral along this closed path vanishes.
Sum over 1 ≤ 𝑖 ≤ 𝑛, we see that integrals over the 𝛾𝑖 ’s cancel and thus
∫ 𝑓(𝑧)𝑑𝑧 = ∫ 𝑓(𝑧)𝑑𝑧.
𝜙 𝜓
Now invoke the technical proposition.

Again this hightlights an idea that has been used throughout this course:
complex analysis alone, i.e. the study of holomorphic functions gives local results.
By using topological machinaries such as connectedness, compactness and in
this section, homotopy, one can then extend them to global results.
27
Corollary 3.6. Let 𝑈 be a simply connected domain and 𝑓 ∶ 𝑈 → C

holomorphic. Then
∫ 𝑓(𝑧)𝑑𝑧 = 0
𝛾
for all closed paths 𝛾 in 𝑈.
Proof. Trivial.
Remark. Let 𝜙, 𝜓 ∶ [𝑎, 𝑏] → 𝑈 be two closed homotopic curves in a domain
𝑈. If we take 𝑤 ∈ C \ 𝑈, then the function 𝑧−𝑤
1
is holomorphic in 𝑈 and by
Homotopy form of Cauchy’s theorem, 𝐼(𝜙, 𝑤) = 𝐼(𝜓, 𝑤).
28
4 Laurent expansion, Singularities and the Residue theorem
4 Laurent expansion, Singularities and the Residue

theorem
4.1 Laurent expansion
Theorem 4.1. Let 𝑓 be a holomorphic function on an annulus
𝐴 = {𝑧 ∈ C ∶ 𝑟 < |𝑧 − 𝑎| < 𝑅}
where 0 ≤ 𝑟 < 𝑅 ≤ ∞. Then

1. 𝑓 has a unique convergent expansion in 𝐴 of the form
∞
𝑓(𝑧) = ∑ 𝑐𝑛 (𝑧 − 𝑎)𝑛 , (∗)
𝑛=−∞
2. for any 𝜌 ∈ (𝑟, 𝑅), the coefficient 𝑐𝑛 is given by
1 𝑓(𝑧)
𝑐𝑛 = ∫ 𝑑𝑧,
2𝜋𝑖 |𝑧−𝑎|=𝜌 (𝑧 − 𝑎)𝑛+1
3. if 𝑟 < 𝜌′ ≤ 𝜌 < 𝑅, then the series converges uniformly on the set
{𝑧 ∈ C ∶ 𝜌′ ≤ |𝑧 − 𝑎| ≤ 𝜌}
and thus locally uniformly on 𝐴.
Proof. As in the proof of Cauchy integral formula, consider the function

𝑓(𝑧)−𝑓(𝑤)
𝑧−𝑤 if 𝑧 ≠ 𝑤
𝑔(𝑧) = { ′
𝑓 (𝑤) if 𝑧 = 𝑤
𝑔 is holomorphic on 𝐴 by a corollary of Morera. Choose
𝑟 < 𝜌2 < |𝑤 − 𝑎| < 𝜌1 < 𝑅
and consider circles
𝐶1 ∶ |𝑧 − 𝑎| = 𝜌1
𝐶2 ∶ |𝑧 − 𝑎| = 𝜌2
which are homotopic on 𝐴. By Homotopy form of Cauchy’s theorem,
∫ 𝑔(𝑧)𝑑𝑧 = ∫ 𝑔(𝑧)𝑑𝑧
𝐶1 𝐶2
so
𝑓(𝑧) 𝑑𝑧 𝑓(𝑧) 𝑑𝑧
∫ 𝑑𝑧 − 𝑓(𝑤) ∫ =∫ 𝑑𝑧 − 𝑓(𝑤) ∫
𝑧−𝑤 𝑧−𝑤 𝑧−𝑤 𝑧−𝑤
𝐶1 ⏟𝐶
⏟2⏟⏟⏟ 𝐶2 ⏟𝐶
⏟2⏟⏟⏟
2𝜋𝑖𝐼(𝐶1 ,𝑤)=2𝜋𝑖 2𝜋𝑖𝐼(𝐶2 ,𝑤)=0
29
so
1 𝑓(𝑧) 𝑓(𝑧)
𝑓(𝑤) = (∫ 𝑑𝑧 − ∫ 𝑑𝑧) = 𝑓1 (𝑤) + 𝑓2 (𝑤).
2𝜋𝑖 𝐶
𝑧−𝑤 𝐶
𝑧−𝑤
1 2
To deal with 𝑓1 , we proceed exactly as in the proof of the Taylor series:

∞
1 (𝑤 − 𝑎)𝑛
=∑
𝑧 − 𝑤 𝑛=0 (𝑧 − 𝑎)𝑛+1
uniformly as ∣ 𝑤−𝑎
𝑧−𝑎 ∣ < 1. Thus use uniform convergence to get
∞
𝑓1 (𝑤) = ∑ 𝑐𝑛 (𝑤 − 𝑎)𝑛
𝑛=0
where
1 𝑓(𝑧)
𝑐𝑛 = ∫ 𝑑𝑧.
2𝜋𝑖 𝐶 (𝑧 − 𝑎)𝑛+1
1
To deal with 𝑓2 we use the same trick with the role of 𝑧 and 𝑤 reversed:
∞
1 1/(𝑤 − 𝑎) (𝑧 − 𝑎)𝑚−1
− = 𝑧−𝑎 = ∑
𝑧−𝑤 1 − 𝑤−𝑎 𝑚=1
(𝑤 − 𝑎)𝑚
uniformly for 𝑧 ∈ 𝐶2 as ∣ 𝑤−𝑎

𝑧−𝑎
∣ < 1. Thus
∞
𝑓2 (𝑤) = ∑ 𝑑𝑚 (𝑤 − 𝑎)−𝑚
𝑚=1
where
1 𝑓(𝑧)
𝑑𝑚 = ∫ 𝑑𝑧.
2𝜋𝑖 𝐶 (𝑧 − 𝑎)−𝑚+1
2
Write 𝑛 = −𝑚, we get (∗).

Suppose we have
∞
𝑓(𝑧) = ∑ 𝑐𝑛 (𝑧 − 𝑎)𝑛
𝑛=−∞
on 𝐴 and let 𝑟 < 𝜌 ≤ 𝜌 < 𝑅, then ∑𝑛=0 𝑐𝑛 (𝑧 − 𝑎)𝑛 must have radius of
′ ∞
convergence ≥ 𝑟, so converges uniformly on |𝑧−𝑎| ≤ 𝜌. Likewise set 𝑢 = (𝑧−𝑎)−1 ,

then ∑∞ 𝑐 𝑢𝑛 must have radius of convergence ≥ 𝑅
𝑛=1 −𝑛
1
, so converges uniformly
on |𝑢| ≤ 𝑅 , hence (∗) converges uniformly in 𝜌 ≤ |𝑧 − 𝑎| ≤ 𝜌.
1 ′
By uniform convergence of (∗), we can integrate term-by-term over every

closed curve in 𝐴, so
∞
𝑓(𝑧)
∫ 𝑑𝑧 = ∑ ∫ (𝑧 − 𝑎)𝑛−𝑚−1 𝑑𝑧 = 2𝜋𝑖𝑐𝑚
|𝑧−𝑎|=𝜌
(𝑧 − 𝑎)𝑚+1 𝑛=−∞ |𝑧−𝑎|=𝜌
so this implies also that the expansion is unique and completes the proof.
4.2 Isolated singularities
30
Definition (Isolated singularity). 𝑓 has an isolated singularity at 𝑎 if 𝑓 ∶

𝐷(𝑎, 𝑅) \ {𝑎} → C is holomorphic.
For such a function, we know we can write it in Laurent expansion

∞
𝑓(𝑧) = ∑ 𝑐𝑛 (𝑧 − 𝑎)𝑛 .
𝑛=−∞
There are three cases:
Definition (Removable singularity, pole & essential singularity).
1. 𝑐𝑛 = 0 for all 𝑛 < 0. In this case we just get a power series which
converges on all 𝐷(𝑎, 𝑅) and defines an analytic function on 𝐷(𝑎, 𝑅).
We say that 𝑓 has a removable singularity at 𝑎.
2. There exists 𝑘 > 0 such that 𝑐−𝑘 ≠ 0 but 𝑐𝑛 = 0 for all 𝑛 < −𝑘. We
say that 𝑓 has a pole of order 𝑘 at 𝑧 = 𝑎.
3. 𝑐𝑛 ≠ 0 for infinitely many negative 𝑛. We say that 𝑓 has an essential

singularity at 𝑧 = 𝑎.
Example.
1. 𝑓(𝑧) = sin𝑧 𝑧 has a removable singularity at 0. To see this, expand sin 𝑧
in Taylor series as usual, and get a (Laurent) series by dividing by 𝑧. As
Luarent series is unique, this is the series representation.
2. 𝑒𝑧
𝑧100and 𝑧19 have a pole at 0, by the same argument as above but replace
“power series” by “Laurent series”.
3. 𝑒1/𝑧 has an essential singularity at 0.
In IID Riemann Surfaces we will learn that poles correpsond to normal points
of a holomorphic function defined on a Riemann surface, and it is only the
essential singularities that stand out.
Proposition 4.2. 𝑓 has a removable singularity at 𝑧 = 𝑎 if and only if
lim (𝑧 − 𝑎)𝑓(𝑧) = 0.
𝑧→𝑎
Proof.
• ⟹ : Write ∞
(𝑧 − 𝑎)𝑓(𝑧) = ∑ 𝑐𝑛 (𝑧 − 𝑎)𝑛+1
𝑛=0
so it vanishes as 𝑧 = 𝑎.
• ⟸ : Consider
(𝑧 − 𝑎)2 𝑓(𝑧) if 𝑧 ≠ 𝑎
𝑔(𝑧) = {
0 if 𝑧 = 𝑎
31
We see that 𝑔 is holomorphic and 𝑔′ (𝑎) = 0 so

∞
𝑔(𝑧) = ∑ 𝑐𝑛 (𝑧 − 𝑎)𝑛
𝑛=2
so ∞
𝑓(𝑧) = ∑ 𝑐𝑛+2 (𝑧 − 𝑎)𝑛
𝑛=0
and hence 𝑓 has a removable singularity at 𝑧 = 𝑎.
Proposition 4.3. 𝑓 has a pole at 𝑧 = 𝑎 if and only if |𝑓(𝑧)| → ∞ as 𝑧 → 𝑎.

Moreover, TFAE
1. 𝑓 has a pole of order 𝑘 at 𝑧 = 𝑎,
2. 𝑓(𝑧) = (𝑧 − 𝑎)−𝑘 𝑔(𝑧) where 𝑔 ∶ 𝐷(𝑎, 𝑅) → C is holomorphic and

𝑔(𝑎) ≠ 0.
3. 𝑓(𝑧) = 1
ℎ(𝑧) where ℎ is holomorphic at 𝑧 = 𝑎 with 𝑎 a zero of order 𝑘.
Proof.
• 1 ⟹ 2: Using Laurent expansion for 𝑓 and multiply by (𝑧 − 𝑎)𝑘 , we get
a power series with non-zero constant term defining 𝑔.
• 2 ⟹ 1: The Laurent series for 𝑓 is obtained by (𝑧 − 𝑎)−𝑘 times the Taylor
series for 𝑔.
• 2 ⟺ 3: 𝑔 is holomorphic at 𝑧 = 𝑎 with 𝑔(𝑎) ≠ 0 if and only if 1
𝑔 is
holomorphic at 𝑧 = 𝑎.
Finally suppose 𝑓 has a pole at 𝑧 = 𝑎. Then by 2 |𝑓| → ∞ as 𝑧 → 𝑎.
Conversely if |𝑓| → ∞ as 𝑧 → 𝑎, then for some 𝑟 > 0, 𝑓 is non-zero for
0 < |𝑧 − 𝑎| < 𝑟. Therefore 1𝑓 is holomorphic for 0 < |𝑧 − 𝑎| < 𝑟 and 1𝑓 → 0 as
𝑧 → 𝑎. By the previous proposition 1𝑓 has a removable singularity at 𝑧 = 𝑎.
Thus there is a holomorphic ℎ on 𝐷(𝑎, 𝑟) with ℎ1 = 𝑓 for 0 < |𝑧 − 𝑎| < 𝑟. As
𝑓 → 0 as 𝑧 → 𝑎, ℎ has a zero at 𝑧 = 𝑎.
1
Corollary 4.4. 𝑓 has an essential singularity at 𝑧 = 𝑎 if and only |𝑓| has

no limit (in [0, ∞]) as 𝑧 → 𝑎.
Proof. Direct consequence of the previous two propositions.

Actually we can say a bit more about how “wild” an essential singuarlity can
be:
Theorem 4.5 (Casorati-Weierstrass). Let 𝑓 ∶ 𝐷(𝑎, 𝑅) \ {𝑎} → C be holo-

morphic with an essential singularity at 𝑧 = 𝑎. Then for any 𝑤 ∈ C and
𝑟 > 0, 𝜀 > 0, there exists 𝑧 such that 0 < |𝑧 − 𝑎| < 𝑟 and |𝑓(𝑧) − 𝑤| < 𝜀.
32
Proof. See example sheet. A long hint: if there eixsts 𝜀 > 0, 𝑟 > 0 such that
𝑓(𝐷(𝑎, 𝑟) \ {0}) ⊆ C \ 𝐷(𝑤, 𝜀),
consider 𝑔(𝑧) = 𝑓(𝑧)−𝑤

1
. We have |𝑔(𝑧)| < 1𝜀 . As it is bounded it must have a
removable singularity at 𝑎, so does 𝑓. Absurd.
With Laurent series defined and shown to be unique and uniformly continuous,
we can integrate a holomorphic function term-by-term. Just as before, along a
closed paths, only the 𝑧−𝑎
1
term survives. We thus define
Definition (Residue). Let 𝑓 ∶ 𝐷(𝑎, 𝑅) \ {𝑎} → C be holomorphic with

Laurent expansion ∑𝑛=−∞ 𝑐𝑛 (𝑧 − 𝑎)𝑛 . The residue of 𝑓 at 𝑧 = 𝑎 is
∞
Res 𝑓 = 𝑐−1 .
𝑧=𝑎
Definition (Principal part). The principal part of 𝑓 at 𝑧 = 𝑎 is the series

−1
∑ 𝑐𝑛 (𝑧 − 𝑎)𝑛
𝑛=−∞
holomorphic in C \ {𝑎}.
Definition (Meromorphic). If 𝐷 is a domain and 𝑆 ⊆ 𝐷 is a set of isolated

points in 𝐷, then a holomorphic function 𝑓 ∶ 𝐷 \ 𝑆 → C with at worst poles
at 𝐷 is called meromorphic.
Remark. As hinted before, a meromorphic function 𝑓 ∶ 𝐷 \ 𝑆 → C can be

extended to a holomorphic function 𝑓 ∶ 𝐷 → C∞ = C ∪ {∞}, sending poles to
∞, between Riemann surfaces.
Proposition 4.6. If 𝛾 is a closed curve in 𝐷(𝑎, 𝑅) \ {𝑎} and 𝑓 ∶ 𝐷(𝑎, 𝑟) \

{𝑎} → C holomorphic then
∫ 𝑓(𝑧)𝑑𝑧 = 2𝜋𝑖𝐼(𝛾, 𝑎) Res 𝑓.

𝑧=𝑎
𝛾
Proof. Using uniform convergence of the Laurent expansion, we can integrate

term-by-term and thus reduce the problem of computing ∫ (𝑧 − 𝑎)𝑛 𝑑𝑧. Since
𝛾
(𝑧−𝑎)𝑛+1
𝑛+1 is an anti-derivative of (𝑧 − 𝑎)𝑛 for 𝑛 ≠ −1, the only term that survives
is
𝑑𝑧
𝑐−1 ∫ = 2𝜋𝑖𝐼(𝛾, 𝑎) Res 𝑓.
𝛾
𝑧−𝑎 𝑧=𝑎
33
Theorem 4.7 (Cauchy’s residue theorem). Let 𝑈 be a simply connected

domain and {𝑧𝑖 }𝑘𝑖=1 be a finite set of points in 𝑈. Let 𝑓 ∶ 𝑈 \ {𝑧𝑖 }𝑘𝑖=1 → C be
holomorphic and let 𝛾 be a closed curve not passing through the 𝑧𝑖 ’s. Then
𝑘
∫ 𝑓(𝑧)𝑑𝑧 = 2𝜋𝑖 ∑ 𝐼(𝛾, 𝑧𝑖 ) Res 𝑓.
𝑧=𝑧𝑖
𝛾 𝑖=1
Proof. The idea is to write the Laurent expansion at a singular point as the sum
of the “problematic part”, i.e. the principal part, and the “good part”, i.e. the
power series part, and apply the previous proposition and Cauchy’s theorem
respectively.
At 𝑧𝑖 , 𝑓(𝑧) = ∑∞ 𝑐𝑖 (𝑧 − 𝑧𝑖 )𝑛 with principal part 𝑔𝑖 (𝑧) = ∑−1
𝑛=−∞ 𝑛
𝑐𝑖 (𝑧 −
𝑛=−∞ 𝑛
𝑧𝑖 )𝑛 . Recall that 𝑔𝑖 defines a holomorphic function on C \ {𝑧𝑖 } and hence
on 𝑈 \ {𝑧𝑖 }. Thus 𝑓 − ∑𝑘𝑖=1 𝑔𝑖 is holomorphic in 𝑈 \ {𝑧𝑖 }𝑘𝑖=1 with removable
singularties at 𝑧𝑖 . Thus by Homotopy form of Cauchy’s theorem,
𝑘
∫(𝑓(𝑧) − ∑ 𝑔𝑖 (𝑧))𝑑𝑧 = 0.
𝛾 𝑖=1
By the proposition above,
∫ 𝑔𝑖 (𝑧)𝑑𝑧 = 2𝜋𝑖𝐼(𝛾, 𝑧𝑖 ) Res 𝑔𝑖 = 2𝜋𝑖𝐼(𝛾, 𝑧𝑖 ) Res 𝑓.

𝑧=𝑧𝑖 𝑧=𝑧𝑖
𝛾
4.3 Application and techniques of integration

In this section we will discuss how to actually use Cauchy’s residue theorem to
compute integrals, which often involve simple geometric curves such as semicircles
and rectangles. We will also solve some trickey real integral by clever use of
Cauchy’s residue theorem.
Firstly, the following observations are useful for computing residues:
1. if 𝑓 has a simple pole (i.e. order 1) at 𝑧 = 𝑎 then the Laurent expansion is
𝑐−1
𝑓(𝑧) = 𝑧−𝑎 + 𝑐0 + ⋯ so
Res 𝑓 = lim (𝑧 − 𝑎)𝑓(𝑧).

𝑧=𝑎 𝑧→𝑎
2. if 𝑓 = ℎ𝑔 where 𝑔 and ℎ are holomorphic at 𝑧 = 𝑎, 𝑔(𝑎) ≠ 0 and ℎ has a

simple zero at 𝑎, then
(𝑧 − 𝑎)𝑔(𝑧) 𝑔(𝑎)
Res 𝑓 = lim = ′ .
𝑧=𝑎 𝑧→𝑎 ℎ(𝑧) ℎ (𝑎)
3. if 𝑓(𝑧) = (𝑧 − 𝑎)−𝑘 𝑔(𝑧) with 𝑔 holomorphic, then
𝑔(𝑘−1) (𝑎)
Res 𝑓 = coefficient of (𝑧 − 𝑎)𝑘−1 in the Taylor series of 𝑔 = .
𝑧=𝑎 (𝑘 − 1)!
34
The main “skill” is to choose the right integrand and the right path. Somes
lemma also help us deform paths into more amenable forms.
Example. Consider
∞
cos 𝑥
∫ 𝑑𝑥.
−∞
1 + 𝑥 + 𝑥2
Let 𝑓(𝑧) = 𝑒𝑖𝑧
1+𝑧+𝑧2and 𝛾 = 𝛾𝑅 + 𝛾0 where 𝛾𝑅 is the semicircle in the upper half
plane around the origin of radius 𝑅. 𝑓 has a simple pole at 𝑤 = 𝑒2𝜋𝑖/3 with
index 𝐼(𝛾, 𝑤) = 1. By Cauchy’s residue theorem,
∫ 𝑓(𝑧)𝑑𝑧 = 2𝜋𝑖 Res 𝑓,

𝑧=𝑤
𝛾𝑅
where the residue can be computed using rules from last time.
Let 𝑧 = 𝑅𝑒𝑖𝑡 where 𝑡 ∈ [0, 2𝜋]. Then
𝜋
∣∫ 𝑓(𝑧)𝑑𝑧∣ = ∣∫ 𝑓(𝑅𝑒𝑖𝑡 )𝑅𝑖𝑒𝑖𝑡 𝑑𝑡∣
𝛾𝑅 0
𝜋
≤ 𝑅 ∫ |𝑓(𝑅𝑒𝑖𝑡 )|𝑑𝑡
0
≤1
⏞
|𝑒−𝑅 sin 𝑡 |
𝜋
= 𝑅∫ 𝑑𝑡
|1 + 𝑅𝑒𝑖𝑡 + 𝑅2 𝑒2𝑖𝑡 |
0 ⏟⏟⏟⏟⏟⏟⏟⏟⏟
≥𝑅2 −𝑅−1
𝑅𝜋
≤ 2
𝑅 −𝑅−1
→0
as 𝑅 → ∞. Therefore
∞
cos 𝑥 2𝜋 1 −√3/2
∫ 𝑑𝑥 = Re(2𝜋𝑖 Res 𝑓) = √ cos 𝑒 .
−∞
1 + 𝑥 + 𝑥2 𝑧=𝑤 3 2
The key idea is that the integral along the semicircle 𝛾𝑅 vanishes as 𝑅 → ∞.
Lemma 4.8. Let 𝑓 be a holomorphic function on 𝐷(𝑎, 𝑟) \ {𝑎} with a simple

pole at 𝑎. If 0 < 𝜀 < 𝑟 and let 𝛾𝜀 ∶ [𝛼, 𝛽] → C, 𝑡 ↦ 𝑎 + 𝜀𝑒𝑖𝑡 . Then
lim ∫ 𝑓(𝑧)𝑑𝑧 = (𝛽 − 𝛼)𝑖 Res 𝑓.

𝜀→0 𝑧=𝑎
𝛾𝜀
This allows us to “jump” over a pole.

Proof. As 𝑓 has a simple pole, write
𝑐
𝑓(𝑧) = + 𝑔(𝑧)
𝑧−𝑎
where 𝑔(𝑧) is holomorphic on 𝐷(𝑎, 𝑟). Thus 𝑐 = Res𝑧=𝑎 𝑓. Then
∣∫ 𝑔(𝑧)𝑑𝑧∣ ≤ (𝛽 − 𝛼)𝜀 sup |𝑔(𝑧)| → 0

𝛾𝜀 𝑧∈𝛾𝜀
35
as 𝜀 → 0 since 𝑔 is bounded on a neighbourhood of 𝑎 by continuity. Then

𝛽
𝑐 𝑖𝜀𝑒𝑖𝑡
∫ 𝑑𝑧 = 𝑐 ∫ 𝑑𝑡 = (𝛽 − 𝛼)𝑖𝑐.
𝛾𝜀
𝑧−𝑎 𝛼
𝜀𝑒𝑖𝑡
Lemma 4.9 (Jordan). If 𝑓 is holomorphic on {|𝑧| > 0} and 𝑧𝑓(𝑧) is bounded

for large |𝑧|, then for all 𝛼 > 0,
∫ 𝑓(𝑧)𝑒𝑖𝛼𝑧 𝑑𝑧 → 0
𝛾𝑅
as 𝑅 → ∞ where 𝛾𝑅 (𝑡) = 𝑅𝑒𝑖𝑡 , 𝑡 ∈ [0, 𝜋].
This comes very handy when one computes Fourier transforms. Also this
makes our first example almost trivial.
Proof. We know |𝑓(𝑧)| ≤ |𝑧|𝑐
for large |𝑧|. Observe that on [0, 𝜋2 ] the function
sin 𝑡
𝑡 is decreasing. Thus sin 𝑡 ≥ 𝜋 for 𝑡 ∈ [0, 2 ]. Then on the path 𝑧 = 𝑅𝑒 ,
2𝑡 𝜋 𝑖𝑡
−𝑅𝛼2𝑡 𝜋
𝑒 𝜋 0≤𝑡≤
|𝑒𝑖𝛼𝑧 | = 𝑒−𝑅𝛼 sin 𝑡 ≤ { −𝑅𝛼2𝑡′
2
𝜋
𝑒 𝜋 0 ≤ 𝑡′ = 𝜋 − 𝑡 ≤ 2
so
𝜋/2 𝜋/2
𝑖𝑡 2𝑅𝑡𝛼
∣∫ 𝑒𝑖𝛼𝑅𝑒 𝑓(𝑅𝑒𝑖𝑡 )𝑖𝑅𝑒𝑖𝑡 𝑑𝑡∣ ≤ ∫ 𝑒− 𝜋 𝑐𝑑𝑡
0 0
𝑐𝜋
= (1 − 𝑒−𝑅𝛼 )
2𝑅𝛼
→0
as 𝑅 → ∞.
Example. Consider
∞
sin 𝑥
∫ 𝑑𝑥.
0
𝑥
Let 𝑓(𝑧) = 𝑒𝑧 . A naïve choice of semicircular path would pass through the pole
𝑖𝑧
at 𝑧 = 0. Instead consider the semicircular path with a cutout 𝛾𝜀 at the origin.

This path contains no singularity so has integral zero. By Jordan, ∫ 𝑓(𝑧)𝑑𝑧 → 0
𝛾𝑅
as 𝑅 → ∞. Also by the previous lemma,
∫ 𝑓(𝑧)𝑑𝑧 → −𝜋𝑖 Res 𝑓 = −𝜋𝑖.

𝑧=0
𝛾𝜀
Putting everything together,

∞
sin 𝑥 𝜋
∫ 𝑑𝑥 = .
0
𝑥 2
36
Example. Consider
𝜋/2
𝑑𝑡
𝐼 =∫ .
0 1 + sin2 𝑡
Note 𝐼 = 1 2𝜋
4 ∫0
𝑑𝑡
1+sin2 𝑡
. On the unit circle 𝑧 = 𝑒𝑖𝑡 ,
𝑒𝑖𝑡 − 𝑒−𝑖𝑡 𝑧 − 1𝑧
sin 𝑡 = = .
2𝑖 2𝑖
Thus
1 𝑑𝑧
𝐼= ∫
4 |𝑧|=1 (1 + (𝑧−1/𝑧)2 )𝑖𝑧
−4
𝑧
= 𝑖∫ 𝑑𝑧
𝑧 4 − 6𝑧 2 + 1
|𝑧|=1 ⏟⏟⏟⏟⏟
𝑓
= 2𝜋𝑖 ⋅ 𝑖( Res√ 𝑓 + Res√ 𝑓)
𝑧=1− 2 𝑧=−1+ 2
√
𝜋 2
=
4
37
5 The Argument principle, Local degree, Open mapping theorem & Rouché’s
theorem
5 The Argument principle, Local degree, Open

mapping theorem & Rouché’s theorem
We will use Cauchy’s residue theorem to develop a better understanding of
holomorphic functions.
Proposition 5.1. Let 𝑓 have a zero (pole, respectively) of order 𝑘 at 𝑧 = 𝑎.

′
Then 𝑓𝑓(𝑧)
(𝑧)
has a simple pole at 𝑧 = 𝑎 with residue 𝑘 (−𝑘, respectively).
Proof. If 𝑓 has a zero of order 𝑘 at 𝑧 = 𝑎, then
𝑓(𝑧) = (𝑧 − 𝑎)𝑘 𝑔(𝑧)
where 𝑔(𝑧) is holomorphic and non-zero at 𝑧 = 𝑎. Then
𝑓 ′ (𝑧) 𝑘 𝑔′ (𝑧)
= + .
𝑓(𝑧) 𝑧−𝑎 𝑔(𝑧)
Hence the result. The pole case is similar.
Theorem 5.2 (Argument principle). Let 𝑈 be a simply connected domain

and suppose 𝑓 is meromorphic on 𝑈 with finitely many zeros {𝑧1 , … , 𝑧𝑘 } and
poles {𝑤1 , … , 𝑤ℓ }. Let 𝛾 be a closed curve in 𝑈 such that 𝑧𝑖 , 𝑤𝑗 ∉ 𝛾. Then
𝑘 ℓ
1 𝑓 ′ (𝑧)
∫ 𝑑𝑧 = ∑ 𝐼(𝛾, 𝑧𝑖 ) ord(𝑓) − ∑ 𝐼(𝛾, 𝑤𝑗 ) ord(𝑓)
2𝜋𝑖 𝛾 𝑓(𝑧) 𝑖=1
𝑧𝑖
𝑗=1
𝑤𝑗
where ord𝑧𝑖 𝑓 and ord𝑤𝑗 𝑓 denote the order of the zeros and poles.
Proof. Apply Cauchy’s residue theorem to 𝑓 .

𝑓′
If 𝑓 has no zero or pole at 𝑧 then
𝑓′
𝑓 has no singularity at 𝑧. Otherwise apply the previous proposition.
Remark. Let Γ = 𝑓 ∘ 𝛾, then 0 ∉ Γ. If we perform the substitution 𝑤 = 𝑓(𝑧) in

the above integral we get
1 𝑑𝑤
∫ = 𝐼(Γ, 0).
2𝜋𝑖 Γ 𝑤
Definition (Bounded domain). Let 𝑈 be a domain and 𝛾 a closed curve

in 𝑈. The curve 𝛾 bounds a domain 𝐷 if 𝐼(𝛾, 𝑤) = 1 for all 𝑤 ∈ 𝐷 and
𝐼(𝛾, 𝑤) = 0 for all 𝑤 ∉ 𝐷 ∪ 𝛾.
As a side note, we have the famous theorem (which we will definitely not
prove) which gives a precise definition of the “interior” of a closed curve:
Theorem 5.3 (Jordan curve theorem). Every simple closed curve admits
an orientation such that it bounds a domain provided 𝑈 is simply connected.
38
theorem
If 𝛾 bounds 𝐷, then we get

𝐼(Γ, 0) = 𝑁 − 𝑃
where
𝑁 = #zeros of 𝑓 in 𝐷 counted with multiplicity
𝑃 = #poles of 𝑓 in 𝐷 counted with multiplicity
Definition (Local degree). If 𝑓 is non-constant and holomorphic at 𝑧 = 𝑎

and 𝑏 = 𝑓(𝑎), we say that the local degree of 𝑓 at 𝑧 = 𝑎 is the order of zero
of 𝑓(𝑧) − 𝑏 at 𝑧 = 𝑎. We denote it by deg𝑎 𝑓, which is a positive integer.
Proposition 5.4. The local degree of 𝑓 at 𝑧 = 𝑎 equals to 𝐼(𝑓 ∘ 𝛾, 𝑓(𝑎)) for

any circle 𝛾(𝑡) = 𝑎 + 𝑟𝑒2𝜋𝑖𝑡 , 𝑡 ∈ [0, 1] of sufficiently small radius.
Proof. Given the previous discussion, just apply Argument principle to 𝑓(𝑧) −
𝑓(𝑎). As it has isolated zeros it is non-zero for 0 < |𝑧 − 𝑎| ≤ 𝑟 if 𝑟 is sufficiently
small.
Theorem 5.5 (Local mapping degree). Let 𝑓 ∶ 𝐷 → C be holomorphic and

non-constant with local degree deg𝑎 𝑓 = 𝑑 > 0. Then if 𝑟 > 0 is sufficiently
small then there exists 𝜀 > 0 such that for every 𝑤 with 0 < |𝑤 − 𝑓(𝑎)| ≤ 𝜀,
the equation
𝑓(𝑧) = 𝑤
has exactly 𝑑 distinct roots in 𝐷(𝑎, 𝑟).
Thus degree is indeed a local property. Consequently in Riemann surfaces,
any holomorphic function has a local model 𝑧 ↦ 𝑧 𝑑 .
Proof. Let 𝑏 = 𝑓(𝑎) and let 𝑟 > 0 be such that 𝑓(𝑧) − 𝑏 and 𝑓 ′ (𝑧) are both
non-zero for 0 < |𝑧 − 𝑎| ≤ 𝑟. Let 𝛾 be the circle of centre 𝑧 = 𝑎 and radius
𝑟. Then Γ = 𝑓 ∘ 𝛾 is a closed curve missing 𝑏. Choose 𝜀 > 0 such that 𝐷(𝑏, 𝜀)
does not intersect Γ. Then if 𝑤 ∈ 𝐷(𝑏, 𝜀), the number of zeros counted with
multiplicity of 𝑓(𝑧) − 𝑤 in 𝐷(𝑎, 𝑟) equals to 𝐼(Γ, 𝑤) by Argument principle. But
𝐼(Γ, 𝑤) = 𝐼(Γ, 𝑏) = 𝑑. Since 𝑟 is chosen so that 𝑓 ′ is non-zero on 𝐷(𝑎, 𝑟) \ {𝑎},
the zeros are all simple.
Corollary 5.6 (Open mapping theorem). A non-constant holomorphic

function 𝑓 ∶ 𝑈 → C maps open sets to open sets.
Proof. Let 𝑉 ⊆ 𝑈 be open. Need to prove that given 𝑎 ∈ 𝑉, there is a disc
around 𝑓(𝑎) inside 𝑓(𝑉 ). But by Local mapping degree, we can choose 𝑟, 𝜀 > 0
such that 𝐷(𝑎, 𝑟) ⊆ 𝑉 and 𝐷(𝑓(𝑎), 𝜀) ⊆ 𝑓(𝑉 ).
Remark. This gives the inverse function theorem: if 𝑓 ′ (𝑎) ≠ 0 then deg𝑎 𝑓 = 1,
and Local mapping degree says that there is a local inverse 𝑔 and Open mapping
theorem implies that 𝑔 is continuous. Moreover, 𝑔 is holomorphic by the argument
in the proof of differentiability of power series, which is the same as in IA Analysis
I.
39
theorem
Theorem 5.7 (Rouché). Let 𝛾 bound a domain 𝐷 and let 𝑓 and 𝑔 be

holomorphic in 𝑈 ⊇ 𝐷. If |𝑓| > |𝑔| on 𝛾, then 𝑓 and 𝑓 + 𝑔 have the same
number of zeros in 𝐷 counted with multiplicity.
Proof. Let ℎ = 𝑓 + 𝑔. Note that by hypothesis 𝑓 and ℎ have no zeros on 𝛾.

Consider 𝐹 = ℎ/𝑓 which is defined on 𝛾. Observe that |ℎ − 𝑓| < |𝑓| on 𝛾 so
|𝐹 − 1| < 1 on 𝛾. Then Γ = 𝐹 ∘ 𝛾 lies inside 𝐷(1, 1). 0 is in the unbounded
component of C \ Im Γ so 𝐼(Γ, 0) = 0. Thus
1 𝐹 ′ (𝑧)
0= ∫ 𝑑𝑧
2𝜋𝑖 𝛾 𝐹 (𝑧)
1 ℎ′ 𝑓′
= ∫ − 𝑑𝑧
2𝜋𝑖 𝛾 ℎ 𝑓
= 𝐼(ℎ ∘ 𝛾, 0) − 𝐼(𝑓 ∘ 𝛾, 0)
= #zeros of ℎ − #zeros of 𝑓
where the last line is by Argument principle.

Example. Let 𝑝(𝑧) = 𝑧⏟4 + 5𝑧
⏟ + 3. On |𝑧| = 2,
𝑓 𝑔
|𝑓| = |𝑧 4 | = 16
|𝑔| = |6𝑧 + 3| ≤ 6|𝑧| + 3 = 15
so |𝑓| > |𝑔|. Thus by Rouché, 𝑝 has 4 zeros inside |𝑧| < 2. Now consider
⏟ on |𝑧| = 1, verify that
4 + 3 + 6𝑧
𝑝(𝑧) = 𝑧⏟
𝑔 𝑓
|𝑓| = |6𝑧| = 6
|𝑔| = |𝑧 4 + 3| ≤ |𝑧 4 | + 3 = 4
so 𝑝 has 1 root inside |𝑧| < 1. Thus we conclude that 𝑝 has 3 roots in 1 ≤ |𝑧| < 2
and 1 root in |𝑧| < 1.
40
Index
analytic continuation, 22 isolated singularity, 31

argument principle, 38
Jordan’s lemma, 36
bounded domain, 38
Liouville’s theorem, 18
Casorati-Weierstrass theorem, 32 local degree, 39
Cauchy integral formula, 17 local mapping degree, 39
for derivatives, 20 locally uniform convergence, 21
Cauchy’s residue theorem, 34 logarithm, 7
Cauchy’s theorem, 15, 27
Cauchy-Riemann equations, 3 maximum principle
choice of argument, 24 global, 23
complex differentiable, 2 local, 19
conformal, 8 mean-value Property, 18
conformal equivalence, 9 meromorphic, 33
Morera’s theorem, 20
domain, 2
star-shaped, 14 open mapping theorem, 39
order of zero, 22
elementary deformation, 26
entire, 3 pole, 31
essential singularity, 31 principal branch of logarithm, 7
exponential, 6 principal part, 33
principle of isolated zeros, 22
Fundamental Theorem of Algebra,
18 radius of convergence, 5
Fundamental Theorem of Calculus, removable singularity, 31
13 residue, 33
Rouché’s theorem, 40
holomorphic, 2
homotopy, 26 simply connected, 26
identity principle, 23 Taylor expansion, 19
index, 24
integration along a cruve, 11 winding number, 24
41

Complex Analysis Notes

Uploaded by

Copyright:

Available Formats

Complex Analysis Notes

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Complex Analysis Notes

Uploaded by

Copyright:

Available Formats

University of

Lectures by Liguori editore City:

Definition (Domain). A domain is an non-empty path-connected open

1.1 Complex differentiation

𝑓(𝑥 + 𝑖𝑦) = 𝑢(𝑥, 𝑦) + 𝑖𝑣(𝑥, 𝑦)

Definition (Differentiable, holomorphic).

1. 𝑓 ∶ 𝑈 → C is differentiable at 𝑤 ∈ 𝑈 if the limit

2. 𝑓 is holomorphic at 𝑤 if there exists 𝜀 > 0 such that 𝑓 is differentiable

Definition (Entire). An entire function is a holomorphic function 𝑓 ∶ C → C.

Theorem 1.1 (Cauchy-Riemann equations). 𝑓 ∶ 𝑈 → C is differentiable at

Proof. From the definition, 𝑓 will be differentiable at 𝑤 with derivative 𝑓 ′ (𝑤) =

1. We could have discovered Cauchy-Riemann as follows: let 𝑧 = 𝑤 + ℎ where

𝑓(𝑤 + 𝑖ℎ) − 𝑓(𝑤)

By symmetry of second derivatives, we get that 𝑢 satisfies the Laplace

Corollary 1.2. Let 𝑓 = 𝑢 + 𝑖𝑣 ∶ 𝑈 → C. Suppose the functions 𝑢 and 𝑣 have

Corollary 1.3. Let 𝑓 ∶ 𝐷 → C be holomorphic on a domain 𝐷 and suppose

1.2 Power series

Theorem 1.4 (Radius of convergence). Let 𝑐𝑛 be a sequence of complex

converges absolutely if |𝑧 − 𝑎| < 𝑅 and diverges if |𝑧 − 𝑎| > 𝑅. If 0 < 𝑟 < 𝑅

Theorem 1.5. Let 𝑓(𝑧) = ∑∞ 𝑐 (𝑧 − 𝑎)𝑛 be a complex power series with

Proof. wlog assume 𝑎 = 0. Claim that ∑∞

Proof. |𝑛𝑐𝑛 | ≥ |𝑐𝑛 | so its radius of convergence 𝑅′ is at most 𝑅.

so by comparison test ∑ 𝑛𝑐𝑛 𝑟𝑛−1 converges and hence 𝑅′ = 𝑅.

for 𝑛 ≥ 1. Consider the series

Thus it converges to a continuous function 𝑔(𝑧, 𝑤). By definition

As 𝑔 is continuous, fixing 𝑤 and letting 𝑧 → 𝑤 we get

Proposition 1.6 (Weierstrass M-test). Let 𝑓𝑛 be a sequence of functions

Definition (Exponential). The complex exponential function is defined as

1. 𝑒𝑧 is an entire function and (𝑒𝑧 )′ = 𝑒𝑧 .

4. exp(𝑧) = 1 if and only if 𝑧 ∈ 2𝜋𝑖Z.

Proof. Similar to the proof in IA Analysis I.

Definition (Logarithm). If 𝑧 ∈ C, we say that 𝑤 ∈ C is a logarithm of 𝑧 if

From above we know 𝑧 has a logarithm if and only if 𝑧 ≠ 0. However, unlike

Definition (Branch of logarithm). Let 𝑈 ⊆ C \ {0} be open. We say that a

Remark. Any branch of logarithm is in fact automatically holomorphic:

Definition (Principal branch of logarithm). Let 𝑈 = C \ {𝑥 ∈ R ∶ 𝑥 ≤ 0}.

Check that this is indeed a branch of logarithm: the projection onto 𝑆 1

exp(log 𝑧) = 𝑒𝑖 log |𝑧| (cos arg(𝑧) + 𝑖 sin arg(𝑧)) = 𝑧.

1. See remarks above.

Thus their difference is constant. Now set 𝑧 = 0.

Remark. There is no way to extend log 𝑧 to a holomorphic (or even just

lim log 𝑒𝑖𝜃 = −𝑖𝜋

Later on we will see that there is no branch of logarithm on C \ {0}.

for 𝑧 ∈ 𝑈. See example sheet.

1.3 Conformal maps

angle(𝛾1 , 𝛾2 ) = arg(𝛾1′ (0)) − arg(𝛾2′ (0)).

angle(𝛿1 , 𝛿2 ) = arg((𝑓 ∘ 𝛾1 )′ (0)) − arg((𝑓 ∘ 𝛾2 )′ (0)).

Now recall that

Definition (Conformal equivalence). If 𝑓 ∶ 𝐷 → C is holomorphic on a

2. 𝑧 ↦ 𝑧 𝑛 , {𝑧 ∈ C \ {0} ∶ 0 < arg 𝑧 < 𝜋/𝑛} → {𝑧 ∈ C ∶ Im 𝑧 > 0}. The

Theorem 1.9 (Riemann mapping theorem). Let 𝐷 ⊆ C be any domain

with equality if and only if 𝑓 is constant.

Proof. Let 𝜃 = arg ∫ 𝑓(𝑡)𝑑(𝑡) and let 𝑀 = sup𝑡∈[𝑎,𝑏] |𝑓(𝑡)|. Clearly

If 𝛾 ∶ [𝑎, 𝑏] → C is a 𝐶 1 curve, we can define its length as follow. Write

Definition (Simple curve). We say 𝛾 is simple if 𝛾(𝑡1 ) ≠ 𝛾(𝑡2 ) unless 𝑡1 = 𝑡2

Definition (Integration along a curve). Suppose 𝑓 ∶ 𝑈 → C is continuous

Some basic properties:

∫(𝑐1 𝑓1 (𝑧) + 𝑐2 𝑓2 (𝑧))𝑑𝑧 = 𝑐1 ∫ 𝑓1 (𝑧)𝑑𝑧 + 𝑐2 ∫ 𝑓2 (𝑧)𝑑𝑧.

2. additivity: if 𝑎 < 𝑎′ < 𝑏,