Sadman Rahman

MATH 399 – FA’22

Complex Analysis
Instructor: Dr. David Schmitz

Main Textbook: Functions of One

Complex Variable I by Conway

April 21, 2023

Springer Nature
Contents

I Elementary Properties of Analytic Functions . . . . . . . . . . . . . . . . . . . . . . 1

II Complex Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

III Singularities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

IV The Maximum Modulus Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

V Compactness and Convergence in the Space of Analytic Functions . . . 27

Module – 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Module – 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Module – 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

VI Harmonic and Entire Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

v
Chapter I
Elementary Properties of Analytic Functions

∞
∑︁
Problem I.1 Find the radius of convergence of the series 𝑎 𝑛 𝑧 𝑛 ; 𝑎 ∈ C.
𝑛=0

Solution I.1 Let, radius of convergence be 𝑅. Here, we set 𝑎 𝑛 = 𝑎 𝑛 and by analogy

∞
∑︁
to the power series 𝑎 𝑛 𝑧 𝑛 , we have
𝑛=0

𝑎𝑛
𝑅 = lim
𝑛→∞ 𝑎 𝑛+1

𝑎𝑛
= lim 𝑛+1
𝑛→∞ 𝑎

1 1
= lim =
𝑛→∞ 𝑎 |𝑎|

1
Thus, 𝑅 =
|𝑎|
if 𝑎 ≠ 0, and 𝑅 = ∞ if 𝑎 = 0. ♦
∞
∑︁ 2
Problem I.2 Find the radius of convergence of the series 𝑎 𝑛 𝑧 𝑛 ; 𝑎 ∈ C.
𝑛=0

2
Solution I.2 Let, radius of convergence be 𝑅. Here, we set 𝑎 𝑛 = 𝑎 𝑛 and by analogy
∞
∑︁
to the power series 𝑎 𝑛 𝑧 𝑛 , we have
𝑛=0

1
2 I Elementary Properties of Analytic Functions

𝑎𝑛
𝑅 = lim
𝑛→∞ 𝑎 𝑛+1
2
𝑎𝑛
= lim
𝑛→∞ 𝑎 (𝑛+1) 2
1 1
= lim = lim
𝑛→∞ 𝑎 2𝑛+1 𝑛→∞ |𝑎| 2𝑛+1


 0, |𝑎| > 1;
1

♦


Thus, 𝑅 = lim = 1, |𝑎| = 1;
𝑛→∞ |𝑎| 2𝑛+1 
 ∞, |𝑎| < 1;


Problem I.3 Show that 𝑓 (𝑧) = |𝑧| 2 is differentiable only at the origin.

Solution I.3 We recall by definition, 𝑓 (𝑧) is differentiable at 𝑧 if

𝑓 (𝑧 + ℎ) − 𝑓 (𝑧)
lim
ℎ→0 ℎ

exists; Substituting 𝑓 (𝑧) = |𝑧| 2 , we have

|𝑧 + ℎ| 2 − |𝑧| 2
𝑓 ′ (𝑧) = lim
ℎ→0 ℎ
(𝑧 + ℎ) ( 𝑧¯ + ℎ) ¯ − 𝑧 𝑧¯
= lim
ℎ→0 ℎ
(𝑧 + ℎ) ( 𝑧¯ + ℎ) ¯ − 𝑧 𝑧¯
= lim
ℎ→0 ℎ
𝑧 𝑧¯ + 𝑧 ℎ¯ + ℎ 𝑧¯ + ℎ ℎ¯ − 𝑧 𝑧¯
= lim
ℎ→0 ℎ
ℎ¯ ¯
 
= lim 𝑧¯ + 𝑧 + ℎ
ℎ→0 ℎ
¯ ¯ By definition, if the limit of 𝑤 exists, then we can approach
Now, let 𝑤 := 𝑧¯ + 𝑧 ℎℎ + ℎ.
from any any direction in C and arrive at the same limit point. In particular, we
choose the real and imaginary axes, to obtain:
Along Real Axis: Since ℎ = ℎ, ¯ we have

lim (𝑤) = 𝑧¯ + 𝑧.
ℎ→0

¯ we have
Along Imaginary Axis: Since ℎ = − ℎ,

lim (𝑤) = 𝑧¯ − 𝑧.
ℎ→0

If we assume 𝑓 is differentiable at 𝑧, then we must have

I Elementary Properties of Analytic Functions 3

𝑧¯ + 𝑧 = 𝑧¯ − 𝑧 =⇒ 𝑧 = 0.

Thus, 𝑓 (𝑧) = |𝑧| 2 is differentiable only at 𝑧 = 0 and 𝑓 ′ (0) = 0. ♦

Problem I.4 Describe the set {𝑧 : 𝑒 𝑧 = 𝑖}.
Solution I.4 Consider the set {𝑧 : 𝑒 𝑧 = 𝑖}. Taking complex logarithm on both sides,
we obtain

𝑧 =: log(e𝑧 ) = log(𝑖)
:= {ln |𝑖| + 𝑖 arg(𝑖) + 2𝜋𝑖𝑘 : 𝑘 ∈ Z}
= {ln(1) + 𝑖𝜋/2 + 2𝜋𝑖𝑘 : 𝑘 ∈ Z}
n 𝜋 o
= 𝑖 + 2𝜋𝑖𝑘 : 𝑘 ∈ Z
2
n 𝜋 o
Thus, {𝑧 : 𝑒 𝑧 = 𝑖} = 𝑖 + 2𝜋𝑖𝑘 : 𝑘 ∈ Z . ♦
2
√
Problem I.5 Give the principal branch of 1 − 𝑧.
Solution I.5 We recall, the principal branch of log(𝑧) is given by

Log(𝑧) = ln |𝑧| + 𝑖 Arg(𝑧),

where Arg(·) denotes the principal argument. We also recall that this branch is ana-
lytic in 𝐺, where 𝐺 = C \{𝑧 ∈ R : ℜ(𝑧) ≤ 0}.
√  
Now, by definition, 1 − 𝑧 := exp 12 Log(1 − 𝑧) . Here, the principal branch of
the logarithm in the exponent is given by

Log(1 − 𝑧) = ln |1 − 𝑧| + 𝑖 Arg(1 − 𝑧)
     
1 1 𝑖
∴ exp Log(1 − 𝑧) = exp ln |1 − 𝑧| · exp Arg(1 − 𝑧)
2 2 2
√ √︁

∴ 1 − 𝑧 = |1 − 𝑧| · cos 𝜃 + 𝑖 sin 𝜃 ,

where 𝜃 := 12 Arg(1 − 𝑧).

√
We notice that 1 − 𝑧 is analytic ∀𝑧 ∈ C \{𝑧 ∈ R : ℜ(𝑧) ≥ 1}. ♦
Problem I.6 Find the image of {𝑧 : ℜ(𝑧) < 0, |ℑ(𝑧)| < 𝜋} under the exponential
function.
Solution I.6 Let 𝑧 := 𝑥 + 𝑖𝑦. Then we have

e𝑧 = e 𝑥+𝑖𝑦 = e 𝑥 e𝑖𝑦 = e 𝑥 (cos 𝑦 + 𝑖 sin 𝑦).

Now, let 𝑟 = e 𝑥 . Since ℜ(𝑧) < 0, we have 0 < 𝑟 < 1. Moreover, we have
𝑦 ∈ (−𝜋, 𝜋). For a fixed value of 𝑟 ∈ (0, 1), letting 𝑦 vary in (−𝜋, 𝜋), the ex-
pression 𝑟 (cos 𝑦 + 𝑖 sin 𝑦) defines a circle with radius 𝑟 centered at (0, 0) but without
4 I Elementary Properties of Analytic Functions

the points (0, 0) and (−𝑟, 0). This is because 𝑦 ∈ (−𝜋, 𝜋) and 𝑟 ≠ 0.

Next, we let 𝑟 vary from 0 to 1 and observe that the expression 𝑟 (cos 𝑦 + 𝑖 sin 𝑦)
defines the open unit disk without its center and also without the negative real axis.
A sketch of the region is given in the following figure.

Fig. I.1 Sketch of the set {e𝑧 : ℜ(𝑧) < 0, |ℑ(𝑧) | < 𝜋 }.

Here, the reddish area indicates the excluded negative real axis and {0} on the
unit disk (in blue). ♦

Problem I.7 Find fixed points of a dilation, a translation and the inversion on C∞ .

Solution I.7 Dilation: We recall a Möbius Transformation 𝑀 is a dilation if

𝑀 𝑧 = 𝑎𝑧. In general, we have
𝑎𝑧 + 𝑏
𝑀𝑧 = .
𝑐𝑧 + 𝑑
So, to get dilation, we set 𝑏 = 0, 𝑐 = 0, 𝑑 = 1. Now, we find its fixed points. For fixed
points, we recall
𝑀 𝑧 = 𝑧 =⇒ 𝑎𝑧 = 𝑧.
Hence, it’s clear that 𝑧 = 0 and 𝑧 = ∞ are fixed points of 𝑀. We verify this as follows:
𝑏 0
𝑀 (0) = = = 0.
𝑑 1
𝑎 𝑎
𝑀 (∞) := = = ∞.
𝑐 0
I Elementary Properties of Analytic Functions 5

Thus, for a dilation, the fixed points are 0 and ∞.

Translation: We recall a Möbius Transformation 𝑀 is a translation if 𝑀 𝑧 = 𝑧 + 𝑏.

In general, we have
𝑎𝑧 + 𝑏
𝑀𝑧 = .
𝑐𝑧 + 𝑑
So, to get translation, we set 𝑎 = 1, 𝑐 = 0, 𝑑 = 1. Now, we find its fixed points. For
fixed points, we recall
𝑀 𝑧 = 𝑧 =⇒ 𝑧 + 𝑏 = 𝑧.
Hence, it’s clear that 𝑧 = ∞ is a fixed point of 𝑀. In fact, this is the only fixed point
of 𝑀. We verify this as follows:
𝑎 1
𝑀 (∞) := = = ∞.
𝑐 0
Thus, for a translation, the only fixed point is ∞.

Inversion: We recall a Möbius Transformation 𝑀 is an inversion if

𝑀 𝑧 = 1/𝑧. In general, we have
𝑎𝑧 + 𝑏
𝑀𝑧 = .
𝑐𝑧 + 𝑑
So, to get inversion, we set 𝑎 = 0, 𝑏 = 1, 𝑐 = 1, 𝑑 = 0. Now, we find its fixed points.
For fixed points, we recall

𝑀 𝑧 = 𝑧 =⇒ 1/𝑧 = 𝑧.

Hence, it’s clear that 𝑧 = 1 and 𝑧 = −1 are fixed points of 𝑀. We verify this as
follows:
𝑎+𝑏 0+1
𝑀 (1) = = = 1.
𝑐+𝑑 1+0
−𝑎 + 𝑏 0+1
𝑀 (−1) = = = −1.
−𝑐 + 𝑑 −1 + 0
Thus, for an inversion, the fixed points are 1 and −1. ♦
𝑎𝑧 + 𝑏
Problem I.8 If 𝑇 𝑧 = find 𝑧2 , 𝑧3 , 𝑧4 (in terms of 𝑎, 𝑏, 𝑐, 𝑑) such that
𝑐𝑧 + 𝑑
𝑇 𝑧 = (𝑧, 𝑧2 , 𝑧3 , 𝑧4 ).

Solution I.8 We recall by definition, (𝑧, 𝑧2 , 𝑧3 , 𝑧4 ) is the image of the unique Möbius
Transformation which maps 𝑧 2 ↦→ 1, 𝑧 3 ↦→ 0 and 𝑧 4 ↦→ ∞.
According to the problem, since (𝑧, 𝑧2 , 𝑧3 , 𝑧4 ) is the image of 𝑇 𝑧, 𝑇 is this unique
transformation. Thus, we are done if we can find the inverse images of 𝑧2 , 𝑧3 , 𝑧4 .
Since all Möbius transformations are invertible with the inverse given by
6 I Elementary Properties of Analytic Functions

𝑑𝑧 − 𝑏
𝑇 −1 (𝑧) = ,
−𝑐𝑧 + 𝑎

we are done if we set 𝑧2 = 𝑇 −1 (1), 𝑧3 = 𝑇 −1 (0) & 𝑧 4 = 𝑇 −1 (∞). We compute these

three values as follows:
𝑑−𝑏
𝑇 −1 (1) = ;
−𝑐 + 𝑎
−𝑏
𝑇 −1 (0) = ;
𝑎
𝑑
𝑇 −1 (∞) = ;
−𝑐
These are the values for which 𝑇 𝑧 = (𝑧, 𝑧2 , 𝑧3 , 𝑧4 ). ♦

Problem I.9 If 𝑢, 𝑣 : R2 → R are C 1 and they satisfy the Cauchy-Riemann equa-

tions, then prove that 𝑓 = 𝑢 + 𝑖𝑣 is conformal, unless 𝑢, 𝑣 are constant functions.

Solution I.9 Suppose 𝑢, 𝑣 : R2 → R are C 1 and they satisfy the CR-equations. Set
𝑓 = 𝑢 + 𝑖𝑣, a complex function in 𝐺. Now, if 𝑢, 𝑣 are constant functions, then the
Jacobian 𝐽 𝑓 of 𝑓 satisfies det 𝐽 𝑓 = 0 ≯ 0 and so 𝑓 doesn’t preserve orientation. So,
we assume 𝑢, 𝑣 are non-constant functions. Using the CR-equations, we can rewrite
the Jacobian of 𝑓 as follows:

©𝑢 𝑥 𝑢 𝑦 ª © 𝑢 𝑥 𝑢 𝑦 ª
𝐽𝑓 = ­ ®=­ ®.
« 𝑣 𝑥 𝑣 𝑦 ¬ « −𝑢 𝑦 𝑢 𝑥 ¬
   
𝑢𝑥 𝑢𝑦
Now, let, 𝛼 := and 𝛽 := be the tangent vectors of two smooth curves
𝑣𝑥 𝑣𝑦
Γ1 , Γ2 respectively. It follows from a direct computation that ⟨𝛼, 𝛽⟩ = 0 =⇒ 𝛼 ⊥ 𝛽.
Thus, 𝐽 𝑓 ∈ 𝑂 (2) and also det(𝐽 𝑓 ) = (𝑢 2𝑥 + 𝑢 2𝑦 ) > 0. So, ∃𝑇 ∈ 𝑆𝑂 (2) such that
√︃
𝐽𝑓 = 𝑢 2𝑥 + 𝑢 2𝑦 𝑇 .
√︃
So, 𝐽 𝑓 is then a composition of rotations and dilations by 𝑢 2𝑥 + 𝑢 2𝑦 and thus preserves
angles. Thus, 𝑓 = 𝑢 + 𝑖𝑣 is conformal and the proof is complete. ♦
Chapter II
Complex Integration

Problem II.1 Let 𝛾∫:= [1, 𝑖] and

∫ 𝜎 := [1, 1 + 𝑖, 𝑖] be two polygons. Express them as
paths and calculate 𝑓 and 𝑓 for 𝑓 (𝑧) = |𝑧| 2 .
𝛾 𝜎

Solution II.1 We recall that we denote the line segment from 𝑧 to 𝑤 by

[𝑧, 𝑤] := {𝑡𝑤 + (1 − 𝑡)𝑧 : 𝑡 ∈ [0, 1]}.

Also, [𝑎, 𝑧1 , . . . , 𝑧 𝑛 , 𝑏] denotes the polygon from 𝑎 to 𝑏 and is defined by the union
of line segments as

[𝑎, 𝑧1 , . . . , 𝑧 𝑛 , 𝑏] := [𝑎, 𝑧1 ] ∪ [𝑧 1 , 𝑧2 ] ∪ · · · ∪ [𝑧 𝑛 , 𝑏].

By these definitions, we have for 0 ≤ 𝑡 ≤ 1,

𝛾 = [1, 𝑖] = {𝑡𝑖 + (1 − 𝑡)1} ;

𝜎 = [1, 1 + 𝑖, 𝑖] = [1, 1 + 𝑖] ∪ [1 + 𝑖, 𝑖]
= {𝑡 (𝑖 + 1) + (1 − 𝑡)1} ∪ {𝑡𝑖 + (1 − 𝑡) (1 + 𝑖)}
= {1 + 𝑖𝑡} ∪ {𝑖 + (1 − 𝑡)}
= 𝜎1 ∪ 𝜎2 ;

where we let 𝜎1 := {𝑡 : 1 + 𝑖𝑡}, 𝜎2 := {𝑡 : 𝑖 + (1 − 𝑡)} with 𝑡 ∈ [0, 2𝜋]. Now, by

definition of line integrals,
∫ ∫ 1
𝑓 := ( 𝑓 ◦ 𝛾)𝛾 ′ (𝑡) 𝑑𝑡.
𝛾 0

7
8 II Complex Integration
∫ ∫ 1
𝑡 2 + (1 − 𝑡) 2 𝑑𝑡
 
∴ 𝑓 = (𝑖 − 1)
𝛾 0
  1
2 2
= (𝑖 − 1) 𝑡 3 − 𝑡 2 + 𝑡 = (𝑖 − 1).
3 3
0
∫ ∫ ∫ ∫
We compute 𝜎 𝑓 similarly by recalling that 𝜎1 ∪𝜎2
𝑓 = 𝜎1
𝑓+ 𝜎2
𝑓 since 𝜎(𝑡) is
of bounded variation.
∫ ∫ ∫
∴ 𝑓 = 𝑓+ 𝑓
𝜎 𝜎1 𝜎2
∫ 1 ∫ 1
2
1 + (1 − 𝑡) 2 𝑑𝑡
   
=𝑖 1+𝑡 𝑑𝑡 −
0 0
1 1
𝑡3 𝑡3
 
4
=𝑖 𝑡+
3
− 2𝑡 − 𝑡 2 +
3
=
3
(𝑖 − 1). ♦
0 0

Problem II.2 Give the power series expansion of log 𝑧 about 𝑧 = 𝑖 and find its radius
of convergence.
Solution II.2 We recall that the power series expansion of a function analytic at 𝑧 0
is given by
∞
∑︁ 𝑓 (𝑛) (𝑧0 )
𝑓 (𝑧) = (𝑧 − 𝑧 0 ) 𝑛 ;
𝑛=0
𝑛!

In this problem, 𝑓 (𝑧) = log 𝑧 and 𝑧 0 = 𝑖. We calculate the derivatives of the complex
logarithm as follows:
1
𝑓 ′ (𝑧) = ;
𝑧
1
𝑓 ′′ (𝑧) = − 2 ;
𝑧
2
𝑓 ′′′ (𝑧) = 3 ;
𝑧
(4) 6
𝑓 (𝑧) = − 4 ;
𝑧
..
.
(𝑛 − 1)!
𝑓 (𝑛) (𝑧) = (−1) 𝑛−1 ;
𝑧𝑛
(𝑛 − 1)!
∴ 𝑓 (𝑛) (𝑖) = (−1) 𝑛−1 .
𝑖𝑛

Substituting this in the formula for the power series expansion, for 𝑘 ∈ Z, we obtain
II Complex Integration 9
∞
∑︁ (−1) 𝑛−1 (𝑛 − 1)!
log 𝑧 = log(𝑖) + 𝑛
(𝑧 − 𝑖) 𝑛
𝑛=1
𝑖 𝑛!
∞
𝑖𝜋 ∑︁ (−1) 𝑛−1
= 2𝜋𝑖𝑘 + + (𝑧 − 𝑖) 𝑛 .
2 𝑛=1 𝑛 𝑖 𝑛

To find the Radius of Convergence, 𝑅, we recall the formula

𝑎𝑛
𝑅 := lim ,
𝑛→∞ 𝑎 𝑛+1

(−1) 𝑛−1
where 𝑎 𝑛 = . Thus,
𝑛 𝑖𝑛
(−1) 𝑛−1 (𝑛 + 1) 𝑖 𝑛+1
𝑅 = lim ·
𝑛→∞ 𝑛 𝑖𝑛 (−1) 𝑛
𝑛+1
= lim (−𝑖) ·
𝑛→∞ 𝑛
𝑛+1
= lim
𝑛→∞ 𝑛
= 1. ♦

∫
sin 𝑧
Problem II.3 Evaluate 𝑑𝑧, where 𝛾(𝑡) = e𝑖𝑡 and 𝑡 ∈ [0, 2𝜋].
𝛾 𝑧3

Solution II.3 Let 𝑓 (𝑧) = sin 𝑧. We recall that sin 𝑧 is analytic on the whole plane C.
¯ 0) ⊂ C and 𝑓 ′′ (𝑧) = − sin 𝑧. Thus,
Now, it is obvious that 𝐵(1;
∫
2! 𝑓 (𝑧)
𝑓 ′′ (0) = 𝑑𝑧
2𝜋𝑖 𝛾 (𝑧 − 0) 2+1
∫
sin 𝑧
∴
𝛾 𝑧
3
𝑑𝑧 = −𝜋𝑖 · sin 0 = 0. ♦

 
1 1 + 𝑖𝑧
Problem II.4 Let 𝑓 (𝑧) = log ; Show that
2𝑖 1 − 𝑖𝑧
∞
∑︁ 𝑧2𝑘+1
𝑓 (𝑧) = (−1) 𝑘
𝑘=0
2𝑘 + 1

for |𝑧| < 1. Note that tan 𝑓 (𝑧) = 𝑧 (i.e. 𝑓 is a branch of arctan 𝑧).

Solution II.4 We consider the principal branch of the logarithm. Then we find the
power series expansion of 𝑓 at 𝑧 = 0 as follows:
10 II Complex Integration

1
𝑓 (𝑧) = [log(1 + 𝑖𝑧) − log(1 − 𝑖𝑧)]
2𝑖
1 h i
= log(𝑖) + log(𝑧 − 𝑖) − log(−𝑖) − log(𝑧 + 𝑖) ;
2𝑖
′ 1 1 − 𝑖𝑧 𝑖(1 − 𝑖𝑧) + 𝑖(1 + 𝑖𝑧) 1
𝑓 (𝑧) = · · 2
= ;
2𝑖 1 + 𝑖𝑧 (1 − 𝑖𝑧) 1 + 𝑧2
2𝑧
𝑓 ′′ (𝑧) = −
2 ;
1 + 𝑧2
2(3𝑧 2 − 1)
𝑓 ′′′ (𝑧) =
3 ;
1 + 𝑧2
..
.
 
(−1) 𝑛−1 1 1
𝑓 (𝑛) (𝑧) = (𝑛 − 1)! − ;
2𝑖 (𝑧 − 𝑖) 𝑛 (𝑧 + 𝑖) 𝑛

Substituting 𝑧 = 0 in the last equation, we obtain

 (𝑛 − 1)! ,
  
(−1) 𝑛−1 1 1 if 𝑛 is odd;


(𝑛)
𝑓 (0) = (𝑛 − 1)! − = 𝑖 𝑛+1
2𝑖 (−𝑖) 𝑛 𝑖 𝑛 
 0, if 𝑛 is even;

Now, we recall that if 𝑛 is odd then 𝑛 + 1 is even and thus 𝑖 𝑛+1 is always ±1. Thus,
for |𝑧| < 1, we obtain
∑︁ (𝑘 − 1)!
𝑓 (𝑧) = 𝑘+1 𝑘!
𝑧𝑘
𝑘 is odd
𝑖
∑︁ 𝑧𝑘
=
𝑘 is odd
𝑖 𝑘+1 · 𝑘
∞
𝑧 2𝑘+1
♦
∑︁
∴ 𝑓 (𝑧) = (−1) 𝑘 .
𝑘=0
2𝑘 + 1

Problem II.5 Let 𝐺 be a region and 𝑓 : 𝐺 → C is an analytic function such that

there is a point 𝛼 ∈ 𝐺 with | 𝑓 (𝛼)| ≤ | 𝑓 (𝑧)| for all 𝑧 ∈ 𝐺. Show that either 𝑓 is
constant or 𝑓 (𝛼) = 0.

Solution II.5 Let 𝑓 be an analytic function in 𝐺 satisfying the above hypotheses and
assume that 𝑓 (𝛼) ≠ 0 for each 𝛼 ∈ 𝐺. It is enough to show that this implies 𝑓 is
constant in 𝐺.
1
Now, consider the function 𝑔 : 𝐺 → C defined by 𝑔(𝑧) = . Since 𝑓 is analytic
𝑓 (𝑧)
and non-zero everywhere in 𝐺, it follows that 𝑔 is analytic in 𝐺. Next, we notice that
for any 𝛼 ∈ 𝐺,
II Complex Integration 11

1
|𝑔(𝑧)| =
𝑓 (𝑧)
1
=
| 𝑓 (𝑧)|
1
≤ = |𝑔(𝛼)|.
| 𝑓 (𝛼)|

Thus, |𝑔(𝑧)| ≤ |𝑔(𝛼)|, ∀𝛼 ∈ 𝐺. But by the Maximum Modulus Principle, this means
that 𝑔 and hence 𝑓 is a constant and we are done. ♦

Problem II.6 Let 𝑓 be an entire function and suppose there is a constant 𝑀, an

𝑅 > 0, and an integer 𝑛 ≥ 1 such that | 𝑓 (𝑧)| ≤ 𝑀 |𝑧| 𝑛 for |𝑧| > 𝑅. Show that 𝑓 is a
polynomial of degree ≤ 𝑛.

Solution II.6 Since 𝑓 is entire, it has a power series around 𝑧 = 0,

∞
∑︁ 𝑓 (𝑘 ) (0) 𝑘
𝑓 (𝑧) = 𝑧 .
𝑘=0
𝑘!

Now, for some 𝑟 > 0, since 𝐵(0; 𝑟) ∈ C, we can apply Corollary – IV.2.13 to get
∫
𝑘! 𝑓 (𝑧)
𝑓 (𝑘 ) (0) = 𝑑𝑧 ,
2𝜋𝑖 𝛾 𝑧 𝑘+1

where 𝛾(𝑡) = 𝑟e𝑖𝑡 for 𝑡 ∈ [0, 2𝜋]. Thus, it is enough to show that 𝑓 (𝑘 ) (0) = 0 for
all 𝑘 ≥ 𝑛 + 1. We note that
∫
𝑘! 𝑓 (𝑧)
𝑓 (𝑘 ) (0) = 𝑑𝑧
2𝜋 𝛾 𝑧 𝑘+1
∫
𝑘! | 𝑓 (𝑧)|
≤ |𝑑𝑧|
2𝜋 𝛾 |𝑧| 𝑘+1
∫
𝑀 𝑘! |𝑧| 𝑛
≤ |𝑑𝑧|
2𝜋 𝛾 |𝑧| 𝑘+1
∫
𝑀 𝑘!
≤ |𝑧| 𝑛−𝑘−1 |𝑑𝑧|
2𝜋 𝛾
∫
𝑀 𝑘! 𝑛−𝑘−1
= ·𝑟 |𝑑𝑧|
2𝜋 𝛾
𝑀 𝑘! 𝑛−𝑘−1 𝑖𝑡 2 𝜋
= ·𝑟 𝑟e 0
2𝜋
𝑀 𝑘! 𝑛−𝑘
= ·𝑟 · 2𝜋 .
2𝜋
12 II Complex Integration

Thus, 𝑓 (𝑘 ) (0) = 𝑀 𝑘! 𝑟 𝑛−𝑘 , which implies that 𝑓 (𝑘 ) (0) → 0 for 𝑘 ≥ 𝑛 + 1 since

𝑟 𝑛−𝑘 → 0. So, 𝑓 (𝑘 ) (0) = 0 for all 𝑘 ≥ 𝑛 + 1 and we are done. ♦

Problem II.7 Give an elementary proof of the Maximum Modulus Principle for
polynomials.

Solution II.7 Since all polynomials are entire, by Liouville’s Theorem, if 𝑓 (𝑧) is a
polynomial satisfying | 𝑓 (𝑧)| ≤ | 𝑓 (𝑎)| for all 𝑧 ∈ C then 𝑓 is constant in C . ♦

Problem II.8 Show that Cauchy’s Integral Formula follows from Cauchy’s Theorem.
∫
Solution II.8 Cauchy’s Theorem states that if 𝛾 ≈ 0 (homologous) then 𝛾 𝑓 = 0
for any analytic function 𝑓 defined in an open connected set 𝐺 ⊆ C. To derive the
Integral Formula from Cauchy’s Theorem, we consider the function Φ : 𝐺 → C
defined by

 𝑓 (𝑧) − 𝑓 (𝑎) ,

for 𝑧 ≠ 𝑎;


Φ(𝑧) = 𝑧−𝑎


 𝑓 ′ (𝑎), for 𝑧 = 𝑎;

where 𝑎 ∉ {𝛾}. If we can show this function Φ(𝑧) is analytic in 𝐺, then we have that
∫ ∫
𝑓 (𝑧) − 𝑓 (𝑎)
0= Φ= 𝑑𝑧
𝛾 𝛾 𝑧−𝑎
∫ ∫
𝑓 (𝑧) 𝑓 (𝑎)
= 𝑑𝑧 − 𝑑𝑧
𝛾 𝑧−𝑎 𝛾 𝑧−𝑎
∫ ∫
𝑓 (𝑧) 𝑑𝑧
= 𝑑𝑧 − 𝑓 (𝑎)
𝛾 𝑧−𝑎 𝛾 𝑧−𝑎
∫
𝑓 (𝑧)
= 𝑑𝑧 − 𝑛(𝛾; 𝑎) 2𝜋𝑖 𝑓 (𝑎)
𝛾 𝑧−𝑎
∫
1 𝑓 (𝑧)
∴ 𝑛(𝛾; 𝑎) 𝑓 (𝑎) = 𝑑𝑧 ,
2𝜋𝑖 𝛾 𝑧 − 𝑎

which is the Cauchy Integral Formula. So, the proof will be complete if we can show
that Φ(𝑧), defined as above, is an analytic function in 𝐺.
If 𝑧 ≠ 𝑎, then Φ(𝑧) is clearly analytic (hence also continuous). Next, we note that
 
𝑓 (𝑧) − 𝑓 (𝑎)
lim [Φ(𝑧)] = lim = 𝑓 ′ (𝑎),
𝑧→𝑎 𝑧→𝑎 𝑧−𝑎

which shows that Φ is continuous at 𝑧 = 𝑎. For showing Φ is differentiable at 𝑧 = 𝑎,

we recall that since 𝑓 is analytic in 𝐺, this implies that 𝑓 is infinitely differentiable
in 𝐺. In particular, 𝑓 ′′ exists and is also analytic in 𝐺. Now,
II Complex Integration 13
 
Φ(𝑧) − Φ(𝑎)
Φ′ (𝑎) = lim
𝑧→𝑎 𝑧−𝑎
" 𝑓 (𝑧) − 𝑓 (𝑎) #
𝑧−𝑎 − 𝑓 ′ (𝑎)
= lim
𝑧→𝑎 𝑧−𝑎
𝑓 (𝑧) − 𝑓 (𝑎) − 𝑓 ′ (𝑎) (𝑧 − 𝑎)
 
= lim
𝑧→𝑎 (𝑧 − 𝑎) 2
′′
= 𝑓 (𝑎)/2,

where the last equality follows from the fact that the analyticity of 𝑓 implies a power
series expansion of 𝑓 around 𝑧 = 𝑎. So, we have Φ′ (𝑎) = 𝑓 ′′ (𝑎)/2, implying that Φ
is differentiable at 𝑧 = 𝑎 and the proof is complete. ♦
Chapter III
Singularities

Problem III.1 The following functions have an isolated singularity at 𝑧 = 0. De-

termine its nature; if it is a removable singularity define 𝑓 (0) so that is analytic at
𝑧 = 0; if it is a pole find the singular part.
sin 𝑧
Solution III.1 Part-(a): 𝑓 (𝑧) = . Using the power series for sin 𝑧 around 𝑧0 = 0,
𝑧
we obtain the following Laurent Series:

sin 𝑧 𝑧2 𝑧4 𝑧6
𝑓 (𝑧) = =1− + − +··· ;
𝑧 3! 5! 7!
Since 𝑎 𝑛 = 0 for all 𝑛 ≤ −1, by Corollary – 1.18, we know that 𝑓 has a removable
singularity at 𝑧 = 0. Next, we define a function 𝑔 : C → C that agrees with 𝑓 in
some 𝐵(0; 𝑅) but is also analytic at 𝑧 = 0. We note that 𝑔 defined as
(
𝑓 (𝑧), 𝑧 ≠ 0;
𝑔(𝑧) =
1, 𝑧 = 0.

is analytic at 𝑧 = 0 and we are done. ♦

log(1 + 𝑧)
Part-(e): 𝑓 (𝑧) = . Using the power series for log(1 + 𝑧) around 𝑧0 = 0,
𝑧2
we obtain the following Laurent Series:

log(1 + 𝑧) 1 1 𝑧 𝑧 2
𝑓 (𝑧) = = − + − + · · · + 2𝜋𝑖𝑘,
𝑧2 𝑧 2 3 4
where 𝑘 ∈ Z. Since 𝑎 𝑛 = 0 for all 𝑛 ≤ −2, by Corollary – 1.18, we know that 𝑓 has
a pole of order 1 at 𝑧 = 0. Next, we compute the singular part S 𝑓 (𝑧) of 𝑓 . From the
above Laurent Series, we have S 𝑓 (𝑧) = 𝑧 −1 . ♦

15
16 III Singularities
 
1
Problem III.2 Find the type of singularity of 𝑓 (𝑧) = exp at 𝑧 = 0. Determine
𝑧
𝑓 ({𝑧 : 0 < |𝑧| < 𝛿}) for arbitrarily small values of 𝛿.

Solution III.2 Using the power  series

 of exp(𝑧) around 𝑧0 = 0, we obtain the
1
following Laurent series for exp :
𝑧
∞
𝑧 −𝑘
  ∑︁
1 1 1
𝑓 (𝑧) = exp = = 1+ + 2 +··· ;
𝑧 𝑘=0
𝑘! 𝑧 2𝑧

Since 𝑎 𝑛 ≠ 0 for infinitely many 𝑛 ≤ −1, by Corollary – 1.18, we know that 𝑓 has
an essential singularity at 𝑧 = 0.
Next, we find 𝑓 ({𝑧 : 0 < |𝑧| < 𝛿}) for arbitrarily small values of 𝛿. Let 𝐴 := {𝑧 :
0 < |𝑧| < 𝛿}. Being motivated by the Casorati-Weierstrass Theorem, we claim that

𝑓 ({𝑧 : 0 < |𝑧| < 𝛿}) = C \{0}.

We now proceed to prove our claim. Let 𝜔, 𝑧 ∈ C such that 𝜔 = e1/𝑧 and 0 < |𝑧| < 𝛿.
If 1/𝑧 = 𝑥 + 𝑖𝑦 (𝑥, 𝑦 ∈ R), then we have |𝜔| = e 𝑥 |e𝑖𝑦 |. This implies that |𝜔| > 0 and
hence 0 ∉ 𝐴. So, 𝐴 ⊆ C \{0}.
Next, we show that any nonzero 𝜔 ∈ C can be expressed as 𝜔 = e1/𝑧 . This will then
imply that C \{0} ⊆ 𝐴. We notice that we can express 𝜔 in polar form as follows:

𝜔 = 𝑟e𝑖 𝜃 = eln 𝑟 e𝑖 𝜃 = e𝑖 𝜃+ln 𝑟 ,

where 𝑟, 𝜃 ∈ R. Now, if we choose a positive integer 𝑘 ∈ Z+ such that

1
| ln 𝑟 + 𝑖𝜃 + 2𝜋𝑖𝑘 | < ,
𝛿
and then set 𝑧 = ln 𝑟 + 𝑖𝜃 + 2𝜋𝑖𝑘, we have 𝜔 = e1/𝑧 for 0 < |𝑧| < 𝛿. Thus, 𝜔 ∈ 𝐴 and
we must have 𝐴 = C \{0} and we are done. ♦

∫ 𝜋
𝑑𝜃
Problem III.3 Compute .
0 (𝑎 + cos 𝜃) 2

Solution III.3 We observe that if 𝑧 = e𝑖 𝜃 , then 𝑧¯ = 1/𝑧. Using this, we obtain

 2
2 𝑧 + 𝑧¯
(𝑎 + cos 𝜃) = 𝑎 +
2
 2
1 1
= 2𝑎 + 𝑧 +
4 𝑧
2

2
𝑧 + 2𝑎𝑧 + 1
= .
4𝑧2
III Singularities 17

We also notice that since the integrand is symmetric about 𝜃 = 𝜋,

∫ 𝜋 ∫ 2𝜋
𝑑𝜃 1 𝑑𝜃
2
=
0 (𝑎 + cos 𝜃) 2 0 (𝑎 + cos 𝜃) 2
∫
−2𝑖𝑧 𝑑𝑧
=
2 ,
𝛾 𝑧2 + 2𝑎𝑧 + 1
√ √
where 𝛾 is the circle |𝑧| = 1. Moreover, 𝛼 := −𝑎 + 𝑎 2 − 1 and 𝛽 := −𝑎 − 𝑎 2 − 1
are each poles of order 2 of the above integrand. Since 𝑎 > 1, this gives |𝛼| < 1
and |𝛽| > 1. Hence 𝑛(𝛾; 𝛼) = 1 and 𝑛(𝛾; 𝛽) = 0. Next, we compute Res( 𝑓 ; 𝛼). By
Proposition – V.2.4, since 𝛼 is a pole of order 2, if 𝑔(𝑧) = (𝑧 − 𝛼) 2 𝑓 (𝑧), we have
√
𝑔 ′ (𝛼) (𝛼 − 𝛽) 2 − 2𝛼 2(𝑎 2 − 1) + 𝑎 − 𝑎 2 − 1
Res( 𝑓 ; 𝛼) = = = .
(2 − 1)! (𝛼 − 𝛽) 4 8(𝑎 2 − 1) 2
By the Residue Theorem, we have
√
2(𝑎 2 − 1) + 𝑎 − 𝑎 2 − 1
∫
−2𝑖𝑧 𝑑𝑧

2 = (𝜋𝑖) · .
𝛾 𝑧2 + 2𝑎𝑧 + 1 4(𝑎 2 − 1) 2

Combining this with the previous result, we finally obtain

√
2(𝑎 2 − 1) + 𝑎 − 𝑎 2 − 1
∫ 𝜋

0
𝑑𝜃
(𝑎 + cos 𝜃) 2
=𝜋·
2(𝑎 2 − 1) 2
. ♦

∫
Problem III.4 Find all possible values of exp(1/𝑧) 𝑑𝑧 for any closed curve 𝛾 not
𝛾
passing through 𝑧 = 0.

Solution III.4 From Problem – III.2, we know that the Laurent series of exp(1/𝑧)
about 𝑧 = 0 is given by
∞
𝑧 −𝑘
  ∑︁
1 1 1
𝑓 (𝑧) = exp = = 1+ + 2 +··· ;
𝑧 𝑘=0
𝑘! 𝑧 2𝑧

Thus, Res( 𝑓 ; 0) = 𝑎 −1 = 1. Using the Residue Theorem, we obtain

∫
exp(1/𝑧) 𝑑𝑧 = 𝑛(𝛾; 0).
𝛾
♦
18 III Singularities

Problem III.5 Suppose that 𝑓 has a simple pole at 𝑧 = 𝑎 and let 𝑔 be analytic in an
open set containing 𝑎. Show that Res( 𝑓 𝑔; 𝑎) = 𝑔(𝑎) Res( 𝑓 ; 𝑎).
Solution III.5 Let 𝑓 , 𝑔 be functions as above.
First assume 𝑔 has a simple zero at 𝑧 = 𝑎 i.e. 𝑔(𝑎) = 0. Then 𝑓 𝑔 has a removable
singularity at 𝑧 = 𝑎. But this implies that Res( 𝑓 𝑔; 𝑎) = 0. We also have that
𝑔(𝑎) Res( 𝑓 ; 𝑎) = 0, since 𝑔(𝑎) = 0. Hence, the identity is true for this case.
Next, assume that 𝑔(𝑎) ≠ 0. This implies that 𝑓 𝑔 has a simple pole at 𝑧 = 𝑎 and we
have

Res( 𝑓 𝑔; 𝑎) = lim [(𝑧 − 𝑎) ( 𝑓 𝑔)] = lim [(𝑧 − 𝑎) 𝑓 (𝑧)] · lim [𝑔(𝑧)] = 𝑔(𝑎) Res( 𝑓 ; 𝑎).
𝑧→𝑎 𝑧→𝑎 𝑧→𝑎

So, we are done. ♦

Problem III.6 Use Problem – III.5 to show that if 𝐺 is a region and 𝑓 is analytic in
𝐺 except for simple poles at 𝑎 1 , . . . , 𝑎 𝑛 ; and if 𝑔 is analytic in 𝐺 then
∫ 𝑛
1 ∑︁
𝑓𝑔 = 𝑛(𝛾; 𝑎 𝑘 ) 𝑔(𝑎 𝑘 ) Res( 𝑓 ; 𝑎 𝑘 ).
2𝜋𝑖 𝛾 𝑘=1

Solution III.6 Since all the 𝑎 𝑖 ’s are simple poles of 𝑓 , using the Residue Theorem
and the formula from the previous Exercise, we have
∫ 𝑛
1 ∑︁
𝑓𝑔 = 𝑛(𝛾; 𝑎 𝑘 ) Res( 𝑓 𝑔; 𝑎 𝑘 )
2𝜋𝑖 𝛾 𝑘=1
𝑛
∑︁
= 𝑛(𝛾; 𝑎 𝑘 ) 𝑔(𝑎 𝑘 ) Res( 𝑓 ; 𝑎 𝑘 ).
𝑘=1

So, we are done. ♦

Problem III.7 Let 𝛾 be the rectangle 𝑛 + 12 + 𝑛𝑖, −𝑛 − 21 + 𝑛𝑖, −𝑛 − 21 − 𝑛𝑖, 𝑛 + 12 − 𝑛𝑖, 𝑛 + 1

 
2 + 𝑛𝑖 .
Evaluate 𝛾 𝜋 𝑧cot(
∫ 𝜋 𝑧)
2 −𝑎 2 𝑑𝑧 and show that the limit goes to zero as 𝑛 → ∞. Consequently,
show that
∞
1 ∑︁ 2𝑎
𝜋 cot(𝜋𝑎) = + .
𝑎 𝑛=1 𝑎 2 − 𝑛2
Solution III.7 The above integrand has simple poles at 𝑧 = 𝑘 for 𝑘 ∈ Z, and it has
two additional simple poles at 𝑧 = ±𝑎. Using the Residue Theorem, we then have
∫ 𝑛
1 ∑︁
𝑓 = Res( 𝑓 ; 𝑎) + Res( 𝑓 ; −𝑎) + Res( 𝑓 ; 𝑘),
2𝜋𝑖 𝛾 𝑘=−𝑛

𝜋 cot(𝜋𝑧)
where 𝑓 (𝑧) := . Now, by definition,
𝑧2 − 𝑎2
III Singularities 19
 
𝜋 cot(𝜋𝑧) 1
Res( 𝑓 ; 𝑘) = lim (𝑧 − 𝑘) 2 2
= 2 .
𝑧→𝑘 𝑧 −𝑎 𝑘 − 𝑎2

Also,  
𝜋 cot(𝜋𝑧) 𝜋 cot(𝜋𝑎)
Res( 𝑓 ; ±𝑎) = lim (𝑧 ± 𝑎) · = ,
𝑧→±𝑎 𝑧2 − 𝑎2 2𝑎
∫
for both cases. Next, we compute lim 𝑓 . But as 𝑛 → ∞, | cot(𝜋𝑧)| ≤ 2 and
𝑛→∞ 𝛾
1
→ 0 for 𝑧 ∈ {𝛾}. Thus,
𝑧2 − 𝑎2
∫
𝜋 cot(𝜋𝑧)
lim 𝑑𝑧 = 0.
𝑛→∞ 𝛾 𝑧2 − 𝑎2

Combining this fact with our previous results, we obtain

∞
𝜋 cot(𝜋𝑎) ∑︁ 1
=−
𝑎 𝑘=−∞
𝑘 − 𝑎2
2

∞
∑︁ 1
= 2 − 𝑘2
𝑘=−∞
𝑎
∞
1 ∑︁ 2
= 2
+
𝑎 𝑘=1
𝑎 − 𝑘2
2

∞
1 ∑︁ 2𝑎
=⇒ 𝜋 cot(𝜋𝑎) = +
𝑎 𝑘=1 𝑎 2 − 𝑘 2
. ♦

Problem III.8 Let 𝑓 be an entire function and let ∫ 𝑎, 𝑏 ∈ C such that |𝑎| < 𝑅 and
|𝑎| < 𝑅. If 𝛾(𝑡) = 𝑅e𝑖𝑡 for 𝑡 ∈ [0, 2𝜋], evaluate 𝛾 [(𝑧 − 𝑎) (𝑧 − 𝑏)] −1 𝑓 (𝑧) 𝑑𝑧. Use
this result to give another proof of Liouville’s Theorem.

Solution III.8 Let 𝑓 be a bounded entire function, say for 𝑀 > 0, | 𝑓 (𝑧)| ≤ 𝑀 for
1
each 𝑧 ∈ C. Now, since 𝑔(𝑧) = has two simple poles at 𝑧 = 𝑎, 𝑏 and
(𝑧 − 𝑎) (𝑧 − 𝑏)
𝑓 is analytic everywhere, using Problem – III.6, we obtain
∫
1
𝑓 𝑔 = 𝑓 (𝑎) Res(𝑔; 𝑎) + 𝑓 (𝑏) Res(𝑔; 𝑏)
2𝜋𝑖 𝛾
𝑓 (𝑎) 𝑓 (𝑏)
= +
𝑎−𝑏 𝑏−𝑎
𝑓 (𝑎) − 𝑓 (𝑏)
= .
𝑎−𝑏
Next, using the fact that 𝑓 is bounded and 𝛾 is a circle, we have
20 III Singularities
∫ ∫ 2𝜋
1 1 𝑓 (𝑅e𝑖𝑡 ) 𝑑𝑡
𝑓𝑔 = 𝑖𝑅e𝑖𝑡 ·
2𝜋𝑖 𝛾 2𝜋 0 (𝑅e𝑖𝑡 − 𝑎) (𝑅e𝑖𝑡 − 𝑏)
∫ 2𝜋
1 | 𝑓 (𝑅e𝑖𝑡 )| 𝑑𝑡
≤ 𝑖𝑅e𝑖𝑡 ·
2𝜋 0 |𝑅e𝑖𝑡 − 𝑎||𝑅e𝑖𝑡 − 𝑏|
∫ 2𝜋
𝑅 𝑀 𝑑𝑡
≤
2𝜋 0 (𝑅 − |𝑎|) (𝑅 − |𝑏|)
𝑀 𝑅(2𝜋 − 0)
=
2𝜋(𝑅 − |𝑎|) (𝑅 − |𝑏|)
𝑀𝑅
= .
(𝑅 − |𝑎|) (𝑅 − |𝑏|)
∫
1
Letting 𝑅 → ∞, we have 𝑓 𝑔 → 0. Combining this with our evaluated
∫ 2𝜋𝑖 𝛾
expression for 2 1𝜋𝑖 𝛾 𝑓 𝑔, we must have that for sufficiently large 𝑅,

𝑓 (𝑎) − 𝑓 (𝑏)
= 0,
𝑎−𝑏

which implies 𝑓 (𝑎) = 𝑓 (𝑏). But since 𝑎, 𝑏, 𝑅 were all arbitrary, we have that 𝑓 is a
constant function. ♦

∞
𝑥 2 𝑑𝑥
∫
Problem III.9 Compute .
0 𝑥4 + 𝑥2 + 1

𝑧2
Solution III.9 We consider the function 𝑓 (𝑧) = 4 . We observe that this
𝑧 + 𝑧2 + 1
1 √ 1 √ 1 √ 1 √
function has simple poles at 𝑧 = (1 + 3𝑖), (−1 + 3𝑖), (−1 − 3𝑖), (1 − 3𝑖).
2 2 2 2
We call them 𝑎 1 , . . . , 𝑎 4 respectively. Since each 𝑎 𝑖 is a simple pole, we have

𝑎 2𝑖
Res( 𝑓 ; 𝑎 𝑖 ) = Î ,
(𝑎 𝑖 − 𝑎 𝑗 )
𝑖≠ 𝑗

for 𝑖, 𝑗 = 1, 2, 3, 4. Here, we consider the curve 𝛾 := {𝑅e𝑖𝑡 : 𝑡 ∈ [0, 𝜋]} ∪ [−𝑅, 𝑅]

for some 𝑅 > 2. Since the region enclosed by 𝛾 only contains the poles 𝑎 1 , 𝑎 2 then
winding number 𝑛(𝛾; 𝑎 3 ) = 0 = 𝑛(𝛾; 𝑎 4 ). Combining these results using the Residue
Theorem, we obtain
∫
1 −𝑖
𝑓 = Res( 𝑓 ; 𝑎 1 ) + Res( 𝑓 ; 𝑎 2 ) = √ .
2𝜋𝑖 𝛾 2 3
But using the definition of line integrals, we also have
III Singularities 21

𝑥 2 𝑑𝑥 𝑅 2 e2𝑖𝑡 𝑑𝑡
∫ ∫ 𝑅 ∫ 𝜋

𝑓 = + 𝑖𝑅e𝑖𝑡 · .
𝛾 −𝑅 𝑥 + 𝑥2 + 1
4
0 𝑅 4 e4𝑖𝑡+ 𝑅 2 e2𝑖𝑡 + 1

This implies that

𝑥 2 𝑑𝑥 𝑖𝑅 3 e3𝑖𝑡 𝑑𝑡
∫ 𝑅 ∫ ∫ 𝜋
= 𝑓− .
−𝑅 𝑥4 + 𝑥2 + 1 𝛾 0 𝑅 4 e4𝑖𝑡 + 𝑅 2 e2𝑖𝑡 + 1
∫ 𝜋
From our previous calculation, we have 𝑓 = √ . Moreover, we notice that
𝛾
3

𝑖𝑅 3 e3𝑖𝑡 𝑑𝑡 𝑖𝑅 3 e3𝑖𝑡 𝑑𝑡
∫ 𝜋 ∫ 𝜋
≤
0 𝑅 4 e4𝑖𝑡 + 𝑅 2 e2𝑖𝑡 + 1 0 𝑅 4 e4𝑖𝑡 + 𝑅 2 e2𝑖𝑡 + 1
𝑅 3 |e3𝑖𝑡 | 𝑑𝑡
∫ 𝜋
=
0 |𝑅 4 e4𝑖𝑡 + 𝑅 2 e2𝑖𝑡 + 1|
𝑅 3 𝑑𝑡
∫ 𝜋
≤
0 𝑅 4 e4𝑖𝑡 + 𝑅 2 e2𝑖𝑡 − 1
𝑅 3 𝑑𝑡
∫ 𝜋
≤
0 𝑅4 − 𝑅2 − 1
𝜋𝑅 3
= 4 .
𝑅 − 𝑅2 − 1
Letting 𝑅 → ∞, we finally obtain that

𝑖𝑅 3 e3𝑖𝑡 𝑑𝑡
∫ 𝜋
→ 0.
0 𝑅 4 e4𝑖𝑡 + 𝑅 2 e2𝑖𝑡 + 1

Thus, we have
𝑥 2 𝑑𝑥
∫ 𝑅
𝜋
lim =√ .
𝑅→∞ −𝑅 𝑥4 + 𝑥2 + 1 3
𝑥2
But since is an even function, we obtain
𝑥4 + 𝑥2 + 1
∫ ∞
𝑥 2 𝑑𝑥 1 ∞ 𝑥 2 𝑑𝑥
∫

0 𝑥 4 + 𝑥2 + 1
=
2 −∞ 𝑥 4 + 𝑥2 + 1
𝜋
= √ .
2 3
♦
22 III Singularities

Problem III.10 State and prove a more general version of Rouche’s Theorem for
curves other than circles in 𝐺.

Solution III.10 We state the following general version of Rouche’s Theorem:

Suppose 𝑓 and 𝑔 are meromorphic in a region 𝐺 with no zeros or poles on

𝑓 𝑓 𝑔 𝑔
a closed rectifiable curve 𝛾 with 𝛾 ∼ 0(𝐺). If 𝑧 𝑘 , 𝑎 𝑘 (𝑧 𝑘 , 𝑎 𝑘 ) are respectively
the zeros and poles of 𝑓 (or 𝑔), and 𝑍 𝑓 , 𝑍 𝑔 (𝑃 𝑓 , 𝑃𝑔 ) are the number of zeros
(poles) of 𝑓 and 𝑔 inside 𝛾 counted according to their multiplicities and if

| 𝑓 (𝑧) + 𝑔(𝑧)| < | 𝑓 (𝑧)| + |𝑔(𝑧)|

on 𝛾, then
𝑍𝑓
∑︁  𝑃𝑓
 ∑︁  𝑍𝑔
 ∑︁  𝑃𝑔
 ∑︁  
𝑓 𝑓 𝑔 𝑔
𝑛 𝛾; 𝑧 𝑘 − 𝑛 𝛾; 𝑎 𝑘 = 𝑛 𝛾; 𝑧 𝑘 − 𝑛 𝛾; 𝑎 𝑘 .
𝑘=1 𝑘=1 𝑘=1 𝑘=1

We now proceed with the proof of our generalization noting most of the original
arguments need no change.
From the hypothesis,
𝑓 𝑓
+1 < +1
𝑔 𝑔
on 𝛾. If 𝜆 = 𝑓 /𝑔 and 𝜆 is a positive real number, then we have 𝜆 + 1 < 𝜆 + 1, which
is false. Hence, the meromorphic function 𝑓 /𝑔 maps 𝛾 into Ω = C \[0, ∞]. We can
thus define an analytic branch of the complex logarithm in Ω, say, ℓ(𝑧). This implies
that ℓ( 𝑓 (𝑧)/𝑔(𝑧)) is a well-defined primitive for ( 𝑓 /𝑔) ′ (𝑔/ 𝑓 ) near {𝛾}. Thus,
∫
1
0= ( 𝑓 /𝑔) ′ (𝑔/ 𝑓 )
2𝜋𝑖 𝛾
∫  ′
𝑔′

1 𝑓
= −
2𝜋𝑖 𝛾 𝑓 𝑔
𝑍𝑓
∑︁ 𝑃𝑓  ∑︁ 𝑍 𝑃𝑔 
 𝑓
∑︁
𝑓   𝑔 𝑔
∑︁
𝑔 
=  𝑛(𝛾; 𝑧 𝑘 ) −
 𝑛(𝛾; 𝑎 𝑘 )  −  𝑛(𝛾; 𝑧 𝑘 ) −
 𝑛(𝛾; 𝑎 𝑘 )  ,
 𝑘=1 𝑘=1   𝑘=1 𝑘=1 
   
where the last equality directly follows from the Argument Principle. ♦
Chapter IV
The Maximum Modulus Theorem

Problem IV.1 Prove the following Minimum Principle. If 𝑓 is a non-constant analytic

function on a bounded open set 𝐺 and is continuous on 𝐺, then either 𝑓 has a zero
in 𝐺 or | 𝑓 | assumes its minimum value on 𝜕𝐺.

Solution IV.1 Since 𝑓 is continuous on 𝐺, we must have | 𝑓 | is continuous on 𝐺.

Then there exists an 𝑎 ∈ 𝐺 such that | 𝑓 (𝑎)| ≤ | 𝑓 (𝑧)| for all 𝑧 ∈ 𝐺. If 𝑎 ∈ 𝜕𝐺, then
| 𝑓 | assumes its minimum value on 𝜕𝐺 and we are done. If 𝑎 ∉ 𝜕𝐺, then 𝑎 ∈ 𝐺 and
Ð
we can write 𝐺 = 𝑋𝑖 , where the 𝑋𝑖 ’s are the components of 𝐺. Then 𝑎 ∈ 𝑋𝑖 for
some 𝑖.
But each 𝑋𝑖 is a region, so using Problem – II.5, either 𝑓 (𝑎) = 0 or 𝑓 is constant in
each 𝑋𝑖 (and hence in 𝐺). But 𝑓 was assumed to be non-constant, hence 𝑓 has to
have a zero in 𝐺. Therefore, either 𝑓 has a zero in 𝐺 or | 𝑓 | assumes its minimum
value on 𝜕𝐺 and we are done. ♦

Problem IV.2 Let 𝐺 be a bounded region and suppose 𝑓 is continuous on 𝐺 and

analytic on G. Show that if there is a constant 𝑐 ≥ 0 such that | 𝑓 (𝑧)| = 𝑐 for all 𝑧 on
the boundary of 𝐺 then either 𝑓 is a constant function or 𝑓 has a zero in 𝐺.

Solution IV.2 Let 𝑐 > 0 be a real constant such that | 𝑓 (𝑧)| = 𝑐 for all 𝑧 ∈ 𝜕𝐺. By
the Maximum Modulus Theorem (Second Version), we obtain

max{| 𝑓 (𝑧)| : 𝑧 ∈ 𝐺} = max{| 𝑓 (𝑧)| : 𝑧 ∈ 𝜕𝐺} = 𝑐.

Thus, there exists a complex number 𝜔 ∈ 𝐺 such that

| 𝑓 (𝑧)| ≤ | 𝑓 (𝜔)| = 𝑐

for all 𝑧 ∈ 𝐺. Then by the Maximum Modulus Theorem (First Version), 𝑓 must be
constant in 𝐺.
Now, if 𝑐 = 0, then by the same arguments, there must exist an 𝛼 ∈ 𝐺 such that
| 𝑓 (𝛼)| = 𝑐 = 0. But since 𝑓 is continuous on 𝐺, this would imply that 𝑓 has a zero
in 𝐺 and we are done. ♦

23
24 IV The Maximum Modulus Theorem

Problem IV.3 Suppose 𝐺 is a region, 𝑓 : 𝐺 → C is analytic, and 𝑀 is a constant

such that whenever 𝑧 ∈ 𝜕∞ 𝐺 and {𝑧 𝑛 } is a sequence in 𝐺 with 𝑧 = lim 𝑧 𝑛 , we have
lim sup | 𝑓 (𝑧 𝑛 )| ≤ 𝑀. Show that | 𝑓 (𝑧)| ≤ 𝑀, for each 𝑧 in 𝐺.

Solution IV.3 The proof is by contrapositive. Let 𝑎 ∈ 𝜕∞ 𝐺 and lim sup | 𝑓 (𝑧)| > 𝑀.
𝑧→𝑎
We will show that this implies lim sup | 𝑓 (𝑧 𝑛 )| > 𝑀.
Now, since 𝑧 𝑛 → 𝑧 as 𝑧 → ∞, and since 𝑓 is analytic (continuous), we must have
𝑓 (𝑧 𝑛 ) → 𝑓 (𝑧). Thus,

lim sup | 𝑓 (𝑧)| > 𝑀 =⇒ lim sup | 𝑓 (𝑧 𝑛 )| > 𝑀.

𝑧→𝑎

Then by negation of this statement, we obtain that

lim sup | 𝑓 (𝑧 𝑛 )| ≤ 𝑀 =⇒ lim sup | 𝑓 (𝑧)| ≤ 𝑀

𝑧→𝑎

for all 𝑎 ∈ 𝜕∞ 𝐺, and we are done. ♦

Problem IV.4 Suppose | 𝑓 (𝑧)| ≤ 1 for |𝑧| < 1 and 𝑓 is a non-constant analytic
function. Considering the function 𝑔 : D → D defined by

𝑓 (𝑧) − 𝑎
𝑔(𝑧) = ,
1 − 𝑎¯ 𝑓 (𝑧)

where 𝑎 = 𝑓 (0), prove that

| 𝑓 (0)| − |𝑧| | 𝑓 (0)| + |𝑧|

≤ | 𝑓 (𝑧)| ≤
1 + | 𝑓 (0)||𝑧| 1 − | 𝑓 (0)||𝑧|

for |𝑧| < 1.

Solution IV.4 Considering the given function 𝑔(𝑧) above, we obtain

𝑓 (0) − 𝑎
𝑔(0) =
1 − 𝑎¯ 𝑓 (0)
𝑎−𝑎
= = 0.
1 − 𝑎𝑎
¯
Again, using the triangle and reverse triangle inequalities, we obtain

𝑓 (𝑧) − 𝑎
|𝑔(𝑧)| =
1 − 𝑎¯ 𝑓 (𝑧)
| 𝑓 (𝑧) − 𝑎|
=
|1 − 𝑎¯ 𝑓 (𝑧)|
| 𝑓 (𝑧)| + |𝑎|
≤ ≤ 1,
|1 − | 𝑎||
¯ 𝑓 (𝑧)||
IV The Maximum Modulus Theorem 25

since | 𝑓 | ≤ 1 on D. So, 𝑔 satisfies the hypotheses of Schwarz’s Lemma, and we have

|𝑔(𝑧)| ≤ |𝑧| for |𝑧| < 1. Using this we can write

=⇒ | 𝑓 (𝑧) − 𝑎| ≤ |𝑧| · |1 − 𝑎¯ 𝑓 (𝑧)|

=⇒ | 𝑓 (𝑧)| − |𝑎| ≤ | 𝑓 (𝑧) − 𝑎| ≤ |𝑧| · |1 − 𝑎¯ 𝑓 (𝑧)| ≤ |𝑧| + |𝑧||𝑎| · | 𝑓 (𝑧)|

=⇒ | 𝑓 (𝑧)| − |𝑎| ≤ |𝑧| + |𝑧||𝑎| · | 𝑓 (𝑧)|

=⇒ −|𝑧| − |𝑧||𝑎| · | 𝑓 (𝑧)| ≤ | 𝑓 (𝑧)| − |𝑎| ≤ |𝑧| + |𝑧||𝑎| · | 𝑓 (𝑧)|
|𝑎| − |𝑧| |𝑎| + |𝑧|
=⇒ ≤ | 𝑓 (𝑧)| ≤ .
1 + |𝑎||𝑧| 1 − |𝑎||𝑧|

Since 𝑎 := 𝑓 (0), we are done. ♦

 
1 3
Problem IV.5 Does there exist an analytic function 𝑓 : D → D such that 𝑓 2 = 4
 
and 𝑓 ′ 12 = 23 ?

Solution IV.5 The proof is by contradiction. Assume such an analytic function 𝑓

exists. By Proposition – V.2.2, we know that if 𝛼 = 𝑓 (𝑎), then

1 − |𝛼| 2
| 𝑓 ′ (𝑎)| ≤ .
1 + |𝑎| 2
3
Setting 𝛼 = 4 and 𝑎 = 12 , we obtain

3 2
1−
 
1 4 7 2
𝑓′ ≤ = < .
2 1 2 12 3
1+ 2
 
This is a contradiction since we assumed 𝑓 ′ 1
2 = 23 . ♦

Problem IV.6 Let 𝑓 : 𝐺 → C be analytic and suppose 𝑀 is a constant such that

lim sup | 𝑓 (𝑧 𝑛 )| ≤ 𝑀 for each sequence {𝑧 𝑛 } in 𝐺 which converges to a point in
𝜕∞ 𝐺. Show that | 𝑓 (𝑧)| ≤ 𝑀. (See Problem – IV.3).

Solution IV.6 We have already proved this result for open and connected regions in
Problem – IV.3. We claim that the result holds even if the region is not required to
be connected. The proof, as before, is by contrapositive.
Here, 𝐺 is assumed to be an open set. But every open set can Ð be written as a
union of its components, which are regions in C. That is, 𝐺 = 𝑋𝑖 , where the
𝑋𝑖 areÐcomponents (maximal open connected subsets) of 𝐺. This also gives that
𝜕𝐺 = (𝜕 𝑋𝑖 ).
Now, assume that there exists an 𝑎 ∈ 𝜕∞ 𝐺 such that for every 𝑀 > 0 we have
26 IV The Maximum Modulus Theorem
h i
lim sup | 𝑓 (𝑧)| := lim+ sup{| 𝑓 (𝑧)| : 𝑧 ∈ 𝐺 ∩ 𝐵(𝑎; 𝑟)} > 𝑀.
𝑧→𝑎 𝑟→0

But then there exists a 𝛿 > 0 such that

h i
lim+ sup{| 𝑓 (𝑧)| : 𝑧 ∈ 𝐺 ∩ 𝐵(𝑎; 𝑟)} > 𝑀 + 𝛿.
𝑟→0

Next, if we have a sequence {𝑧 𝑛 } with lim 𝑧 𝑛 = 𝑎 and let 𝑎 𝑛 := 2|𝑧 𝑛 − 𝑎|, then we
have lim 𝑎 𝑛 = 0. Now consider the sequence

𝑟 𝑛 := sup{| 𝑓 (𝑧)| : 𝑧 ∈ 𝐺 ∩ 𝐵(𝑎; 𝑎 𝑛 )}.

Then lim 𝑟 𝑛 > 𝑀 +𝛿. But since {𝑟 𝑛 } is non-increasing, this implies that 𝑟 𝑛 > 𝑀 +𝛿 for
each 𝑛. Thus, there exist {𝑦 𝑛 } such that each 𝑦 𝑛 ∈ 𝐺 ∩ 𝐵(𝑎; 𝑎 𝑛 ) and | 𝑓 (𝑦 𝑛 )| > 𝑀 + 𝛿
for every 𝑛. Then we finally have

lim sup | 𝑓 (𝑦 𝑛 )| ≥ 𝑀 + 𝛿 ≥ 𝑀,
𝑛→∞

where lim 𝑦 𝑛 = 𝑎. By negation, we obtain the desired result. ♦

Chapter V
Compactness and Convergence in the Space of
Analytic Functions

Module – 1

Problem V.1 If (𝑆, 𝑑) is a metric space and if

𝑑 (𝑠, 𝑡)
𝜇(𝑠, 𝑡) = (V.1)
1 + 𝑑 (𝑠, 𝑡)

then prove that (𝑆, 𝜇) is also a metric space.

Solution V.1 Now, since (𝑆, 𝑑) is a metric space, we have

𝑑 (𝑠, 𝑡) ≥ 0
𝑑 (𝑠, 𝑡) = 𝑑 (𝑡, 𝑠)
𝑑 (𝑠, 𝑡) = 0 ⇐⇒ 𝑠 = 𝑡
𝑑 (𝑠, 𝑡) ≤ 𝑑 (𝑠, 𝑢) + 𝑑 (𝑢, 𝑡),

for all 𝑠, 𝑡, 𝑢 ∈ 𝑆. To show (𝑆, 𝜇) is a metric space we must show that 𝜇(𝑠, 𝑡) satisfies
these properties as well. Now, since 𝑑 (𝑠, 𝑡) ≥ 0, we must have

𝑑 (𝑠, 𝑡)
𝜇(𝑠, 𝑡) = ≥ 0.
1 + 𝑑 (𝑠, 𝑡)

Moreover, if 𝜇(𝑠, 𝑡) = 0, then

𝑑 (𝑠, 𝑡)
= 0 =⇒ 𝑑 (𝑠, 𝑡) = 0 ⇐⇒ 𝑠 = 𝑡.
1 + 𝑑 (𝑠, 𝑡)
Again, for any 𝑠, 𝑡 ∈ 𝑆, we have

𝑑 (𝑠, 𝑡) 𝑑 (𝑡, 𝑠)
𝜇(𝑠, 𝑡) = = = 𝜇(𝑡, 𝑠).
1 + 𝑑 (𝑠, 𝑡) 1 + 𝑑 (𝑡, 𝑠)

27
28 V Compactness and Convergence in the Space of Analytic Functions

We are done if we can show that 𝜇(𝑠, 𝑡) satisfies the triangle inequality. Let 𝑠, 𝑡, 𝑢 ∈ 𝑆.
Now, since 𝜇(𝑠, 𝑡) ≥ 0, if 𝑑 (𝑠, 𝑡) ≤ max{𝑑 (𝑠, 𝑢), 𝑑 (𝑢, 𝑡)} then

𝜇(𝑠, 𝑡) ≤ max{𝜇(𝑠, 𝑢), 𝜇(𝑢, 𝑡)}.

This would then imply that

𝜇(𝑠, 𝑡) ≤ 𝜇(𝑠, 𝑢) + 𝜇(𝑢, 𝑡).

Again, if 𝑑 (𝑠, 𝑡) > max{𝑑 (𝑠, 𝑢), 𝑑 (𝑢, 𝑡)} then

 
1 1 1
< min , .
1 + 𝑑 (𝑠, 𝑡) 1 + 𝑑 (𝑠, 𝑢) 1 + 𝑑 (𝑢, 𝑡)

Since 𝑑 (𝑠, 𝑡) also satisfies the triangle inequality, this implies

𝑑 (𝑠, 𝑡) 𝑑 (𝑠, 𝑢) + 𝑑 (𝑢, 𝑡)

𝜇(𝑠, 𝑡) = ≤
1 + 𝑑 (𝑠, 𝑡) 1 + 𝑑 (𝑠, 𝑡)
𝑑 (𝑠, 𝑢) 𝑑 (𝑢, 𝑡)
= +
1 + 𝑑 (𝑠, 𝑡) 1 + 𝑑 (𝑠, 𝑡)
𝑑 (𝑠, 𝑢) 𝑑 (𝑢, 𝑡)
≤ +
1 + 𝑑 (𝑠, 𝑢) 1 + 𝑑 (𝑢, 𝑡)
= 𝜇(𝑠, 𝑢) + 𝜇(𝑢, 𝑡).

So, (𝑆, 𝜇) is a metric space. ♦

Problem V.2 Recall that Proposition – VII.1.2 states that if 𝐺 is an open set then
Ð
there exists a sequence {𝐾𝑛 } of compact subsets of 𝐺 such that 𝐾𝑛 = 𝐺. Moreover,
the sets 𝐾𝑛 can be chosen to satisfy some other properties. Find such sets 𝐾𝑛 for:
(a) 𝐺 an open disk;
(d) 𝐺 an infinite strip.

Solution V.2 Part-(a): Let 𝑎 ∈ C and 𝑟 > 0 such that 𝐺 = 𝐵(𝑎; 𝑟). Consider the
sets  
1
𝐾𝑛 := 𝐵 𝑎; 𝑟 − .
𝑛
1
Here, if 𝑟 − ≤ 0, then we set 𝐾𝑛 = ∅. Then each 𝐾𝑛 is compact and satisfies
𝑛 Ð
𝐾𝑛 ⊂ int 𝐾𝑛+1 . Moreover, we trivially have that 𝐾𝑛 = 𝐺 and so we are done.

Part-(d): Without loss of generality, we assume 𝐺 = {𝑧 ∈ C : |ℑ(𝑧)| < 1}. Let

𝑧 ∈ C is of the form 𝑧 = 𝑥 + 𝑖𝑦, where 𝑥, 𝑦 ∈ R. Consider
 
1
𝐾𝑛 := {𝑧 ∈ C : |𝑥| ≤ 𝑛} ∩ 𝑧 ∈ C : |𝑦| ≤ 1 − .
𝑛
Module – 1 29

Now, since the intersection of two closed sets is closed, each 𝐾𝑛 is closed. Moreover,
Ð 𝐾𝑛 is bounded and hence 𝐾𝑛 is compact for all 𝑛 ∈ N. We also have that
each
𝐾𝑛 = 𝐺 and so we are done. ♦

Problem V.3 (Dini’s Theorem) Consider C(𝐺, R) and suppose that { 𝑓𝑛 } is a se-
quence in C(𝐺, R) which is monotonically increasing (i.e., 𝑓𝑛 (𝑧) ≤ 𝑓𝑛+1 (𝑧) for all
𝑧 ∈ 𝐺) and lim 𝑓𝑛 (𝑧) = 𝑓 (𝑧) for all 𝑧 in 𝐺, where 𝑓 ∈ C(𝐺, R). Show that 𝑓𝑛 → 𝑓 .

Ð we know that there exists a sequence {𝐾𝑛 }

Solution V.3 By Proposition – VII.1.2,
of compact subsets of 𝐺 such that 𝐾𝑛 = 𝐺. Without loss of generality, we can
thus assume that 𝐺 itself is compact. We consider the sequence of functions

𝑔𝑛 := 𝑓 − 𝑓𝑛 .

Since 𝑓𝑛 is monotonically increasing and lim 𝑓𝑛 (𝑥) = 𝑓 (𝑥), 𝑔𝑛 is monotonically

decreasing and lim 𝑔𝑛 (𝑥) = 0. That is, for every 𝜀 > 0, there exists an 𝑁 ∈ N such
that
𝜀
0 ≤ 𝑔 𝑁 (𝑥) <
2
for all 𝑥 ∈ 𝐺. Moreover, each 𝑔𝑛 is continuous since each 𝑓𝑛 and 𝑓 are in C(𝐺, R).
By the continuity of 𝑔𝑛 , we have that for every 𝜀 > 0, there exists an 𝑁 ∈ N such
that
𝑔 𝑁 (𝑦) < 𝜀
for all 𝑦 ∈ 𝐵(𝑥; 𝑟) with 𝑟 < ∞ and any 𝑥 ∈ 𝐺. Since 𝑔 𝑁 is monotonically decreasing,
we must have 𝑔𝑛 < 𝜀 for every 𝑛 ≥ 𝑁 and 𝑦 ∈ 𝐵(𝑥; 𝑟). These open balls form an open
cover for 𝐺. But 𝐺 is compact and hence we must have finitely many 𝑥 1 , 𝑥2 , . . . , 𝑥 𝑘
such that for every 𝜀 > 0, 𝐵(𝑥 𝑖 ; 𝑟) forms a subcover that covers 𝐺 whenever 𝑛 ≥ 𝑁𝑖 .
Thus, we finally have that 𝑔𝑛 < 𝜀 for every 𝑛 ≥ max{𝑁1 , . . . , 𝑁 𝑘 }. This implies that

| 𝑓𝑛 (𝑥) − 𝑓 (𝑥)| < 𝜀

for every 𝜀 > 0 and thus, 𝑓𝑛 → 𝑓 . ♦

Problem V.4 Let { 𝑓𝑛 } ⊂ C(𝐺, Ω) and suppose that { 𝑓𝑛 } is equicontinuous at each

point of 𝐺. If 𝑓 ∈ C(𝐺, Ω) and 𝑓 (𝑧) = lim 𝑓𝑛 (𝑧) for each 𝑧 then show that 𝑓𝑛 → 𝑓 .
Solution V.4 Using the same argument as in the previous exercise, we can assume
without loss of generality that 𝐺 is compact. Since 𝑓𝑛 is equicontinuous at each
𝑧 ∈ 𝐺, we have that for every 𝜀 > 0 there exists a 𝛿 > 0 such that
𝜀
𝑑 ( 𝑓𝑛 (𝑥), 𝑓𝑛 (𝑦)) <
3
whenever 𝑑 (𝑥, 𝑦) < 𝛿. Since 𝑓 (𝑧) = lim 𝑓𝑛 (𝑧), we also have
𝜀
𝑑 ( 𝑓 (𝑥), 𝑓 (𝑦)) <
3
30 V Compactness and Convergence in the Space of Analytic Functions

whenever 𝑑 (𝑥, 𝑦) < 𝛿. As in the previous exercise, we can find finitely many points
𝑦 1 , . . . , 𝑦 𝑘 ∈ 𝐺 such that the open balls 𝐵(𝑦 𝑖 ; 𝛿) cover 𝐺, since 𝐺 is compact. Since
𝑓𝑛 → 𝑓 (pointwise), we thus have for every 𝑛 ≥ 𝑁𝑖
𝜀
𝑑 ( 𝑓𝑛 (𝑦 𝑖 ), 𝑓 (𝑦 𝑖 )) < .
3
Finally, for an arbitrary 𝑧 ∈ 𝐺 and 𝑛 ≥ max{𝑁1 , . . . , 𝑁 𝑘 }, we have
𝜀
𝑑 ( 𝑓𝑛 (𝑧), 𝑓 (𝑧)) ≤ 𝑑 ( 𝑓𝑛 (𝑧), 𝑓𝑛 (𝑦 𝑖 )) + 𝑑 ( 𝑓𝑛 (𝑦 𝑖 ), 𝑓 (𝑦 𝑖 )) + 𝑑 ( 𝑓 (𝑦 𝑖 ), 𝑓 (𝑧)) < 3 · = 𝜀.
3
This then means 𝑓𝑛 → 𝑓 and we are done. ♦

Problem V.5 Prove Vitali’s Theorem: If 𝐺 is a region and { 𝑓𝑛 } ⊂ H (𝐺) is locally

bounded and 𝑓 ∈ H (𝐺) that has the property that 𝐴 = {𝑧 ∈ 𝐺 : lim 𝑓𝑛 (𝑧) = 𝑓 (𝑧)}
has a limit point in 𝐺 then 𝑓𝑛 → 𝑓 . [H (𝐺) is the space of all analytic functions
from 𝐺 to C.]

Solution V.5 The proof is by contradiction. We assume 𝑓𝑛 ̸→ 𝑓 in H (𝐺). The

sequence { 𝑓𝑛 } is a locally bounded sequence in H (𝐺). Then by Montel’s Theorem,
{ 𝑓𝑛 } is normal. So, there is a subsequence { 𝑓𝑛𝑘 } ⊂ { 𝑓𝑛 } that converges to 𝑓 in
H (𝐺). Now, since 𝑓𝑛 ̸→ 𝑓 , there must be a compact set 𝐾 ⊂ 𝐺 such that the
convergence is not uniform on 𝐾. Thus, there must be an 𝜀 > 0 such that for all
𝑛 ∈ N there is 𝑧 𝑛 ∈ 𝐾 with

| 𝑓𝑛 (𝑧 𝑛 ) − 𝑓 (𝑧 𝑛 )| ≥ 𝜀.

By compactness of 𝐾 we choose a subsequence of 𝑧 𝑛 , say, 𝑧 𝑛𝑚 so that it converges

to a point 𝑧 0 ∈ 𝐾.
Here, since { 𝑓𝑛 } is locally bounded, the subsequence { 𝑓𝑛𝑚 } is locally bounded. Again
by Montel’s Theorem, there is an analytic function in H (𝐺) such that 𝑓𝑛𝑚 → 𝑔 in
H (𝐺). If 𝐴 is the set of points of pointwise convergence 𝑓𝑛 (𝑧) → 𝑓 and 𝑓𝑛 (𝑧) → 𝑔.
Now since 𝐺 is a region and 𝐴 has a limit point in 𝐺, by Corollary – IV.3.8, we have
𝑓 ≡ 𝑔 on 𝐺, which gives a contradiction on the set 𝐾 and the point 𝑧0 because

| 𝑓𝑛𝑚 (𝑧 𝑛𝑚 ) − 𝑓 (𝑧 𝑛𝑚 )| → |𝑔(𝑧0 ) − 𝑓 (𝑧 0 )| ≥ 𝜀.

So, we must have that 𝑓𝑛 → 𝑓 in H (𝐺). ♦

Problem V.6 Let D = 𝐵(0; 1) and for 0 < 𝑟 < 1, let 𝛾𝑟 (𝑡)∫= 𝑟e2 𝜋𝑖𝑡 , 𝑡 ∈ [0, 1]. Show
that a sequence { 𝑓𝑛 } in H (D) converges to 𝑓 if and only if 𝛾 | 𝑓 (𝑧)− 𝑓𝑛 (𝑧)| |𝑑𝑧| → 0
𝑟
as 𝑛 → ∞ for each 𝑟, 0 < 𝑟 < 1.

Solution V.6 We first assume lim 𝑓𝑛 (𝑧) = 𝑓 (𝑧) for all 𝑧 ∈ D. Then we have
Module – 1 31
∫ ∫
| 𝑓 (𝑧) − 𝑓𝑛 (𝑧)| |𝑑𝑧| ≤ sup {| 𝑓 (𝑧) − 𝑓𝑛 (𝑧)|} |𝑑𝑧|
𝛾𝑟 𝑧 ∈𝛾𝑟 𝛾𝑟
= 2𝜋𝑟 · sup {| 𝑓 (𝑧) − 𝑓𝑛 (𝑧)|}.
𝑧 ∈𝛾𝑟

But sup {| 𝑓 (𝑧) − 𝑓𝑛 (𝑧)|} → 0 and thus, we have

𝑧 ∈𝛾𝑟
∫
| 𝑓 (𝑧) − 𝑓𝑛 (𝑧)| |𝑑𝑧| → 0.
𝛾𝑟
∫
Next, we assume 𝛾 | 𝑓 (𝑧) − 𝑓𝑛 (𝑧)| |𝑑𝑧| → 0 as 𝑛 → ∞ for each 𝑟, 0 < 𝑟 < 1. Let
𝑟
𝐾 be some compact subset of D and choose 0 < 𝑟 < 1 such that 𝐾 ⊂ 𝐵(0; 𝑟) and
define 𝛿 > 0 by
𝛿 = min{𝑑 (𝑥, 𝑦) : 𝑥 ∈ 𝐾, 𝑦 ∈ 𝐵(0; 𝑟)}.
By the assumption, we have
∫
| 𝑓 (𝑧) − 𝑓𝑛 (𝑧)| |𝑑𝑧| < 2𝜋𝜀𝛿.
𝜕𝐵

If 𝑎 ∈ 𝐾, by Cauchy’s Integral Formula,

∫
1 𝑓 (𝑧) − 𝑓𝑛 (𝑧)
𝑓 (𝑎) − 𝑓𝑛 (𝑎) = 𝑑𝑧.
2𝜋𝑖 𝜕𝐵 𝑧−𝑎
Thus, we have
∫
1 | 𝑓 (𝑧) − 𝑓𝑛 (𝑧)|
| 𝑓 (𝑎) − 𝑓𝑛 (𝑎)| ≤ |𝑑𝑧|
2𝜋 𝜕𝐵 |𝑧 − 𝑎|
∫
1
≤ | 𝑓 (𝑧) − 𝑓𝑛 (𝑧)||𝑑𝑧|
2𝜋𝛿 𝜕𝐵
2𝜋𝜀𝛿
< = 𝜀.
2𝜋𝛿
Thus, 𝑓𝑛 (𝑎) → 𝑓 (𝑎) for all 𝑎 ∈ 𝐾. Since 𝐾 was arbitrary, we must have that
{ 𝑓𝑛 (𝑧)} → 𝑓 (𝑧) for all 𝑧 ∈ D . ♦

Problem V.7 Let { 𝑓𝑛 } ⊂ H (𝐺) be a sequence of one-one functions which converge

to 𝑓 . If 𝐺 is a region, show that either 𝑓 is one-one or 𝑓 is a constant function.

Solution V.7 Assume { 𝑓𝑛 } ⊂ H (𝐺) is a sequence of one-one functions which

converge to 𝑓 where 𝐺 is a region. Suppose 𝑓 is not the constant function. Then,
we have to show that 𝑓 is one-one. Choose an arbitrary point 𝑧 0 ∈ 𝐺 and define the
sequence
𝑔𝑛 (𝑧) = 𝑓𝑛 (𝑧) − 𝑓𝑛 (𝑧0 ).
32 V Compactness and Convergence in the Space of Analytic Functions

It is clear that the sequence {𝑔𝑛 } converges to 𝑔(𝑧) = 𝑓 (𝑧) − 𝑓 (𝑧0 ) on the open
connected set 𝐺 \ {𝑧 0 }, which is again a region. We also have that 𝑔𝑛 never vanishes
on 𝐺 \ {𝑧0 }, since 𝑓𝑛 are one-to-one functions. Therefore, the sequence {𝑔𝑛 } satisfies
the hypotheses of Corollary – VII.2.6, where the region is 𝐺 \ {𝑧 0 }. Thus, either
𝑔 ≡ 0 or 𝑔 never vanishes. Since 𝑔(𝑧) = 𝑓 (𝑧) − 𝑓 (𝑧0 ) is not identically zero on
𝐺 \ {𝑧0 }, we must have 𝑔(𝑧) = 𝑓 (𝑧) − 𝑓 (𝑧 0 ) has no zeros in 𝐺 \ {𝑧0 }. This implies

𝑓 (𝑧) ≠ 𝑓 (𝑧 0 )

for all 𝑧 ∈ 𝐺 \ {𝑧 0 }. But this in turn implies that 𝑓 is one-to-one on 𝐺. ♦

Problem V.8 Prove Proposition – 3.3(b): If 𝜌 > 0 is given and 𝑎 ∈ C then there is a
number 𝑟 > 0 such that 𝐵(𝑎; 𝑟) ⊂ 𝐵∞ (𝑎; 𝜌).
𝛿
Solution V.8 Let 𝑎 ∈ C. Assume 𝛿 > 0 is given. We pick an 𝑟 > 0 such that 𝑟 < .
2
Now, if 𝑧 ∈ 𝐵(𝑎; 𝑟) we have

|𝑧 − 𝑎| < 𝑟
𝛿
< ·1
2
𝛿 √︁ √︁
≤ 1 + |𝑧| 2 1 + |𝑎| 2
2
2|𝑧 − 𝑎|
=⇒ 𝛿 > √︁ √︁
1 + |𝑧| 2 1 + |𝑎| 2
∴ 𝑑 (𝑧, 𝑎) < 𝛿.

This implies that 𝑧 ∈ 𝐵∞ (𝑎; 𝜌), which implies that

𝐵(𝑎; 𝑟) ⊂ 𝐵∞ (𝑎; 𝜌). ♦

Problem V.9 Prove Proposition – 3.3(d): If a compact set 𝐾 ⊂ C is given, there is a

number 𝑝 > 0 such that 𝐵∞ (∞; 𝜌) ⊂ C∞ \𝐾.

Solution V.9 Let 𝑟 > 0 and 𝐾 ⊂ C, where 𝐾 is compact with 𝐾 ⊂ 𝐵(0; 𝑟). We
trivially have
C∞ \𝐵(0; 𝑟) ⊂ C∞ \𝐾.
So, it is enough to show that for a given 𝑟 > 0, there exists a 𝛿 > 0 such that

𝐵∞ (∞; 𝛿) ⊂ C∞ \𝐵(0; 𝑟).

2
To show this we choose 𝛿 < √ . We then have
1 + 𝑟2
Module – 2 33

𝐵∞ (∞; 𝛿) = {𝑧 ∈ C : 𝑑 (∞, 𝑧) < 𝛿} ∪ {∞}

( )
2
= 𝑧 ∈ C : √︁ < 𝛿 ∪ {∞}
1 + |𝑧| 2
( )
2 2
⊂ 𝑧 ∈ C : √︁ < √ ∪ {∞}
1 + |𝑧| 2 1 + 𝑟2
= {𝑧 ∈ C : 𝑟 ≤ |𝑧|} ∪ {∞}
= C ∪{𝑧 ∈ C : |𝑧| < 𝑟} 𝑐 ∪ {∞}
= C∞ \{𝑧 ∈ C : |𝑧| < 𝑟}
= C∞ \𝐵(0; 𝑟).

We then have
𝐵∞ (∞; 𝛿) ⊂ C∞ \𝐵(0; 𝑟)
and so we are done. ♦

Module – 2

Problem V.10 Let 𝑓 be analytic on 𝐺 = {𝑧 : ℜ(𝑧) > 0}, one-one, with ℜ[ 𝑓 (𝑧)] > 0
for all 𝑧 in 𝐺, and 𝑓 (𝑎) = 𝑎 for some real number 𝑎. Show that | 𝑓 ′ (𝑎)| ≤ 1.

Solution V.10 Here, we notice 𝐺 is a ’proper’ simply connected region in C. If

𝑎 ∈ 𝐺 then by the Riemann Mapping Theorem, there is a unique analytic function
𝑔 : 𝐺 → D with the properties:
(i) 𝑔(𝑎) = 0 and 𝑔 ′ (𝑎) > 0;
(ii) 𝑔 is one-to-one;
(iii) 𝑔(𝐺) = D.
Now, since 𝑔 : 𝐺 → D is bijective, 𝑔 −1 : D → 𝐺 is also bijective, and since
𝑓 : 𝐺 → 𝐺 is also one-to-one, we can define a one-to-one analytic function ℎ(𝑧) by

ℎ(𝑧) = 𝑔 ◦ 𝑓 ◦ 𝑔 −1 (𝑧).

We note ℎ(𝑧) is also bijective with ℎ(D) = D. Thus, we have

ℎ(0) = 𝑔 ◦ 𝑓 ◦ 𝑔 −1 (0) = 𝑔 ◦ 𝑓 (𝑎) = 𝑔(𝑎) = 0,

where we have used the facts that 𝑔 −1 (0) = 𝑓 (𝑎) = 𝑎. By definition, we also have

|ℎ(𝑧)| ≤ 1

for all 𝑧 ∈ D. Applying Schwarz’s Lemma to ℎ(𝑧), we now have

|ℎ′ (0)| ≤ 1.
34 V Compactness and Convergence in the Space of Analytic Functions

Taking the derivative of ℎ(𝑧), using the Chain Rule, we obtain

 ′
ℎ′ (𝑧) = 𝑔 ◦ 𝑓 ◦ 𝑔 −1 (𝑧)
h i h i h i′
= 𝑔 ′ ◦ 𝑓 ◦ 𝑔 −1 (𝑧) · 𝑓 ′ ◦ 𝑔 −1 (𝑧) · 𝑔 −1 (𝑧)
h i h i 1
= 𝑔 ′ ◦ 𝑓 ◦ 𝑔 −1 (𝑧) · 𝑓 ′ ◦ 𝑔 −1 (𝑧) · ′ −1 ,
𝑔 (𝑔 (𝑧))

where the last equality follows from Proposition – III.2.20 when 𝑔 ′ (𝑔 −1 (𝑧)) ≠ 0.
Next, we evaluate ℎ′ at 𝑧 = 0 as follows
h i h i 1
ℎ′ (0) = 𝑔 ′ ◦ 𝑓 ◦ 𝑔 −1 (0) · 𝑓 ′ ◦ 𝑔 −1 (0) · ′ −1
𝑔 (𝑔 (0))
h i h i
𝑔 ′ ◦ 𝑓 (𝑎) · 𝑓 ′ (𝑎)
=
𝑔 ′ (𝑎)
𝑔 ′ (𝑎) · 𝑓 ′ (𝑎)
= = 𝑓 ′ (𝑎).
𝑔 ′ (𝑎)

We note the last equality is well-defined since 𝑔 ′ (𝑎) > 0. But combining this result
with our earlier result, we have |ℎ′ (𝑎)| = | 𝑓 ′ (𝑎)| ≤ 1 and we are done. ♦

Problem V.11 Let 𝐺 1 and 𝐺 2 be simply connected regions neither of which is the
whole plane. Let 𝑓 be a one-one analytic mapping of 𝐺 1 onto 𝐺 2 . Let 𝑎 ∈ 𝐺 1 and put
𝛼 = 𝑓 (𝑎). Prove that for any one-one analytic map ℎ of 𝐺 1 into 𝐺 2 with ℎ(𝑎) = 𝛼 it
follows that |ℎ′ (𝑎)| ≤ | 𝑓 ′ (𝑎)|. What can be said if ℎ is not assumed to be one-one?

Solution V.11 First, we define a one-to-one analytic function 𝐹 (𝑧) : 𝐺 1 → 𝐺 1 by

𝐹 (𝑧) = 𝑓 −1 ◦ ℎ(𝑧).

Here, 𝐹 (𝑧) well-defined since 𝑓 : 𝐺 1 → 𝐺 2 is bijective and ℎ : 𝐺 1 → 𝐺 2 is

one-to-one. We also have for any 𝑎 ∈ 𝐺 1

𝐹 (𝑎) = 𝑓 −1 ◦ ℎ(𝑎) = 𝑓 −1 (𝛼) = 𝑎.

Thus, by the previous exercise (Problem – V.10), we obtain |𝐹 (𝑎)| ≤ 1. But we also
have by differentiating 𝐹 (𝑧),
 ′
𝐹 ′ (𝑧) = 𝑓 −1 ◦ ℎ(𝑧)
h i′
= 𝑓 −1 (ℎ(𝑧)) · ℎ′ (𝑧)
ℎ′ (𝑧)
= ,
𝑓 ′ ( 𝑓 −1◦ ℎ(𝑧))
Module – 2 35

where the last equality follows from Proposition – III.2.20. Evaluating 𝐹 ′ at 𝑧 = 𝑎,

we obtain
ℎ′ (𝑎)
𝐹 ′ (𝑎) =
𝑓 ′ ( 𝑓 −1
◦ ℎ(𝑎))
ℎ′ (𝑎)
= ′ −1
𝑓 ( 𝑓 (𝛼))
ℎ′ (𝑎)
= ′ ,
𝑓 (𝑎)

which is well-defined since 𝑓 ′ (𝑎) ≠ 0 (by injectivity). Finally, combining this with
our previous result, we directly have |ℎ′ (𝑎)| ≤ | 𝑓 ′ (𝑎)|.
If ℎ(𝑧) is not injective, we still must have the same result since we do not require the
assumption that 𝐹 (𝑧) is injective in Problem – V.10. ♦

Problem V.12 Let 𝑟 1 , 𝑟 2 , 𝑅1 , 𝑅2 be positive numbers such that 𝑅1 /𝑟 1 = 𝑅2 /𝑟 2 ; Show

that ann(0; 𝑟 1 , 𝑅1 ) and ann(0; 𝑟 2 , 𝑅2 ) are conformally equivalent.

Solution V.12 We notice that the function 𝑓 : ann(0; 𝑟 1 , 𝑅1 ) → ann(0; 𝑟 2 , 𝑅2 ) de-

fined by
𝑟2
𝑓 (𝑧) = 𝑧,
𝑟1
is analytic and maps the annulus ann(0; 𝑟 1 , 𝑅1 ) bijectively onto the annulus
ann(0; 𝑟 2 , 𝑅2 ). In particular, 𝑓 maps the inner boundary of the domain to the in-
ner boundary of the range, that is,

𝑓 ({𝑧 : |𝑧| = 𝑟 1 }) = {𝑧 : |𝑧| = 𝑟 2 }.

𝑅2 𝑟 1
As for the outer boundaries, we note that since 𝑅1 = , if |𝑧| = 𝑅1 , then
𝑟2
𝑟 2 𝑅2 𝑟 1
| 𝑓 (𝑧)| = · = 𝑅2 .
𝑟1 𝑟2
So, 𝑓 maps the outer boundary of the domain onto the outer boundary of the range.
Now, the inverse function of 𝑓 is
𝑟1
𝑓 −1 (𝑧) = 𝑧.
𝑟2

We observe that 𝑓 −1 is also a non-constant analytic function. Therefore, the annulus

ann(0; 𝑟 1 , 𝑅1 ) is conformally equivalent to the annulus ann(0; 𝑟 2 , 𝑅2 ). ♦
36 V Compactness and Convergence in the Space of Analytic Functions

log(1 + 𝑧)
Problem V.13 Prove that lim = 1.
𝑧→0 𝑧
log(1 + 𝑧)
Solution V.13 We first note that the function 𝑓 (𝑧) := has a removable
𝑧
singularity at 𝑧 = 0 since

lim 𝑧 𝑓 (𝑧) = lim log(1 + 𝑧) = log(1) = 0.

𝑧→0 𝑧→0

So, 𝑓 (𝑧) can be defined to be analytic at 𝑧 = 0. We know

∞
∑︁ 𝑧𝑘
log(1 + 𝑧) = (−1) 𝑘+1 ,
𝑘=1
𝑘

where |𝑧| < 1. Then we have

∞ ∞ ∞
log(1 + 𝑧) ∑︁ 𝑧 𝑘−1 ∑︁ 𝑧 𝑘−1 ∑︁ 𝑧𝑘
= (−1) 𝑘+1 =1+ (−1) 𝑘+2 =1+ (−1) 𝑘 .
𝑧 𝑘=1
𝑘 𝑘=2
𝑘 𝑘=1
𝑘 +1

We note that the above sum is an analytic function. Thus,

∞ ∞
𝑧𝑘 0
♦
∑︁ ∑︁
lim 𝑓 (𝑧) = 1 + lim (−1) 𝑘 =1+ (−1) 𝑘 = 1.
𝑧→0 𝑧→0
𝑘=1
𝑘 +1 𝑘=1
𝑘 +1

∞  
Ö 1 1
Problem V.14 Prove that 1− 2 = .
𝑛=2
𝑛 2

Solution V.14 First, we consider the finite product,

𝑛   Ö 𝑛   
Ö 1 1 1
𝑃𝑛 = 1− 2 = 1− 1+
𝑘=2
𝑘 𝑘=2
𝑘 𝑘
" 𝑛   #
Ö 1 1
= exp log 1− 1+
𝑘=2
𝑘 𝑘
" 𝑛     #
∑︁ 1 1
= exp log 1 − 1+
𝑘=2
𝑘 𝑘
" 𝑛   #
∑︁ 𝑘 −1 𝑘 +1
= exp log
𝑘=2
𝑘 𝑘
" 𝑛 # " 𝑛 #
∑︁ ∑︁
= exp [log(𝑘 − 1) − log(𝑘)] + exp [log(𝑘 + 1) − log(𝑘)]
𝑘=2 𝑘=2
Module – 2 37
   
= exp log(1) − log(𝑛) + exp log(𝑛 + 1) − log(2)
 
= exp log(𝑛 + 1) − log(2𝑛)
  
𝑛+1 𝑛+1
= exp log = .
2𝑛 2𝑛

Now, by definition, the infinite product is the limit of 𝑃𝑛 as 𝑛 → ∞. Taking the limit,
we obtain
∞  
1 𝑛+1 1
♦
Ö
1 − 2 = lim 𝑃𝑛 = lim = .
𝑘=2
𝑘 𝑛→∞ 𝑛→∞ 2𝑛 2

For the next few exercises, we first state a few definitions. Let 𝐺 be a region and
I ⊆ H (𝐺).
(i) I is called an ideal iff : (a) 𝑓 , 𝑔 ∈ I implies 𝑎 𝑓 + 𝑏𝑔 ∈ I for all 𝑎, 𝑏 ∈ C;
(b) 𝑓 ∈ I and 𝑔 ∈ H (𝐺) implies 𝑓 𝑔 ∈ I .
(ii) I is called a proper ideal iff I ≠ (0) and I ≠ H (𝐺).
(iii) I is a maximal ideal iff I is a proper ideal and whenever J (fancy version
of the letter ’J’) is an ideal with I ⊆ J , this implies I = J or J = H (𝐺).
(iv) I is a prime ideal if whenever 𝑓 , 𝑔 ∈ H (𝐺) and 𝑓 𝑔 ∈ I then either 𝑓 ∈ I
or 𝑔 ∈ I .
(v) If 𝑓 ∈ H (𝐺) then Z ( 𝑓 ) denotes the set of zeros of 𝑓 counted according to
multiplicity.
(vi) If S ⊂ H (𝐺) then Z (S ) := {Z ( 𝑓 ) : 𝑓 ∈ S }, where the zeros are again
Ñ
counted according to multiplicity.
(vii) If 𝐴 ⊂ 𝐺, then I ( 𝐴) := { 𝑓 ∈ H (𝐺) : 𝐴 ⊂ Z ( 𝑓 )}.
(viii) If 𝑓 , 𝑔 ∈ H (𝐺) then 𝑓 divides 𝑔 (in symbols, 𝑓 |𝑔) if there is an ℎ in H (𝐺)
such that 𝑔 = 𝑓 ℎ.
(ix) If S ⊂ H (𝐺) and S contains a non-zero function then 𝑓 is the greatest
common divisor of S if: (a) 𝑓 |𝑔 for each 𝑔 ∈ S ; (b) whenever ℎ|𝑔 for all
𝑔 ∈ S , we have ℎ| 𝑓 . We write 𝑓 = 𝑔.𝑐.𝑑.S .

Problem V.15 Prove that 𝑓 |𝑔 iff Z ( 𝑓 ) ⊂ Z (𝑔).

Solution V.15 First, we assume Z ( 𝑓 ) ⊂ Z (𝑔). Then every zero of 𝑓 of multiplicity

𝑚 is also a zero of 𝑔 of multiplicity at least 𝑚. We observe that to show 𝑓 |𝑔 it is
𝑔
enough to show that the function ℎ = is analytic in 𝐺. But by the above reasoning,
𝑓
ℎ is analytic in 𝐺. Hence, 𝑔 = 𝑓 ℎ and we have 𝑓 |𝑔.
Next, we assume that 𝑓 |𝑔. This implies that there exists an analytic function in 𝐺
with 𝑔 = 𝑓 ℎ. But then we have

Z (𝑔) = Z ( 𝑓 ) ∪ Z (ℎ),

which implies that

Z ( 𝑓 ) ⊂ Z (𝑔). ♦
38 V Compactness and Convergence in the Space of Analytic Functions

Problem V.16 Prove that 𝑓 = 𝑔.𝑐.𝑑.S iff Z ( 𝑓 ) = Z (S ).

Solution V.16 Let 𝑓 = 𝑔.𝑐.𝑑.S . Then by definition, 𝑓 |𝑔 for all 𝑔 ∈ S . By Problem

– V.15, this implies
Z ( 𝑓 ) ⊂ Z (𝑔)
for all 𝑔 ∈ S . But then we have
Ù
Z (𝑓) = {Z (𝑔) : 𝑔 ∈ S } =: Z (S ).

Next, assume that Z ( 𝑓 ) = Z (S ). Then Z ( 𝑓 ) ⊂ Z (𝑔) for all 𝑔 ∈ S . By Problem

– V.15, this gives 𝑓 |𝑔. Now, let ℎ|𝑔 for all 𝑔 ∈ S . We are done if we can show that
ℎ| 𝑓 . Since ℎ|𝑔, ∀𝑔 ∈ S , by Problem – V.15,

Z (ℎ) ⊂ Z (𝑔)

for all 𝑔 ∈ S . By definition, this implies

Ù
Z (ℎ) = {Z (𝑔) : 𝑔 ∈ S } =: Z (S ).

Problem V.17 Show that every maximal ideal in H (𝐺) is a prime ideal.

Solution V.17 The proof is by contrapositive. Let I be a maximal ideal in H (𝐺).

Assume that 𝑓 , 𝑔 ∈ H (𝐺) with 𝑓 , 𝑔 ∉ I . Then since I is a maximal ideal, we can
write
H (𝐺) = I + { 𝑓 ℎ1 : ℎ1 ∈ H (𝐺)},
and
H (𝐺) = I + {𝑔ℎ2 : ℎ2 ∈ H (𝐺)}.
We now notice that

1 = 𝑚 + 𝑓 ℎ1
1 = 𝑛 + 𝑔ℎ2 ,

where 𝑚, 𝑛 ∈ I . Multiplying, we obtain

1 = 𝑚𝑛 + 𝑚𝑔ℎ2 + 𝑛 𝑓 ℎ1 + 𝑓 𝑔ℎ1 ℎ2 .

Here, the analytic function 𝑚𝑛 + 𝑚𝑔ℎ2 + 𝑛 𝑓 ℎ1 is in I . Since 1 ∉ I , we must then

have that
𝑓 𝑔ℎ1 ℎ2 ∉ I .
By contraposition, we get the desired result. ♦
Module – 2 39

Problem V.18 Give an example of an ideal which is not a prime ideal.

Solution V.18 Let 𝑧1 , 𝑧2 ∈ 𝐺. Consider the set

J := { 𝑓 ∈ H (𝐺) : 𝑓 (𝑧1 ) = 0 and 𝑓 (𝑧2 ) = 0}.

We notice that if 𝑓 , 𝑔 ∈ J and 𝑎, 𝑏 ∈ C, then (𝑎 𝑓 + 𝑏𝑔) ∈ J since

(𝑎 𝑓 + 𝑏𝑔) (𝑧1 ) = 𝑎 𝑓 (𝑧1 ) + 𝑏𝑔(𝑧1 ) = 0 + 0 = 0,

and
(𝑎 𝑓 + 𝑏𝑔) (𝑧 2 ) = 𝑎 𝑓 (𝑧2 ) + 𝑏𝑔(𝑧2 ) = 0 + 0 = 0.
Again, if 𝑓 ∈ J and 𝑔 ∈ H (𝐺), then 𝑓 𝑔 ∈ J since

( 𝑓 𝑔) (𝑧 1 ) = 𝑓 (𝑧 1 )𝑔(𝑧1 ) = 0 · 0 = 0,

and
( 𝑓 𝑔) (𝑧2 ) = 𝑓 (𝑧2 )𝑔(𝑧 2 ) = 0 · 0 = 0.
Thus, J is an ideal in H (𝐺). Next, we observe that if 𝑓 𝑔 ∈ J , then none of them
is necessarily in the ideal J . For instance, consider 𝑓 (𝑧) = 𝑧 − 𝑧 1 and 𝑔(𝑧) = 𝑧 − 𝑧2 .
Then 𝑓 (𝑧)𝑔(𝑧) = (𝑧 − 𝑧1 ) (𝑧 − 𝑧2 ) is in J , but clearly neither 𝑓 nor 𝑔 is in J .
Hence, J is an ideal which is not a prime ideal. ♦

Problem V.19 Find an entire function 𝑓 such that 𝑓 (𝑛 + 𝑖𝑛) = 0 for every integer 𝑛.

Solution V.19 Let (𝑎 𝑛 ) := 𝑛 + 𝑖𝑛. First, we observe that

∞  2 ∞
∑︁ 𝑟 ∑︁ 𝑟2
= <∞
𝑛=1
|𝑛 + 𝑖𝑛| 𝑛=1
2𝑛2

for all 𝑟 > 0. Then by the Weierstrass Factorization Theorem for 𝐺 = C, we obtain
that the function
Ö∞
𝑓 (𝑧) = 𝐸 2 (𝑧/𝑎 𝑛 )
𝑛=1

is entire and have zeros at 𝑧 = 𝑎 𝑛 for each integer 𝑛. Here, 𝐸 2 (𝑧/𝑎 𝑛 ) is the second
order Elementary Factor evaluated at 𝑧 0 = 𝑧/𝑎 𝑛 . By definition,

𝑧2
   
𝑧 𝑧
𝐸 2 (𝑧/𝑎 𝑛 ) = 1 − exp + .
𝑎𝑛 𝑎 𝑛 2𝑎 2𝑛

Thus, our desired function is given by

∞ 
𝑧2
  
𝑧 𝑧
♦
Ö
𝑓 (𝑧) = 1− exp + .
𝑛=1
𝑎𝑛 𝑎 𝑛 2𝑎 2𝑛
40 V Compactness and Convergence in the Space of Analytic Functions

Problem V.20 Find a factorization for cos(𝜋𝑧).

Solution V.20 We recall that the complex sine and cosine functions satisfy the well-
known double-angle identity

sin(2𝜋𝑧) = 2 sin(𝜋𝑧) cos(𝜋𝑧).

Since a factorization for sin(𝜋𝑧) is known, we can use the above identity to derive a
factorization for cos(𝜋𝑧). We recall
∞ 
𝑧2
Ö 
sin(𝜋𝑧) = 𝜋𝑧 1− 2 .
𝑛=1
𝑛
∞ 
4𝑧2
Ö 
∴ sin(2𝜋𝑧) = 2𝜋𝑧 1− 2 .
𝑛=1
𝑛

Substituting these in the double-angle identity, we obtain

∞  ∞ 
4𝑧 2 𝑧2
Ö  Ö 
2𝜋𝑧 1− = 2𝜋𝑧 cos(𝜋𝑧) 1 −
𝑛=1
𝑛2 𝑛=1
𝑛2
∞  ∞   ∞ 
𝑧2 4𝑧2 Ö 4𝑧2
Ö  Ö 
=⇒ cos(𝜋𝑧) 1− 2 = 1− 1 − ,
𝑛=1
𝑛 𝑚=1
(2𝑚) 2 𝑚=1 (2𝑚 − 1) 2

where the last equality holds since we know that rearrangement of terms does not
alter the convergence of an infinite product. Continuing, we further obtain
∞  ∞   ∞ 
𝑧2 𝑧2 Ö 4𝑧 2
Ö  Ö 
=⇒ cos(𝜋𝑧) 1−= 1 − 1 −
𝑛=1
𝑛2 𝑛=1
𝑛2 𝑛=1 (2𝑛 − 1) 2
∞ 
4𝑧2

♦
Ö
⇐⇒ cos(𝜋𝑧) = 1− .
𝑛=1
(2𝑛 − 1) 2
Module – 2 41

Problem V.21 Find a factorization for sinh 𝑧.

Solution V.21 We first note that

1 𝑧
sinh 𝑧 = (𝑒 − 𝑒 −𝑧 )
2
1  −𝑖2 𝑧 2

= 𝑒 − 𝑒 − (−𝑖 ) 𝑧
2
−𝑖  𝑖 (𝑖𝑧) 
= 𝑒 − 𝑒 −𝑖 (𝑖𝑧)
2𝑖
∴ sinh 𝑧 = (−𝑖) sin(𝑖𝑧).

Now, by Equation – VII.6.2 (from the text), we know

∞ 
𝑧2
Ö 
sin(𝜋𝑧) = 𝜋𝑧 1− .
𝑛=1
𝑛2

This implies that

∞ 
𝑧2
Ö 
sin 𝑧 = 𝑧 1− 2 2 .
𝑛=1
𝜋 𝑛
Substituting 𝑖𝑧 into the above identity, we obtain
∞  ∞ 
𝑖2 𝑧2 𝑧2
 
♦
Ö Ö
sinh 𝑧 = (−𝑖) sin(𝑖𝑧) = (−𝑖) (𝑖𝑧) 1− 2 2 = 𝑧 1+ 2 2 .
𝑛=1
𝜋 𝑛 𝑛=1
𝜋 𝑛

Problem V.22 Prove Wallis’ Formula:

∞
𝜋 Ö 4𝑛2
= .
2 𝑛=1 4𝑛2 − 1

Solution V.22 Substituting 𝑧 = 1/2 into the factorization of the sin(𝜋𝑧) function
(Equation – VII.6.2), we obtain
42 V Compactness and Convergence in the Space of Analytic Functions
∞  
𝜋Ö 1/4
1 = sin(𝜋/2) = 1− 2
2 𝑛=1 𝑛
∞
𝜋 Ö 𝑛2 − 1/4
 
=
2 𝑛=1 𝑛2
∞ 
𝑛2

𝜋 Ö
⇐⇒ =
2 𝑛=1 𝑛2 − 1/4
∞ 
𝑛2
Ö 
=
𝑛=1
(𝑛 + 1/2) (𝑛 − 1/2)
∞
!
Ö 𝑛2
= (2𝑛+1) (2𝑛−1)
𝑛=1 4
∞ 
4𝑛2
Ö 
=
𝑛=1
(2𝑛 + 1) (2𝑛 − 1)
∞
4𝑛2
 
𝜋 Ö
∴ =
2 𝑛=1 4𝑛2 − 1
. ♦

Module – 3

Problem V.23 Show that 0 < 𝛾 < 1, where 𝛾 is the Euler-Mascheroni constant.

Solution V.23 We consider the (real) integral

∫ 1/𝑘  
𝑡 1 𝑘 +1
𝑑𝑡 = − log ,
0 1+𝑡 𝑘 𝑘

where we used the substitution 𝑢 = 1 + 𝑡 to evaluate it. We note that

∫ 1/𝑘 ∫ 1/𝑘
𝑡 1
𝑑𝑡 ≤ 𝑡 𝑑𝑡 = 2 .
0 1 + 𝑡 0 2𝑘
And, ∫ 1/𝑘 ∫ 1/𝑘
𝑡 𝑡 1
𝑑𝑡 ≥ 𝑑𝑡 = .
0 1+𝑡 0 1 + 1/𝑘 2𝑘 (𝑘 + 1)
From this we have the estimates
∞  ∞
𝜋2
   ∑︁
∑︁ 1 𝑘 +1 1
𝛾= − log ≤ 2
= ≈ 0.8225 < 1.
𝑘=1
𝑘 𝑘 𝑘=1
2𝑘 12
Module – 3 43

We also have
∞    ∞
∑︁ 1 𝑘 +1 ∑︁ 1
𝛾= − log ≥ > 0.
𝑘=1
𝑘 𝑘 𝑘=1
2𝑘 (𝑘 + 1)

Combining these results, we obtain 0 < 𝛾 < 1. ♦

Problem V.24 Show that √ Γ(𝑧)Γ(1 − 𝑧) = 𝜋 csc(𝜋𝑧) for 𝑧 not an integer. Deduce
from this that Γ(1/2) = 𝜋.
Solution V.24 If 𝑧 is not an integer, then by Gauss’ Formula (VII.7.6), we obtain

𝑛! 𝑛 𝑧 𝑚! 𝑚 1−𝑧
Γ(𝑧)Γ(1 − 𝑧) = lim lim
𝑛→∞ 𝑧(𝑧 + 1) · · · (𝑧 + 𝑛) 𝑚→∞ (1 − 𝑧) (2 − 𝑧) · · · (𝑚 + 1 − 𝑧)

𝑛! 𝑛! 𝑛 𝑧 𝑛1−𝑧
= lim
𝑛→∞ 𝑧(1 + 𝑧) (1 − 𝑧) · · · (𝑛 + 𝑧) (𝑛 − 𝑧) (𝑛 + 1 − 𝑧)

(𝑛!) 2 · 𝑛
= lim
𝑛→∞ 𝑧(12 − 𝑧 2 ) · · · (𝑛2 − 𝑧 2 ) (𝑛 + 1 − 𝑧)
(𝑛!) 2 · 𝑛
= lim  2
  
𝑧2
𝑧 (𝑛!) 2 1 −
𝑛→∞ 𝑧
12
··· 1− 𝑛2
(𝑛 + 1 − 𝑧)
1
= h   i  
𝑧2 𝑧2 𝑛+1−𝑧
𝑧 · lim 1− 12
··· 1− 𝑛2
· lim 𝑛
𝑛→∞ 𝑛→∞
1
= h   i .
𝑧2 𝑧2
𝑧 · lim 1− 12
··· 1− 𝑛2
𝑛→∞

Now, using the definition of infinite products we know that

∞ 
𝑧2 𝑧2 𝑧2
    
Ö sin(𝜋𝑧)
lim 1− 2
· · · 1 − 2
:= 1 − 2
= ,
𝑛→∞ 1 𝑛 𝑛=1
𝑛 𝜋𝑧

where the last equality follows from Equation – VII.6.2. Combining this with our
previous result, we obtain
1 𝜋
Γ(𝑧)Γ(1 − 𝑧) = h   i = = 𝜋 csc(𝜋𝑧).
𝑧 · lim 1− 𝑧2
··· 1− 𝑧2 sin(𝜋𝑧)
𝑛→∞ 12 𝑛2

Now, substituting 𝑧 = 1/2, we get

[Γ(1/2)] 2 = 𝜋 csc(𝜋/2).

This implies that

√
Γ(1/2) = 𝜋. ♦
44 V Compactness and Convergence in the Space of Analytic Functions

Problem V.25 Use Theorem – VII.8.17 to prove that 𝑝 𝑛−1 = ∞, if {𝑝 𝑛 } is a

Í
sequence of primes. (Note that this implies there are infinitely many primes.)
Solution V.25 The proof is by contradiction. Assume that 𝑝 𝑛−1 is finite. Euler’s
Í
Theorem (VII.8.7) states that for ℜ(𝑧) > 1,
∞  
Ö 1
𝜁 (𝑧) = .
𝑛=1
1 − 𝑝 𝑛−𝑧

Here, as ℜ(𝑧) → 1, both Í sides of the equality becomes unbounded. Now, since
𝑝 𝑛 > 0 for all 𝑛, the sum 𝑝 𝑛−1 , if finite, also converges absolutely. By Proposition
– VII.5.4, we then have that the series
∞  
∑︁ 1
log 1 −
𝑛=1
𝑝𝑛

converges absolutely. But this implies

∞   ∞  
∑︁ 1 Ö 1
− log 1 − = log = log 𝜁 (1) < ∞.
𝑛=1
𝑝𝑛 𝑛=1
1 − 𝑝𝑛

This is a contradiction since 𝜁 (𝑧) has a simple pole at 𝑧 = 1. Hence, there are in-
finitely many primes. ♦

∞
∑︁ 𝑑 (𝑛)
Problem V.26 Prove that 𝜁 2 (𝑧) = , for ℜ(𝑧) > 1, where 𝑑 (𝑛) is the number
𝑛=1
𝑛𝑧
of divisors of 𝑛.
Solution V.26 We start with the definition of the zeta function and expanding the
product as follows
∞
! ∞ !   
2
∑︁ 1 ∑︁ 1 1 1 1 1 1 1
𝜁 (𝑧) = = 1 + + + + · · · 1 + + + · · ·
𝑛=1
𝑛 𝑧 𝑛=1 𝑛 𝑧 2𝑧 3𝑧 4𝑧 2𝑧 3𝑧 4𝑧
1 1 1 1
=1+ + 𝑧 + 𝑧 + 𝑧 +···
2 𝑧 3 4 5
1 1 1 1
+ 𝑧 + 𝑧 + 𝑧 + 𝑧 +···
2 4 6 8
1 1 1 1
+ 𝑧 + 𝑧 + 𝑧 + 𝑧 +···
3 6 9 12
1 1 1 1
+ 𝑧 + 𝑧 + 𝑧 + 𝑧 +···
4 8 12 16
..
.
2 2 3 2 4
= 1+ 𝑧 + 𝑧 + 𝑧 + 𝑧 + 𝑧 +··· ;
2 3 4 5 6
Module – 3 45

1
Here, we observe that each term of the form 𝑧 appears exactly as many times as
𝑛
there are divisors of 𝑛, counting order. For instance, since 2 = 1 × 2 = 2 × 1, it
appears twice. Same for any prime 𝑛. For composite 𝑛, for instance for 𝑛 = 4, since
4 = 1 × 4 = 2 × 2 = 4 × 1, it appears thrice. Hence, we have
∞
𝑑 (𝑛)
♦
∑︁
𝜁 2 (𝑧) = .
𝑛=1
𝑛𝑧

√ √
Problem V.27 Find a meromorphic function with poles √ of order 2 at 1, 2, 3, . . .
2
√ (𝑧 − 𝑚) 𝑓 (𝑧) = 1 for all 𝑚.
such that the residue at each pole is 0 and lim
𝑧→ 𝑚

Solution V.27 We claim that the function 𝑓 defined by

∞ √
∑︁ (3 𝑛 − 2𝑧)𝑧 2
𝑓 (𝑧) = √ √ 2
3
𝑛=1 𝑛 (𝑧 − 𝑛)

has the required

√ properties. To verify this we observe that the only poles of 𝑓
occur at 𝑧 = 𝑛, ∀𝑛 ∈ N. These are all poles √ of multiplicity 2. Moreover, 𝑓 is
meromorphic in√ C. Next, we verify that Res( 𝑓 ; 𝑛) = 0, ∀𝑛.√By Proposition – V.2.4,
if 𝑔(𝑧) = (𝑧 − 𝑛) 2 𝑓 (𝑧), then it is enough to show that 𝑔 ′ ( 𝑛) = 0, ∀𝑛 ∈ N.
√
3 𝑛 − 2𝑧 𝑧2


𝑔(𝑧) = √
𝑛3
√

′ 2𝑧 3 𝑛 − 2𝑧 2𝑧2
∴ 𝑔 (𝑧) = √ −√
𝑛3 𝑛3
√

6𝑧 𝑧 − 𝑛
=− √ ;
𝑛3
√ √
=⇒ 𝑔 ′ ( 𝑛) = Res( 𝑓 ; 𝑛) = 0.
√ 2
√ (𝑧 − 𝑚) 𝑓 (𝑧) = 1 for all 𝑚. We can write
Finally, we verify that lim
𝑧→ 𝑚

∞ √ √
√ 2 ∑︁ (𝑧 − 𝑚) 2 (3 𝑛 − 2𝑧)𝑧2
lim
√ (𝑧 − 𝑚) 𝑓 (𝑧) = lim
√ √ √
𝑧→ 𝑚 𝑧→ 𝑚
𝑛=1 𝑛3 (𝑧 − 𝑛) 2
" √ √ √ √ #
(𝑧 − 𝑚) 2 (3 𝑛 − 2𝑧)𝑧2 ∑︁ (𝑧 − 𝑚) 2 (3 𝑛 − 2𝑧)𝑧2
= lim√ √ √ + √ √
𝑧→ 𝑚 𝑛3 (𝑧 − 𝑛) 2 𝑛≠𝑚 𝑛3 (𝑧 − 𝑛) 2
√ √
3 𝑚3 − 2 𝑚3
= √ + 0 = 1.
𝑚3

Thus, our claimed function satisfies all the requirements and we are done. ♦
Chapter VI
Harmonic and Entire Functions

Problem VI.1 Show that if 𝑢 is harmonic then so are 𝑢 𝑥 and 𝑢 𝑦 .

Solution VI.1 By definition, if 𝑢 is harmonic then

𝑢 𝑥 𝑥 + 𝑢 𝑦𝑦 = 0.

Again since 𝑢 is harmonic, 𝑢 is infinitely differentiable. In particular, the third partial

derivatives of 𝑢 exist and are continuous. So, we have

𝑢 𝑥 𝑦 𝑦 = 𝑢 𝑦𝑦 𝑥 and 𝑢 𝑦 𝑥 𝑥 = 𝑢 𝑥 𝑥 𝑦 .

To show that 𝑢 𝑥 is harmonic, we notice that

(𝑢 𝑥 ) 𝑥 𝑥 + (𝑢 𝑥 ) 𝑦𝑦 = 𝑢 𝑥 𝑥 𝑥 + 𝑢 𝑥 𝑦𝑦 = (𝑢 𝑥 𝑥 + 𝑢 𝑦𝑦 ) 𝑥 = 0.

Thus, by definition, 𝑢 𝑥 is harmonic. Similarly, to show that 𝑢 𝑦 is harmonic, we note

that
(𝑢 𝑦 ) 𝑥 𝑥 + (𝑢 𝑦 ) 𝑦𝑦 = 𝑢 𝑦 𝑥 𝑥 + 𝑢 𝑦𝑦𝑦 = (𝑢 𝑥 𝑥 + 𝑢 𝑦𝑦 ) 𝑦 = 0.
Thus, 𝑢 𝑦 is also harmonic and we are done. ♦
Problem VI.2 If 𝑢 is harmonic, show that 𝑓 = 𝑢 𝑥 − 𝑖𝑢 𝑦 is analytic.

Solution VI.2 By definition, if 𝑢 is harmonic then

𝑢 𝑥 𝑥 + 𝑢 𝑦𝑦 = 0.

Again since 𝑢 is harmonic, 𝑢 is infinitely differentiable. In particular, the second

partial derivatives of 𝑢 exist and are continuous, that is, 𝑢 𝑥 𝑦 = 𝑢 𝑦 𝑥 . Now, to show
that 𝑓 = 𝑢 𝑥 − 𝑖𝑢 𝑦 is analytic, it is enough to show the following three conditions are
true:
• ℜ( 𝑓 ) = 𝑢 𝑥 is harmonic;
• ℑ( 𝑓 ) = −𝑢 𝑦 is harmonic;

47
48 VI Harmonic and Entire Functions

• 𝑢 𝑥 and −𝑢 𝑦 satisfy the Cauchy – Riemann Equations.

By Problem – VI.1, we already have shown that 𝑢 𝑥 and 𝑢 𝑦 (hence −𝑢 𝑦 ) are harmonic.
To show that 𝑢 𝑥 and −𝑢 𝑦 satisfy the Cauchy – Riemann Equations, we note that

(𝑢 𝑥 ) 𝑥 = −(𝑢 𝑦 ) 𝑦 ⇐⇒ 𝑢 𝑥 𝑥 + 𝑢 𝑦𝑦 = 0.

Again, we also notice

(𝑢 𝑥 ) 𝑦 = −(−𝑢 𝑦 ) 𝑥 ⇐⇒ 𝑢 𝑥 𝑦 = 𝑢 𝑦 𝑥 .

Thus, the Cauchy – Riemann Equations are satisfied and so we are done. ♦
Problem VI.3 Let 𝑢 be harmonic in 𝐺 and suppose 𝐵(𝑎; 𝑅) ⊂ 𝐺. Show that
∬
1
𝑢(𝑎) = 𝑢(𝑥, 𝑦) 𝑑𝑥 𝑑𝑦 .
𝜋𝑅 2 𝐵(𝑎;𝑅)

Solution VI.3 Let D be a disk such that 𝐵(𝑎; 𝑅) ⊂ D ⊂ 𝐺. There exist an 𝑓 ∈ H (D)
such that ℜ( 𝑓 ) = 𝑢 on D. From Cauchy’s Integral Formula, we have
∫ ∫ 2𝜋
1 𝑓 (𝑤) 1
𝑓 (𝑎) = 𝑑𝑤 = 𝑓 (𝑎 + 𝑟e𝑖 𝜃 ) 𝑑𝜃,
2𝜋𝑖 𝛾 𝑤−𝑎 2𝜋 0

where 𝛾(𝑡) = 𝑎 + 𝑟e𝑖 𝜃 and 𝑡 ∈ [0, 2𝜋]. Now, multiplying by 𝑟 and then integrating
both sides w.r.t. 𝑟, we obtain
∫ 𝑅 ∫ 𝑅 ∫ 2𝜋
1
𝑟 𝑓 (𝑎) 𝑑𝑟 = 𝑟 𝑓 (𝑎 + 𝑟e𝑖 𝜃 ) 𝑑𝜃 𝑑𝑟
0 2𝜋 0 0
∫ 𝑅 ∫ 2𝜋
1 𝑅 1
⇐⇒ 𝑟 2 𝑓 (𝑎) = 𝑟 𝑓 (𝑎 + 𝑟e𝑖 𝜃 ) 𝑑𝜃 𝑑𝑟
2 0 2𝜋 0 0
∫ 𝑅 ∫ 2𝜋
1
⇐⇒ 𝑓 (𝑎) = 𝑟 𝑓 (𝑎 + 𝑟e𝑖 𝜃 ) 𝑑𝜃 𝑑𝑟
𝜋𝑅 2 0 0
∬
1
⇐⇒ 𝑓 (𝑎) = 𝑓 (𝑥, 𝑦) 𝑑𝑥 𝑑𝑦,
𝜋𝑅 2 𝐵(𝑎;𝑅)

where the last equality follows by a change of coordinates from polar to rectangular.
Now, we note
∬ ∬
1 1
𝑢(𝑎) = ℜ( 𝑓 (𝑎)) =
𝜋𝑅 2 𝐵(𝑎;𝑅)
ℜ( 𝑓 (𝑥, 𝑦)) 𝑑𝑥 𝑑𝑦 =
𝜋𝑅 2 𝐵(𝑎;𝑅)
𝑢(𝑥, 𝑦) 𝑑𝑥 𝑑𝑦 . ♦
Problem VI.4 Let 𝑢 : 𝐺 → R be harmonic and let 𝐴 = {𝑧 ∈ 𝐺 : 𝑢 𝑥 (𝑧) = 𝑢 𝑦 (𝑧) =
0}. Can 𝐴 have a limit point in 𝐺 ?
Solution VI.4 Let 𝑢 be a nonconstant harmonic function. By Problem – VI.1, we
know that 𝑢 𝑥 and 𝑢 𝑦 are harmonic, and by Problem – VI.2, we know that 𝑓 = 𝑢 𝑥 −𝑖𝑢 𝑦
is analytic. Define
VI Harmonic and Entire Functions 49

𝐵 = {𝑧 ∈ 𝐺 : 𝑓 (𝑧) = 0} = {𝑧 ∈ 𝐺 : 𝑢 𝑥 (𝑧) − 𝑖𝑢 𝑦 (𝑧) = 0}.

Then we note that 𝐴 = 𝐵 since 𝑢 𝑥 (𝑧) = 𝑢 𝑦 (𝑧) = 0 ⇐⇒ 𝑓 (𝑧) = 0. Since 𝑓 is

analytic, the zeros are isolated and therefore 𝐵 cannot have a limit point in 𝐺. This
implies that 𝐴 cannot have a limit point in 𝐺.
But if 𝑢 is a constant harmonic function, then we must have 𝐴 = 𝐵 = 𝐺, and 𝑓 ≡ 0.
Thus, 𝐴 must have a limit point in 𝐺. ♦
Problem VI.5 In the hypothesis of Jensen’s Formula, do not suppose that 𝑓 (0) ≠ 0.
Show that if 𝑓 has a zero at 𝑧 = 0 of multiplicity 𝑚 then
∫ 2𝜋
𝑓 (𝑚) (0)
𝑛  
∑︁ 𝑟 1  
𝑚 log 𝑟 + log =− log + log 𝑓 𝑟e𝑖 𝜃 𝑑𝜃 .
𝑚! 𝑘=1
|𝑎 𝑘 | 2𝜋 0

𝑟 (𝑧 − 𝑎 𝑘 )
Solution VI.5 We know that the map maps 𝐵(0; 𝑟) onto itself. We let
𝑟 2 − 𝑧 𝑎¯ 𝑘
𝑛
Ö 𝑟 2 − 𝑧 𝑎¯ 𝑘
𝑟𝑚
𝐹 (𝑧) = 𝑓 (𝑧) .
𝑧𝑚 𝑘=1
𝑟 (𝑧 − 𝑎 𝑘 )

Then 𝐹 ∈ H (𝐺), 𝐹 (𝑧) ≠ 0 ∀𝑧 ∈ 𝐵(0; 𝑟), and |𝐹 (𝑧)| = | 𝑓 (𝑧)| whenever |𝑧| = 𝑟.
Thus, by Jensen’s Formula, we have
∫ 2𝜋 ∫ 2𝜋
1 
𝑖𝜃
 1  
log |𝐹 (0)| = log 𝐹 𝑟e 𝑑𝜃 = log 𝑓 𝑟e𝑖 𝜃 𝑑𝜃 .
2𝜋 0 2𝜋 0

Since 𝑓 has a zero at 𝑧 = 0 of multiplicity 𝑚, we also notice that

∞
𝑓 (𝑧) 1 ∑︁ 𝑓 (𝑘 ) (0) 𝑘
𝑔(𝑧) ≡ = 𝑧
𝑧𝑚 𝑧 𝑚 𝑘=0 𝑘!
∞
1 ∑︁ 𝑓 (𝑘 ) (0) 𝑘
= 𝑧
𝑧 𝑚 𝑘=𝑚 𝑘!
∞
∑︁ 𝑓 (𝑘 ) (0) 𝑘−𝑚
= 𝑧 .
𝑘=𝑚
𝑘!
𝑛  
𝑚
Ö 𝑟
∴ |𝐹 (0)| = |𝑔(0)| 𝑟 .
𝑘=1
|𝑎 𝑘 |

Substituting this into our previous equation for log |𝐹 (0)|, we obtain
∫ 2𝜋
𝑓 (𝑚) (0) ∑︁
𝑛  
𝑟 1  
𝑚 log 𝑟 + log + log = log 𝑓 𝑟e𝑖 𝜃 𝑑𝜃
𝑚! 𝑘=1
|𝑎 𝑘 | 2𝜋 0
∫ 2𝜋
𝑓 (𝑚) (0)
𝑛  
1
♦
∑︁ 𝑟  
⇐⇒ 𝑚 log 𝑟 + log =− log + log 𝑓 𝑟e𝑖 𝜃 𝑑𝜃 .
𝑚! 𝑘=1
|𝑎 𝑘 | 2𝜋 0