COMPLEX ANALYSIS

Prof. Pranav Haridas

Mathematics
IIT Madras
 INDEX

S.NO TOPICS PAGE.NO

Week 1
1 Field of Complex Numbers 4
2 Conjugation and Absolute value 23
3 Topology on Complex plane 33
4 Topology on Complex Plane (Contd) 47

Week 2
5 Problem Session 56
6 Isometries on the Complex Plane 71
7 Functions on the Complex Plane 79
8 Complex differentiability 88

Week 3
9 Power Series 96
10 Differentiation of power series 110
11 Problem Session 119
12 Cauchy-Riemann equations 127

Week 4
13 Harmonic functions 134
14 Möbius transformations 144
15 Problem session 153
16 Curves in the complex plane 161

Week 5
18 First Fundamental theorem of Calculus 185
19 Second Fundamental theorem of Calculus 192
20 Problem session 199

1
 Week 6
21 Homotopy of curves 208
22 Cauchy-Goursat theorem 215
23 Cauchy's theorem 223
24 Problem Session 229

Week 7
25 Cauchy Integral Formula 237
26 Principle of analytic continuation and Cauchy estimates 242
27 Further consequences of Cauchy Integral Formula 249
28 Problem session 257

Week 8
29 Winding number 266
30 Open mapping theorem 273
31 Schwarz reflection principle 280
32 Problem session.. 287

Week 9
33 Singularities of a holomorphic function 292
34 Pole of a function 299
35 Laurent Series 302
36 Casorati Weierstrass theorem 310

Week 10
37 Problem -Session 317
38 Residue theorem 324
39 Argument principle 331
40 Problem- Session 338

Week 11
41 Branch of the Complex logarithm 345

2
42 Automorphisms of the Unit disk 352
43 Phragmen Lindelof method 358
44 Problem- -Session 367

Week 12
45 Lifting of maps 374
46 Covering spaces 381
47 Bloch's theorem 388
48 Little Picard's theorem 394

3
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics

Lecture – 1
Field of Complex Numbers

Welcome to this course on complex analysis. This is a first course on complex analysis. The
prerequisites that will be needed for this course would be sound understanding of basic linear
algebra and basic real analysis. If you have seen some abstract algebra that will certainly help,
but other than the first lecture the rest of the cause will not use much material or much
knowledge from abstract algebra.
(Refer Slide Time: 00:47)

Regarding the textbook, so let me note down a few books. I should warn that we do not have a
prescribed textbook for this course; however, most of the material that will be covered in this
course can be found in almost all the classical books written in complex analysis. However, I
would like to refer the following books for this course. The first one being Complex Analysis
by Stein and Shakarchi.

The second is titled “Functions of One Complex Variable”. There are 2 parts in this book, so
the material that we will cover in this course will be found in the first part and it is by John B.
Conway. Of course, there are many, many, many beautiful books written on this subject, maybe
I should give a few other references.
(Refer Slide Time: 01:59)

4
There is this age old classic it is called “Complex Analysis” by Lars Ahlfors. There is also
“Complex Analysis” by Theodore Gamelin. There is also this fantastic book “Real and
Complex Analysis” with a slightly different approach by Walter Rudin. Of course, there are
many more books, let me not write down all of them, but I would also suggest that you refer to
these other references maybe a second or a third reading. Okay that is about textbooks.

There will be weekly assignments in this course and you are really strongly encouraged to work
on these problems on your own, you should spend some time sitting and thinking about these
problems that will give you much better clarity on the subject material that will be covered.
More or less everything introduced the course to you, so let us now begin the study of the
subject. In a course in real analysis you would have started by studying rational numbers.

You would have seen that rational numbers have certain deficiencies. For example if you look
at some Cauchy sequences in rational numbers they do not converge and real numbers were
constructed precisely to address this particular problem. Real numbers are the complete field
which contain the rational numbers and it is unique up to some field isomorphisms.

You would have developed and studied an entire rich and beautiful theory of real analysis on
this field of real numbers, but then from an algebraic point of view, real numbers also have
certain drawbacks. So there are polynomials, for example, which do not have roots in the real
numbers. For example, 𝑥 2 + 1 is a polynomial which does not have a root in the field of real
numbers. Complex numbers was historically constructed in order to address this particular
problem. So let us start this course by recalling what a field is.

5
(Refer Slide Time: 04:50)

So, a field F is a set with 2 operations. Let us call these operations addition which are denote
by (+) and multiplication which is being denoted by (×), which satisfy the following properties.
(Refer Slide Time: 05:37)

First one is commutativity. If you take 2 elements 𝑥, 𝑦 𝜖 F, 𝑥 + 𝑦 = 𝑦 + 𝑥. The order in

which we take the sum does not matter and 𝑥 × 𝑦 = 𝑦 × 𝑥, the order in which you multiply
also does not matter. Second one is associativity. For 𝑥, 𝑦, 𝑧 𝜖 F,(𝑥 + 𝑦) + 𝑧 = 𝑥 + (𝑦 + 𝑧).
Similarly,(𝑥 × 𝑦) × 𝑧 = 𝑥 × (𝑦 × 𝑧). Associativity is a property which holds for both addition
and multiplication.
(Refer Slide Time: 07:11)

6
The third property is existence of additive and multiplicative identities. There exists an element
0 𝜖 F such that 𝑥 + 0 = 𝑥 for all 𝑥 𝜖 F. This is the additive identity in F. There exists an
element 1𝜖 F such that 𝑥 × 1 = 𝑥 for all 𝑥 𝜖 F. So basically, there are two special elements
in the field F.
Fourth property is distributivity. The addition and the multiplication operations, they interact
with each other. For 𝑥, 𝑦, 𝑧 𝜖 F, 𝑥 × (𝑦 + 𝑧) = 𝑥 × 𝑦 + 𝑥 × 𝑧.
(Refer Slide Time: 08:51)

Finally, existence of inverses. For 𝑥 𝜖 F, there exists 𝑦 𝜖 F such that 𝑥 + 𝑦 gives you the
additive identity 0 and for 𝑥 𝜖 F\{0}, there exists 𝑧 𝜖 F such that 𝑥 × 𝑧 = 1. The multiplicative
inverse exists only for nonzero elements in the field F. This is the set of all properties which
needs to be satisfied by the two operations for our given set to be a field. So that is the definition
of a field and we are familiar with the field of real numbers.

7
(Refer Slide Time: 09:59)

The set ℝ is a field with the usual addition and multiplication operation. However, the algebraic
structure of ℝ has certain drawbacks. For example, all polynomials need not have roots in the
field of real numbers.
(Refer Slide Time: 11:03)

For every 𝑥 𝜖 ℝ, 𝑥 2 ≥ 0 and hence the polynomial 𝑥 2 + 1 does not have root in ℝ. Field of
complex numbers is constructed to address this drawback. Let me just start with definition for
a field of complex numbers.
(Refer Slide Time: 12:09)

8
Definition: A field of complex numbers ℂ is a field which contains ℝ as subfield and a root 𝑖
to the polynomial 𝑥 2 + 1. Furthermore, the field of complex numbers is the smallest such field
which contains ℝ and 𝑖.

Rephrasing it, there does not exist a proper subfield of ℂ containing ℝ and 𝑖. When put
differently, this says that ℂ is generated by ℝ and 𝑖 or if you have ℂ′, a subfield of ℂ which
contains ℝ and 𝑖, then ℂ′ would necessarily be equal to ℂ.
(Refer Slide Time: 14:57)

This definition poses the following questions. The first one being does there exist such a field
of complex numbers and the second one is can we say anything about the uniqueness.
(Refer Slide Time: 15:55)

9
In order to address the existence of such a field of complex numbers, we will be using some
notions from abstract algebra, more precisely from ring theory. If you have not seen a course
on abstract algebra, there is no problem, you may skip the remaining part of the lectures.

You can assume the existence and the uniqueness of such a field of complex numbers, move
over to the next lecture. From the next lecture onwards, there will be more focus on the analysis
on such a field of complex numbers, you will not be needing too much of background in
abstract algebra. However, in this lecture, we will be assuming some amount of knowledge in
ring theory.

Let ℝ[𝑥] be the collection of all formal polynomials over ℝ, the field of real numbers. A formal
polynomial is an expression of the type 𝑎0 + 𝑎1 𝑥 + ⋯ + 𝑎𝑑 𝑥 𝑑 , where 𝑎𝑖 are real numbers and
𝑑 is a non-negative integer.

We will be familiar with the addition of polynomials, multiplication of polynomials. Let me

not give all those things as definitions again.
(Refer Slide Time: 18:28)

10
Let me just note that with the usual addition and multiplication of polynomials, ℝ[𝑥] is
commutative ring with identity. So if you have not seen abstract algebra before, a commutative
ring is set with two operations with almost all these properties satisfied except the existence of
multiplicative inverse.

What are the identities here? The zero polynomial is the additive identity and the constant
polynomial 1 will be the multiplicative identity. The first thing to note is that ℝ[𝑥] is a
commutative ring which is not a field, not all elements can be inverted.
(Refer Slide Time: 19:54)

In fact any non-constant polynomial is not a unit. There does not exist a multiplicative inverse
for any non-constant polynomial in ℝ[𝑥]. Why is that the case? Because we have a notion of
the degree of a polynomial, 𝑑 and if you look at the product of two polynomials, the degree

11
adds up. If you want an inverse for a polynomial 𝑝(𝑥), there should exist a 𝑞(𝑥) such that
𝑝(𝑥)𝑞(𝑥) = 1.

The constant polynomial 1 has degree 0. However if 𝑝(𝑥) has degree greater than 0, then degree
of 𝑝(𝑥)𝑞(𝑥) will always be greater than or equal to 1 and hence it cannot ever be a constant
polynomial.
Since the degree of the product of two polynomials is equal to the sum of the degree of the
polynomials and the degree is always a non-negative number. Therefore non-constant
polynomial can never have an inverse.
(Refer Slide Time: 22:03)
Let us take the polynomial 𝑥 2 + 1. Since 𝑥 2 + 1 is a non-constant polynomial, it is not
invertible, the ideal generated by 𝑥 2 + 1, ⟨𝑥 2 + 1⟩ will not be the entire ring.
If 𝑎 is an element in the ideal, for any element 𝑏 in the ring, 𝑎𝑏 will be in the ideal and the
ideals are the right objects with which we take quotients in a ring.
(Refer Slide Time: 23:11)

Let us define ℂ := ℝ[𝑥]/⟨𝑥2 + 1⟩ .

The objects here will be cosets of the ideal ⟨𝑥2 + 1⟩. The operation in C are given by;
for the cosets 𝑝(𝑥) +⟨𝑥 2 + 1⟩ 𝝐 ℂ and 𝑞(𝑥) +⟨𝑥 2 + 1⟩ 𝜖 ℂ, 𝑝(𝑥) +⟨𝑥 2 + 1⟩ = 𝑞(𝑥) +
⟨𝑥 2 + 1⟩ if 𝑝(𝑥) − 𝑞(𝑥) 𝜖 ⟨𝑥 2 + 1⟩ if and only if 𝑥 2 + 1 being a factor of 𝑝(𝑥) – 𝑞(𝑥). This
is the equivalence relation of the cosets involved.
(Refer Slide Time: 24:48)

12
The sum of two elements in ℂ is defined by, for 𝑝(𝑥) +⟨𝑥 2 + 1⟩ , 𝑞(𝑥) +⟨𝑥 2 + 1⟩ 𝜖 ℂ,
𝑝(𝑥) +⟨𝑥 2 + 1⟩ + 𝑞(𝑥) +⟨𝑥 2 + 1⟩ = (𝑝(𝑥) + 𝑞(𝑥)) +⟨𝑥 2 + 1⟩ and the multiplication of two
elements in C is defined as, for 𝑝(𝑥) +⟨𝑥 2 + 1⟩ , 𝑞(𝑥) +⟨𝑥 2 + 1⟩ 𝜖 ℂ,
(𝑝(𝑥) +⟨𝑥 2 + 1⟩) × (𝑞(𝑥) +⟨𝑥 2 + 1⟩) = (𝑝(𝑥)𝑞(𝑥)) +⟨𝑥 2 + 1⟩

(Refer Slide Time: 26:01)

Let me invoke the knowledge from ring theory to say that ℂ is a commutative ring with identity
with these operations. So we have found a commutative ring which is a candidate. So why is
this a candidate?
(Refer Slide Time: 26:30)

13
Define 𝜑 ∶ ℝ → ℂ given by, for 𝑎 𝜖 ℝ, 𝜑(𝑎): = 𝑎 +⟨𝑥 2 + 1⟩. So easy to see that this ring
homomorphism.
Claim: 𝜑 is injective, 𝜑 embeds ℝ into ℂ.
To check it is injective what do we have to do?
Proof of the claim,
𝜑(𝑎) = 𝜑(𝑏) ⇒ 𝑎 +⟨𝑥 2 + 1⟩= 𝑏 +⟨𝑥 2 + 1⟩ ⇒ 𝑥 2 + 1 divides 𝑎 − 𝑏, but 𝑎 – 𝑏 is a real
number. Any divisible, any factor or any multiple of 𝑥 2 + 1 should have at least degree 2 or
else it should be the 0 polynomial. Since degree of 𝑎 – 𝑏 = 0. We have 𝑎 – 𝑏 = (𝑥 2 +
1) × 0 = 0 ⇒ 𝑎 = 𝑏.
(Refer Slide Time: 28:28)

Immediately we see that the map 𝜑 from ℝ to ℂ is an embedding, it actually gives us a copy of
ℝ in the commutative ring ℂ. That is one aspect of the definition solved and like it may let us
now that we can get hold of a root of 𝑥 2 + 1 in this particular commutative ring. So let 𝑖 be

14
defined to be, 𝑖 ≔ 𝑥 +⟨𝑥 2 + 1⟩ in ℂ.

Recall that ℂ is ℝ[𝑥]/⟨𝑥2 + 1⟩, so you look at the coset represented by 𝑥, then 𝑖 2 + 1 =
(𝑥 +⟨𝑥 2 + 1⟩)2 + (1 +⟨𝑥 2 + 1⟩) in ℂ. By the definitions of multiplication and addition that
we have defined in the quotient, 𝑖 2 + 1 = (𝑥 2 + 1) +⟨𝑥 2 + 1⟩ = 0 + ⟨𝑥2 + 1⟩ ⇒ 𝑖2 + 1 = 0
in ℂ.
Hence, we have both solution to 𝑥 2 + 1 = 0 and a copy of ℝ sitting inside ℂ. So we are at least
two of the aspects in the definition of field of complex numbers.
(Refer Slide Time: 31:05)

In fact, we also can say something more about the characterization, namely that this is actually
generated by ℝ and 𝑖 and we say that any element, you look to any element in ℂ, this is
represented by 𝑝(𝑥) +⟨𝑥 2 + 1⟩, for some polynomial 𝑝(𝑥). Then this is equal to 𝑝(𝑖).
That is, any element of ℂ can be written as polynomial expression of 𝑖 with coefficients in ℝ.
But what can we say about any field of ℂ which contains ℝ and 𝑖? It will certainly have any
polynomial expression in 𝑖 with coefficients because it is a field that is closed under both
multiplication and addition. Hence any subring of ℂ which contains ℝ and 𝑖 should necessarily
be equal to ℂ.
The only thing that is to be checked is whether it is a field. As of now whatever we have
constructed ℂ, it is just a commutative ring with the identity.

If we establish that it is indeed a field as well, we would have proved the existence of field of
complex numbers.

15
(Refer Slide Time: 34:40)

So let us now work towards proving that this particular commutative ring is indeed a field, but
to do that, let us focus on the polynomial 𝑥 2 + 1 in ℝ[𝑥]. The polynomial 𝑥 2 + 1 is an
irreducible element in the commutative ring ℝ[𝑥]. So the first observation would be that an
irreducible element in ℝ [𝑥] would be prime, more specifically 𝑥 2 + 1 is hence prime element.
(Recall what was a prime element was: in a ring 𝐴, 𝑝 is defined to be a prime element if
whenever 𝑝 divides 𝑎𝑏, the product of two elements 𝑎 and 𝑏 in ring, 𝑝 either divides 𝑎 or 𝑝
divides 𝑏.)
So, the claim here is that 𝑥 2 + 1 is a prime element in ℝ [𝑥]. Let us just check that very quickly.
Let 𝑥 2 + 1 divide 𝑝(𝑥)𝑞(𝑥), let us prove that it divides one of them. So without loss of
generality if 𝑥 2 + 1 divides 𝑝(𝑥), then we have already proved.

If 𝑥 2 + 1 does not divide 𝑝(𝑥), 𝑥 2 + 1 is an irreducible polynomial and therefore if you look
at the greatest common divisor of 𝑥 2 + 1 and 𝑝(𝑥), it should necessarily be 1. So it will be a
unit, so it will be 1 here. Let me rephrase it this way since 𝑥 2 +1 is irreducible, there exist
polynomials 𝛼(𝑥) and 𝛽(𝑥) such that 𝛼(𝑥)(𝑥 2 + 1) + 𝛽(𝑥)𝑝(𝑥) = 1.
(Refer Slide Time: 36:56)

16
Multiplying 𝑞(𝑥) to the equation above, we have 𝛼(𝑥)(𝑥 2 + 1)𝑞(𝑥) + 𝛽(𝑥)𝑝(𝑥)𝑞(𝑥) =
𝑞(𝑥), but if you focus on the left hand side, the first term contain the factor of 𝑥 2 + 1 and
second term contain the product 𝑝(𝑥)𝑞(𝑥) and hence 𝑥 2 + 1 divides both terms of left hand
side, therefore 𝑥 2 + 1 divides 𝑞(𝑥).
(Refer Slide Time: 37:51)

Then 𝑥 2 + 1 is a prime element and therefore the ideal, ⟨𝑥2 + 1⟩ is a prime ideal. We know
that when we go modulo a prime ideal, we get an integral domain. Since ⟨𝑥2 + 1⟩ is a prime
ideal, we have ℂ is an integral domain.
The fact that ℂ is a field is now established only after showing that every nonzero element can
be inverted.
(Refer Slide Time: 39:24)

17
Notice that ℝ[𝑥], because there is a copy of ℝ is in ℝ[𝑥], is a vector space over ℝ with
generating set given by {1, 𝑥, 𝑥 2 , … } and since ℂ is a quotient of our ℝ[𝑥] by an ideal, ℂ is also
a vector space over ℝ with generating set given by{1, 𝑖, 𝑖 2 , … }.
(Refer Slide Time: 40:42)

Since 𝑖 2 = −1, we have {1, 𝑖} is a spanning set of ℂ. So we have a set consisting of 2 elements,
which is a spanning set of ℂ. Also note this that 𝑖 does not belong to ℝ. Why is that? It is
because if 𝑖 belongs to ℝ, then there is a real number 𝑎 in ℝ such that 𝑥 +⟨𝑥 2 + 1⟩= 𝑎 +
⟨𝑥 2 + 1⟩ ⇒ 𝑥 − 𝑎 𝜖 ⟨𝑥 2 + 1⟩ which is not possible since 𝑥 − 𝑎 will have degree one.
So, 𝑖 does not belong to ℝ and therefore 1 and 𝑖 turn out to be linearly independent. So hence
{1, 𝑖} is a basis of ℂ over ℝ.
(Refer Slide Time: 41:59)

18
We have now just established a lemma.
Lemma: ℂ is a two-dimensional vector space over ℝ.

We have still not proved that ℂ is a field, so in order to do that let us take a non-zero arbitrary
element 𝑧 in ℂ. Define 𝑀𝑧 : ℂ → ℂ given by,𝑀𝑧 (𝑤): = 𝑧𝑤. We have already checked that ℂ is
an integral domain, this is left multiplication by 𝑧 and you should check that 𝑀𝑧 is actually an
ℝ linear map, it is a linear transformation.
(Refer Slide Time: 42:57)

This is an easy check and you will immediately note that the null space of 𝑀𝑧 is {0}, since null
space contain those 𝑤 such that 𝑧𝑤 = 0, but we already checked that ℂ is an integral domain
and this can happen only if 𝑤 = 0, as 𝑧 ≠ 0.
Therefore, 𝑀𝑧 is an injective linear transformation from a two-dimensional vector space to

19
itself. We know that an injective linear transformation from a finite dimensional vector space
to itself should necessarily be surjective by the rank-nullity theorem.
(Refer Slide Time: 44:42)

That means there exists some 𝑤′ such that 𝑧𝑤 ′ = 1. Hence 𝑧 is invertible. Therefore ℂ is a
field.
(Refer Slide Time: 45:08)

We have established every aspect of ℂ being a field of complex numbers. Now have one such
field of complex numbers. We have answered one of the questions that arose from the
definitions satisfactorily.

Let us now move over to the second one, that is more easier than this, the question of
uniqueness.

20
(Refer Slide Time: 46:20)

Theorem: Let ℂ′ be a field of complex numbers with 𝑖′ a root of 𝑥 2 + 1. Then ℂ′ is isomorphic

to ℂ.
(Refer Slide Time: 47:25)

Proof: Define 𝜓 ∶ ℂ → ℂ′ given by for 𝑝(𝑖) 𝜖 ℂ, 𝜓(𝑝(𝑖)) = 𝑝(𝑖 ′ ). So notice that it will be a
field homomorphism from ℂ to ℂ′.

𝜓(ℂ) is a subfield of ℂ′, but this 𝜓(ℂ) has certain characteristics , it contains ℝ because its
identity on ℝ it will send ℝ to ℝ. It also contains 𝑖′ because 𝜓(𝑖) = 𝑖′ by the very definition of
𝜓.
(Refer Slide Time: 49:04)

21
Since ℂ’ is field of complex numbers, this forces 𝜓(ℂ) to be entire ℂ′ and hence 𝜓 is a field
isomorphism. So we have established remarkable aspect here. It tells us that any two field that
satisfy the properties in the definition of complex number, they should necessarily be
isomorphic to each other and therefore we could study analysis over any one such field that we
can construct. In the next lecture, we will get hold of another such construction of a field of
complex numbers and that will be more handy to work with and we will be doing analysis on
that field of complex numbers some more.

22
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics

Lecture – 2
Conjugation and Absolute Value

In the last lecture, we constructed an explicit field of complex numbers, a field which contains
ℝ as a subfield and a root 𝑖 to the polynomial 𝑥 2 + 1. We also showed that if there is a subfield
of the field of complex numbers which contains ℝ and 𝑖, it should necessarily be the entire set
ℂ. We also showed a uniqueness up to field isomorphism of the field of complex numbers, so
hence we can talk about the field of complex numbers here.
(Refer Slide Time: 00:41)

In the last lecture, we proved that ℂ is a two dimensional vector space over ℂ. We, in fact, get
hold of an explicit basis of ℂ over ℝ, which is {1, 𝑖}.

Let us define ℂ ≔ {𝑎 + 𝑖𝑏 ∶ 𝑎, 𝑏 𝜖 ℝ} and since {1, 𝑖} is a basis, 𝑎 + 𝑖𝑏 = 𝑐 + 𝑖𝑑 if and only if

𝑎 = 𝑐, 𝑏 = 𝑑. We even have an explicit expression for the addition and multiplication. So the
addition of two complex numbers is given by, (𝑎 + 𝑖𝑏) + (𝑐 + 𝑖𝑑) = (𝑎 + 𝑐) + 𝑖(𝑏 + 𝑑) and
we also defined the multiplication as being the product of cosets, that is, (𝑎 + 𝑖𝑏)(𝑐 + 𝑖𝑑) =
(𝑎𝑐 − 𝑏𝑑) + 𝑖(𝑎𝑑 + 𝑏𝑐). If you have not seen a course on abstract algebra or ring theory you
may as well start the course, this course complex analysis here by taking this as our definition
of field of complex numbers. It is a collection of all elements of the type 𝑎 + 𝑖𝑏 where 𝑎 and
𝑏 are real numbers.
Addition and multiplication are defined as above and you can check that ℂ is a field of complex

23
numbers with these operations.
We know that any two dimensional vector space over ℝ going to be isomorphic as vector
spaces to ℝ2 .
(Refer Slide Time: 04:22)

By looking at the linear transformation from ℂ → ℝ which sends 1 to (1, 0) and 𝑖 to (0, 1), we
could also define a field structure on ℝ2 and we can identify ℂ with ℝ2 .

Now we can define multiplication on ℝ2 . For (𝑎, 𝑏), (𝑐, 𝑑) 𝜖 ℝ2 , define

(𝑎, 𝑏)(𝑐, 𝑑) = (𝑎𝑐 − 𝑏𝑑, 𝑎𝑑 + 𝑏𝑐). This is the definition for the multiplication in ℝ2 and this
particular definition is something which we had obtained from considering the multiplication
of cosets in ℝ[𝑥]/⟨𝑥 2 + 1⟩.
(Refer Slide Time: 07:34)

24
In this course if 𝑧 is a complex number then 𝑧 = 𝑎 + 𝑖𝑏 for 𝑎, 𝑏 𝜖 ℝ. Then 𝑎 is called the real
part of z and is denoted ℜ𝔢(𝑧) and at the same time 𝑏 is called the imaginary part of 𝑧 and is
denoted ℑ𝔪(𝑧).

(Refer Slide Time: 09:12)

If 𝑎 = 0, then 𝑧 is called purely imaginary and if 𝑏 = 0 it is a real then 𝑧 is purely real or is a

real number.
The vector spaceℝ2 is something which you might be very familiar with, it is the plane which
we are usually accustomed to and we are familiar with geometry there and because ℂ can be
identified with ℝ2 , the field of complex numbers is also sometimes referred to as the complex
plane.
(Refer Slide Time: 11:09)

25
We have the 𝑥-axis and the 𝑦-axis on ℝ2 and since ℂ can be identified with ℝ2 by the linear
transformation sending 1 to (1, 0) and therefore the 𝑥-axis consists of real numbers and is called
the real axis. Hence when we say real axis, we mean the 𝑥-axis.

And in 𝑎 + 𝑖𝑏, if 𝑎 is 0, then 𝑖𝑏 will be mapped to (0, 𝑏) and that will be the 𝑦-axis.
The 𝑦-axis consists of purely imaginary numbers and is called the imaginary axis.
(Refer Slide Time: 12:36)

By convention for any complex number 𝑧, 𝑧 0 ≔ 1 and also 00 : = 1. For 𝑛, a non-negative

integer, define 𝑧 𝑛+1 = 𝑧 ∙ 𝑧 𝑛 , assuming that 𝑧 𝑛 has been already defined.
(Refer Slide Time: 13:33)

1
For 𝑛 < 0, let 𝑛 = – 𝑚, 𝑚 > 0, then for 𝑧 ≠ 0, 𝑧 𝑛 ≔ .
𝑧𝑚

26
These are the usual notations which we have borrowed from field theory that is how we will
define the exponentiation. In this course, we will be defining exponents with arbitrary
exponents, but we will come to that later.
(Refer Slide Time: 14:31)

Let us next introduce the important concept of conjugation. Let 𝑧 = 𝑎 + 𝑖𝑏 𝜖 ℂ. We define the
conjugate of 𝑧 to be the complex number 𝑧̅ ≔ 𝑎 − 𝑖𝑏.
(Refer Slide Time: 15:34)

It is immediate from the definition that ̅̅̅̅̅̅̅̅

𝑧 + 𝑤 = 𝑧̅ + 𝑤
̅ and ̅̅̅̅̅̅
𝑧 ∙ 𝑤 = 𝑧̅ ∙ 𝑤
̅, the proof of which
is left as an exercise. You could write it down explicitly in terms of coordinates, compute both
left hand side and right hand side and see that they do match. If 𝑧 = 𝑎 + 𝑖𝑏 𝜖 ℂ satisfies
𝑧 = 𝑧̅ ,then 𝑧 = 𝑧̅ ⇒ 𝑎 + 𝑖𝑏 = 𝑎 − 𝑖𝑏 ⇒ 2𝑖𝑏 = 0 ⇒ 𝑏 = 0.
Since 1 and 𝑖 are forming a basis, 𝑏 = 0 tells us that 𝑧 is a real. If that is real, then naturally

27
𝑏 = 0, and therefore𝑧 = 𝑧̅.
(Refer Slide Time: 17:05)

Lemma: A complex number 𝑧 satisfies 𝑧 = 𝑧̅ if and only if 𝑧 𝜖 ℝ.

𝑧+𝑧̅
Now observe that 𝑧 + 𝑧̅ = (𝑎 + 𝑖𝑏) + (𝑎 − 𝑖𝑏) = 2𝑎 ⇒ ℜ𝔢(𝑧) = . Similarly
2
𝑧−𝑧̅
ℑ𝔴(𝑧) = .
2𝑖

(Refer Slide Time: 18:49)

The conjugation is a notion which you will be familiar with from linear algebra. Geometrically,
conjugation is the linear transformation which is the reflection along the real axis.
Now let us take the first step towards studying analysis and to do that let us try to see the
topology on the field of complex numbers.

28
But before we explore all that, let us define a function. Let us take any complex number 𝑧,
define𝑁(𝑧) = 𝑧 ∙ 𝑧̅. This is the function from ℂ to ℂ, but there are a few observations
immediately, which we can make.
̅̅̅̅̅̅ = ̅̅̅̅̅
Notice that for 𝑧 𝜖 ℂ, 𝑁(𝑧) 𝑧 ∙ 𝑧̅ = 𝑧̅ ∙ 𝑧̅̅ = 𝑧̅ ∙ 𝑧 = 𝑁(𝑧)

Then we just observed a few minutes back that a complex number is equal to its conjugate if
and only if it is a real number. This implies that 𝑁(𝑧) belongs to ℝ and therefore 𝑁 ∶ ℂ ⟶ ℝ.
(Refer Slide Time: 23:04)

Also check that 𝑁(𝑧̅) = 𝑁(𝑧) and 𝑁(𝑧 ∙ 𝑤) = 𝑁(𝑧) ∙ 𝑁(𝑤). This function is many times
referred to as the field norm and it is an algebraic concept. It is a field norm of ℂ over ℝ.
Our next goal would be to explicitly get hold of the metric with which will be working on the
field of complex numbers. For that for the linear algebra we can use the fact that ℂ is vector
space over ℝ and we know that ℂ has an inner product over ℝ defined by 〈𝑧, 𝑤〉 ≔ ℜ𝔢(𝑧𝑤
̅).

This is the standard inner product which we generally consider on ℂ considered over the field
of real numbers.
(Refer Slide Time: 25:21)

29
With this inner product, ℂ is an inner product space. You must check explicitly that this defines
an inner product on ℂ over ℝ by checking all the properties of an inner product.
(Refer Slide Time: 26:05)

This inner product can be used to define a norm given by |𝑧| = 〈𝑧, 𝑧〉1/2 = (ℜ𝔢(𝑧 ∙ 𝑧̅))1/2, but
we have already defined 𝑁(𝑧) = 𝑧 ∙ 𝑧̅ and hence |𝑧| = [ℜ𝔢(𝑁(𝑧))]1/2. Since we have proved
that 𝑁(𝑧) is a real number ℜ𝔢(𝑁(𝑧)) = 𝑁(𝑧). Hence |𝑧| = 𝑁(𝑧)1/2 .
(Refer Slide Time: 27:57)

30
Now let us see what happens in coordinates. If 𝑧 = 𝑎 + 𝑖𝑏, 𝑁(𝑧) = (𝑎 + 𝑖𝑏)(𝑎 − 𝑖𝑏) = 𝑎2 +
𝑏 2 . Then we can define norm in terms of the coordinates of 𝑧 as |𝑧| = (𝑎2 + 𝑏 2 )1/2
(Refer Slide Time: 28:28)

If 𝑧 is real, then 𝑏 = 0 and this would imply that the norm of 𝑧, |𝑧| = (𝑎2 + 0)1/2 = |𝑎|.
Therefore, the norm which we just defined naturally extends the absolute value which we are
familiar with on the field of real numbers and hence the norm on ℂ will also be called the
absolute value.
In any normed vector space, we can talk about a metric.
(Refer Slide Time: 30:28)

31
Let us now define a metric on this particular normed vector space. In other words, let us define
a metric on ℂ using the norm which we just defined. We define a metric 𝑑 ∶ ℂ × ℂ → ℝ given
by, for 𝑧, 𝑤 𝜖 ℂ, 𝑑(𝑧, 𝑤) = |𝑧 − 𝑤|. This will turn out to be a metric on ℂ. I will leave it as an
exercise for you to check that this indeed coincides with the standard metric on ℝ2 .
The standard metric on ℝ2 is 𝑑((𝑎, 𝑏), (𝑐, 𝑑)) = √(𝑎 − 𝑐)2 + (𝑏 − 𝑑)2 .

Hence the basic topological notions of open sets, closed sets, boundary, convergence, limits,
continuity all these things are the usual topological notions we are familiar with from a course
on real analysis and hence the topological notions on ℂ coincide with ℝ2 .

The metric 𝑑 is defined using the absolute value which in turn is being defined using the field
norm, which is an algebraic property.
If you go to different construction of the field of complex numbers, there will be a
corresponding field norm there and we can do all these things again. So, basically the topology
and the analysis that follows is going to be invariant under which choice of complex numbers
you pick.

32
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture – 3
Topology on the Complex Plane

In the last lecture, we defined a metric on the field of complex numbers. The metric was
defined using an absolute value, which was a natural extension of the notion of absolute
value on the field of real numbers. The absolute value turned out to be the square root
of the field norm and we remarked that because of this, it does not matter which copy of
the complex field that we work on, the analysis will be consistent.
(Refer Slide Time: 00:46)

Let us start by recalling the metric that we defined. Recall that the metric defined
on C is given by d (z, w) := |z − w|, z, w ∈ C. In this lecture, we will do a quick recap of
the familiar topological notions on the complex plane. The material will be something
which you would have already seen in multivariate real analysis course. The intention
is also to set the notations for the rest of this course. First thing to do is describing the
balls in C. For a point z 0 ∈ C and > 0, denote the ball of radius r around z 0 to be D(z 0 , r ).
Recall that D(z 0 , r ) := {z ∈ C : |z − z 0 | < r }. The D here is actually for disc, since balls in C
1

33
2

are also called discs.

If we are to work with general metric space, which we will do sometimes in this lecture
itself, we will use the word B instead of D, which should mean the same thing.
(Refer Slide Time: 03:24)

A subset D ⊆ C is said to be open if for every z ∈ D, ∃r > 0 such that D(z, r ) ⊆ D.

This can be restated in another way as, if every point in D is an interior point of D, we
say that D is open.
A subset F ⊆ C is said to be closed if C, the complement of F , C \ F is open in C.
(Refer Slide Time: 04:54)

34
 3

A point z ∈ C is said to be a limit point of a subset D ⊆ C, if for every ² > 0, D(z, ²) ∩ D

contains a point other than z.
In a metric space, this is also the same as demanding that you have a sequence in D
converging to z, where the sequence consists of points other than z.

E XERCISE 1. A subset F ⊆ C is closed if and only if it contains all its limit points.

(Refer Slide Time: 07:17)

35
4

Properties of open sets

(1) C and ∅(empty set) are open subsets of C.

(2) If Ω1 , Ω2 , . . . Ωn are open sets in C, then Ω1 ∩ Ω2 ∩ . . . Ωn are also open. That is
finite intersection of open sets is open.
(3) If {Ωα }α∈A is a collection of open sets of C, then Ωα is open in C.
S
α∈A

Verification of these properties left as an exercise to the reader. Check that each of the
sets mentioned satisfy the definition of open set.
(Refer Slide Time: 09:26)

Hence we can say that the collection of all open sets on C form a topology on C.

Let D ⊆ C, we define the interior of D, denoted by D o , to be the union of all open

sets in C which are contained in D.
That is, interior of D, D o := {Ω : Ω ⊆ D and Ω is open in C}.
S

From the properties stated above, it is clear that D o is an open set.

We define closure of D to be the set, D := {F : F ⊆ D and F is closed in C.
T

Using De Morgan’s law, we can say that the arbitrary intersection of closed sets will
again be a closed set. Hence D is a closed set. The interior is the biggest open set con-
tained in D and closure, D, is the smallest closed set which contains D.
(Refer Slide Time: 13:22)

36
 5

Let E ⊆ D. We say that E is dense in D, if the closure of E in D is D, where the closure

of E in D, in general metric space, is with respect to the collection of open and closed
sets in D with metric on D obtained by restricting the metric on C to D.
Anyway in the case of our complex plane, closure of E in D is the intersection of closure
of E in C and D.
(Refer Slide Time: 16:06)

A sequence {z n }n∈N in C is said to converge to z if |z n − z| −→ 0 as n −→ ∞.

Let Ω ⊆ C be open and a function, f : Ω −→ C is said to be continuous if f −1 (D) is open
in Ω whenever D is open in C. It is the same as saying that if there is a sequence {z n }n∈N

37
6

in Ω which converges to z in Ω, then f (z n ) −→ f (z).

(Refer Slide Time: 17:55)

We know that if {z n }n∈N , {w n }n∈N are two convergent complex sequences, then
lim (z n + w n ) = lim z n + lim w n and lim (z n × w n ) = lim z n × lim w n . From here
n→∞ n→∞ n→∞ n→∞ n→∞ n→∞
we can conclude that the addition and multiplication operation on complex plane is
continuous. This can be checked by considering the triangle inequality and the fact that
|zw| = |z||w|.
(Refer Slide Time: 19:52)

38
 7

Also notice that lim z n = lim z n which tells us that conjugation is also a continuous
n→∞ n→∞
function, which follows form the observation that |z| = |z|.
Our next aim is to show that the field of complex number is complete. For that we use
the fact that R is complete.
p
If z = a + i b, then the absolute value of z, |z| = a 2 + b 2 . Then
p
|a|, |b| ≤ a 2 + b 2 ≤ |a| + |b| =⇒ |Re(z)|, |Im(z)| ≤ |z| ≤ |Re(z)| + |Im(z)|.
(Refer Slide Time: 23:05)

39
8

Because of these inequalities, if {z n } is a sequence of complex numbers which con-

verges to z, then Re(z n ) → Re(z), similarly Im(z n ) → Im(z n ).
We can say it another way also, that if {z } is sequence of complex numbers converges to
a point z ∈ C if and only if Re(z n ) → Re(z) and Im(z n ) → Im(z).

C OROLLARY 2. The field of complex number C is complete with respect to the metric
defined as above.

If {z n } is a Cauchy sequence, then {Re(z n )} and {Im(z n )} both are Cauchy and by the
completeness of R, there exist a, b such that Re(z n ) → a and Im(z n ) → b. Let z = a + i b.
Then z n → z. Hence every Cauchy sequence has a limit, therefore C is complete.
(Refer Slide Time: 26:25)

40
 9

Next important notion is connectedness. We may work with the general metric space
rather than C for defining the connectedness.
Let (X , d ) be metric space. We say that X is separated if there exist disjoint non-
S
empty open subsets U ,V of X such that X = U V .
For example, let X ⊆ C, X = D(0, 1) ∪ D(4, 1) and let us use the metric on C to restrict
it to X , then U = D(0, 1) and V = D(4, 1) is a separation.
Also, if we had taken closed discs instead of open discs, that is if X 0 = D(0, 1) D(4, 1)
S

and look at a new metric space. Then again, the reader should verify that U = D(0, 1),V =
D(4, 1), that are not open in C, are open in X 0 in the subspace topology, then U ∩ V = ∅
and the union will give us X 0 .Hence X 0 is also separated.
(Refer Slide Time: 32:10)

41
10

We say that a metric space (X , d ) is connected if there does not exist a separation of
X.
We can reformulate the definition as, X is connected if and only if X does not contain
a proper subset which is both open and closed and simultaneously in X .
The examples we have mentioned above for the spaces that are separated, X = D(0, 1) ∪
D(4, 1), we can see that D(0, 1) is open and closed in X , since D(0, 1) is open in C, under
subspace topology it is open in X and also the compliment of D(0, 1) in X is an open set
in X , thus D(0, 1) is closed. Similarly D(4, 1) is both open and closed in X which implies
that X is separated. Similarly in the case of X 0 , D(0, 1) and D(4, 1) are both open and
closed in X 0 .
Example: X ⊆ R is connected if and only if X is an interval.
(Refer Slide Time: 35:36)

42
 11

So, here we have given a characterization of how the connected sets in R behave. We
will not be able to give this kind of a characterization on the complex plane. However,
we will give an alternate description of how connected open sets in C will behave.

P ROPOSITION 3. Let X , Y are metric spaces and f : X −→ Y is a continuous function.

Then f (x) is connected if X is connected.

P ROOF. The proof is by contradiction. Let Z = f (X ) ⊆ Y . Suppose Z is not con-

S
nected.Then Z = A B where A, B are non-empty, open in Z and A ∩ B = ∅.
Claim: f −1 (A) and f −1 (B ) are open in X .
Since A = U ∩ Z , U open in Y , f −1 (A) = f −1 (U ), and f is continuous =⇒ f −1 (U ) =
f −1 (A) is open in X . Similarly f −1 (B ) is open in X . f is surjective onto Z and A, B are
non-empty =⇒ f −1 (A), f −1 (B ) are non-empty. Also Z = A ∪ B =⇒ f −1 (A) ∪ f −1 (B ) = X
also f −1 (A) ∩ f −1 (B ) = ∅. Then X is separated by f −1 (A) and f −1 (B ) which is a contra-
diction to our assumption that X is connected. Hence f (X ) is connected. 

(Refer Slide Time: 41:47)

43
12

A path in a metric space from a point x ∈ X to y ∈ X is a continuous map

γ : [0, 1] −→ X such that γ(0) = x and γ(1) = y.
(Refer Slide Time: 42:36)

Example: Consider z ∈ D(0, 1) and γ : [0, 1] −→ C such that γ(t ) = t z. Then γ is a

straight line from 0 to z.
Let γ1 be a path from x to y and γ2 be a path from y to z in a metric space X . Define


γ1 (2s) if, s ∈ [0, 1 ]

2
σ(s) =
γ (2s − 1) if, s ∈ [ 1 , 1]

2 2

Then σ is a continuous map from [0, 1] to X such that σ(0) = x, σ(1) = z.

(Refer Slide Text: 46:44)

44
 13

T HEOREM 4. A open subset Ω ⊆ C is connected if and only if there exists a path for
every pair of points z, w ∈ Ω there exists a path from z to w.

P ROOF. (⇐) Suppose Ω is separated. Then ∃U ,V such that Ω = U ∪ V where U ,V

are non-empty open subsets of Ω. Let z ∈ U and w ∈ V . If there exists a path γ form
z to w, then γ : [0, 1] −→ X is continuous and γ(0) = z and γ(1) = w. Then [0, 1] =
f −1 (U ) f −1 (V ), which is a contradiction since [0,1] is connected. Therefore, there does
S

not exists a separation of Ω, hence Ω is connected.

(⇒) Let us assume that Ω is connected. Fix z 0 ∈ Ω. Let A := {z ∈ Ω : ∃ a path from z to z 0 }.
Claim: A is open.
Let z ∈ A. Since Ω is open ∃ r > 0 such that D(z, r ) ⊆ Ω. Since z ∈ A, ∃ a path γ from z 0
to z. For any w ∈ D(z, r ) consider the path γ1 : [0, 1] −→ Ω given by γ1 (s) = (1 − s)z + sw.
Then γ1 is a straight line path from z to w and is contained in D(z, r ). Define


γ(2s) if, s ∈ [0, 1 ]

2
σ(s) =
γ (2s − 1) if, s ∈ [ 1 , 1]

1 2
Hence σ is a path from z 0 to w =⇒ w ∈ A =⇒ D(z, r ) ⊆ A =⇒ A is open.
Claim: A is closed.
Let z ∈ Ω \ A. Since Ω is open, we have r > 0 such that D(z, r ) ⊆ Ω. If D(z, r ) ∩ A 6= ∅,

45
14

then ∃w ∈ D(z, r ) ∩ A. This would give a path from z 0 to z through w contradiction to

our assumption that z ∈ Ω \ A. Hence D(z, r ) ∩ A = ∅ =⇒ Ω \ A is open =⇒ A is closed.
Hence A is both open and closed. Therefore either A = Ω or A = ∅. Since z 0 ∈ A, A 6=
∅ =⇒ A = Ω. Thus any pair of points in a can be connected by a path. 

46
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture – 4
Topology on the Complex Plane(Continued)

(Refer Slide Time: 00:16)

Let (X , d ) be a metric space. Let U be a collection of open sets in X . We say that U

is an open cover of a subset K ⊆ X if K ⊆ {U : U ∈ U }.
S

(Refer Slide Time: 02:19)

We say that a subset K ⊆ X is compact if for every open cover U of K , there exists
n
finitely many elements U1 ,U2 , . . .Un ∈ U such that K ⊆
S
Uk .
k=1
(Refer Slide Time: 03:33)
1

47
2

Example:

• A finite subset of a metric space is compact.

• ∅ is compact.
• Let {x n } be a sequence of points converging to x 0 . Then A = {x n : n ∈ N ∪ {0}} is
compact in X .

(Refer Slide Time: 05:03)

48
 3

P ROPOSITION 1. In a metric space, a compact set is closed.

P ROOF. Let K ⊆ X be compact and x 0 ∈ X \ K . Define B n := {x ∈ X : d (x, x 0 ) ≤ n1 }. Let

U = {Un }n∈N , where Un = X \ B n .
∞
S ∞
T ∞
T
Then X \ ( Un ) = (X \Un ) = B n = {x 0 }. Since we choose x 0 not in K =⇒ K ⊆
n=1 n=1 n=1
∞
Un . Hence U is an open cover of K . Since K is compact, there exists a finite subcover
S
n=1
of U . i.e, ∃n 0 ∈ N such that K ⊆ U1 ∪U2 ∪ · · · ∪Un0 .
(Refer Slide Time: 08:02)

n
S0
1 1
Let m > n 0 , now reader can verify that B (x 0 , m )∩( Un ) = ∅ =⇒ B (x 0 , m ) ⊂ X \K.
n=1
Hence X \ K is open =⇒ K is closed.


E XERCISE 2. A closed subset of a compact set is closed.

(Refer Slide Time: 11:10)

49
4

Let X be a metric space. We say that X is limit point compact if every infinite subset
of X has a limit point.

P ROPOSITION 3. Let X be a compact metric space, then X is limit point compact.

P ROOF. Let A be an infinite subset of X . Suppose A does not have a limit point, then
A is closed in X =⇒ X \ A is open.
Refer Slide Time: 12:52

Given x ∈ A, then x is not a limit point of A. Hence ∃ a neighborhood U x of x such

that A ∩U x = {x}. Define U := {U x : x ∈ A} {X \ A}. Then U is an open cover of X . Since
S

X is compact ∃ x 1 , x 2 , . . . x n ∈ A such that X = U x1 ∪U x2 ∪ · · · ∪U xn ∪ (X \ A).

(Refer Slide Time: 16:04)

50
 5

But A ∩U xi = {x i } and A ∩ (X \ A) = ∅ =⇒ A is finite which is a contradiction. Hence

∃ a limit point of A. 

(Refer Slide Time: 17:15)

We say that metric space is sequentially compact if every sequence in the metric
space has that convergent subsequence.

P ROPOSITION 4. Let X be a metric space. If X is limit point compact, then X is se-

quentially compact.

P ROOF. Let {x n } be a sequence in X and A = {x n : n ∈ N}. Then A can either be finite

or infinite.
(Refer Slide Time: 19:48)

51
6

If A is finite, ∃n k such that n k < n k+1 , k ∈ N and x nk = x 0 for some x 0 ∈ X which

repeats infinitely many times in the sequence. Then lim x nk = x 0 Hence {x n } has a con-
k→∞
vergent subsequence. If A is infinite, then A has a limit point x 0 , since X is limit point
compact.
Claim: There exists a subsequence of {x n } converging to x 0 .
Let n 1 ∈ N be such that x n1 ∈ B (x 0 , 1). Let n 2 > n 1 be such that x n2 ∈ B (x 0 , 12 ).
(Refer Slide Time: 21:58)

52
 7

Reader should verify the existence of such an n 2 .(Hint: If there does not exists such an
n 2 such that n 2 > n 1 and x n2 ∈ B (x 0 , 12 ). Then d (x k , x 0 ) > 12 ∀k > n 1 . Let δ = min{d (x 0 , x 1 ),
d (x 0 , x 2 ), . . . d (x 0 , x n1 ), 21 }. Then consider B (x 0 , δ).)
Construct inductively x nk such that n k > n k−1 and x nk ∈ B (x 0 , k1 ). Then {x nk } con-
verges to x 0 . Hence {x nk } is a convergent subsequence of {x n }. Hence X is sequentially
compact.


Now we have proved that if a metric space is compact then it is limit point compact.
If a metric space is limit point compact then it is sequentially compact. We can complete
’the cycle’ by proving that any sequentially compact metric space is a compact metric
space. For that we need to use a theorem called Lebesgue number lemma.
(Refer Slide Time: 27:48)

L EMMA 5 (Lebesgue number lemma). Let X be sequentially compact and U be an

open cover of X . Then ∃ δ > 0 such that for x ∈ X , ∃U ∈ U such that B (x, δ) ⊂ U .

P ROOF. Suppose there does not exists such a δ > 0. Then for every n ∈ N, ∃ x n ∈ X
such that B (x n , n1 ) is not contained in any element of U . Since X is sequentially com-
pact, ∃ a subsequence {x nk } such that x nk −→ x 0 , x 0 ∈ X . U is an open cover of X , x 0 ∈
X =⇒ ∃ ² > 0 such that B (x o , ²) ⊂ U for some U ∈ U . Let n k be large enough such that
1
nk < 2² and d (x nk , x o ) < 2² .
Claim: B (x nk , n1 ) ⊂ U .
k

53
8

Let y ∈ B (x nk , n1 ), then d (x 0 , y) ≤ d (x 0 , x nk ) + d (x nk , y) < ² =⇒ y ∈ B (x 0 , ²) ⊂ U =⇒

k

B (x nk , n1 ) ⊂ U .
k

(Refer Slide Time: 31:22)

This is a contradiction to the assumption that B (x n , n1 ) is not contained in any ele-

ment of U . Hence ∃ δ > 0 satisfying the condition in the lemma. 

T HEOREM 6. Let X be sequentially compact. Then X is compact metric space.

P ROOF. Claim: Given ² > 0, ∃ finitely many points x 1 , x 2 , . . . x n such that X = B (x 1 , ²)∪
B (x 2 , ²) ∪ · · · ∪ B (x n , ²).
Suppose ² > 0 is such that there does not exists finitely many points x 1 , x 2 , . . . x n
such that X = B (x 1 , ²) ∪ B (x 2 , ²) ∪ · · · ∪ B (x n , ²). Pick x 1 ∈ X . Inductively x n+1 6∈ B (x 1 , ²) ∪
B (x 2 , ²)∪· · ·∪B (x n , ²). Consider the sequence {x n }. But X is sequentially compact, hence
∃ a converging subsequence of {x n }. But from our assumption d (x j , x i ) > ² and hence
no subsequence will be Cauchy, which is a contradiction.
Hence ∃ finitely many points x 1 , x 2 , . . . x n such that X = B (x 1 , ²)∪B (x 2 , ²)∪· · ·∪B (x n , ²)(This
particular notion sometimes also called as total boundedness).
(Refer Slide Time: 38:49)

54
 9

Let U be an open cover of X . Then by using Lebesgue number lemma, ∃ δ > 0 such
that given x ∈ X , B (x, δ) ⊂ U x for some U x ∈ U . By the above claim ∃ finitely many points
such that X = B (x 1 , δ) ∪ B (x 2 , δ) ∪ · · · ∪ B (x n , δ) =⇒ X ⊂ U x1 ∪U x2 ∪ · · · ∪U xn . Hence X is
compact. 

With this, we complete our circle of ideas. From now on, because our complex plane
is metric space with respect to the metric we have define, these notions coincide in our
complex plane and we will freely be using these notions interchangeable.

55
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture – 2.1
Problem Session

(Refer Slide Time: 00:25)

P ROBLEM 1. Let n > 1 and suppose c 0 > c 1 > · · · > c n > 0 be real numbers and let
p(z) = c 0 +c 1 z +· · ·+c n z n . Then prove that there does not exist a root of p whose absolute
value does not exceed 1.

S OLUTION 1. We want to show that if p(z 0 ) = 0 for some z 0 ∈ C, then |z 0 | > 1.

1

56
2

Consider the polynomial (1 − z)p(z),

(1 − z)p(z) = (c 0 + c 1 z + · · · + c n z n ) − (c 0 z + · · · + c n z n+1 )

= c 0 + (c 1 − c 0 )z + · · · + (c n − c n−1 )z n − c n z n+1

= c 0 − (c 0 − c 1 )z + · · · + (c n−1 − c n )z n + c n z n+1 .
¡ ¢

Then note that all the coefficients of the terms inside the parentheses are positive num-
bers.
Let z 0 ∈ C such that p(z 0 ) = 0. Then we want to prove that |z 0 | > 1. We may consider
the case when z 0 is in the unit disc, |z 0 | ≤ 1
(Refer Slide Time: 05:50)

57
 3

c 0 = |c o | = |(c 0 − c 1 )z 0 + · · · + (c n−1 − c n )z 0n + c n z 0n+1 |

= |z 0 | |(c 0 − c 1 ) + · · · + (c n−1 − c n )z 0n−1 + c n z n |

¡ ¢

≤ |z 0 | |(c 0 − c 1 )| + · · · + |(c n−1 − c n )z 0n−1 | + |c n z n |

¡ ¢

= |z 0 | (c 0 − c 1 ) + · · · + (c n−1 − c n )|z 0n−1 | + c n |z n |

¡ ¢

≤ (c 0 − c 1 ) + (c 2 − c 1 ) + · · · + (c n−1 − c n ) + c n

= c0

Since we started with c 0 and ended with c 0 , all the inequalities in the above equations
will be equalities.
(Refer Slide Time: 10:31)

58
4

Therefore, we have

|z 0 | |(c 0 − c 1 ) + · · · + (c n−1 − c n )z 0n−1 + c n z n | = |z 0 | |(c 0 − c 1 )| + · · · + |(c n−1 − c n )z 0n−1 | + |c n z n |

¡ ¢ ¡ ¢

If z 0 = 0, then p(z 0 ) = 0 =⇒ c 0 = 0, which is a contradiction. Hence z 0 6= 0 =⇒

|(c 0 − c 1 ) + · · · + (c n−1 − c n )z 0n−1 + c n z n | = |(c 0 − c 1 )| + · · · + |(c n−1 − c n )z 0n−1 | + |c n z n |

¡ ¢ ¡ ¢

(Refer Slide Time: 12:53)

Reader should verify using Cauchy–Schwarz inequality that if |a + b| = |a| + |b|, then
a = λb for λ ∈ R.
Then, by an induction argument that reader should verify,
(Refer Slide Time: 14:44)

59
 5

c 0 −c 1
for λ ∈ R, (c 0 − c 1 ) = λ(c 2 − c 1 )z 0 =⇒ z 0 = λ(c 2 −c 1 )
∈ R.
Note that if z 0 > 0, then p(z 0 ) = c 0 + c 1 z 0 + · · · + c n z 0n ≥ c 0 6= 0. Hence z 0 cannot be root.
If z 0 < 0, z 0 = −a, a > 0.
If n is odd, p(z 0 ) = (c 0 − c 1 a) + a 2 (c 2 − c 3 a) + · · · + a n−1 (c n−1 − c n a), then each term inside
the parentheses of the sum on RHS is positive. Hence in this case also p(z 0 ) > 0.
If n is even, p(z 0 ) = (c 0 −c 1 a)+a 2 (c 2 −c 3 a)+· · ·+a n−2 (c n−2 −c n−1 a)+a n c n , then here
also each term inside the parentheses of the sum on RHS is positive =⇒ p(z 0 ) 6= 0.

Hence we have proved that if |z 0 | ≤ 1 =⇒ p(z 0 ) 6= 0.

P ROBLEM 2. Let z, w ∈ C such that (1 + |z|2 )w = (1 + |w|2 )z. Then either z = w or

zw = 1.

60
6

S OLUTION 2. We know that |z|2 = zz. Then,

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

w + zzw = z + w w z

w(1 − w z) = z(1 − zw).

If z = 0 =⇒ w = o, hence z = w.
Suppose z 6= 0 and zw 6= 1, then
w
z
= (1−zw)
(1−w z)
=⇒ | wz | = | 1−w
1−zw
z
| = |1−zw|
|1−w z|
= 1 =⇒ |w| = |z| =⇒ (1 + |w|2 ) = (1 + |z|2 ).
Hence (1 + |w|2 )z = (1 + |z|2 )w =⇒ z = w.
(Refer Slide Time: 25:46)

D EFINITION 3. A path γ from z to w is a polygonal if ∃ z = z 0 , z 1 , . . . z n−1 , z n = w ∈ C

such that γ = γz0 ,z1 · γz1 ,z2 · · · · γzn−1 ,zn where γzi ,zi +1 (s) = (1 − s)z i + sz i +1 .

61
 7

That is polygonal path between two points in the complex plane is obtained by con-
catenating straight line paths between finite number of points.
(Refer Slide Time: 29:18)

P ROBLEM 3. Let Ω ⊆ C be a connected open set. Then prove that given z, w ∈ Ω, ∃ a

polygonal path from z to w consisting of straight lines parallel to the real or imaginary
axis.

(Refer Slide Time: 30:56)

62
8

S OLUTION 4. Suppose z 0 ∈ C and r > 0. Then we shall first prove that D(z 0 , r ) satisfies
the given condition.
If z 00 = a 0 +i b 0 and z = a +i b be two points inside D(z 0 , r ). Let z 00 = a +i b 0 ∈ D(z 0 , r ).
(Reader is strongly suggested to draw a picture with given details to get a good idea.)
γz 0 ,z 00 (s) = (1 − s)z 00 + sz 00 (paralell to real axis)
0

γz 00 ,z (s) = (1 − s)z 00 + sz(paralell to imaginary axis)

Define γ = γz 0 ,z 00 · γz 00 ,z , then γ will be a polygonal path from z 00 to z with desired
0

condition.
(Refer Slide Time: 36:10)

63
 9

Hence the disc D(z 0 , r ) satisfies the condition of the problem.

Now, fix z 0 ∈ Ω, define A = {z ∈ Ω : z can be joined to z 0 by a polygonal path parallel
to axes}. Now it is left to reader to verify that A is both open and closed. Since z 0 ∈ A =⇒
A 6= ∅ =⇒ A = Ω, as Ω is connected.

(Refer Slide Time: 38:28)

64
10

D EFINITION 5. Let X , Y be metric spaces. A function f : X −→ Y is said to be a closed

map if f (C ) is closed in Y whenever C is closed in X .

P ROBLEM 4. Let f : X −→ Y be a continuous mapping from a compact metric space

to another metric space. Then f is a closed map.

(Refer Slide Time: 40:30)

65
 11

S OLUTION 6. Let E ⊂ X be a closed subset of X . Since X is compact, E is compact.

Claim: f (E ) is compact.
Let U = {Uα }α∈A be an open cover of f (E ). Define V := { f −1 (Uα )}α∈A , then V is an
open cover of E .
(Refer Slide Time: 42:14)

66
12

Now it is left as an exercise to reader to conclude that f (E ) is also compact. Then

f (E ) is compact in Y =⇒ f (E ) is closed.

T HEOREM 5 (Heine-Borel). A subset K ⊆ Rn is compact if and only if K is closed and

bounded in Rn .

P ROOF. (⇒) Let K be compact, then K is closed. Consider U = (B (0, n))n∈N .

Then U is an open cover of K . Since K is compact, ∃ n ∈ N such that

K ⊂ B (0, 1) ∪ · · · ∪ B (0, n)

. Hence K is bounded.
(⇐) Assume that K is closed and bounded.
(Refer Slide Time: 47:26)

67
 13

K is bounded =⇒ K ⊂ [a 1 , b 1 ] × · · · × [a n , b n ].
Claim: [a, b] is sequentially compact.
(Refer Slide Time: 49:14)

68
14

Let {x n }n∈N be a sequence in [a, b]. Let A = {x n : n ∈ N}, then ∃ a subsequence con-
verging to sup(A) by using Bolzano- Weierstrass theorem. Since [a, b] is a closed set,
supremum will belong [a, b]. Hence [a, b] is sequentially compact.
Claim: [a, b] × [c, d ] is sequentially compact.
Let {(x n , y n )}n∈N be a sequence in [a, b] × [c, d ]. Let {x nk }k∈N be a convergent subse-
quence of {x n }n∈N in [a, b], whose existence is obtained by above claim. Similarly, let
{y nk }k∈N be a convergent subsequence of {y n }n∈N in [c, d ].
(Refer Slide Time: 53:18)

69
 15

Now we obtain a subsequence {(x nk , y nk )}k∈N of {(x n , y n )}n∈N . Hence [a, b] × [c, d ] is
also sequentially compact.
By induction [a 1 , b 1 ] × · · · × [a n , b n ] is also sequentially compact. We know that in
a metric space sequentially compactness coincides with compactness, hence [a 1 , b 1 ] ×
· · ·×[a n , b n ] is compact set. Since K is a closed subset of a compact set, K is also compact.


70
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 2.2
Isometries on the Complex Plane

In the last week we defined the field of complex numbers and proved that any com-
plex field will be isomorphic to it. We also defined a metric on the complex numbers
using the absolute value, which happened to be the square root of the norm function.
And we remarked that the explicit construction of the complex plane should not affect
the study of analysis on the field of complex numbers. Thereafter we studied some im-
portant topological notions on the complex plane.
In this week we will study some functions on the complex plane and some of its
geometric properties. Let us begin this lecture by discussing isometries on the field of
the complex numbers.
(Refer Slide Time: 01:00)

Definition
A function f : C −→ C is called an isometry if | f (z) − f (w)| = |z − w| ∀z, w ∈ C.
1

71
2

That is, isometry is a function which preserves the distance between two points.
Claim: Let f be an isometry on C such that f (0) = 0. Then 〈 f (z), f (w)〉 = 〈z, w〉∀z, w ∈ C.
(Refer Slide Time: 04:56)

P ROOF. | f (z)− f (w)|2 = 〈 f (z)− f (w), f (z)− f (w)〉 = | f (z)|2 +| f (w)|2 −2〈 f (z), f (w)〉 =⇒
| f (z)|2 +| f (w)|2 −| f (z)− f (w)|2 2
+|w|2 −|z−w|2
〈 f (z), f (w)〉 = 2
= |z| 2
= 〈z, w〉. 

P ROPOSITION 1. Let f : C −→ C be an isometry such that f (0) = 0. Then f is a linear

map.

P ROOF. (Refer Slide Time: 09:26)

72
 3

For every α ∈ R,
| f (αz + w) − α f (z) − f (w)|2 = 〈 f (αz + w) − α f (z) − f (w), f (αz + w) − α f (z) − f (w)〉 =
| f (αz+w)|2 +|α|2 | f (z)|2 +| f (w)|2 −2α〈 f (αz+w), f (z)〉−2〈 f (αz+w), f (w)〉+2α〈 f (z), f (w)〉 =
|αz+w|2 +|α|2 |z|2 +|w|2 −2〈αz+w, αz〉−2〈αz+w, w〉+2〈αz, w〉 = |(αz+w)−αz−w|2 = 0.
Hence f (αz + w) = α f (z) + f (w). 

(Refer Slide Time: 10:47)

Let us now look at how f acts on {1, i }. If z ∈ C, then z = a + i b and f (z) = a f (1) +
b f (i ). Suppose f (1) = α, where α ∈ C. Then, 1 = |1 − 0| = | f (1) − f (0)| = | f (1)| = |α| =⇒
αᾱ = 1.
(Refer Slide Time: 13:26)

73
4

Define M ᾱ (w) := ᾱw. Then |M ᾱ (z)−M ᾱ (w)| = |ᾱ(z −w)| = |ᾱ||z −w| = |z −w|. Notice
that M ᾱ (0) = 0. Hence M ᾱ is a linear isometry.
Now reader should verify that the composition of two linear isometry is again a linear
isometry.
Define T := M ᾱ f . Then T is a linear isometry. T (1) = M ᾱ ( f (1)) = ᾱ f (1) = ᾱα = 1.
(Refer Slide Time: 16:03)

Let T (i ) = a + i b, |T (i ) − T (0)| = |i | = 1 =⇒ |a + i b| = 1 =⇒ a 2 + b 2 = 1 −→ (∗)

|T (i ) − T (1)|2 = |i − 1|2 =⇒ |a + i b − 1|2 = 2 =⇒ (a − 1)2 + b 2 = 2 −→ (∗∗)

74
 5

Now by (∗) and (∗∗), (a − 1)2 − a 2 = 1 =⇒ −2a + 1 = 1 =⇒ a = 0 =⇒ b 2 = 1 =⇒ b = ±1.

Hence, either T (i ) = i or T (i ) = −i . Therefore either T (z) = z or T (z) = z̄ ∀z ∈ C.
Now, we have defined T = M ᾱ f , notice that M ᾱ M α (w) = ᾱ(αw) = |α|2 w = w = M α M ᾱ (w).
Thus either f (z) = αz ∀z ∈ C or f (z) = αz̄ ∀z ∈ C.
Let S 1 = {z ∈ C : |z| = 1}. Then α ∈ S 1 .
We have already noticed that, for each element in S 1 we can find an isometry corre-
sponding to that element. That is, if z 0 ∈ S 1 ,define f (w) = z 0 w, then f is an isometry.
Observe that the set S 1 can be interpreted geometrically as a unit circle, S 1 = {z ∈ C : z =
a + i b, a 2 + b 2 = 1}.
(Refer Slide Time: 21:45)

Polar representation of a complex number

³ ´
z z
Let z ∈ C be a non-zero complex number, then z = |z| |z|
. Let r = |z|, w = |z|
, then
r > 0 and w ∈ S 1 . Such a representation of z is unique and is called polar decomposition
of z. Since w ∈ S 1 , w = cos θ + i sin θ.
Let f be an isometry which 0. Then, let α ∈ S 1 be such that f (z) = αz and α = cos θ +
i sin θ.
(Refer Slide Time: 26:43)

75
6

f (z) = (cos θ + i sin θ)(r (cos φ + i sin φ))

= r (cos (θ + φ) + i sin (θ + φ)).

Then f turns out to be a rotation by θ.

(Refer Slide Time: 28:08)

Similarly we can see the case of f (z) = αz̄, then f turns out to be a reflection along a
line through origin, since on taking the conjugate we are reflecting z along the real line
and then we are rotating it.

76
 7

Polar decomposition is easy when multiplying complex numbers. Let z 1 = r 1 (cos θ1 +

i sin θ1 ) and z 2 = r 2 (cos θ2 + i sin θ2 ), then z 1 z 2 = r 1 r 2 (cos (θ1 + θ2 ) + i sin (θ1 + θ2 )). That
is, geometrically, multiplication of complex numbers is a kind of dilation and rotation.
(Refer Slide Time: 32:29)

If z = r (cos θ + i sin θ), then r is called magnitude of z, (cos θ + i sin θ) ∈ S 1 is called

phase of z.
If θ ∈ R such that for z ∈ C, z = r (cos θ + i sin θ), then θ is called an argument of z.
Notice that, if θ is an argument of z, the basic trigonometric rule tells that, θ +2kπ is also
an argument of z, where k ∈ Z.
(Refer Slide Time: 36:43)

77
8

We shall denote argument of z by arg(z). From the observation we made on argu-

ment, in the algebraic point of view, argument is a coset of R/2πZ.
If we pick the semi-open interval (−π.π] and demand that the argument of given
complex number z lies in this interval, then the argument is called the standard argu-
ment of z and is denoted by Arg(z).
(Refer Slide Time: 38:54)

78
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 2.3
Functions on the complex plane

In the last lecture we discussed isometrics on the complex plane; isometries were
distance preserving complex valued functions which were defined on the entire com-
plex plane. Throughout this course we will be interested in studying complex valued
functions which are defined on open connected subsets of the complex plane.
(Refer Slide Time: 00:35)

We shall refer to open connected subsets of C by the terms, domain or region.

(Refer Slide Time: 01:47)
1

79
2

Let Ω ⊆ C be a domain and let f : Ω −→ C be a complex valued function. Define

u(z) := Re( f (z)) and v(z) := Im( f (z)) for z ∈ Ω. Then f (z) = u(z) + i v(z).

E XAMPLE 1. Consider p : C −→ C given by p(z) = a 0 + a 1 z + · · · + a d z d . Since C can

be identified with R2 , we have z = x + i y and thus p(z) can be identified as p(x, y),then
p(z) = p(x, y) = p 1 (x, y) + i p 2 (x, y). But notice that, every polynomial in two variable
z+z
may not be written as a polynomial in z itself. If p(x, y) = x, then p(z) = 2
. Here p is a
polynomial which involves both z and z.

(Refer Slide Time: 05:29)

80
 3

E XAMPLE 2 (Rational functions). Let p(z) and q(z) be polynomials with no common
factors and let Z (q) = {z ∈ C : q(z) = 0}. Note that q(z) is a polynomial it has a degree,
say d , then the number of points where q vanishes (counting multiplicities) will be less
than or equal to d . Then Z (q) is finite and hence closed. Define Ω := C \ Z (q). Then,
since Z (q) is finite and closed, Ω is an open connected subset of C.
p(z)
Define R : Ω −→ C given by R(z) := q(z)
. Then such a function is called rational func-
tion.

(Refer Slide Time: 09:15)

81
4

D EFINITION 1 (Uniform convergence). Let Ω ⊆ C and f , f n :−→ C be a collection of

functions defined on Ω. We say that the sequence { f n }n∈N converges to f uniformly on
Ω if given ² > 0, ∃N ∈ N such that | f n (x) − f (x)| < ² ∀x ∈ Ω and n ≥ N .

E XERCISE 3. Let { f n }n∈N is a sequence of continuous functions on Ω ⊆ C which con-

verges uniformly to a function f on Ω. Then f is a continuous function on Ω.

∞
P
D EFINITION 2 (Absolute Convergence). We say that a series z n converges abso-
n=1
∞
P
lutely if |z n | converges.
n=1

(Refer Slide Time: 15.23)

82
 5

Complex exponential
We are familiar with the real exponential function exp : R −→ R whose Taylor series
is given by,
∞
P xn
exp(x) = n! .
n=0

exp(x) is absolutely convergent for x ∈ R.

(Refer Slide Time: 17:10)

83
6

Let us define the define the complex exponential exp : C −→ C given by

∞
P zn
exp(z) := n!
.
n=0

For every z ∈ C, exp(z) converges absolutely. Let us denote the function exp(z) by e z .

E XERCISE 4. Prove that, for every z, w ∈ C, e z+w = e z · e w .

(Refer Slide Time: 19:08)

Trigonometric functions We know that in the real variable cases, sine and cosine
functions have a Taylor series given by

X∞ (−1)n x 2n+1
sin x =
n=0 (2n + 1)!

x3 x5
=x− + −...
3! 5!
X∞ (−1)n x 2n
cos x =
n=0 (2n)!
x2 x4
= 1− + −...
2! 4!

(Refer Slide Time: 21:43)

84
 7

∞
(i x)n
For x ∈ R, e i x = = cos x +i sin x. Hence, for θ ∈ R, e i θ are points the unit circle.
P
n!
n=0

E XERCISE 5.

(1) Find the values of

• e 2πi
• e πi
π
• ei 2
(2) Prove that e i θ · e i φ = e i (θ+φ) and e i θ = e −i θ .

ix
+e −i x ix
−e −i x
Notice that cos x = e 2
and sin x = e 2i
.
(Refer Slide Time: 25:08)

85
8

Let us define the complex trigonometric functions sin : C −→ C and cos : C −→ C

given by,

e i z + e −i z
cos z =
2
e i z − e −i z
sin z =
2i

E XERCISE 6. Prove that cos2 z + sin2 z = 1.

(Refer Slide Time : 27.06)

86
 9

Hyperbolic trigonometric functions

Define hyperbolic trigonometric functions cosh : C −→ C and sinh : C −→ C given by,
e z + e −z
cosh z =
2
e z − e −z
sinh z =
2
Now, observe that;
e z + e −z
cos(i z) = = cosh(z)
2
e −z − e z
sin(i z) = = i sinh(z) =⇒ sinh(z) = −i sin(i z)
2i

E XERCISE 7. Prove that cosh2 z − sinh2 z = 1

87
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 2.4
Complex differentiability

At the very heart of the study of Real analysis, lies the notion of differentiability and
integration and these two notions are tied together by the very beautiful fundamental
theorem of calculus. In Complex analysis also, the things are quite similar. At the very
heart of the study of Complex analysis, the notion of complex differentiability and com-
plex line integrals lie and they are also tied together by a variant of the fundamental
theorem of calculus.
There is a further variant of this fundamental theorem called the Cauchy’s theo-
rem, which wends certain amount of rigidity to complex differentiable functions thereby
making the theory very beautiful. In this lecture we will define complex differentiability
and explore some of its properties.
(Refer Slide Time: 01:02)

88
2

Recall that a function f : U ⊆ R −→ R is said to be differetiable at a point x 0 ∈ U if x 0

f (x)− f (x 0 )
is an interior point of U and lim x−x 0
exists. The limit is denoted by f 0 (x 0 ) which
x→x 0
x∈U \{x 0 }
is called the derivative of f at x 0 .
Complex differentiability
Let Ω ⊆ C and f : Ω −→ C. We say that f is complex differentiable at a point z 0 ∈
f (z)− f (z 0 )
Ω if z 0 is an interior point and lim z−z 0
exists. The limit is deoted by f 0 (z 0 ) or
z→z 0
z∈Ω\{z 0 }
df
(z ).
dz 0
If f is complex differentiable at every point z ∈ Ω, then we say that f is complex
differentiable on Ω or holomorphic on Ω.
(Refer Slide Time: 06:30)

There also many other variants for the definition of complex differentiability. An-
other important variant is ² − δ definition for complex differentiability:
We say that f is complex differentaible at z 0 , if ∃ f 0 (z 0 ) ∈ C such that, given ² > 0, ∃δ > 0
¯ ¯
¯ f (z)− f (z
such that D(z 0 , δ) ⊆ Ω and ¯ z−z0 0 − f 0 (z 0 )¯ < ² whenever 0 < |z − z 0 < δ.
¯

(Refer Slide Time: 07:55)

89
 3

Another definition is in terms of linear approximation,

We say that f is complex differentiable at z 0 if ∃ f 0 (z 0 ) ∈ C such that

f (z) = f (z 0 ) + f 0 (z 0 )(z − z 0 ) + o(z − z 0 ), where o(z − z 0 ) = (z − z 0 )e(z), e(z) → 0 as z → z 0 .

L EMMA 1. Let Ω ⊆ C. If f : Ω −→ C is complex differentiable at z 0 , then f is continuous

at z 0 .

The proof the lemma is immediate from the definition of complex differentiability.

E XERCISE 2. Let f (z) = z n be defined on C. Fix z 0 ∈ C. Then for z 6= z 0 ,

f (z) − f (z 0 ) z n − z 0n
=
z − z0 z − z0
= z n−1 + z n−2 z 0 + · · · + z 0n−1
f (z) − f (z 0 )
lim = z 0n−1 + z 0n−1 · · · + z 0n−1 (n − terms)
z→z 0 z − z0
= nz 0n−1

(Refer Slide Time: 13:22)

90
4

Laws of Calculus are satisfied:

• Linearity
If f , g : Ω ⊆ C −→ C are complex differentiable at z 0 ∈ Ω, then so is f + g with
complex derivative f 0 (z 0 ) + g 0 (z 0 ). If c ∈ C, then c f is complex differentiable at
z 0 and (c f )0 (z 0 ) = c f 0 (z 0 ).
• Product rule
If f , g : Ω ⊆ C −→ C be complex differentiable at z 0 ∈ Ω. Then ( f g ) is complex
differentiable at z 0 with derivative f 0 (z 0 )g (z 0 ) + f (z 0 )g 0 (z 0 ).
• Quotient rule
If f , g : Ω ⊆ C −→ C be complex differentiable at z 0 ∈ Ω and suppose that g does
not vanish at z 0 . Then by the continuity, g does not vanish in a neighborhood
f
of D of z 0 . Then is complex differentiable at z 0 .
g

µ ¶0
f f 0 (z 0 )g (z 0 ) − g 0 (z 0 ) f (z 0 )
(z 0 ) = .
g (g (z 0 ))2

(Refer Slide Time: 20:05)

91
 5

• Chain rule Let f : Ω −→ C be complex differentiable at z 0 ∈ Ω and suppose g :

D −→ C such that g is complex differentiable at f (z 0 ) and f (Ω) ⊆ D, then g ◦ f
is complex differentiable at z 0 and (g ◦ f )0 (z 0 ) = g 0 ( f (z 0 )) f 0 (z 0 ).
(Refer Slide Time: 22:06)

P ROOF. Since f is complex differentiable at z 0 , ∃ f 0 (z 0 ) and e 1 (z) such that

f (z) = f (z 0 ) + (z − z 0 ) f 0 (z 0 ) + (z − z 0 )e 1 (z), where e 1 (z) → 0 as z → z 0 .

92
6

Let w 0 = f (z 0 ). Since g is complex differentiable at w 0 , ∃ g 0 (w 0 ) and e 2 (w)

such that, g (w) = g (w 0 ) + (w − w 0 )g 0 (w 0 ) + (w − w 0 )e 2 (w), where e 2 (w) → 0 as
w → w0.
Since f is complex differentiable at z 0 , it is continuous at z 0 . Hence f (z) → w 0
whenever z → z 0 .

g ( f (z)) = g ( f (z 0 )) + f (z) − f (z 0 ) g 0 ( f (z 0 )) + ( f (z) − f (z 0 ))e 2 ( f (z))

¡ ¢

= f (z) − f (z 0 ) g 0 ( f (z 0 )) + e 2 ( f (z))
¡ ¢¡ ¢

= (z − z 0 ) f 0 (z 0 ) + e 1 (z) g 0 ( f (z 0 )) + e 2 ( f (z))
¡ ¢¡ ¢

g ( f (z)) − g ( f (z 0 ))
= lim f 0 (z 0 ) + e 1 (z) g 0 ( f (z 0 )) + e 2 ( f (z))
¡ ¢¡ ¢
lim
z→z 0 z − z0 z→z 0
z∈Ω\{z 0 } z∈Ω\{z 0 }

= f 0 (z 0 )g 0 ( f (z 0 ))

(Refer Slide Time: 27:06)

E XERCISE 3. Prove the linearity, product rule and quotient rule for complex deriva-
tive.

93
 7

Functions which are complex differentiable on C are called entire functions. For
example, z → z n is an entire function.
By using the laws of Calculus,we have, p(z) = a 0 + a 1 z + · · · + a d z d is an entire function.
Moreover, we know that, for z 0 ∈ C, p 0 (z 0 ) = a 1 + 2a 2 z 0 + · · · + d a d z 0d −1 .
(Refer Slide Time: 30:51)

p(z)
Let R(z) = on Ω such that q(z) 6= 0∀z ∈ Ω. By the quotient rule R(z) is holomor-
q(z)
phic on Ω.

f (z) − f (z 0 ) z − z0
• Consider f (z) = z. Fix z 0 ∈ C, then = . Let z = z 0 + h, z =
z − z0 z − z0
f (z) − f (z 0 ) h
z 0 + h, where h 6= 0. lim = = lim . Suppose h → 0 along the real
z→z 0 z − z0 h→0 h
h h
axis, then lim = 1. Suppose h → 0 along the imaginary axis, then lim = −1.
h→0 h h→ h
Hence the limit does not exists =⇒ f (z) is not holomorphic.
(Refer Slide Time: 35:27)

94
8

• Let f (z) = |z|2 = zz. Fix z 0 ∈ C and z = (z 0 + h).Then,

f (z 0 + h) = (z 0 + h)(z 0 + h) = |z 0 |2 + |h|2 + hz 0 + hz 0 =⇒

f (z 0 + h) − f (z 0 ) hh + hz 0 + hz 0
=
h h
h
= h + z0 + z0
h
Now on taking the limit h → 0 along real axis will obtain z 0 + z 0 and along the
imaginary axis will obtain z 0 − z 0 . Since the limits are not equal, the limit does
not exists. Hence f is not holomorphic.

95
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 3.1
Power Series

In the last week, we defined the notion of complex differentiability. We saw that
complex differentiable functions also satisfy the laws of calculus, namely linearity, the
product rule, quotient rule and also the chain rule. Thereafter, we saw a few examples
of complex differentiable functions and in particular we noted that polynomials in the
variable z are complex differentiable in the entire complex plane; they are entire func-
tions.
This week we begin by discussing the notion of power series. Power series is an in-
finite degree variant of a polynomial inside its disk of convergence. A power series be-
haves very similar to polynomials both analytically and algebraically. Let us start this
lecture by defining a power series around a point z 0 in the complex plane.
(Refer Slide Time: 01:11)

96
2

Let z 0 ∈ C. A formal power series around z 0 with complex coefficients is a formal

expansion,
∞
a n (z − z 0 )n
X
n=0
where a n ∈ C and z is an indeterminate.
We could ask whether a formal power series converges at a given point z ∈ C.
∞
a n (z − z 0 )n converges absolutely.
P
For example, at z 0 , the formal power series
n=0
∞
z n , then for any z with |z| ≥ 1,
P
Another example is,consider the geometric series,
n=0
∞
n
z n does not converge.
P
the summands, |z | > 1 and does not converge to 0. Hence
n=0

(Refer Slide Time: 05:18)

Radius of Convergence
∞
a n (z − z 0 )n be a formal power series around z 0 . We define the radius of con-
P
Let
n=0
vergence R of the formal power series to be the number in [0, ∞] given by

R := lim inf |a n |−1/n

n→0

97
 3

The radius of convergence, as the name suggests, is a quantity which tells us some-
thing about the convergence of the given formal power series which is captured in the
next proposition.
(Refer Slide Time: 07:41)

∞
a n (z − z 0 )n be a power series around z 0 . Let R be the radius of
P
P ROPOSITION 1. Let
n=0
∞
a n (z − z 0 )n converges
P
convergence of the power series. Then for z ∈ D(z 0 , R), the series
n=0
absolutely. For r < R, the series converges uniformly on D(z 0 , r ). Furthermore, if |z − z 0 | >
∞
a n (z − z 0 )n diverges.
P
R, then
n=0

(The disc D(z 0 , R) is called the disc of convergence of the formal power series).

P ROOF. Let z ∈ C such that |z − z 0 | > R. Then ∃ infinitely many n ∈ N such that

|a n |−1/n < |z − z 0 | =⇒ |a n |−1/n |z − z 0 | > 1 =⇒ |a n (z − z 0 )n | > 1

for infinitely many n ∈ N → (∗). Since the summands does not converge to 0(by (∗)), we
∞
a n (z − z 0 )n does not converge.
P
can conclude that
n=0
Let z ∈ D(z 0 , R) =⇒ |z − z 0 | < r < R for some r > 0. Since R = lim inf |a n |−1/n , ∃ N ∈ N
n→∞
−1/n n
such that ∀n > N , |a n | > r =⇒ |a n |r < 1∀n > N .

98
4

(Refer Slide Time: 17:19)

For n > N ,

|z − z 0 |n
|a n (z − z 0 )n | = |a n r n |
rn
¶n
|z − z 0 |
µ
≤
r
|z − z 0 | n
µ ¶
n
X X
a n (z − z 0 ) ≤ .
n>N n>N r

∞ ∞
|a n (z − z 0 )n | converges. That is a n (z − z 0 )n converges absolutely.
P P
Hence
n=0 n=0
Let r < R and R 1 be such that r < R 1 < R. ∃N ∈ N such that ∀n > N , |a n |−1/n > R 1 > r . For
n > 1 and z ∈ D(z 0 , r ),

n ¶n
|z − z 0 | r
µ
n
|a n (z − z 0 ) | = |a n R 1n | ≤
R 1n R1

∞ r n X r n
µ ¶ µ ¶
is a convergent series, given ε > 0, ∃ N0 > N such that < ε.
X
. Since
n=0 R 1 n≥N0 R 1
(Refer Slide Time: 23:24)

99
 5

Then,

∞ N0
a n (z − z 0 )n − a n (z − z 0 )n | = | a n (z − z 0 )n |
X X X
|
n=0 n=0 n>N0
¶n
r
µ
X
<
n≥N0 R1

< ε.

(Refer Slide Time: 24:26)

100
6

∞
a n (z − z 0 )n be a formal power series with radius of conver-
P
P ROPOSITION 2. Let
n=0
gence R. Assume that a n is non-zero for n sufficiently large. Then

|a n | |a n |
lim inf ≤ R ≤ lim sup .
n→∞ |a n+1 | n→∞ |a n+1 |

|a n |
P ROOF. Let R 1 = lim inf . Let r < R 1 . Then ∃N ∈ N such that ∀n > N ,
n→∞ |a n+1 |

|a n |
> r =⇒ |a n+1 |r < |a n |.
|a n+1 |

(Refer Slide Time: 27:45)

101
 7

For z ∈ D(z 0 , r ) and n > N ,

|z − z 0 | n+1
µ ¶
n+1
|a n+1 (z − z 0 ) | = |a n+1 |
r
¶n+1
n |z − z 0 |
µ
< |a n r |
r
..
.
¶n+1
N ¯ |z − z 0 |
µ
¯ ¯
< aN r
¯ .
r
¯ X |z − z 0 | n
µ ¶
n N¯
P ¯
Hence |a n (z − z 0 ) | ≤ a N r
¯ .
n≥N n≥N r
∞
a n (z − z 0 )n converges =⇒ D(z 0 , r ) ⊆ {z : |z − z 0 | ≤ R} =⇒ r ≤ R. Since this is
P
Hence
n=0
true for all r < R 1 , we have R 1 ≤ R → (∗).

|a n |
Let R 2 = lim sup . Let r > R 2 and let z ∈ C be such that |z − z 0 | > r . Since r >
n→∞ |a n+1 |
R 2 , ∃N ∈ N such that ∀n > N ,
|a n |
< r =⇒ |a n+1 r | > |a n |.
|a n+1 |

102
8

(Refer Slide Time: 35:38)

Then for n > N ,

|a n+1 (z − z 0 )n+1 | > |a n+1 r n+1 | > |a n r n | · · · > |a N r N | = M (say)

∞
a n (z − z 0 )n does not converge.
P
Since the summands do not converge to 0,
n=0
That is, {z : |z − z 0 | > r } ⊆ {z : |z − z 0 | ≥ R} =⇒ D(z 0 , R) ⊆ {z : |z − z 0 | ≤ r } =⇒ R ≤ r .
Since this is true for all r > R 2 =⇒ R ≤ R 2 → (∗∗).
|a n | |a n |
By (∗) and (∗∗), lim inf ≤ R ≤ lim sup . 
n→∞ |a n+1 | n→∞ |a n+1 |

E XAMPLE 3.
∞ zn
• ez =
X
n=0 n!
1 |a n | (n + 1)!
Here a n = . Now, = = (n + 1). Hence the radius of conver-
n! |a n+1 | n!
gence is infinite.
(Refer Slide Time: 38:12)

103
 9

∞
n!z n .
X
•
n=0
|a n | (n! 1
Then = = → 0 as n → ∞. Hence the series converges
|a n+1 | (n + 1)! (n + 1)
only at the origin.
∞
z n has radius of convergence 1. For |z| = 1, the series diverges. For z ∈
P
•
n=0
D(0, 1), we have

∞ ∞
n
z n+1
X X
z z =
n=0 n=0
∞
zn
X
=
n=1
∞
z n − 1.
X
=
n=0

∞ 1
zn =
P
On D(0, 1), we have .
n=0 1−z
(Refer Slide Time: 41:10)

104
10

1 1
On C \ {1}, the function is holomorphic. Now note that is a func-
1−z 1−z
∞
zn.
X
tion which extends beyond D(0, 1) which is the disc of convergence of
n=0
X∞ zn
• 2
has a radius of convergence 1.
n=0 n
∞ zn
For z ∈ C such that |z| = 1,
X
2
converges.
n=0 n

∞
a n (z − z 0 )n be a power series with a positive radius
P
Abel’s Theorem: Let F (z) =
n=0
of convergence R and suppose z 1 = z 0 + Re i θ be a point such that F (z 1 ) converges. Then
lim F (z 0 + r e i θ ) = F (z 1 ).
r →R −
(Refer Slide Time: 45:18)

105
 11

P ROOF. We may assume that F (z 1 ) = 0.

∞
b n (z − z 0 )n where b n = a n ∀n > 0, b 0 = a 0 − F (z 1 ). If
P
Define G(z) = F (z) − F (z 1 ) =
n=0
lim− G(z 0 + r e i θ ) = 0, then lim− (F (z 0 + r e i θ ) − F (z 1 )) = 0. Hence we shall assume that
r →R r →R
F (z 1 ) = 0.
We can also assume z 0 = 0.
(Refer Slide Time: 48:34)

106
12

∞
a n z n . If G(r e i θ ) = 0 and lim− G(r e i θ ) = 0, then reader should verify
P
Define G(z) =
n=0 r →R
that lim− G(r e i θ ) = lim− F (z 0 + r e i θ ) = 0. Hence we shall assume that z 0 = 0.
r →R r →R
Thus if we prove for the power series around 0, we can translate it can conclude that it is
true for all power series around any arbitrary point.
(Refer Slide Time: 50:40)

We may also assume that R = 1 and θ = 0.

Let G(z) = a n R n e −i nθ z n = F (Re −i nθ z).
X

1
The radius of convergence of G = lim inf |a n |−1/n R −n/n = lim inf |a n |−1/n = 1. If G(1) = 0
n→∞ R n→∞
and lim− G(r ) = 0, then lim− F (r e i θ ) = 0.
r →1 r →R
Hence to prove the theorem, it is enough to prove the following:
Let a n be a series converging to 0. Then lim− a n r n = 0.
P P
r →1
(Refer Slide Time: 54:42)

107
 13

Let A n = a 0 + a 1 + · · · + a n . Given ε > 0, ∃ N such that ∀n > N , |A n | < 2ε .

m
a n r n = A 0 + (A 1 − A 0 )r + · · · + (A m − A m−1 )r m
X
n=0
m−1
A n (r n − r n+1 ) + A m r m
X
=
n=0
m−1
An r n + Am r m .
X
= (1 − r )
n=0

∞ m−1
a n r n = (1 − r ) A n r n → (∗).
P P
Hence
n=0 n=0
For r < 1 and given ε > 0 as above, R.H.S of (∗) becomes,
¯ ¯ ¯ ¯
¯ N ∞ ¯ ¯X N ¯ ∞
A n r n + (1 − r ) A n r n ¯ ≤ (1 − r ) ¯ A n r n ¯ + (1 − r ) |A n |r n
X X X
¯(1 − r )
¯ ¯ ¯ ¯
n=0 n=N n=0 n=N
¯ ¯ ¯ ¯
¯ ¯
¯X N ¯ ε rN
< (1 − r ) ¯ A n r n ¯ + (1 − r )
¯ ¯
¯n=0 ¯ 2 (1 − r )
¯ ¯
¯X N ¯ ε
< (1 − r ) ¯ An r n ¯ + .
¯ ¯
¯n=0 ¯ 2

(Refer Slide Time: 01:00:55)

108
14

N
A n r n is a continuous on [0, 1]. Hence ∃M such that
X
Then the function
¯ ¯ n=0
N
A n r n ¯¯ < M ∀ r ∈ [0, 1].
¯P ¯
¯
n=0
¯
ε ε ε
¯ ¯
∞
n¯
¯P ¯
Let r be such that (1 − r ) < . Then (1 − r ) ¯
¯ An r ¯ < M= .
2Mµ n=0
¶ 2M 2
∞
Hence lim− a n r n = lim− (1 − r ) A n r n = 0.
P P
r →1 r →1 n=0


109
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 3.2
Differentiation of Power Series

In the last lecture we defined what is meant by a power series and we proved that
a power series converges absolutely in its disk of convergence we also remarked that a
power series behaves very similar to polynomials in its disc of convergence. We begin
this lecture by proving that a power series is holomorphic in its disc of convergence and
that just like in the case of polynomials we can obtain the derivative by differentiating
the power series term by term.
(Refer Slide Time: 01:11)

a n (z − z 0 )n be a power series around z 0 with a radius of

P
T HEOREM 1. Let F (z) =
convergence R. Then F is holomorphic in D(z 0 , R) and the derivative F 0 (z) is given by the
power series,
∞
na n (z − z 0 )n−1
X
n=1
1

110
2

which has a radius of convergence R.

P ROOF. Consider the power series

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

Then R 0 = lim inf ((n + 1)|a n+1 |)−1/n .

n→∞
Now it is left to reader as an exercise to verify that lim (n + 1)1/n = 0.
n→∞
−1/n
Claim: Let R 0 = lim inf (|a n+1 |) . Then R 0 = R.
n→∞
For z ∈ D(z 0 , R 0 ),

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

∞ ∞
|a n (z − z 0 )n | ≤ |a 0 | + |(z − z 0 )| |a n ||(z − z 0 )|n−1 =⇒ a n (z − z 0 )n converges
P P P
Hence,
n=0 n=1
in D(z 0 , R 0 ) =⇒ R 0 ≤ R.
(Refer Slide Time: 07:56 )

111
 3

∞
a n (z − z 0 )n converges absolutely.
P
Let z ∈ D(z 0 , R) and z 6= z 0 . We know
n=0

¯∞ ¯ ∞
n−1 ¯
|a n (z − z 0 )n− |
¯X ¯ X
¯
¯ a n (z − z 0 ) ¯≤
n=1 n=1
1 ∞
|a n ||(z − z 0 )|n
X
=
|z − z 0 | n=1
1 ∞
|a n ||(z − z 0 )|n
X
≤
|z − z 0 | n=1

∞
a n (z − z 0 )n converges in D(z 0 , R) =⇒ R ≤ R 0 . Hence R = R 0 .
P
Hence
n=0
Exercise: Let a n and b n be positive real numbers such that a n → a and where a > 0
and lim inf b n = b. Then lim inf a n b n = ab.
n→∞ n→∞
By using this exercise, we can conclude that the radius of convergence of the power
∞
na n (z − z 0 )n−1 is R.
P
series
n=1

(Refer Slide Time: 15:25 )

112
4

Fix z 1 ∈ D(z 0 , R). Now, consider the Newton quotient,

∞ n n
F (z) − F (z 1 ) X (z − z 0 ) − (z 1 − z 0 )
= an
(z − z 1 ) n=0 (z − z 1 )
∞
a n (z − z 0 )n−1 + (z − z 0 )n−2 (z 1 − z 0 ) + · · · + (z 1 − z 0 )n−1 .
X ¡ ¢
=
n=0

We want to show that,

∞ ∞
a n (z − z 0 )n−1 + (z − z 0 )n−2 (z 1 − z 0 ) + · · · + (z 1 − z 0 )n−1 = na n (z 1 −z 0 )n−1 .
X ¡ ¢ X
lim
z→z 1
z∈D(z 0 ,R)\{z 1 } n=1 n=1

∞
a n (z − z 0 )n−1 + (z − z 0 )n−2 (z 1 − z 0 ) + · · · + (z 1 − z 0 )n−1 .
P ¡ ¢
Define G(z) =
n=1
Let R 1 , R 2 be such that |z 1 − z 0 | < R 1 < R 2 < R and ε > 0 be such that D(z 1 , ε) ⊂ D(z 0 , R 1 ).

(Refer Slide Time: 20:32 )

Let z ∈ D(z 1 , ε),

|a n (z − z 0 )n−1 + (z − z 0 )n−2 (z 1 − z 0 ) + · · · + (z 1 − z 0 )n−1 | ≤ |a n nR 1n−1 |

¡ ¢

113
 5

Since lim inf |a n |−1/n = R and R 2 < R, we have N ∈ N such that ∀n > N ,
n→∞
−1/n 1
|a n | > R 2 =⇒ |a n | < n
R2
µ ¶n
¡ n−1 n−2 n−1
¢ R1
=⇒ |a n (z − z 0 ) + (z − z 0 ) (z 1 − z 0 ) + · · · + (z 1 − z 0 ) |≤n .
R2
Hence by Weierstrass M-test, a n (z − z 0 )n−1 + (z − z 0 )n−2 (z 1 − z 0 ) + · · · + (z 1 − z 0 )n−1 con-
P ¡ ¢

verges uniformly in D(z, ε) to G(z). Since G(z) is continuous, we have

lim G(z) = G(z 1 )

z→z 1

.
(Refer Slide Time: 25:59 )

That is,
F (z) − F (z 1 ) ∞
na n (z 1 − z 0 )n−1
X
lim = G(z 1 ) =
z→z 1 (z − z 1 )
z∈D(z 0 ,R)\{z 1 } n=1

a n (z − z 0 )n is holomorphic on D(z 0 , R).

P
. Thus F (z) =


a n (z − z 0 )n on D(z 0 , R), where R is the radius of conver-

P
C OROLLARY 2. Let F (z) =
F n (z 0 )
gence of the power series. Then a n = .
n!

114
6

∞
P ROOF. The reader should verify that F m (z) = n(n−1) . . . (n−m+1)a n (z −z 0 )n−m .
P
n=m
m
F (z 0 )
Now, F (z 0 ) = m(m − 1) . . . 2 · 1a m = m!a m =⇒ a m = 
m!

(Refer Slide Time: 32:24 )

C OROLLARY 3. If G(z) is a power series b n (z − z 0 )n such that G(z) = F (z) in a neigh-

P

borhood of z 0 , where F (z) = a n (z − z 0 )n . Then a n = b n .

P

P ROOF. Since F (z) = G(z) in D(z 0 , ε) for some ε > 0, we have F n (z) = G n (z)∀n ∈
N =⇒ a n = b n 

Now lets calculate the derivative of one of the important power series we discussed ear-
lier, the exponential function.
∞ zn d z X ∞ z n−1 ∞ zn
Consider e z = = ez.
X X
. Then, e = =
n=0 n! d z n=1 (n − 1)! n=0 n!
e i z − e −i z d e i z + e −i z
Now consider, sin(z) = . Then, sin(z) = = cos(z).
2i dz 2
d
Similarly, cos(z) = − sin(z).
dz
(Refer Slide Time: 36:39 )

115
 7

∞ ∞
a n (z − z 0 )n and G(z) = b n (z − z 0 )n be power series
P P
P ROPOSITION 4. Let F (z) =
n=0 n=0
∞
c n (z − z 0 )n where
P
which converges in D(z 0 , R). Then F (z)G(z) is given by power series
n=0
n
P
cn = a k b n−k which also has a radius of convergence at least R.
k=0

∞ ∞
a n (z−z 0 )n and b n (z−z 0 )n converges abso-
P P
P ROOF. Notice that for z ∈ D(z 0 , R),
n=0 n=0
n n n
a k (z−z 0 )k , B n = b k (z−z 0 )k C n = c k (z−z 0 )k and D n = B n −G(z).
P P P
lutely. Let A n =
k=0 k=0 k=0
(Refer Slide Time: 41:39)

116
8

Then

C n = a 0 b 0 + (a 1 b 0 + a 0 b 1 )(z − z 0 ) + · · · + (a n b 0 + a n−1 b 1 + · · · + a 0 b n )(z − z 0 )n

= B n a 0 + B n−1 a 1 (z − z 0 ) + B n−2 a 2 (z − z 0 )2 + · · · + B 0 a n (z − z 0 )n .

= G(z) (a 0 + a 1 (z − z 0 ) + · · · + a n (z − z 0 ))n + (D n a 0 + D n−1 a 1 (z − z 0 ) + · · · + D 0 a n (z − z 0 )n )

C n = G(z)A n + (D n a 0 + D n−1 a 1 (z − z 0 ) + · · · + D 0 a n (z − z 0 )n )

(Refer Slide Time: 45:33 )

117
 9

Consider D n a 0 + D n−1 a 1 (z − z 0 ) + · · · + D 0 a n (z − z 0 )n . For a given ε > 0, ∃N such that

|D n | < ε ∀n ≥ N .

|D n a 0 + · · · + D 0 a n (z − z 0 )n | ≤ |D 0 a n (z − z 0 )n + · · · + D N a n−N (z − z 0 )n−N |

+ |D N +1 a n−N +1 (z − z 0 )n−N +1 + · · · + D n a 0 |

≤ |D 0 ||a n (z − z 0 )|n + · · · + |D N ||a n−N (z − z 0 )n−N |

∞
+ε |a n |(z − z 0 )|n
X
n=0

|D n a 0 + · · · + D 0 a n (z − z 0 )n | ≤ |D 0 ||a n (z − z 0 )|n + · · · + |D N ||a n−N ||z − z 0 |n−N + εα

lim |D n a 0 + · · · + D 0 a n (z − z 0 )n | ≤ εα
n→∞

lim sup |D n a 0 + · · · + D 0 a n (z − z 0 )n | = 0
n→∞

⇒ lim |D n a 0 + · · · + D 0 a n (z − z 0 )n | = 0
n→∞

⇒ lim C n = F (z)G(z).
n→∞

118
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 3.3
Problem Session

P ROBLEM 1. Let T : C → C be an isometry of the complex plane. Then show that T is

given by either

T (z) = z 0 + w z for z ∈C

or

T (z) = z 0 + w z̄ for z ∈C

where z 0 ∈ C is fixed and w ∈ S 1 .

(Refer Slide Time: 01:53)

S OLUTION 1. Let z 0 = T (0). Define T−z0 := z − z 0 . Now it is left to reader to verify that
T−z0 is an isometry. Then T−z0 T (0) = T−z0 (z 0 ) = z 0 − z 0 = 0. Hence T−z0 T is an isometry
fixing origin.
1

119
2

Then by the theorem proved earlier, we have either

T−z0 T (z) = w z for z ∈ C, (where w ∈ S 1 )

or

T−z0 T (z) = w z̄ for z ∈ C, (where w ∈ S 1 ).

Define T z0 (z) = z + z 0 . Then observe that T z0 is also an isometry. Moreover

T z0 T−z0 (z) = T z0 (z − z 0 ) = z = i d (z) =⇒ T z0 T−z0 = i d .
(Refer Slide Time: 05:53)

Hence, either

T z0 (T−z0 T )(z) = T z0 (w z) = z 0 + w z

or

T z0 (T−z0 T )(z) = T z0 (w z̄) = z 0 + w z̄.

We know that function composition is associative, hence T z0 (T−z0 T ) = (T z0 T−z0 )T =

idT = T .
Hence for z ∈ C either

T (z) = z 0 + w z

120
 3

or

T (z) = z 0 + w z̄.

P ROBLEM 2. Let n be a positive integer. Show that the complex number solutions to
2πi k
the equation z n = 1 are given by z k = e n for k = 0, 1, . . . , n −1. These roots are called the
n th roots of unity.

(Refer Slide Time: 10:10)

S OLUTION 2. Consider z n = 1. Then |z n | = 1 =⇒ |z|n = 1. Since |z| ≥ 0, we have

|z| = 1. Hence z ∈ S 1 =⇒ z = e 2πi θ for some θ ∈ [0, 1).

z n = 1 =⇒ e 2πi (nθ) = 1

=⇒ 2π(nθ) = 2πk, where k ∈ Z

k
=⇒ θ = , where k ∈ Z.
n
Notice that if `1 and `2 are such that
`1 `2 (`1 −`2 ) `1 − `2
e 2πi n = e 2πi n =⇒ e 2πi n = 1 =⇒ ∈ Z.
n

121
4

`1 `2
Thus if `1 = `2 + kn, then e 2πi n = e 2πi n .
2πi k
Hence every complex number z such that z n = 1 is given by e n where k = 0, 1, . . . n −1.
(Any other integer ` can be written as ` = d n + k.)
(Refer Slide Time: 16:44)

2πi k 1 2πi k 2
Now it is left for reader to check that e n 6= e n for 0 < k 1 < k 2 < n.
2πi k
Hence we have concluded that z n = 1 has n distinct roots given by e n where k =
0, 1, . . . , n − 1 and are called the n th roots of unity.

P ROBLEM 3. Show that if w is a non-zero complex number and n a positive integer,

then there exists n distinct roots of the equation z n = w. Any two roots differ by multi-
plication by a root of unity.

S OLUTION 3. Consider z n = w. In polar coordinates, let w = r e i 2πθ . Then |z|n = |w| =

r =⇒ |z| = r 1/n . If z = r 1/n e i 2πϕ

zn = w
³ ´n
1/n i 2πϕ
=⇒ r e = r e i 2πθ

=⇒ e i 2π(nϕ−θ ) = 1

122
 5

=⇒ nϕ − θ = k ∈ Z

k +θ
=⇒ ϕ = where k ∈ Z.
n
(Refer Slide Time: 22:10)

i 2π(θ+k) i 2πθ i 2πk

Hence z = r 1/n e n where k ∈ Z =⇒ z = r 1/n e n ·e n where k = 0, 1, . . . n − 1.
Check that product of two n th roots of unity is again an n th root of unity and the inverse
of an n th root of unity is again an n th root of unity.

P ROBLEM 4. Let w 1 , w 2 be distinct complex numbers, and let λ be a positive real

number.
|z − w 1 |
½ ¾
(i) Show that the set z : = λ gives a circle when λ 6= 1 and a straight line
|z − w 2 |
when λ = 1.
(ii) Conversely, show that every circle and straight line can be written in this manner.

|z − w 1 |
S OLUTION 4. Consider = λ. Then,
|z − w 2 |

|z − w 1 |2 = λ2 |z − w 2 |2

|z|2 − z w¯1 − z̄w 1 + |w 1 |2 = λ2 (|z|2 − z w¯2 − z̄w 2 + |w 2 |2 )

123
6

If λ = 1,

z(w¯2 − w¯1 ) + z̄(w¯2 − w¯1 ) = |w 2 |2 − |w 1 |2

ᾱz + αz̄ = c, where c ∈ R

which is a line.
When λ 6= 1,

(1 − λ2 )|z|2 − z(w¯1 − λ2 w¯2 ) − z̄(w 1 − λ2 w 2 ) = λ2 |w 2 |2 − |w 1 |2

λ2 |w 2 |2 − |w 1 |2
|z|2 − ᾱz − αz̄ + |α|2 = + |α|2
(1 − λ2 )

|z − α|2 = r 2 , where r ∈ R

which is a circle.
(Refer Slide Time: 31:59)

Converse part is left as an exercise to reader.

Definition: Let a 1 , a 2 , . . . , a n ∈ C. The convex hull of a 1 , a 2 , . . . , a n is given by the set

{ λi a i : λi ∈ R, λi ≥ 0 and λi = 1} .
P P

124
 7

P ROBLEM 5. Let P (z) be a polynomial given by P (z) = c(z−z 1 )(z−z 2 ) · · · (z−z n ) where
z 1 , z 2 , . . . z n ∈ C(not necessarily distinct). Then the roots of P 0 (z) lies in the convex hull of
z1 , z2 , . . . zn .

Pn
S OLUTION 5. By product rule, P 0 (z) = j =1 c(z − z 1 ) · · · (z − z j ) · · · (z − z n ), here the hat
à

indicates absence of the particular term.

(Refer Slide Time: 36:59)

At the points away from the roots of P , we have,

n n
P 0 (z) c(z − z 1 ) · · · (z − z j ) · · · (z − z n ) X 1
X à
= = . → (∗)
P (z) c(z − z 1 )(z − z 2 ) · · · (z − z n ) j =1
(z − z j )
j =1

Let w be a root of P 0 (z) such that P (w) 6= 0. Then by (∗), we have

n
X 1
=0
j =1
w − zj

n
X w̄ − z¯j
=⇒ =0
j =1
|w − z j |2
X 1 X z¯j
=⇒ w̄ =
|w − z j |2 |w − z j |2

125
8

X 1
|w−z j |2
w= ³P ´zj.
1
|w−z j |2
1
|w−z j |2
´ = λ j , then λ j = 1 and hence w = λj z j .
P P
Put ³P
1
|w−z j |2

126
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 3.4
Cauchy Riemann Equations

Up till now, we have seen that complex differentiability is not very different from the
notion of differentiability which was developed in Real Analysis course. The definitions
were similar and some of the properties of like the laws of Calculus pertains to this lec-
ture are satisfied by the differentiable functions and also the properties which we explore
in complex differentiability. In this lecture we will see how complex differentiable func-
tion different from real differential function. We will prove that Complex differentiable
functions are real differentiable functions which have some rigidity conditions.

(Refer Slide Time: 00:44)

Recall that a function f : U ⊆ R2 → R2 is said to be differentiable at z 0 ∈ U if ∃ a R

linear map d f (z 0 ) such that f (z 0 + h) − f (z 0 ) = d f (z 0 )h + o(h) , h ∈ R2 .
1

127
2

The function is said to be complex differentiable at z 0 ∈ U if ∃ a complex number

f 0 (z 0 ) such that f (z 0 + h) − f (z 0 ) = f 0 (z 0 )h + o(h).
(Refer Slide Time: 06:14)

L EMMA 1. A complex differentiable function is differentiable.

P ROOF. Let f be complex differentiable. If f 0 (z 0 ) = a + i b and h = x + i y, then

   
x ax − b y
d f (z 0 )   := (a + i b)(x + i y) =  .
y a y + bx

Now, with d f (z 0 )h := f 0 (z 0 )h, the function is differentiable. 

Notice that a function T : C → C is C linear then T (z) = T (z · 1) = zT (1) = αz, where

α = T (1).
If d f (z 0 ) is a C− linear map, then ∃ a complex number f 0 (z 0 ) such that

d f (z 0 )h = f 0 (z 0 )h

where h ∈ C.

T HEOREM 2. Let f : U ⊆ C → C be a differentiable function. Then f is complex differ-

entiable if and only if d f (z 0 ) is C linear.

128
 3

(Refer Slide Time: 14:21)

Let T : R2 → R2 be R linear. Suppose T is also C linear as a function

  from C → C.
a b x
We have a, b, c, d ∈ R such that T (x, y) = (ax + b y, cx + d y) =    .
c d y
If T is C− linear, then

T (i (x, y)) = i T (x, y) ∀(x, y) ∈ R2 .

Note that (x, y) can be identified as x +i y, with this identification, i (x, y) corresponds to
i (x + i y) = −y + i x. Hence, T (i (x, y)) = T (−y, x) = i T (x, y).
If (x, y) = (1, 0); T ((0, 1)) = i T ((1, 0)).
Thus (b, d ) = i (a, c) = (−c, a) =⇒ b = −c, d = a → (∗).
Thus, from here we can observe that any linear map T : R2 → R2 given by

T (x, y) = (ax + b y, −bx + a y)

is C− linear.
Let f : U ⊆ R2 → R2 be a differentiable function and f = u + i v. Then
 
u x (z 0 ) u y (z 0 )
d f (z 0 ) =  
v x (z 0 ) v y (z 0 )

129
4

∂u
where u x (z 0 ) = (z 0 ) and other quantities defined similarly.
∂x
(Refer Slide Time: 22:51)

T HEOREM 3. Let f : U ⊆ C → C be differentiable at z 0 ∈ U . Then f is complex differen-

tiable if and only if, for f = u + i v, we have


u x (z 0 ) = v y (z 0 ) 

→ (∗)
u y (z 0 ) = −v x (z 0 )


The equations (∗) are called the Cauchy- Riemann equations.

C OROLLARY 4. Let f : U ⊆ C → C be a differentiable function on U . Then f is holo-

morphic on U if and only if f satisfies the Cauchy-Riemann equations at every point of
U.

(Refer Slide Time: 27:57)

130
 5

f (z) − f (z 0 )
Recall that f is C− differentiable at z 0 if lim exists.
z→z 0 z − z0
z∈U \{z 0 }
Consider the limit along the direction parallel to x− axis,

∂f f (z 0 + x) − f (z 0 )
(z 0 ) = lim = ux + i v x
∂x x→0 x
x6=0

Consider the along the imaginary axis,

f (z 0 + i y) − f (z 0 ) 1 f (z 0 + i y) − f (z 0 )
lim = lim
i y→0 iy i y→0 y
y6=0 y6=0

1 ∂f 1
= (z 0 ) = (u y + i v y )
i ∂y i

(Refer Slide Time: 33:01)

131
6

Hence f is C differentiable =⇒

∂f 1 ∂f
f 0 (z 0 ) = (z 0 ) = (z 0 ).
∂x i ∂y

Suppose U ∩ R 6= ∅ and g = f U ∩R . i.e. g : U ∩ R → C. The above derivation tells us that

dg df
(z 0 ) = (z 0 ).
dx dz

Suppose f : U → C be a differentiable function at z 0 ∈ U . Define the Wirtinger derivatives

∂f ∂f
(z 0 ) and (z 0 ) by the following formulae:
∂z ∂z̄

∂f 1 ∂f 1 ∂f
µ ¶
(z 0 ) = (z 0 ) + (z 0 )
∂z 2 ∂x i ∂y

∂f 1 ∂f 1 ∂f
µ ¶
(z 0 ) = (z 0 ) − (z 0 ) .
∂z̄ 2 ∂x i ∂y

E XERCISE 5. The function f satisfies the Cauchy-Riemann equations at z 0 ∈ U if and

∂f
only if (z 0 ) = 0.
∂z̄

(Refer Slide Time: 39:06)

132
 7

E XAMPLE 6.

• Consider the function f : C → C, f (z) = z̄. Then f (x, y) = x − i y =⇒ u(x, y) =

x, v(x, y) = −y. Here u x = 1 and v y = −1. Hence f does not satisfy the Cauchy-
Riemann equations.
• f (z) = Re(z). Here u(x, y) = x and v(x, y) = 0.
• f (z) = |z|2 = x 2 + y 2 ; u(x, y) = x 2 + y 2 , v(x, y) = 0.

133
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Module No - 3
Lecture No – 13
Harmonic Functions

In the last week we explore the Cauchy–Riemann equations where the right set of con-
ditions which ensured that a differentiable function also turned out to be complex dif-
ferentiable. Given a holomorphic function, the Cauchy–Riemann equations tied the real
part of the holomorphic function with imaginary part of the holomorphic function. We
begin this week by observing that the Cauchy–Riemann equations imparts an extra set
of condition on these functions, real part and the imaginary part of a given holomorphic
function which is captured by the more common harmonic functions.
(Refer Slide Time: 00:59)

In this lecture we will always assume that our holomorphic functions are twice con-
tinuously differentiable. So I would like to note that this is a redundant condition. It is
1

134
2

redundant because holomorphic functions will always satisfy this criteria. We will prove
that later. But since we have not proved it yet let me just put this extra condition.
Let f : U → C be a holomorphic function. The Cauchy- Riemann equation tells us
that if f = u + i v, then
∂u ∂v

(z) = (z)


∂x ∂y


→ (∗)
∂u ∂v 
(z) = − (z)

∂y ∂x


From (∗), we get

∂2 u ∂2 v
(z) = (z)
∂x 2 ∂x∂y

∂2 u ∂2 v
(z) = − (z).
∂y 2 ∂y∂x

(Refer Slide Time: 04:20)

By Clairaut’s theorem,
∂2 v ∂2 v
=
∂x∂y ∂y∂x

135
 3

and hence we get,

∂2 u ∂2 u
+ = 0.
∂x 2 ∂y 2

Similarly,

∂2 v ∂2 v
+ = 0.
∂x 2 ∂y 2

∂2 ∂2
Let ∆ := + . Then ∆ is called the Laplacian operator.
∂x 2 ∂y 2

∆u = ∆v = 0 =⇒ ∆ f = 0.

D EFINITION 1. Let u : Ω → R be a twice continuously differentiable function. We say

u is harmonic if ∆u = 0.

E XAMPLE 1.

• Given any holomorphic function f . Consider Re( f ) and Im( f ), both will be
harmonic functions.
• Consider f (z) = z̄. Then f is a harmonic function which is not holomorphic.
• Let u(z) = x 2 − y 2 , which then will be harmonic.

Recall that, given Ω ⊂ X , where X is a metric space, the boundary of Ω, ∂Ω is given

by ∂Ω = Ω ∩ X \ Ω.

E XAMPLE 2. ∂B (x 0 , r ) = {x ∈ X : d (x, x 0 ) = r }.

(Refer Slide Time: 12:56)

136
4

T HEOREM 3 (Maximum Principle for Harmonic Functions). Let Ω be an open con-

nected subset of C and u : Ω → R be a twice differentiable harmonic function. Let K ⊂ Ω
be a compact subset of Ω. Then,

sup u(z) = sup u(z)

z∈K z∈∂K

and

inf u(z) = inf u(z)

z∈K z∈∂K

P ROOF. We shall prove that

sup u(z) = sup u(z).

z∈K z∈∂K

(By considering −u which is also a harmonic function, the second statement follows.)
We already have

sup u(z) ≤ sup u(z).

z∈∂K z∈K

(Refer Slide Time: 18:18)

137
 5

Assume that

sup u(z) < sup u(z).

z∈∂K z∈K

Let z 0 be a point in K such that u(z 0 ) = sup u(z).

z∈K

∂2 u ∂2 u
Since z 0 attains the maximum, we have (z 0 ) ≤ 0 and (z 0 ) ≤ 0. Let δ = sup u(z) −
∂x 2 ∂y 2 z∈K
sup u(z).
z∈∂K
Consider the function x 2 + y 2 , then this is bounded above in K by a constant M , as K is
compact.
δ
Define u ϵ = u(z) + ϵ(x 2 + y 2 ), where ϵ < .
2M
(Refer Slide Time: 22:10)

138
6

Now it is left as an exercise to reader to check that sup u ϵ (z) < sup u ϵ (z).
z∈∂K z∈K
Let z ϵ be a point in K where u ϵ attains its maximum.
Hence
∂2 u ϵ

(z ϵ ) ≤ 0

∂x 2


→ (∗∗)
∂2 u ϵ 
(z ϵ ) ≤ 0

∂y 2


But we know that,

∂2 u ϵ ∂2 u ϵ ∂2 u ∂2 u
+ = + 2ϵ + + 2ϵ
∂x 2 ∂y 2 ∂x 2 ∂y 2

(Refer Slide Time: 24:38)

And the Laplacian of u ϵ ,

∆u ϵ (z ϵ ) = ∆u(z ϵ ) + 4ϵ =⇒ ∆u ϵ (z ϵ ) > 0

which is a contradiction to (∗∗).

Hence

sup u(z) = sup u(z).

z∈K z∈∂K

139
 7

T HEOREM 4 (Maximum Principle for Holomorphic Functions ). Let U ⊆ C be open

and connected and let f : U → C be a holomorphic function. Then for K ⊂ U compact, we
have

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

z∈K z∈∂K

P ROOF. Let z ∈ C, then

|z| = sup Re(ze i θ )

θ∈R
³ ´
sup | f (z)| = sup sup Re f (z)e i θ
z∈K z∈K θ∈R
³ ´
iθ
= sup sup Re f (z)e .
θ∈R z∈K

Define f θ (z) = e i θ f (z). Then f θ is holomorphic. Hence Re( f θ ) is harmonic.

(Refer Slide Time: 30:51)

By the maximum principle for harmonic functions, we have

³ ´ ³ ´
sup Re e i θ f (z) = sup Re e i θ f (z) .
z∈K z∈∂K

Hence, we then have

³ ´
sup | f (z)| = sup sup Re e i θ f (z)
z∈K θ∈R z∈∂K
³ ´
= sup sup Re e i θ f (z)
z∈∂K θ∈R

= sup | f (z)|.
z∈∂K

140
8

D EFINITION 2 (Harmonic Conjugate). Let u : Ω ⊆ C → R be a twice differentiable

harmonic function. We say that v : Ω → R is a harmonic conjugate of u if f := u + i v is
holomorphic.

(Refer Slide Time: 34:34)

P ROPOSITION 5. Let u : C → R be a harmonic function. Then there exists a harmonic

function v : C → R such that v is a harmonic conjugate of u. Moreover v is determined
uniquely up to addition by constants.

P ROOF. First we may assume the existence of such a harmonic conjugate and prove
the uniqueness. That is, we will prove if there exists two harmonic conjugates, v 1 and
v 2 , then v 1 − v 2 is a constant.
Let v 1 be a harmonic conjugate. By using fundamental theorem of calculus, we have

∂v 1 ∂v 1
Z x Z y
v 1 (x + i y) = v 1 (0) + (t )d t + (x + i t )d t
0 ∂x 0 ∂y
∂u ∂u
Z x Z y
= v 1 (0) − (t )d t + (x + i t )d t .
0 ∂y 0 ∂x

Similarly, if v 2 is any other harmonic conjugate of u, then

∂u ∂u
Z x Z y
v 2 (x + i y) = v 2 (0) − (t )d t + (x + i t )d t .
0 ∂y 0 ∂x

(Refer Slide Time: 40:01)

141
 9

Hence,
v 1 (x + i y) − v 2 (x + i y) = v 1 (0) − v 2 (0) = c ∈ C.

Now let us prove the existence part of the proposition.

Define
x ∂u y ∂u
Z Z
v(x + i y) = − (t )d t + (x + i t )d t .
0 ∂y 0 ∂x
Then
∂v ∂u
(x + i t ) = (x + i t )
∂y ∂x
which is one of the expressions in the Cauchy-Riemann equations.
(Refer Slide Time: 43:53)

142
10

∂v ∂u ∂ ∂u
µZ y ¶
(x + i y) = − (x) + (x + i t )d t
∂x ∂y ∂x 0 ∂x
∂u ∂ u
Z y 2
= − (x) + (x + i t )d t .
∂y 0 ∂x
2

Since u is harmonic, we have

∂ u ∂ u
Z y 2 Z y 2
(x + i t )d t = − (x + i t )d t
0 ∂x 0 ∂y
2 2

∂u ∂u
µ ¶
=− (x + i y) − (x) .
∂y ∂y
∂v ∂u ∂u ∂u ∂u
(x + i y) = − (x) − (x + i y) + (x) = − (x + i y).
∂x ∂y ∂y ∂y ∂y
Hence u, v satisfy the Cauchy- Riemann equations and therefore f = u + i v is holomor-
phic. □

143
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Module No - 3
Lecture No – 14
Möbius transformations

In this lecture, we will discuss a very important class of function called Möbius trans-
formation. Möbius transformations are special rational function which naturally occurs
in study of various domains of complex analysis. But before we get into the study of
Möbius transformations, it will be useful to study the extended complex plane. The ex-
tended complex plane is nothing but the complex plane with 1 point attached to it.
If you have seen course on topology it is just a 1 point compactification of the com-
plex plane. Let me begin by what an extended complex plane is and we will give a con-
crete description of what that is.
(Refer Slide Time: 01:02)

Extended Complex Plane

We define extended complex plane to be the set Ĉ := C ∪ {∞}.
(Refer Slide Time: 02:06)
1

144
2

Recall that the sphere in R3 is given by S 2 := {(x 1 , x 2 , x 3 ) : x 12 + x 22 + x 32 = 1}.

Let P = (0, 0, 1) and let z = (x, y, 0) be a point in C, then the line joining z and P is
given by L = {(1 − t )z + t P : t ∈ R}. That is L = { (1 − t )x, (1 − t )y, t z) : t ∈ R}.
¡ ¢

Now let us find where this line meets the unit sphere.

(1 − t )2 x 2 + (1 − t )2 y 2 + t 2 = 1

(1 − t )2 |z|2 = (1 − t 2 )

For t ̸= 1,
(1 + t ) |z|2 − 1
|z|2 = =⇒ t = 2
(1 − t ) |z| + 1

|z|2 − 1
µ ¶
2 2x 2y
Hence L and S meets at , , .
|z|2 + 1 |z|2 + 1 |z|2 + 1

Define ϕ : C −→ S 2 given by

2Re(z) 2Im(z) |z|2 − 1

µ ¶
ϕ(z) = , , .
|z|2 + 1 |z|2 + 1 |z|2 + 1

145
 3

By defining ϕ(∞) = P , we have a bijection from Ĉ to S 2 . The map ϕ defined is called

the steriographic projection.
(Refer Slide Time: 10:09)

We define a distance on Ĉ as follows:

d (z, z ′ ) = Euclidean distance of ϕ(z) to ϕ(z ′ ) in S 2

2|z − z ′ |
=¡ ¢1/2 .
(1 + |z|2 )(1 + |z ′ |2 )

If z ′ = ∞, then
2
d (z, z ′ ) = .
1 + |z|2
Möbius Trasformation
Möbius transformations are special class of functions which naturally occur at various
stages in complex analysis.

az + b
D EFINITION 1. A map S(z) := is called a Möbius transformation if ad −bc ̸= 0,
cz + d
where a, b, c, d ∈ C.

E XAMPLE 1.

• I d (z) = z is a Möbius transformation with a = c = 1 and b = d = 0.

146
4

z −i
• S(z) = is a Möbius transformation with a = c = 1, b = −i and d = i . Here
z +i
we can see that S(z) is holomorphic at C \ {−i }.

The observation we made on the last example can be stated in general as Möbius
transformations are rational functions.
az + b
½ ¾
−d
A Möbius transformation S(z) = is holomorphic at C \ .
cz + d c
(Refer Slide Time: 14:48)

dz −b
Consider T (z) = . Now at the domain of definition of T , it satisfies the con-
−c z + a
dition that T ◦ S = I d .
Observe that Möbius transformation can be composes to obtain another Möbius
transformation. Hence Möbius transformations form a group.
We shall consider a Möbius transformation to be defined on Ĉ rather than on C by
az + b
defining for S(z) = ,
cz + d
a
µ ¶
−d
S = ∞ and S(∞) = . When c = 0, then S(∞) = ∞.
c c

E XERCISE 2. S is a bijection on Ĉ.

E XAMPLE 3.

• Translation: S(z) = z + b, for b ∈ C.

• Dilation: S(z) = az, for a ̸= 0. For a = e i θ , S is an isometry on C where θ ∈ R.

147
 5

1
• Inversion: S(z) = .
z

P ROPOSITION 4. Any Möbius transformation can be written as a composition of trans-

lations, dilation and inverson.

In the proposition, note that it is not necessary that all must appear. For example if
you take a dilation, we doesn’t need to compose with any translation and inversion.
(Refer Slide Time: 23:19)

az + b
P ROOF. Consider any Möbius transformation S(z) = .
cz + d
³a´ b
µ ¶
b
µ ¶ ³a´
If c = 0, S(z) = z+ = S 1 ◦ S 2 (z), where S 1 (z) = z + and S 2 (z) = z.
d d d µ ¶ d
d 1
If c ̸= 0, we can see that S = S 4 ◦ S 3 ◦ S 2 ◦ S 1 , where S 1 (z) = z + , S 2 (z) = , S 3 (z) =
c z
(bc − ad ) ³a´
z and S 4 (z) = z + . □
c2 c

az + b aw + b
Let S(z) = and w be a fixed point S =⇒ S(w) = w =⇒ = w =⇒
cz + d cw + d
aw + b = w(cw + d ) =⇒ cw 2 + (d − a)w − b = 0. Hence if S is not the identity, then S can
have a maximum of two fixed points.
(Refer Slide Time: 30:10)

148
6

Let a, b, c be distinct points in Ĉ and α, β, and γ be such that S(a) = α, S(b) = β and
S(c) = γ. Let T be any other Möbius transformation such that T (a) = α, T (b) = β and
T (c) = γ. Then T −1 ◦ S is a Möbius transformation which fixes a, b, c ∈ Ĉ. But we have
already observed that all non-identity Möbius transformation fixes maximum of two
points, which forces T −1 ◦ S = I d =⇒ T = S since a, b, c are distinct points.
Let a, b, c ∈ Ĉ be distinct.
If a, b, c ∈ C, define

(z − a)
µ ¶

(b − a)
S(z) := µ ¶.
(z − c)
(b − c)

Then S(a) = ∞, S(b) = 1 and S(c) = ∞.

If a = ∞, then define

(b − c)
S(z) := .
(z − c)

Then S(∞) = 0, S(b) = 1, S(c) = ∞.

(Refer Slide Time: 34:59)

149
 7

If b = ∞, then
(z − a)
S(z) = .
(z − c)
If c = ∞, then
(z − a)
S(z) = .
(b − a)
Thus from the observations we made, we have the following proposition:

P ROPOSITION 5. Given distinct points a, b, c ∈ Ĉ, then there exists a unique Möbius
transformation S such that S(a) = 0, S(b) = 1 and S(c) = ∞.

C OROLLARY 6. Given distinct points a, b, c ∈ Ĉ and distinct points α, β, γ ∈ Ĉ, then

there exists a unique Möbius transformation S such that S(a) = α, S(b) = β and S(c) = γ.

P ROOF. Let S 1 be a Möbius transformation such that S 1 (a) = 0, S 1 (b) = 1 and S 1 (c) =
∞. Similarly, let S 2 be a Möbius transformation such that S 2 (α) = 0, S 2 (β) = 1 and S 2 (γ) =
∞. Now define, T := S 2−1 ◦ S 1 . Then T maps a to α, b to β and c to γ =⇒ T = I d .
□

One of the important and also interesting property of Möbius transformation is, it
maps generalized circles to generalized circles.
(Refer Slide Time: 40:30)

150
8

¯ z − ω1 ¯
½ ¯ ¯ ¾
Recall that a generalized circle in the complex plane is a set of points given by z ∈ C : ¯
¯ ¯ =λ
z − ω2 ¯
where λ is a positive real number.

P ROPOSITION 7. Let S be a Möbius transformation. Then S maps the set R ∪ {∞} to a

generalized circle.

(Refer Slide Time: 42:43)

az + b
P ROOF. We want to study w ∈ S(R ∪ {∞}). Let S −1 (z) = . Now S −1 (w) ∈ R
cz + d
⇐⇒ S −1 (w) = S −1 (w)

aw + b ā w̄ + b̄
⇐⇒ =
cw + d c̄ w̄ + d¯

151
 9

⇐⇒ (a c̄ − āc)|w|2 + (a d¯ − b̄c)w + (b c̄ − ād )w̄ + b d¯ − b̄d = 0

If a c̄ = āc,
αw − ᾱw̄ = b̄d − b d¯ where α = a d¯ − b̄c

2i Im(αw) = 2i Im(b̄d )

Im(αw) = Im(b̄d ) = c ∈ R.

If a c̄ ̸= āc
a d¯ − b̄c b d¯ − b̄d
⇐⇒ |w|2 + αw + ᾱw̄ + γ = 0, where α = ,γ =
a c̄ − āc a c̄ − āc
⇐⇒ |w − α|2 = |α|2 − γ > 0 (Why?).

T HEOREM 8. A Möbius transformation maps generalized circle to generalized circles.

(Refer Slide Time: 50:28)

P ROOF. Pick three distinct points a, b, c on the given generalized circle. Let S be a
Möbius transformation and α = S(a), β = S(b) and γ = S(c).
Let S 1 and S 2 be Möbius transformations such that S 1 (a) = 0 = S 2 (α), S 1 (b) = 1 =
S 2 (β) and S 1 (c) = ∞ = S 2 (γ). Then S = S 2−1 ◦ S 1 . S 1 maps a generalized circle containing
a, b, c to R ∪ {∞} and by the previous proposition S 2−1 maps R ∪ {∞} to a generalized
circle. □

152
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Module No - 3
Lecture No – 15
Problem Session

In this problem session, we will address some problems on the Cauchy- Riemann equa-
tions and its implications. Recall that given Ω ⊆ C a domain and f : Ω → C which is
differentiable. Then the Wirtinger derivatives were defined as

∂f 1 ∂f 1 ∂f
µ ¶
(z 0 ) := (z 0 ) + (z 0 )
∂z 2 ∂x i ∂y

∂f 1 ∂f 1 ∂f
µ ¶
(z 0 ) := (z 0 ) − (z 0 ) .
∂z̄ 2 ∂x i ∂y

P ROBLEM 1. Let f be a polynomial given by

c n,m z n z̄ m .
X
f (z) =
n,m≥0
n+m≤d

Show that

∂f
c n,m (nz n−1 )z̄ m
X
(z) =
∂z n,m≥0
n+m≤d

and

∂f
c n,m z n (m z̄ m−1 ).
X
(z) =
∂z̄ n,m≥0
n+m≤d

(Refer Slide Time: 04:38)

1

153
2

S OLUTION 1. Before going to the solution of the problem, we first observe some of
the properties satisfied by the Wirtinger derivatives.

∂ 1 ∂ 1 ∂
µ ¶
(f + g) = (f + g)+ (f + g)
∂z 2 ∂x i ∂y
1 ∂f 1 ∂f 1 ∂g 1 ∂g
µ ¶ µ ¶
= + + +
2 ∂x i ∂y 2 ∂x i ∂y
∂ f ∂g
= + .
∂z ∂z

Similarly
∂ ∂ f ∂g
(f + g) = + .
∂z̄ ∂z̄ ∂z̄

Wirtinger derivatives satisfies the product rule also:

∂ 1 ∂ 1 ∂
µ ¶
(f g) = (f g)+ (f g)
∂z 2 ∂x i ∂y
1 ∂g ∂ f 1 ∂g ∂f
µ µ ¶¶
= f +g + f +g
2 ∂x ∂x i ∂y ∂y
1 ∂g 1 ∂g 1 ∂f 1 ∂f
µ µ ¶¶ µ µ ¶¶
=f + +g +
2 ∂x i ∂y 2 ∂x i ∂y
∂g ∂f
=f +g .
∂z ∂z

154
 3

Similarly
∂ ∂g ∂f
(f g) = f +g .
∂z̄ ∂z̄ ∂z̄
Now observe that
∂z̄ 1 ∂z̄ 1 ∂z̄
µ ¶
1
= + = (1 − 1) = 0.
∂z 2 ∂x i ∂y 2
m
∂z̄
Hence by applying product rule, we have = 0.
∂z
We also know that
∂z n
= nz n−1 .
∂z
c n,m z n z̄ m .
X
Let f (z) =
n,m≥0
n+m≤d

(Refer Slide Time: 12:21)

∂f ∂
z n z̄ m
X ¡ ¢
(z) = c n,m
∂z n,m≥0 ∂z
n+m≤d

m ∂z n ∂ z̄
X
µ n m¶
= c n,m z̄ +z
n,m≥0 ∂z ∂z
n+m≤d

c n,m (nz n−1 )z̄ m .

X
=
n,m≥0
n+m≤d

155
4

∂z ∂z n
In order to prove the second expression, observe that = 0 and hence = 0. Also
∂z̄ ∂z̄
∂z¯m
= m z̄ m−1 .Now it is an easy exercise for the reader to complete the solution.
∂z̄

Before we do the next problem, let us also look at the Laplacian operator in terms of
the Wirtinger derivatives.
Recall that the Laplacian operator was given by

∂2 ∂2
∆= + .
∂x 2 ∂y 2

Now by the definition of Wirtinger derivatives, we have

∂f ∂f ∂f
= +
∂x ∂z ∂z̄

and

∂f ∂f ∂f
i = − .
∂y ∂z ∂z̄

(Refer Slide Time: 15:44)

156
 5

Hence
∂2 f ∂ ∂f ∂f ∂ ∂f ∂f
µ ¶ µ ¶
= + + +
∂x 2 ∂z ∂z ∂z̄ ∂z̄ ∂z ∂z̄
∂2 f ∂2 f ∂2 f
= + + 2
∂z 2 ∂z̄ 2 ∂z∂z̄
∂ f
2
1 ∂ 1 ∂f ∂f ∂ 1 ∂f ∂f
µ µ µ ¶¶ µ µ ¶¶¶
= − − −
∂y 2 i ∂z i ∂z ∂z̄ ∂z̄ i ∂z ∂z̄
∂ f ∂2 f ∂2 f
µ 2 ¶
=− + −2 .
∂z 2 ∂z̄ 2 ∂z∂z̄
Hence
∂2 f ∂2 f ∂2 f
∆f = + = 4 .
∂x 2 ∂y 2 ∂z∂z̄
c n,m z n z̄ m is holo-
X
From the previous problem we can see that a polynomial f (z) =
n,m≥0
n+m≤d
morphic if and only if c n,m = 0 whenever m > 0. Just like how we can give a characteri-
zation of what holomorphic polynomials will look like, we will also be able to talk about
how harmonic polynomials will be look like.
(Refer Slide Time: 19:34)

P ROBLEM 2. Let f be the polynomial

c n,m z n z̄ m .
X
f (z) =
n,m≥0
n+m≤d

157
6

Then the polynomial f is harmonic on C if and only if c n,m = 0 whenever both n, m are
positive.

S OLUTION 2.
c n,m ∆ z n z̄ m .
X ¡ ¢
∆ f (z) =
n,m≥0
n+m≤d
Now,
∂2 (z n ) ∂ ∂z n
µ ¶
n
∆(z ) = 4 =4 =0
∂z∂z̄ ∂z ∂z̄
and
∂2 (z̄ m ) ∂ ∂z̄ m
µ ¶
m
∆(z̄ ) = 4 =4 = 0.
∂z̄∂z ∂∂z̄ ∂z
(Refer Slide Time: 22:19)

By the properties of Wirtinger derivatives we saw,

∂2 ¡ n m ¢
∆ z n z̄ m = 4
¡ ¢
z z̄
∂z∂z̄
∂ ∂ ¡ n m¢
=4 z z̄
∂z ∂z̄
∂ ¡
(m − 1)z n z̄ m−1
¢
=4
∂z
= 4(m − 1)(n − 1)z n−1 z̄ m−1 .

(Refer Slide Time: 24:02)

158
 7

Therefore,
c n,m 4(m − 1)(n − 1)z n−1 z̄ m−1 .
X
∆ f (z) =
n,m≥0
n+m≤d
Hence ∆ f (z) = 0 if and only if c n,m = 0 ∀ n, m > 0.

P ROBLEM 3. Let f : C → C be an entire function. Then f is real valued only if f is

constant.

S OLUTION 3. Let f = u + i v be real valued =⇒ v = 0.

(Refer Slide Time: 27:34)

159
8

Then,
x ∂u y ∂u
Z Z
u(x + i y) = u(0) + (t )d t + (x + i t )d t
0 ∂x 0 ∂y
x ∂v y ∂v
Z Z
= u(0) + (t )d t − (x + i t )d t
0 ∂y 0 ∂x

= u(0).

Hence u(z) = u(0).

160
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Module No - 3
Lecture No – 16
Curves in the Complex Plane

In the last few weeks we explore the notion of differentiability in great detail. Just
like in the real analysis setting, the notion of complex differentiability is also tied down
together with notion of integration. And among the notions of integration, we will be
most interested in integral along curves or rather line integrals. We will plunge into that
in the next few weeks but in this lecture let us review and recall the notion of curve and
the various properties of curves. Let us begin this lecture by recalling what are curves?
(Refer Slide Time: 00:59)

D EFINITION 1. A continuous parametrized curve is a continuous map γ : [a, b] −→ C.

If a = b, then the curve is trivial. The point γ(a) is called the initial point of γ and
γ(b) is called the terminal point.
γ is said to be a closed curve if γ(a) = γ(b). We say that γ is a simple curve if γ(t ) ̸= γ(t ′ )
with the exception of t = a, t ′ = b.
The image γ([a, b]) of γ is called the image of the curve.
1

161
2

E XAMPLE 1.

• Let z 1 , z 2 ∈ C,

γz1 →z2 := (1 − t )z 1 + t z 2 , t ∈ [0, 1].

If z 1 = z 2 , then γz1 →z2 is a closed curve but not simple.

• Define γ(θ) = z 0 + r e i θ , θ ∈ [0, 2π]. Then image of curve is a circle of radius r
centered z 0 . Here the initial and terminal points is z 0 + r . Hence γ is a simple
closed curve.
Now consider γ1 and γ2 defined by

γ1 (θ) = z 0 + r e 2πi θ , θ ∈ [0, 1]

γ2 (θ) = z 0 + r e 2i θ θ ∈ [0, 2π].

Then γ1 is a simple closed curve having initial and terminal point z 0 + r and
image of γ1 is same as that of γ, but still we treat them as different curves as
domain of γ and γ1 are different. In the case of γ2 , it looks similar to γ, however
γ2 is not a simple closed curve as for any point in the image of γ2 we have two
preimages.

Now we would like to somehow identify γ and γ1 , and the right notion to look at for this
purpose of curves is the continuous reparametrization.

(Refer Slide Time: 10:21)

162
 3

D EFINITION 2. We say that a curve γ2 : [a 2 , b 2 ] −→ C is a continuous reparametriza-

tion of γ1 : [a 1 , b 1 ] −→ C if ∃ a homeomorphism ϕ : [a 1 , b 1 ] −→ [a 2 , b 2 ] such that ϕ(a 1 ) =
a 2 and ϕ(b 1 ) = b 2 and such that γ2 (ϕ(t )) = γ1 (t ), ∀ t ∈ [a 1 , b 1 ].

By the definition, reparametrized curves must have the same initial and terminal
point and also the image must be same. Hence γz2 →z1 is not a reparametrization of
γz1 →z2

E XERCISE 2. Continuous reparametrization is an equivalence relation.

D EFINITION 3. We say that a curve −γ is a reversal of a curve γ : [a, b] → C if −γ :

[−b, −a] → C and −γ(t ) = γ(−t )

(Refer Slide Time: 16:54)

163
4

E XAMPLE 3. Let γz2 →z1 be the straight line joining z 2 to z 1 , then

γz2 →z1 ≡ −γz1 →z2

where γz1 →z2 is a straight line joining z 1 to z 2 .

Now let us define the notion of concatenation.

(Refer Slide Time: 18:31)

Let γ1 : [a 1 , b 1 ] −→ C and γ2 : [a 2 , b 2 ] −→ C be two curves such that the terminal point

of γ1 coincides with the initial point of γ2 . Let γ̃2 be a reparametrization of γ2 through
ϕ : [a 2 , b 2 ] −→ [b 1 , b 2 + (b 1 − a 2 )]. That is, γ̃2 : [b 1 , b 2 + (b 1 − a 2 )] −→ C given by

γ̃2 (t ) = γ2 (t − (b 1 − a 2 )).

Define

γ1 + γ2 : [a 1 , b 2 + b 1 − a 2 ] −→ C

164
 5

by

γ1 (t ),

a1 ≤ t ≤ b1
(γ1 + γ2 )(t ) :=
γ̃ (t ),

b1 ≤ t ≤ b2 + b1 − a2 .
2

(Refer Slide Time: 23:23)

E XERCISE 4.

• γ1 + γ2 is continuous.
• Let γ1 , γ2 , γ3 be three curves such that terminal point of γ1 is the initial point of
γ2 and terminal point of γ2 is the initial point of γ3 . Then,

(γ1 + γ2 ) + γ3 = γ1 + (γ2 + γ3 ).

P ROPOSITION 5. Let γ1 , γ2 , γ̃1 , γ̃2 be curves such that terminal point of γ1 coincides
with the initial point of γ2 . Let γ̃1 and γ̃2 be continuous reparametrization of γ1 and γ2
respectively. Then,

γ1 + γ2 ≡ γ̃1 + γ̃2

and

−γ1 = −γ̃1 .

P ROOF. Let γ1 : [a 1 , b 1 ] −→ C and γ̃1 : [c 1 , d 1 ] −→ C. Let ϕ1 : [a 1 , b 1 ] −→ [c 1 , d 1 ] be

the homeomorphism such that ϕ1 (a 1 ) = c 1 and ϕ1 (b 1 ) = d 1 and such that γ̃1 (ϕ1 (t )) =
γ1 (t ), ∀ t ∈ [a 1 , b 1 ]. Similarly, let γ2 : [a 2 , b 2 ] −→ C and γ̃2 : [c 2 , d 2 ] −→ C. Let ϕ2 : [a 2 , b 2 ] −→
[c 2 , d 2 ] be homeomorphism.
(Refer Slide Time: 27:38)

165
6

Now γ1 + γ2 : [a 1 , b 2 + b 1 − a 2 ] −→ C and γ̃1 + γ̃2 : [c 1 , d 2 + d 1 − c 2 ] −→ C.

Let us define ψ : [a 1 , b 2 + b 1 − a 2 ] −→ [c 1 , d 2 + d 1 − c 2 ] by


ϕ1 (t ),

a1 ≤ t ≤ b1
ψ(t ) :=
ϕ (t − b + a ) + d − c ,

b1 ≤ t ≤ b2 + b1 − a2 .
2 1 2 1 2

Check that ψ is a homeomorphism. Hence γ1 + γ2 ≡ γ̃1 + γ̃2 .

Now to prove other result,
−γ1 : [−b 1 , −a 1 ] −→ C and −γ̃1 : [−d 1 , −c 1 ] −→ C. Since −ϕ : [−b 1 , −a 1 ] −→ [−d 1 , −c 1 ]
and −ϕ is a homeomorphism (Why?) we have the result. □

E XERCISE 6. −(γ1 + γ2 ) ≡ (−γ1 ) + (−γ2 ).

(Refer Slide Time: 33:08)

166
 7

D EFINITION 4 (Arc-length). Let γ : [a, b] −→ C be a curve. We define the arc-length

of the curve γ to be
n
X
|γ| := sup |γ(t j ) − γ(t j −1 )|
j =0
where the supremum is over n and all partitions a = t 0 < t 1 < · · · < t n = b.

We say that the curve γ is rectifiable if |γ| is finite.

E XERCISE 7. |γ1 + γ2 | = |γ1 | + |γ2 |.

D EFINITION 5. A curve γ : [a, b] −→ C is said to be continuously differentiable if for

each t 0 ∈ (a, b),
γ(t ) − γ(t 0 )
γ′ (t 0 ) = lim
t →t 0 t − t0
t ∈[a,b]\{t 0 }
exists and is continuous.

(Refer Slide Time: 37:35)

167
8

L EMMA 8. Let g : [a, b] −→ C be continuous. Then

b b
¯Z ¯ Z
¯ ¯
¯
¯ g (t )d t ¯¯ ≤ |g (t )|d t .
a a

Z b Z b Z b
Here, g (t )d t := Re(g (t ))d t + i Im(g (t ))d t .
a a a

P ROOF. Fix θ ∈ R,

b ¶¯ ¯Z b
¯ µ Z ³ ´ ¯¯
¯Re e i θ iθ
¯ ¯ ¯
¯ g (t )d t ¯ = ¯
¯ ¯ Re e g (t ) d t ¯¯
a a
Z b¯ ³ ´¯
iθ
≤ ¯Re e g (t ) ¯ d t
¯ ¯
a
Z b¯ ¯
¯ iθ
≤ ¯e g (t )¯ d t
¯
a
Z b
= |g (t )|d t .
a

Taking supremum, we get

b b
¯Z ¯ Z
¯ ¯
¯
¯ g (t )d t ¯¯ ≤ |g (t )|d t .
a a

(Refer Slide Time: 41:43)

168
 9

T HEOREM 9. Let γ : [a, b] −→ C be a continuous differentiable curve. Then γ is rectifi-

able and
Z b
|γ| = |γ′ (t )|d t .
a

Z b
The formula |γ| = |γ′ (t )|d t is called the arc-length formula.
a

P ROOF. Let a = t 0 < t 1 < · · · < t n = b be a partition of [a, b]. Using the fundamental
theorem of calculus,
¯ ¯
n ¯ n ¯Z t j ¯
¯γ(t j ) − γ(t j −1 )¯ = γ′ (t )d t ¯
X ¯ X ¯ ¯
¯
j =0 j =0 t j −1
¯ ¯
n Z tj ¯
¯γ′ (t )¯ d t
X ¯
≤
j =0 t j −1
Z b
= |γ′ (t )|d t .
a

Hence taking supremum over partitions, we get

Z b
|γ| ≤ |γ′ (t )|d t .
a

From this, we get |γ| is finite and hence a continuous differentiable curve is rectifiable.
(Refer Slide Time: 45:39)

169
10

To prove the other inequality, consider the following set.

Given ϵ > 0, let
t′
( Z )
E ϵ = t ∈ [a, b] : ¯γ ↾[a,t ′ ] ¯ ≥ |γ′ (t )|d t − ϵ(t ′ − a), ∀t ′ ≤ t .
¯ ¯
a

Claim: E ϵ is closed.
Let t n ∈ E ϵ be a sequence converging to t 0 . We want to show that
Z t′
¯γ ↾[a,t ′ ] ¯ ≥ |γ′ (t )d t | − ϵ(t ′ − a), ∀ t ′ ≤ t 0 .
¯ ¯
a

For, t < t 0 , this is satisfied. Because t n is a sequence which converges to t 0 . So, if t ′ < t 0
′

is fixed then there will be t m for some m which will be greater than t ′ . Since t n ∈ E ϵ and
t ′ < t m =⇒ t ′ satisfies the condition.
Now let’s check the condition for t ′ = t 0 . If {t n }n∈N is a sequence converging to t 0 from
the right, again we are done by above argument.
Let us assume that {t n }n∈N is a sequence converging from the left to t 0 . Then we have
following observation,
Z tn
¯γ ↾[a,t ] ¯ ≥ ¯γ ↾[a,t ] ¯ ≥ |γ′ (t )|d t − ϵ(t n − a).
¯ ¯ ¯ ¯
0 n
a

Hence t 0 ∈ E ϵ .

170
 11

(Refer Slide Time: 52:39)

Claim: E ϵ is open.
Let t 0 ∈ E ϵ . We already know that [t 0 − δ, t 0 ] ⊂ E ϵ for δ small enough.
Since γ is continuously differentiable at t 0 , given ϵ > 0, ∃ δ > 0 such that
¯ γ(t ) − γ(t ′ ) ¯ ϵ
¯ ¯
′
¯
¯ t − t′ − γ (t 0 )¯¯ ≤ , whenever (t − t 0 ) < δ.
2
Then,
ϵ
|γ′ (t 0 )(t − t 0 )| − |γ(t ) − γ(t 0 )| ≤ (t − t 0 ).
2
′ ϵ
|γ(t ) − γ(t 0 )| ≥ |γ (t 0 )|(t − t 0 ) − (t − t 0 ). → (∗)
2
Since γ is continuous, we can pick δ > 0 small enough so that for t ∈ (t 0 , t + δ),
′

ϵ´ ϵ
Z t Z t³
|γ′ (t )|d t ≤ |γ′ (t 0 )| + d t ≤ |γ′ (t 0 )|(t − t 0 ) + (t − t 0 ). → (∗∗)
t0 t0 2 2
Using (∗) and (∗∗),
t ϵ
Z
|γ′ (t )|d t ≤ |γ(t ) − γ(t 0 )| + (t − t 0 ).
t0 2
Now,
t ϵ
Z
¯γ ↾[t ,t ] ≥ |γ(t ) − γ(t 0 )| ≥ |γ′ (t )|d t − (t − t 0 ).
¯ ¯
¯
0
t0 2
Since t 0 ∈ E ϵ ,
t0 ϵ
Z
¯γ ↾[a,t ] ¯ ≥ |γ′ (t )|d t − (t 0 − a).
¯ ¯
0
a 2

171
12

(Refer Slide Time: 01:00:27)

Hence,
Z t
¯γ ↾[a,t ] ¯ ≥ |γ′ (t )|d t − ϵ(t − a), ∀ t ∈ [a, t 0 + δ).
¯ ¯
a
Then, E ϵ is open and [a, b] is connected =⇒ E ϵ = [a, b].
Hence ∀ ϵ > 0,
Z b
|γ| ≥ |γ′ (t )|d t − ϵ(b − a).
a
Z b
Hence, |γ| ≥ |γ′ (t )|d t .
a Z b
Therefore |γ| = |γ′ (t )|d t . □
a

E XERCISE 10.
¯ ¯
• ¯γz1 →z2 ¯ = |z 2 − z 1 |.
• Let γ(θ) = z 0 + r e i θ for θ ∈ [0, 2π]. Then |γ| = 2πr.

172
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 17
Complex Integration Over Curves

In the last few lectures, we discussed the notion of complex differentiability in great
detail. Just like in the case of real analysis, complex differentiability is tied together in-
tricately with the notion of complex integration. When we say complex integration here,
we mean integration of complex valued functions over rectifiable curves. We will define
the integration of complex valued functions over rectifiable curves in this lecture. There-
after, we will prove fundamental theorems of calculus for such functions. We will also
later on will prove a special version of the fundamental theorem of calculus for holomor-
phic functions defined on nice domains, it is called the Cauchy’s theorem. Let us recall
the notion of the Riemann integral of real valued functions defined on closed interval
[a, b]. We will be defining the complex line integral over rectifiable curves similarly.
(Refer Slide Time: 01:17)

Recall that if f : [a, b] −→ R is continuous, then the Riemann integral of f over [a, b]
is the limit of the Riemann sums

n
f (t ∗j )(t j − t j −1 )
X
j =1

173
2

where the limit is taken as the partition size |P | converges to 0.

Here P : a = t 0 < t 1 < · · · < t n = b and |P | = max (t j − t j −1 ) and t ∗j ∈ [t j −1 , t j ].
1≤ j ≤n
Z b
This limit is denoted by f (t )d t .
a
(Refer Slide Time: 04:06)

Let γ be a rectifiable curve. That is, γ : [a, b] −→ C continuous and |γ| is finite. Let
f : γ([a, b]) −→ C be a continuous function.
Then for a partition P of [a, b], a = t 0 < t 1 < · · · < t n = b, we define the Riemann sum
n ³ ´¡
f γ(t ∗j ) γ(t j ) − γ(t j −1 )
X ¢
j =1

where t J∗ ∈ [t j −1 , t j ].

P ROPOSITION 1. The above Riemann sum converge to a complex limit as the partition
Z
size converge to 0. The limit is denoted by f (z)d z.
γ
More precisely, given ϵ > 0, ∃ δ > 0 such that
¯Z ¯
X ³ ´¡
¯ f (z)d z − f γ(t ∗ ) γ(t j ) − γ(t j −1 ) ¯ < ϵ
¯ ¢¯
j
γ
¯ ¯

whenever |P | < δ.

(Refer Slide Time: 09:23)

174
 3

P ROOF. We shall prove that given ϵ > 0, ∃ δ > 0 such that if P 1 : a = t 0 < t 1 < · · · < t n = b
and P 2 : a = s 0 < s 1 , · · · < s m = b be partitions of size less than δ, then

¯ ¯
¯X n ³ ´¡ m ¢¯¯
∗
¡ ∗ ¢¡
f γ(t j ) γ(t j ) − γ(t j −1 ) − f γ(s k ) γ(s k ) − γ(s k−1 ) ¯ < ϵ.
¯ ¢ X
¯
¯ j =1 k=1
¯

n ³ ´¡ m
f γ(t ∗j ) γ(t j ) − γ(t j −1 ) and R(P 2 ) = f γ(s k∗ ) γ(s k ) − γ(s k−1 ) .
X ¢ X ¡ ¢¡ ¢
Let R(P 1 ) =
j =1 k=1

If Q is a refinement of P 1 , P 2 , then |Q| < min(|P 1 |, |P 2 |).

|R(P 1 ) − R(P 2 )| ≤ |R(P 1 ) − R(Q)| + |R(Q) − R(P 2 )|.

Hence we may assume that P 2 is a refinement of P 1 .

We thus have integers 0 = m 0 < m 1 < · · · < m n = m such that t j = s m j .

(Refer Slide Time: 14:26)

175
4

Then,

¯ ¯
¯X n ³ ´¡ m ¢¯¯
∗
¡ ∗ ¢¡
f γ(t j ) γ(t j ) − γ(t j −1 ) − f γ(s k ) γ(s k ) − γ(s k−1 ) ¯
¯ ¢ X
¯
¯ j =1 k=1
¯
¯ Ã m !¯
¯X n ³ ´¡ n j ¯
f γ(t ∗j ) γ(t j ) − γ(t j −1 ) − f γ(s ℓ∗ ) γ(s ℓ ) − γ(s ℓ−1 ) ¯
¢ X X ¡ ¢ ¡ ¢
=¯
¯ ¯
j =1 ℓ=m j −1 +1
¯ j =1 ¯
mj
ï ¯!
n ¯ ³ ´¡ ¢¯¯
∗
¡ ∗ ¢¡
¯ f γ(t j ) γ(t j ) − γ(t j −1 ) − f γ(s ℓ ) γ(s ℓ ) − γ(s ℓ−1 ) ¯
X ¢ X
≤
¯
j =0 ℓ=m j −1 +1
¯ ¯
à m !
n Xj ¯ ³ ´ ¯ ¯¡
¯ f γ(t ∗ ) − f γ(s ∗ ) ¯ ¯ γ(s ℓ ) − γ(s ℓ−1 ) ¯ .
X ¡ ¢ ¢¯
≤
¯ ¯
j ℓ
j =1 ℓ=m j −1 +1

By uniform continuity of f ◦ γ, we have a δ > 0 such that

¯ ³ ´ ¢¯¯ ϵ
¯ f γ(t ∗j ) − f γ(s ℓ∗ ) ¯ < whenever |t ∗j − s ℓ∗ | < δ.
¯ ¡
|γ|

Thus for |P 1 | < δ, we have

176
 5

mj mj
à !
n ¯ ³
∗
´ ³ ´¯ ¯¡
∗ ¯¯ ϵ X n
¯ f γ(t j ) − f γ(s ℓ ) ¯ γ(s ℓ ) − γ(s ℓ−1 ) ≤ ¯γ(s ℓ ) − γ(s ℓ−1 )¯
X X ¯ ¢¯ X ¯ ¯
¯
j =1 ℓ=m j −1 +1 |γ| j =1 ℓ=m j −1 +1

ϵ X m ¯
¯γ(s k ) − γ(s k−1 )¯
¯
≤
|γ| k=1

ϵ
≤ |γ| = ϵ.
|γ|
□

Properties
Let γ : [a, b] −→ C be a rectifiable and f : γ([a, b]) −→ C be continuous.

(1) Let γ̃ : [c, d ] −→ C be a continuous reparametrization of γ. Then

Z Z
f (z)d z = f (z)d z.
γ γ̃

P ROOF. Let ϕ : [a, b] −→ [c, d ] be a homeomorphism such that γ̃(ϕ(t )) =

γ(t ).
(Refer Slide Time: 23:57)

It is enough to prove that given ϵ > 0, ∃δ > 0 such that if P : a = t 0 < t 1 < · · · <
t n = b is a partition of [a, b] such that |P | < δ, then
¯Z ¯
¯ n ³ ´¡ ¯
f γ(t ∗j ) γ(t j ) − γ(t j −1 ) ¯ < ϵ.
X ¢
¯ f (z)d z −
¯ ¯
¯ γ̃ j =0
¯

177
6

Let ϵ > 0 be given. We know that ∃ δ′ > 0 such that if Q : c = s 0 < s 1 < · · · < s m = b
such that Q < δ′ , then

¯Z ¯
¯ m ³ ´¡ ¢¯¯
∗
f γ̃(s j ) γ̃(t j ) − γ̃(t j −1 ) ¯ < ϵ.
X
¯ f (z)d z −
¯
¯ γ̃ j =0
¯

Since ϕ is continuous on a compact set, it is uniformly continuous.

Hence ∃ δ > 0 such that ¯ϕ(x) − ϕ(y)¯ < δ′ whenever |x − y| < δ.
¯ ¯

(Refer Slide Time: 27:21)

Let P : a = t 0 < t 1 < · · · < t n = b such that |P | < δ.

Now it is left as an exercise for the reader to check that ϕ is monotonically in-
creasing.
Hence, we have s 0 = ϕ(t 0 ) < s 1 Then since γ̃ ◦ ϕ = γ,

178
 7
¯Z ¯
¯ n ³ ´¡ ¯
f γ(t ∗j ) γ(t j ) − γ(t j −1 ) ¯
X ¢
¯ f (z)d z −
¯ ¯
¯ γ̃ j =0
¯
¯Z ¯
¯ n ¯
f γ̃(ϕ(t j )∗ ) γ(t j ) − γ(t j −1 ) ¯
X ¡ ¢ ¡ ¢
= ¯ f (z)d z −
¯ ¯
¯ γ̃ j =0
¯
Now,
s 0 = ϕ(t 0 ) < s 1 < · · · < s n = ϕ(t n )

where s j = ϕ(t j ). Moreover ¯ϕ(t j ) − ϕ(t j −1 )¯ < δ′ .

¯ ¯

Since |P | < δ,

¯Z ¯
¯ n ¢¯¯
∗
f γ̃(ϕ(t j ) ) γ(t j ) − γ(t j −1 ) ¯
X ¡ ¢¡
¯ f (z)d z −
¯
¯ γ̃ j =0
¯
¯Z ¯
¯ n ³ ´¡ ¯
f γ̃(s ∗j ) γ̃(s j ) − γ̃(s j −1 ) ¯ < ϵ.
X ¢
= ¯ f (z)d z −
¯ ¯
¯ γ̃ j =0
¯

As ϵ can be arbitrary small, we have

Z Z
f (z)d z = f (z)d z. □
γ γ̃

(Refer Slide Time: 33:18)

179
8

Z Z
(2) f (z)d z = − f (z)d z.
−γ γ

P ROOF. Let γ : [a, b] −→ C. Then (−γ) : [−b, −a] −→ C given by (−γ)(t ) =

γ(−t ). Let P : a = t 0 < t 1 < · · · < t n = b be a partition of [a, b] =⇒ P ′ : −b = s 0 <
s 1 < · · · < s n = −a where s j = −t n− j .
Z Z
Given ϵ > 0, let δ > 0 be such that f (z)d z and f (z)d z can be ϵ− ap-
(−γ) γ
proximated by partition of size less than δ.
Let P be as above with |P | < δ,

n
¯Z Z ¯ ¯Z ³ ´¡
f γ(t ∗j ) γ(t j ) − γ(t j −1 )
¯ ¯ ¯ X ¢
¯ f (z)d z + f (z)d z ¯¯ = ¯¯ f (z)d z −
γ
¯
−γ −γ j =1
n ³ ´¡ Z ¯
γ(t ∗j ) γ(t j ) − γ(t j −1 ) + f (z)d z ¯¯
X ¢ ¯
+ f
j =1 γ

n
¯Z ³ ´¡ Z
∗
f γ(t j ) γ(t j ) − γ(t j −1 ) +
¯ X ¢
= ¯ f (z)d z −
¯ f (z)d z
γ j =1 (−γ)
n ³ ´¡ ¯
∗
X ¢¯
+ f (−γ)(s n− j ) (−γ)(s n− j ) − (−γ)(s n− j +1 ) ¯¯
j =1

180
 9

n
¯Z ³ ´¡ ¯
∗
≤ ϵ + ¯¯
¯ X ¢¯
f (z)d z − f (−γ)(s j ) (−γ)(s j ) − (−γ)(s j −1 ) ¯¯
(−γ) j =1

< 2ϵ.
Since ϵ is arbitrary, we have
Z Z
f (z)d z = − f (z)d z.
(−γ) γ

(3) Let f be defined on γ1 ([a, b]) γ2 ([c, d ]) where γ1 and γ2 are rectifiable curves
S

such that γ1 (b) = γ2 (c). Then

Z Z Z
f (z)d z = f (z)d z + f (z)d z.
γ1 +γ2 γ1 γ2

Proof of property (3) is left as an exercise for the reader.

(Refer Slide Time: 43:21)

(4) Change of variable.

Let γ be a continuously differentiable curve γ : [a, b] −→ C. Then

Z Z b
f γ(t ) γ′ (t )d t .
¡ ¢
f (z)d z =
γ a

181
10

P ROOF. Given ϵ > 0, let δ > 0 be such that if |P | < δ,

¯Z ¯
¯ n ³ ´¡ ¢¯¯
∗
f γ(t j ) γ(t j ) − γ(t j −1 ) ¯ < ϵ
X
¯ f (z)d z −
¯
¯ γ j =1
¯

for a partition P : a = t 0 < t 1 < · · · < t n = b of [a, b]. Let γ = ϕ + i ψ. Then,

n ³ ´¡ n ³ ´¡
f γ(t ∗j ) γ(t j ) − γ(t j −1 ) = f γ(t ∗j ) ϕ(t j ) − ϕ(t j −1 )
X ¢ X ¢
j =1 j =1
n ³ ´¡
f γ(t ∗j ) ψ(t j ) − ψ(t j −1 ) .
X ¢
+i
j =1
By Mean Value Theorem,
n ³ ´¡ n ³ ´
f γ(t ∗j ) ϕ(t j ) − ϕ(t j −1 ) = f γ(t ∗j ) ϕ′ (s ∗j )(t j − t j −1 )
X ¢ X
j =1 j =1

where s ∗j ∈ [t j −1 , t j ].
Hence,
¯ ¯ ¯
¯X n ³ ´¡ ¯ ¯X n ³ ´³ ´
f γ(t ∗j ) ϕ(t j ) − ϕ(t j −1 ) ¯ = ¯¯ f γ(t ∗j ) ϕ′ (s ∗j ) − ϕ′ (t ∗j ) (t j − t j −1 )
¯ ¢¯
¯
¯ j =1 ¯ j =1

n ³ ´ ¯
γ(t ∗j ) ϕ ′
(t ∗j )(t j
X ¯
+ f − t j −1 )¯¯
j =1
¯ ¯
¯X n ³ ´³ ´ ¯
≤¯ f γ(t ∗j ) ϕ′ (s ∗j ) − ϕ′ (t ∗j ) (t j − t j −1 )¯
¯ ¯
¯ j =1 ¯
¯ ¯
¯X n ³ ´ ¯
∗ ′ ∗
+¯ f γ(t j ) ϕ (t j )(t j − t j −1 )¯.
¯ ¯
¯ j =1 ¯
By uniform continuity of ϕ′ , let δ > 0 be small enough so that
¯ ¯
¯ ′ ∗ ′ ∗ ¯
¯ϕ (s j ) − ϕ (t j )¯ < ϵ.

If M = sup f ◦ γ then we have,

[a,b]
¯ ¯
¯X n ³ ´³ ´ ¯ X n
f γ(t ∗j ) ϕ′ (s ∗j ) − ϕ′ (t ∗j ) (t j − t j −1 )¯ ≤ M ϵ(t j − t j −1 ) = ϵM (b − a).
¯ ¯
¯
¯ j =1 ¯ j =0

(Refer Slide Time: 52:19)

182
 11

¯ ¯
¯ n
¯P ³ ´³ ´ ¯
Therefore, as |P | → 0, ¯ f γ(t ∗j ) ϕ′ (s ∗j ) − ϕ′ (t ∗j ) (t j − t j −1 )¯ → 0.
¯
¯ j =1 ¯
Hence,

n ³ ´¡ n ³ ´
f γ(t ∗j ) ϕ(t j ) − ϕ(t j −1 ) = lim f γ(t ∗j ) ϕ′ (t ∗j )(t j − t j −1 )
X ¢ X
lim
|P |→0 j =1 |P |→0 j =1

Z b
= f (γ(t ))ϕ′ (t )d t .
a
Similarly,

n ³ ´¡ Z b
∗
f γ(t j ) ψ(t j ) − ψ(t j −1 ) = f (γ(t ))ψ′ (t )d t .
X ¢
lim
|P |→0 j =1 a

Hence,

Z Z b Z b
f γ(t ) ϕ′ (t )d t + i f γ(t ) ψ′ (t )d t
¡ ¢ ¡ ¢
f (z)d z =
γ a a
Z b
f γ(t ) γ′ (t )d t .
¡ ¢
=
a

(Refer Slide Time: 55:01)

183
12

(5) If f : γ([a, b]) −→ C is bounded by M , i.e., sup | f (z)| = M . Then

z∈γ([a,b])
Z
f (z)d z ≤ M |γ|.
γ

(6) Let γ : [a, b] −→ C and f , g be defined on γ([a, b]). Then for c ∈ C,

Z Z Z
(c f + g )(z)d z = c f (z)d z + g (z)d z.
γ γ γ

E XERCISE 2. Prove property (5) and property (6).

184
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 18
First Fundamental Theorem of Calculus

In this lecture, we will study the complex analog of the fundamental theorem of calculus,
rather the first fundamental theorem of calculus. In a course on real analysis, you would
have seen a version of the first fundamental theorem of calculus. in a course on real
analysis, you would have seen a version of the first fundamental theorem of calculus.
The complex analog is quite similar.
(Refer Slide Time: 00:54)

T HEOREM 1. Let Ω be an open subset of C and f : Ω −→ C be a continuous function.

Let F : Ω −→ C be an antiderivative of f , i.e., F is holomorphic on Ω and F ′ (z) = f (z) ∀ z ∈
Ω. Suppose γ is a rectifiable curve defined on Ω. Then
Z
f (z)d z = F (z 1 ) − F (z 0 )
γ

where z 0 is the initial point of γ and z 1 is the terminal point of γ.

1

185
2

P ROOF. Let γ : [a, b] −→ Ω. Suppose γ is continuously differentiable. Then

Z Z b
f γ(t ) γ′ (t )d t .
¡ ¢
f (z)d z =
γ a

Since F ′ (z) = f (z)∀z ∈ Ω,

Z Z b
F ′ γ(t ) γ′ (t )d t .
¡ ¢
f (z)d z =
γ a

Now it is left as an exercise to the reader to check that

¢′
F ′ γ(t ) γ′ (t ) = F ◦ γ (t )
¡ ¢ ¡

which follows from chain rule and Cauchy-Riemann equations.

Hence,

Z Z b¡ ¢′
f (z)d z = F ◦ γ (t )d t
γ a

= F γ(b) − F γ(a)
¡ ¢ ¡ ¢

= F (z 1 ) − F (z 0 ).

So if the curve γ is continuously differentiable, our work becomes extremely easy and we
will be able to conclude the complex analog of fundamental theorem of calculus from
the fundamental theorem of calculus proved in the real analysis setting. However, we
have put more generality here. We are only assuming that our curve is rectifiable. So we
will have to work a bit more to prove our result in the complex analog case.
(Refer Slide Time: 08:25)

186
 3

Given ϵ > 0, let

½ ¯Z ¯ ¾
Ωϵ = t ∈ [a, b] : ¯¯ f (z)d z − F γ(t 0 ) − F γ(a) ¯ ≤ ϵ|γ ↾[a,t0 ] |, ∀a ≤ t 0 ≤ t .
¯ ¡ ¡ ¢ ¡ ¢¢¯
¯
γ↾[a,to ]

It is enough to show that Ωϵ is both open and closed.

Let t 0 be a limit point of Ωϵ , i.e., ∃ {t n }n∈N ⊂ Ωϵ such that t n → t 0 . If t n > t 0 , then
t 0 ∈ Ωϵ .
Hence,
¯Z we may assume that t n ↑ t 0 . ¯
f (z)d z − F γ(t 0 ) − F γ(a) ¯¯
¯ ¡ ¡ ¢ ¡ ¢¢¯
¯
γ↾[a,to ]
¯
¯Z ¯
¯ ¢¢¯¯
f (z)d z − F γ(t 0 ) − F γ(t n ) + F γ(t 0 ) − F γ(a) ¯
¡ ¡ ¢ ¡ ¢ ¡ ¢ ¡
=¯
¯
¯ γ↾[a,tn ] +γ↾[tn ,t ] ¯
0
¯Z ¯ ¯Z ¯
f (z)d z − F γ(t n ) − F γ(a) ¯¯ + ¯¯ f (z)d z − F γ(t 0 ) − F γ(t n ) ¯¯
¯ ¡ ¡ ¢ ¡ ¢¢¯ ¯ ¡ ¡ ¢ ¡ ¢¢¯
≤ ¯¯
γ↾[a,tn ] γ↾[tn ,to ]

≤ ϵ ¯γ ↾[a,tn ] ¯ + M ¯γ ↾[tn ,t0 ] ¯ + ¯F γ(t 0 ) − F γ(t n ) ¯

¯ ¯ ¯ ¯ ¯ ¡ ¢ ¡ ¢¯

Given ϵ′ > 0, we can pick t n arbitrarily close to t 0 such that

ϵ ¯γ ↾[a,tn ] ¯ + M ¯γ ↾[tn ,t0 ] ¯ + ¯F γ(t 0 ) − F γ(t n ) ¯ ≤ ϵ ¯γ ↾[a,tn ] ¯ + (M + 1)ϵ′ .

¯ ¯ ¯ ¯ ¯ ¡ ¢ ¡ ¢¯ ¯ ¯

Hence,
¯Z ¯
f (z)d z − F γ(t 0 ) − F γ(a) ¯¯ ≤ ϵ ¯γ ↾[a,tn ] ¯ + (M + 1)ϵ′
¯ ¡ ¡ ¢ ¡ ¢¢¯ ¯ ¯
¯
γ↾[a,to ]
¯

187
4

Since ϵ′ is arbitrary,
¯Z ¯
f (z)d z − F γ(t 0 ) − F γ(a) ¯¯ ≤ ϵ ¯γ ↾[a,t0 ] ¯ .
¯ ¡ ¡ ¢ ¡ ¢¢¯ ¯ ¯
=⇒ ¯¯
γ↾[a,to ]

Hence Ωϵ is closed.

(Refer Slide Time: 18:46)

Claim: Ωϵ is open.
Let t 0 ∈ Ωϵ . For δ > 0 such that (t 0 − δ, t 0 + δ) ⊂ [a, b], we have (t 0 − δ, t 0 ] ⊆ Ωϵ .
For t ∈ [t 0 , t 0 + δ).
Since F is holomorphic at γ(t 0 ), given ϵ > 0, ∃ δ > 0 such that
¯ F γ(t ) − F γ(t 0 ) ¢¯ ϵ
¯ ¡ ¢ ¡ ¢ ¯
− f γ(t 0 ) ¯¯
¡
¯
¯ γ(t ) − γ(t 0 ) 2
whenever t ∈ (t 0 , t 0 + δ).
Then

¯F γ(t ) − F γ(t 0 ) − f γ(t 0 ) γ(t ) − γ(t 0 ) ¯ ≤ ϵ ¯γ(t ) − γ(t 0 )¯

¯ ¡ ¢ ¡ ¢ ¡ ¢¡ ¢¯ ¯ ¯
2
ϵ¯
≤ ¯γ ↾[t0 ,t ] ¯ .
¯
2
Now,
Z
f γ(t 0 ) γ(t ) − γ(t 0 ) = f γ(t 0 ) d z.
¡ ¢¡ ¢ ¡ ¢
γ↾[t0 ,t ]

(Refer Slide Time: 23:17)

188
 5

Let δ be small enough such that

¯ f γ(t ) − f γ(t 0 ) ¯ < ϵ .

¯ ¡ ¢ ¡ ¢¯
2

Then,
¯Z ¯ Z
¯ ¢¯¯
f (z)d z − f γ(t 0 ) γ(t ) − γ(t 0 ) ¯ ≤ ¯ f (z) − f γ(t 0 ) ¯ d z
¯ ¡ ¢¡ ¯ ¡ ¢¯
¯
¯ γ↾[t γ↾[t
0 ,t ] 0 ,t ]
¯
ϵ¯
≤ ¯γ ↾[t0 ,t ] ¯ .
¯
2

Hence by triangle inequality, we have

¯Z ¯
f (z)d z − F γ(t ) − F γ(t 0 ) ¯¯ ≤ ϵ ¯γ ↾[t0 ,t ] ¯ → (∗).
¯ ¡ ¡ ¢ ¡ ¢¢¯ ¯ ¯
¯
γ↾[to ,t ]
¯

We know that, since t 0 ∈ Ωϵ , we have

¯Z ¯
f (z)d z − F γ(t 0 ) − F γ(a) ¯¯ ≤ ϵ ¯γ ↾[a,t0 ] ¯ → (∗∗).
¯ ¡ ¡ ¢ ¡ ¢¢¯ ¯ ¯
¯
γ↾[a,to ]
¯

(Refer Slide Time: 28:19)

189
6

Hence,
¯Z ¯
f (z)d z − F γ(t ) − F γ(a) ¯¯ ≤ ϵ ¯γ ↾[a,t0 ] ¯ + ϵ ¯γ ↾[t0 ,t ] ¯
¯ ¡ ¡ ¢ ¡ ¢¢¯ ¯ ¯ ¯ ¯
¯
γ↾[a,t ]
¯

≤ ϵ ¯γ ↾[t0 ,t ] ¯ .
¯ ¯

This is true ∀ t ∈ (t 0 , t 0 + δ) =⇒ (t 0 − δ, t 0 + δ) ⊆ Ωϵ .
Hence Ωϵ is open.
Since a ∈ Ωϵ , we have by connectedness of [a, b], Ωϵ = [a, b].
Therefore, ∀ ϵ > 0, we have
¯Z ¯
¯ f (z)d z − F γ(b) − F γ(a) ¯ ≤ ϵ|γ|.
¯ ¡ ¡ ¢ ¡ ¢¢¯
γ
¯ ¯
Z
=⇒ f (z)d z = F (z 1 ) − F (z 0 )
γ
where z 1 = γ(b) and z 0 = γ(a).
□

E XAMPLE 2.
dez
• We know that = e z . Hence, if γ is a curve from z 1 to z 2 , we have
dz
Z
e z d z = e z2 − e z1 .
γ

190
 7

d ³z ´
• = z. Then,
dz 2
z 22 − z 12
Z
zd z =
γ 2
where z 1 and z 2 are initial
µ and ¶ terminal point of γ respectively.
d 1 1
• On C \ {0}, we have − = 2.
dz z z
1 1 1
Z
2
dz = −
γz z1 z2
where z 1 and z 2 are initial and terminal point of γ respectively.
(Refer Slide Time: 34:11)

∞
a n (z − z 0 )n be a power series converging in D(z 0 , R).
X
• Let F (z) =
n=0
∞
X an
Let G(z) = (z − z 0 )n+1 on D(z 0 , R). Then G ′ (z) = F (z).
n=0
n +1

∞ ∞ ∞
an an
Z X X X
n n+1
a n (z − z 0 ) = (z 2 − z 0 ) − (z 1 − z 0 )n+1 .
γ n=0 n=0
n +1 n=0
n +1

191
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 19
Second Fundamental Theorem of Calculus

In this lecture, we will be proving a second fundamental theorem of calculus, which is

the complex analog of the second fundamental theorem of calculus, which you would
have already seen in the real setting. But before we do that, let us recall the equivalent
formulation of open connectedness in the complex plane.
(Refer Slide Time: 01:03)

Recall: For Ω ⊆ C open, the following are equivalent

i Ω is connected.
ii Ω is path-connected.
iii Ω is polygonally path-connected.

Polygonal path is the concatenation of straight line curves. γz1 →z2 →···→zn used to denote
a polygonal path obtained by concatenating line joining z j to z j +1 .

T HEOREM 1 (Second Fundamental Theorem of Calculus). Let Ω ⊆ C be a non-empty

open connected subset and let f : Ω −→ C be a continuous function such that
Z
f (z)d z = 0
η
1

192
2

whenever η is a closed polygonal path contained in Ω. Fix z 0 ∈ Ω and define

Z
F (z 1 ) = f (z)d z
γ

where γ is a polygonal path from z 0 to z 1 . Then F is well-defined holomorphic function

such that F ′ (z 1 ) = f (z 1 ) ∀z 1 ∈ Ω.

(Refer Slide Time: 09:47)

P ROOF. First, let us prove that the F defined as in the theorem is a well defined func-
tion. That is it is independent of the polygonal path we choose.
Suppose γ1 and γ2 be polygonal paths from z 0 to z 1 . Then γ1 +(−γ) is a closed polyg-
onal path.
Hence,
Z Z Z Z Z
0= f (z)d z = f (z)d z + f (z)d z = f (z)d z − f (z)d z
γ1 +(−γ) γ1 −γ2 γ1 γ2

Z Z
=⇒ f (z)d z = f (z)d z.
γ1 γ2

Therefore F (z 1 ) is well defined.

Let z 1 ∈ Ω and let D(z 1 , r ) ⊆ Ω. Let γ be a polygonal path from z 0 to z 1 .
Pick z 2 ∈ D(z 1 , r ) and γz1 →z2 be the straight line from z 1 to z 2 . Then γ + γz1 →z2 is a
polygonal path from z 0 to z 2 .

193
 3

Now,
Z Z
F (z 2 ) − F (z 1 ) = f (z)d z − f (z)d z
γ+γz1 →z2 γ
Z
= f (z)d z
γz1 →z2

(Refer Slide Time: 15:55)

¯Z ¯
¯ ¯ ¯¯ ¯
¯F (z 2 ) − F (z 1 ) − f (z 1 )(z 2 − z 1 )¯ = ¯ f (z)d z − f (z 1 )(z 2 − z 1 )¯
¯
¯ γz →z ¯
1 2
¯Z ¯
¯ ¡ ¢ ¯¯
=¯ f (z) − f (z 1 ) d z ¯ .
¯
¯ γz →z ¯
1 2

Let ϵ > 0 be given. Then by picking r small enough, we have

¯F (z 2 ) − F (z 1 ) − f (z 1 )(z 2 − z 1 )¯ ≤ ϵ|z 2 − z 1 |.
¯ ¯

Then,
¯ ¯
¯ F (z 2 ) − F (z 1 )
¯ ≤ ϵ.
¯
lim ¯ − f (z 1 )
z 2 →z 1 ¯ z2 − z1 ¯
z 2 ∈Ω\{z 1 }

Hence
F (z 2 ) − F (z 1 )
lim = f (z 1 )
z 2 →z 1 z2 − z1
z 2 ∈Ω\{z 1 }

194
4

That is,

F ′ (z 1 ) = f (z 1 ).

P ROPOSITION 2. Let Ω ⊆ C be a non-empty connected and f : Ω −→ C be continuous

function. Let γ : [a, b] −→ Ω be a rectifiable curve in Ω from z 1 to z 2 . Then given ϵ > 0,
there exists a polygonal path σ : [a, b] −→ Ω from z 1 to z 2 such that

¯Z Z ¯
¯ f (z)d z − f (z)d z ¯ < ϵ.
¯ ¯
γ σ
¯ ¯

P ROOF. We know that γ([a, b]) is compact set. Let U = {D(z 0 , r 0 ), D(z 1 , r 1 ), · · · , D(z n .r n )}
n
be a finite cover such that γ([a, b]) ⊂ D(z k , r k ) and D(z k , r k ) ⊂ Ω for each k = 0, 1, · · · , n.
S
k=0
n
D(z k , r k ). Then K ⊂ Ω and by Heine-Borel theorem, K is compact.
S
Let K =
k=0
Since K is compact, f is uniformly continuous on K . Hence for given ϵ > 0, we have
δ > 0 such that

¯ f (z) − f (w)¯ < ϵ

¯ ¯
2|γ|

whenever |z − w| < δ in K .
Let P : a = t 0 < t 1 < · · · < t n = b be a partition such that |P | < min(δ, Lebesgue number of U )
and
¯Z ¯
¯ n ¯ ϵ
f γ(t j −1 γ(t j ) − γ(t j −1 ) ¯ < .
X ¡ ¢ ¡ ¢
¯ f (z)d z −
¯ ¯
¯ γ j =1
¯ 2

(Refer Slide Time: 29:52)

195
 5

Let σ be a polygonal path such that

σ ↾[t j −1 ,t j ] ≡ γγ(t j −1 )→γ(t j ) .

¯Z Z ¯ ¯¯Z n
¯
¢¯¯
f γ(t j −1 γ(t j ) − γ(t j −1 ) ¯
¯ ¯ ¯ X ¡ ¢¡
¯ f (z)d z − f (z)d z ¯ ≤ ¯ f (z)d z −
γ σ γ
¯ ¯ ¯
j =1
¯
¯ ¯
¯X n Z ¯
f γ(t j −1 γ(t j ) − γ(t j −1 ) − f (z)d z ¯
¡ ¢ ¡ ¢
+¯
¯ ¯
¯ j =1 σ ¯

¯Z Z ¯ ¯¯Z n
¯
¢¯¯
f γ(t j −1 γ(t j ) − γ(t j −1 ) ¯
¯ ¯ ¯ X ¡ ¢¡
¯ f (z)d z − f (z)d z ¯ ≤ ¯ f (z)d z −
γ σ γ
¯ ¯ ¯
j =1
¯
¯ ¯
¯X n Z ¯
f γ(t j −1 γ(t j ) − γ(t j −1 ) − f (z)d z ¯ .
¡ ¢ ¡ ¢
+¯
¯ ¯
¯ j =1 σ ¯

196
6

Now,

¯Z ¯ ¯Z ¯
¯ n ¯ ¯ n ¯
f γ(t j −1 γ(t j ) − γ(t j −1 ) ¯ = ¯ f (z)d z − f γ(t j −1 σ(t j ) − σ(t j −1 ) ¯
X ¡ ¢ ¡ ¢ X ¡ ¢ ¡ ¢
¯ f (z)d z −
¯ ¯ ¯ ¯
¯ σ j =1
¯ ¯ σ j =1
¯
¯Z ¯
¯ n Z ¢ ¯¯
f γ(t j −1 ) d z ¯
X ¡
= ¯ f (z)d z −
¯
¯ σ j =1 σ↾[t j −1 ,t j ]
¯
¯ Z ¯
¯X n ¢¢ ¯¯
f (z) − f γ(t j −1 ) d z ¯
¡ ¡
=¯
¯
¯ j =1 σ↾[t ,t ] ¯
j −1 j
¯ ¯
n ¯Z ¢¢ ¯¯
f (z) − f γ(t j −1 ) d z ¯
X ¡ ¡
≤
¯
¯
j =1 σ↾[t ,t ]
¯ ¯
j −1 j

n ϵ ¯¯
σ(t j ) − σ(t j −1 )¯
X ¯
≤
j =1 2|γ|

ϵ X n ¯
¯γ(t j ) − γ(t j −1 )¯
¯
≤
2|γ| j =1
ϵ
≤ .
2

Hence,

¯Z Z ¯
¯ f (z)d z − f (z)d z ¯ < ϵ.
¯ ¯
γ σ
¯ ¯

(Refer Slide Time: 38:21)

197
 7

Let us define two classes of curves, which will be of greatest interest. They are very
special curves with a much better regularity.

D EFINITION (Smooth Curves). We say that a curve γ : [a, b] −→ C is a smooth curve

if γ is continuously differentiable and γ′ (t ) ̸= 0 ∀ t ∈ [a, b].

The straight line path, for example, is a smooth curve. The most of the curves we
have actually encountered are smooth curves. In fact the equation of a circle is a smooth
curve.

D EFINITION (Contours). We say that a curve γ : [a, b] −→ C is a contour if it is the

concatenation of finitely many smooth curves.

For example, the polygonal paths are contours, which are obtained by concatenation
of finitely many straight line paths.

198
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 20
Problem Session

(Refer Slide Time: 00:14)

P ROBLEM 1. Let γ1 : [a, b] −→ C and γ2 : [c, d ] −→ C be such that γ1 (b) = γ2 (c). Then

|γ1 + γ2 | = |γ1 | + |γ2 |.

S OLUTION 1. The first observation is that if you look at a continuous re-parametrization

of a given curve, then the re-parameterized curve also will have the same arc length. This
was given as an exercise to the reader to verify.
Assuming this exercise, without loss of generality, we may assume that b = c.
(Refer Slide Time: 04:12)
1

199
2

Let ϵ > 0 be given. Let P : a = t 0 < t 1 < · · · < b = t k = c < t k+1 < · · · < t n = b be a
partition of [a, d ] = [a, b] ∪ [b, d ] such that

¯(γ1 + γ2 )(t j ) − (γ1 + γ2 )(t j −1 )¯ < ϵ

n ¯
|γ1 + γ2 | −
X ¯
(1)
j =1 3

and such that

¯γ1 (t j ) − γ1 (t j −1 )¯ < ϵ ,
X k ¯ ¯
(2) |γ1 | −
j =1 3

¯γ2 (t j ) − γ2 (t j −1 )¯ < ϵ .
Xn ¯ ¯
(3) |γ2 | −
j =k+1 3

Existence of such a partition can always ensure because there exists a partition Q 1 such
that (1) is satisfied, similarly there exists partitions Q 2 and Q 3 of [a, b] and [b, d ] respec-
tively such that (2) and (3) are satisfied. Now if we take a common refinement of Q 1 ,Q 2
and Q 3 , then all these conditions are satisfied.

200
 3

Then,
¯ ¯
¢¯ ¯¯ n ¯ ¯¯¯
¯|γ1 + γ2 | − |γ1 | + |γ2 | ¯ ≤ ¯|γ1 + γ2 | − ¯(γ1 + γ2 )(t j ) − (γ1 + γ2 )(t j −1 )¯¯
¯ ¡ X
j =1
¯ ¯
¯ ¯
¯ n ¯ ¯¯¯
¯(γ1 + γ2 )(t j ) − (γ1 + γ2 )(t j −1 )¯¯
X
+ ¯|γ1 | + |γ2 | −
¯
j =1
¯ ¯
¯ ¯ ¯ ¯
ϵ ¯¯ X k ¯ ¯¯¯ ¯¯
¯γ1 (t j ) − γ1 (t j −1 )¯¯ + ¯|γ2 | −
Xn ¯ ¯¯¯
¯γ2 (t j ) − γ2 (t j −1 )¯¯
< + ¯|γ1 | −
3 ¯ j =1
¯ ¯ j =k+1
¯
ϵ ϵ ϵ
< + + =ϵ
3 3 3

Since ϵ that was chosen was arbitrary, we have

|γ1 + γ2 | = |γ1 | + |γ2 |.

(Refer Slide Time: 11:26)

P ROBLEM 2. Let γ : [a, b] −→ C be a curve. Define ℓ : [a, b] −→ R to be ℓ(t ) := ¯γ ↾[a,t ] ¯.

¯ ¯

Then ℓ is a continuous function.

S OLUTION 2. Given t ∈ [a, b], we shall prove that ℓ is left continuous at t . This is
enough since right continuity follows by considering the length corresponding to (−γ).

201
4

Let ϵ > 0 be given. Since γ is uniformly continuous on [a, b], ∃ δ′ such that

ϵ
|γ(s) − γ(t )| <
2

whenever |t − s| < δ′ .
Let P : a = t 0 < t 1 < · · · < t n = t be such that

¯γ ↾[a,t ] ¯ −
n ϵ
|γ(t j ) − γ(t j −1 )| < .
¯ ¯ X
j =1 2

(Refer Slide Time: 16:20)

Let δ < min(δ′ , |t n − t n−1 |) and t ′ ∈ (t − δ, t ].

By considering the refinement, a = t 0 < t 1 < · · · < t n = t ′ < t n+1 = t ,

¯γ(t j ) − γ(t j −1 )¯ < ϵ .

¯ n+1
¯γ ↾[a,t ] ¯ −
¯ X¯ ¯
j =1 2

Hence,
à !
¯γ(t j ) − γ(t j −1 )¯ − ¯γ(t ) − γ(t ′ )¯ < ϵ .
n ¯
¯γ ↾[a,t ′ ] ¯ + ¯γ ↾[t ′ ,t ] ¯ −
¡¯ ¯ ¯ ¯¢ X ¯ ¯ ¯
j =1 2
Now regrouping the above equation,
à !
¯γ(t j ) − γ(t j −1 )¯ + ¯γ ↾[t ′ ,t ] ¯ − ¯γ(t ) − γ(t ′ )¯ < ϵ .
n ¯
¯γ ↾[a,t ′ ] ¯ −
¯ ¯ X ¯ ¡¯ ¯ ¯ ¯¢
j =1 2

n ¯ n ¯
Since ¯γ ↾[a,t ′ ] ¯ > ¯γ(t j ) − γ(t j −1 )¯, we have ¯γ ↾[a,t ′ ] ¯ − ¯γ(t j ) − γ(t j −1 )¯ = c > 0.
¯ ¯ X ¯ ¯ ¯ X ¯
j =1 j =1

202
 5

Thus,

¯γ ↾[t ′ ,t ] ¯ − ¯γ(t ) − γ(t ′ )¯ < ϵ − c < ϵ .

¯ ¯ ¯ ¯
2 2
Hence,

¯γ ↾[t ′ ,t ] ¯ < ϵ + ϵ = ϵ.
¯ ¯
2 2
Therefore,

¯γ ↾[t ′ ,t ] ¯ < ϵ.
¯ ¯

=⇒ ¯γ ↾[a,t ] ¯ − ¯γ ↾[a,t ′ ] ¯ < ϵ.

¯ ¯ ¯ ¯

That is,

ℓ(t ) − ℓ(t ′ ) < ϵ.

(Refer Slide Time: 23:03)

1
P ROBLEM 3. Let f (z) = and γ : [0, 2π] −→ C be the curve given by γ(t ) = 2e i t .
z2 − 1
Compute
Z
f (z)d z.
γ

203
6

S OLUTION 3. We can decompose f into partial fractions,

µ ¶ µ ¶
1 1 1 1 1
f (z) = 2 = − .
z −1 2 z −1 2 z +1

Since γ(t ) = 2e i t , we have

1 1 1 1
Z Z Z
f (z)d z = dz − dz
γ 2 γ z −1 2 γ z +1
1 2π 2i e i t 2i e i t
Z 2π
1
Z
= d t − d t dt
2 0 (2e i t − 1) 2 0 2e i t + 1

(Refer Slide Time: 27:57)

1 2π 2i e i t
Z
Consider d t . Since 2e i t − 1 = 2 cos t − 1 + i 2 sin t , we have
2 0 (2e i t − 1)

2i e i t e i t 2e −i t − 1
¡ ¢
1
Z 2π Z 2π
dt = i ¯ dt
2 (2e i t − 1) ¯2e i t − 1¯2
¯
0 0
Z 2π
(2 − cos t ) − i sin t
=i dt
0 5 − 4 cos t
Z 2π Z 2π
sin t (2 − cos t )
= dt +i dt.
0 5 − 4 cos t 0 5 − 4 cos t

Now by computing the integral of real valued functions on the RHS, we will get,
Z 2π sin t
dt = 0
0 5 − 4 cos t

204
 7

and
Z 2π (2 − cos t )
d t = π.
0 5 − 4 cos t
Hence,
1 1 1 2π 2i e i t
Z Z
dz = d t = i π.
2 γ z −1 2 0 (2e i t − 1)
By a similar computation,
1 1
Z
d z = i π.
2 γ z +1
Hence,
1
Z
d z = i π − i π = 0.
γ z2 − 1
(Refer Slide Time: 32:29)

E XERCISE 4. Let γ(t ) = 1 + e i t . Compute

1
Z
2
d z.
γ z −1

1
P ROBLEM 5. Prove that the function f (z) = does not have an anti-derivative in
z
C \ {0}.

S OLUTION 4. Let us consider γ(t ) = e i t on [0, 2π].

If f (z) had an anti-derivative in C \ {0}, say F , then
Z
f (z)d z = F (1) − F (1) = 0.
γ

205
8

(Refer Slide Time: 37:56)

Z
Let us compute f (z)d z.
γ

1 2π i ei t 2π
Z Z Z
dz = dt = i d t = 2πi ̸= 0.
γz 0 ei t 0

1
Therefore f (z) = does not have an anti-derivative.
z

P ROBLEM 6. Let Ω ⊆ C be an open subset of C and f , g : Ω −→ C be holomorphic. Let

γ : [a, b] −→ Ω be a rectifiable curve. Then

Z Z
′
f (z)g (z)d z = f (z 1 )g (z 1 ) − f (z 0 )g (z 0 ) − f ′ (z)g (z)d z.
γ γ

where z 0 and z 1 are the initial and terminal point of γ respectively.

S OLUTION 5. Let F (z) = ( f (z)g (z))′ . By the fundamental theorem of calculus,

Z
(4) F (z)d z = f (z 1 )g (z 1 ) − f (z 0 )g (z 0 )
γ

(Refer Slide Time: 42:27)

206
 9

By product rule, F (z) = f ′ (z)g (z) + f (z)g ′ (z).

Z Z Z
(5) F (z)d z = f (z)g (z)d z + f (z)g ′ (z)d z.
′
γ γ γ

By (4) and (5) , we have

Z Z
f (z)g (z)d z = f (z 1 )g (z 1 ) − f (z 0 )g (z 0 ) − f ′ (z)g (z)d z.
′
γ γ

207
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 21
Homotopy of Curves

We will soon discuss one of the fundamental theorems in complex analysis called
Cauchy’s theorem. Very broadly speaking Cauchy’s theorem says that if we start off with
a domain Ω and if γ0 and γ1 are two curves in Ω such that γ0 can be continuously de-
Z Z
formed to the curve γ1 then, f (z)d z = f (z)d z if f is a holomorphic function on
γ0 γ1
Ω. The notion of continuously deforming a curve γ0 to obtain a curve γ1 is made precise
by the topological notion called homotopy.
(Refer Slide Time: 00:57)

Let Ω be an open subset of C. Let γ0 : [a, b] −→ Ω and γ1 : [a, b] −→ Ω be curves.

(1) (Homotopy of two curves with fixed end points) Suppose γ0 and γ1 have the
same initial point, z 0 and the same terminal point, z 1 . We say that γ0 is homo-
topic to γ1 with fixed end-point if there exists a continuous map
1

208
2

H : [0, 1] × [a, b] −→ Ω such that ∀ t ∈ [a, b],

H (0, t ) = γ0 (t )

H (1, t ) = γ1 (t )

and ∀ s ∈ [0, 1],

H (s, a) = z 0

H (s, b) = z 1 .

Let us try to understand what this definition actually means. H is a function

defined on [0, 1]×[a, b], where [a, b] is common domain of both γ0 and γ1 . Now
we will think of first parameter from [0, 1] as the stage at which we are in. For
example, if

H (s, t ) = γs (t )

then γs is a continuous map from z 0 to z 1 .

(Refer Slide Time: 05:40)

(2) (Homotopy of closed curves) Suppose γ0 is closed curve at z 0 and γ1 is a closed

curve at z 1 . We say that γ0 is homotopic to γ1 as closed curves if there exists a

209
 3

continuous map H : [0, 1] × [a, b] −→ Ω such that ∀ t ∈ [a, b],

H (0, t ) = γ0 (t )

H (1, t ) = γ1 (t )

and

H (s, a) = H (s, b) ∀ s ∈ [0, 1].

(Refer Slide Time: 10:21)

If γs (t ) = H (s, t ), then γs is a closed curve.

In either case, the map H is called a homotopy from γ0 to γ1 .

E XAMPLE 1. Let Ω be a convex domain (i.e., if z 0 , z 1 ∈ Ω, then (1 − t )z 0 + t z 1 ∈ Ω ∀ t ∈

[0, 1]). Suppose γ0 : [a, b] −→ Ω and γ1 : [a, b] −→ Ω be any two curves from z 0 to z 1 .
Define H : [0, 1] × [a, b] −→ Ω by

H (s, t ) = (1 − s)γ0 (t ) + sγ1 (t ).

Then H is a homotopy from γ0 to γ1 .

Properties
Let Ω ⊆ C be an open subset.

210
4

(1) The relation of γ0 being homotopic to γ1 with fixed end points is an equivalence
relation.

P ROOF.
(a) Let γ0 : [a, b] −→ Ω. Consider H : [0, 1] × [a, b] −→ Ω given by

H (s, t ) := γ0 (t ).

Then H is a homotopy from γ0 to itself with fixed end points.

(b) Let γ0 : [a, b] −→ Ω and γ1 : [a, b] −→ Ω be such that γ0 is homotopic to γ1
with fixed end points. Then H : [0, 1] × [a, b] −→ Ω be a homotopy from γ0
to γ1 .
Define H1 : [0, 1] × [a, b] −→ Ω given by

H1 (s, t ) = H (1 − s, t ).

Then H1 is a homotopy with fixed end points from γ1 to γ0 .

(c) Let γ0 , γ1 , γ2 : [a, b] −→ Ω be such that γ0 is homotopic with fixed end points
to γ1 and γ1 is homotopic with fixed end points to γ2 .
That is, there exist H1 : [0, 1] × [a, b] −→ Ω and H2 : [0, 1] × [a, b] −→ Ω such
that

H1 (0, t ) = γ0 (t ), H1 (1, t ) = γ1 (t )

and

H2 (0, t ) = γ1 (t ), H2 (1, t ) = γ2 (t ).

Define H : [0, 1] × [a, b] −→ Ω given by


 1
 H1 (2s, t )
 0≤s ≤
H (s, t ) := 2
1
 H2 (2s − 1, t ) ≤ s ≤ 1.


2
Then H is a homotopy with fixed end points from γ0 to γ2 .
□

211
 5

A very similar proof tells us that the relation of homotopy as closed curve is
also an equivalence relation.
(2) Let γ0 , γ1 : [a, b] −→ Ω be closed curves at z 0 . Then γ0 ∼ γ1 as closed if and only
if there exists a closed curve γ2 at z 0 such that γ0 is homotopic with fixed end
points to a reparametrization of γ2 + γ1 + (−γ2 ).
(Refer Slide Time: 29:18)

P ROOF. (⇒) Suppose H : [0, 1] × [a, b] −→ Ω be a homotopy of closed curves

from γ0 to γ1 .
Define γ2 : [0, 1] −→ Ω by

γ2 (s) = H (s, a).

Claim: γ0 is homotopic to a reparametrization of γ2 + γ1 + (−γ2 ) with fixed

points.
Define

σs = γ2 ↾[0,s] +γs + (−γ2 ) ↾[−s,0]

where γs (t ) = H (s, t ).

212
6

Then σs : [0, (b − a) + 2s] −→ Ω. Now reparametrize σs to obtain σ̃s ,

σ̃s : [a, b] −→ Ω. This can be done by composing σs with a linear homeomor-
phism between [0, (b − a) + 2s] and [a, b].
Define

H̃ (s, t ) := σ̃s (t ).

Now it is left as an exercise to the reader to verify that H̃ is continuous

and also a homotopy with fixed end points from γ0 to a reparametrization of
γ2 + γ1 + (−γ2 ).
(⇐) It is enough to check that γ1 is homotopic as a closed curves to a reparametriza-
tion of γ2 + γ1 + (−γ2 ), because of the transitivity property of homotopy.
Define

σs := γ2 ↾[1−s,1] +γ1 + (−γ2 ) ↾[−1,s−1]

Let σ̃s : [a, b] −→ Ω be a reparametrization of σs . Define

H : [0, 1] × [a, b] −→ Ω

by

H (s, t ) = σ̃s (t ).

Then H is a homotopy of closed curves from γ1 to a reparametrization of γ2 +

γ1 + (−γ2 ). □

(3) Let γ0 : [a, b] −→ Ω be a curve from z 0 to z 1 and γ1 : [a, b] −→ Ω be such that

γ1 (t ) = z 0 ∀ t ∈ [a, b]. Then a reparametrization of γ0 +(−γ0 ) is homotopic to γ1 .
(Refer Slide Time: 46:02)

213
 7

P ROOF. For s > 1, let

σs := γ0 ↾[0,s] +(−γ0 ) ↾[−s,0]

and σ̃s be a reparametrization of σs to [a, b]. For s = 0, σ̃0 (t ) = z 0 ∀ t ∈ [a, b].

Define
H (s, t ) := σ̃s (t ).

Then H is a homotopy from γ1 to a reparametrization of γ0 + (−γ0 ). □

E XERCISE 2. Let γ0 , γ̃0 : [a, b] −→ Ω and γ1 , γ̃1 : [c, d ] −→ Ω be such that γ0 ∼ γ̃0 and
γ1 ∼ γ̃1 . Then
γ0 + γ1 ∼ γ̃0 + γ̃1 .

Definition. Let γ0 : [a, b] −→ Ω be a closed curve at z 0 . Then γ0 is null-homotopic if γ0

is homotopic to the constant curve where γ1 : [a, b] −→ Ω given by γ1 (t ) = z 0 as closed
curves.

214
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 22
Cauchy-Goursat Theorem

In the last lecture we laid the topological framework to state Cauchy’s theorem.
(Refer Slide Time: 00:30)

Cauchy’s Theorem
Let Ω ⊆ C be open and f : Ω −→ C be holomorphic on Ω.

(i) If γ0 : [a, b] −→ Ω and γ1 : [c, d ] −→ Ω are rectifiable curves from z 0 to z 1 and suppose
γ0 is homotopic to a reparametrization of γ1 , then
Z Z
f (z)d z = f (z)d z.
γ0 γ1

(ii) If γ0 : [a, b] −→ Ω and γ1 : [c, d ] −→ Ω are closed rectifiable curves such that γ0 is
homotopic as closed curves to a reparametrization of γ1 , then
Z Z
f (z)d z = f (z)d z.
γ0 γ1
1

215
2

Cauchy’s theorem is a very powerful tool in the sense that it can be used to com-
pute the integral of f over any rectifiable curve by shifting the curve via homotopy to a
contour over which the integral can be computed easily.
We will now state variant of the Cauchy’s theorem. It essentially says the same.
Cauchy’s Theorem-II
Let Ω ⊆ C be open and f : Ω −→ C is holomorphic on Ω. Suppose γ0 : [a, b] −→ Ω is a
rectifiable curve which is null-homotopic. Then

Z
f (z)d z = 0.
γ0

Let us compare the two variance of Cauchy’s theorem.

Cauchy’s theorem =⇒ Cauchy’s theorem-II is immediate, because in the part (ii) of
Cauchy’s theorem, if we take γ1 to be a constant curve, then the hypothesis of Cauchy’s
Z
theorem-II is satisfied. Conclusion follows from the fact that f (z)d z = 0, where γ1 is
γ1
a constant curve.
Proving the other implication is non-trivial. We may put that as a proposition.

P ROPOSITION 1. Cauchy’s theorem-II =⇒ Cauchy’s theorem.

P ROOF. Here we will prove part (i) of Cauchy’s theorem and leave part (ii) as an exer-
cise to the reader.
(Refer Slide Time: 07:26)

216
 3

Let γ0 : [a, b] −→ Ω, γ1 : [a, b] −→ Ω such that γ0 ∼ γ1 with fixed end points.

γ0 + (−γ1 ) ∼ γ1 + (−γ1 ) with fixed end points

∼ γ2 with fixed end points

where γ2 : [a, b] −→ Ω is such that γ2 (t ) = z 0 = γ0 (a).

By Cauchy’s theorem-II, we have
Z Z Z
f (z)d z = 0 ⇐⇒ f (z)d z = f (z)d z.
γ0 +(−γ1 ) γ0 γ1

□
1
E XAMPLE 2. Let f : C \ {0} −→ C given by f (z) = . Consider γ(t ) = e i t , t ∈ [0, 2π].
z
Then
Z
f (z)d z = 2πi .
γ

T HEOREM 3 (Goursat’s Theorem).

Let Ω ⊆ C, f : Ω −→ C be holomorphic and z 1 , z 2 , z 3 ∈ Ω such that the convex hull of
z 1 , z 2 , z 3 is contained in Ω. Then
Z
f (z)d z = 0.
γz1 →z2 →z3 →z1

217
4

(Refer Slide Time: 16:40)

P ROOF. Let us denote the curve γz1 →z2 →z3 →z1 by T0 .

Suppose
¯Z ¯
f (z)d z ¯¯ ≥ ϵ
¯ ¯
¯
¯
T0

for some ϵ > 0. Let z 12 be the midpoint of the straight line joining z 1 and z 2 . Similarly
z 23 , z 31 be the midpoints of straight-line joining z 2 and z 3 , and z 3 and z 1 respectively.
Then
¯Z ¯ ¯Z Z Z Z ¯
¯ ¯ ¯ ¯
¯ f (z)d z ¯¯ = ¯¯ f (z)d z + f (z)d z + f (z)d z + f (z)d z ¯¯ ,
γ1 γ2 γ3 γ4
¯
T0

where,
γ1 = γz1 →z12 →z31 →z1 , γ2 = γz2 →z23 →z12 →z2 , γ3 = γz3 →z31 →z23 →z3 and γ4 = γz12 →z23 →z31 →z12 .
Thus,
¯Z ¯ ¯Z ¯ ¯Z ¯ ¯Z ¯ ¯Z ¯
ϵ ≤ ¯¯
¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯ ¯
f (z)d z ¯¯ ≤ ¯¯ f (z)d z ¯¯ + ¯¯ f (z)d z ¯¯ + ¯¯ f (z)d z ¯¯ + ¯¯ f (z)d z ¯¯ .
T0 γ1 γ2 γ3 γ4

Hence at least one of γ1 , γ2 , γ3 and γ4 , which shall be denoted as T1 , must satisfy

¯ ϵ
¯Z ¯
¯
¯
¯ f (z)d z ¯≥ .
¯ 4
T1

|T0 |
Note that |T1 | = .
2

218
 5
© ª
Let T̂i denote the convex hull of Ti and let diam(T̂i ) = sup |z − w| : z, w ∈ T̂i .
(Refer Slide Time: 25:31)

diam(T̂0 )
Now, diam(T̂1 ) = . By repeating the process above, we get T0 , T1 , T2 , · · · such
2
that
ϵ
¯Z ¯
¯ ¯
¯ f (z)d z ¯¯ ≥ n −→ (∗).
¯
Tn 4

Also observe that T̂0 ⊃ T̂1 ⊃ · · · .

|T0 | diam(T̂0 )
|Tn | = n
, diam(T̂n ) = .
2 2n

Then |Tn | → 0 and diam(T̂n ) → 0 as n → ∞.

Pick z n ∈ T̂n and then {z n }n∈N is a Cauchy sequence by our choice of T̂i . By com-
∞
T
pleteness, z n → z 0 where z 0 ∈ T̂i .
i =0
Since f is complex differentiable at z 0 , given ϵ′ > 0, ∃ δ > 0 such that

¯ f (z) − f (z 0 ) − f ′ (z 0 )(z − z 0 )¯ < ϵ′ |z − z 0 |

¯¡ ¢ ¯

whenever |z − z 0 | < δ.

219
6

diam(T̂0 )
For large n, T̂n ⊂ D(z 0 , δ). Hence |z − z 0 | < diam(T̂n ) = .
2n
That is,
¯ f (z) − f (z 0 ) − f ′ (z 0 )(z − z 0 )¯ < ϵ′ diam(T̂0 ) .
¯¡ ¢ ¯
2n
Hence
¯Z ¯
diam(T̂0 ) diam(T̂0 )
′
f (z) − f (z 0 ) + f (z 0 )(z − z 0 ) d z ¯¯ ≤ ϵ′ |Tn | = ϵ′ |T0 |
¯ ¡ ¡ ¯ ¢¢
¯
n
.
¯
Tn 2 4n

That is,
¯Z ¯
diam(T̂0 )
Z
′
f (z 0 ) + f (z 0 )(z − z 0 ) d z ¯¯ ≤ ϵ′ |T0 |
¯ ¡ ¢ ¯
¯ f (z)d z − .
¯
Tn Tn 4n

(Refer Slide Time: 34:17)

Now it is left to reader to verify by using first fundamental theorem of calculus that,
Z
f (z 0 ) + f ′ (z 0 )(z − z 0 ) d z = 0.
¡ ¢
Tn

Hence
¯Z ¯
diam(T̂0 )
f (z)d z ¯¯ ≤ ϵ′ |T0 |
¯ ¯
¯ .
¯
Tn 4n

220
 7

diam(T̂0 )
Then for ϵ′ small enough such that ϵ′ |T0 | ϵ,
<

ϵ
¯Z ¯
¯ ¯
¯ f (z)d z ¯¯ < n
¯
Tn 4

which is a contradiction to (∗).

Thus,
¯Z ¯
f (z)d z ¯¯ < ϵ ∀ ϵ > 0
¯ ¯
¯
¯
T0

Z
=⇒ f (z)d z = 0.
T0

T HEOREM 4 (Cauchy’s Theorem of Polygonal Paths).

Let Ω ⊆ C be a convex and γz1 →z2 →···→zn →z1 be a closed polygonal path and suppose
f : Ω −→ C be holomorphic. Then
Z
f (z)d z = 0.
γz1 →z2 →···→zn →z1

(Refer Slide Time: 40:22)

221
8

P ROOF. Proof is by using induction. For the base case, we take n = 3 because for
n = 1, 2 we have nothing to prove. For n = 3 Goursat’s theorem tells us that
Z
f (z)d z = 0
γ

where γ is a closed polygonal path with 3 vertices.

Assume the result is proved for up to n − 1. Then,
Z Z Z
f (z)d z = f (z)d z + f (z)d z
γz1 →z2 →···→zn →z1 γz1 →z2 →···→zn−1 →z1 γzn−1 →zn →z1 →zn−1

= 0+0 = 0

222
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 23
Cauchy’s Theorem

In the last lecture as a corollary to Goursat theorem, we proved that if Ω is a convex

domain and if γ is a closed polygonal path in Ω, then for any holomorphic function f
on Ω, γ f (z)d z = 0. By using a complex analog of the second fundamental theorem of
R

calculus, which was proved in the last week, we can immediately conclude a version of
Cauchy’s theorem in the case of convex domains. This is sometimes also called the local
version of the Cauchy’s theorem.
(Refer Slide Time: 00:51)

T HEOREM 1 (Cauchy’s Theorem for Convex Domains).

Let Ω ⊆ C be a convex open set and f : Ω −→ C be holomorphic on Ω. Then f has an
anti-derivative in Ω and if γ is closed rectifiable curve in Ω, then

Z
f (z)d z = 0.
γ
1

223
2
Z
P ROOF. By Cauchy’s theorem of polygonal paths, we have f (z)d z = 0 for any closed
σ
polygonal path σ in Ω. Then by the second fundamental theorem of calculus, we have
an explicit anti-derivative F of f on Ω which was given by
Z
F (z 1 ) = f (z)d z
γ

where γ is the straight line from z 0 to z 1 for z 0 ∈ Ω fixed.

By the first fundamental theorem of calculus,
Z
f (z)d z = 0
γ

for every rectifiable curve γ in Ω since f has an anti-derivative. □

T HEOREM 2 (Cauchy’s Theorem).

Let Ω ⊆ C be open and f : Ω −→ C be holomorphic on Ω. Let γ0 : [a, b] −→ Ω and γ1 :
[c, d ] −→ Ω be rectifiable curves on Ω from z 0 to z 1 and such that γ0 is homotopic with
fixed end-points to a reparametrization of γ1 , then
Z Z
f (z)d z = f (z)d z.
γ0 γ1

P ROOF. Since the integral is invariant under reparametrization, we may assume that
γ1 is defined on [a, b].
Let H : [0, 1] × [a, b] −→ Ω be a homotopy with fixed end-points from γ0 to γ1 . Since
[0, 1] × [a, b] is compact, we have H ([0, 1] × [a, b]) is compact in Ω.
Let U = {U1 ,U2 , . . . ,Un } be open subsets of Ω such that U j ⊆ Ω and H ([0, 1] ×
n
U j . Let r be the Lebesgue number corresponding to U . Then for any (s, t ) ∈
S
[a, b]) ⊂
j =1
[0, 1] × [a, b], D(H (s, t ), r ) ⊆ Ω.
Since [0, 1] × [a, b] is compact, H is uniformly continuous on [0, 1] × [a, b] and hence
∃ δ > 0 such that
r
|H (s, t ) − H (s ′ , t ′ )| < whenever |s − s ′ | < δ and |t − t ′ | < δ.
4
(Refer Slide Time: 17:03)

224
 3

Let P 1 : 0 = s 0 < s 1 < · · · < s n = 1 be a partition of [0, 1] such that |P 1 | < δ and let
P 2 : a = t 0 < t 1 < · · · < t m = b be a partition of [a, b] such that |P 2 | < δ.
(Refer Slide Time: 20:19)

Let C i , j denote the polygonal path γH (si ,t j )→H (si ,t j −1 )→H (si −1 ,t j −1 )→H (si −1 ,t j )→H (si ,t j ) . By
our choice of δ, we can ensure that diam(C i , j ) < r . Hence C i , j ⊂ D(H (s i , t j ), r ).

225
4

By Cauchy’s theorem for polygonal paths on convex open sets, we have

Z
f (z)d z = 0.
Ci , j

Now it is left as an exercise to the reader to verify that

Xn Xm Z Z
f (z)d z = f (z)d z
i =1 j =1 C i , j C

where C := γH (1,tm )→H (1,tm−1 )→···→H (1,t0 )→H (0,t1 )→···→H (0,tm ) . Then C is a reparametrization
of (−σ1 ) + σ2 , where
σ1 := γH (1,t0 )→···→H (1,tm )

σ2 := γH (0,t0 )→···→H (0,tm ) .

Then
Z Z Z Z
f (z)d z = f (z)d z = f (z)d z − f (z)d z.
C (−σ1 )+σ2 σ2 σ1
Since
Z
f (z)d z = 0 ∀ i, j,
Ci , j

we have
Z Z Z
f (z)d z = 0 =⇒ f (z)d z = f (z)d z.
C σ1 σ2
(Refer Slide Time: 30:06)

226
 5

Notice that γ0 ↾[t j ,t j +1 ] ([t j , t j +1 ]) ⊆ D(H (0, t j ), r ) by our choice of partition. Similarly,
γH (0,t j )→H (0,t j +1 ) ([0, 1]) ⊆ D(H (0, t j ), r ).
By Cauchy’s theorem on convex sets,
Z Z
f (z)d z = f (z)d z.
γ0 ↾[t j ,t j +1 ] γH (0,t j )→H (0,t j +1 )

Z Z
f (z)d z = f (z)d z
σ2 γH (0,t0 )→···→H (0,tm )
m−1
XZ
= f (z)d z
j =0 γH (0,t j )→H (0,t j +1 )

m−1
XZ
= f (z)d z
j =0 γ0 ↾[t j ,t j +1 ]
Z
= f (z)d z
γ0 ↾[0,t1 ] +···+γ[tm−1 ,tm ]
Z
= f (z)d z.
γ0

Similarly we can establish that

Z Z
f (z)d z = f (z)d z.
σ1 γ1

Hence we can conclude that

Z Z
f (z)d z = f (z)d z.
γ0 γ1

D EFINITION 1 (Simply Connected Domains). An open set Ω ⊆ C is said to be simply

connected if every closed curve γ at z 0 ∈ Ω is null homotopic to the constant curve at z 0 .

Now we can state a special case of Cauchy’s theorem for simply connected domains.
(Refer Slide Time: 38:39)

227
6

T HEOREM 3. Let Ω ⊆ C be simply connected and f : Ω −→ C be holomorphic on Ω,

then
Z
f (z)d z = 0
γ
for every closed rectifiable curve γ in Ω.

Proof of the theorem is immediate from the Cauchy’s theorem and the definition of
simply connected domain.

228
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 24
Problem Session

P ROBLEM 1. Let Ω = C \ {0} and γ : [a, b] −→ Ω be a closed curve. Prove that there
exists a closed curve σ such that γ is homotopy as closed curves to σ and image of σ is
contained in S 1 .

(Refer Slide Time: 00:30)

S OLUTION 1. Given γ : [a, b] −→ Ω is closed curve. That is γ(a) = γ(b). Define

γ(t )
σ(t ) := .
|γ(t )|
Then
γ(a) γ(b)
σ(a) = = = σ(b).
|γ(a)| |γ(b)|
Hence we have a closed curve whose image is contained in S 1 , which is σ.
It remains to prove that γ is homotopy as closed curves to σ.
(Refer Slide Time: 03:43)
1

229
2

Define

H : [0, 1] × [a, b] −→ Ω

given by

H (s, t ) = (1 − s)γ(t ) + sσ(t )

Now it is left as an exercise to the reader to verify that H is indeed a homotopy from γ to
σ as closed curve.

P ROBLEM 2. Compute the integral:

Z ∞
cos(t 2 )d t .
0

³ 2´
S OLUTION 2. Notice that cos(t 2 ) = Re e i t and

Z ∞ Z R
= lim cos(t 2 )d t .
0 R→∞ 0

(Refer Slide Time: 06:58)

230
 3

Now, let γ1 (t ) = t for t ∈ [0, R]. Then, by the change of variable,

Z Z R
i z2 2
e dz = ei t d t .
γ1 0

Now lets define another path from 0 to R. Let

γ2 (t ) = t e i π/4 t ∈ [0, R]

γ3 (t ) = Re i t t ∈ [0, π/4].

Then γ2 (t ) + (−γ3 ) is a curve from 0 to R.

2
Since e i z is an entire function, by Cauchy’s theorem,
Z Z
i z2 2
e dz = e i z d z.
γ1 γ2 +(−γ3 )

Z Z Z
i z2 i z2 2
e dz = e dz − ei z d z
γ2 +(−γ3 ) γ2 γ3
Z R ¢2 Z π/4 ¢2
i t e i π/4 i π/4 Re i t
¡ ¡
= e e dt − ei Ri e i t d t .
0 0
Z R µ
1+i
¶ Z π/4 2 2i t
−t 2
= e p dt − eiR e
Ri e i t d t .
0 2 0

231
4

Now,

(1 + i )
Z R 1+i
Z ∞
−t 2 2
lim p e dt = p e −t d t
R→∞ 2 0 2 0
µp ¶
1+i π
= p
2 2
Z π/4 2 2i t
Now let’s consider the term eiR e
Ri e i t d t .
0
(Refer Slide Time: 15:30)

We know that e 2i t = cos(2t ) + i sin(2t ). Now,

¯Z π/4 ¯ Z π/4 ¯ ¯
i R 2 e 2i t it ¯ i R 2 e 2i t it¯
¯ ¯
¯
¯ e Ri e d t ¯ ≤
¯ ¯e Ri e ¯ d t
0 0
Z π/4 2
sin(2t )
≤R e −R dt
0
Z π/4 2
≤R e −R t d t
0

R −R 2 t ¯¯π/4
µ ¶¯
=− 2 e
R ¯
0
1³ 2
´
= 1 − e −R π/4
R
1
≤ .
R

232
 5

Hence,
Z
2
lim e z d z = 0.
R→∞ γ3

Hence
p
∞ 1+i π
Z
= p .
0 2 2
p p
∞ ∞ π ∞ ∞ π
Z Z Z Z
2 i t2 2 i t2
cos(t )d t = Re e dt = p ; sin(t )d t = Im e dt = p .
0 0 2 2 0 0 2 2

P ROBLEM 3. Prove that

Z ∞ Z ∞
−(x+i α)2 2
e dx = e −x d x
−∞ ∞

S OLUTION 3. (Refer Slide Time: 21:33)

We can rewrite the L.H.S in the problem as,

Z ∞ Z R
−(x+i α)2 2
e d x = lim e −(x+i α) d x.
−∞ R→∞ −R

2
Let f (z) = e −z . Then f is an entire function.
(Refer Slide Time: 23:37)

233
6

Let γ1 , γ2 , γ3 , γ4 be straight lines such that,

γ1 : −R −→ R

γ2 : R −→ R + i α

γ3 : R + i α −→ −R + i α

γ4 : −R −→ −R + i α.

Then we can parametrize these curves as follows:

γ1 (t ) = t t ∈ [−R, R]

γ2 (t ) = R + i t t ∈ [0, α]

γ3 (t ) = −t + i α t ∈ [−R, R]

γ4 (t ) = −R + i t t ∈ [0, α].

Let us compute the integral of f along all these curves,

Z Z R 2
f (z)d z = e −t d t .
γ1 −R

234
 7
Z Z α 2
−t 2 )+2i Rt )
f (z)d z = e −((R idt
γ2 0
¯Z ¯ ¯Z α ¯
2 2
−( (R −t )+2i Rt )
¯ ¯ ¯ ¯
=⇒ ¯¯ f (z)d z ¯¯ = ¯¯ e i d t ¯¯
γ2 0
Z α
2 2
≤ e −(R −t ) d t
0
Z α
2 2
≤ e (R −α ) d t
0
2
αe α
= 2
−→ 0 as R −→ ∞.
eR

Let us consider the integral of f over −γ3 ,

Z Z R 2
f (z)d z = e −(t +i α) d t .
(−γ3 ) −R

Z Z ∞ 2
lim f (z)d z = e −(t +i α) d t .
R→∞ (−γ3 ) −∞

Since the case of integral over γ4 is also going to similar as in the case of γ2 , we have
Z
f (z)d z −→ 0 as R −→ ∞.
γ4

(Refer Slide Time: 31:19)

235
8

By Cauchy-Goursat theorem,
Z Z Z Z
f (z)d z + f (z)d z + f (z)d z + f (z)d z = 0
γ1 γ2 γ3 (−γ4 )
Z Z
=⇒ f (z)d z + f (z)d z = 0 as R −→ ∞
γ1 γ3
Z Z
f (z)d z = f (z)d z
γ1 (−γ3 )
Z ∞ 2
Z ∞
2
e −x d x = e −(x+i α) d x.
−∞ −∞

P ROBLEM 4. Let Ω be a simply connected domain and f : Ω −→ C be a holomorphic

function such that f (z) ̸= 0 ∀ z ∈ Ω and f ′ is also holomorphic. Then there exists g :
Ω −→ C such that
f (z) = e g (z) .

f ′ (z) f ′ (z)
S OLUTION 4. Consider the function . Since f is non-vanishing on Ω, is
f (z) f (z)
holomorphic on Ω. Since Ω is simply connected, there exists an anti-derivative g 0 (z)
f ′ (z)
such that g 0′ (z) = . Define h(z) = e (g 0 (z)) .
f (z)
Consider
f (z) ′ f ′ (z)h(z) − f (z)h ′ (z)
µ ¶
= .
h(z) (h(z))2
Now,
f ′ (z)
f ′ (z)h(z) − f (z)h ′ (z) = f ′ (z)e g 0 (z) − f (z)e g 0 (z) = f ′ (z)e g 0 (z) − e g 0 (z) f ′ (z) = 0.
f (z)
Hence, on Ω,
¶′
f (z)
µ
=0
h(z)
f (z) = ch(z) = ce g 0 (z)
′ ′
Let c ′ ∈ C be such that e c = c, then f (z) = e c +g 0 (z) . Put g (z) = c ′ + g 0 (z), then g is holo-
morphic on Ω and f = e g .

236
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 25
Cauchy Integral Formula

We have stated and proven Cauchy’s theorem; let us realize some of its consequences.
As noted earlier, Cauchy’s theorem is one of the most fundamental and one of the most
powerful theorems in complex analysis. We will do justice to this claim by proving some
very beautiful and striking consequences: the first one among them being the Cauchy
integral formula. Cauchy integral formula is a powerful theorem in itself and many times
it forms that tool for application of Cauchy’s theorem to conclude various things about
holomorphic functions.
(Refer Slide Time: 00:49)

237
2

T HEOREM 1 (Cauchy Integral Formula). Let Ω ⊆ C be an open subset of the complex

plane and f : Ω −→ C be holomorphic. Fix z 0 ∈ Ω and let r > 0 be such that D(z 0 , r ) ⊆ Ω.
Suppose γ is a closed curve in Ω \ {z 0 } such that γ is homotopic as closed curve up to a
reparametrization to γ1 in Ω \ {z 0 } where γ1 (t ) = z 0 + r e i t for t ∈ [0, 2π]. Then

1 f (z)
Z
f (z 0 ) = d z.
2πi γ z − z0

f (z)
P ROOF. Since f is holomorphic on Ω, we have is holomorphic on Ω\{z 0 } and
(z − z 0 )
γ is homotopic as closed curve to γ1 in Ω \ {z 0 }, by Cauchy’s theorem,

f (z) f (z)
Z Z
dz = d z.
γ (z − z 0 ) γ1 (z − z 0 )

(Refer Slide Time: 07:00)

238
 3

In D(z 0 , r ) \ {z 0 }, the circle, γ2 (t ) = z 0 + ϵe i t for t ∈ [0, 2π], of radius ϵ > 0 around z 0 is

homotopic as closed curves to the circle γ1 . Hence by Cauchy’s theorem,
f (z) f (z) f (z)
Z Z Z
dz = dz = d z.
γ (z − z 0 ) γ1 (z − z 0 ) γ2 (z − z 0 )

Since f is complex differentiable at z 0 , ∃ ϵ′ > 0 such that for ϵ < ϵ′ and M > 0,
¯ ¯
¯ f (z) − f (z 0 ) ¯ n
it
o
¯
¯ z −z
¯<M
¯ for z ∈ z 0 + ϵe : t ∈ [0, 2π] .
0

We can rewrite this as, ¯Z ¯

f (z) − f (z 0 ) ¯¯
d z ¯ ≤ 2M πϵ
¯
¯
¯
γ2 z − z0
¯Z ¯
f (z) f (z 0 ) ¯¯
Z
d z ¯ ≤ 2M πϵ.
¯
(1) ¯ dz −
¯
γ2 z − z0 γ2 z − z 0

Since
f (z 0 ) dz
Z Z
d z = f (z 0 ) = 2πi f (z 0 )
γ2 z − z0 γ2 z − z0
(1) can be written as,
¯Z ¯
f (z)
¯ z − z d z − 2πi f (z 0 )¯ ≤ 2M πϵ
¯ ¯
¯ ¯
γ 0
¯ ¯
¯ 1 f (z)
Z
d z − f (z 0 )¯¯ ≤ M ϵ.
¯
¯
γ z − z0
¯ 2πi

1 f (z)
Z
Since ϵ > 0 was arbitrary, we have f (z 0 ) = d z. □
2πi γ z − z 0
D EFINITION 1. We say that a function f : Ω −→ C is Complex analytic if given z 0 ∈ Ω,
∞
there exists a disc D(z 0 , r ) ⊆ Ω such that the power series a n (z − z 0 )n converges in
X
n=0
D(z 0 , r ) to the function f .

T HEOREM 2. Let f : Ω −→ C be holomorphic on Ω and z 0 ∈ Ω. Suppose D(z 0 , r ) ⊆ Ω,

then for every n ∈ N, let
1 f (z)
Z
an = n+1
dz
2πi γ (z − z 0 )
∞
where γ(t ) = z 0 + r e i t for t ∈ [0, 2π]. Then the power series a n (z − z 0 )n converges in
X
n=0
D(z 0 , r ) to f (z).

239
4

P ROOF. Since f is continuous on γ, by compactness, ∃ M > 0 such that | f (z)| ≤ M on

γ.
¯ ¯
¯ 1 f (z) ¯ ≤ 1 M 2πr = M 1 .
Z
¯
|a n | = ¯
¯
n+1
d z ¯ 2π r n+1
2πi γ (z − z 0 ) rn
Hence
µ ¶1/n
1/n 1 1
lim sup |a n | = lim sup M n = .
r r
∞
a n (z − z 0 )n converges in D(z 0 , r ).
X
Hence
n=0
(Refer Slide Time: 22:00)

Let w ∈ D(z 0 , r ) and γ2 be the curve γ2 (t ) = w +ϵe i t for t ∈ [0, 2π] where ϵ > 0 be such
that D(w, ϵ) ⊆ D(z 0 , r ). In D(z 0 , r ) \ {w}, we have γ2 is homotopic to γ and hence by the
Cauchy integral formula,

1 f (z)
Z
(2) f (w) = d z.
2πi γ z −w

240
 5

We have,
1 1
=
(w − z 0 )
µ ¶
z −w
(z − z 0 ) 1 −
(z − z 0 )
¯ ¯
¯ (w − z 0 ) ¯
On γ, we have ¯¯ ¯ < 1. Hence,
(z − z 0 ) ¯
1 1 X ∞ µ w − z ¶n
0
(3) =
z − w z − z 0 n=0 z − z 0
Substituting (3) in (2), we have
1
Z µ ∞ (w − z )n ¶
X 0
f (w) = f (z) n+1
d z.
2πi γ n=0 (z − z 0 )
Now,
¯ n ¯¯
¯ f (z) (w − z 0 ) ¯ ≤ M |w − z 0 |n
¯
¯ (z − z )n+1 ¯ r n+1
0
M
≤ ρn where 0 < ρ < 1.
r n+1
∞
X (w − z 0 )n
By Weierstrass M −test, the series f (z) converges uniformly and hence we
n=0 (z − z 0 )n+1
have,
∞ (w − z 0 )n ∞ µ 1 Z f (z)
Z µX ¶ ¶
1
d z (w − z 0 )n .
X
f (w) = f (z) n+1
dz = n+1
2πi γ n=0 (z − z 0 ) n=0 2πi γ (z − z 0 )
∞
a n (z − z 0 )n in D(z 0 , r ).
X
Hence we have f (w) = □
n=0

C OROLLARY 3. If f : Ω −→ C is holomorphic, then f ′ : Ω −→ C is holomorphic.

C OROLLARY 4. If f : Ω −→ C is holomorphic, then f is infinitely differentiable.

241
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 26
Principle of Analytic Continuation and Cauchy Estimates

We saw in the last lecture that holomorphic functions can be identified with complex
analytic functions. We will now look at one more corollary to the theorem, which states
that a holomorphic function can be written as a power series on sufficiently small disc
in the domain of definition, called the factor theorem for analytic functions.
(Refer Slide Time: 00:59)

T HEOREM 1 (Factor Theorem for Analytic Functions). Let Ω ⊆ C be an open set and
f : Ω −→ C be a complex analytic function such that f (z 0 ) = 0 at z 0 ∈ Ω. Then there exists
a unique complex analytic function g : Ω −→ C such that f (z) = (z − z 0 )g (z).
1

242
2

1 f (z)
P ROOF. Since is holomorphic on Ω \ {z 0 } we have , is holomorphic on
z − z0 z − z0
Ω \ {z 0 }.
By the hypothesis, f is complex analytic on D(z 0 , r ), where D(z 0 , r ) ⊆ Ω, we have the
following power series expansion,
∞
a n (z − z 0 )n .
X
f (z) =
n=1

For z ∈ D(z 0 , r ), define

∞
a n+1 (z − z 0 )n .
X
g (z) :=
n=0

It is left as an exercise to the reader to verify that g converges in D(z 0 , r ).

On D(z 0 , r ), we have
µ ∞ ¶ ∞
n
a n (z − z 0 )n = f (z).
X X
(z − z 0 )g (z) = (z − z 0 ) a n+1 (z − z 0 ) =
n=0 n=1

(Refer Slide Time: 07:59)

243
 3

Hence on D(z 0 , r ), f (z) = (z − z 0 )g (z). Then, on D(z 0 , r ) \ {z 0 }, g (z) coincides with

f (z)
.
z − z0
f (z)
Since is complex analytic on Ω \ {z 0 }, g is complex analytic at every point in
z − z0
Ω \ {z 0 }. Also g has a power series expansion around z 0 by the definition.
∞
a n+1 (z − z 0 )n at z 0 .
P
Uniqueness of the function g follows from the continuity of
n=0
□

Let us prove one more consequence. This consequence is classically called as the
principle of analytic continuation.

T HEOREM 2 (Principle of Analytic Continuation). Let Ω ⊆ C be an open connected

subset and f , g : Ω −→ C be complex analytic on Ω. Suppose f and g agree on a non-
empty open subset of Ω. Then f ≡ g on Ω.

P ROOF. Let us define a subset E of Ω to be

E := z ∈ Ω : f n (z) = g n (z) ∀n ∈ N .
© ª

Since f and g agrees on an open set, E is non-empty. If we manage to prove that E is

both open and closed, then by the connectedness of Ω implies that E = Ω.
We shall prove first that E is closed. For n ∈ N, let E n = {z ∈ Ω : f n (z) = g n (z)}. Since
both f n and g n are continuous, E n is closed for each n ∈ N. Further E =
T
E n , which is
n∈N
closed.
Next we shall prove that E is open. Let z 0 ∈ E . Then f n (z 0 ) = g n (z 0 ) ∀ n ∈ N. Hence
f and g have the same power series expansion in D(z 0 , r ), where D(z 0 , r ) ⊆ Ω. Then for
w ∈ D(z 0 , r ), we have f n (w) = g n (w) ∀ n ∈ N. Hence D(z 0 , r ) ⊆ E .
Hence E is both open and closed. □

We can prove that the zero sets of a non-trivial holomorphic function will always be
isolated. In some sense, we are proving a stronger version of what we have just proved. If
two functions defined on a connected open set agree on a subset that has a limit point,

244
4

we will be able to show that those two functions are identical in the domain of definition.
This is stronger than the principle of analytic continuation in the sense that here we are
not demanding that these two functions should agree on an open set for being identical
but a condition weaker than that. This theorem is also sometimes called the identity
theorem.
(Refer Slide Time: 17:41)

T HEOREM 3 (Identity Theorem). Let Ω ⊆ C be an open connected set and f : Ω −→ C

be a holomorphic function which is not identically zero and such that f (z 0 ) = 0. Then
there exists ϵ > 0 such that f (z) ̸= 0 on D(z 0 , ϵ) \ {z 0 }.

P ROOF. If f n (z 0 ) = 0 for every n ∈ N, then f ≡ 0 on Ω by Theorem 2. Since by the

hypothesis f ̸≡ 0, ∃ m ∈ N such that f m (z 0 ) ̸= 0. Let n 0 be the smallest positive integer
such that f n0 (z 0 ) ̸= 0.

245
 5

By the choice of n 0 , we have f k (z 0 ) = 0 for every k < n 0 . Applying Theorem 1 itera-

tively, we have

f (z) = (z − z 0 )n0 g (z).

If g (z 0 ) = 0, then f n0 (z 0 ) = 0 which is a contradiction. Hence g (z 0 ) ̸= 0. By the continuity

of g , there exists a ϵ > 0 such that g (z) ̸= 0 for every z ∈ D(z 0 , ϵ). Also (z − z 0 ) does not
vanish on D(z 0 , r )\{z 0 }. Hence f (z) = (z −z 0 )n0 g (z) does not vanish on D(z 0 , r )\{z 0 }. □

E XAMPLE 4. Consider f (z) = sin2 (z) + cos2 (z). We know that sin2 (x) + cos2 (x) = 1 for
every x ∈ R. Since R × {0} is a closed subset of C, by Theorem 3, we have f (z) ≡ 1. That is
sin2 (z) + cos2 (z) = 1 on C.

T HEOREM 5 (Higher Order Cauchy Integral Formula). Let f : Ω −→ C be complex

analytic on an open set Ω ⊆ C an z 0 ∈ Ω with D(z 0 , r ) ⊆ Ω. Let γ be a closed curve in
Ω \ {z 0 } homotopic as closed curves to a reparametrization of γ1 where γ1 (t ) = z 0 + r e i t ,
t ∈ [0, 2π]. Then for each n ∈ N

n! f (z) n! f (z)
Z Z
n
f (z 0 ) = n+1
dz = d z.
2πi γ (z − z 0 ) 2πi γ1 (z − z 0 )n+1

P ROOF. Since f is complex analytic at z 0 and the coefficients a n in the power series
expansion of f is given by,

1 f (z)
Z
an = d z.
2πi γ1 (z − z 0 )n+1

(Refer Slide Time: 31:42)

246
6

f n (z 0 )
We know that a n = . Hence,
n!
n! f (z)
Z
n
f (z 0 ) = n+1
d z.
2πi γ (z − z 0 )

And as a consequence, we directly have the Cauchy estimates using the higher order
Cauchy integral formula. By this we will be able to place bounds on the derivatives of f
in Ω.

C OROLLARY 6 (Cauchy Estimates). Let Ω ⊆ C be an open subset and f : Ω −→ C be

a holomorphic function. Suppose z 0 ∈ Ω be such that D(z 0 , r ) ⊆ Ω for some r > 0. Let
γ(t ) = z 0 + r e i t for t ∈ [0, 2π]. Suppose | f (z)| ≤ M for every z ∈ γ([0, 2π]). Then for every
n ∈ N,
n!
f n (z 0 ) ≤ M .
rn

247
 7

P ROOF. By Theorem 5,
¯ ¯
¯ ¯ n! f (z)
Z
¯ n ¯
¯ f (z 0 )¯ = ¯ d z ¯
n+1
γ (z − z 0 )
¯ 2πi ¯

n! M n!
µ ¶
≤ n+1
2πr = M n .
2π r r
□

248
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 27
Further Consequences of Cauchy Integral Formula

In the last lecture we established inequalities on the higher order derivatives of a holo-
morphic function defined on a domain Ω. They were called the Cauchy estimates or the
Cauchy inequalities. And we will be using the Cauchy inequalities to establish now that a
non-constant holomorphic function defined on the complex plane cannot be bounded.
This result is called Liouville’s theorem.
(Refer Slide Time: 00:45)

T HEOREM 1 (Liouville’s Theorem). Let f be an entire function which is bounded.

Then f is a constant function.
1

249
2

P ROOF. Since f is bounded, there exists M > 0 such that f (z) ≤ M for every z ∈ C. Fix
z 0 ∈ C. For r > 0, let σr (t ) = z 0 + r e i t , t ∈ [0, 2π]. Then

| f (z)| < M ∀ z ∈ σr ([0, 2π]) and ∀ r > 0.

By Cauchy inequality,

M
| f ′ (z 0 )| ≤ .
r

Taking the limit as r → ∞, we get

f ′ (z 0 ) = 0.

Since the point z 0 was arbitrary, we have f ′ ≡ 0. By fundamental theorem of calculus, f

is a constant function. □

So bounded entire functions are necessarily constants. This is actually in stark con-
trast to what can be seen in the Real analysis setting. If we were to look at the real line,
we have functions like sine which is real analytic and bounded by 1. This cannot happen
in the Complex setting. The moment we have an entire function that is bounded, it is
forced to be a constant function. Sine function certainly has an extension to the complex
plane. The reader should convince themselves that sine function away from the real line
is not bounded.

T HEOREM 2 (Fundamental Theorem of Algebra). Let p(z) = a 0 + a 1 z + · · · + a n z n be

a non-constant polynomial with a j ∈ C ∀ 0 ≤ j ≤ n such that a n ̸= 0. Then there exist
z 1 , . . . , z n (not necessarily distinct) such that p(z) = a n (z − z 1 ) . . . (z − z n ).

P ROOF. We shall prove this theorem by using induction.

(Refer Slide Time: 08:26)

250
 3

For n = 1, the theorem follows directly. Assume that the result has been established
up to n − 1.
Let p(z) = a 0 + a 1 z +· · ·+ a n z n be a polynomial of degree n such that p does not have
1
a root in C, i.e., p(z) ̸= 0 for every z ∈ C. Then is a holomorphic function in C.
p(z)
For |z| > R > 1,

|p(z)| = ¯a n z n + a n−1 z n−1 + · · · + a 0 ¯

¯ ¯

≥ |a n ||z|n − ¯a n−1 z n−1 + · · · + a 0 ¯

¯ ¯

a 0 ¯¯
µ ¶
n 1 ¯¯
≥ |z| |a n | − ¯a n−1 + · · · + n−1 ¯ .
|z| z

251
4

Now let us focus on the second term in the RHS. For R large enough, we have
a 0 ¯¯ 1
µ ¶
1 ¯¯ |a 0 |
¯a n−1 + · · · + n−1 ¯ ≤ |a n−1 | + · · · + n−1
|z| z |z| |z |
1
≤ (|a n−1 | + · · · + |a 0 |)
|z|
< |a n |.

Hence
1 ¯¯ a 0 ¯¯
|a n | − ¯a n−1 + · · · + n−1 ¯ > M ′ > 0.
|z| z
That is, if we choose R > 1 large enough, we have

|p(z)| > |z|n M ′ > R n M ′ .

¯ ¯
¯ 1 ¯
Also, given M > 1, there exists R > 1 such that |p(z)| > M or ¯
¯ ¯ < M for every z with
p(z) ¯
|z| > R.
(Refer Slide Time: 17:07)

252
 5
µ ¶³
1 1 ´
Since D(0, R) is compact and is continuous, D(0, R) is compact. Hence,
p(z) p
¯ ¯
¯ 1 ¯
¯ p(z) ¯ < M 1 ∀ z ∈ D(0, R).
¯ ¯

Let M 2 = max{M , M 1 }. Then we have,

¯ ¯
¯ 1 ¯
¯ p(z) ¯ < M 2
¯ ¯ ∀ z ∈ C.

1
By Liouville’s theorem, = c, where c ∈ C and therefore p(z) is a constant function
p(z)
which is a contradiction.
Hence there exists at least one root z n ∈ C such that p(z n ) = 0. Then we have, p(z) =
q(z)(z − z n ), where q(z) = a n z n−1 + b n−2 z n−2 + · · · + b 0 . By induction,

q(z) = a n (z − z 1 ) . . . (z − z n−1 ).

Hence, p(z) = a n (z − z 1 ) . . . (z − z n−1 )(z − z n ). □

We will now prove a form of converse to the Cauchy’s theorem. Cauchy’s theorem
stated that, if you have a function f which is holomorphic on a given domain Ω and if
you have a closed path γ which is null homotopic, then γ f = 0. A form of converse
R

would be to demand that, if γ f = 0 for any closed curve γ, then our function f is holo-
R

morphic. So that is going to be the type of statement we will be proving. This theorem is
called Morera’s theorem.

T HEOREM 3 (Morera’s Theorem). Let Ω ⊆ C be an open set and f : Ω −→ C be a con-

tinuous function such that
Z
f (z)d z = 0
γ

for every closed polygonal path γ in Ω. Then f is holomorphic on Ω.

P ROOF. Let z 0 ∈ Ω and r > 0 be such that D(z 0 , r ) ⊆ Ω.

(Refer Slide Time: 23:10)

253
6

Now any closed polygonal path γ in D(z 0 , r ) is a closed polygonal path in Ω and
hence
Z
f (z)d z = 0.
γ

By second fundamental theorem of calculus, there exists an anti-derivative F on D(z 0 , r )

which is holomorphic. Since F is complex analytic, F ′ is holomorphic on D(z 0 , r ), i.e., f
is holomorphic on D(z 0 , r ). Hence f is holomorphic on Ω. □

Morera’s theorem help us to conclude, if a sequence of holomorphic functions con-

verges uniformly on compact sets to a function, then the limit is also holomorphic.

T HEOREM 4. Let Ω ⊆ C be open and f n : Ω −→ C be holomorphic on Ω for each n ∈ N

such that f n converges uniformly on compact sets to a function f . Then f is holomorphic.

P ROOF. Let z 0 ∈ Ω and r > 0 be such that D(z 0 , r ) ⊆ Ω. Since f n converges to f uni-
formly on compact sets in Ω, we have f is continuous on Ω. In particular, f n converge

254
 7

to f uniformly on compact sets in D(z 0 , r ). Let σ : [0, 1] −→ C be a closed polygonal path

in D(z 0 , r ). Since σ([0, 1]) is compact, f n converges to f uniformly on σ([0, 1]). Then we
have,
Z Z Z
lim f n (z)d z = lim f n (z)d z = f (z)d z.
n→∞ σ σ n→∞ σ
But
Z
f n (z)d z = 0 ∀ n ∈ N,
σ
hence
Z
f (z)d z = 0.
σ
By Morera’s theorem, we have f is holomorphic on Ω. □

(Refer Slide Time: 31:30)

Before we conclude, let us summarize what are the things that we have concluded
about complex valued functions on domains in C till now. Let f be a function such that
f : (a, b) −→ R. If f is continuous then it belongs to a broader class of functions. If we put

255
8

more and more regularity to the function f , the size of set to which it belong will become
’small’. That is, f can be k−times differentiable but it is not k + 1−times differentiable.
Then we can put f in a subclass of continuous functions called differentiable functions.
If f has a regularity of C ∞ −smooth, then it will fall into the subclass of differentiable
functions called the smooth functions. We have a further subclass called analytic func-
tions. There are functions which are smooth but not analytic. For example, consider the
function,

−1
e x

if x > 0
f (x) =
if x ≤ 0.

0

Then f is smooth but it is not analytic. The reader should verify the details.
But if we look into the functions on an open subset Ω of C, we however do not have
as many classifications as we have in the real analysis setting. There will certainly be
the broad class of continuous functions, but the moment we go down to a subclass of
complex differentiable functions on Ω that is going to be the same as complex analytic
functions or holomorphic functions. But this is the same as complex analytic which we
have proved.
Of course, there is one more interesting class of functions; the set of all those func-
tions, f : Ω −→ C such that γ f = 0 where γ is a closed curve. Now, notice that this is
R

actually a special class of functions because if we consider Ω = C\{0} and f (z) = 1z . Then
f is holomorphic on Ω but σ f (z) = 2πi ̸= 0, where σ(t ) = e i t for t ∈ [0, 2π].
R

256
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 28
Problem Session

(Refer Slide Time: 00:23)

P ROBLEM 1. Evaluate
e z − e −z
Z
dz
γ zn
where n ∈ N and γ(t ) = e i t , t ∈ [0, 2π].

S OLUTION 1. Let f (z) = e z − e −z . Then f is an entire function. By higher order

Cauchy integral formula we have,
(n − 1)! f (z)
Z
n−1
f (0) = d z.
2πi γ zn
1

257
2

Hence,

e z − e −z 2πi f n−1 (0)

Z
d z = .
γ zn (n − 1)!

Now it is an easy check to verify that,


e z − e −z

if k is even
f k (z) =
e z + e −z

if k is odd.

Therefore,

e z − e −z 2πi f n−1 (0)

Z
d z =
γ zn (n − 1)!

 4πi

 if n is even
= (n − 1)!

0
 if n is odd.

P ROBLEM 2. Evaluate

z2 + 1
Z
2
dz
γ z(z + 4)

where γ(t ) = r e i t , t ∈ [0, 2π], when 0 < r < 2 and r > 2.

(Refer Slide Time: 07:09)

258
 3

S OLUTION 2. We may spilt the integrand by using the partial fractions.

z2 + 1 a b c
2
= + + .
z(z + 4) z (z + 2i ) z − 2i

By our routine calculation, we will get the values of the coefficients a, b and c as,
1 3
a = ,b = c = .
4 8
Now we have,

z2 + 1 1 dz 3 dz 3 dz
Z Z Z Z
2
dz = + + .
γ z(z + 4) 4 γ z 8 γ z + 2i 8 γ z − 2i

First we shall consider the case when 0 < r < 2 and we have γ(t ) = r e i t , t ∈ [0, 2π].
1
We know that is holomorphic on C \ {−2i }. Also, γ is null-homotopic on C \
z + 2i
{−2i }. By Cauchy’s theorem,
dz
Z
= 0.
γ z + 2i

259
4

Similarly,
dz
Z
= 0.
γ z − 2i

Hence, when 0 < r < 2, we have,

z2 + 1 1 d z πi
Z Z
2
dz = = .
γ z(z + 4) 4 γ z 2
Now, let us consider the case when r > 2.
(Refer Slide Time: 14:49)

Let ϵ > 0 be small enough such that D(2i , ϵ) ⊆ D(0, r ) and let γ1 (t ) = 2i + ϵe i t , t ∈
[0, 2π]. Now γ is homotopic γ1 as closed curves by the straight line homotopy,

H (s, t ) = (1 − s)γ(t ) + sγ1 (t ), (s, t ) ∈ [0, 1] × [0, 2π].

Hence by Cauchy’s theorem, we have

1 dz 1 dz
Z Z
=
2πi γ z − 2i 2πi γ1 z − 2i

260
 5

and by the Cauchy integral formula,

3 dz 3 1 dz 3πi
Z Z
= · 2πi · = .
8 γ z − 2i 8 2πi γ1 z − 2i 4
Similarly, we have
3 dz 3πi
Z
=
8 γ z + 2i 4
and
1 d z πi
Z
= .
4 γ z 2
Hence,
z2 + 1 πi 3πi
Z
2
dz = + = 2πi .
γ z(z + 4) 2 2

P ROBLEM 3. Let f be a bounded entire function and a, b ∈ C be two distinct complex

numbers. Let R > max(|a|, |b|). Then evaluate,
f (z)
Z
IR = dz
γ (z − a)(z − b)

where γ(t ) = Re i t , t ∈ [0, 2π]. Also evaluate lim I R .

R→∞

S OLUTION 3.
f (z)
Z
IR = dz
γ (z − a)(z − b)
f (z) f (z)
µ ¶
1
Z
= − dz
γ (a − b) (z − a) (z − b)
1 f (z) 1 f (z)
Z Z
= dz − dz
(a − b) γ z − a (a − b) γ z − b
2πi 2πi
= f (a) − f (b)
(a − b) (a − b)
f (a) − f (b)
µ ¶
I R = 2πi .
a −b
Notice that f is bounded, thus there exists an M > 0 such that | f (z)| < M for every z ∈ C.
Since, |z − a| ≥ |z| − |a| ≥ R − |a| and |z − b| ≥ R − |b|,
¯ ¯
¯ f (z) ¯ M
¯ (z − a)(z − b) ¯ ≤ (R − |a|)(R − |b|)
¯ ¯

261
6

and hence

M
|I R | ≤ 2πR.
(R − |a|)(R − |b|)

Now,

M
lim |I R | ≤ lim 2πR = 0.
R→∞ R→∞ (R − |a|)(R − |b|)

f (a) − f (b)
µ ¶
Hence lim I R = 0 =⇒ 2πi = 0 =⇒ f (a) = f (b), i.e., f is a constant.
R→∞ a −b

(Refer Slide Time: 28:11)

P ROBLEM 4. Let f be an entire function such that | f ′ (z)| < | f (z)| for every z ∈ C. Then
there exists a positive real number a such that

| f (z)| ≤ ae |z| .

262
 7

1
S OLUTION 4. Since 0 ≤ | f ′ (z)| < | f (z)|, we have f is does not vanish on C. Hence is
f
f ′ (z)
an entire function. Let g (z) = . Then g is holomorphic on C and also |g (z)| < 1 for
f (z)
every z ∈ C. By Liouville’s theorem, g ≡ c for some c ∈ C, i.e., f ′ (z) = c f (z).
We know that f is an entire function, hence there exists a power series expansion,

∞
an z n
X
f (z) =
n=0

in C. Also we know that,

∞
f ′ (z) = na n z n−1 .
X
n=1

Since f ′ (z) = c f (z), we have

∞ ∞
na n z n−1 = ca n z n .
X X
n=1 n=0

Hence
c k a0
ca 0 = a 1 , ca 1 = 2a 2 , . . . , ca k = (k + 1)!a k+1 =⇒ a k = .
k!

Then we have,
∞ a cn zn ∞ (cz)n
0
= a0 · e c z .
X X
f (z) = = a0
n=0 n! n=0 n!

P ROBLEM 5. Suppose f is an entire function such that | f (z)| ≤ a + b|z|k for every
z ∈ C, where a, b, k ∈ N. Then f is a polynomial.

S OLUTION 5. Let z 0 ∈ C and γ : [0, 2π] −→ C be a curve given by γ(t ) = z 0 + Re i t .

Now by the higher order Cauchy integral formula,
¯ ¯ ¯¯ (k + 1)! Z f (z) ¯¯
¯
¯ (k+1)
(1) ¯f (z 0 )¯ = ¯ .
¯ ¯
2πi k+2 ¯
γ (z − z 0 )

Since | f (z)| ≤ a + b|z|k , there must exists positive integers a ′ and b ′ such that | f (z)| ≤
a ′ + b ′ |z − z 0 |k for every z ∈ C.

263
8

Then in (1), we have

f (z) ¯¯ (k + 1)! (a ′ + b ′ R k )
¯ ¯
¯ (k + 1)!
Z
¯
k+2 ¯
≤ 2πR
¯ 2πi
γ (z − z 0 ) 2π R k+2
µ ′
a b′
¶
= (k + 1)! k+1 + .
R R

(Refer Slide Time: 37:35)

Hence, if R → ∞, we have | f k+1 (z 0 )| = 0 =⇒ f k+1 (z) = 0 for every z ∈ C. Thus f ℓ ≡ 0

for every ℓ ≥ k + 1. Then,
∞ k
an z n = ak z k
X X
f (z) =
n=0 n=0

which is a polynomial.

P ROBLEM 6. Does there exists a holomorphic function on D, the unit disc, such that
f (z n ) = 0 where {z n } is a countable set in D consisting of distinct points.

264
 9
µ ¶
1 1
S OLUTION 6. Consider z n = 1 − and the function defined by f (z) = sin .
nπ 1−z
Then, Ã !
1
f (z n ) = sin ¡ 1
¢ = sin(nπ) = 0.
1 − 1 − nπ

265
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 29
Winding Number

In the version of the Cauchy integral formula that we have proved, we have imposed a
strict condition on the curve in the question. We demanded that in C \ {z 0 }, the curve is
homotopic as closed curve to a circle centered at z 0 with radius r , for some r > 0. This
is a bit stringent condition to impose on the curve and we would like to circumvent this
condition. The route to getting a more general Cauchy integral formula is by introducing
the notion of the winding number of a curve.
(Refer Slide Time: 00:54)

266
2

D EFINITION 1 (Winding Number). Let γ : [a, b] −→ C be a closed curve and let z 0 ∈

C \ γ([a, b]). We define the winding number of γ around z 0 to be

1 dz
Z
Wγ (z 0 ) := .
2πi γ z − z0

(Refer Slide Time: 02:33)

E XAMPLE 1. Let γ0 : [0, 2π] −→ C be given by γ0 (t ) = z 0 + r e i t for some r > 0. Then by

Cauchy integral formula, we have

1 dz
Z
Wγ0 (z 0 ) = = 1.
2πi γ z − z0

More generally, let γ1 : [0, 2π] −→ C be given by γ1 (t ) = z 0 + r e i mt for some r > 0 and
m ∈ N. Then we have Wγ1 (z 0 ) = m.

267
 3

Let γ and σ be two closed curves in C \ {z 0 } such that γ is homotopic as closed curves to
a reparametrization of σ in C \ {z 0 }. Then by Cauchy’s theorem, we have
dz dz
Z Z
= .
γ z − z0 σ z − z0

1
Since is a holomorphic function on C \ {z 0 },
z − z0
1 dz 1 dz
Z Z
Wγ (z 0 ) = = = Wσ (z 0 ).
2πi γ z − z 0 2πi σ z − z 0

P ROPOSITION 2. Let γ0 : [a, b] −→ C be a closed curve and z 0 be a point not in γ0 .

Suppose γ1 : [a, b] −→ C be a closed curve such that

¯γ0 (t ) − γ1 (t )¯ < ¯γ0 (t ) − z 0 ¯ .

¯ ¯ ¯ ¯

Then Wγ0 (z 0 ) = Wγ1 (z 0 ).

P ROOF. Define H : [0, 1] × [a, b] −→ C by

H (s, t ) = (1 − s)γ0 (t ) + sγ1 (t ).

Then H is continuous function into C and notice that at each point s ∈ [0, 1], γs : [a, b] −→
C given by γs (t ) = H (s, t ) is a closed curve. Hence H is a homotopy from γ0 to γ1 .
Claim: z 0 ̸∈ H ([0, 1] × [a, b]).

¯ ¯
|H (s, t ) − z 0 | = ¯(1 − s)γ0 (t ) + sγ1 (t ) − z 0 ¯

= ¯s γ1 (t ) − γ0 (t ) + γ0 (t ) − z 0 ¯
¯ ¡ ¢ ¯

≥ ¯γ0 (t ) − z 0 ¯ − s ¯γ1 (t ) − γ0 (t )¯
¯ ¯ ¯ ¯

≥ ¯γ0 (t ) − z 0 ¯ − ¯γ1 (t ) − γ0 (t )¯
¯ ¯ ¯ ¯

> 0.

Hence H (s, t ) ̸= z 0 for any (s, t ) ∈ [0, 1] × [a, b]. Therefore H is a homotopy of closed
curves in C \ {z 0 } from γ0 to γ1 =⇒ Wγ0 (z 0 ) = Wγ1 (z 0 ). □

268
4

(Refer Slide Time: 14:21)

Let γ : [a, b] −→ C be a closed curve and z 0 be a point in C such that

dist(γ, z 0 ) > diam(γ),

where dist(γ, z 0 ) := inf ¯γ(t ) − z 0 ¯ and diam(γ) := sup ¯γ(t ) − γ(t ′ )¯.
¯ ¯ ¯ ¯
t ∈[a,b] t ,t ′ ∈[a,b]
Let us see what happens when we have such a scenario. Define γ1 to a constant
curve as γ1 (t ) = z = γ(a) for every t ∈ [a, b]. Then,

¯γ(t ) − γ1 (t )¯ < diam(γ) < dist(γ, z 0 ) ≤ ¯γ(t ) − z 0 ¯ .

¯ ¯ ¯ ¯

Hence by Proposition 2, we have Wγ (z 0 ) = Wγ1 (z 0 ). But since γ1 is a constant curve,

Wγ1 (z 0 ) = 0 = Wγ (z 0 ). Geometrically it tells that, if z 0 is a point on the complex plane
which is far away from the curve γ, then the curve does not wind around z 0 , as to be
expected.

269
 5

P ROPOSITION 3. Let γ : [a, b] −→ C be a closed curve and z 0 be a point not in γ. Then

there exists r > 0 such that for z ∈ D(z 0 , r ), Wγ (z) = Wγ (z 0 ).

P ROOF. Since γ([a, b]) is a closed subset of C and z 0 ̸∈ γ([a, b]), there exists r > 0 such
that D(z 0 , r ) ∩ γ([a, b]) = ∅. Let h ∈ C be such that |h| < r . Define γh (t ) = γ(t ) − h for
t ∈ [a, b]. Then,
¯γh (t ) − γ(t )¯ = |h| < r < ¯γ(t ) − z 0 ¯ .
¯ ¯ ¯ ¯

Hence

(1) Wγh (z 0 ) = Wγ (z 0 ).

Now it is left as an exercise to the reader to verify the following equality,

1 dz 1 dz
Z Z
Wγh (z 0 ) = = = Wγ (z 0 + h).
2πi γh z − z 0 2πi γ z − (z 0 + h)

(Refer Slide Time: 22:29)

270
6

From (1), for each h with |h| < r , we have

Wγ (z 0 ) = Wγ (z 0 + h).

Hence Wγ is constant on D(z 0 , r ). □

T HEOREM 4. Let γ : [a, b] −→ C be a closed curve and z 0 be a point not in the image of
γ. Then Wγ (z 0 ) is an integer.

P ROOF. Pick a partition P : a = t 1 < t 2 < · · · < t n = b of [a, b] such that γ ↾[t j −t j +1 ]
is homotopic with fixed end points to γz j →z j +1 in C \ {z 0 } where z j = γ(t j ). Then γ is
homotopic to γz1 →z2 →···→zn →z1 as closed curves. Since Wγ (z 0 ) = Wγz1 →z2 →···→zn →z1 (z 0 ), we
may start with the assumption that γ is a closed polygonal path.
Let us now prove the result by induction. Let z 1 , . . . , z n be as above. Suppose z 0 ∈
γzn−1 →z1 . Since Wγ (z) = Wγ (z 0 ) for z ∈ D(z 0 , r ), with r as described in Proposition 3, pick
a point z ′ ∈ D(z 0 , r ) and z ′ ̸∈ γzn−1 →z1 .
The induction is on the number of points {z 1 , . . . , z n }. The case where n = 1 and n = 2
are trivial, since the polygonal path with one or two points does not form a closed curve.
When n = 3, γ = γz1 →z2 →z3 →z1 will be a triangle, T . Let T̂ denote the convex hull of the
triangle T . Then T̂˚ denote the interior of T̂ . If z ∈ T̂˚ , then by Cauchy integral formula,
0

1 dz
Z
= 1.
2πi γ z − z0

If z 0 ∈ C \ T̂ , then γ is null-homotopic in C \ {z 0 } and hence

1 dz
Z
= 0.
2πi γ z − z0

Hence Wγ (z 0 ) is an integer.
Assume the result to be proved for up to n − 1.

271
 7

Let γ = γz1 →z2 →···→zn →z1 . Now we have

Wγ (z ′ ) = Wγz1 →z2 ···→zn−1 →z1 (z ′ ) + Wγz1 →zn−1 →zn →z1 (z ′ )

1 dz 1 dz
Z Z
= ′
+
2πi γz1 →z2 ···→zn−1 →z1 z − z 2πi γz1 →zn−1 →zn →z1 z − z ′
Now by the induction hypothesis, first and second terms on the R.H.S are integers. Hence
Wγ (z ′ ) is an integer which implies Wγ (z 0 ) is also an integer. □

T HEOREM 5 (General Cauchy Integral Formula). Let f : Ω −→ C be holomorphic

on open set Ω and γ : [a, b] −→ Ω be a closed curve which is null-homotopic. Then for
z 0 ̸∈ γ([a, b]), we have
1 f (z)
Z
f (z 0 )Wγ (z 0 ) = .
2πi γ (z − z 0 )

P ROOF. By factorization theorem for holomorphic functions,

f (z) − f (z 0 ) = (z − z 0 )g (z)

and hence,
1 f (z) − f (z 0 ) 1
Z Z
= g (z).
2πi γ (z − z 0 ) 2πi γ
Since γ is null homotopic on Ω, by Cauchy’s theorem,
1
Z
g (z) = 0.
2πi γ
Therefore,
1 f (z) 1 f (z 0 )
Z Z
= = f (z 0 )Wγ (z 0 ).
2πi γ (z − z 0 ) 2πi γ (z − z 0 )
□

272
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 30
Open Mapping Theorem

In this lecture we will explore the local behavior of a non-constant holomorphic function
f defined on an open connected set Ω. We will prove that given a point z 0 ∈ Ω there exists
a neighborhood U of z 0 such that in the punctured neighborhood U \ {z 0 }, our function
f is an m to 1 mapping for some m ∈ N.
We will come to that. But, before that, let us prove a special case of the inverse func-
tion theorem which states that, if the derivative of the function f does not vanish at a
point z 0 , then the function f is locally invertible in a neighborhood of z 0 and further we
have a holomorphic inverse.
Let us begin this lecture by proving a preparatory lemma which is needed to prove
the inverse function theorem.
(Refer Slide Time: 00:59)

273
2

L EMMA 1. Let f : Ω −→ C be a holomorphic function on an open set Ω ⊆ C. Then

define G : Ω × Ω −→ C given by
f (z) − f (w)


 when z ̸= w
G(z, w) := z −w
 f ′ (z)

when z = w.

Then G is continuous on Ω × Ω.

P ROOF. Note that on the points (z, w) ∈ Ω × Ω where z ̸= w, continuity of G is trivial.

We shall prove the continuity of G on the points z = w. Given z 0 ∈ Ω, since f ′
is continuous, if given an ϵ > 0, we have an r > 0 such that for z ∈ D(z 0 , r ), we have
¯ f (z) − f (z 0 )¯ < ϵ.
¯ ¯

Pick z, w ∈ D(z 0 , r ) where z ̸= w. Let γ(t ) = (1 − t )z + t w for t ∈ [0, 1]. Notice that
γ′ (t ) = w − z. Then,
Z 1 1
Z 1
′
γ(t ) d t = f ′ γ(t ) γ′ (t )d t
¡ ¢ ¡ ¢
f
0 w −z 0
1
Z
= f ′ (z)d z
w −z γ
f (w) − f (z)
=
w −z
= G(z, w).

Consider
¯Z 1 Z 1 ¯
′ ′
f γ(t ) d t −
¯ ¡ ¢ ¯
|G(z, w) −G(z 0 , z 0 )| = ¯¯ f (z 0 )d t ¯¯
0 0
¯Z 1 ¯
¯ ¡ ′¡ ′
f γ(t ) − f (z 0 ) d t ¯¯
¢ ¢ ¯
=¯¯
0

≤ ϵ.

Hence G is continuous at (z 0 , z 0 ) ∈ Ω × Ω. □

T HEOREM 2. Let Ω ⊆ C be an open set and f : Ω −→ C be a holomorphic function.

Suppose z 0 ∈ Ω be such that f ′ (z 0 ) ̸= 0. Then there exists a neighborhood U of z 0 in Ω such

274
 3

that f ↾U is injective. Furthermore, V = f (U ) is an open set and the inverse g : V −→ U of

f is holomorphic.

| f ′ (z 0 )|
P ROOF. Since f ′ (z 0 ) ̸= 0, let ϵ = . By Lemma 1, there exists a neighborhood U
2
of z 0 such that for z, w ∈ U and z ̸= w, we have
¯ ¯¯ ′ ¯
f (z 0 )¯
¯
¯ f (z) − f (w) ′
¯
¯
¯ z −w − f (z 0 )¯¯ <
2
¯ ′ ¯
¯ f (z 0 )¯ ¯ ¯ ¯
¯ ¯ f (z) − f (w) ¯
′
> ¯ f (z 0 )¯ − ¯¯ ¯
2 z −w ¯

¯ ′ ¯
¯ f (z 0 )¯
(1) =⇒ | f (z) − f (w)| > |z − w|.
2

If z ̸= w we have | f (z) − f (w)| > 0. That is f (z) ̸= f (w). Hence f ↾ U is injective.

(Refer Slide Time: 14:33)

Next we want to show that V , the image of U under f , is an open set. Let a ∈ U . Then
there exists s > 0 such that D(a, s) ⊆ U . Then by (1), for every θ ∈ [0, 2π],
¯ ³ ´ ¯ | f ′ (z )|s
¯ f a + se i θ − f (a)¯ >
0
=c
¯ ¯
2

275
4

where c is a positive constant. That is f (a + se i θ ) is outside the disc of radius c around

f (a) for every θ ∈ [0, 2π].
Let w 0 ∈ D f (a), 2c . Then for every θ ∈ [0, 2π], we have,
¡ ¢

¯ ³ ´¯ c
¯ f (w 0 ) − f a + se i θ ¯ > .
¯ ¯
2

Suppose there does not exists z ∈ U such that f (z) = w 0 . Then consider f (z)− w 0 , which
1
is holomorphic and non-vanishing on U . Hence g (z) = is holomorphic on U .
f (z) − w 0
By maximum principle applied to g and compact set D(a, s),

¯ ³ ´¯
|g (z)| ≤ sup ¯g a + se i θ ¯ for all z ∈ D(a, s).
¯ ¯
θ∈[0,2π]

Hence,

1 1
≤ inf ¯ ¡
| f (z) − w 0 | θ∈[0,2π] ¯ f a + se i θ − w 0 ¯
¢ ¯
¯ ³ ´ ¯
=⇒ inf ¯ f a + se i θ − w 0 ¯ ≤ | f (z) − w 0 |
¯ ¯
θ∈[0,2π]

¢¯ c
Since for each θ ∈ [0, 2π], ¯ f (w 0 ) − f a + se i θ ¯ > , we have,
¯ ¡
2

¯ ³ ´¯ c
inf ¯ f (w 0 ) − f a + se i θ ¯ ≥
¯ ¯
θ∈[0,2π] 2

Then,

c
| f (a) − w 0 | ≥
2

which is a contradiction to our choice of w 0 ∈ D f (a), 2c . Hence f (a) is an interior point.

¡ ¢

(Refer Slide Time: 25:22)

276
 5

Since a ∈ U was arbitrary, V = f (U ) is an open set. Therefore f ↾U : U −→ V is bijec-

tive. Let g : V −→ U be its inverse. Now it remains to show that this map g is holomor-
phic on V .
Let w ′ ∈ V . For w ∈ V and w ̸= w ′ , let z, z ′ ∈ U be such that f (z) = w, f (z ′ ) = w ′ ,
g (w) − g (w ′ ) z − z′ 1
= =µ ¶.
w −w ′ ′
f (z) − f (z ) f (z) − f (z ′ )
z − z′
Claim: If w → w ′ , then z → z ′ where f (z) = w.
By (1),
¯ ′ ¯
¯ f (z 0 )¯
′
| f (z) − f (z )| > |z − z ′ |
2
|z − z ′ | < C |w − w ′ |

2
where C = . Then the claim is verified, if w → w ′ , then z → z ′ .
| f (z 0 )|
Thus,
g (w) − g (w ′ ) 1 1
lim ′ ′
= lim′ µ ′ ¶= ′ ′ .
w→w w −w z→z f (z) − f (z ) f (z )
w∈V \{w ′ } z∈U \{z ′ } ′
z −z
Hence g is holomorphic on V . □

277
6

E XAMPLE 3. Consider the map πm : C −→ C given by πm (z) = z m where m is a positive

integer. Then, for z 0 ̸= 0, π′m (z 0 ) ̸= 0. Then by Theorem 2, πm is a local biholomorphic
function, i.e, πm is bijective holomorphic function on a neighborhood of each non-zero
point with a holomorphic inverse function.
In particular, If V is an open set which does not contain the origin, then U = πm (V )
is also an open set. ³ ´
θ 2πk
iθ m 1/m i m+ m
Recall that if w = r e ̸= 0, then z = w has roots given by z k = r e ,k=
0, 1 . . . , m − 1. Hence πm (D(0, r )) = D(0, r m ).

T HEOREM 4 (Local Behaviour of Holomorphic Functions). Let f : Ω :−→ C be a non-

constant holomorphic function on an open connected set Ω . Let z 0 ∈ Ω and w 0 = f (z 0 ).
Then there exists a neighborhood U of z 0 and bijective holomorphic function φ on U such
¢m
that f (z) = w 0 + φ(z) for z ∈ U and some integer m > 0. Moreover φ maps U onto
¡

D(0, r ) for some r > 0.

(Refer Slide Time: 38:51)

278
 7

P ROOF. Since f (z 0 ) = w 0 , we know that f (z) − w 0 vanishes at z 0 . By using the factor-

ization theorem,
f (z) − w 0 = (z − z 0 )m g (z)

for some m > 0 and such that g (z 0 ) ̸= 0.

Let s > 0 be such that g (z) ̸= 0 on D(z 0 , s). Since D(z 0 , s) is simply connected and g
does not vanish on D(z 0 , s), there exists a holomorphic function h on D(z 0 , s) such that
g (z) = e h(z) on D(z 0 , s).
Define φ : D(z 0 , s) −→ C to be
h(z)
φ(z) = (z − z 0 )e m .

Then
¢m
φ(z) = (z − z 0 )m e h(z) = (z − z 0 )m g (z) = f (z) − w 0 .
¡

Also note that φ′ (z 0 ) ̸= 0. Let U ′ be a neighborhood of z 0 in D(z 0 , s) such that φ ↾U ′ is

injective and φ(U ′ ) is an open set. Since φ(z 0 ) = 0, there exists r > 0 such that D(0, r ) ⊆
φ(U ′ ). Define U = φ−1 (D(0, r )).
On U , we have
¢m
f (z) − w 0 = φ(z) .
¡

Also φ is bijective and holomorphic on U with φ(U ) = D(0, r ). □

Notice that f (z) = w 0 + πm ◦ φ (z), hence f is an m to 1 mapping on U \ {z 0 }.

¡ ¢

T HEOREM 5 (Open Mapping Theorem). Let f : Ω −→ C be a non-constant holomor-

phic function on an open connected set Ω. Then f (Ω) is an open set.

P ROOF. Let z 0 ∈ Ω and w 0 = f (z 0 ). By Theorem 4, there exists a neighborhood U of

¡ ¢m
z 0 such that f (z) = w 0 + φ on U and such that φ(U ) = D(0, r ). Hence πm ◦ φ(U ) =
D(0, r m ). Then f (U ) = D(w 0 , r m ) =⇒ w 0 is an interior point of f (Ω). Therefore f (Ω) is
open. □

279
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 31
Schwarz Reflection Principle

In Real Analysis we have seen many instances where we had to extend our given function
which is defined on an interval to a larger interval. So a function f defined on I , an
interval on R had to be extended to a function on J , a larger interval containing I . In
fact, if we start off with a continuous function we would have demanded many times
to get hold of an extension which was also continuous. If we started off with a smooth
function on I , the requirement was that we get a smooth extension to J . So the regularity
had to be preserved.
In this lecture, we will address a similar problem for holomorphic functions. Holo-
morphic functions, as we have already seen by now, is a very rigid class of functions. It is
not very easy to answer the question of extension many times. There would not be an ex-
1
tension possible at all. For example, on the unit disk, if we take the function f (z) = z−1 ,

then it does not extend past the boundary at all.

Here we will prove a theorem called the Schwarz Reflection Principle which tells us
certain special circumstances when we will be able to talk about the extension of a holo-
morphic function. Let us try to set up the circumstances where we can indeed talk about
such an extension.
(Refer Slide Time: 01:28)
1

280
2

Let Ω be an open connected set. Define Ω := z ∈ C : z ∈ Ω . Suppose f : Ω −→ C be a

© ª

holomorphic function. Then define f ∗ : Ω −→ C by f ∗ (z) = f (z).

E XERCISE 1. f ∗ is holomorphic on Ω.

We say that Ω is symmetric with respect to R if Ω = Ω.

E XERCISE 2. If Ω is symmetric with respect to R, then Ω intersects the real line.

We shall now focus on open connected sets Ω which are symmetric with respect to
R. Let f be a holomorphic function on Ω. Since Ω = Ω, the domain of definition of both
f and f ∗ are the same and are holomorphic on Ω.
Define g : Ω −→ C given by

g (z) = f (z) − f ∗ (z).

Then g is holomorphic on Ω.
Now let us look into the case that our function f satisfies the condition that, f (x) ∈ R
whenever x ∈ R. Then, for x ∈ Ω ∩ R, we have f (x) = f ∗ (x). Hence g (x) = 0. Since
Ω is open, there exist a, b ∈ R such that [a, b] ⊆ Ω and g (x) = 0 for x ∈ [a, b]. But by

281
 3

the identity theorem, we know that zeroes of a non-constant holomorphic function are
isolated. Hence g ≡ 0 on Ω. That is, f (z) = f ∗ (z) on Ω.
(Refer Slide Time: 08:52)

We will be using the following notations:

Let Ω be an open connected set which is symmetric with respect to R. Then

Ω+ = {z ∈ Ω : Im(z) > 0}

Ω− = {z ∈ Ω : Im(z) < 0}

I = {z ∈ Ω : Im(z) = 0} .

Hence Ω = Ω+ ∪ I ∪ Ω− .

T HEOREM 3 (Schwarz Reflection Principle). Let Ω be symmetric with respect to R.

Suppose f : Ω+ ∪ I −→ C be a function which is continuous on Ω+ ∪ I and holomorphic on
Ω+ . Suppose f (x) ∈ R for every x ∈ I , then there exists g : Ω −→ C holomorphic on Ω such
that g (z) = f (z) for each z ∈ Ω+ ∪ I .

282
4

P ROOF. Define the function g as,


 f ∗ (z) for z ∈ Ω−

g (z) =
 f (z) for z ∈ Ω+ ∪ I .


Now it is left as an exercise for the reader to check that the function g is continuous on
Ω.
We shall prove that g is holomorphic on Ω. Since f is holomorphic on Ω+ , it is easy
to verify that g is holomorphic on Ω+ ∪ Ω− . Hence to establish that g is holomorphic on
Ω, it is enough to check complex differentiability at points on I .
Let z 0 ∈ I and let r > 0 be such that D(z 0 , r ) ⊆ Ω. One can verify that by Morera’s
R
theorem, for proving g is holomorphic, it is enough to check T g = 0 for every triangle
T in D(z 0 , r ).
(Refer Slide Time: 17:25)

For this, it is enough to show that

Z
g =0 for T ⊆ Ω+ ∪ I or T ⊆ Ω− ∪ I .
T

283
 5

We will establish this statement by proving that

Z
g =0
T

where T is the triangular path γz1 →z2 →z3 →z1 where z 1 , z 2 , z 3 ∈ D(z 0 , r ) such that z 2 , z 3 ∈ I .
We know that the convex hull T̂ is a compact set and g is uniformly continuous on T̂ .
Let M = sup |g (z)| and d = sup |z − w|. By uniform continuity, given an ϵ > 0, there
z∈T̂ z,w∈T
exists ϵ > δ > 0 such that |g (z) − g (w)| < ϵ whenever |z − w| < δ on T̂ . Pick w 2 and w 3 on
γz1 →z2 and γz1 →z3 respectively such that |w 2 − z 2 | < δ and |w 3 − z 3 | < δ. Then
Z Z Z
g (z)d z = g (z)d z + g (z)d z.
T γz1 →w 2 →w 3 →z1 γw 3 →w 2 →z2 →z3 →w 3

But by Cauchy’s theorem,

Z
g (z)d z = 0
γz1 →w 2 →w 3 →z1

since γz1 →w 2 →w 3 →z1 ⊂ Ω+ . Hence

Z Z
g (z)d z = g (z)d z
T γw 3 →w 2 →z2 →z3 →w 3

Z Z Z Z Z
(1) g (z)d z = g (z)d z + g (z)d z + g (z)d z + g (z)d z.
T γz2 →z3 γz3 →w 3 γw 3 →w 2 γw 2 →z2

Now let us consider first and third terms of (1).

Z Z Z Z
(2) g (z)d z + g (z)d z = g (z)d z − g (z)d z.
γz2 →z3 γw 3 →w 2 γz2 →z3 γw 2 →w 3

Since γz2 →z3 = (1 − t )z 2 + t z 3 and γw 2 →w 3 = (1 − t )w 2 + t w 3 . Then, (2) can be written as,

Z Z Z 1
g (z)d z + g (z)d z = (z 3 − z 2 ) g ((1 − t )z 2 + t z 3 ) d t
γz2 →z3 γw 3 →w 2 0
Z 1
− (w 3 − w 2 ) g ((1 − t )w 2 + t w 3 ) d t
0

284
6

Now if |((1 − t )z 2 + t z 3 ) − ((1 − t )w 2 + t w 3 )| < δ, we have,

¯Z Z ¯ ¯ Z 1 ¯
¯ ¯ ¯ ¯
g (z)d z − g (z)d z ¯ ≤ ¯(z 3 − z 2 ) g ((1 − t )z 2 + t z 3 ) − g ((1 − t )w 2 + t w 3 ) d t ¯¯
¯ ¯ ¯
¯
¯ γz γw 2 →w 3 0
2 →z 3
¯
¯ Z 1 ¯
¯ ¯
+ ¯ ((w 3 − w 2 ) − (z 3 − z 2 ))
¯ g ((1 − t )w 2 + t w 3 ) d t ¯¯
0

≤ |z 3 − z 2 |ϵ + (|w 3 − z 3 | + |w 2 − z 2 |) M

≤ d ϵ + 2ϵM = (d + 2M )ϵ.

Now let us consider second and fourth terms in (1).

¯Z ¯ ¯ Z 1 ¯
¯ ¯ ¯ ¯
g (z)d z ¯ = ¯(w 3 − z 3 ) g ((1 − t )z 3 + t w 3 )d t ¯¯
¯ ¯ ¯
¯
¯ γz 0
3 →w 3
¯

≤ ϵM .

Similarly,

¯Z ¯
¯ ¯
g (z)d z ¯ ≤ M ϵ.
¯ ¯
¯
¯ γw
2 →z 2
¯

Hence,

¯Z ¯
¯ ¯
¯ g (z)d z ¯ ≤ cϵ
¯ ¯
T

for some positive constant c ∈ R.

(Refer Slide Time: 34:22)

285
 7

Since ϵ > 0 was arbitrary, we have,

Z
g (z)d z = 0.
T

Hence g is holomorphic in D(z 0 , r ). □

286
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 32
Problem Session

P ROBLEM 1. Let p be a polynomial of degree n with complex coefficients. Suppose

the roots of the polynomial p are contained in D(0, R) for large R. Then prove that

1 p ′ (z)
Z
dz = n
2πi γ p(z)

where γ(t ) = Re i t for t ∈ [0, 2π].

S OLUTION 1. By fundamental theorem of algebra, we have

p(z) = a n (z − z 1 ) · · · (z − z n ).

Now,
n V

p ′ (z) =
X
a n (z − z 1 ) . . . (z − z j ) . . . (z − z n ),
j =1

where (z − z j ) means (z − z j ) does not appear in the expression.

Away from z 1 , . . . , z n , we have the function,

n
P V

a n (z − z 1 ) . . . (z − z j ) . . . (z − z n )
p ′ (z) j =1
=
p(z) a n (z − z 1 ) · · · (z − z n )
n
X 1
= .
j =1 z − z j

(Refer Slide Time: 04:50)

1

287
2

Hence,

Z
p ′ (z) n
Z X
1
dz = dz
γ p(z) γ j =1 z − z j

X n Z dz
=
j =1 γ z − z j
n
X
= Wγ (z j )
j =1

= n.

P ROBLEM 2. Let f : Ω −→ C be a non-constant holomorphic function defined on an

open set Ω. Suppose a 1 , a 2 , . . . , a n be points in Ω such that f (a i ) = α for some α ∈ C.
Let γ be a closed continuous differentiable curve which is null homotopic and such that
a j ̸∈ γ. Then prove that

1
Z
f ′ (z) X n
dz = m j Wγ (a j )
2πi γ f (z) − α j =1

where m j are positive integers such that f (z) − α = (z − a j )m j g j (z), where g j (a j ) ̸= 0.

288
 3

S OLUTION 2. We have

f (a j ) = α for j = 1, 2, . . . , n.

If h(z) = f (z) − α, then h(a j ) = 0 for j = 1, 2, . . . , n. Hence we have,

h(z) = (z − a 1 )m1 g 1 (z).

Also,

h(a 2 ) = 0 =⇒ g 1 (z) = (z − a 2 )m2 g 2 (z).

..
.

h(a n ) = 0 =⇒ g n−1 = (z − a n )mn g (z).

i.e., h(z) = (z−a 1 )m1 · · · (z−a n )mn g (z), where g does not vanish in Ω. Then in Ω\{a 1 , a 2 , . . . , a n },
Z ′
h (z) f ′ (z)
Z
dz = d z.
γ h(z) γ f (z) − α

Now it is left as an exercise to the reader to verify that

h ′ (z) m1 m2 mn g ′ (z)
= + +···+ + .
h(z) (z − a 1 ) (z − a 2 ) (z − a n ) g (z)
(Refer Slide Time: 14:02)

289
4

Now,
Z ′
h (z) m1 m2 mn g ′ (z)
Z µ ¶
1 1
dz = + +···+ + dz
2πi γ h(z) 2πi γ (z − a 1 ) (z − a 2 ) (z − a n ) g (z)
Z ′
dz dz dz g (z)
µ Z ¶
1
Z Z
= m1 + m2 + · · · + mn + dz .
2πi γ (z − a 1 ) γ (z − a 2 ) γ (z − a n ) γ g (z)

By Cauchy’s theorem, we have

g ′ (z)
Z
dz = 0
γ g (z)

Also,
mj dz
Z
= m j Wγ (a j ) for j = 1, 2 . . . , n.
2πi γ (z − a j )
Hence
1
Z
f ′ (z) 1
Z
h ′ (z) X n
dz = dz = m j Wγ (a j ).
2πi γ f (z) − α 2πi γ h(z) j =1

P ROBLEM 3. Let Ω be an open connected set containing the origin and f : Ω −→ C

be holomorphic such that f ′ (0) ̸= 0. Then prove that there exists a neighborhood U of 0
and a function g holomorphic on U such that

f (z n ) = f (0) + (g (z))n for n > 0.

S OLUTION 3. Let us assume n > 1. Put h(z) = f (z n )− f (0). Then h ′ (z) = f ′ (z n ) nz n−1
¡ ¢

and hence h ′ (0) = 0. Now we are interested in finding the order of zero of h at 0. That is,
by Leibniz rule,

dk dk ¡
¯ ¯
′ ′ n n−1
¯ ¡ ¢¢ ¯
k
(h (z))¯ = f (z ) nz ¯
dz ¯
z=0 d zk z=0
¯
ℓ
à !
X k d ¡ ′ n ¢ d k−ℓ ¡ n−1 ¢ ¯
k
¯
= ℓ
f (z ) nz ¯ .
ℓ=0 ℓ d z d z k−ℓ ¯
z=0

Thus, h k+1 (0) = 0 for k < n − 1. In other words, for k < n, we have h k (0) = 0.
Let us now compute h n (0). By the computation above,

h n (0) = f ′ (0)n! =⇒ h n (0) ̸= 0.

290
 5

Hence h(z) = z n φ(z), where φ(0) ̸= 0. By a theorem proved earlier, we have

¡ ¢n
h(z) − h(0) = g (z)

for some function g holomorphic in a neighborhood of 0.

(Refer Slide Time: 24:18)

Hence,
¡ ¢n
f (z) − f (0) = g (z) .

P ROBLEM 4. Let f : Ω −→ C be a non-constant holomorphic function on an open

connected set Ω. Then | f (z)| does not attain a maximum in Ω.

S OLUTION 4. Let z 0 ∈ Ω. By open mapping theorem, f (Ω) is and hence f (z 0 ) is an

interior point. Let r > 0 be such that D( f (z 0 ), r ) ⊂ f (Ω).
Suppose f (z 0 ) = Re i θ . Then for 0 < s < r , f (z 0 ) + se i θ ∈ D( f (z 0 ), r ). Now,
¯ ¯ ¯ ¯
iθ¯ ¯ it it¯
(1) ¯ f (z 0 ) + se ¯ = ¯Re + se ¯ = R + s > | f (z 0 )|.
¯

Since D( f (z 0 ), r ) ⊆ Ω, there exists w ∈ Ω such that f (w) = f (z 0 ) + se i θ . By (1), we have

| f (w)| > | f (z 0 )|. Since the choice of z 0 was arbitrary, for any point z ∈ Ω, we can get hold
of a w ∈ Ω with | f (w)| > | f (z)|. Hence the maximum cannot be attained in Ω.

291
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 33
Singularities of a Holomorphic Function

Often times we are interested in studying functions f which are defined and holomor-
phic on a set Ω \ S where the set S is called the set of singularities of f . Here we will
be discussing the notion of singularities of a holomorphic function. Singularities come
in varying levels of severity and we will be exploring them one by one. Let us begin by
defining what is meant by a singularity of a holomorphic function.
(Refer Slide Time: 00:46)

D EFINITION 1 (Isolated Singularity). We say that a point z 0 is an isolated singularity

of a function f if f is holomorphic on D(z 0 , R) \ {z 0 } for some R > 0.

E XAMPLE 1.
ez − 1
(1) Let f (z) = on C \ {0}.
z

292
2

cos(z)
(2) Let g (z) = on C \ {0}.
z
1
(3) Let h(z) = e 1 − z on C \ {1}.

Here the functions f and g have isolated singularity at z 0 = 0 and the function h has
an isolated singularity at z 0 = 1.

D EFINITION 2 (Removable Singularity). Let z 0 be an isolated singularity of a holo-

morphic function f , i.e., there exists an R > 0 such that f is holomorphic on D(z 0 , R) \
{z 0 }. We say that z 0 is removable singularity of f if there exists a function g holomorphic
on D(z 0 , R) and such that f (z) = g (z) on D(z 0 , R) \ {z 0 }.

Let us revisit the examples.

ez − 1
(1) Let f (z) = . Then f has an isolated singularity at z 0 = 0. Let R > 0 be any
z
real number. Note that in D(0, R), we have
µ ∞ n¶ ∞ zn
z
X z X
e −1 = −1 = .
n=0 n! n=1 n!

∞ z n−1
P
Define F (z) := . Then F is holomorphic on D(0, R). (Why?)
n=1 n!
Hence on D(0, R), we have
µ ∞ n−1 ¶ ∞ n
X z X z
zF (z) = z = .
n=1 n! n=1 n!

Hence zF (z) = e z − 1. For z ̸= 0, in particular on D(0, R) \ {0}, we have

ez − 1
F (z) = = f (z).
z

Note that if an isolated singularity z 0 of f is a removable singularity, then lim f (z)

z→z 0
z̸=z 0
exists.
cos(z)
(2) Let g (z) = . Then g has an isolated singularity at 0. Note that cos(0) = 1.
z
We know that cos(z) is a continuous function and hence there exists R > 0 such

293
 3

that on D(0, R), | cos(z)| < M for some M > 0. Observe that,

¯ ¯
¯ cos(z) ¯
lim ¯ ¯ = ∞.
z→0 ¯ z ¯

¯ ¯
¯ cos(z) ¯
If 0 were to be a removable singularity, lim ¯ ¯ ¯ should exists. Hence 0 is not
z→0 z ¯
a removable singularity.
1
(3) Let h(z) = e 1 − z . Then h has an isolated singularity at 1.
1
Let z n = 1 − , so that z n −→ 1 as n −→ ∞. Then
2πi n

1
1 − 1 − 2πi1 n
¡ ¢
h(z n ) = e = e 2πi n = 1.

Hence lim h(z n ) = 1.

n→∞
1
Let z n′ = 1 − . Then z n′ −→ 1 as n −→ ∞. Now,
2πi n + πi
2

 
πi
µ ¶
1
h(z n′ ) = exp 
 
µ ¶  = exp 2πi n +
 = i.

1 2
1− 1−
2πi n+ πi
2

Here, lim h(z n′ ) = i . Hence h does not have a removable singularity at z = 1.

n→∞

(Refer Slide Time: 22:07)

294
4

T HEOREM 2 (Riemann Removable Singularity Theorem). Let z 0 be an isolated singu-

larity of a function f , i.e., there exists R > 0 such that f is holomorphic on D(z 0 , R) \ {z 0 }.
Then f has a removable singularity at z 0 if and only if f is locally bounded around z 0 (i.e.
there exist ϵ > 0 and M > 0 such that | f (z)| ≤ M for each z ∈ D(z 0 , ϵ) \ {z 0 }).

P ROOF. (⇒) Assume that f has a removable singularity at z 0 . Then there exists a
holomorphic function g on D(z 0 , R) that agrees with f on D(z 0 , R) \ {z 0 } and since g is
continuous at z 0 , g is bounded in a neighborhood D(z 0 , ϵ) of z 0 . Hence f is bounded in
D(z 0 , ϵ) \ {z 0 }.
(⇐) Let us now assume that f is locally bounded in a neighborhood D(z 0 , R) \ {z 0 }.
Define h : D(z 0 , R) −→ C by

(z − z 0 ) f (z),

z ∈ D(z 0 , R) \ {z 0 }
h(z) :=
z = z0 .

0,

Then h is continuous at z 0 (verify). If h was holomorphic on D(z 0 , R), then there exists a
power series expansion of h on D(z 0 , R) given by
∞
a n (z − z 0 )n
X
h(z) =
n=0

295
 5

and since h(z 0 ) = 0, we have

∞
a n (z − z 0 )n .
X
h(z) =
n=1

That is, on D(z 0 , R) we have,

µ ∞ ¶
n−1
X
h(z) = (z − z 0 ) a n (z − z 0 )
n=1

∞
a n (z − z 0 )n−1 . Then on D(z 0 , R) \ {z 0 }, we have
P
Define g (z) :=
n=1

h(z) = (z − z 0 ) f (z) = (z − z 0 )g (z) =⇒ f (z) = g (z).

Hence f has a removable singularity at z 0 .

(Refer Slide Time: 35:02)

Claim: h is holomorphic on D(z 0 , R).

We shall use Morera’s theorem for proving above claim.
R
If a triangular curve T satisfies z 0 ̸∈ T̂ , then T is hull homotopic and we have T h = 0.
Let us consider the case when z 0 is a vertex of T = γz0 →z1 →z2 →z0 , where z 1 , z 2 ∈
D(z 0 , R). Let d = ¯γz0 →z1 →z2 →z0 ¯. Now if w 1 ∈ γz0 →z1 and w 2 ∈ γz2 →z0 , then by Cauchy’s
¯ ¯

296
6

theorem we have
Z
h(z)d z = 0.
γw 1 →z1 →z2 →w 2 →w 1

Hence,
Z Z Z
h(z)d z = h(z)d z + h(z)d z
T γz0 →w 1 →w 2 →z0 γw 1 →z1 →z2 →w 2 →w 1
Z
= h(z)d z.
γz0 →w 1 →w 2 →z0

ϵ
Since h is continuous, given ϵ > 0, let δ > 0 be such that |h(w) − h(z 0 )| < , whenever
d
ϵ
|w − z 0 | < δ. But h(z 0 ) = 0, then |h(w)| < , whenever |w − z 0 | < δ.
d
Now pick w 1 , w 2 above such that γ̂z0 →w 1 →w 2 →z0 ⊆ D(z 0 , δ). Then,

¯Z ¯ ¯¯Z ¯
¯ ϵ¯
h(z)d z ¯ ≤ ¯γz0 →w 1 →w 2 →z0 ¯ = ϵ.
¯ ¯ ¯
¯ h(z)d z ¯ = ¯¯ ¯
¯
T
¯ ¯
γz0 →w 1 →w 2 →z0 ¯ d

(Refer Slide Time: 44:27)

¯Z ¯ Z
Hence ¯¯ h(z)d z ¯¯ < ϵ. Since the ϵ was arbitrary, we have h(z)d z = 0.
¯ ¯
T T

297
 7

Now let us consider the case where z 0 is an edge of T . If T = γz1 →z2 →z3 →z1 and with-
out loss of generality, let z 0 ∈ γz1 →z2 . Then,
Z Z Z
h(z)d z = h(z)d z + h(z)d z.
T γz1 →z0 →z3 →z1 γz2 →z3 →z0 →z2

But we have already proved the case where z 0 is a vertex of the triangle. Hence,
Z Z
h(z)d z = 0, h(z)d z = 0.
γz1 →z0 →z3 →z1 γz2 →z3 →z0 →z2

Therefore,
Z
h(z) = 0.
T
Finally, if z 0 is in the interior of T̂ , where T = γz1 →z2 →z3 →z1 . Here also we can apply the
R
case where z 0 is a vertex by decomposing the integral T h(z)d z as
Z Z Z Z
h(z)d z = h(z)d z + h(z)d z + h(z)d z.
T γz1 →z0 →z3 →z1 γz2 →z0 →z1 →z2 γz3 →z0 →z2 →z3
Z
But each of the integral in the RHS is 0, hence h(z)d z = 0. □
T

298
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 34
Pole of a Function

In the last lecture we saw what was meant by a removable singularity of a function f .
We observed that an isolated singularity of f is a removable singularity if and only if f
is locally bounded around that point. We also noted that if an isolated singularity z 0 is a
removable singularity then the limit lim f (z) should exist. By considering the behavior
z→z 0
of f (z) as z goes to the singularity, let us now study the other types of singularities that
can occur.
(Refer Slide Time: 00:46)

D EFINITION 1 (Pole of a function). Let z 0 be an isolated singularity of f . We say that

f has a pole at z 0 if lim | f (z)| = ∞. That is, given M > 0, there exists ϵ > 0 such that
z→z 0
| f (z)| > M for each z ∈ D(z 0 , ϵ) \ {z 0 }.

E XAMPLE 1.
1

299
2
¯ ¯
cos(z) ¯ cos(z) ¯
• Let f (z) = . Then 0 is an isolated singularity of f . Also, lim ¯
¯ ¯ = ∞.
z z→0 z ¯
Hence f has a pole at z 0 = 0.
1
• Recall that if f (z) = e 1 − z , then 1 is an isolated singularity of f and lim f (z)
z→1
does not exists. Hence f does not have a pole at 1.

D EFINITION 2 (Essential Singularity of a function). Let z 0 be an isolated singularity

of f . Then z 0 is called an essential singularity of f if it is neither a removable singularity
nor a pole of f .

Let z 0 be a pole of f . Then lim | f (z)| = ∞. Hence there exists R > 0 such that on
z→z 0
1
D(z 0 , R) \ {z 0 }, f (z) ̸= 0. In particular, on D(z 0 , R) \ {z 0 }, g (z) := is holomorphic.
f (z)
Then z 0 is an isolated singularity of g .
Let R > 0 be such that | f (z)| > M on D(z 0 , R) \ {z 0 }. That is,on D(z 0 , R) \ {z 0 }, we have
1
|g (z)| < . By Riemann removable singularity theorem, z 0 is a removable singularity of
M
g . Now it is left as an exercise for the reader to verify that lim g (z) = 0.
z→z 0
Hence let us define h to be

g (z)

, on D(z 0 , R) \ {z 0 }
h(z) :=
, for z = z 0 .

0

Then h is holomorphic. By factorization and principle of analytic continuation, on

D(z 0 , R),

h(z) = (z − z 0 )m h 1 (z) where h 1 (z) ̸= 0 and m ≥ 1.

Hence on D(z 0 , R) \ {z 0 }, we have

1
= g (z) = (z − z 0 )m h 1 (z).
f (z)

(Refer Slide Time: 13:32)

300
 3

1
Since h 1 (z) ̸= 0 on D(z 0 , R), let h 0 (z) = , which is holomorphic on D(z 0 , R).
h 1 (z)
Thus, we have, on D(z 0 , R) \ {z 0 },
h 0 (z)
f (z) = .
(z − z 0 )m
Since h 0 is complex analytic, on D(z 0 , R), we have

h 0 (z) = a 0 + a 1 (z − z 0 ) + · · · + a m−1 (z − z 0 )m−1 + a n (z − z 0 )n

X
where a 0 ̸= 0.
n≥m

Hence, on D(z 0 , R) \ {z 0 },
a0 a1 a m−1
a n (z − z 0 )n−m , a 0 ̸= 0.
X
f (z) = + + · · · + +
(z − z 0 )m (z − z 0 )m−1 (z − z 0 ) n≥m
a n (z − z 0 )n−m . Observe that g 1 is a holomorphic function on D(z 0 , R).
P
Put g 1 (z) =
n≥m
The function S given by
a0 a1 a m−1
S(z) = m
+ m−1
+···+
(z − z 0 ) (z − z 0 ) (z − z 0 )
is called the singular part of f around the pole z 0 and the integer m is called the order of
the pole at z 0 .

301
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 35
Laurent series

In this lecture, our goal would be to generalize the expression that we gave for a function
f which has a pole at z 0 and get hold of a more general such series expansion which
is called the Laurent series expansion. In order to do that we will be now considering
functions which are defined on an annulus and we will be considering doubly infinite
series. So, before we really start developing our theory with Laurent series, let us set
certain notations that we will be using.

D EFINITION 1 (Doubly Infinite Series). Let {z n : n = 0, ±1, ±2, . . . } be a doubly infi-

∞
X ∞
X
nite sequence of complex numbers. The expression z n is said to converge if zn
n=−∞ n=0
∞
X
converges and z −n converges.
n=1
∞
X ∞
X ∞
X
We say that z n converges absolutely if z n converges absolutely and z −n
n=−∞ n=0 n=1
converges absolutely.
Then we say that

∞
X ∞
X ∞
X
zn = zn + z −n
n=−∞ n=0 n=1

We define uniform convergence also very similarly. Let u n , for n = 0, ±1, ±2, . . . , be
∞
X ∞
X
functions defined on a set U . We say that u n converges uniformly on U if u n (z)
n=−∞ n=0
∞
X
converges uniformly on U and u −n (z) converges uniformly on U .
n=1
(Refer Slide Time: 05:41)
1

302
2

D EFINITION 2 (Annulus). An annulus A(z 0 , R 1 , R 2 ) around a point z 0 ∈ C, for o ≤ R 1 <

R 2 , is the set

A(z 0 , R 1 , R 2 ) := {z ∈ C : R 1 < |z − z 0 | < R 2 }.

We now have all the ingredients to talk about the Laurent series development in an
annulus.

T HEOREM 1 (Laurent Series). Let f be a function holomorphic on an annulus A(z 0 , R 1 , R 2 ).

Then there exist a n ∈ C for n ∈ Z such that
∞
a n (z − z 0 )n
X
f (z) =
n=∞

called the Laurent series of f around z 0 , where the doubly infinite series converges abso-
lutely and uniform in A(z 0 , r 1 , r 2 ) where R 1 < r 1 < r 2 < R 2 . Moreover, if γ(z) = z 0 + r e i t for
t ∈ [0, 2π], where R 1 < r < R 2 , then

1 f (z)
Z
an = n+1
d z.
2πi γ (z − z 0 )

(Refer Slide Time: 12:33)

303
 3

P ROOF. For R 1 < r 1 < r 2 < R 2 , let

γ1 (t ) = z 0 + r 1 e i t for t ∈ [0, 2π]

and

γ2 (t ) = z 0 + r 2 e i t for t ∈ [0, 2π].

Notice that γ1 is homotopic to γ2 as closed curves in A(z 0 , R 1 , R 2 ).

Hence,

f (z) f (z)
Z Z
dz = d z.
γ1 (z − z 0 )n+1 γ2 (z − z 0 )n+1

(Refer Slide Time: 15:41)

304
4

On D(z 0 , R 2 ), let γr (t ) = z 0 + r e i t . Then define

f (w)
Z
g (z) := dz
γr w − z

where |z − z 0 | < r and R 1 < r < R 2 .

Claim: g is continuous and holomorphic on D(z 0 , R 2 ).
Let z ′ ∈ D(z 0 , R 2 ). Let R be such that R 1 < R < R 2 and |z ′ − z 0 | < R. Then, on D(z ′ , ϵ) for
ϵ > 0 small, there exists δ > 0 such that |w − z| > δ, |w − z ′ | > δ for every w ∈ γR .
Then,
¯ ¯
¯ 1 f (w) 1 f (w)
Z Z
′
¯
|g (z) − g (z )| = ¯¯ dz − d z ¯
2πi γR w − z 2πi γR w − z ′ ¯
¯ ′ ¯
¯ 1 f (w)(z − z )
Z
¯
= ¯¯ d w ¯
2πi γR (w − z)(w − z ′ ) ¯

′ M
µ ¶
≤ |z − z | 2 2πR .
δ

Hence g is continuous at z ′ .
For z ̸= z ′ ,
g (z) − g (z ′ ) 1 f (w)
Z
′
= d w.
z −z 2πi γR (w − z)(w − z ′ )

305
 5

Now it is left as an exercise to the reader to verify that

g (z) − g (z ′ ) 1 f (w)
Z
lim′ = d w.
z→z z −z ′ 2πi γR (w − z ′ )2
z̸=z ′

Hence g is holomorphic on D(z 0 , R 2 ).

Now, define a function h on {z : |z − z 0 | > R 1 } by

1 f (w)
Z
h(z) = − dw
2πi γr w −z

where r ≤ |z − z 0 | and R 1 < r < R 2 .

(Refer Slide Time: 24:14)

Let z ∈ A(z 0 , R 1 , R 2 ) and R 1 < r 1 < r 2 < R 2 be such that r 1 < |z − z 0 | < r 2 . Consider a
curve γ from γr 1 (0) to γr 2 (0) such that z ̸∈ γ.
Let σ = γr 2 + (−γ) + (−γr 1 ) + γ. Then σ is a closed curve in A(z 0 , R 1 , R 2 ). Note that σ is
null homotopic in A(z 0 , R 1 , R 2 ).

306
6

Then, on A(z 0 , R 1 , R 2 ), we have

1 f (w)
Z
Wσ (z) f (z) = dw
2πi σ w − z
ÃZ !
1 f (w) f (w) f (w) f (w)
Z Z Z
= dw + dw + dw + dw
2πi γr 2 w − z −γ w − z (−γr 1 ) w − z γ w −z

= g (z) + h(z)

We know that g (z) can be written as a power series expansion on D(z 0 , R 2 ). Let us see
what happens to h(z).
On D (0, 1/R 1 ) \ {0}, define
µ ¶
1
h 1 (z) = h z 0 + .
z
Since
¯ ¯ ¯ ¯
¯z 0 + 1 − z 0 ¯ = ¯ 1 ¯ > R 1 ,
¯ ¯ ¯ ¯
¯ z ¯ ¯z ¯

h 1 is well defined as {z : |z − z 0 | > R 1 } is the domain of definition of h.

Claim: h 1 is locally bounded around 0.
For |z| > s, where s > 0 is some large real,
¯ ¯
¯ 1 Z f (w) ¯
|h(z)| = ¯− dw¯
¯ ¯
¯ 2πi γr w − z ¯
1

M 2πr 1
≤
2πd (z, γr 1 )
M r1
=
d (z, γr 1 )

< 1.

Since h 1 (z) = h z 0 + 1z where |z| < ϵ for ϵ > 0 small, we have ¯z 0 + 1z ¯ > s and |h 1 (z)| =
¡ ¢ ¯ ¯

¯h z 0 + 1 ¯ < 1. Hence h 1 has a removable singularity at 0.

¯ ¡ ¢¯
z
∞
b n z n . Since h 1 (0) = 0, we can rewrite the series expan-
X
That is, we have, h 1 (z) =
n=0
sion for h 1 on D(0, 1/R 1 ) as
∞
bn z n .
X
h 1 (z) =
n=1

307
 7
³ ´
Let z 0 + 1z = w. Then z = 1
w−z 0
. Hence h(w) = h 1 1
w−z 0
.
Thus, if z ∈ {z ∈ C : |z − z 0 | > R 1 },
µ ¶
1
h(z) = h 1
z − z0
∞ µ ¶n
X 1
= bn
n=1 z − z0
−∞
a n (z − z 0 )n .
X
=
n=−1

Thus,

f (z) = g (z) + h(z)

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

(Refer Slide Time: 47:41)

308
8

Suppose f has an isolated singularity at z 0 . Then, for some R > 0, we have

∞
a n (z − z 0 )n
X
f (z) = on D(z 0 , R) \ {z 0 }.
n=−∞

Suppose z 0 is a removable singularity of f , then we have

∞
b n (z − z 0 )n
X
f (z) = on D(z 0 , R).
n=0

Hence a n = 0 for every n < 0.

The converse of above observation is also true. This can be proved if one back track
what we did above. That is if a n = 0 for every n < 0, then f has a power series expansion
around z 0 . Hence z 0 is a removable singularity. These observation will the following
proposition.

P ROPOSITION 2. Let z 0 be an isolated singularity of the function f . Then f has a

removable singularity at z 0 if and only if a n = 0 for n < 0 in the Laurent series expansion
of f around z 0 .

If f has a pole at z 0 of order m, then (z − z 0 )m f has a removable singularity at z 0 .

Thus, on D(z 0 , R) \ {z 0 }, we have,
b −m b −1 ∞
a n (z − z 0 )n .
X
f (z) = + · · · + +
(z − z 0 )m (z − z 0 ) n=0
Hence a n = 0 for each n < −m in the Laurent series expansion around z 0 of f .
Here also, the converse of the above observation is true. That is, if a n = 0 for n < −m
in the Laurent series expansion around z 0 of f , then f has a pole of order m at z 0 . The
proof is trivial, as it follows from the definition.

P ROPOSITION 3. Let z 0 be an isolated singularity of the function f . Then f has a pole

of order m at z 0 if and only if a n = 0 for n < −m in the Laurent series expansion of f in
D(z 0 , R) \ {z 0 } = A(z 0 , 0, R).

P ROPOSITION 4. Let z 0 be an isolated singularity of the function f . Then f has an

essential singularity at z 0 if and only if a n ̸= 0 for infinitely many negative integers n.

309
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 36
Casorati-Weierstarss Theorem

We had explored the notions of removable singularity and the pole of a function f at an
isolated singularity z 0 in great detail. We also studied the behaviour of the function as we
approached these isolated singularities. For example, when we approach a removable
singularity z 0 of f , we notice that the limit should exists and when we approach a pole
of the function f at z 0 , we noticed that the absolute value of the function blow up.
We also explored a Laurent series expansion of a function f defined on an annulus
and we classified our singularities bases on how the negative coefficients of the Lau-
rent series behave. However, we have still not really looked into what happens when a
given isolated singularity is an essential singularity and how the function behaves as we
approach an essential singularity.
1
Recall that f (z) = e 1−z has essential singularity at z 0 = 1. Then, for

1
zn = 1 − , we have f (z n ) −→ 1 as n −→ ∞
2πi n

and for

1
z n′ = 1 − , we have f (z n′ ) −→ i as n −→ ∞.
2πi n + i2π

We will now prove a result which tells us that in fact the behaviour of essential singular-
ities can get as bad as once can expect.
(Refer Slide Time: 02:05)
1

310
2

T HEOREM 1 (Casorati-Weierstrass). Let z 0 be an essential singularity of a function

f . Then given α ∈ C, there exists a sequence z n ∈ D(z 0 , R) \ {z 0 } such that z n −→ z 0 and
f (z n ) −→ α.

P ROOF. Suppose α ∈ C be such that there does not exists a z n such that z n −→ z 0 and
f (z n ) −→ α. That is, there exists ϵ > 0 such that D(α, ϵ) ∩ f (D(z 0 , R) \ {z 0 }) = ∅.
1
Consider g (z) = f (z)−α. Then g does not vanish on D(z 0 , R)\{z 0 }. Let h(z) = g (z)
and
1
h i holomorphic on D(z 0 , R) \ {z 0 }. Since |g (z)| > ϵ, we have h(z)| < ϵ on D(z 0 , R) \ {z 0 }.
Hence h is bounded on D(z 0 , R) \ {z 0 }. By the Riemann removable singularity theorem,
z 0 is a removable singularity of h. Since h is not a constant function, on D(z 0 , R) we have

h(z) = (z − z 0 )m h 1 (z)

where m ≥ 0 and h 1 (z 0 ) ̸= 0. We may assume that h 1 (z) ̸= 0 on D(z 0 , R). Then we have
1
= (z − z 0 )m h 1 (z).
f (z) − α
That is,
1
h 1 (z)
f (z) = α + .
(z − z 0m

311
 3

Thus f has a pole of order m at z 0 , which is a contradiction to the fact that z 0 is an

essential singularity of f . Hence there exists a sequence z n −→ z 0 such that f (z n ) −→
α. □

Let us now define what is meant by a meromorphic function. But before we do that,
let us recall the notion of the order of a pole of the function f at z 0 .
Let f be a function with an isolated singularity at z 0 . Let z 0 be a pole of order m at
z 0 . That is,
g (z)
f (z) =
(z − z 0 )m
in D(z 0 , R) \ {z 0 } where g is holomorphic on D(z 0 , R) and g (z 0 ) ̸= 0.
We will similarly define the notion of order of zero of a holomorphic function f
which vanishes at a point z 0 . Suppose f be a non-constant holomorphic function in
a neighborhood of a point z 0 such that f (z 0 ) = 0. Hence we have,

f (z) = (z − z 0 )m g (z)

where g is holomorphic and g (z 0 ) ̸= 0. We then say that f has a zero of order m at z 0 .

(Refer Slide Time: 15:59)

312
4

D EFINITION 1 (Meromorphic Function). Let Ω be an open connected subset and

S ⊂ Ω be a subset of Ω. Let f : Ω \ S −→ C be holomorphic on Ω \ S. We say that f is a
meromorphic function on Ω if

(1) S is a discrete set.

(2) f has either a removable singularity or a pole at points of S.

We say that two meromorphic functions f and g on Ω are equivalent if f : Ω\S 1 −→ C

and g : Ω \ S 2 −→ C satisfies f (z) = g (z) on Ω \ (S 1 ∪ S 2 ). It is left to the reader to check
that if S 1 and S 2 are discrete sets, then so is S 1 ∪ S 2 .
Define M (Ω) := { Equivalence classes of meromorphic functions on Ω}.
Let f , g ∈ Ω, i.e., f : Ω\S 1 −→ C and g : Ω\S 1 −→ C be such that f = g on Ω\(S 1 ∪S 2 ).
Define f + g to be the equivalence class of ( f + g ) : Ω\(S 1 ∪S 2 ) −→ C given by ( f + g )(z) =
f (z) + g (z). Similarly define f g to be the equivalence class of ( f g ) : Ω \ (S 1 ∪ S 2 ) −→ C
given by ( f g )(z) = f (z)g (z). Let z 0 ∈ S 1 ∪ S 2 . Then z 0 is either a removable singularity of
f g or a pole of f g .
We know the behaviour of the functions f and g in a neighborhood of z 0 ,

f (z) = (z − z 0 )m1 f 1 (z) where f 1 (z 0 ) ̸= 0,

g (z) = (z − z 0 )m2 g 1 (z) where g 1 (z ) ̸= 0

and m 1 , m 2 are non-negative or negative based on whether z 0 is a removable singularity

or pole of f , g respectively. Then,

f (z)g (z) = (z − z 0 )m1 +m2 f 1 (z)g 1 (z).

(Refer Slide Time: 27:43)

313
 5

P ROPOSITION 2. The space of meromorphic function M (Ω) on Ω is a field with oper-

ations defined above.

P ROOF. Let f ∈ M (Ω)∗ , i.e., that is f is a non-zero meromorphic function on Ω. Then

f : Ω \ S −→ C is a non-zero holomorphic function. Let S 1 = {z ∈ C : f (z) = 0}. By identity
theorem, S 1 is discrete.
1
Define g : Ω \ (S ∪ S 1 ) −→ C by g (z) = . Then g is holomorphic on Ω \ (S ∪ S 1 ). Let
f (z)
z 0 ∈ S 1 . Then f (z 0 ) = 0. Thus we have f (z) = (z − z 0 )m1 f 1 (z 0 ), where f 1 (z 0 ) ̸= 0. Now, in
D(z 0 , R) \ {z 0 }, we have f 1 (z) ̸= 0 for z ∈ D(z 0 , R) \ {z 0 } and
1
1 f 1 (z)
g (z) = = .
f (z) (z − z 0 )m1
1
That is, g = has a pole of order m 1 at z 0 .
f
If z 0 ∈ S, then z 0 is either a pole or a removable singularity of f . First let us consider
the case when z 0 is a pole of f .
If z 0 is a pole of f , then we have
f 2 (z)
f (z) =
(z − z 0 )m2

314
6

where f 2 (z) ̸= 0 on D(z 0 , R) \ {z 0 }. Hence,

µ ¶
1 m2 1
g (z) = = (z − z 0 ) ,
f (z) f 2 (z)
1
and g = has a removable singularity at z 0 .
f
Now it is left as an exercise to the reader to check if z 0 is a removable singularity of
f , then z 0 is either a removable singularity or a pole of g . Hence g is meromorphic on Ω
and f is a unit in M (Ω). □

D EFINITION 2 (Order of a meromorphic function at z 0 ). Let f be a meromorphic

function on Ω. Then for z 0 ∈ Ω, define the order of f at z 0 to be:

(i) if z 0 ∈ S and z 0 is a removable singularity, then order of f at z 0 is the order of the

zero at z 0 of f , i.e., if f (z) = (z − z 0 )m g (z), then or d z0 ( f ) = m. (Note that here m is
non-negative.)
(ii) if z 0 ∈ S and is a pole of order m, then or d z0 ( f ) = −m.
(iii) if z 0 ̸∈ S and we have f (z) = (z − z 0 )m g (z), where m ≥ 0, then or d z0 ( f ) = m.
(iv) if f ≡ 0, then or d z0 ( f ) = ∞.

(Refer Slide Time: 39:44)

315
 7

E XERCISE 3. Prove that or d z0 : M (Ω) −→ Z satisfies

(1) or d z0 ( f g ) = or d z0 ( f ) + or d z0 (g )
¡ ¢
(2) or d z0 ( f + g ) ≥ min or d z0 ( f ), or d z0 (g ) .

316
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 37
Problem Session

(Refer Slide Time: 00:30)

P ROBLEM 1. Let f : D(z 0 , r )\{z 0 } −→ C be a holomorphic function such that Im( f (z)) >
0 for every z ∈ D(z 0 , r ) \ {z 0 }. Then prove that z 0 is a removable singularity.

z −i
S OLUTION 1. Consider the function φ(z) = . Then φ is holomorphic with inverse
z +i
1+z
given by φ−1 (z) = i .
1−z
Now the reader should verify that φ maps the upper half plane H := {z ∈ C : Im(z) > 0}
onto the unit disc D centred at the origin. Since we are given that Im( f (z)) > 0 for every
z ∈ D(z 0 , r ) \ {z 0 }, we have, f : D(z 0 , r ) \ {z 0 } −→ H.
By composition with φ, we have φ ◦ f : D(z 0 , r ) \ {z 0 } −→ D. Hence φ ◦ f is bounded
and by Riemann removable singularity theorem, there exists a holomorphic function g
1

317
2

on D(z 0 , r ) such that

φ ◦ f (z) = g (z) on D(z 0 , r ) \ {z 0 }.

By open mapping theorem, we have g : D(z 0 , r ) −→ D and |g (z 0 )| < 1. Since φ is invert-

ible, we can compose g with φ−1 and we have

f (z) = φ−1 ◦ g (z) on D(z 0 , r ) \ {z 0 }

and φ−1 ◦ g is holomorphic on D(z 0 , r ) . Therefore z 0 is a removable singularity of f .

P ROBLEM 2. Let z 0 be a pole of a function f : D(z 0 , r ) \ {z 0 } −→ C. Then prove that

there exists R > 0 such that

{z ∈ C : |z| > R} ⊆ f (D(z 0 , r ) \ {z 0 }) .

P ROOF. Let z 0 be a pole of order m. Let us assume that f (z) ̸= 0 on D(z 0 , r )\{z 0 }. Then
there exists a holomorphic function g : D(z 0 , r ) −→ C with g (z) ̸= 0 on D(z 0 , r ) such that
g (z)
on D(z 0 , r ) \ {z 0 }, we have f (z) = .
(z − z 0 )m
On D(z 0 , r ) \ {z 0 }, let us define h 1 to be,

1 (z − z 0 )m
h 1 (z) = = .
f (z) g (z)

Then h 1 has a removable singularity at z 0 and that h(z 0 ) = 0 where h extends h 1 to

D(z 0 , r ). By the open mapping theorem, since h is not a constant function, there exists
1 1
ϵ > 0 such that D(0, ϵ) ⊆ h(D(z 0 , r )). If w ∈ C be such that |w| > ϵ
= R, then w
∈ D(0, ϵ)
and since w ̸= 0, there exists z ∈ D(z 0 , r ) \ {z 0 } such that

1 1
= h(z) = h 1 (z) = .
w f (z)

Hence f (z) = w and {w ∈ C : |w| > R} ⊆ f (D(z 0 , r ) \ {z 0 }). □

(Refer Slide Time: 15:00)

318
 3

P ROBLEM 3. Let f : D(z 0 , r ) \ {z 0 } −→ C be a holomorphic function such that the real

¡ ¢
part of f , Re f (z) < M for some real number M and for every z ∈ D(z 0 , r ) \ {z 0 }. Then f
has a removable singularity at z 0 .

S OLUTION 2. Suppose z 0 is a pole of f . Then by Problem 2, there exists R > 0 such

that

(1) {z ∈ C : |z| > R} ⊆ f (D(z 0 , r ) \ {z 0 }) .

Let w ∈ C such that |w| > R and such that Re(w) > M . By (1), there exists z ∈ D(z 0 , r ) \
¡ ¢
{z 0 } such that f (z) = w. Then Re f (z) = Re(w) > M for z ∈ D(z 0 , r ) \ {z 0 }, which is a
contradiction. Hence z 0 cannot be a pole.
Suppose z 0 is an essential singularity. Then by Casorati-Weierstrass theorem, f (D(z 0 , r )\
{z 0 }) is dense in C. Let w ∈ C be such that Re(w) > M and let {z n }n∈N be a sequence in
D(z 0 , r ) \ {z 0 } such that z n −→ z 0 and f (z n ) −→ w. Then there exists N ∈ N such that for
¡ ¢
every n > N , we have Re f (z n ) > M , which is again a contradiction. Hence z 0 cannot
be an essential singularity.

319
4

P ROBLEM 4. Let z 0 be an isolated singularity of f . Then e f does not have a pole at

z0 .

S OLUTION 3. We will consider the all three cases of z 0 being an isolated singularity
of f .
Case-I: Let z 0 be a removable singularity of f . That is, there exists a holomorphic
function g : D(z 0 , r ) −→ C such that f (z) = g (z) on D(z 0 , r ) \ {z 0 }. Hence, we have e f (z) =
e g (z) for each z ∈ D(z 0 , r ) \ {z 0 }. Since e g is a holomorphic function on D(z 0 , r ), e f has a
removable singularity at z 0 .
(Refer Slide Time: 26:01)

Case-II: Let z 0 be a pole of f . Suppose z 0 is a removable singularity of e f . Then,

¯ ¯
¯ f (z) ¯
¯e ¯<M for z ∈ D(z 0 , r ) \ {z 0 }

=⇒ e Re( f (z)) < M for z ∈ D(z 0 , r ) \ {z 0 }

=⇒ Re f (z) < M ′ for z ∈ D(z 0 , r ) \ {z 0 } and for some M ′ .

¡ ¢

Hence by Problem 3, f has a removable singularity at z 0 which is a contradiction to our

assumption that z 0 is a pole of f . Therefore z 0 cannot be a removable singularity of e f .

320
 5

Suppose z 0 is a pole of e f . Then e − f has a removable singularity at z 0 . By the argument

above, we have z 0 is a removable singularity of − f and hence is a removable singularity
of f , again a contradiction. Thus z 0 is an essential singularity of e f .
Case-III- Let z 0 be an essential singularity of f . By Casorati-Weierstrass theorem, we
have f (D(z 0 , r ) \ {z 0 }) = C. Since exponential map is continuous, we have,
³ ´ ¡ ¢
exp f (D(z 0 , r ) \ {z 0 }) ⊆ exp f (D(z 0 , r ) \ {z 0 }) .

Hence we have

C \ {0} ⊆ exp f (D(z 0 , r ) \ {z 0 }) .

¡ ¢

Therefore exp f (D(z 0 , r ) \ {z 0 }) = C as it is closed and contains C \ {0} implies that z 0 is

¡ ¢

an essential singularity of e f .

P ROBLEM 5. Let f : Ω \ {z 0 , z 1 , z 2 . . . } −→ C be a holomorphic function, where Ω is an

open connected subset of C and z n be a sequence of distinct points such that z n −→ z 0
in Ω. Let z 1 , z 2 , . . . be poles of f . Then given ϵ > 0, prove that f (D(z 0 , ϵ) \ {z 0 , z 1 , z 2 , . . . }) is
dense in C.

S OLUTION 4. Consider α ∈ C such that α ̸∈ f (D(z 0 , ϵ) \ {z 0 , z 1 , . . . }). Then there exists

δ > 0 such that D(α, δ) ∩ f (D(z 0 , ϵ) \ {z 0 , z 1 , . . . }) = ∅.
1
On Ω \ {z 0 , z 1 , z 2 . . . }, define g (z) = . Then,
f (z) − α

1 1
|g (z)| = ¯ ¯< on Ω \ {z 0 , z 1 , z 2 . . . }.
¯ f (z) − α¯ α

Since z n for n > 0 are isolated singularities of f and g is bounded by δ1 on Ω\{z 0 , z 1 , z 2 . . . },

g has a removable singularity at z n for every n > 0.
Since z n is a pole of f , we have g (z n ) = 0 and therefore g : D(z 0 , r ) \ {z 0 } −→ C has a
removable singularity at z 0 and moreover by the continuity of g , we have g (z 0 ) = 0.
(Refer Slide Time: 41:19)

321
6

Notice that g is a non-zero function such that g (z n ) = 0 for every n ≥ 0 and such that
{z n : n ∈ N} has a limit point, which is a contradiction to the identity theorem. Hence
f (D(z 0 , ϵ) \ {z 0 , z 1 , . . . }) = C.

1
P ROBLEM 6. Consider f (z) = . Find the Laurent series expansion of f in the
z(z − 1)
following annuli:

(a) D(0, 1) \ {0}

(b) D(1, 1) \ {1}
(c) {z : |z| > 1}.

S OLUTION 5.

(a) On D(0, 1), we know that

1 −1 ∞
zn.
X
= =−
z −1 1−z n=0

Then for z ̸= 0, we have,

µ ∞ ¶ ∞ ∞
1 1 X n −1 X
z n−1 = zn.
X
= − z =− −
z(z − 1) z n=0 n=0 z n=0

(Refer Slide Time: 48:05)

322
 7

(b) On D(1, 1), we have

1 1 ∞
(1 − z)n .
X
= =
z 1 − (1 − z) n=0
For z ̸= 1,
µ∞ ¶ ∞ ∞
1 1 n 1
(1 − z)n−1 = (1 − z)n .
X X X
= (1 − z) = − −
z(z − 1) (z − 1) n=0 n=0 z − 1 n=0

(c) On {z : |z| > 1}, we have,

1 1 1 X∞ µ 1 ¶n ∞ µ 1 ¶n+2 ∞
z −n .
X X
= ¡ 1
¢ = 2
= =
2 z n=0 z
(z − 1)z z 1 − z n=0 z n=2

323
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 38
Residue Theorem

Let us explore the behavior of holomorphic functions around its isolated singularities a
bit further. Here we will state and prove the residue theorem. Residue theorem broadly
answers the following question: suppose you have a function f which is holomorphic
on an open set Ω \ S, where S is a discrete set of singularities of f , then what can we say
about γ f (z)d z, where γ is some closed curve in Ω \ S.
R

Suppose S was empty and γ was null homotopic, then this is just the Cauchy’s theo-
rem. The Cauchy’s theorem tells us that this integral is equal to 0. So in some sense, the
residual theorem can be thought of as a generalization of the Cauchy’s theorem. How-
ever the presence of these singularities makes the answer to this question a bit more
complicated.
(Refer Slide Time: 01:30)

324
2

D EFINITION 1 (Residue of a function at a point z 0 ). Let f : Ω \ S −→ C be a holomor-

phic function where Ω is an open set and S is a discrete subset of Ω. For z 0 ∈ S, let r > 0
be such that D(z 0 , r ) ⊆ Ω and D(z 0 , r ) ∩ S = {z 0 }. Then in D(z0, r ) \ {z 0 }, consider the
Laurent series expansion of f given by,

∞
a n (z − z 0 )n .
X
f (z) =
n=−∞

We define the residue of f at z 0 to be

¡ ¢
Res f , z 0 := a −1 .

If z 0 is a removable singularity, then a −n = 0 for every n ∈ N. Hence Res f , z 0 = 0.

¡ ¢

If z 0 is a pole of f of order 1, then

g (z)
f (z) = in D(z 0 , r ) \ {z 0 }
(z − z 0 )

where g ̸= 0. Thus, we have,

(z − z 0 ) f (z) = g (z) = a 0 + a 1 (z − z 0 ) + a 2 (z − z 0 )2 . . .

a0
=⇒ f (z) = + a 1 + a 2 (z − z 0 ) + . . . .
(z − z 0 )
¡ ¢
Hence Res f , z 0 = a 0 = g (z 0 ).
If z 0 is a pole of order m, then

(z − z 0 )m f (z) = g (z)

where g (z) ̸= 0 on D(z 0 , r ) \ {z 0 }.

(Refer Slide Time: 09:56)

325
 3

Hence we have
a0 a m−1
f (z) = m
+···+ + a m + a m+1 (z − z 0 ) + . . .
(z − z 0 ) (z − z 0 )
∞ g (m−1) (z 0 )
a n (z − z 0 )n . Therefore, Res f , z 0 = a m−1 =
P ¡ ¢
where g (z) = .
n=0 (m − 1)!

T HEOREM 1 (Residue Theorem). Let Ω be an open subset of C and S be a finite subset

of Ω. Suppose f : Ω\S −→ C is a bounded function. Let γ be a null homotopic closed curve
in Ω. Then
1
Z k
X ¡ ¢ ¡ ¢
f (z)d z = Wγ z j Res f , z j
2πi γ j =1
where S = {z 1 , z 2 , . . . , z k } and Wγ is the winding number.

P ROOF. Define
k ¡ ¢
X Res f , z j
g (z) = .
(z − z j )
j =1
Note that g is holomorphic on Ω \ S.
Fix z j and r > 0 be such that D(z j , r ) ⊆ Ω and D(z j , r ) ∩ S = {z j }. We have
¡ ¢
Res f , z j
g (z) = + g 1 (z)
(z − z j )

326
4

where g 1 is the functions consists of the remaining terms of g . Then g 1 is holomorphic

¡ ¢ ¡ ¢
on D(z j , r ). Note that Res g , z j = Res f , z j .
(Refer Slide Time: 18:20)

R
Define F (z) = f (z) − g (z). If γ F (z)d z = 0, then

Z
¡ ¢
f − g (z)d z = 0
γ
1 1
Z Z
=⇒ f (z)d z = g (z)d z
2πi γ 2πi γ
Z X k ¡ ¢
1 Res f , z j
= dz
2πi γ (z − z j )
j =1
k
dz
µ ¶
¢ 1
X Z
¡
= Res f , z j
j =1
2πi γ (z − z j )
k
X ¡ ¢ ¡ ¢
= Wγ z j Res f , z j .
j =1

R
Hence the proof will be completed if we manage to show that γ F (z)d z = 0.

327
 5

Let us try to figure out the Laurent series expansion of F . Since F (z) = f (z) − g (z),
then in D(z j , r ) \ {z j },
∞
a n (z − z j )n
X
f (z) =
n=−∞

and
¡ ¢
Res f , z j ∞ a −1 ∞
b n (z − z j )n = b n (z − z j )n
X X
g (z) = + +
(z − z j ) n=0 (z − z j ) n=0
where the neighborhood D(z j , r ) of z j is same as above.
On D(z j , r ) \ {z 0 }, we have
∞ a −1 ∞
a n (z − z j )n − b n (z − z j )n
X X
F (z) = +
n=−∞ (z − z j ) n=0
∞ ∞
a −n (z − z j )−n + a n′ (z − z j )n .
X X
=
n=2 n=0

Now, let us define a function G on D(z j , r ) \ {z j } by

∞ a (z − z )−n+1 ∞ a ′ (z − z )n+1
X −n j X n j
G(z) = + .
n=2 −n + 1 n=0 (n + 1)

Then, G ′ (z) = F (z) and hence on D(z j , r ) \ {z j }, F has an anti-derivative. Therefore, if C

is a closed curve in D(z j , r ) \ {z 0 }, we have
Z
(1) F (z)d z = 0.
C

If z ∈ Ω \ S, then r 1 > 0 be such that D(z, r 1 ) ⊆ Ω and D(z, r 1 ) ∩ S = ∅. Then by Cauchy’s

theorem, for any closed curve C 1 in D(z 0 , r 1 ),
Z
(2) F (z)d z = 0.
C1

Let γ : [a, b] −→ Ω \ S be null-homotopic in Ω. That is, there exists H : [0, 1] × [a, b] −→ Ω

such that γ0 = γ and γ1 = γz0 and such that γs is closed curve for every s ∈ [0, 1], where
γz0 (t ) = z 0 for each t ∈ [a, b].
(Refer Slide Time: 30:55)

328
6

Since [0, 1] × [a, b] is compact, so is H ([0, 1] × [a, b]) in Ω. Let

U′ := {D(z, r z ) : z ∈ H ([0, 1] × [a, b]) \ S, D(z, r z ) ⊆ Ω and D(z, r z ) ∩ S = ∅}

[

{D(z j , r j ) : 1 ≤ j ≤ k, z j ∈ H ([0, 1] × [a, b]), D(z j , r j ) ∩ S = {z j } and D(z j , r j ) ⊆ Ω}.

Then U′ will be an open cover of H ([0, 1]×[a, b]). By compactness, let U = {U1 ,U2 , . . . ,Un }
be a finite subcover and let ϵ be the Lebesgue number corresponding to U.
By uniform continuity, there exists δ > 0 such that

ϵ
|H (s, t ) − H (s ′ , t ′ )| < whenever |s − s ′ | < δ and |t − t ′ | < δ.
4

Refer Slide Time: 37:26

329
 7

Consider the partitions P 1 : 0 = a 0 < s 1 < · · · < s n = 1 of [0, 1] and P 2 : a = t 0 < t 1 < · · · <
t m = b of [a, b] such that the partition size is less than δ.
For every pair (i , j ) such that 1 ≤ i ≤ n and 1 ≤ j ≤ m,

C i j = γH (si ,t j )→H (si ,t j −1 )→H (si −1 ,t j −1 )→H (si ,t j )

does not intersect S.

By a similar argument as given in the Cauchy’s theorem, if i ntC i , j F (z)d z = 0 then
Z Z m X
X n Z
(3) F (z)d z = F (z) = F (z)d z.
γ=γ0 γz0 =γ1 j =1 i =1 C i , j
R
Claim: F (z)d z = 0.
Ci , j

Note that ¯C i , j ¯ < ϵ. Thus diam C i , j < ϵ and we have Cˆi , j ⊂ U for some U ∈ U.
¯ ¯ ¡ ¢

Hence by (1) and (2) and the fact that C i , j ∩ S = ∅, we have

Z
F = 0.
Ci , j

Therefore by (3), we have

Z Z
F (z) = F (z)d z.
γ γz 0
□

330
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 39
Argument Principle

We have not defined what is meant by the logarithm of a complex number however if we
have a logarithm of complex number say log z, then it is desirable that the derivative of
log z is equal to 1z and by the chain rule we would like to have the derivative of log f (z)
¡ ¢

f ′ (z)
to be f (z) .

D EFINITION 1 (Logarithmic Derivative). Let f : Ω −→ C be a holomorphic function.

f ′ (z)
We will define the log-derivative of f to be the meromorphic function .
f (z)
(Refer Slide Time: 01:40)

331
2

E XAMPLE 1.

(1) Let f : Ω −→ C and g : Ω −→ C be any two holomorphic functions on Ω. Then

the log-derivative of the product f g will be,
¡ ¢′
f g (z) f ′ (z)g (z) f (z)g ′ (z) f ′ (z) g ′ (z)
¡ ¢ = + = + .
f g (z) f (z)g (z) f (z)g (z) f (z) g (z)

If g is non-zero on Ω, then we can talk about the log-derivative of the quotient

f
g and which will be,
³ ´′
f
g
(z) f ′ (z)g (z) − f (z)g ′ (z) f ′ (z) g ′ (z)
³ ´ = = − .
f f (z)g (z) f (z) g (z)
g (z)

(2) Consider p(z) = a(z − z 1 )d1 . . . (z − z n )dn and q(z) = b(z − w 1 )e 1 . . . (z − w m )e m ,

p(z)
where z i ’s are distinct and w j ’s are distinct. Let R(z) = q(z)
. Then

R ′ (z) d1 d2 dn e1 e2 em
= + +···+ − − −···− .
R(z) (z − z 1 ) (z − z 2 ) (z − z n ) (z − w 1 ) (z − w 2 ) (z − w m )

(3) Let f : Ω \ S −→ C be a holomorphic function.

f′
• If z 0 ∈ Ω\S is such that f (z 0 ) ̸= 0, then f
is holomorphic in a neighborhood
of z 0 .
• If z 0 ∈ Ω \ S and f (z 0 ) = 0 with order m, then f (z) = (z − z 0 )m g (z) where
g (z 0 ) ̸= 0 and the log-derivative of f will be

f ′ (z) m(z − z 0 )m−1 g (z) (z − z 0 )m g ′ (z)

= +
f (z) (z − z 0 )m g (z) (z − z 0 )m g (z)
m g ′ (z)
= + .
(z − z 0 ) g (z)
g′
Note that g
is holomorphic on z 0 .
f′
• If z 0 ∈ S and z 0 is a removable singularity of f , then f behaves as above.
• If z 0 ∈ S and z 0 is a pole of order m, then

g (z)
f (z) =
(z − z 0 )m

332
 3

where g (z) ̸= 0 in a neighborhood of z 0 and we have

g ′ (z) mg (z)
f ′ (z) (z−z 0 )m − (z−z m+1 g ′ (z) m
0)
= g (z)
= − .
f (z) g (z) (z − z 0 )
(z−z 0 )m

T HEOREM 2 (Argument Principle). Let Ω be an open set in C and f be a meromorphic

function defined on Ω such that f has zeroes of order d 1 , . . . , d n at z 1 , . . . , z n respectively, af-
ter removing the removable singularities, and f has poles of order e 1 , . . . , e m at w 1 , . . . , w m
respectively. Let γ be a closed curve which is null-homotopic in Ω such that zeroes and
poles don’t lie on γ.Then
1
Z
f ′ (z) X n Xm
dz = d i Wγ (z i ) − e i Wγ (w j )
2πi γ f (z) i =1 j =1

where Wγ is the winding number.

(Refer Slide Time: 20:20)

P ROOF. We have
(z − z 1 )d1 . . . (z − z n )dn
f (z) = g (z)
(z − w 1 )e 1 . . . (z − w m )e m
where g has neither zeros nor poles in Ω.

333
4

Then,

f ′ (z) d1 dn e1 em
Z µ ¶
1 1
Z
dz = +···+ − −−···− d z+
2πi γ f (z) 2πi γ (z − z 1 ) (z − z n ) (z − w 1 ) (z − w m )
1 g ′ (z)
Z
dz
2πi γ g (z)
n
X m
X
= d i Wγ (z i ) − e j Wγ (w j ).
i =1 j =1

E XAMPLE 3. Let γ be a contour in Ω and f be a holomorphic function on Ω. Define

σ : [a, b] −→ C given by σ(t ) = f ◦ γ(t ). Then

f ′ γ(t ) γ′ (t )d t
¡ ¢
1 f ′ (z) 1 b
Z Z
dz =
f γ(t )
¡ ¢
2πi γ f (z) 2πi a
¢′
f ◦ γ (t )d t
¡
1
Z b
=
f ◦ γ (t )
¡ ¢
2πi a

1 dz
Z
=
2πi σ z

= Wσ (0).

Suppose z 1 , z 2 , . . . , z n be the zeroes of f on Ω. If γ is a null-homotopic simple closed

1 R f′
curve which does not pass through z i ’s, then gives the number of zeroes of f
2πi γ f
counting multiplicities in H ([0, 1] × [a, b]), where H is the homotopy between γ and the
constant curve.

(Refer Slide Time: 27:40)

334
 5

T HEOREM 4 (Stability of zeroes). Let Ω be an open set and γ0 : [a, b] −→ C be a null-

homotopic closed curve in Ω. Suppose H : [0, 1] × [a, b] −→ Ω be a homotopy of closed
curves in Ω from γ0 to γ1 . Suppose f 0 and f 1 are holomorphic functions on Ω such that
there exists a continuous function F : [0, 1]×Ω −→ C such that F (0, z) = f 0 (z) and F 1 (1, z) =
f 1 (z). Further, for each s ∈ [0, 1], F (s, H (s, t )) ̸= 0 (i.e., F (s, .) does not vanish on H (s, t ) =
γs (t )). Then the number of zeroes of f 0 in the interior of γ0 counting multiplicities is the
same as the number of zeroes of f 1 in the interior of γ1 .

(Refer Slide Time: 32:23)

335
6

P ROOF. The number of zeroes with counting multiplicities of f 0 in the interior of γ0

is given by

1 f 0′ (z)
Z
(1) d z = Wσ0 (0)
2πi γ0 f 0 (z)

where σ0 = f 0 ◦ γ0 .
Similarly, the number of zeroes with multiplicities of f 1 in the interior of γ1 is given
by

1 f 1′ (z)
Z
(2) d z = Wσ1 (0)
2πi γ1 f 1 (z)

where σ1 = f 1 ◦ γ1 .
Define G : [0, 1]×[a, b] −→ C\{0} given by G(s, t ) := F (s, H (s, t )). Then G is a homotopy
of closed curves in C \ {0} from σ0 to σ1 which gives us Wσ0 (0) = Wσ1 (0) and hence the
result. □

E XAMPLE 5. Let f (z) = z 2 and γ(t ) = e i t for t ∈ [0, 2π]. Then z = 0 is the only zero of f
and the number of zeroes of f , counting multiplicities, is 2. Now consider f ϵ (z) = z 2 + ϵ.

336
 7
p p
Then the zeroes of f are z = i ϵ and z = −i ϵ. Note that the homotopy here will be the
constant homotopy, H (s, t ) = γ(t ).

T HEOREM 6 (Rouche’s Theorem). Let γ be a closed curve which is null-homotopic in

Ω. Suppose f and g are holomorphic in Ω and |g (z)| < | f (z)| on γ. Then f and f + g have
the same number of zeroes counting multiplicities on the interior of Γ([0, 1]×[a, b]) where
Γ is the null-homotopy from γ to a constant path.

P ROOF. Define F (s, t ) = f (t ) + sg (t ). Also define H (s, t ) = γ(t ). By the stability of

zeroes, since F (s, H (s, t )) ̸= 0, we have the number of zeroes of f and f + g in the interior
of γ are equal upto multiplicity. □

337
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 40
Problem Session

P ROBLEM 1. Compute the integral

2π dθ
Z

0 a + cos θ

where a > 1.

it θ
S OLUTION 1. Let γ(tµ ) = e¶ for t ∈ [0, 2π]. Now, for z ∈ γ, we have |z| = 1. If z = e ,
z + z̄ 1 1
then cos θ = = z+ .
2 2 z
(Refer Slide Time: 01:50)

338
2

2π dθ 1 2π i eiθd θ
Z Z
=
a + cos θ i
³ ³ ´´
0 0 a + 12 e i θ + e1i θ e i θ

1 2π γ′ (t )d t
Z
= ³ ³ ´´
i 0 a + 12 γ(t ) + γ(t1
γ(t )
)

1 dz
Z
=
a + z + 1z z
1
¡ ¡ ¢¢
i γ 2
2 dz
Z
= .
i γ z 2 + 2az + 1
p p
We know that z 2 +2az +1 has roots at −a ± a 2 − 1. Note that, since a > 1, −a − a 2 − 1 <
p p
−1 and hence −a − a 2 − 1 ̸∈ D. Also it is an easy verification that −1 < −a + a 2 + 1 < 0.
By residue theorem, we have

1 dz
Z p ³ p ´
= W (−a − a 2 − 1)Res f , −a − a 2 −1 +
2 γ
2πi γ z + 2az + 1
p ³ p ´
Wγ (−a + a 2 − 1)Res f , −a + a 2 − 1
³ p ´
= Res f , −a + a 2 − 1 .

(Refer Slide Time: 06:40)

339
 3

Now, on D

  
1 1 1 1
= ³ p p ´ p ´− p ´
z 2 + 2az + 1
´ ³ ³ ³
−a − a 2 − 1 − −a + a 2 − 1 z − −a + a 2 − 1 z − −a − a 2 − 1
 
1 1
= p  ³ p ´ + h(z)
2 a 2 − 1 z − −a + a 2 − 1

³ p ´ 1
where h is a function holomorphic on D. Hence, Res f , −a + a 2 − 1 = p .
2 a2 − 1
Therefore,

2π dθ dz dz
µ ¶
2 1 2π
Z Z Z
= = 4π = p .
0 a + cos θ i 2
γ z + 2az + 1
2
2πi γ z + 2az + 1 2 a2 − 1

P ROBLEM 2. Compute the integral

Z ∞ 1
¡ ¢2 d x.
0 1 + x2

1
S OLUTION 2. Since ¡ ¢2 is an even function,
1 + x2

Z ∞ 1 1
Z ∞ 1 1
Z R 1
¢2 d x = ¢2 d x = lim ¢2 d x.
2 2 R→∞
¡ ¡ ¡
0 1 + x2 −∞ 1 + x 2 −R 1 + x2

1 1
Consider the function f (z) = ¡ ¢2 = .
1 + z2 (z + i )2 (z − i )2
(Refer Slide Time: 12:28)

340
4

Consider C to be the curve γ−R→R + γ, where γ(t ) = Re i t for t ∈ [0, 2π] and R ∈ R+ .
Then,
à !
1 1 1
Z
¢2 d z = Res ¡ ¢2 , i .
2πi
¡
C 1 + z2 1 + z2
à !
1
Let us now try to find out what will be Res ¡ ¢2 , i .
1 + z2
We have,
µ ¶
1 1 1
¢2 = .
(z − i )2 (z + i )2
¡
1 + z2
Suppose
1 ∞
b n (z − i )n .
X
=
(z + i )2 n=0
Then
1 b0 b1
¢2 = 2
+ + h(z)
(z − i ) (z − i )
¡
1 + z2
where h is a holomorphic function on i . Therefore,
à !
1
Res ¡ ¢2 , i = b 1 .
1 + z2

341
 5

1
Since b 1 is the coefficient of (z − i ) in the power series expansion of around i ,
(z + i )2
¶¯
d
µ
1 ¯ 1
b1 = ¯ = .
d z (z + i )2 z=i 4i
¯

Thus,
1 1 1
Z
¢2 d z = .
2πi 4i
¡
C 1 + z2
Now, notice that

1 1 1
Z Z Z
¡ ¢2 d z = ¡ ¢2 d z + ¡ ¢2 d z
C 1 + z2 γ−R→R 1 + z2 γ 1 + z2

and also
Z
1
Z R 1
¡ ¢2 d z = ¡ ¢2 d x.
γ−R→R 1 + z 2 −R 1 + x2
1 1
R
Now, let us concentrate on the term I = 2πi γ d z.
¯ (1+z 2 )2
¯
¯ 2
¯ 2
¯ 1 ¯ 1
By triangle inequality, we have ¯1 + z ¯ > R − 1 and hence ¯ ¡ < .
¯ ¯
¯ 1 + z 2 2 ¯ R 4 − 2R 2 + 1
¢ ¯

Therefore,
¯ ¯
¯ 1 Z 1 ¯ 2πR 1 1
|I | = ¯ d z¯ < · 4 =
¯ ¯
¯ 2πi γ ¡1 + z 2 ¢2 ¯ 2π R − 2R + 1 R − 2R + R1
2 3

and we have |I | −→ 0 as R −→ ∞.
Hence taking the limit as R −→ ∞,

1
Z ∞dx 1
¢2 = .
2πi 4i
¡
−∞ 1 + x 2

Hence
∞ 1 1 ∞ 1 π
Z Z
¢2 d x = ¢2 d x = .
2 4
¡ ¡
0 1 + x2 −∞ 1 + x2

P ROBLEM 3. Evaluate the integral

∞ ei x
Z
¡ ¢2 d x.
−∞ 1 + x2

342
6

S OLUTION 3. Here the problem is similar to that of Problem 2. Hence the setup for
solving this problem is also similar to that of Problem 2 that we consider the curve C =
γ−R→R + γ, for γ(t ) = Re i t for t ∈ [0, 2π].
ei z ei z
µ ¶
1 ¡ ¢
If f (z) = ¡ ¢2 = , then Res f , i will be the coefficient of (z − i )
1 + z2 (z − i )2 (z + i )2
ei z
in the power series expansion of around z = i ,
(z + i )2

ei z
¶¯
¡ ¢ d
µ
¯ −i
Res f , i = 2
¯ = .
d z (z + i ) z=i 2e
¯

Hence,
R ei x π
µ ¶
−i
Z
lim ¢2 d x = 2πi = .
2e e
¡
R→∞ −R 1 + x2

(Refer Slide Time: 23:00)

P ROBLEM 4. Let Ω be an open set containing D and let f be non-constant holomor-

phic function on Ω which satisfies | f (z)| = 1 whenever |z| = 1. Then D is contained in
f (Ω).

S OLUTION 4. Claim: There exists z 0 ∈ D such that f (z 0 ) = 0.

343
 7

1
If not, then g (z) = f (z)
is holomorphic in a neighborhood of D. Since | f (z)| = 1 for
|z| = 1, by the maximum modulus principle, we have

(1) | f (z)| ≤ 1 ∀ z ∈ D.

For |z| = 1, note that |g (z)| = 1. By a similar argument as above, we have |g (z)| ≤ 1. Hence

(2) | f (z)| ≥ 1 ∀ z ∈ D.

By (1) and (2), we have | f (z)| = 1 for z ∈ D. By open mapping theorem, f (D) is open and
thus | f (z)| = 1 is a contradiction. Hence there exists z 0 ∈ D such that f (z 0 ) = 0.
Let w ∈ D. Define a constant function g on Ω given by g (z) = −w. Then |g (z)| < | f (z)|
on the circle |z| = 1. By Rouche’s theorem, f and f + g have equal number of zeroes in D
counted with multiplicity.
We know that f vanishes on D. Hence there exists z 1 ∈ D such that ( f + g )(z 1 ) = 0.
That is, f (z 1 ) = w. Since w ∈ D was arbitrary, we have D ⊆ f (Ω).

344
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 41
Branch of the Logarithm

When we proved the argument principle, we defined the notion of the log-derivative of a
f′
holomorphic function f as the meromorphic function f
. We even checked that the log-
derivative satisfies some of the properties, which the complex logarithm is expected to
satisfy. We did all that without really defining what is meant by the complex logarithm in
the complex setting. We are familiar with the same notion in the real setting. However,
the notion has to be made precise in the complex setting.
(Refer Slide Time: 02:00)

Recall that the logarithm ln : (0, ∞) −→ R to be the function which inverts the ex-
ponential function on R. This definition cannot be generalized to define the complex
logarithm. Notice that the exponential function exp : C −→ C \ {0} satisfy the property
that

e z = e w ⇐⇒ z = w + 2πi k for k ∈ Z.

Thus exp is not injective.

For z ∈ C∗ = C \ {0}, let us denote log(z) to be the set

log(z) = {w ∈ C : exp(w) = z}.

1

345
2

If w = ln |z| + i θ for θ ∈ arg(z) = θ ∈ R : z = |z|e i θ , then e w = |z|e i θ = z.

© ª

That is an alternate way of looking at it, but this is not something which we will be
satisfied with. We would like to really get hold of an ‘honest’ function f , which inverts
the exponential function. Let us try to define one such function and in order to do that
let us revisit from the real analysis setting; what we did, when we encountered real val-
ued functions?
Let us consider a function on R which is not injective. One such function is f : R −→
p p
[0, ∞) given by f (x) = x 2 . Given x > 0, there exist two real numbers x, − x such that
¡p ¢2 ¡ p ¢2
x = − x = x. By picking one of the square roots, we worked freely with the square
root.
By a branch of the square root, we mean a function g : [0, ∞) −→ R such that
¡ ¢2
g (x) = x.
Can we have a g which is continuous?
p p
Define g (x) = x, where x > 0, then g is continuous branch of the square root. We
p p
could also define other branches as well. For example, define g (x) = − x where x > 0,
then also g is continuous branch of the square root.
Observe that, we could also define a branch of the square root which is not continu-
ous. For example, define,
p

 x

0 ≤ x ≤ 10
g (x) :=
−px

x > 10

then g will be a branch of the square root, but g is not a continuous function as g has a
discontinuity at x = 10.

D EFINITION 1 (Branch of the logarithm). Let Ω be an open connected subset of C∗ .

By a branch of the logarithm on Ω, we mean a function f : Ω −→ C such that

for each z ∈ Ω
¡ ¢
exp f (z) = z

346
 3

Let Ω = C \ {0} = C∗ . Suppose f is a branch of logarithm on C∗ . That is f : C∗ −→ C be

¡ ¢
a function such that exp f (z) = z.
(Refer Slide Time: 13:44)

One can ask this question: Can f be a holomorphic function?

¡ ¢
If f were holomorphic, then by chain rule applied to the equation exp f (z) = z, we
have
d ¡
exp f (z) = f ′ (z) exp f (z) = z f ′ (z)
¡ ¢¢ ¡ ¢
1=
dz
1
=⇒ f ′ (z) = .
z
1
Let γ(t ) = e i t for t ∈ [0, 2π] in C∗ . Since f is the anti-derivative of z in C∗ , by the funda-
mental theorem of calculus, we have
Z
f ′ (z)d z = 0.
γ

But,
dz
Z Z
′
f (z)d z = = 2πi ̸= 0.
γ γ z
Hence if we have a holomorphic branch of the logarithm on C∗ , then we have a contra-
diction.
Let Ω be an open connected subset of C∗ . Further let Ω be simply connected. Con-
1
sider the function z which is holomorphic on Ω. Let γ be any simply closed curve. By
Cauchy’s theorem,
dz
Z
= 0.
γ z

347
4

By the fundamental theorem of calculus, there exists a function f : Ω −→ C such that

f ′ (z) = 1z .
Let f be an anti-derivative such that given z 0 and w 0 with e z0 = w 0 , we have f (w 0 ) =
z 0 . Since f is holomorphic on Ω, by chain rule
¡ ¢
d ¡ ¡ ¢¢ ′
¡ ¢ exp f (z)
exp f (z) = f (z) exp f (z) = .
dz z

Now, by the quotient rule

¡ ¢¶ ¡ ¢ ¡ ¢
d exp f (z) exp f (z) − exp f (z)
µ
= =0 ∀ z ∈ Ω.
dz z z2
¡ ¢
exp f (z)
=⇒ = C , a constant.
z
For z = w 0 , we have
¡ ¢
exp f (w 0 ) = exp (z 0 ) = w 0 .

Hence ¡ ¢
exp f (w 0 )
= C = 1.
w0
¡ ¢
=⇒ exp f (z) = z.

(Refer Slide Time: 26:47)

Hence f is a branch of the logarithm on Ω.

348
 5

E XAMPLE 1. Let Ω = C \ {x ∈ R : x ≤ 0}. Any point of Ω can be connected to 1 be a

straight line in Ω. If γ is a closed curve in Ω, then

H (s, t ) = (1 − s)γ(t ) + s

is a homotopy of closed curves in Ω from γ to the constant curve γ1 . Hence Ω is simply

connected.
Therefore, there exists a holomorphic function f : C\{x ≤ 0} −→ C such that exp f (z) =
¡ ¢

z.

Suppose f and g are two branches of the logarithm on Ω which are continuous func-
tions. We know that exp f (z) = exp g (z) for each z ∈ Ω. That is, for each z ∈ Ω, we have
¡ ¢ ¡ ¢

exp f (z) − g (z) = 1. Hence f (z) = g (z) = 2πi k for k ∈ Z. Since f and g are continuous
¡ ¢

functions on Ω, we have f − g is continuous on Ω and hence f − g (Ω) is connected.

¡ ¢ ¡ ¢

Note that the only connected subsets of {2πi k : k ∈ Z} are singletons (Why?). Since the
image of f −g will be connected and are 2πi k for k ∈ Z, we must have a fixed k 0 ∈ Z such
that
¡ ¢
f − g (Ω) = 2πi k 0 .

That is,

f (z) = g (z) + 2πi k 0 for some fixed k 0 ∈ Z.

Recall that Arg(z) = θ, where θ ∈ (−π, π] and z = |z|e i θ .

D EFINITION 2 (Standard branch of Logarithm). Let Ω = C \ {x ∈ R : x ≤ 0}. Define the

standard branch of the logarithm to be

Log(z) = ln |z| + i Arg(z).

E XERCISE 2. Log(z) is continuous on C \ {x ∈ R : x ≤ 0} and also exp Log(z) = z.

¡ ¢

349
6

Hence Log(z) is a continuous branch of the logarithm on Ω.

Since there exists a holomorphic branch of the logarithm f on Ω, by the observation
made above, we have

Log(z) = f (z) + 2πi k 0 for some fixed k 0 ∈ Z.

Hence Log(z) is a holomorphic function.

Notice that C \ {x ∈ R : x ≤ 0} is just one example of the simply connected domain in
C∗ for which we have a holomorphic branch of the logarithm. We can get hold of another
examples of simple domains in C∗ that admits a continuous branch of the logarithm.
(Refer Slide Time: 40:28)

If Ω = D(1, 1), then Ω is simply connected and contained in C∗ . Hence in this do-
main also, we have a continuous branch of the logarithm and moreover, since D(1, 1) is
contained in C \ {x ∈ R : x ≤ 0} as well, the restriction of the holomorphic branch of the
logarithm Log(z) to D(1, 1) will again turn out to a holomorphic function on D(1, 1).
If Ω = D(−2, 1), then Ω is a simply connected domain in C∗ and hence there exists a
holomorphic branch of the logarithm in Ω. However, the Log(z) (which can be defined
on C∗ ) is not even continuous in Ω as Log(z) is discontinuous on {x ∈ R : x ≤ 0} ∩ Ω.
Thus the point here to notice is that, when we talk about the holomorphic branch of
the complex logarithm, it matters which domain we are considering.

350
 7

Recall that, given w ∈ C∗ , there exists n roots to the equation z n = w.

We would like to obtain a holomorphic branch of the n t h root function. That is, we
want to address the question: does there exists a holomorphic function g : Ω −→ C such
¢n
that g (z) = z for each z ∈ Ω, where Ω ⊆ C∗ ?
¡

Let Ω be a simply connected domain. and let f be a holomorphic branch of the

logarithm on Ω. Define
f (z)
µ ¶
g (z) = exp .
n
Being a composition of holomorphic functions on Ω, g is holomorphic on Ω. Also,
f (z) n
µ µ ¶¶
¡ ¢n ¡ ¢
g (z) = exp = exp f (z) = z
n
and this is precisely what we are trying to get hold off.
Hence, on simply connected domains, we have a holomorphic branch of n t h root
function.

351
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 42
Automorphisms of the Unit Disk

(Refer Slide Time: 01:01)

We have already seen the maximum modulus principle for holomorphic functions. It
states that if f : Ω −→ C is a holomorphic function on an open set Ω ⊆ C and K ⊆ Ω is a
compact set, then

sup | f (z)| ≤ sup | f (z).

z∈K z∈∂K

Here we will be giving a phenomenal application of the maximum modulus principle in

proving the very famous Schwarz’s lemma and thereafter, we will give a characterization
of the automorphisms of the unit disk. Automorphisms are holomorphic functions from
the unit disk to itself, which has a holomorphic inverse.

L EMMA 1 (Schwarz’s Lemma). Let f : D −→ D be a holomorphic function such that

f (0) = 0. Then | f (z)| ≤ |z| for every z ∈ D and | f ′ (0)| ≤ 1.
Furthermore, if | f (z)| = |z| for some z ∈ D \ {0} or if | f ′ (0)| = 1, then there exists λ ∈ C with
|λ| = 1 such that f (z) = λz.

f (z)
P ROOF. Since f (0) = 0, we have g (z) = z has a removable singularity at 0.
1

352
2

Let 0 < r < 1. By the maximum modulus principle on D(0, r ) applied to g , we have

sup | f (z)|
|z|=r
|g (z)| ≤ ∀ z ∈ D(0, r )
r
1
≤ .
r

Since this is true for 0 < r < 1, taking limit r → 1, we have |g (z)| ≤ 1 for each z ∈ D(0, 1).
That is,

| f (z)|
≤1 for every z ∈ D(0, 1).
|z|

Hence,

| f (z) ≤ |z| for every z ∈ D(0, 1).

Notice that f ′ (0) = g (0) and since |g (0)| ≤ 1 we have | f ′ (0)| ≤ 1.

If |g (z)| = 1 for some z ∈ D, then g is a constant function. That is,

g (z) = λ, for every z ∈ D and for some λ such that |λ| = 1.

Hence, | f (z)| = λz for every z ∈ D. Similarly, if |g (0)| = 1, we have f (z) = λz for each z ∈ D
and some λ such that |λ| = 1. □

Now, let us consider some special functions on the unit disk D. For α ∈ D, define
ϕα : D −→ C given by,

z −α
ϕα (z) := .
1 − ᾱz
1
Notice that ϕα is a Möbius transformation with a pole on ᾱ
. But since α ∈ D, we have
1
ᾱ ̸∈ D. Hence this function is holomorphic in a neighborhood of the closure of the unit
disk for each α ∈ D.
(Refer Slide Time: 14:15)

353
 3

Now,
z+α
1+ᾱz
−α
ϕα ϕ−α (z) =
¡ ¢
z+α
1 − ᾱ 1+ ᾱz
z + α − α − |α|2 z
=
1 + ᾱz − ᾱz − |α|2
1 − |α|2 z
¡ ¢
= ¡ ¢
1 − |α|2

= z.

Hence, ϕ−α and ϕα are holomorphic functions which are inverses of each other. Observe
that,
¯ ³ ´¯ ¯¯ e i θ − α ¯¯
¯ϕα e i θ ¯ = ¯¯
¯ ¯ ¯
1 − ᾱe i θ ¯
¯ iθ
¯e − α¯
¯
= ¯ i θ ¯ ¯ −i θ
− ᾱ¯
¯
¯e ¯ ¯e
¯ iθ
¯e − α¯
¯
= ¯ iθ
¯e − α¯
¯

= 1.

That is, ¯ϕα (z)¯ = 1 for each z ∈ ∂D. By maximum modulus principle, we have ¯ϕα (z)¯ ≤ 1
¯ ¯ ¯ ¯

for each z ∈ D. That is image of ϕα is contained in D, also, since it is not a constant func-
tion ϕα (D) will be an open subset of D. By a very similar argument, we have ¯ϕ−α (z)¯ ≤ 1
¯ ¯

for each z ∈ D. Combining these two facts, we have ϕα (D) = D.

354
4

That is, ϕα : D −→ D is a holomorphic map of D to itself which has a holomorphic

inverse.

D EFINITION 1 (Automorphism). We say that a function f : Ω −→ Ω is an automor-

phism if f is holomorphic and has a holomorphic inverse.

By what we have done above, ϕα is an automorphism of D for each α ∈ D. Observe

that ϕα has some nice properties like,

ϕα (α) = 0, ϕα (0) = −α, ϕ−α (0) = α.

Also,
1
ϕ′α (0) = 1 − |α|2 ϕ′α (α) = .
1 − |α|2
Now, let us try to answer the following question:
Question: Let f : D −→ D be a holomorphic function such that f (α) = β, where α, β ∈
D. What is the maximum value of | f ′ (α)|?
That is, we are asking for sup | f ′ (α)| : f : D −→ D holomorphic and f (α) = β .
© ª

Define g (z) = ϕβ ◦ f ◦ ϕ−α (z). Then g is a holomorphic function, g : D −→ D and

g (0) = 0. Then by Schwarz’s lemma, ¯g ′ (0)¯ ≤ 1.
¯ ¯

Check that, by chain rule,

g ′ (0) = ϕ′β (β) f ′ (α)ϕ′−α (0).

Then,
¯ ′ ¯
¯g (0)¯ = 1 ¯ ′ ¯¡ 2
¢
¯ ¯2 ¯ f (α)¯ 1 − |α| ≤ 1.
1 − ¯β ¯
¯ ¯2
′ 1 − ¯β ¯
=⇒ | f (α)| ≤ .
1 − |α|2
This is the upper bound of what | f ′ (α)| can have in terms of α and β. Let us now check
whether we can attain this value of some function.
(Refer Slide Time: 26:28)

355
 5

If |g ′ (0)| = 1, then by Schwarz’s lemma g (z) = λz for some λ with |λ| = 1. That is,

g (z) = λz

=⇒ ϕβ ◦ f ◦ ϕ−α (z) = λz

=⇒ f (z) = ϕ−β λϕα (z) .

¡ ¢

Hence if we consider the function f : D −→ D given by f (z) = ϕ−β λϕα (z) , then
¡ ¢

¯ ′ ¯ 1 − ¯β ¯2
¯ ¯
¯ f (α)¯ = .
1 − |α|2

T HEOREM 2. Let f : D −→ D be an automorphism. Then there exist α ∈ D and λ ∈ ∂D

such that f (z) = λϕα (z).

P ROOF. Since f is an automorphism, there exists α ∈ D such that f (α) = 0. Let g D −→

D be the holomorphic function such that

for z ∈ D.
¡ ¢
g f (z) = z

By chain rule, we have

(1) g ′ (0) f ′ (α) = 1.

356
6

We know from the previous discussion that,

¯g (0)¯ ≤ 1 − |α|2
¯ ′ ¯

and
1
| f ′ (α) ≤ .
1 − |α|2
Hence
¯ ′
¯g (0) f ′ (α)¯ ≤ 1.
¯

(Refer Slide Time: 32:29)

By observing the equality in (1), it is possible only if the above inequalities are equal-
ities. Then,
1
| f ′ (α)| =
1 − |α|2
and it is possible only if
f (z) = λϕα (z).

357
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 43
Phragmen-Lindelöf Method

Now we will take a closer look at the maximum modulus principle. The Phragmen-
Lindelöf method is, in some sense, a generalization of the maximum modulus principle
to the unbounded domains in the complex plane.
Let Ω be a bounded open connected set in C and f : Ω −→ C be continuous on Ω and
holomorphic on Ω. If f is non-constant, then by the open mapping theorem, there does
not exists z 1 ∈ Ω such that f attains a local maximum at z 1 . Since f is continuous on Ω,
there exists z 0 ∈ Ω such that

| f (z 0 )| = sup | f (z)|.
z∈Ω

Since z 0 ̸∈ Ω, we have z 0 ∈ ∂Ω.

Hence sup | f (z)| = sup | f (z)|. If the supremum is attained at an interior, then f is
z∈Ω z∈∂Ω
constant.
But, when we ask a similar question on an unbounded domain, we do not have such
an answer. In fact, this exact statement, when considered in unbounded domains, is
going to be false. Let us discuss an example to illustrate that this particular statement
will not go through for unbounded domains.
(Refer Slide Time: 05:37)
1

358
2

π
Consider Ω = z ∈ C : −π
2 < Im(z) < 2 . Then notice that Ω is an unbounded set,
© ª

which is an infinite strip that contains all points in between the horizontal ray that
π
passes through i −π
2 and the horizontal ray that passes through i 2 . In this region, con-

sider the function f given by f (z) = exp(exp(z)). Then, on the boundary of this domain,
an element z ∈ ∂Ω will be of the form z = x ± i π2 . Observe that
³ π´
exp x ± i = ±i e x
2
and hence
¯ ³ π ´¯¯ ¯
¯ f x ± i ¯ = ¯exp ±i e x ¯ = 1.
¯ ¡ ¢¯
2
That is, on the boundary, | f | is 1. But,

lim | f (x)| = ∞.
x→∞

Clearly, the maximum modulus principle fails in this particular setup. So, the fact that
Ω is bounded was playing a crucial role.
We dropped the condition that Ω is bounded and we ended up with this kind of a
situation. So, we will ask the next question, what is the next best thing that can be done?

359
 3

That is where the Phragmen-Lindelöf method comes into our picture. But before going
to that, let us describe a couple of application of the method.
When we looked at the Liouville’s theorem, it told us that bounded entire functions
are necessarily constant. But if we impose the following condition: suppose f is entire
and | f (z)| ≤ M +|z|1/2 , then we claim that all entire functions with this growth condition
will also be constant. The proof is actually very similar to how we proved Liouville’s
theorem.
On D(0, R), by Cauchy estimates, we have

¯ (n) ¯ n! M + R 1/2
¡ ¢
¯ f (0)¯ ≤ .
Rn

As R → ∞, we have f (n) (0) = 0 for each n ≥ 1. Since f is an entire function, f has power
series expansion around 0. That is,

∞
an z n
X
f (z) =
n=0

f n (0)
around 0. But, then a n = n! = 0 for each n ≥ 1. Hence f (z) = a 0 .
That is, even with a weaker growth condition like this, we will be able to conclude
the same result as Liouville’s theorem.

T HEOREM 1. Let Ω = {z ∈ C : a < Re(z) < b}. Let f : Ω −→ C be such that f is continu-
ous on Ω and holomorphic on Ω. Suppose ¯ f (z)¯ < B , for some large B > 0, and let
¯ ¯

©¯ ¯ ª
M (x) = sup ¯ f (x + i y)¯ : −∞ < y < ∞ .

Then,

M (x)b−a ≤ M (a)b−x M (b)x−a .

(Refer Slide Time: 13:49)

360
4

The theorem tells us that, for z ∈ Ω, z = x + i y, then

| f (z)| ≤ M (x) ≤ max{M (a), M (b)} = sup | f (z)|.

z∈∂Ω

P ROOF. Let us assume that M (a) = M (b) = 1. For ϵ > 0, define an auxiliary function
h ϵ to be,
1
h ϵ (z) = .
1 + ϵ(z − a)

Note that,

|1 + ϵ(z − a)| > Re(1 + ϵ(z − a)) = 1 + ϵ(x − a) ≥ 1.

Hence,

|h ϵ (z)| ≤ 1 ∀z ∈ Ω.

By our assumption that M (a) = 1 = M (b), for each z ∈ ∂Ω, we have | f (z)| ≤ 1. Hence,

∀ z ∈ ∂Ω.
¯ ¯
¯ f (z)h ϵ (z)¯ ≤ 1

361
 5

Consider the set E = z = x + i y ∈ Ω : |y| ≥ Bϵ , then B

© ª
ϵ|y|
≤ 1. If z = x +i y ∈ E , then observe
that Im(1 + ϵ(z − a)) = ϵy

=⇒ |1 + ϵ(z − a)| ≥ ϵ|y|

1
=⇒ |h ϵ (z)| ≤
ϵ|y|
¯ ¯ B
=⇒ ¯ f (z)h ϵ (z)¯ ≤ ≤ 1.
ϵ|y|
(Refer Slide Time: 23:31)

Let R be the rectangle on C bounded by the lines x = a, x = b, y = Bϵ and y = −B ϵ

.
Observe that ¯ f (z)h ϵ (z)¯ ≤ 1 for z ∈ ∂R and hence by the maximum modulus princi-
¯ ¯
¯ ¯ ¯ ¯
ple, we have ¯ f (z)h ϵ (z)¯ ≤ 1 for z ∈ R. Since ¯ f (z)h ϵ (z)¯ ≤ 1 for z ∈ E , and by the above
observation,
for z ∈ Ω.
¯ ¯
¯ f (z)h ϵ (z)¯ ≤ 1

Now, as ϵ −→ 0, we have f (z)h ϵ (z) −→ f (z). Hence ¯ f (z)¯ ≤ 1.

¯ ¯

That is, with our assumption that M (a) = M (b) = 1, we have

M (x) ≤ 1 =⇒ M (x)b−a ≤ 1 = 1b−x · 1x−a = M (a)b−x M (b)x−a .

Now, let us try to prove the more general case.

362
6

Define

b−z ³z −a
µ ¶ ´
g (z) = exp ln M (a) exp ln M (b) .
b−a b−a

Now, let us see what happens when we look at g on the boundary of Ω.

¯ ¶¯ ¯
b − (a + i y) (a + i y) − a
µ µ ¶¯
¯ ¯¯ ¯
|g (a + i y)| = ¯exp
¯ ln M (a) ¯ ¯exp
¯ ¯ ln M (b) ¯¯
b−a b−a
¯ ³³ y ´ ´¯ ¯ ³ y ´¯
= ¯exp 1 − i ln M (a) ¯ ¯exp i ln M (b) ¯
¯ ¯¯ ¯
b−a b−a
¯ ³ ³ −y ´´¯
= ¯exp ln M (a) + i ln M (a) ¯ · 1
¯ ¯
b−a
= exp(ln(M (a))) = M (a).

Similarly,

|g (b + i y) = M (b).

Now, it is an exercise for the reader to check that g1 is bounded on Ω.

¯ ¯
f ¯ f (z) ¯
Consider g and then we have ¯ g (z) ¯ ≤ 1 for each z ∈ Ω.
From here one could conclude the result in the general case. (Exercise.) □

So, what did we do in this entire theorem? We somehow introduced an auxiliary

function h ϵ and using h ϵ , we reduced the problem partly to a problem of the maximum
modulus principle being applied on some bounded function a bounded domain. And
that is precisely what is referred to as the Phragmen-Lindelöf method.
Let us look at one more example of the Phragmen-Lindelöf method on another un-
bounded domain which now is going to be a horizontal strip rather than a vertical strip.
(Refer Slide Time: 33:22)

363
 7

T HEOREM 2. Let Ω = z ∈ C : − π2 < Im(z) < π2 . Let f : Ω −→ C be continuous on Ω

© ª

and holomorphic on Ω. Suppose | f (z)| ≤ 1 for z ∈ ∂Ω, | f (z)| ≤ exp(A · exp(α|x|)), where
z = x + i y, A < ∞ and α < 1, then | f (z)| ≤ 1 for each z ∈ Ω.

P ROOF. Let β > 0 be such that α < β < 1. For ϵ > 0, define an auxiliary function h ϵ (z)
to be

³ ³ ´´
h ϵ (z) := exp −ϵ e βz + e −βz .

Now, notice that for z = x + i y ∈ Ω,

³ ³ ´´ ³ ´
Re −ϵ e βz + e −βz = −ϵ e βx + e −βx cos βy.

Since −π
2 < y < π2 , and β < 1, we have

³ π´
cos βy > cos β = δ.
¡ ¢
2

364
8

Hence, on Ω, we have
¯ ³ ³ ´ ´¯ ¯ ³ ³ ´ ´¯
βx
|h ϵ (z)| = ¯exp −ϵ e + e −βx
cos βy ¯ ¯exp i ϵ e −βx − e βx sin βy ¯
¯ ¯¯ ¯
³ ³ ´ ´
= exp −ϵ e βx + e −βx cos βy
³ ³ ´ ´
= exp −ϵ e βx + e −βx cos β
³ ³ ´´
< exp −ϵδ e βx + e −βx

<1

Consider the function f h ϵ , then | f (z)h ϵ (z)| ≤ 1 for z ∈ ∂Ω. For z ∈ Ω,

³ ³ ´´
| f (z)h ϵ (z)| ≤ exp(A · exp(α|x|)) exp −ϵδ e βx + e −βx
³ ³ ´´
= exp A · exp(α|x|) − ϵδ e βx + e −βx .

Check that A · exp(α|x|) − ϵδ e βx + e −βx −→ −∞ as x −→ ∞ or x −→ −∞.

¡ ¡ ¢¢

Hence | f (z)h ϵ (z)| < 1 for |x| ≥ x 0 for some x 0 > 0.

(Refer Slide Time: 44:10)

Let R the rectangle with sides x = ±x 0 and y = ± π2 . Then on ∂R, we have | f (z)h ϵ (z)| ≤
1. By the maximum modulus principle, | f (z)h ϵ (z)| ≤ 1 on R.

365
 9

Therefore, | f (z)h ϵ (z)| ≤ 1 for every z ∈ Ω. Since ϵ > 0 was arbitrary, h ϵ (z) −→ 1 as
ϵ −→ 0 for each z ∈ Ω. Hence | f (z)| ≤ 1 on Ω. □

366
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 44
Problem Session

P ROBLEM 1. Let Ω be a simply connected domain and f : Ω −→ C∗ be a function

which is holomorphic on Ω. Then prove that there exists a function h : Ω −→ C such that
h is holomorphic and (h(z))n = f (z).

(Refer Slide Time: 02:29)

f′
S OLUTION 1. Since f is non-vanishing, f is holomorphic on Ω. Then for any closed
curve γ on Ω, we have
f ′ (z)
Z
d z = 0.
γ f (z)
f′
By the fundamental theorem of calculus, there exists an anti-derivative g of f
on Ω.
Let z 0 ∈ Ω and w 0 ∈ C∗ be such that f (z 0 ) = w 0 and g be picked in such a manner
that e g (z0 ) = w 0 (This can be done as the anti-derivative is determined uniquely up to
addition by a constant.).
1

367
2

exp g (z)
Consider the function . Then,
f (z)

d exp g (z) f (z)g ′ (z) exp(g (z)) − exp(g (z)) f ′ (z)

µ ¶
= ¢2
dz f (z)
¡
f (z)
f ′ (z) exp(g (z)) − f ′ (z) exp(g (z))
= ¡ ¢2
f (z)

= 0.

exp g (z) exp g (z 0 ) w 0

Therefore, = constant. For z = z 0 , we have = = 1.
f (z) f (z 0 ) w0
Hence,

exp g (z)
= 1 =⇒ exp(g (z)) = f (z).
f (z)

g (z)
µ ¶
Now, let us define h(z) = exp . Then,
n

(h(z))n = f (z).

P ROBLEM 2. Let f : D −→ D be a holomorphic map with two fixed points. Then prove
that f is identity.

S OLUTION 2. Let α and β be points in D such that f (α) = α and f (β) = β. Define
g : D −→ D, given by

g (z) = ϕα ◦ f ◦ ϕ−α (z),

z −α
where ϕα (z) = . Then g (0) = 0.
1 − ᾱz
(Refer Slide Time: 10:23)

368
 3

Let β′ = ϕα (β). Then,

g (β′ ) = ϕα ◦ f ◦ ϕ−α (β′ )

= ϕα f (β)
¡ ¢

= ϕα (β)

= β′ .

Hence by Schwarz’s lemma, g (z) = λz where |λ| = 1. Since g (β′ ) = β′ , we have λβ′ = β′ .
Hence λ = 1. That is,

g (z) = z

ϕα ◦ f ◦ ϕ−α (z) = z

=⇒ f (z) = ϕ−α ◦ ϕα (z)

= z.

Hence f is the identity.

Before going to the next problem, let us have a definition which will be using shortly. For
z, w ∈ D, define
¯ z −w ¯
ρ(z, w) = ¯ ¯.
¯ ¯
1 − w̄ z

369
4

P ROBLEM 3 (Schwarz-Pick Theorem). Let f : D −→ D be a holomorphic map. Then

ρ f (z), f (w) ≤ ρ(z, w) ∀ z, w ∈ D.

¡ ¢

Furthermore, ¯ ′ ¯
¯ f (z)¯ 1
¯2 ≤ ∀ z ∈ D.
1 − |z|2
¯
1 − ¯ f (z)¯

S OLUTION 3. For z, w ∈ D, define a map g given by,

g (ζ) = ϕ f (w) ◦ f ◦ ϕ−w (ζ).

Then, notice that g : D −→ D is holomorphic and g (0) = 0. By Schwarz’s lemma, we have,

∀ ζ ∈ D.
¯ ¯
(1) ¯g (ζ)¯ ≤ |ζ|

Let z ′ ∈ D be such that z ′ = ϕw (z).

(Refer Slide Time: 18:49)

By (1), ¯g (z ′ )¯ ≤ |z ′ |. Note that,

¯ ¯

g (z ′ ) = ϕ f (w) ◦ f ◦ ϕ−w ϕw (z)

¡ ¢

= ϕ f (w) f (z)
¡ ¢

f (z) − f (w)
= .
1 − f (w) f (z)

370
 5

Then,
¯ ¯
¯ f (z) − f (w) ¯
¯g (z ′ )¯ = ¯ ¯ = ρ f (z), f (w) .
¯ ¯ ¯ ¯ ¡ ¢
¯ 1 − f (w) f (z) ¯

Also,
¯ ¯¯ z − w ¯¯
|z ′ | = ¯ϕw (z)¯ = ¯ ¯ = ρ(z, w).
¯
1−wz

Hence,

ρ f (z), f (w) ≤ ρ(z, w).

¡ ¢

It is an exercise for the reader to check that, if f is an automorphism of D, then the above
inequality is an equality.
Thus, for all z, w ∈ D, we have
¯ ¯
¯ f (z) − f (w) ¯ ¯ z − w ¯
¯≤¯ ¯.
¯ ¯ ¯ ¯
¯
¯ 1 − f (w) f (z) ¯ 1−wz

For z ̸= w,
¯ ¯¯ ¯ ¯
¯ ¯ 1 ¯¯
¯ f (z) − f (w) ¯ ¯¯ 1
¯ ¯¯ ¯≤¯
¯ ¯ ¯.
¯ z −w 1−wz ¯
1 − f (w) f (z) ¯
¯¯

By taking the limit as w −→ z, we have

¯ ′ ¯
¯ f (z)¯ · 1 1
¯2 ≤ .
1 − |z|2
¯
1 − ¯ f (z)¯

P ROBLEM 4. Let f be a non-constant entire function such that | f (z)| = 1 for all |z| = 1.
Describe f .

S OLUTION 4. By the maximum modulus principle, we have | f (z)| ≤ 1 for each z ∈ D.

Claim: f has at least one zero in D.
1
If f does not have a zero in D, then f
is holomorphic on D and continuous on D.
(Refer Slide Time: 27:59)

371
6

¯ ¯
¯ 1 ¯
Note that, ¯ f (z) ¯ = 1 for |z|, and hence by maximum modulus principle,
¯ ¯
¯ 1 ¯
¯ f (z) ¯ ≤ 1
¯ ¯ ∀ z ∈ D.

Hence | f (z)| ≥ 1 on D. Since we already have | f (z)| ≤ 1 on D =⇒ | f (z)| = 1 on D. By open

mapping theorem, f is constant, which is a contradiction. Therefore f has at least one
zero in D.
Claim: f has finitely many zeroes in D.
If f has infinitely many zeroes in D, then the zeroes of f has a limit point by com-
pactness of D. Then, by the identity theorem, we have f ≡ 0, which is a contradiction as
f is non-constant. Hence f has only finitely many zeroes in D.
Let α1 , . . . , αn be zeroes of f of order d 1 , . . . , d n in D. Then, we have

f (z) = (z − α1 )d1 · · · (z − αn )dn h(z),

where h is a non-vanishing holomorphic function defined on D.

z − αj
Consider the automorphism of unit disc ϕα j (z) = . Since ϕα j being an auto-
1 − α¯j z
morphism it has unique zero at α j of order 1. Hence, we have

ϕα j = (z − α j )ψ j (z),
³ ´d j ¡ ¢d ¡ ¢d
where ψ j is non-vanishing on D. Then ϕα j = z − α j j ψ j (z) j .

372
 7

(Refer Slide Time: 35:05)

Define
f (z)
g (z) = ¡ ¢d1 ¡ ¢d .
ϕα1 · · · ϕαn n
Then, note that g has isolated singularity at α1 , . . . , αn .
Since α1 is a zero of f order d 1 , we have
(z − α1 )d1 f 1 (z)
g (z) = ¢d ³¡ ¢d ¢d ´ .
(z − α1 )d1 ψ1 (z) 1 ϕα2 (z) 2 · · · ϕαn (z) n
¡ ¡

Now, it is an exercise for the reader to check that g has a removable singularity at α1 .
Similarly, prove that g has a removable singularity at α j for 1 ≤ j ≤ n.
For |z| = 1, we have
¯ ¯
¯ ¯ ¯¯
¯g (z)¯ = ¯ ¡ f (z) ¯ | f (z)|
¯= ¯ ¯d = 1.
¯
¯ ϕ d d d
· · · ϕαn n ¯ ¯ϕα1 ¯ 1 · · · ¯ϕαn ¯ n
¢ 1
¡ ¢ ¯ ¯
α1

Since g does not vanish in D (why?), by a similar argument above, we have g (z) is a
constant function. Hence g (z) = λ, where |λ| = 1. Thus,
¢d ¢d
f (z) = λ ϕα1 (z) 1 · · · ϕαn (z) n .
¡ ¡

373
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 45
Lifting of maps

Picard’s theorem is one of the classical and powerful results in complex analysis. Very
broadly, Picard’s theorem tells us that if you consider a non-constant entire function,
then its image does not miss too many points. We will make it concrete and prove Pi-
card’s theorem in the due course. However, before that let us develop some machinery
that is needed to indeed prove Picard’s theorem namely that of ’covering spaces’.
First, we will define what is meant by the lifting property of maps and we will prove
some properties of lifts.
(Refer Slide Time: 00:56)

D EFINITION 1. Let X , Y and Z be open subsets of C and let f : Y −→ X and g : Z −→

X be continuous maps. We say that a map g̃ : Z −→ Y is a lift of g with respect to f if
f ◦ g̃ = g on Z .
1

374
2

E XAMPLE 1. Consider the exponential function exp : C −→ C∗ and let Ω be a simply

connected subset of C∗ . Then let i ˜d be a branch of the logarithm holomorphic on Ω.
³ ´
That is, exp If
d (z) = z = I d (z). Hence a continuous branch of the logarithm is a lift of
the identity map with respect to the exponential map.

T HEOREM 2 (Uniqueness of Lifts). Let X , Y be open subsets of C and f : Y −→ X be a

local homeomorphism. (That is, given y 0 ∈ Y , there exists a neighborhood V of y 0 in Y
and U of f (y 0 ) in X such that f ↾V : V −→ U is a homeomorphism onto U.) Let be Z be a
connected open subset of C and g : Z −→ X be a continuous map. Let g̃ 1 and g̃ 2 be lifts of
g with respect to f and suppose g̃ 1 (z 0 ) = g̃ 2 (z 0 ) for some z 0 ∈ Z . Then g̃ 1 ≡ g̃ 2 .

© ª
P ROOF. Let E = z ∈ Z : g̃ 1 (z) = g̃ 2 (z) . Notice that E ̸= ∅ as z 0 ∈ E and also E is closed
¡ ¢−1
since E = g̃ 1 − g̃ 2 (0).
Claim: E is open in Z .
Let z ∈ E . Then g̃ 1 (z) = g̃ 2 (z) = y (say). Let x = f (y). Since f is a local homeomor-
phism, there exists a neighborhood V of y and U of x in Y and X respectively such that
f ↾V : V −→ U is a homeomorphism. Let φ : U −→ V be the inverse of f ↾V .
(Refer Slide Time: 17:25)

375
 3

Since g̃ 1 and g̃ 2 are continuous, we have g̃ 1−1 (V ) ∩ g̃ 2−1 (V ) is open in Z and contains
z. Let W be a neighborhood of z such that W ⊆ g̃ 1−1 (V )∩ g̃ 2−1 (V ). Observe that g̃ 1 (W ) ⊆ V
and g̃ 2 (W ) ⊆ V . Then, by the construction, we have

g ↾W = f ◦ g̃ 1 ↾W = f ↾V ◦g̃ 1 ↾W

and similarly

g ↾W = f ◦ g̃ 2 ↾W = f ↾V ◦g̃ 2 ↾W .

Then by composing with φ, we have

g̃ 1 ↾W = φ ◦ g ↾W = g̃ 2 ↾W .

Hence g̃ 1 = g̃ 2 on W =⇒ W ⊆ E . Therefore E is open in Z . Since Z is connected,

E = Z. □

Till now, we have not addressed the case of holomorphic maps, we have only dealt
with continuous maps. Let us address that issue now.

T HEOREM 3. Let f : Y −→ X be a holomorphic map such that f ′ (y) ̸= 0 for each y ∈ Y .

Let g : Z −→ X be a holomorphic map such that g̃ : Z −→ Y is a lift of g with respect to f .
Then g̃ is holomorphic.

(Refer Slide Time: 24:21)

376
4

P ROOF. Since f ′ (y) ̸= 0 for every y ∈ Y , f is a local homeomorphism which is holo-

morphic with holomorphic inverse.
Let z ∈ Z , x = g (z) and y = g̃ (z). Since f ′ (y) ̸= 0, by inverse function theorem, there
exists a neighborhood V of y and U of x such that f ↾V −→ U is a holomorphic map onto
U with holomorphic inverse φ : U −→ V .
Let W = f −1 (V ). On W , we have g̃ = φ ◦ g . Since g and φ are holomorphic, we have
g̃ is holomorphic on W =⇒ g̃ is holomorphic on Z . □

Let us now try to see what happens when we lift curves. Curves are very special
objects for us. And we would like to see what happens when we are lifting curves under
such good maps like say local homeomorphisms. The uniqueness of the lifting tells us
that if you have if you prescribe the initial point of the lifting then the lifting has to be
unique of a curve. That is, let f : Y −→ X be a local homeomorphism. Suppose γ :
[a, b] −→ X be a curve and γ̃ : [a, b] −→ Y be a lift of γ with respect to f . Then γ̃ is unique
if we prescribe the initial point.

T HEOREM 4. Let f : Y −→ X be a local homeomorphism between open subsets of C.

Let γ0 and γ1 be curves in X from z 1 to z 2 which are homotopic with fixed end points

377
 5

through H : [0, 1] × [a, b] −→ X . Suppose for every s ∈ [0, 1], if γs (t ) = H (s, t ) can be lifted
to a path γ̃s : [a, b] −→ Y with respect to f such that γ̃s (a) = z̃ 1 for every s ∈ [0, 1]. Then γ̃0
and γ̃1 are homotopic with fixed end points in Y .

(Refer Slide Time: 35:07)

P ROOF. Define H e (s, t ) = γ̃s (t ). Let us try to prove this

e : [0, 1] × [a, b] −→ Y given by H

map is continuous, which is the important part.

Let V,U be neighborhoods of z̃ 1 and z 1 respectively such that f ↾V : V −→ U is a
homeomorphism with inverse φ : U −→ V . Since H is continuous, H −1 (U ) is an open
subset of [0, 1] × [a, b].
Check that, by the compactness of [0, 1] and continuity of H , we have an ϵ > 0 such
that H ([0, 1] × [a, a + ϵ)) ⊆ U .
For (s, t ) ∈ [0, 1] × [a, a + ϵ), we have f ◦ γ̃s (t ) = γs (t ). By composing with φ, we have

e (s, t ) = φ ◦ γs (t ) = φ ◦ H (s, t ).
H

Hence H ↾[0,1]×[a,a+ϵ) is continuous. We will prove it is continuous on [0, 1] × [a, b].

Suppose H
e is not continuous. Given s 0 ∈ [0, 1], let t 0 ∈ [a, b] be such that H
e is not

continuous on at (s 0 , t 0 ) and H
e is continuous at (s 0 , t ) for each t < t 0 .

378
6

Let U ′ and V ′ be neighborhoods of γs0 (t 0 ) and γ̃s0 (t 0 ) such that f ↾V ′ : V ′ −→ U ′ is a

homeomorphism and φ′ : U ′ −→ V ′ be its continuous inverse.
Since H is continuous, there exists δ > 0 such that

H ((s 0 − δ, s 0 + δ) × (t 0 − δ, t 0 + δ)) ⊆ U ′ .

(Refer Slide Time: 44:28)

Let t 1 ∈ (t 0 − δ, t 0 ). Then H
e is continuous at (s 0 , t 1 ). Then there exists a δ′ > 0 with

δ′ < δ such that for s ∈ (s 0 − δ′ , s 0 + δ′ ), we have H

e (s, t 1 ) ∈ V ′ .

On (s 0 −δ′ , s 0 +δ′ )×(t 0 −δ, t 0 +δ), γ̃s ↾(t0 −δ,t0 +δ) is a lift of γs ↾(t0 −δ,t0 +δ) . Also note that
φ ◦ γs ↾(t0 −δ,t0 +δ) is a lift of γs ↾(t0 −δ,t0 +δ) .
Observe that, at (s 0 , t 1 ), we have

γ̃s (s 0 , t 1 ) = φ ◦ γs (s 0 , t 1 )

=⇒ γ̃s ↾(t0 −δ,t0 +δ) = φ ◦ γs ↾(t0 −δ,t0 +δ)

e (s, t ) = φ ◦ H (s, t )
=⇒ H

for (s, t ) ∈ (s 0 − δ′ , s 0 + δ′ ) × (t 0 − δ, t 0 + δ).

379
 7

Since φ ◦ H is continuous at (s 0 , t 0 ), we have a contradiction. Hence H

e is continuous

on [0, 1] × [a, b].

Notice that f ◦ H e (s, b) ∈ f −1 {z 1 }. Since f −1 {z 1 }
e (s, b) = H (s, b) = z 1 . In particular, H

is discrete in Y and H
e is continuous on [0, 1] × {b}, we have H
e ([0, 1] × {b}) is connected

subset of Y contained in f −1 {z 1 }. Hence H

e ([0, 1] × {b}) is a constant. □

380
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 46
Covering Spaces

We defined the notion of a lift of a map g with respect to a continuous map f . We also
discussed some of the properties of such a lift and we saw that if we have a couple of
curves γ1 , γ2 which are homotopic to each other and at every stage γs could be lifted,
then the homotopy could also be lifted. However, at no point of time, we did talk about
the existence of such a lift. So let us define a ‘covering map’ which gives us a sufficient
condition to talk about when a curve could be lifted.

D EFINITION 1 (Covering map). Let X , Y be open subsets of C. We say that a con-

tinuous map f : Y −→ X is a covering map if given x ∈ X , there exists a neighborhood
U of x and open sets {Vα }α∈A in Y such that f −1 (U ) =
`
Vα and f ↾Vα : Vα −→ U is a
α∈A
homeomorphism. Then Y is called a cover of X . The set U is called as an evenly covered
neighborhood.

(Refer Slide Time: 4:34)

E XAMPLE 1.
1

381
2

• Let f : C∗ −→ C∗ be given by f (z) = z k for k ∈ N. Since f ′ (z) ̸= 0 for each z ∈ C∗ ,

we have f is a local biholomorphism. Given z 0 ∈ C∗ , ∃ w 1 , . . . , w k ∈ C∗ such that
z k = z 0 has w i as solutions. Let Vi be neighborhoods of w i and Ui be neighbor-
n
hood of z 0 such that f ↾Vi : Vi −→ Ui is a homeomorphism. Define U =
T
Ui
i =1
ǹ
and Vi′ =f −1
(U ) ∩ Vi . Then, f −1
(U ) = Vi′ . Hence f is a covering map.
i =1
• Consider exp : C −→ C∗ . Since exp is a local biholomorphism, given z 0 ∈ C∗ ,
there exists w 0 ∈ C such that e w 0 = z 0 . By inverse function theorem, we have
neighborhoods V and U of w 0 and z 0 respectively such that exp ↾V V :−→ U is
a biholomorphism. Now, the reader should verify that if

Vn := V + 2πi n = {z + 2πi n : z ∈ V, n ∈ Z},

then {Vn }n∈Z is mutually disjoint. Then exp ↾Vn : Vn −→ U is a homeomorphism

and exp−1 (U ) =
`
Vn . Hence exp is a covering map
n∈Z

T HEOREM 2. Let f : Y −→ X be a covering map and γ : [a, b] −→ X be a curve from x 0

to x 1 in X . Suppose y 0 ∈ f −1 ({x 0 }). Then there exists unique lift γ̃ : [a, b] −→ Y of γ with
respect to f such that γ̃(a) = y 0 .

(Refer Slide Time: 12:00)

Given below is a corrected version of the proof given in the lecture.

382
 3

P ROOF. Given x ∈ γ([a, b]). We have a neighborhood U x evenly covered for the cov-
ering map f .
Let U := {U x : x ∈ γ([a, b])} be the collection of evenly covered neighborhoods corre-
sponding to each point in the image of γ. Then U is an open cover for γ([a, b]). Since
γ is continuous, U ′ = {γ−1 (U x ) : U x ∈ U } is an open cover for the compact set [0, 1]. Let
δ > 0 be the Lebesgue number of U ′ .
Consider a partition P : a = t 0 < t 1 < · · · < t n = b of [a, b] with partition size less than
δ
2
, so that image of the consecutive points under γ lie in one evenly covered neighbor-
hood.
We will construct a lift by induction on the index t i . For t 0 , t 1 , there exists an evenly
covered neighborhood U1 ∈ U such that γ(t 0 ), γ(t 1 ) ∈ U1 . Let f −1 (U1 ) =
`
Vα .
α∈A
Let α0 ∈ A be such that y 0 ∈ Vα0 and ϕ : U1 −→ Vα0 be the inverse of f ↾Vα0 . For
t ∈ [t 0 , t 1 ], define

γ̃(t ) = ϕ ◦ γ(t ).

(Refer Slide Time: 21:23)

Then γ̃ is a lift of γ ↾[t0 ,t1 ] . Suppose γ̃ has been constructed inductively up to γ(t k−1 ).
Let Uk ∈ U be an evenly covered neighborhood containing γ(t k−1 ) and γ(t k ). Let us
put γ̃(t k−1 ) = z̃ k−1 . Suppose f −1 (Uk ) = Vα and α′0 ∈ A ′ be such that z̃ k−1 ∈ Vα′ ′ . Let
` ′
α′ ∈A ′ 0

383
4

ϕ′ : Vα′ ′ −→ Uk be the inverse of f ↾V ′ ′ . For t ∈ [t k−1 , t k ], define

0 α
0

γ̃(t ) = ϕ′ ◦ γ(t ).

Also γ̃ : [a, t k ] −→ Y is a lift of γ ↾[a,tk ] . Hence, inductively, we have γ̃ : [a, b] −→ Y is a lift

of γ. □

E XAMPLE 3. Consider the function f : C∗ −→ C∗ given by f (z) = z 2 . Let γ(t ) = e i t for

t
t ∈ [0, 2π]. For t ∈ [0, 2π] γ̃(t ) = e i 2 . Then, note that f γ̃(t ) = e i t = γ(t ). Hence γ̃ is a lift
¡ ¢

of γ with respect to f with initial point 1.

P ROPOSITION 4. Let X be path connected and f : Y −→ X be a covering map. Suppose

x 0 , x 1 ∈ X . Then the cardinality of f −1 (x 0 ) is the same as the cardinality of f −1 (x 1 ).

P ROOF. Since X is path connected, there exists a path γ : [a, b] −→ X from x 0 to x 1 .

Let y 0 ∈ f −1 (x 0 ). Then, by Theorem 2, there exists a lift γ̃ : [a, b] −→ Y of γ with initial
point y 0 .
Define ϕ : f −1 (x 0 ) −→ f −1 (x 1 ) given by ϕ(y) = γ̃(b). By uniqueness, this map is injec-
tive. Now, it is left as an exercise for the reader to prove that ϕ is in fact a bijection. □

T HEOREM 5. Let X , Y be open subsets of C and f : Y −→ X be a covering map from

Y to X . Let Z be an open connected subset of C, which is simply connected and locally
connected. Suppose g : Z −→ X be a continuous function. If given z 0 ∈ Z and y 0 ∈ Y such
that g (z 0 ) = f (y 0 ), then there exists a unique lift g̃ : Z −→ Y of g with respect to f .

(Refer Slide Time: 36:47)

384
 5

P ROOF. Since every connected open subsets of C are path connected, we have Z is
path connected. Let z ∈ Z and let γ0 : [a, b] −→ Z be a curve from z 0 to z. Let σ0 :
[a, b] −→ X be the curve g ◦ γ0 from g (z 0 ) = x 0 to g (z) = x.
By Theorem 2, there exists a unique curve σ̃0 : [a, b] −→ Y in Y with initial point y 0
and lifting σ0 with respect to f . Define g̃ (z) := σ̃0 (b).
Let us try to prove that g̃ is well defined. Let γ1 be another curve in Z from z 0 to z.
Since Z is simply connected, γ0 is homotopic to γ1 with fixed end points through the
homotopy H : [0, 1] × [a, b] −→ Z . Now one can easily verify that g ◦ H is a homotopy
with fixed end points from σ0 to σ1 = g ◦ γ1 .
(Refer Slide Time: 45:28)

385
6

If σs (t ) := g ◦ H (s, t ) is a curve from x 0 to x, then there exists a lift σ̃s in Y of σs with

initial point y 0 (by Theorem 2). Also, g ◦ H can be lifted to a homotopy with fixed end
points in Y from σ̃0 to σ̃1 . In particular, σ̃1 (b) = g̃ (z). Hence g̃ is well defined.
¡ ¢
Now it is an easy check that f g̃ (z) = g (z).
Claim: g̃ is continuous.
To prove that g̃ is continuous, it is enough to show that, given z ∈ Z , let V be a neigh-
borhood of g̃ (z) = y. Then z is an interior point of g̃ −1 (V ).
By shrinking V , if needed, we may assume that f ↾V : V −→ U is a homeomorphism
with inverse ϕ : U −→ V , where U is an evenly covered neighborhood.
Since g is continuous, consider g −1 (U ) which is a neighborhood of of z. By local
connectedness of Z , there exists a neighborhood W of z such that W ⊆ g −1 (U ) and W is
connected.
Claim: W ⊆ g̃ −1 (V ).
(Refer Slide Time: 52:38)

Let z ′ ∈ W , then there exists a path δ : [a, b] −→ W from z to z ′ in W . Then g ◦ δ is a

curve in U from x to x ′ = g (z ′ ). Now we will lift the curve g ◦ δ to a curve in V . Define
◦ δ = ϕ ◦ g ◦ δ.
g‚
A simple check will tell us that g (γ0 +δ) = g ◦γ0 +g ◦δ = σ0 +g ◦δ and further we have
σ̃0 + g‚
◦ δ = σ̃0 + ϕ ◦ g ◦ δ is a lift of σ0 + g ◦ δ.

386
 7

By the definition of g̃ ,

g̃ (z ′ ) = σ̃0 + ϕ ◦ g ◦ δ (b) = ϕ ◦ g ◦ δ(b) ∈ V.

¡ ¢

Hence z ′ ∈ g̃ −1 (V ) =⇒ W ⊆ g̃ −1 (V ). □

C OROLLARY 6. Let Ω ⊆ C be an open connected subset which is simply connected and

locally connected and g : Ω −→ C∗ be a holomorphic map. Then there exists a lift g̃ : Ω −→
C such that exp(g̃ ) = g .

387
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 47
Bloch’s Theorem

The theorem due to Bloch is a prerequisite for proving the ‘Little Picard’s theorem’. Bloch’s
theorem tries to answer the following question: suppose we have an open connected set
Ω which contains D and consider the family

F = f : Ω −→ C : f is holomorphic on Ω, f (0) = 0 and f ′ (0) = 1 .

© ª

Then what is the largest disk that can be fitted up in f (D) for any f ∈ F .
To begin with, let us try to prove a special case of Bloch’s theorem.
(Refer Slide Time: 01:01)

P ROPOSITION 1. Let f : D −→ C be such that f (0) = 0, f ′ (0) = 1 and | f (z)| ≤ M for each
¡ 1 ¢
z ∈ D. Then D 0, 6M ⊆ f (D).

P ROOF. Consider the power series of f around 0,

an z n .
X
f (z) = z +
n≥2
1

388
2

Note that
¯ ¯
n
¯ ¯ ¯X ¯
¯ f (z)¯ ≥ |z| − ¯ a n z ¯¯ .
¯
n≥2

By Cauchy estimates, we have

¯ ¯
sup ¯ f (z)¯
z∈∂D(0,r ) M
|a n | ≤ ≤ for each 0 < r < 1.
rn rn

Taking the limit as r −→ 1− , we have |a n | ≤ M . Then,

|z|2
¯ ¯
n
¯X ¯ X n
¯ a n z ¯ ≤ M |z| = M .
¯
n≥2
¯
n≥2 1 − |z|

Hence,
µ 2 ¶
¯ f (z)¯ ≥ |z| − M |z|
¯ ¯
.
1 − |z|
Our first observation is the number M should necessarily be at least 1. For proving this,
let us first assume M < 1, to arrive at a contradiction.
If M < 1, then f : D −→ D with f (0) = 0 and f ′ (0) = 1, then by Schwarz’s lemma tells
that f (z) = λz for |λ| = 1. Then for z ∈ D such that M < |z| < 1, we have | f (z)| = |z| > M ,
which is a contradiction. Hence M ≥ 1.
1
Now, for |z| = 4M < 1,
1
|z|2 1 M · 16M 2
|z| − M = − 1
1 − |z| 4M 1 − 4M
1 1
= −
4M 12M
2
=
12M
1
= .
6M
1 1
¯ ¯ © ª
Hence ¯ f (z)¯ ≥ 6M on z : |z| = 4M .
¡ 1 ¢ 1
© 1
ª
Let w ∈ D 0, 6M . Then, |w| < 6M ≤ | f (z)| on z : |z| = 4M . By Rouché’s theorem,
¡ 1 ¢
f (z) − w and f (z) has the same number of zeroes in D 0, 4M . Then there exists z ∈
¡ 1 ¢ ¡ 1 ¢
D 0, 4M such that f (z) = w. Hence D 0, 6M ⊆ f (D). □

389
 3

Let us now generalize this result to bit higher generality by stating and proving the
following proposition.

P ROPOSITION 2. Let R > 0, f : D(0, R) −→ C be a holomorphic function such that

f (0) = 0, ¯ f ′ (0)¯ = µ > 0 and ¯ f (z)¯ ≤ M for each z ∈ D(0, R). Then
¯ ¯ ¯ ¯

R 2 µ2
µ ¶
D 0, ⊆ f (D(0, R)).
6M

(Refer Slide Time: 11:20)

P ROOF. Consider the function g on D given by,

g (z) = ψ2 ◦ f ◦ ψ1 (z),

z
where ψ1 (z) = Rz and ψ2 (z) = Rµ . Then note that

M
µ ¶
ψ1 f ψ2
g : D(0, 1) −−→ D(0, R) − → D(0, M ) −−→ D 0, .
Rµ
³ ´
M
¯ M
. Now, g (0) = 0, g ′ (0) = 1 and ¯g (z)¯ ≤ Rµ for z ∈ D. Then
¯
That is g is a map into D 0, Rµ
³ ´
Rµ
by Proposition 1, we have D 0, 6M ⊆ g (D). Now it is left as an exercise to the reader to
conclude the result. □

Let us know prove the Bloch’s theorem which is in slightly more generality.

390
4

T HEOREM 3 (Bloch’s theorem). Let Ω be an open connected set in C such that D ⊂ Ω.

Suppose f : Ω −→ C be such that f (0) = 0 and f ′ (0) = 1. Then there exists a disk D 1
contained in D such that f ↾D 1 is injective and
µ ¶
1
D 0, ⊆ f (D 1 ) ⊆ f (D).
72

P ROOF. Define M (r ) = sup ¯ f ′ (z)¯. Then M (r ) is continuous on [0, 1] (Why?). Now,

¯ ¯
|z|=r
define h(r ) := (1 − r )M (r ). Hence h is continuous on [0, 1]. Moreover, we have h(0) = 1
and h(1) = 0.
Let r 0 = sup{r : h(r ) = 1}. Then h(r ) < 1 for r > r 0 . Since M (r 0 ) = sup ¯ f ′ (z)¯ and f is
¯ ¯
|z|=r 0
also continuous on the compact set {z : |z| = r 0 }, there exists z 0 such that |z 0 | = r 0 with
¯ f (z 0 )¯ = M (r 0 ). Now, h(r 0 ) = 1 =⇒ ¯ f ′ (z 0 )¯ = 1 .
¯ ′ ¯ ¯ ¯
1−r 0

(Refer Slide Time: 20:02)

Let ρ = 12 (1 − r 0 ) and consider D(z 0 , ρ). Let z ∈ D(z 0 , ρ). Then,

1 1
|z| ≤ r 0 + (1 − r 0 ) = (1 + r 0 ).
2 2

Since 1 > 21 (1 + r 0 ) > r 0 =⇒ h 12 (1 + r 0 ) < 1. That is,

¡ ¢

µ ¶ µ ¶ µ ¶
1 1 1 1 1
1 > 1 − (1 + r 0 ) M (1 + r 0 ) =⇒ M (1 + r 0 ) < 1 = .
2 2 2 2
(1 − r 0 ) ρ

391
 5

By maximum modulus principle,

¯ ′ ¯ 1
¯ f (z)¯ < for z ∈ D(z 0 , ρ).
ρ

Define g (z) = f ′ (z) − f ′ (z 0 ) on D(z 0 , ρ). Then,

¯ 1 1 3
|g (z)| ≤ ¯ f ′ (z)¯ + ¯ f ′ (z 0 )¯ < +
¯ ¯ ¯
= .
ρ 2ρ 2ρ

Consider the holomorphic function given by h(z) = ψ2 ◦ g ◦ ψ1 (z) that is

µ ¶
ψ1 f 3 ψ2
h : D −−→ D(z 0 , ρ) −
→ D 0, −−→ D,
2ρ
2ρ
where ψ1 (z) = z 0 +ρz and ψ2 (z) = z. Then h(0) = 0 and by Schwarz’s lemma, we have
3
¯ ¯
¯ψ2 ◦ g ◦ ψ1 (z)¯ ≤ |z| =⇒ ¯ g (z)¯ ≤ ¯ψ−1 (z)¯ = |z − z 0 |
¯ ¯ ¯ 2ρ ¯ ¯ ¯
1
¯ 3 ¯ ρ

3
=⇒ |g (z)| ≤ |z − z 0 |.
2ρ 2
Note that
¯ f (z 0 ) − f ′ (z)¯ = |g (z)| < 3 |z − z 0 | < 3 · ρ = 1 = | f ′ (z 0 )|.
¯ ′ ¯
ρ2 ρ 2 3 2ρ
³ ρ´
Put D 1 := D z 0 , . Let z 1 , z 2 ∈ D 1 . Then,
3
¯Z ¯
¯ ¯¯ ¯
f ′ (z)d z ¯
¯
¯ f (z 2 ) − f (z 1 )¯ = ¯ ¯
¯ γz →z ¯
1 2
¯Z ¯ ¯Z ¯
¯ ¯ ¯ ¯
f ′ (z 0 )d z ¯ − ¯ f ′ (z 0 ) − f ′ (z) d z ¯
¡ ¢
≥¯
¯ ¯ ¯ ¯
¯ γz →z ¯ ¯ γz →z ¯
1 2 1 2
¯Z ¯
¯ ′ ¯ ¡ ′ ¢ ¯¯
′
¯
≥ f (z 0 ) |z 1 − z 2 | − ¯ f (z 0 ) − f (z) d z ¯
¯ ¯ ¯
¯ γz →z ¯
1 2

> ¯ f ′ (z 0 )¯ |z 1 − z 2 | − ¯ f ′ (z 0 )¯ |z 1 − z 2 | = 0.
¯ ¯ ¯ ¯

That is, for each z 1 , z 2 ∈ D 1 , we have | f (z 2 ) − f (z 1 )| > 0. Hence f is injective on D 1 .

Thus we proved one part of the Bloch’s theorem. Let us now try to prove the remain-
¡ 1¢
ing part of the theorem, which says that D 0, 72 ⊆ f (D 1 ).

392
6
¡ ρ¢
Define g̃ (z) = f (z + z 0 ) − f (z 0 ) on D 0, 3 . Then, we have g̃ (0) = 0 and ¯g̃ ′ (0)¯ =
¯ ¯

¯ f (z 0 )¯ = 1 .
¯ ′ ¯
2ρ

(Refer Slide Time: 31:40)

¡ ρ¢
For w ∈ D 0, 3 , γz0 →z0 +w ⊂ D 1 . Then
¯Z ¯
¯ ′ ¯ ¯¯ ¯ 1 1
f ′ (z)d z ¯ < |w| < .
¯ ¯
¯g̃ (w)¯ = ¯ f (w + z 0 ) − f (z 0 )¯ = ¯ ¯
¯ γz ¯ ρ 3
0 →z 0 +w

¡ ρ¢ 1 1
Hence g̃ : D 0, 3 −→ C is such that g̃ (0) = 0, |g̃ (0)| = and |g̃ | < for z ∈ D. Then by
2ρ 3
Proposition 1, we have  ¡ ¢ ³ ´2 
ρ 2 1
3 2ρ 
³ ³ ρ ´´
D 0, ⊆ g̃ D 0,

6 · 13 3


µ
1
¶ ³ ³ ρ ´´
=⇒ D 0, ⊆ g̃ D 0,
72 3
µ ¶
1
=⇒ D 0, ⊆ f (D 1 ).
72
□

393
 Complex Analysis
Prof. Pranav Haridas
Kerala School of Mathematics
Lecture No – 48
Little Picard’s Theorem

The little Picard’s theorem tells us that if you have a entire function which is non-constant
then it cannot omit 2 points. We have developed all the machinery needed to prove the
theorem, but we will first give a characterization of holomorphic functions defined on a
simply connected domain which omits 2 points.

P ROPOSITION 1. Let Ω be an open connected subset of C which is simply connected.

Let f : Ω −→ C be a holomorphic function which omits 0 and 1. Then there exists a holo-
morphic function g : Ω −→ C such that

f (z) = − exp πi cosh 2g (z) .

¡ ¡ ¢¢

(Refer Slide Time: 02:45)

394
2

P ROOF. Since f : Ω −→ C∗ , we have a lift of f with respect to exp. Let f˜ be a holo-

f˜(z)
morphic function on Ω such that exp f˜(z) = f (z). Define F (z) =
¡ ¢
.
2πi
If F (z 0 ) = n for some z 0 ∈ Ω, then

f (z 0 ) = exp f˜(z 0 ) = exp (2πi n) = 1

¡ ¢

which is a contradiction as f (Ω) ⊆ C \ {0, 1}. Hence F does not take integer values.
(Refer Slide Time: 04:05)

Now, consider the covering map p : C∗ −→ C∗ given by p(z) = z 2 . Then we have a lift
p
p F
F of F with respect to p, Ω −−→ C∗ . Also since F does not take integer values, F (z) − 1
p
does not vanish and is also holomorphic. Hence we have a lift F (z) − 1 of F (z) − 1 with
p
F (z)−1
respect to p, Ω −−−−−−→ C∗ .
p p
Define H : Ω −→ C given by H (z) = F (z) − F (z) − 1. Notice that H (z) ̸= 0 for every
z ∈ Ω. That is, H : Ω −→ C∗ .

395
 3
¡ ¢
Then we have a lift g of H with respect to exp. That is exp g (z) = H (z). Note that

¡ ¢ e 2g (z) + e −2g (z)

cosh 2g (z) + 1 = +1
2
¡ g (z) ¢2
e + e −g (z)
=
2
³ ´2
1
¡ ¢ H (z) + H (z)
=⇒ cosh 2g (z) + 1 =
2
= 2F (z).

Thus,
˜
f (z) = e f (z) = e 2πi F (z) = e πi cosh(2g (z)) e πi = −e πi cosh(2g (z)) .

P ROPOSITION 2. The function g in the Proposition 1 does not contain any disk of ra-
dius 1 in its image.

¡p p
P ROOF. Let S = ± ln n + n − 1 + 21 i mπ : n ≥ 1, m ∈ Z . Then notice that the points
© ¢ ª

in the set S can be represented in complex plane as the darkened dots in the following
grid.

396
4

Claim: g (Ω) ∩ S = ∅.
Assume that claim is true.
The height of each rectangle can be bounded by

¯ (m + 1)π − i mπ¯ = π < 3.

p
¯ ¯
¯i ¯
¯2 2 ¯ 2
p ¢ ¡p p
Also, if we define φ(x) = ln x + 1 + x − ln x + x − 1 , then one can verify that φ is
¡p ¢

a decreasing function and φ(1) < 1. Thus, the width of each rectangle can be bounded
by
¯ ³p p ´ ³p p ´¯
¯ln n + 1 + n − ln n + n − 1 ¯ < 1.
¯ ¯

p 2
q
Hence the diagonal of each rectangle is bounded by 3 + 1 = 2.
(Refer Slide Time: 16:36)

Hence any disk of radius 1 intersects some point in S and by our assumption propo-
sition follows. Thus it remains to prove the claim.
Let z 0 ∈ Ω be such that
³p p ´ i
g (z 0 ) = ± ln n + n − 1 + mπ.
2

397
 5

Now,

2 cosh 2g (z 0 ) = e 2g (z0 ) + e −2g (z0 )

¡ ¢

i mπ p
p ´2 ³p p
µ³ ´2 ¶
=e n + n −1 + n − n −1

= (−1)m 2(2n − 1)

cosh 2g (z 0 ) = (−1)m (2n − 1).

¡ ¢

Then,

f (z 0 ) = − exp πi cosh 2g (z)

¡ ¡ ¢¢

= − exp (2n − 1)πi (−1)m

¡ ¢

=1

which is a contradiction. □

What we have proved is, if f is holomorphic on Ω which omits 0 and 1, then there
exists a holomorphic function g on Ω which does not have a disk of radius 1 contained
in g (Ω).

T HEOREM 3 (Little Picard’s Theorem). If f is an entire function which omits two

points, then f is a constant function.

P ROOF. Let z 0 , z 1 be two distinct points which f omits. We may assume without loss
of generality that z 0 = 0 and z 1 = 1,

f (z) − z 0
f 1 (z) = .
z1 − z0

That is, f 1 is an entire function omits 0 and 1.

(Refer Slide Time: 23:42)

398
6

By the Proposition 1 and Proposition 2, there exists an entire function g such that

f (z) = − exp πi cosh 2g (z)

¡ ¡ ¢¢

which does not have a disk of radius 1 in the image of g .

Assume f is non-constant, then g is also non-constant.
Let z 0 ∈ C be such that g ′ (z 0 ) ̸= 0. Define h(z) = g (z + z 0 ) − g (z 0 ). Then h(0) = 0 and
h ′ (0) = g ′ (z 0 ). Now, it is left as an exercise to check that h is an entire function which also
does not contain a disk of radius 1 in its image. Let R > 0 be some positive number and
for 0 < r < R, define
rz
µ ¶
1
ψ(z) = h ′ .
r h (0)
Then ψ(0) = 0 and ψ′ (0) = r1 h ′ (0) h ′r(0) = 1.
By Bloch’s theorem,
µ ¶
1
D 0, ⊆ ψ(D).
72
¡ r ¢
Thus, for w ∈ D 0, 72 , there exists z ∈ D such that

w 1 rz
µ ¶
= h ′ =⇒ w = h(z ′ ) where z ′ ∈ C
r r h (0)

399
 7
³ r ´
=⇒ D 0, ⊆ h(C).
72
But since r was arbitrary, we can choose r > 72, which is a contradiction.
Hence f is a constant function. □

400
 THIS BOOK
IS NOT FOR
SALE
NOR COMMERCIAL USE

(044) 2257 5905/08

nptel.ac.in
swayam.gov.in