Complex Variables

Lecture Notes

Victor Ivrii

Department of Mathematics,
University of Toronto

© by Victor Ivrii, 2021,

Toronto, Ontario, Canada
Contents

Contents i

1 The Complex Plane 2

1.1 Complex Numbers . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Some Geometry . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Subsets of the Plane . . . . . . . . . . . . . . . . . . . . . . 16
1.4 Functions and Limits . . . . . . . . . . . . . . . . . . . . . . 21
1.5 Exponential, Logarithm, and Trigonometric Functions . . . . 26
1.6 Line Integrals and Green’s Theorem . . . . . . . . . . . . . . 36

2 Basic Properties of Analytic Functions 47

2.1 Analytic and Harmonic Functions; Cauchy-Riemann Equations 47
2.2 Power series . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.3 Cauchy’s Theorem and Cauchy’s Formula . . . . . . . . . . . 62
2.4 Consequences of Cauchy’s Formula . . . . . . . . . . . . . . 66
2.5 Isolated Singularities . . . . . . . . . . . . . . . . . . . . . . 76
2.6 The Residue Theorem and its Application to the Evaluation
of Definite Integrals . . . . . . . . . . . . . . . . . . . . . . . 93

3 Analytic Functions as Mappings 112

3.1 The Zeros of an Analytic Function . . . . . . . . . . . . . . 112
3.2 Maximum Modulus and Mean Value . . . . . . . . . . . . . 124
3.3 Linear Fractional Transformations . . . . . . . . . . . . . . . 132
3.4 Conformal Mapping . . . . . . . . . . . . . . . . . . . . . . . 140
3.5 The Riemann Mapping Theorem and Schwarz-Christoffel
Transformations . . . . . . . . . . . . . . . . . . . . . . . . . 149

i
Preface

These are Lecture Notes for MAT 334 “Complex Variables” at Faculty of
Arts and Science, University of Toronto. This is a junior class for all but
Math Specialist students.
I was teaching it for several years and the last time it at Fall of 2020.
This time the class was taught online due to COVID-19 pandemic and I
made beamer slides for lectures. These slides were reformatted to a book
format.
These Lecture Notes are addition rather than substitution for our stan-
dard textbook Complex Variables, 2nd Edition, by Stephen D. Fisher (re-
ferred as Textbook).
We cover Chapters 1–3 from Textbook, however some material removed
and other material added, exposition is different.

c b a This work is licensed under a Creative Commons Attribution-ShareAlike 4.0

International License.

1
Chapter 1

The Complex Plane

Introduction to MAT334
We start a class called “Complex Variables” but more precisely it should be
called Functions of a Complex Variable and even more precisely Functions
of One Complex Variable.
In late 17-th century I. Newton diskovered that the decomposition of
functions of one real variable into power series (which we call today Taylor
series) is a very powerful tool:
∞
X
f (x) = an (x − x0 )n (1.1.1)
n=0
with
1 (n)
an = f (x0 ). (1.1.2)
n!
Starting from I. Newton mathematicians began to use this tool extensively
and for more than 100 years a function was something that is given by a
power series.
They did not care where this series converges (if it converges at all,
except for x = x0 ), and if it converges to f (x). Today we call these functions
analytic or, more precisely real analytic.
There were some problems with this approach: many functions are not
differentiable, even for infinitely differentiable functions series (1.1.1) may
not converge (except at x = x0 ), or converge not to f (x). Also comparison
of decompositions at different points was not very obvious. So, if a function
f (x) depends only on x, why we need to deal with x0 ?

2
Chapter 1. The Complex Plane 3

Also, if f (x) is defined on the whole real line R = (−∞, ∞) we cannot

say where such series converges.
F.e. for f (x) = ex , cos(x), sin(x) Taylor series at x = 0 converges on R,
1
but for f (x) = 2 it converges only at [−1, 1].
x +1
Note that if f (x) is given by the power series
∞
X
f (x) = an (x − x0 )n (1.1.1)
n=0

then one can plug instead of x a complex number z and get a function of
the complex variable
∞
X
f (z) = an (z − x0 )n (1.1.3)
n=0

(assuming that it converges somewhere).

We will get an complex analytic function f (z) which for real z = x
coincides with f (x). We say that f (z) is an analytic continuation of f (x) to
the complex domain.
However we are still talking about power series, and they depend on the
point, where we decompose f (z):
∞
X
f (z) = an (z − z0 )n (1.1.4)
n=0

(now z0 could be a complex number as well).

So, we have

Goal #1. Characterize functions of the complex variable z = x + iy which

are analytic, so that they can be decomposed into series (1.1.4).
The theory of analytic functions of a complex variable became a workhorse
of analysis with multiple applications to mathematics itself, physics, engi-
neering and through many mathematical uses to many other fields.

Why “Complex variables” is a hard class

So far you know real valued and vector valued (in particular, complex valued)
functions of one real variable (Calculus I), real valued and vector valued (in
Chapter 1. The Complex Plane 4

particular, complex valued) functions of several real variables (Calculus II).

Now we have a theory of the functions of the complex variable. But such
functions could be considered as functions of two real variables, x and y,
right?
However only some of them could be decomposed into power series with
respect to z and such functions have many properties which general functions
of x, y do not posses.
So the theory of such functions is different from both a theory of functions
of one real variable and a theory of functions of two real variables.
One needs to think in the framework of theory of functions of a complex
variable! For example, while function sin(z) is uniquely defined as an
analytic continuation of sin(x), they are very different. In particular, the
latter is bounded, and the former is not.

Think differently!!
Think like a complex analyst!!

We start slowly, but do not relax.

1.1 Complex Numbers

1.1.1 Complex Numbers: Definition
Complex numbers are pairs of real numbers (x, y) written as z = x + iy.
When x, y run a real line R, corresponding complex number z runs a complex
plane C. We have a one-to-one correspondence between C and a real plain
R2 : C ∋ z = x + yi ↔ z = xi + yj ∈ R2 (here i and j are coordinate
vectors on the real plane).

Definition 1.1.1. x is called a real part of z and is denoted as z = Re(z)

(sometimes denoted as x = ℜ(z)) and y is called an imaginary part of z and
is denoted as z = Im(z) (sometimes denoted as x = ℑ(z)).
We identify x + i0 with a real number x and 0 + iy with an imaginary
number 0 + iy. In particular, we write 0 + i0 simply as 0.

Therefore a complex plane C inherits from the the real plane R2 an

addition (x1 + iy1 ) + (x2 + iy2 ) = (x1 + x2 ) + i(y1 + y2 ) and multiplication by
a real number α : α(x + iy) = (αx) + i(αy). All the usual properties hold.
Chapter 1. The Complex Plane 5

1.1.2 Complex Conjugate and Multiplication of

Complex Numbers
There are two more operations:
Complex conjugate to z = x + iy is z = x − iy (which is the reflection of
z with respect to real line):

y
z

One can see easily that

z1 + z2 = z 1 + z 2 ,
αz = αz for real α,
z = z.

Multiplication of complex numbers For complex numbers we intro-

duce multiplication:

(x1 + iy1 )(x2 + iy2 ) := (x1 x2 − y1 y2 ) + i(x1 y2 + x2 y1 ). (1.1.5)

One can check easily that

z1 z2 = z2 z1 , (1.1.6)
z(z1 + z2 ) = zz1 + zz2 , (1.1.7)
z1 (z2 z3 ) = (z1 z2 )z3 , (1.1.8)
z1 z2 = z 1 z 2 . (1.1.9)

Actually (1.1.5) is introduced to have these properties and

1 · 1 = 1, i · 1 = 1 · i = i, i · i = −1.
Chapter 1. The Complex Plane 6

1.1.3 Polar Representation of Complex Numbers

p
Let us introduce polar coordinates: polar radius r = x2 + y 2 and polar
angle θ:
y
z
r
θ x

- r is called an absolute value (or module) of z and denoted by |z|.

- θ is called an argument of z and denoted by arg(z).
Remark 1.1.1. (a) arg(z) is defined only as z ̸= 0 ( ⇐⇒ r ̸= 0).
(b) arg(z) is defined modulo 2πn, n ∈ Z and thus is a multivalued function
of z.
(c) Arg(z) is a single-valued function of z; usually we define it so that
0 ≤ Arg(z) < 2π.
Observe that
r2 = |z|2 = zz, (1.1.10)
|z| = |z|, arg(z) = − arg(z). (1.1.11)
Therefore z −1 := z
̸ 0: z −1 z = zz −1 = 1; z −1 does not
is an inverse to z =
|z|2
exist for z = 0.
Since x = r cos(θ), y = r sin(θ) we conclude that
z = r(cos(θ) + i sin(θ)) (1.1.12)
with r ≥ 0 defined uniquely and θ defined modulo 2πn, n ∈ Z. It is a polar
representation of complex number z.
If z1 = r1 (cos(θ1 ) + i sin(θ1 )) and z2 = r2 (cos(θ2 ) + i sin(θ2 )), then
z1 z2 =r1 r2 cos(θ1 ) + i sin(θ1 ))(cos(θ2 ) + i sin(θ2 ))


=r1 r2 cos(θ1 ) cos(θ2 ) − sin(θ1 ) sin(θ2 )


+ i cos(θ1 ) sin(θ2 ) + sin(θ1 ) cos(θ2 )
 
=r1 r2 cos(θ1 + θ2 ) + i sin(θ1 + θ2 )
Chapter 1. The Complex Plane 7

(remember trigonometric formulas!) and therefore

|z1 z2 | = |z1 | · |z2 | (1.1.13)

and
arg(z1 z2 ) = arg(z1 ) + arg(z2 ). (1.1.14)

Using the formula for inverse we get

Theorem 1.1.1 (De Moivre’s Theorem). As z2 ̸= 0

z1 |z1 |
= , (1.1.15)
z2 |z2 |
z1

arg = arg(z1 ) − arg(z2 ). (1.1.16)
z2
Corollary 1.1.2. As z ̸= 0, n ∈ Z

|z n | = |z|n , (1.1.17)
arg(z n ) = n arg(z). (1.1.18)

1.1.4 Exponential Representation of Complex

Numbers
Let us introduce the following complex function of the real variable θ ∈ R:

eiθ := cos(θ) + i sin(θ). (1.1.19)

From the above it follows that

ei(θ1 +θ2 ) = eiθ1 eiθ2 , (1.1.20)

−iθ
eiθ = e . (1.1.21)

Further,

′
(eiθ )′ = cos(θ) + i sin(θ) = − sin(θ) + i cos(θ) = i cos(θ) + i sin(θ)

= ieiθ

Thus
(eiθ )′ = ieiθ . (1.1.22)
Chapter 1. The Complex Plane 8

Properties (1.1.20) and (1.1.22) show that this function behaves like an
exponent, which justifies this definition.
Furthermore, if we define for λ = α + iβ, α, β ∈ R,

eλθ := eαθ eiβθ = eαθ cos(βθ) + i sin(βθ) ,

(1.1.23)
then

′
eλθ = λeλθ (1.1.24)

which we used in ODE class.

It follows from the definition and properties of eiθ that
1 iθ
e + e−iθ = Re(eiθ ),


cos(θ) = (1.1.25)
2
1 iθ
e − e−iθ = Im(eiθ )


sin(θ) = (1.1.26)
2i
It follows from the trigonometric representation and (1.1.19)

z = reiθ , r = |z|, θ = arg(z). (1.1.27)

It is called exponential representation of the complex number.

1.1.5 Roots and Fractional Powers

Let us find all complex roots z of equation

z n = w, w = ρeiφ . (1.1.28)

We look at it also in the exponential form z = reiθ . Then

z n = rn einθ = ρeiφ . (1.1.29)

Therefore rn = ρ and nθ = φ+2πm, m ∈ Z and r = ρ1/n , θ = n1 (φ+2πm)

and
1
z = w1/n = |w|1/n e n (Arg(w)+2πm) , m ∈ Z. (1.1.30)

One can see easily that only for m = 0, 1, . . . , n − 1 formula (1.1.30)

gives different values, and then they repeat. Therefore, for w ̸= 0 equation
z n = w has n simple roots defined by (1.1.30) (for w = 0 it has just one
root of multiplicity n).
Chapter 1. The Complex Plane 9

Definition 1.1.2. (a) If p/q is an irreducible fraction (which means that

p, q ∈ Z, q > 0 do not have common divisors except ±1), then we
define q-valued function
p
z p/q := |z|p/q ei q (Arg(z)+2mπ) , m = 0, 1, . . . , q − 1. (1.1.31)

(b) If t is an irrational number, then we define ∞-valued function

z t := |z|t eit(Arg(z)+2mπ) , m ∈ Z. (1.1.32)

Subsection 1.1.1∗ A Formal View of the Complex Numbers. Read

it by yourself (optional)

1.2 Some Geometry

1.2.1 Norm and Inner Product
Let us explore the geometry of the complex plane C. Recall that a complex
number z = x + iy ∈ C matches to a vector z = xi + yj (later we will not
distinguish them!)
One can see easily that

z1 · z 2 = x1 x2 + y1 y2 = Re(z1 z 2 ) = Re(z 1 z2 ) (1.2.1)

where z 1 · z 2 is an the inner product of z 1 and z 2 . In particular

p q p
∥z∥ = ⟨z, z⟩ = Re(zz) = |z|2 = |z| (1.2.2)
and
|z1 + z2 | ≤ |z1 | + |z2 | (1.2.3)

where inequality become an equality if and only if either z2 = αz1 with

α ≥ 0 (and therefore α ∈ R) or z1 = 0. This is a triangle inequality.
Chapter 1. The Complex Plane 10

1.2.2 Straight Lines

Therefore straight line ℓ in C is defined by equation

Re(nz) = C n = A + iB ̸= 0. (1.2.4)

Indeed, we know that straight line ℓ in R2 is defined by equation

n·z =C (1.2.5)

where n = Ai + Bj is a normal to ℓ and

C
√ =d (1.2.6)
A2 + B 2
is the distance (with the sign) from the origin to ℓ.

y ℓ
n

d x

Figure 1.1: d > 0 above dashed line, d < 0 below dashed line.

1.2.3 Circles
A circle with a center at p ∈ C and radius r ≥ 0 is defined by equation

|z − p| = r. (1.2.7)

We say “a circle” meaning “circumference” (a line). The figure will be called

“a disk ”.
Chapter 1. The Complex Plane 11

Consider now two distinct points, p, q ∈ C and a curve which is the locus
of the points, satisfying

|z − p| = ρ|z − q| 0 < ρ < ∞. (1.2.8)

If ρ = 1 we have points z equidistant from p and q. The locus of these

points is precisely the straight line that is the perpendicular bisector of the
line segment joining p to q.
However, if ρ ≠ 1, the locus will be a circle. To see this, suppose that
0 < ρ < 1 (otherwise, divide both sides of the equation by ρ). Let z = w + q
and c = p − q; then (1.2.8) becomes |w − c| = ρ|w|.
Upon squaring and transposing terms, this can be written as

(1 − ρ2 )|w|2 − 2 Re(cw) + |c|2 = 0.

Completing the square on the left side we find that

c 2 ρ2 |c|2
(1 − ρ2 ) w − =
1 − ρ2 1 − ρ2
c ρ|c|
⇐⇒ w − 2
= ;
1−ρ 1 − ρ2
ρ|c| c
Thus, w is on the circle of radius R = 1−ρ 2 centered at the point w0 = 1−ρ2
and so z lies on the circle of the same radius R centered at the point

p ρ2 q
z0 = − . (1.2.9)
1 − ρ2 1 − ρ2
Point z0 belongs to a straight line passing through p and q, on the
ρ2 |c|
distance 1−ρ 2 from p and further from q than p. Recall that |z − p| = ρ|z − q|
p 2
ρ q
and z0 = 1−ρ 2 − 1−ρ2 . When ρ ↘ 0, circles became smaller and smaller

and shrink to p, and when ρ ↗ 1 then circles become larger and larger,
approximating L and their centers go to the right infinity (in our picture),
when ρ jumps over 1 we get circles on the other side, and their centers far
away to the left, and when ρ ↗ ∞ these circles shrink to q and their centers
move to from the left to q.
Chapter 1. The Complex Plane 12

q p L′

q p L′

These are Circles of Appolonius.

The family of red circles (including straight line L′ ) and the family of
blue circles (including straight line L) are mutually orthogonal : each red
Chapter 1. The Complex Plane 13

circle is orthogonal to each blue circle at the point of intersection (two lines
are orthogonal at the point of intersection if their tangent lines at this point
are orthogonal).
See the proof on pp. 18–20 of the Textbook.
Remark 1.2.1. (a) In this class we will encounter many mutually orthogo-
nal families of lines.
(b) We see that straight lines could be also included in the family of circles.
We will see that a straight line is a circle, passing from the infinity
(the exact meaning will be explained later).

1.2.4 Transformations: Translations, Scalings,

Rotations, Inversion
Consider some important geometrical transformations.

Translations.

C ∋ z 7→ w = z + a ∈ C, a ∈ C, (1.2.10)

is a translation by a (also called shift by a). Obviously, it transforms straight

lines and circles into straight lines and circles respectively.

Scalings.

C ∋ z 7→ w = λz ∈ C, λ > 0, (1.2.11)

is a scaling by the factor λ. Obviously, it transforms straight lines and circles

into straight lines and circles respectively. If 0 < λ < 1 it is a compression
by the factor λ−1 .

Rotations.

C ∋ z 7→ w = zeiφ ∈ C, φ ∈ R, (1.2.12)

is a rotation by angle φ in the counter-clockwise direction about the origin.

Indeed, it preserves a module |z| but adds φ to a polar angle arg(z). Obvi-
ously, it transforms straight lines and circles into straight lines and circles
respectively.
Chapter 1. The Complex Plane 14

If φ < 0 it is a rotation by angle −φ in the clockwise direction about

the origin.
In particular, z 7→ bz with b ∈ C \ {0} is a scaling by the factor |b| and
rotation by the angle arg(b).
Consider a more complicated map:

Inversion.

C \ {0} ∋ z 7→ w = z −1 ∈ C \ {0}; (1.2.13)

it is called an inversion.
Later we will prove that it transforms circles and straight lines into
circles and straight lines (but some straight lines can become circles and
some circles can become straight lines–think, which are suspect!)
We can solve easier problems today:
z2
w3 z1
Show that a straight line passing through
0 (with an angle φ) becomes the straight z3
line passing through 0 (with an angle −φ). w2
w1

Show that a circle about 0 (with a radius r)

becomes the circle about 0 (with a radius
r−1 ) and direction changes (from counter-
clockwise to clockwise).

Consider a circle passing through 0. We claim that it becomes a straight

line, not passing through 0.
Let us make first a rotation z 7→ eiφ z so after this we get a circle with a
center on the positive half-line of R and passing through 0: {z : |z − r| = r}.
But w = z −1 after this also rotates albeit in the opposite direction: w 7→
e−iφ z.
Chapter 1. The Complex Plane 15

Therefore
1 (1 + e−it )
z = r(1 + eit ) =⇒ w = =
r(1 + eit ) r(1 + eit )(1 + e−it )
(1 + e−it ) (1 + cos(t) − i sin(t)) 1 i sin(t)
= = = −
2r(1 + cos(t)) 2r(1 + cos(t)) 2r 2r(1 + cos(t))
1
and Re(w) = 2r
and
sin(t) 1
Im(w) = − = − tan(t/2)
2r(1 + cos(t)) 2r
(check it!) and when t runs from 0 to ±π, Im(w) runs from 0 to ∓∞.
So, we got a vertical straight line {z : Re(z) = 1/(2r)}.
Since inversion is self-inverse we conclude also that a straight line, not
passing through 0 becomes a circle passing through 0.

r(1 + eit )

0 x

r(1 + e−it )

1
2r (1 − i tan( 2t )

1
Figure 1.2: r = 2

Example 1.2.1. Find the locus of points satisfying

 
z−p
arg =φ (1.2.14)
z−q
with fixed φ.
Chapter 1. The Complex Plane 16

Solution. Note that (1.2.14) does not change under translations, rotations
and even scalings. So without any lost of generality one can assume that
p = r and q = −r (shifting their midpoint to 0 and then rotating). Assume
that 0 < φ < π (otherwise we permute p and q).
Also note that arg(z − p) is an angle between vector y
⃗ and the positive direction of axis x, and arg(z − q)
pz
is an angle between vector qz
⃗ nd the positive direction z
of axis x. Therefore (1.2.14) means that
  φ
z−p
φ = arg =
z−q
arg(z − p) − arg(z − q) = ∠(qzp)

(in counter-clockwise direction) and from geometry it

is known that the locus of such points is an arc. q p x
 
z−p
Remark 1.2.2. The remaining arc of the circle is defined by arg z−q
= φ−π.

1.3 Subsets of the Plane

1.3.1 Neighbourhoods and Open Sets
We study today the topology of complex plane C. Except the last part,
covering infinity ∞ and topology of the extended complex plane C, b we
already know it: topology on C coincides with topology on the ordinary
plane R2 .
A neigbourhood (or vicinity) of z0 ∈ C is any open disk Dε (z0 ) :=
{z : |z − z0 | < ε} of radius ε > 0 centered at z0 . Sometimes we say explicitly
ε-neigbourhood (or ε-vicinity).
◦
Definition 1.3.1. (a) Interior of set M ⊂ C is a set M of points z, which
are contained in M together with some vicinity Dε (z), ε depends on z.
◦
(b) M ⊂ C is an open if M = M , which means that some neighbourhood
every point z ∈ M belongs to M .
Chapter 1. The Complex Plane 17

z z
Dε (z) Dε (z)

(a) This set is open (b) This set is not

Remark 1.3.1. The notion of an open set depends where we consider it (and
the same will be true for the notions of the interior of a set, a closed set, a
closure of the set). For example Open interval (a, b) ⊂ R is an open subset
of R but not of C and its interior in C is empty:

See other examples on page 23 of the Textbook.

1.3.2 The Boundary of a Set

Definition 1.3.2. (a) A point z is a boundary point of a set M if every
open disk centered at z contains both points of M and points not in
M.
(b) The set of all boundary points of a set M is called the boundary of M
and denoted ∂M :
∂M = {z : ∀ε > 0 Dε (z) ∩ M ̸= ∅, Dε (z) \ M ̸= ∅}.
Theorem 1.3.1. (i) Set M is open if and only if it contains no boundary
point.
(ii) Interior of the set M is a set of all points, belonging to M but not to
its boundary.
(iii) The boundaries of M and of its complement C \ M = {z : z ∈
/ M}
coincide: ∂M = ∂(C \ M ).
See examples on pages 24–25 of the Textbook.
Chapter 1. The Complex Plane 18

1.3.3 Closed Sets

Closed sets

Definition 1.3.3. (a) A set M is closed if it contains its boundary: M ⊃

∂M .

(b) A closure of the set M is a union of M and its boundary: M = M ∪∂M .

Theorem 1.3.2. A set M is closed if and only if its complement C \ M =

{z : z ∈
/ M } is open.

1.3.4 Connected Sets

Definition 1.3.4. (a) A polygonal curve is the union of a finite number
of directed line segments P1 P2 , P2 P3 , . . . , Pn−1 Pn where the terminal
point of one is the initial point of the next (except for the last one).

(b) An open set M is connected if each pair of z ∈ M , w ∈ M may be

joined by a polygonal curve lying entirely with M . That is, there are
points P2 , . . . , Pn−1 such that all the line segments such that all line
segments P1 P2 ,. . . , Pn−1 Pn belong to M , P1 = z, Pn = w.

(a) This set is connected (b) This set is connected

(c) This set is not connected (d) This set is not connected

See other examples on page 26 of Textbook.

Chapter 1. The Complex Plane 19

Remark 1.3.2. We introduced a notion of connectivity only for open sets.

Definition 1.3.5. An open connected set is called a domain.

Domains are the natural setting for the study of analytic and harmonic
functions.

Definition 1.3.6. A set M is convex if the line segment zw joining each

pair of points z ∈ M and w ∈ M also lies in M .

Definition 1.3.7. A set M is starlike if there is a point z ∈ M such that

line segment zw joining each pair of points z with any point w ∈ M also
lies in M .

Obviously, any convex open set is also starlike, and any starlike set is
also connected.

(a) Convex set (b) Starlike but not con- (c) Connected but not star-
vex set like set

1.3.5 The Point at Infinity

Finally, let us introduce the point at infinity (∞) which will be treated
mostly as an ordinary point.
Let us start from the toy-model. Consider a circle S1 , touching at 0 a
straight line R. The opposite point N will be a “North Pole”.
For each point K ∈ R we draw a straight line KN ; denote by K ′ it’s
intersection with S1 . We get a bijection R and S1 \ N . We add to R a point
∞ which in this bijection corresponds to N . So we get a bijection between
S1 and R b = R ∪ {∞} which we call extended real line. Note that in this
construction there is just one infinity ∞, not two +∞ and −∞.
Now let us consider a sphere S2 , touching at 0 a plane R2 ≃ C. The
opposite point N will be a “North Pole”.
Chapter 1. The Complex Plane 20

K′

K′

For each point K ∈ C we draw a straight line KN ; denote by K ′ it’s

intersection with S2 . We get a bijection C and S2 \ N . We add to C a
point ∞ which in this bijection corresponds to N . So we get a bijection
between S2 and C b = C ∪ {∞} which we call extended complex plane. Also
a parallel of S passing through K ′ corresponds to a circle centered at 0
2

passing through K. Again, in this construction is just one infinity ∞.

N
Chapter 1. The Complex Plane 21

Note that in this correspondence to the neighbourhood of N on S2 (a

cap above red parallel) corresponds the exterior of the blue circle while to
the rest of S2 corresponds an interior of the blue circle DR (0).
Therefore, the neighbourhood of ∞ is {z : |z| > R} with arbitrarily large
R.

Recall the inversion z 7→ w = z1 . Now we can say that it maps C

b onto C,
b
−1
and 0 to ∞, and ε-neighbourhood of 0 onto ε –neighbourhood of ∞, and
∞ to 0, and R-neighbourhood of ∞ onto R−1 –neighbourhood of 0.
In particular, set M ⊂ C b containing ∞, is open only if it contains
neighbourhood of ∞.

1.4 Functions and Limits

1.4.1 Definitions
Definition 1.4.1. (a) A function of a complex variable z is a map from
domain D ⊂ C to C: f : D → C, or more detaily, D ∋ z → w =
f (z) ∈ C. Reminder: domain is an open connected set.

(b) D is called domain of definition of f .

(c) The set of all possible values is called a range of f .

Remark 1.4.1. (a) If a function is given by some formula, but domain is

not indicated, we consider a natural domain, where f is defined by
this formula.

(b) We are not interested in all functions. Severe restrictions will be

imposed late.
Chapter 1. The Complex Plane 22

Example 1.4.1. (a) f (z) = z 3 − 1 has a domain C, and the range C as

well.
1
(b) f (z) = has a domain C \ {i} (all C except i) and a range C \ {0}
z−i
(all C except 0).

(c) f (z) = z has a domain C and a range C.

(d) f (z) = Re(z) has a domain C and a range R.

i
(e) f (z) = has a domain C \ iR (all C except of the imaginary axis)
Re(z)
and a range iR \ {0} (all imaginary axis exept 0).
1+z
Example 1.4.2. Function f (z) = with a domain (now domain is
1−z
indicated!) D = {z : |z| < 1} has a range {w : Re(w) > 0}.
Indeed,

1 − |z|2
   
1+z (1 + z)(1 − z)
Re = Re =
1−z (1 − z)(1 − z) |1 − z|2

which is positive for |z| < 1.

1+z w−1
On the other hand, solving w = we get z = . If Re(w) > 0
1−z 1+w
then |w − 1| < |w + 1| and therefore |z| < 1.

1.4.2 Graphs
We would like to plot functions of complex variables, but this is not really
possible. Indeed, the plot of a real-valued function of a real variable is a
two-dimensional picture.
The plot of a real-valued function of the complex variable (that is of two
real variables) is a three-dimensional picture which could be plotted.
But the plot of a complex-valued function of a complex variable is four
dimensional picture, so we draw domain and range. F.e. in Example 1.4.2
we show a picture like this:
Chapter 1. The Complex Plane 23

1+z
w=
1−z

{z : |z| < 1} {w : Re(w) > 0}

Remark 1.4.2. (a) Wolfram’s Mathematica and Waterloo Maple provide

visualization for complex functions using ComplexPlot but the result
is colour-coded and really difficult to decipher.
(b) On the other hand, Wolfram’s Alpha (https://wolframalpha.com)
interprets it as as Plot and creates 3D pictures of Re(f (z)), Im(f (z))
and makes a plots like in the previous picture.
Try there Plot zˆ2 , Plot zˆ3, Plot zˆ3 - 3z.

1.4.3 Limits
The notions of limits and continuity is due to the same notions from Calculus
I and Calculus II.
Definition 1.4.2. (a) A sequence of complex numbers {zn }∞
n=1 converges
to A (equivalently, has a limit A),

zn → A as n → ∞ or lim zn = A
n→∞

if for any ε > 0 there exists N = N (ε) such that n ≥ N =⇒ |zn −A| <
ε.
(b) Otherwise we say that {zn }∞
n=1 diverges.

See examples on pages 33–34 of the Textbook.

Theorem 1.4.1 (Theorem 1 from page 34 of the Textbook). Let zn → A,
wn → B, α ∈ C. Then
(i) αzn → αA, zn + wn → A + B, zn wn → AB;
Chapter 1. The Complex Plane 24

(ii) z n → A, |zn | → |A|, Re(zn ) → Re(A), Im(zn ) → Im(A);

zn A
(iii) If B ̸= 0 then → .
wn B
Proof. Proof is trivial. See on page 34 of the Textbook.

Definition 1.4.3. Consider f (z) in the domain D and let z0 belong to the
closure of M , that is is either in M or on its boundary. We say that f (z)
tends to L as z tends to z0 , or f (z) → L as z → z0 , or f (z) has a limit L
at z0 , or

lim f (z) = L
z→z0

if for any ε > 0 there exists δ = δ(ε) > 0, such that

z ∈ M, |z − z0 | < δ =⇒ |f (z) − L| < ε.

See examples from page 35 of the Textbook.

Consider now a limit at ∞.

Definition 1.4.4. Assume that ∞ belongs to D, or at least, D is unbounded

(domain is bounded if it is contained in some large disk: D ⊂ {z : |z| <
R} for some R, domain is unbounded if it is not contained in any disk:
D ∩ {z : |z| > R} ≠ ∅ for any R.
We say that f (z) tends to L as z tends to ∞, or f (z) → L as z → ∞ ,
or f (z) has a limit L at ∞, or

lim f (z) = L
z→∞

if for any ε > 0 there exists R = R(ε), such that

z ∈ M, |z| > R =⇒ |f (z) − L| < ε.

See examples from pages 35–36 of the Textbook.

z
Example 1.4.3. Function f (z) = has no limits as z → 0 or z → ∞.
z
Indeed, as arg(z) = θ, f (z) = e2iθ so the limit exists along straight lines
but depend on the direction.
Chapter 1. The Complex Plane 25

Theorem 1.4.2 (Theorem 2 from page 36 of the Textbook). Let f (z) → A

and g(z) → B as z → z0 (where now z ∈ C
b to cover also z0 = ∞), α ∈ C.
Then

(i) αf (z) → αA, f (z) + g(z) → A + B, f (z)g(z) → AB as z → z0 ;

(ii) f (z)n → A, |f (z)| → |A|, Re(f (z)) → Re(A), Im(f (z)) → Im(A) as
z → z0 ;
f (z) A
(iii) If B ̸= 0 then → .
g(z) B

1.4.4 Continuity
Definition 1.4.5. Function f (z) is continuous at point z0 if f is defined in
z0 and

lim f (z) = f (z0 ).

z→z0

See examples on page 36 of the Textbook.

Theorem 1.4.3 (Theorem 3 from page 37 of the Textbook). Let f (z) and
g(z) be continuous at z0 , α ∈ C. Then

(i) αf (z), f (z) + g(z), f (z)g(z) are continuous at z0 ;

(ii) f (z), |f (z)|, Re(f (z)), Im(f (z)) are continuous at z0 ;

f (z)
(iii) If g(z0 ) ̸= 0 then is continuous at z0 .
g(z)

Corollary 1.4.4. (i) Polynomial P (z) = a0 z n + a1 z n−1 + . . . + an is

continuous in C.
P (z)
(ii) Rational function where P (z) and Q(z) are two polynomials is
Q(z)
continuous in C except of roots of Q(z).
Chapter 1. The Complex Plane 26

1.4.5 Infinite Series

The notions of the converging, absolutely converging and diverging series
repeat those from Calculus I.
Definition 1.4.6. Consider series ∞
P
n=1 an . Then

(a) It is convergent
PN and its sum its sum is S if the sequence of partials
sums SN = n−1 an converges to S.
(b) It is divergent if the sequence of partial sums diverges.
See Examples on pages 38–40 of the Textbook.
The following theorem is proven the same way as in Calculus I:
Theorem 1.4.5. Let series ∞
P
n=1 an converges absolutely, that is
∞
X
|an | < ∞.
n=1

Then it converges.
Power series will play a crucial role in this class.

1.4.6 Some Sums

The geometric series:
N
X 1 − z N +1
zn = , z ̸= 1
n=0
1−z
∞
X 1
=⇒ zn = , |z| < 1.
n=0
1−z

We can differentiate and integrate it.

1.5 Exponential, Logarithm, and

Trigonometric Functions
In this section we introduce functions of complex variable ez , and related
hyperbolic and trigonometric functions cosh(z), sinh(z), cos(z) and sin(z).
Chapter 1. The Complex Plane 27

These functions will coincide with the standard functions of real variable ex ,
cosh(x), sinh(x), cos(x) and sin(x) when z = x ∈ R, and satisfy the same
identities. It will turn out later that these functions are analytic.
We also introduce the inverse functions log(z), arccos(z) and arcsin(z),
which will be multi-valued functions of complex variable z, and their single-
valued variants Log(z), Arccos(z) and Arcsin(z). These functions also
coincide with the corresponding functions of the real variable when z = x ∈ R
and it will turn out later that these functions are analytic.

1.5.1 Exponential Function

We define exponential function

ez := ex eiy = ex cos(y) + i sin(y)

which is completely natural as we want

ez+w = ez ew (1.5.1)

and we already defined eiy = cos(y) + i sin(y). Sometimes instead of ez we

write exp(z). One can check easily, using ordinary identities for cosine and
sine of the sum (for real arguments) that (1.5.1) holds.
Further,

|ez | = eRe(z) (1.5.2)

1/2
Indeed, |eiy | = cos2 (y) + sin2 (y) = 1 =⇒ ex+iy = |ex | · |eiy | =
ex · 1 = ex .
Using Wolfram Alpha we can plot Re(ez ) = ex cos(y) and Im(ez ) =
ex sin(y):
https://www.wolframalpha.com/input/?i=Plot+e%5Ez
Let us analyze complex map in details. Since this function is 2πi-periodic:

ez+2πi = ez (1.5.3)

we restrict ourselves to the strip 0 ≤ Im(z) < 2π:

Since ez ̸= 0, 0 does not belong to range of ez but all other values could
be achieved. So range of ez is C \ {0}.
Chapter 1. The Complex Plane 28

0 ≤ Im(z) < 2π

Observe that the horizontal lines {z : Im(z) = const}, that is z = x + yi

with −∞ < x < ∞, y = const, become rays w = ex (cos(y) + i sin(y)) that
is {w : arg(w) = y = const}.
Meanwhile, vertical segments {z : Re(z) = const, 0 < z < 2π}, that is
z = x + yi with x = const, 0 ≤ y < 2π, become circles w = ex (cos(y) +
i sin(y)) that is {w : |w| = ex }.

(a) z (b) w = ez

Remark 1.5.1. (a) When x in the horizontal lines runs from (−∞, ∞),
|w| runs (0, ∞) (so rays are outward),

(b) When y in the vertical segments runs from 0 to 2π, arg(w) runs
(0, 2π) (so circles go counter-clockwise).

(c) One can see that half-strip {z : 0 ≤ Im(z) < 2π, Re(z) < 0} is
mapped to the unit disk with a punched center {0 < w : |w| < 1}, while
half-strip {z : 0 ≤ Im(z) < 2π, Re(z) > 0} is mapped to the exterior of the
unit disk {w : |w| > 1}.

(d) This function is 2πi-periodic: ez+2πi = ez . In complex variables

periods are not necessary real (may be complex) and there are analytic
functions which have two non-proportional complex periods. For example:
Chapter 1. The Complex Plane 29

https://mathworld.wolfram.com/WeierstrassEllipticFunction.html

(it is way too advanced for our class, we are not covering it, but look at this
beauty).

1.5.2 Logarithm Function

We want to consider an inverse function to ez . From the definition of ez it
follows immediately, that the inverse function log(z) is defined up to 2πin,
n ∈ Z, thus it is multivalued, and its domain is C \ {0} (that is the plane
with a punched {0}),
log(z) := ln(|z|) + i arg(z). (1.5.4)
We can make it single-valued
Log(z) := ln(|z|) + i Arg(z) (1.5.5)
but it will be discontinuous.
Going around the origin once in the counter-clockwise direction adds
2πi to log(z), so to make Log(z) continuous we need not to allow it.
We do it by imposing a cut (−∞, 0] ⊂ R, that is, we consider C\(−∞, 0].

(a) z (b) w = Log(z)

and take Arg(z) ∈ (−π, π) (we are opportunistic here and change the
definition as we see fit!); then Log(z) maps C \ (−∞, 0] onto strip {w : − π <
Im(w) < π}.
We selected a branch Arg(z) of arg(z). Two different sides of the cut
go to two different horizontal straight lines. Cuts always have two sides and
in mappings they usually separate.
Chapter 1. The Complex Plane 30

(a) z (b) w = Log(z)

1.5.3 Fractional pPowers

Using logarithm function we can define any power of z:
z α = eα log(z) (1.5.6)
with α ∈ C. This function is single-valued only when α ∈ Z (real integer).
Indeed, since log(z) is defined up to 2πin, z α is defined up to a factor e2πiαn
and it is 1 for all n ∈ Z if and only if α ∈ Z.
Furthermore, if α ∈ Q, that is α = p/q is a real rational number, with
p/q irreducible ratio (so integers p and q have only ±1 as common divisors,
q > 0) this is q-valued function, with q different branches, while otherwise it
has an infinite number of different branches.
Consider mappings. We want them to be bijective. We consider only
real α ̸= 0. Let us assume that 0 < α < 1. Let us consider C \ (−∞, 0]
which is mapped to a sector {w : | Arg(w)| < απ}.

(a) z (b) w = z α

Again different sides of the cut are mapped onto different pieces of the
boundary. Blue rays and red arcs on both pictures have same orientations.
Chapter 1. The Complex Plane 31

Remark 1.5.2. (a) Case α > 1 is reduced to this one since z α is inverse to
z 1/α . Then picture (a) and (b) interchange.

(b) One can consider −1 < α < 0, pictures remain the same, except the
sector now is {w : | Arg(w)| < −απ} and blue rays and red arcs on
both pictures have opposite orientations.

(c) Case α < −1 is reduced to the previous one.

Exercise 1.5.1. Think, what happens with in the cases 0 < α < 1 and
−1 < α < with

(a) The disk with a cut {z : |z| < R, | Arg(z)| < π},

(b) The exterior of the disk with a cut {z : |z| > r, | Arg(z)| < π},

(c) The ring with the cut {z : r < |z| < R, | Arg(z)| < π}.

1.5.4 Trigonometric Functions

Now we want to introduce trigonometric functions of complex variable. They
should satisfy the same identities as the corresponding functions of the real
variable, and coincide with them as z = x ∈ R. Let

eiz + e−iz
cos(z) = ,
2
and
eiz − e−iz
sin(z) =
2i
because we have these formulas for real z.
Let also
sin(z) cos(z)
tan(z) = , cot(z) = ,
cos(z) sin(z)
and
1 1
sec(z) = csc(z) =
cos(z) sin(z)

also because we have these formulas for real z.

Chapter 1. The Complex Plane 32

Plugging expression for eiz = e−y (cos(x) + i sin(x)) with z = x + iy we

get

cos(z) = cos(x) cosh(y) − i sin(x) sinh(y),

sin(z) = sin(x) cosh(y) + i cos(x) sinh(y)

with hyperbolic functions of real vatriable

ey + e−y ey − e−y
cosh(y) = , sinh(y) = .
2 2
We can plot cos(z), tan(z) and sec(z) using
https://www.wolframalpha.com/input/?i=Plot+cos(z)
https://www.wolframalpha.com/input/?i=Plot+tan(z)
https://www.wolframalpha.com/input/?i=Plot+sec(z)
while plots sin(z), cot(z) and csc(z) are obtained by shifts and/or reflections.
Let us consider complex plot of cos(z). We restrict z to a strip {z : 0 <
Re(z) < π}: and look where it maps to.

(a) Vertical lines {z : x = Re(z) = const, y = Im(z)} are mapped to

w = u + iv with v = − sin(x) sinh(y), u = cos(x) cosh(y) which are
branches of hyperbolas satisfying
u2 v2
− = 1,
A2 B 2
with A = | cos(x)|, B = sin(x).
(b) Horizontal segments {z : 0 < x = Re(z) < π, y = Im(z) = const} are
mapped to w = u + iv with v = − sin(x) sinh(y), u = cos(x) cosh(y)
which are halves of ellipses satisfying
u2 v 2
+ 2 =1
a2 b
However
withthe
a =same w =bcos(z)
cosh(y), restricted to a half-strip {z : 0 < Re(z) <
= | sinh(y)|.
2π, Im z > 0}, maps it onto the same plane with a different cut: C \ [−1, ∞].

(a) Vertical half-lines {z : x = Re(z) = const, y = Im(z)} are mapped to

w = u + iv with v = − sin(x) sinh(y), u = cos(x) cosh(y) which are
half-branches of the same hyperbolas as in before.
Chapter 1. The Complex Plane 33

(a) z (b) w = cos(z)

(a) z (b) w = cos(z)

(b) Horizontal segments {z : 0 < x = Re(z) < 2π, y = Im(z) = const} are
mapped to w = u + iv with v = − sin(x) sinh(y), u = cos(x) cosh(y)
which are the same ellipses as before.

The same is true for sin(z), except strips are shifted and flipped due to
sin(z) = cos( π2 − z).
sin(z)
Finally, tan(z) = cos(z) gives us a picture of Circles of Appolonius with
= i and q = −i, but explanation why it happens and the discussion of
Chapter 1. The Complex Plane 34

possible domains and ranges will wait until Chapter 3.

-i

1.5.5 Inverse Trigonometric Functions

Consider inverse functions:
eiz + e−iz
w = cos(z) = =⇒ e2iz + 1 − 2weiz = 0 =⇒
2 √ √
eiz = w ± w2 − 1 =⇒ iz = log w ± i 1 − w2 .

Therefore
√

arccos(w) = − i log w ± 1 − w2 .
Similarly
√

arcsin(w) = − i log iw ± 1 − w2
Chapter 1. The Complex Plane 35

and
i  1 − iw 
arctan(w) = log .
2 1 + iw
In particular, range of cos(z) and sin(z) in C (all values are achieved), and
range of tan(z) is C \ {i, −i} (all values except ±i are achieved).
For single-valued functions we have formulas
√

Arccos(w) = − i log w ± 1 − w2 ,
√

Arcsin(w) = − i log iw + 1 − w2
and
i  1 − iw 
Arctan(w) = Log
2 1 + iw
√
where on the domain (with cuts removed) Re( 1 − w2 ) > 0,
which defines branch uniquely.

Exercise 1.5.2. Check these formulas!

On the other hand, usual formulas hold
arccos(w) = ± Arccos(w) + 2πn,
arcsin(w) =(−1)n Arcsin(w) + πn,
and
arctan(w) = Arctan(w) + πn, n ∈ Z.
Exercise 1.5.3. Calculate Arccos(2).

1.5.6 Hyperbolic and Inverse Hyperbolic Functions

We also have hyperbolic functions
ez + e−z ez − e−z sinh(z)
cosh(z) = , sinh(z) = , tanh(z) = ,
2 2 cosh(z)
Chapter 1. The Complex Plane 36

but due to the formulas

cos(z) = cosh(iz), sin(z) = −i sinh(iz), tanh(z) = −i tan(iz)

their properties and properties and formulas for inverse hyperbolic functions
follow from the properties of trigonometric functions.
Example 1.5.1. Range of tanh(z) is C \ {1, −1}.

1.6 Line Integrals and Green’s Theorem

1.6.1 Curves
In this section we simply repeat what was studied in Calculus II, but with
the “Complex Variables flavour”.

Definition 1.6.1. (a) A curve γ is a continuous complex-valued function

γ(t) defined for t in some interval [a, b] in the real axis; in other words,
γ : [a, b] → C.

(b) The curve γ is simple if

a ≤ t1 < t2 ≤ b &γ(t1 ) = γ(t2 ) =⇒ t1 = a & t2 = b,

that is, it does not have self-intersections (but may be closed)

(c) The curve γ is closed if γ(a) = γ(b), that is the initial point γ(a)
coincides with the end-point γ(b).

(a) This curve is simple (b) This curve is simple

and closed but not closed

(c) This curve is closed but (d) This curve is neither

not simple closed nor simple
Chapter 1. The Complex Plane 37

Remark 1.6.1. The famous Jordan Curve Theorem asserts that the comple-
ment of the range of a curve, which is simple and closed, consists of two
disjoint open connected sets, one bounded and the other unbounded.
The bounded piece is the inside of the curve and the unbounded piece
the outside.
Despite the almost painful obviousness of this statement, the theorem is
hard to prove. We shall accept it as true.

outside
inside

Definition 1.6.2. (a) Curve is smooth if γ(t) is continuously differen-

tiable (that is, both Re(γ(t)) and Im(γ(t)) are continuously differen-
tiable).
(b) Curve is piecewise smooth if γ(t) is piecewise continuously differentiable
(that is, there exist a = t0 < t1 < t2 < . . . < tn−1 < tn = b such that
γ(t) is continuously differentiable on [t0 , t1 ], [t1 , t2 ],. . . ,[tn−1 , tn ].

(a) This curve is smooth (b) This curve is piece-

wise smooth

Remark 1.6.2. (a) It is very common and convenient to refer to the range
of γ(t) as the curve γ and to γ(t) itself as the parametrization of the
curve.
(b) With this use of the word curve, a curve becomes a concrete geometric
object such as a circle or a straight line segment and hence is easily
Chapter 1. The Complex Plane 38

visualized. The difficulty with this view is that a particular curve has
many different parameterizations. However, our results under very
broad assumptions would not depend on parametrizations.

(c) What is more, for closed curves the results would not depend on the
choice of start-point (which is also an end-point).

(d) We extend the notion of the curve, requiring γ(t) to be only piecewise
continuous, that is, consisting of several curves in the old understand-
ing.

(e) Then we can even select a parametrization on each piece separately.

Definition 1.6.3. (a) Standard parametrization of a straight segment

from z0 to z1 : z(t) = z0 (1 − t) + z1 t, 0 ≤ t ≤ 1.

(b) Standard parametrization of a circular arc with a center at c and

radius R: z(t) = c + Reit with θ0 < t < θ1 .

θ1 θ0
c
z1

z0

1.6.2 Line Integral

Definition 1.6.4. Let γ = γ(t) be an oriented smooth curve and let f (z)
be a complex-valued function on this curve. Then
Z Z t1
f (z) dz := f (γ(t))γ ′ (t) dt
γ t0

is a line integral of f along γ.

Chapter 1. The Complex Plane 39

Remark 1.6.3. This line integral could be written as

Z Z t1
f (x(t) + iy(t)) (dx + idy) := f (x(t), y(t))(x′ (t) + iy ′ (t)) dt
γ t0

where γ(t) = x(t) + iy(t), dγ = dx + idy.

So we reduced the notion of the complex line integral to the notion of the
real line integral (complexity of f is not important) and it has the following
properties

(a) γ f (z) dz does not depend on parametrization, that means, if γ ′ (s) =

R

γ(t(s)) where t = t(s) is continuously differentiable and t(s0 ) = t0 ,

t(s1 ) R= t1 (where now we do not even assume that t0 < t1 or s0 < s1
then f (z) dz does not change.
(b) On the other hand,
Z Z
f (z) dz = − f (z) dz
−γ γ

where −γ is the same curve but with the opposite parametrization:

Z Z t0
f (z) dz = f (γ(t))γ ′ (t) dt.
−γ t1

(c) Also
Z Z Z
f (z) dz = f (z) dz + f (z) dz
γ1 +γ2 γ1 γ2

provided γ1 = γ(t) with t running from t0 to t1 , γ2 = γ(t) with t

running from t1 to t2 , and γ = γ(t) with t running from t0 to t2 .
(d) It allows us to introduce
Z Z Z
f (z) dz := f (z) dz + . . . + f (z) dz
γ γ1 γn

for any piecewise smooth curve γ = γ1 + . . . + γn , even consisting of

several disjoint segments, because the right-hand expression does not
depend on breaking γ into segments γ1 , . . . , γn . However, orientation
matters!
Chapter 1. The Complex Plane 40

(e) Integral over closed contour does not depend which point is start- and
end-point. Indeed, if such points are a and b then the path with start-
and end-point a is γ = γ1 + γ2 , and the path with start- and end-point
b is γ ′ = γ2 + γ1 :

γ1 b

γ2

I
(f) For integral over closed contour a special notation often is used:
γ
But recall: orientation matters!

Remark 1.6.4. (a) Line integral differs from length integral

Z Z t1 p
f (z) |dz| := f (x(t), y(t)) (x′ (t))2 + (y ′ (t))2 dt
γ t0
p
where |dz| = (x′ (t))2 + (y ′ (t))2 dt is a length element and here we
assume that t0 < t1 . We will (almost) never use the length integral in
our course.
(b) Recall that length integral also was studied in Calculus II (even in
Calculus I) and this integral does not change sign when we change an
orientation (which means, parmute star- and end-points).
(c) We can also consider
Z Z
g(z) dz := g(z) (dx − idy).
γ γ

We will use the following important inequality, which follows from the
standard properties of integral :
Theorem 1.6.1.
Z Z
| f (z) dz| ≤ |f (z)| |dz|.
γ γ
Chapter 1. The Complex Plane 41

1.6.3 Green’s Formula: Real Variables

Let us recall the following from Calculus II:
Theorem 1.6.2 (Green’s formula, real variables).
I ZZ

∂N ∂M

M dx + N dy = − dxdy, (1.6.1)
γ D ∂x ∂y

where D is a bounded domain in R2 and γ = ∂D is it’s border, properly

oriented, M and N are smooth in the closure of D̄ = D ∪ γ (recall that D̄
is a closure of domain D.

1.6.4 Green’s Formula: Discussion

(a) What is properly oriented? If domain D is simply-connected (with no
holes inside), then γ is a simple closed curve and it means counter-
clockwise.
(b) If D is not a simply connected (with holes inside) then γ = γ0 + γ1 +
. . . + γN , where γ0 is the external boundary, counter-clockwise oriented
and and γ1 are inner boundaries (that means boundaries of the holes),
clock-wise oriented.
γ γ0

D γ1

(a) Simply connected do- (b) Not simply connected

main domain

Why? Let us make an infinitely thin cut (if N > 1 we need N cuts)
so that domain D∗ after cuts will be connected and simply-connected. Its
boundary γ ∗ = γ0 + ℓ + ℓ′ + γ1 is connected and the counter-clockwise
orientation of γ0 implies clockwise orientation of γ1 (and γ2 , . . . , γN ) and
the proper orientations of ℓ′ and ℓ.
Chapter 1. The Complex Plane 42

γ0 γ0
ℓ′ ℓ

γ1 γ1

D∗ D

∗ ∗
RR RR
Applying to D , γ Green’s formula we see that D ∗ = D
, and
′
H R R R R R R H
γ∗
= γ0
+ ℓ
+ ℓ′
+ =
γ1 R γ0R
+ γ1
= γ
because ℓ = −ℓ (have opposite
directions) and therefore ℓ + ℓ′ = 0, so corresponding integrals cancel one
another.

(a) Permute x and y, M and N , then orientation becomes opposite so

the left hand expression of (1.6.1) changes sign, and so does the right
hand expression.
I ZZ

∂N ∂M

M dx + N dy = − dxdy. (1.6.1)
γ D ∂x ∂y

(b) Let
∂N ∂M
− = 0. (1.6.2)
∂x ∂y
Then if a closed curve γ ′ is properly oriented and bounds a domain
D′ ⊂ D then
I ZZ

∂N ∂M

M dx + N dy = − dxdy = 0. (1.6.3)
γ′ D′ ∂x ∂y

(c) In simple connected domains every simple closed curve bounds some
subdomain; it is not so in domains which are not simply-connected:
(d) Therefore in simply-connected domains (1.6.2) implies that
I


M dx + N dy = 0 (1.6.4)
γ

for any simple closed counter-clockwise oriented curve γ.

Chapter 1. The Complex Plane 43

(e) We do not need to have γ counter-clockwise oriented or even simple:

for clockwise oriented integral will be still 0 since it just changes sign
and closed curves with self-intersections could be broken into simple
closed curves:

(we slightly cut corners here, since there could be an infinite number
of self-intersections).
(f) Then integral over curve which is not closed depends only on it’s start
and end-points:

γ2

γ1

Indeed, γ1 − γ2 is a closed curve, so

Z Z Z Z Z
0= = − =⇒ = .
γ1 −γ2 γ1 γ2 γ1 γ2

(g) Fix point (x0 , y0 ) ∈ D and define

Z 
U (x, y) = M dx + N dy + C (1.6.5)
γ

with an arbitrary constant C = U (x0 , y0 ) with integral taken over

any curve γ from (x0 , y0 ) to (x, y) because in framework of simply-
connected D and assumption (1.6.2) this integral does not depend on
the choice of γ.
Chapter 1. The Complex Plane 44

(h) Then dU = M dx + N dy. Indeed

Z (x+dx,y+dy) Z (x,y)
dU = (M dx + N dy) − (M dx + N dy) =
(x0 ,y0 ) (x0 ,y0 )
Z (x+dx,y+dy)
(M dx + N dy) = M dx + N dy.
(x,y)

Therefore

Theorem 1.6.3. If domain D is simply-connected and M dx + N dy is

closed that means
∂N ∂M
− = 0. (1.6.2)
∂x ∂y
then M dx + N dy is exact that means for some M dx + N dy = dU for some
function U and therefore
∂U ∂U
M= , N= . (1.6.6)
∂x ∂y

Remark 1.6.5. (a) U is defined up to constant C.

(b) Conversely, if M dx + N dy is exact, then it is closed (and we do not

need D to be simply-connected.

(c) On the other hand, if D is not simply-connected, then some closed

forms M d + N dy are not exact.
Example 1.6.1. Indeed, consider C \ {0} and

x dy − y dx
= d θ, θ = arg(x + yi);
x2 + y 2

so it is closed. However polar angle arg(x+yi) is not a single-valued function

on C \ {0} and

x dy − y dx
I
= 2π
γ x2 + y 2

where γ winds exactly once in the counter-clockwise direction around 0.

Chapter 1. The Complex Plane 45

1.6.5 Green’s Formula: Complex Variables

Theorem 1.6.4 (Green’s theorem, complex variables).
I ZZ

∂f ∂g

f dz + g dz = 2i − dxdy, (1.6.7)
γ D ∂z ∂z
where D is a bounded domain in C and γ = ∂D is it’s border, properly
oriented, f and g are smooth in D̄ = D ∪ γ and
∂f 1 ∂f ∂f

:= +i , (1.6.8)
∂z 2 ∂x ∂y
∂g 1 ∂g ∂g

:= −i . (1.6.9)
∂z 2 ∂x ∂y
Proof. Indeed, we need to apply “real” Green’s formula; since

f dz + g dz = f (dx + idy) + g(dx − idy)

= (f + g) dx + i(f − g) dy

we need to plug M = f + g, N = i(f − g), so we need to calculate

∂N ∂M ∂(if − ig) ∂(f + g) ∂f ∂f
∂g ∂g

− = − =i +i −i −i .
∂x ∂y ∂x ∂y ∂x ∂y ∂x ∂y

Remark 1.6.6. Why these notations (1.6.8) and (1.6.9)? Let us write
∂f ∂f
df = dx + dy. (1.6.10)
∂x ∂y
On the other hand dz = dx + idy, dz = dx − idy imply that dx = 12 (dz + dz)
and dy = − 2i (dz − dz) and therefore
1 ∂f ∂f
1 ∂f ∂f

df = −i dz + +i dz
2 ∂x ∂y 2 ∂x ∂y
and in these notations we extend the usual formula to complex variables:
∂f ∂f
df = dz + dz (1.6.11)
∂z ∂z
∂f ∂f
even if ∂z
and ∂z
are not partial derivatives in the sense of Calculus II.
Chapter 1. The Complex Plane 46

We will use only

Corollary 1.6.5. I ZZ
∂f
f dz = 2i dxdy, (1.6.12)
γ D ∂z
where D is a bounded domain in C and γ = ∂D is it’s border, properly
oriented.

Remark 1.6.7. If you took MAT257 (Analysis II = Calculus II on steroids),

then you know that

(a) M dx + N dy is called 1-form;

(b) one can write dxdy as dx ∧ dy.

(c) On the other hand dz ∧ dz = (dx + idy) ∧ (dx − idy) = −2idx ∧ dy

and therefore one can rewite (1.6.7) as
I ZZ

∂f ∂g

f dz + g dz = − − dz ∧ dz. (1.6.13)
L D ∂z ∂z

This is Green’s formula in (perfectly) complex form.

Chapter 2

Basic Properties of Analytic

Functions

2.1 Analytic and Harmonic Functions;

Cauchy-Riemann Equations
2.1.1 Analytic functions
Now we want to introduce our main definition.
Recall formulas (1.6.11), (1.6.8), (1.6.9):

∂f ∂f
df = dz + dz. (2.1.1)
∂z ∂z
with
∂f 1 ∂f ∂f

:= +i , (2.1.2)
∂z 2 ∂x ∂y
∂f 1 ∂f ∂f

:= −i . (2.1.3)
∂z 2 ∂x ∂y
Looks like the usual formula from Calculus II?
∂f ∂f
df = dx + dy. (2.1.4)
∂x ∂y

Think again! In (2.1.4) x and y are independent variables, so dx and dy

are independent, but in (2.1.1) z and z are not independent, and dz and dz

47
Chapter 2. Basic Properties of Analytic Functions 48

are not independent. In fact, z uniquely defines z, and dz uniquely defines

dz.
However,

Lemma 2.1.1. dz and dz are linearly independent: for any complex num-
bers A and B

A dz + B z = 0 ∀dz =⇒ A = B = 0.

Proof. Indeed, plugging dz = dx + idy, dz = dx − idy we get

(A + B) dx + i(A − B) dy = 0 ∀dx, dy

which implies A + B = A − B = 0 and then A = B = 0.

Therefore,

Corollary 2.1.2.

df = A dz + Bdz ∀dz
∂f ∂f
implies that A = ∂z
, B= ∂z
defined by (2.1.2)–((2.1.3).

However, we want to have a formula from Calculus I:

df (z) = f ′ (z) dz (2.1.5)

and we arrive to our main definition:

Definition 2.1.1. (a) Function f (z) is analytic in the domain D ⊂ C if

df (z) = A(z) dz holds for all dz and all z ∈ D.

(b) Then we call A(z) derivative of analytic f (z) function f (z) and denote
df
it by f ′ (z) and dz .

(c) Function analytic in C is called entire analytic function or, simply,

entire function.

Then we have:
Chapter 2. Basic Properties of Analytic Functions 49

Theorem 2.1.3. f (z) is analytic in domain D if and only if there

∂f 1 ∂f ∂f

:= +i = 0, (2.1.6)
∂z 2 ∂x ∂y

in which case (2.1.5) holds with

df ∂f
f ′ (z) = := . (2.1.7)
dz ∂z
Remark 2.1.1. It is equivalent to definition on page 77 of the Textbook:

(a) A function f defined for z in a domain D is differentiable at a point

z0 in D if the following limit exists:

f (z) − f (z0 ) f (z0 + h) − f (z0 )

lim = lim .
z→z0 z − z0 h→0 h

(b) This limit, if it exists, is denoted by f ′ (z0 ).

(c) If f is differentiable at each point of the domain D, then f is called

analytic in D.

(d) A function analytic on the whole complex plane is called entire.

Remark 2.1.2. Analytic functions are often called holomorphic.

Theorem 2.1.4. Let f and g be analytic in domain D. Then

(i) αf is analytic in D (α ∈ C and (αf )′ = αf ′ ;

(ii) f + g is analytic in D and (f + g)′ = f ′ + g ′ ;

(iii) f g is analytic in D and (f g)′ = f ′ g + f g ′ ;

f f ′ f ′ g−f g ′


(iv) g
is analytic in {z ∈ D : g(z) ̸= 0} and g
= g2
.

Proof. The proof is the same as in Calculus I: f.e.

d(f g) = g df + f dg = gf ′ dz + f g ′ dz = (gf ′ + f g ′ ) dz,

which implies Statement (iii).

Chapter 2. Basic Properties of Analytic Functions 50

Theorem 2.1.5. Let f be analytic in D and g be analytic in G, where

z ∈ D =⇒ f (z) ∈ G.
Then g ◦ f is analytic in D and (g ◦ f )′ (z) = (g ′ ◦ f )f ′ .
Proof. Indeed, like in Calculus I, let w = f (z), then

dg(w) = g ′ (w) dw = g ′ (w) · f ′ (z) dz.

Example 2.1.1. Polynomial

P (z) = an z n + an−1 z n−1 + . . . + a1 z + a0

is entire analytic function and

P ′ (z) = nan z n−1 + (n − 1)an−1 z n−2 + . . . + an−1 .

Proof. Proof follows from Theorem 2.1.4.

P (z)
Example 2.1.2. A rational function Q(z) where P (z) and Q(z) are polynomials,
is an analytic function on the set {z : Q(z) ̸= 0}.
Example 2.1.3. ez is an entire analytic function and (ez )′ = ez .
Proof. Indeed

dez =d ex cos(y) + i sin(y)




= cos(y) + i sin(y) ex dx + ex − sin(y) + i cos(y) dy

=ex cos(y) + i sin(y) dx + ex cos(y) + i sin(y) i dy

=ez (dx + idy) = ez dz.

Example 2.1.4. (a) cos(z), sin(z), cosh(z) and sinh(z) are entire analytic
functions and usual expression for derivatives hold.

(b) tan(z), sec(z) are analytic functions except when cos(z) = 0 (so except
{z = π2 + πn, n ∈ Z}) and usual expression for derivatives hold.

(c) cot(z), csc(z) are analytic functions except when sin(z) = 0 (so except
{z = πn, n ∈ Z} and usual expression for derivatives hold.
Chapter 2. Basic Properties of Analytic Functions 51

(d) tanh(z), is analytic function except when cosh(z) = 0 (so except

{z = iπ2 + iπn, n ∈ Z}) and usual expression for derivatives hold.
Example 2.1.5. Log(z) is an analytic function on C \ [0, ∞) (domain with a
cut) and (Log(z))′ = z1 .
Proof. Indeed, w = Log(z) is defined there and z = ew . Then dz = ew dw =
z dw =⇒ dw = dz z
.
Example 2.1.6. z α is an analytic function on C \ [0, ∞) (domain with a cut)
and (z α )′ = αz α−1 .
Proof. Indeed, z α = eα Log(z) is defined there and

′
′
z α = eα Log(z) = eα Log(z) × α × z −1 = αz α × z −1 = αz α−1 .

Remark 2.1.3. In these examples we can make a cut along any ray from 0
to ∞ and select a corresponding branch.
Later we will prove

Theorem 2.1.6 (Inverse function theorem). Let f be an analytic function

in the vicinity of z0 and f ′ (z0 ) ̸= 0. Let w0 = f (z0 ).
Then in the vicinity of w0 is uniquely determined inverse function g(w),
1
f (g(w)) = w and g(f (z)) = z, and g ′ (w) = ′ .
f (g(w))
Then we will prove analyticity of Arcsin(z) and Arccos(z) in C\((−∞, −1]∪
[1, ∞)) and Arctan(z) and Arccot(z) in C \ ((−i∞, −i] ∪ [i, i∞)) and usual
formulas for derivatives.

(a) Domain of (b) Domain of

Arcsin and Arccos Arctan and Arccot
Chapter 2. Basic Properties of Analytic Functions 52

2.1.2 Cauchy-Riemann equations

In the previous Subsection we derived Cauchy-Riemann equation in the
complex form (also we have not called them so):
∂f 1 ∂f ∂f

:= +i = 0. (2.1.8)
∂z 2 ∂x ∂y
Since f (z) = u(x, y) + iv(x, y) with u = Re(f ) and v = Im(f ) (here we write
argument z as a pair (x, y)) we get
∂ ∂
∂u ∂v
∂v ∂u

+i (u + iv) = − +i +
∂x ∂y ∂x ∂y ∂x ∂y
and therefore (2.1.8) (which is equivalent to analyticity of f ) is equivalent
to the pair of real equations. Thus we arrive to the following theorem:
Theorem 2.1.7. Function f = u + iv is an analytic function of the complex
variable z if and only if u and v satisfy the following Cauchy-Riemann
equations in the real form (or simply Cauchy-Riemann equations):
∂u ∂v
= , (2.1.9)
∂x ∂y
∂u ∂v
=− . (2.1.10)
∂y ∂x
Differentiating the first equation by x and the second by y (or by y and
x and substracting) and adding we get
∂ 2u ∂ 2u
∆u := + = 0, (2.1.11)
∂x2 ∂y 2
∂ 2v ∂ 2v
∆v := + = 0. (2.1.12)
∂x2 ∂y 2
2 2
Definition 2.1.2. Operator ∆ = ∂∂xu2 + ∂∂yu2 is called Laplace operator (or
2 2
Laplacian), equation ∆u := ∂∂xu2 + ∂∂yu2 = 0 is called Laplace equation, and
functions satisfying it harmonic functions.
Therefore we have proven
Theorem 2.1.8. The real and imaginary parts of complex analytic function
f are harmonic functions. Also f itself is a harmonic function.
Definition 2.1.3. Harmonic functions u and v, satisfying Cauchy-Riemann
equations, are called conjugate harmonic functions.
Chapter 2. Basic Properties of Analytic Functions 53

Question. Knowing harmonic function u (or v) can we restore a conjugate

harmonic function v (u respectively) and therefore an analytic function f ?
To answer this question assume that we know u and look at equations
(2.1.9) and (2.1.10):

∂v ∂u ∂v ∂u
= M := − , = N := .
∂x ∂y ∂y ∂x

We know that it is necessary to have ∂N ∂x

− ∂M
∂y
= 0 and in the simply-
∂N ∂M
connected domain it is also sufficient. Since ∂x − ∂y = ∆u we arrive to the
following theorem:

Theorem 2.1.9. Let u be a harmonic function in D. Then

(i) If D is simply-connected then there exists a conjugate harmonic func-

tion v.

(ii) This function v is defined up to additive constant.

It immediately implies

Corollary 2.1.10. (i) In the simply-connected domain an analytic func-

tion could be restored by its real (or imaginary) part.

(ii) This analytic function is defined uniquely up to an imaginary (corre-

spondingly real) additive constant.

Example 2.1.7. One can check easily that u = x3 − 3xy 2 is a harmonic

function in C. To find a conjugate harmonic function we write
∂v ∂v
= 6xy, = 3x2 − 3y 2 .
∂x ∂y
Let us recall how we solve such overdetermined systems in MAT244 ODE:
Integrating the first equation by x we get v = 3x2 y + h(y) with unknown
function h(y). Plugging it to the second equation we get h′ (y) = −3y 2 .
Therefore h(y) = −y 3 + C. Then v(y) = 3x2 y − y 3 + C and

f = x3 − 3xy 2 + i(3x2 y − y 3 + C) = (x + iy)3 + Ci = z 3 + Ci.

Chapter 2. Basic Properties of Analytic Functions 54

Example 2.1.8. Consider multivalued analytic function f (z) = log(z) :=

ln |z| + i arg(z) in C \ {0}. It’s real part is ln |z| = 12 ln(x2 + y 2 ). Since f is
analytic its real part u = 12 ln(x2 + y 2 ) must be harmonic in R2 \ {0}, but
you can check it independently.
If we consider C with a proper cut (from 0 to ∞ along some ray), we can
select here a branch v = arg(x + iy) as a single valued harmonic function.
F.e. if we make a cut {(0, y) : y ≤ 0} we can select v = Arctan(y/x) for
x > 0 and v = π + Arctan(y/x) for x < 0 and v = π/2 for x = 0, y > 0 (but
there will be a jump at x = 0, y < 0).
However, a single valued function v (and thus f ) does not exist. One
can explain it calculating
Z 2π
x dy − y dx
I I
(M dx + N dy) = = dθ = 2π ̸= 0.
γ γ x2 + y 2 0

where γ is a circular curve centered in 0 and counter-clockwise oriented.

Finally, let us prove

Theorem 2.1.11. Let f be an analytic function in D. Then

(i) Re(f ) = const =⇒ f = const and Im(f ) = const =⇒ f = const.

(ii) |f | = const =⇒ f = const and arg(f ) = const =⇒ f = const.

Proof. (i) Indeed, if u = Re(f ) = const, then ∂u

∂x
= ∂u
∂y
= 0 and from
∂v ∂v
Cauchy-Riemann equations we see ∂x = ∂y = 0, and v = const.
The second part of Statement (i) is proven in the same way.

(ii) If |f | = 0 then f = 0. Let |f | = c ̸= 0. Then g := log(f ) is an

analytic function with a constant real part ln(|f |) and therefore according
to Statement 1 it is constant: g = const. Then f = eg = const.
The second part of Statement (ii) is proven in the same way.
Chapter 2. Basic Properties of Analytic Functions 55

2.2 Power series

2.2.1 Definition
Definition 2.2.1. A power series is an infinite series
X∞
an (z − z0 )n , (2.2.1)
n=0

where a0 , a1 , . . . are complex numbers, called coefficients of the series, z0 is

a fixed point called a center of the series.
Example 2.2.1. (a) A polynomial
N
X
P (z) = an z n
n=0

(b) Geometric series

∞
X 1
f (z) = zn = , |z| < 1.
n=0
1−z

(c) Other important series, converging everywhere

∞ ∞ ∞
X zn X z 2n X z 2n+1
ez = , cos(z) = (−1)n , sin(z) = (−1)n .
n=0
n! n=0
(2n)! n=0
(2n + 1)!
Theorem 2.2.1. Suppose there is some z1 ̸= z0 such that the series (2.2.1)
converges.
Then this series converges absolutely and uniformly in every disk
{z : |z − z0 | ≤ r} with r < |z1 − z0 |.
Proof. Since series ∞ n
P
n=0 an (z1 − z0 ) converges, then

lim |an | |z1 − z0 |n = 0.

n→∞
In particular, for some constant M
|an | |z1 − z0 |n ≤ M ∀n
and therefore
|z − z0 |
|an | |z − z0 |n ≤ M ρn , ρ=
|z1 − z0 |
with ρ < 1 ⇐⇒ |z − z0 | ≤ r < |z1 − z0 |.
for all n, P
Since ∞ n
n=0 M ρ < ∞, our series converges absolutely and uniformly.
Chapter 2. Basic Properties of Analytic Functions 56

2.2.2 Radius and disk of convergence

Therefore, there are exactly three mutually exclusive cases:

Case 1. Series
X
S(z) := an (z − z0 )n (2.2.1)
n=1

converges for z = z0 only.

Case 2. Series (2.2.1) converges for all z ∈ C.

Case 3. Series (2.2.1) converges for some z ̸= z0 but not for all of them.

Explore Case 3. So, there are z1 ̸= z0 for each series (2.2.1) converges, and
z2 ̸= z0 for which it diverges.
By Theorem 2.2.1 it is impossible that |z1 − z0 | > |z2 − z0 | and therefore
|z1 − z0 | ≤ |z2 − z0 |. Therefore, there exists R : 0 < R < ∞ such that

- For all z : |z − z0 | > R series (2.2.1) diverges.

- For all z : |z − z0 | < R series (2.2.1) converges.

Here R is the largest number such that |z − z0 | < R =⇒ convergence.

Also R is the smallest number such that |z − z0 | > R =⇒ divergence.

Definition 2.2.2. (a) In Case 1 (series converges for z = z0 only) the

radius of convergence is R = 0 .

(b) In Case 2 (series converges for all z ∈ C only) the radius of convergence
is R = ∞ and the disk of convergence is C.

(c) In Case 3 the radius of convergence is R (0 < R < ∞) and the disk of
convergence is {z : |z − z0 | < R}.

Further, in the disk {z : |z − z0 | ≤ r} converges uniformly and absolutely

for any r < R.

Theorem 2.2.2. Consider a power series (2.2.1) with the radius of conver-
gence R.
Chapter 2. Basic Properties of Analytic Functions 57

(i) If the limit on the right side exists, then

1 |an+1 |
= lim . (2.2.2)
R n→∞ |an |

(ii) If the limit on the right side exists, then

1 p
= lim n |an |. (2.2.3)
R n→∞

Proof. Proof follows from the ratio test and root test for numerical series.
∞
X (z − 1)n
Example 2.2.2. (a) The radius of convergence of is R =
n=0
n!
∞ and it converges in C.
∞
X n(z + 1)n
(b) The radius of convergence of is R = 5 and the disk
n=0
5n
of convergence is {z : |z + 1| < 5}.
∞
X
(c) The radius of convergence of n!!(z − i)n is R = 0 and it
n=0
converges only in {i}.

Question. What happens on the border of the disk of convergence {z : |z −

z0 | = R} (when 0 < R < ∞)?
Remember, in Calculus I, where power series of x ∈ R were considered,
there was an interval of convergence (x0 − R, x0 + R) and the boundary
consisted of two extreme points, x0 − R, and x0 + R. Then you investigated
these two extreme points.
Now question is way more difficult, because the boundary of the disk
{z : |z − z0 | < R} is a circle {z : |z − z0 | = R}, not just two points.
Example 2.2.3. Consider the following series. One can check easily, that all of
them have the radius of convergence, equal 1. So, consider C = {z : |z| = 1}.
∞
X
(a) z n diverges for all z : |z| = 1 (common term does not tend to 0).
n=1
Chapter 2. Basic Properties of Analytic Functions 58

∞
X zn P∞ 1
(b) converges for all z : |z| = 1 ( n=1 n2 < ∞).
n=1
n2

The next examples are much more difficult:

∞
X zn
Example 2.2.4. (a) converges for all z ∈ C, z ̸= 1 and diverges for
n=1
n
z = 1. One can prove easily the convergence as z = −1.
∞
X z pn
(b) Then with p = 2, 3, . . . diverges for zm = e2πmi/p with m =
n=1
n
0, 1, . . . , p (which are roots of degree p from 1) and converges for all
other z ∈ C.

(c) Further, if K ⊂ C is a union of enumerable family of closed subsets

of C, then there exists a power series, which converges at z ∈ K and
diverges in z ∈ C \ K.
So, the correct and complete answer to the Question (in the Test 1) Find
∞
X zn
where series converges should indicate that
n=1
n

(a) it converges as |z| < 1,

(b) diverges as |z| > 1,

(c) converges as z = −1, diverges as z = 1, and ¡4-¿there are other points

in z : |z| = 1 but you don’t know about convergence in these points.

The last part of this answer is important, because it shows, that you
understand, that the boundary of the disk is a circle, not just two points.

2.2.3 Analyticity of the Sum a Power Series

Theorem 2.2.3. Assume that a power series
∞
X
f (z) = an (z − z0 )n (2.2.4)
n=0

converges in the disk D = {z : |z − z0 | < R}, R > 0. Then

Chapter 2. Basic Properties of Analytic Functions 59

(i) f (z) is an analytic function in this disk D;

(ii) f (z) is differentiable in this disk D and

∞
X
f ′ (z) = nan (z − z0 )n−1 . (2.2.5)
n=1

this series also converges in D.

Proof. Without any loss of the generality we can assume that z0 = 0.

Consider z : |z| < r = R − 2δ and h : |h| < δ. Then for |z| ≤ r = R − 2δ
n
!
(z + h)n − z n n−1

X n n−j j−1
| − nz | =| z h |
h j=2 j
n
!
X n
≤|h|δ −2 (R − 2δ)n−j δ j
j=2 j
n
!
X n
<|h|δ −2 (R − 2δ)n−j δ j
j=0 j
=|h|δ −2 (R − 2δ) + δ)n = |h|δ −2 (R − δ)n .

Then
∞
f (z + h) − f (z) −2
X
| − g(z)| ≤ |h|δ |a|n (R − δ)n
h n=2

where g(z) is the right-hand expression of (2.2.5). Since |an | ≤ M (R−δ/2)−n

for some M = M (δ), the right-hand expression does not exceed M1 |h| → 0
as |h| → 0.
We claim that series g(z) converges. Indeed,
∞
X ∞
X
n−1 −1
n|an ||z − z0 | ≤ Mδ nρn < ∞
n=1 n=1

with ρ = (R − 2δ)/(R − δ).

Then f (z) is an analytic function and f ′ (z) = g(z).
Chapter 2. Basic Properties of Analytic Functions 60

Corollary 2.2.4. Assume that a power series

∞
X
f (z) = an (z − z0 )n (2.2.4)
n=0

converges in the disk D = {z : |z − z0 | < R}, R > 0. Then

(i) f (z) is infinitely differentiable in D and
∞
(m)
X n!
f (z) = an (z − z0 )n−m . (2.2.6)
n=m
(n − m)!

(ii) In particular,

f (m) (z0 ) = m!am . (2.2.7)

2.2.4 Primitive of the sum of a power series

Corollary 2.2.5. Assume that a power series (2.2.4) converges in the disk
D = {z : |z − z0 | < R}, R > 0. Then
∞
X an
F (z) = (z − z0 )n+1 (2.2.8)
n=0
n+1

converges in D and is a primitive function of f (z): F ′ (z) = f (z) and

F (z0 ) = 0.
∞
1 X
Example 2.2.5. (a) Integrating = z n , converging as |z| < 1,
1−z n=0
we get
∞
X zn
Log(1 − z) =
n=1
n

∞
1 X
(b) Integrating 2
= (−1)n z 2n , converging as |z| < 1, we get
1+z n=0

∞
X (−1)n z 2n+1
Arctan(z) =
n=0
2n + 1
Chapter 2. Basic Properties of Analytic Functions 61

(c) Consider (1 − z)−α ; it’s Taylor decomposition is

α(α + 1) 2
(1 − z)−α = 1 + αz + z + ...
2
α(α + 1) · · · (α + n − 1) n
+ z + . . . (2.2.9)
n!
Unless α is a non-positive integer we get an infinite series. In particular,
1
plugging α = 12 and z := z 2 we get a decomposition for √ .
1 − z2
Integrating, we get a decomposition for Arcsin(z).
Remark 2.2.1. One can prove that
(a) In Series (2.2.9) coefficient an has a magnitude nα−1 and therefore it
converges absolutely on C = {z : |z| = 1} for α < 0; diverges on C for
α ≥ 1 and converges on C except z = 1 for 0 < α < 1;
(b) Series for Log(1 − z) converges on C except z = 1;
(c) Series for Arctan(z) converges on C except z = ±i;
(d) Series for Arcsin(z) converges on C but not absolutely.
Results “converges but not absolutely” are more difficult and beyond our
reach.

2.2.5 Multiplication and division of powers series

Theorem 2.2.6. Let
∞
X ∞
X
f (z) = an (z − z0 )n and g(z) = bn (z − z0 )n
n=0 n=0
converge in the disk D = {z : |z − z0 | < R}. Then
X∞
(f g)(z) = cn (z − z0 )n (2.2.10)
n=0
with
n
X
cn = am bn−m (2.2.11)
m=0

also converges in D.
Chapter 2. Basic Properties of Analytic Functions 62

2.3 Cauchy’s Theorem and Cauchy’s

Formula
2.3.1 Cauchy’s theorem
Together with Definition of analyticity Cauchy’s Theorem is the linchpin of
complex variables.
We shall first prove Cauchy’s Theorem with the added assumption that
′
f is continuous. As it turns out, an analytic function automatically has a
continuous derivative. However, it is somewhat technical to prove this.
For that reason, most of this work is isolated as a special starred Subsec-
tion 2.3.1∗ of the Textbook; it is optional and we do not cover this in our
Notes.
The final step is Theorem 1 of Section 4 (also not covered). Throughout
this section, we shall assume that f ′ is continuous.
Theorem 2.3.1 (Cauchy’s Theorem). Suppose that f is analytic on a
domain D. Let γ be a piecewise simple smooth counter-clockwise oriented
closed curve in D whose inside Q also lies in D. Then
I
f (z) dz = 0. (2.3.1)
γ

Proof. In contrast to Textbook we already made all the necessary calculation

in Subsection 1.6.5 “Green’s Formula: Complex Variables”.
Namely, we proved that
I ZZ
∂f
f (z) dz = 2i dxdy (1.6.7)
γ Q ∂z

∂f
and since = 0 is our definition of the analyticity, this is 0.
∂z
Remark 2.3.1. We also proved in Subsection 1.6.4 “Green’s Formula: Dis-
cussion” that
- formula (1.6.7) holds in the case when γ consists of several pieces,
provided they are properly oriented
- and if the integrand in the double integral is 0 then the orientation
does not matter,
Chapter 2. Basic Properties of Analytic Functions 63

- and if domain D is simply-connected (that means, whenever γ is a

simple closed curve in D, the inside of γ is also a subset of D then
this integral is 0 for any closed piecewise smooth curve γ.

(we did it for Green’s formula in real variables, but Green’s formula in
Complex variables simply inherits these properties).
Still, read Theorem 2 of Section 2.3 in the Textbook (pages 108–109).

2.3.2 Curve deformation

Corollary 2.3.2. Let f be analytic function in a simply-connected domain
D, and let γ and γ ′ be two curves in D with the same start and end points.
Then
Z Z
f (z) dz = f (z) dz. (2.3.2)
γ γ′

Proof. Indeed, γ − γ ′ is a closed curve in a simply-connected domain and

therefore
Z Z Z
f (z) dz − f (z) dz = f (z) dz = 0
γ γ′ γ−γ ′

Question. What to do if D is not simply-connected?

Definition 2.3.1. (a) Curve γ ′ is deformation of curve γ if there exists

a continuous family of curves γs (s ∈ [0, 1]), contained in D, with the
same start and end points such that γ0 = γ and γ1 = γ ′ .

(b) Let γ and γ ′ be two closed curves in D. Curve γ ′ is deformation of

curve γ if there exists a continuous family of closed curves γs (s ∈ [0, 1]),
contained in D, such that γ0 = γ and γ1 = γ ′ .

Corollary 2.3.3. Let f be analytic function in a domain D, and let γ and

γ ′ be either two curves in D with the same start and end points or two closed
curves. Let γ ′ is a deformation of γ. Then (2.3.2) holds.

This corollary allows us to substitute a curve by its deformation.

Chapter 2. Basic Properties of Analytic Functions 64

Figure 2.1: Green lines are deformations of blue lines, but red lines are not.

2.3.3 A primitive of the analytic function

Theorem 2.3.4. Let f be an analytic function in simply-connected domain
D. Then there exists a function F , analytic in D, such that F ′ (z) = f (z).
Proof. We define F (z) as
Z z
F (z) = f (z) dz (2.3.3)
z0

where the right-hand expression means integral over any curve γ, with a
start-point z0 (some fixed point in D) and an end-point z.
This is justified since the right-hand expression does not depend on the
choice of γ with a start-point z0 (a fixed point in D) and an end-point z.
Indeed, if γ1 and γ2 are tw such points, then
Z Z Z
f (z) dz − f (z) dz = f (z) dz = 0
γ1 γ2 γ1 −γ2

because γ1 − γ2 is a closed curve.

Then
Z z+dz Z z Z z+dz
dF (z) = f (z) dz − f (z) dz = f (z) dz = f (z) dz
z0 z0 z

and therefore F ′ (z) = f (z).

(see also Theorem 3, Section 2.3 of the Textbook, page 109).
Like in Calculus I,
Chapter 2. Basic Properties of Analytic Functions 65

Definition 2.3.2. Function F such that F ′ = f is called a primitive of f .

What happens, if domain is not simply-connected? Then the primitive
may be a multiple-valued function.
Example 2.3.1. f (z) = z1 is analytic in C \ {0} which is not simply-connected.
It’s primitive F (z) = log(z) is multiple-valued.
Sure, if we make a cut, making domain simply-connected (f.e. (−∞, 0],
then we can select a single-valued branch of log(z).
Theorem 2.3.5. Let f be an analytic function in domain D. Let F be a
primitive of f (if D is not simply-connected), then F may not exist for f ,
but we assume that it exists). Then
(i) F is defined up to an additive constant.
(ii)
Z
f (z) dz = F (z1 ) − F (z0 ), (2.3.4)
γ

where γ is any curve with a start-point z0 and an end-point z1 , so the

result is the same for all such curves.
(iii) In particular, this integral is 0 for all closed curves.

Proof. (i) If F ′ = f , then (F + C)′ = F ′ + C ′ = f . on the other hand, if

F1 and F2 have the same derivative f , then (F1 − F2 )′ = F1′ − F2′ =
0 =⇒ F1 − F2 = const.
(ii) If f = F ′ then f dz = dF and
Z Z
f dz = dF = F (z1 ) − F (z0 )
γ γ

if γ is any curve with a start-point z0 and an end-point z1 .

(iii) If γ is closed, then z1 = z0 and the result is 0.

1
Example 2.3.2. (a) If f (z) = and γ is a circular curve, centered at 0
z
with radius r, with counter-clockwise orientation, then
Z Z Z 2π
dz
f (z) dz = = idθ = 2πi ̸= 0
γ γ z 0
Chapter 2. Basic Properties of Analytic Functions 66

and therefore a single-valued primitive does not exist if we do not

make a cut, preventing from going around 0.
1
(b) On the other hand, if f (z) = n , n = 2, 3, . . ., and γ is a circular curve,
z
centered at 0 with radius r, with counter-clockwise orientation, then
Z Z Z 2π
dz
f (z) dz = n
= ie−i(n−1)θ dθ = 0.
γ γ z 0

2.4 Consequences of Cauchy’s Formula

2.4.1 Analyticity and Power Series
Until the end of Chapter 2 we are going to consider different applications of
Cauchy’s formula. The first major application is that analytic function can
be decomposed into converging power series:
Theorem 2.4.1. Suppose that f (z) is analytic on a domain D and z0 ∈ D.
Assume that {z : |z − z0 | < R} ⊂ D.
Then
(i) f has a power series decomposition
∞
X
f (z) = an (z − z0 )n , (2.4.1)
n=0

converging in this disk.

(ii) Furthermore, the coefficients are given by
Z
1 f (ζ) dζ
an = , (2.4.2)
2πi γ (ζ − z0 )n+1

where γ is the properly oriented circle {z : |z − z0 | = r} with any

r : 0 < r < R.
Proof. Let us apply Cauchy’s formula
Z
1 f (ζ dζ
f (z) =
2πi γ ζ −z
Chapter 2. Basic Properties of Analytic Functions 67

with
1 1 1 1
= =
ζ −z (ζ − z0 ) − (z − z0 ) (ζ − z0 ) 1 − z − z0
h i
ζ − z0
and decompose
∞
X z − z0 n
1
z − z0 =
1− n=0
ζ − z0
ζ − z0
z − z0
which converges as |z − z0 | < r and |ζ − z0 | ≥ r and therefore < 1.
ζ − z0
∞ Z
X 1 f (ζ) dζ
f (z) = n+1
×(z − z0 )n .
n=0
2πi γ (ζ − z0 )
=an

This concludes the proof of (2.4.1)–(2.4.2).

For more justification details look at the end of the proof of Theorem 1
of Section 2.4 of the Textbook (pages 124–125).
Remark 2.4.1. We know already, that the sum of the converging power series
is an analytic function. Now we proved the converse statement.

Corollary 2.4.2. If f (z) is entire analytic function (that means, analytic

in C) then it can be decomposed into power series at any point z0 ∈ C and
the radius of convergence is R = ∞.

Example 2.4.1.
∞ ∞ ∞
z
X zn X z 2n X z 2n+1
e = , cosh(z) = , sinh(z) = ,
n=0
n! n=0
(2n)! n=0
(2n + 1)!
∞ n 2n ∞
X (−1) z X (−1)n z 2n+1
cos(z) = , sin(z) =
n=0
(2n)! n=0
(2n + 1)!

have R = ∞.
Chapter 2. Basic Properties of Analytic Functions 68

Example 2.4.2. (a)

∞ ∞
1 X X zn
= zn, − Log(1 − z) = ,
1−z n=0 n=1
n
∞
1 X (−1)m−1 (m + n − 1)!z n
=
(1 − z)m n=0
n!

with singularity at z = 1 have radius of convergence R = 1.

(b)
∞ ∞
1 X X (−1)n z 2n+1
= (−1)n z 2n , Arctan(z) =
1 + z2 n=0 n=0
2n + 1

with singularities at z = ±i, have radius of convergence R = 1.

Corollary 2.4.3. Let f be an analytic function in D. Then f (z) is infinitely

differentiable in D and, furthermore, its derivatives satisfy

|f (n) (z)| ≤ C(r)n!r−n ∀n = 0, 1, . . . (2.4.3)

for any r < dist(z, ∂D), which is the distance from z to ∂D–the boundary
of D.

Proof. Immediately from Theorem 2.3.1 and equality f (n) (z0 ) = n!an (where
in the last moment we plug z instead of z0 ). Indeed, since the radius of
convergence is at least R = dist(z, ∂D), then |an | ≤ C(r)r−n for r < R.
Remark 2.4.2. (a) If we consider functions of one real variable, then exis-
tence of n derivatives does not imply existence of (n + 1) derivatives.

(b) f (n) (x) does not satisfy (2.4.3) or any other inequality. In fact, for any
sequence bn there exists an infinitely smooth function f (x) such that
f (n) (x0 ) = bn .

(c) In this case series ∞ bn n

P
n=0 n! (x − x0 ) does not necessarily converges at
all, and even if it converges, the sum is not necessarily f (x).
Chapter 2. Basic Properties of Analytic Functions 69

Theorem 2.4.4 (Fatou’s theorem (optional)). Let f (z) be analytic in

domain D. Consider power series decomposition at point z0 ∈ D:
∞
X
f (z) = an (z − z0 )n . (2.4.1)
n=0

Assume that the radius of convergence is exactly R and consider circle

{z : |z − z0 | = R}. Let Γ be a closed arc of this circle, and also Γ ⊂ D.
Assume that

lim an Rn = 0. (2.4.4)
n→∞

Then (2.4.1) converges (uniformly) on Γ to f (z).

We do not give a proof (it is way beyond our reach).

∞
X zn
Example 2.4.3 (optional). (a) Consider − Log(1 − z) = . We know,
n=1
n
it converges for |z| < 1, and diverges for z = 1. What happens, when
|z| = 1, but z ̸= 1? Fatou’s theorem implies that it converges because
condition (2.4.4) is fulfilled.
∞
X (−1)n z 2n+1
(b) Consider Arctan(z) = . We know, it converges for
n=0
2n + 1
|z| < 1, and diverges for z = ±i. What happens, when |z| = 1, but
z ̸= ±i? Fatou’s theorem implies that it converges because condition
(2.4.4) is fulfilled.
−1
P∞ n
(c) On
P∞ the other hand, for (1 − z) = n=1 z and (1 + z 2 )−1 =
n 2n+1
n=0 (−1) z condition (2.4.4) fails and these series diverge at
any point of {z : |z| = 1}.
Example 2.4.4 (optional). Consider
∞
X
−α
(1 − z) = an z n , (2.4.5)
n=0
n
α(α + 1) · · · (α + n − 1) Y α − 1

an = = 1+ . (2.4.6)
n! k=0
k
Chapter 2. Basic Properties of Analytic Functions 70

Assuming tht α ̸= 0, −1, −2, . . . (in which case we get a polynomial

instead of series) conside
n
X h  α − 1
i
| ln(an )| = Re log 1 +
k=0
k
n
X Re(α) − 1
∼ ∼ (Re(α) − 1) ln(n)
k=0
k

(these transitions could be proven easily) and therefore |an | ∼ nRe(α)−1 .

So, |an | ∼ nRe(α)−1 . Then

(a) series (2.4.5)–(2.4.6) is absolutely converging as |z| = 1 if and only if

Re(α) < 0.

(b) for Re(α) ≥ 1 it does not converge at any point z : |z| = 1 (common
term does not tend to 0)

(c) for 0 ≤ Re(α) < 1 by Fatou’s theorem it converges at all points

z : |z| = 1, except z = 1.

Example 2.4.5. Plugging into previous example z := z 2 and α = − 12 we get

1
that |an | ∼ n− 2 ; however, after integration we get
∞
3
X
Arcsin(z) = bn z 2n+1 |bn | ∼ n− 2
n=0

and this series is converging absolutely as |z| = 1.

2.4.2 Continuation of Analytic Functions

Theorem 2.4.5. Suppose that f (z) is analytic on a domain D and, further,
at some point z0 ∈ D

f (n) (z0 ) = 0, ∀n = 0, 1, 2, . . . (2.4.7)

Then f (z) = 0 in D.
Chapter 2. Basic Properties of Analytic Functions 71

z0

z1

Proof. Consider some other point z1 ∈ D and connect z0 and z1 by some

curve (remember, domain is connected!). Let ε > 0 be the minimal distance
between this curve and the boundary of D. Consider disk centered at z0 of
radius δ, then f = 0 in it.
Consider another disk of radius δ, centered inside of the first one, down
this curve. Then f = 0 inside of it due to the same reason, and another,
and another . . . Eventually we will be able to reach point z1 and then f = 0
in its neighbourhood.
Corollary 2.4.6. Let f and g be two analytic functions in domain D,
coinciding in non-empty subdomain D′ . Then f and g coincide in D.
Proof. Apply Theorem 2.4.5 to their difference.
Remark 2.4.3. Again, difference from functions of real variable:
( 1
e− x2 x > 0,
f (x) =
0 x ≤ 0,
has the property: f (n) (0) = 0 for all n, but it is not identically 0.
This method allows sometimes to continue f as analytic function to
a larger domain. However for every domain there is an analytic function,
which cannot be continued to any larger domain.
Example 2.4.6. Function
∞
n
X
f (z) = z2
n=0

is analytic in the unit disk, but cannot be continued analytically to any

larger domain.
πki
Indeed, as z = e 2m for any m = 1, 2, . . ., k = 0, . . . , 2m+1 − 1 this series
is diverging to ∞ since all terms from n = m + 1 are just 1. And these
points are dense on the circle {z : |z| = 1}.
Chapter 2. Basic Properties of Analytic Functions 72

√
Example 2.4.7. Consider f (z) = z as Re(z) > 0.

Using the same method we can expand it to the red disk. Then to blue,
and to brown, and to green
√ but where brown and green intersect, we got
two different branches of z!

2.4.3 The Order of Zero

Suppose that f (z) is analytic and not identically 0 in domain D, and
f (z0 ) = 0 for some z0 ∈ D. Then the corresponding power series has the
first term missing:

f (z) = a1 (z − z0 ) + a2 (z − z0 )2 + . . . , because a0 = f (z0 ) = 0.

We know that not all the coefficients here are 0 (otherwise f (z) would
be identically 0), so there is an integer m, such that

a0 = a1 = . . . = am−1 = 0, am ̸= 0
m m+1
⇐⇒ f (z) = am (z − z0 ) + am+1 (z − z0 ) + ..., am ̸= 0
(k) (m)
⇐⇒ f (z0 ) = 0 k = 0, . . . , m − 1, but f (z0 ) ̸= 0.

Definition 2.4.1. (a) In this case we say that z0 is a zero of f (z) and m
is an order of zero.

(b) Simple zeroes are zeroes of order 1, double zeroes are zeroes of order 2,
triple zeroes are zeroes of order 3, and so on.

(c) For convenience we say that z0 is zero of order 0 if it is not a zero at

all: f (z0 ) ̸= 0.
Chapter 2. Basic Properties of Analytic Functions 73

Let z0 be a zero of order m. Then

 f (z)

z ∈ D \ {z0 },
g(z) = (z − z0 )m

am z = z0
is an analytic function in D and g(z0 ) = am ̸= 0. Indeed,
g(z) = am + am+1 (z − z0 ) + am+2 (z − z0 )2 + . . .
and one can prove easily that this series is converging in the same disk as
power series for f (z).
Theorem 2.4.7. f (z) has a zero of order m at z0 ∈ D if and only if
f (z) = (z − z0 )m g(z) where g(z) is analytic in D and g(z0 ) ̸= 0.
Example 2.4.8. (a) sin(z) and tan(z) have simple zeroes in πn, n ∈ Z,
cos(z) and cot(z) have simple zeroes in π(n + 12 ), n ∈ Z,
(b) cos(z) − 1 has a double zero at πn, n ∈ Z,
(c) z sin(z) has a double zero at 0 and simple zeroes at πn, n ∈ Z \ {0},

 sin(z) z ̸= 0,
(d) f (z) = z has simple zeroes in πn, n ∈ Z \ {0}.
f (0) = 1


2.4.4 Morera’s Theorem (optional)

Cauchy’s Theorem, on which so much of the development of complex vari-
ables depends, has a converse. This theorem, given precisely below, states
that if the integral of a continuous function f over every triangle in some
domain is zero, then f must be analytic in that domain.
Theorem 2.4.8 (Morera’s Theorem). If f is a continuous function on a
domain D and if
Z
f (z) dz = 0
γ

for every triangle γ that lies, together with its interior, in D, then f is
analytic on in D.
We skip the proof, see proof of Theorem 2 on page 129 of the Textbook.
Chapter 2. Basic Properties of Analytic Functions 74

2.4.5 Liouville’s Theorem

Theorem 2.4.9. Let F (z) be an entire analytic function and let F be
bounded, which means that there is a constant M such that |F (z)| ≤ M for
all z ∈ C. Then F (z) is identically constant.
F (z) − F (0)
Proof. Let g(z) = (properly defined at z = 0); then g(z) is an
z
entire analytic function as well. Then for z : |z| = R > 0 we estimate

|F (z)| + |F (0)| 2M
|g(z)| ≤ ≤ .
R R
For any ζ ∈ C let us take R > 2|ζ|. By Cauchy’s formula
Z
1 g(z) dz
g(ζ) = .
2πi |z|=R z − ζ

Then
1 2M 2πR
|g(ζ)| ≤ →0 as R → ∞.
2π R R − |ζ|

Then g(ζ) = 0 (identically) and F (z) = F (0).

Theorem 2.4.10 (optional). Let F (z) be an entire analytic function and
let F satisfy

|F (z)| ≤ M (|z| + 1)m ∀z ∈ C.

Then F (z) is a polynomial of degree not exceeding m.

Proof. Similarly, consider
m
F (z) − P (z) X 1 (k)
g(z) = , P (z) = ak z k , ak = F (0).
z m+1 k=0
k!

Then g(z) is an entire analytic function, and it satisfies

|F (z)| + |P (z)| 2M (|z| + 1)m

|g(z)| ≤ ≤ ≤ 4M R−1
|z|m+1 |z|m+1

for |z| + 1 ≤ R and the rest of the proof remains unchanged.

Chapter 2. Basic Properties of Analytic Functions 75

2.4.6 Analytic Logarithms

Suppose f is an analytic function in a simply-connected domain D. Assume
that f has no zeroes in D.
Fix z0 ∈ D and define
Z z ′
f (w)
h(z) = dw
z0 f (w)

with integral taken along any piecewise smooth curve from z0 to z; since
f ′ (w)
f (w)
is an analytic in D (think, why) this integral does not depend on the
curve.
′
Then h′ = ff and

(e−h(z) f (z))′ = − h′ e−h(z) f (z) + e−h(z) f (z)


=e−h(z) −f ′ (z) + f ′ (z) = 0.

Then e−h(z) f (z) = c with non-zero constant c and finally f (z) = ceh(z) .
Since h(z0 ) = 0 we have c = Log(f (z0 )) and g(z) = h(z) − Log(f (z0 ))
we proved

Theorem 2.4.11. Suppose f is an analytic function in a simply-connected

domain D. Assume that f has no zeroes in D. Then there is an analytic
function g(z) in D such that

eg(z) = f (z), z ∈ D.

2.4.7 Multiplication of Power Series (Optional)

We know that the product of two convergent power series is again a power
series, and the coefficients of the product series may be obtained from those
of the original series by a rather simple rule.
Theorem 2.4.7 of this section allows us to give a simple proof of that
result. Let
∞
X ∞
X
n
f (z) = an z and g(z) = bn z n
n=0 n=0

be the two series, each with radius of convergence at least R, R > 0.

Chapter 2. Basic Properties of Analytic Functions 76

Then the product h(z) = f (z)g(z) is analytic in the disk {z : |z| < R}.
Hence, by Theorem 2.4.1 , h has a power series expansion in this same disk
∞
X h(n) (0)
h(z) = cn z n with cn = .
n=0
n!

However, the product rule for differentiation gives

n
(n)
X n!
h (0) = f (k) (0)g (n−k) (0)
k=0
(n − k)!k!

But
f (k) (0) g (n−k) (0)
= ak , = bn−k
k! k!
and therefore
n
X
ck = ak bn−k .
k=0

2.5 Isolated Singularities

2.5.1 Classification
In this section we study isolated singularities which will play a crucial role
in the rest of the class.

Definition 2.5.1. An analytic function f (z) has an isolated singularity at a

point z0 if f is defined and analytic in the punctured disk {z : 0 < |z−z0 | < r},
for some r > 0.

Three examples of isolated singularities are:

z3 sin(z)
(a) 0 is an isolated singularity for f1 (z) = and f2 (z) = ;
z z
2
(b) 0 is an isolated singularity for g1 (z) = and g2 (z) = cot(z);
z3
1

(c) 0 is an isolated singularity for h(z) = exp .
z
Chapter 2. Basic Properties of Analytic Functions 77

As these examples show, there are three possible modes of behaviour for
f (z) when 0 < |z − z0 | < r:
(a) |f (z)| remains bounded as z → z0 ; then we call z0 a removable singu-
larity.
(b) lim |f (z)| = ∞; then we call z0 a pole.
z→z0

(c) Neither of these two cases. Then we call z0 an essential singularity.

As we are going to show,
(a) a removable singularity is not really a singularity at all: we can remove
it, (re)defining f (z0 ) and then f (z) will be analytic in the whole disk
{z : |z − z0 | < r};
(b) a pole is in some sense a “regular singularity” (not a standard termi-
nology);
(c) an essential singularity is in some sense a “wild singularity” (also not
a standard terminology).
Example 2.5.1. The following singularities are not isolated singularities:
1
1
(a) z0 = 0 and f (z) = cot because f (z) is singular at points zn =
z πn
with n ∈ Z \ {0}, so there is a sequence of singular points zn → 0 as
n → ∞;

√ √
(b) z0 = 0 and f (z) = z, or 3 z, or z α with α ∈ / Z because it cannot
be defined as a single-valued analytic function near 0. It is called a
branching point.
Chapter 2. Basic Properties of Analytic Functions 78

(c) z0 = 0 and f (z) = log(z) (also a branching point).

We do not cover these points in this section.

2.5.2 Removable singularities

Suppose that |f (z)| ≤ M in {z : 0 < |z − z0 | < r}. Let
(
f (z)(z − z0 )2 0 < |z − z0 | < r,
g(z) =
0 z = z0 .

This function is analytic in {z : 0 < |z − z0 | < r} and also is differentiable

at z0 because
g(z) − g(z0 )
lim = lim(z − z0 )f (z) = 0.
z→z0 z − z0
Then

g(z) = b0 + b1 (z − z0 ) + b2 (z − z0 )2 + b3 (z − z0 )3 + . . .

with b0 = g(z0 ) = 0 and b1 = g ′ (z0 ) = 0. Then

g(z) = b2 (z − z0 )2 + b3 (z − z0 )3 + . . .
=⇒ f (z) = b2 + b3 (z − z0 ) + . . . (2.5.1)

in {z : 0 < |z − z0 | < r}.

Finally, let us set f (z0 ) = b2 and then (2.5.1) holds in the whole disk
{z : |z − z0 | < r}. Thus f (z) is analytic there. We removed a singularity at
z0 and we call z0 a removable singularity.
z3
Example 2.5.2. (a) Consider f (z) := = z 2 and we set f (0) = 0.
z
(b) Consider
P∞ (−1)n z 2n+1 ∞
sin(z) n=0 (2n+1)!
X (−1)n z 2n
f (z) := = =
z z n=0
(2n + 1)!

and we set f (0) = 1.

Chapter 2. Basic Properties of Analytic Functions 79

Remark 2.5.1. Compare with Calculus I: if f (x) is bounded it does not mean
that it has a limit, and even if it has a limit, it’s derivative could be really
bad:
1

Example 2.5.3. 1. f (x) = sin is bounded but has no limit as x → 0.
x
1

However in complex variables f (z) = sin has an essential singularity
x
at 0.

2. f (x) = x sin( x1 ) is continuous, but f ′ (x) = sin( x1 ) − x1 cos( x1 ) is un-

1

bounded. Again, in complex variables f (z) = z sin has an essential
x
singularity at 0.

2.5.3 Poles
Pole are genuine singularities, but we can “tame” them. Assume that

lim |f (z)| = ∞. (2.5.2)

z→z0

Without any loss of the generality we can assume that

|f (z)| ≥ 1 for z : |z − z0 | < r. (2.5.3)

1
Indeed, if needed we can reduce r. Then g(z) = is analytic and
f (z)
bounded in the punctured disk {z : 0 < |z − z0 | < r} and therefore for g(z)
point z0 is a removable singularity.
So we can define g(z0 ) = limz→z0 g(z) = 0 (because of (2.5.2)) and g(z)
becomes analytic in the whole disk {z : |z − z0 | < r}.
However, since g(z0 ) = 0 point z0 is also a zero of g(z). Let m = 1, 2, . . .
be an order of zero. Then

g(z) = (z − z0 )m h(z)

where h(z) is analytic in the disk {z : |z − z0 | < r} and h(z0 ) ̸= 0. Since

g(z) does not vanish in the punctured disk {z : 0 < |z − z0 | < r}, neither
does h(z), so it does not vanish in the whole disk {z : |z − z0 | < r}.
Chapter 2. Basic Properties of Analytic Functions 80

1
Therefore H(z) = is analytic in {z : |z − z0 | < r} and H(z0 ) ̸= 0.
h(z)
Then
1 1 H(z)
f (z) = = m
= .
g(z) (z − z0 ) h(z) (z − z0 )m
Since
H(z) = b0 + b1 (z − z0 ) + b2 (z − z0 )2 + . . . , b0 ̸= 0
we conclude that
f (z) = a−m (z − z0 )−m + a−m+1 (z − z0 )−m+1 + a−m+2 (z − z0 )−m+2 + . . .
(2.5.4)
with an = bn+m and a−m ̸= 0.
Remark 2.5.2. Decomposition (2.5.4) is an example of what we call later
Laurent’s series.
1
We see that f (z) has a pole at z0 iff and only if g(z) = (with
f (z)
g(z0 ) = 0) has a zero at z0 .
Definition 2.5.2. (a) We say, that f (z) has a pole of order (multiplicity)
1
m at z0 if g(z) = has a zero of order (multiplicity) m at z0 .
f (z)
(b) Poles of multiplicity 1 are called simple poles, poles of multiplicity 2
are called double poles, poles of multiplicity 3 are called triple poles,
and so on.
We the got the following
Theorem 2.5.1. f (z) has a pole of order m at z0 if and only if decomposition
(2.5.4) holds with a−m ̸= 0 holds in {z : 0 < |z − z0 < r}.
Example 2.5.4. (a) cot(z) has simple poles at zn = πn.
(b) cot2 (z) has double poles at zn = πn.
(c) z cot2 (z) has double poles at zn = πn, n =
̸ 0 and a simple pole at
z0 = 0.
(d) z −1 cot(z) has simple poles at zn = πn, n ̸= 0 and a double pole at
z0 = 0.
Chapter 2. Basic Properties of Analytic Functions 81

2.5.4 Essential singularities

Recall that isolated singularities which are neither removable, nor poles are
called essential singularities. To show how wild these singularities are we
give without proof the following

Theorem 2.5.2 (Great Picard’s Theorem). If an analytic function f (z) has

an essential singularity at a point z0 , then on any punctured neighbourhood
of z0 , f (z) takes all possible complex values, with at most a single exception,
infinitely often.

Instead we prove much weaker

Theorem 2.5.3 (Little Picard’s Theorem). If an analytic function f (z) has

an essential singularity at a point z0 , then on any punctured neighbourhood
of z0 , f (z) approaches all possible complex values.

Proof. Assume that it does not approach w. Then in {z : |z − z0 | < r} we

1
have |f (z) − w| ≥ ε for some ε > 0. Then in this disk g(z) =
f (z) − w
m
is an analytic function, and g(z) = (z − z0 ) h(z) with m = 0, 1, 2, . . . and
analytic h(z), h(z0 ) ̸= 0.
Next,
1
f (z) − w = = (z − z0 )−m H(z)
g(z)
1
with analytic H(z), H(z0 ) = ̸= 0. As m = 0 we conclude that f (z)
h(z0 )
has a removable singularity at z0 , and as m = 1, 2, . . . we have a pole.
Contradiction.
1

Example 2.5.5. (a) exp has an essential singularity at 0. It takes any
z
1
value w ̸= 0 at points , n ∈ Z.
Log(w) + 2πin
1

(b) cos has an essential singularity at 0. It takes any value w at points
z
1
, n ∈ Z.
± Arccos(w) + 2πin
Chapter 2. Basic Properties of Analytic Functions 82

2.5.5 Laurent’s Series

We start from Laurent’s series, and then cover residues.

Theorem 2.5.4. Assume that f (z) is an analytic function in the annulus

(ring) D = {z : r < |z − z0 | < R}, with r < R. Then the following
decomposition holds in D:
∞
X
f (z) = an (z − z0 )n (2.5.5)
n=−∞
with Z
1 f (z) dz
an = (2.5.6)
2πi γ (z − z0 )n+1

where γ is any closed curve in D, winging exactly once in the counter-

clockwise direction around z0 .

z0

Remark 2.5.3. Series

∞
X
an (z − z0 )n
n=−∞
converges in D if and only if both series
X∞ ∞
X
an (z − z0 )n and a−n (z − z0 )−n
n=0 n=1

converge in D; these series converge in {z : |z−z0 | < R} and {z : |z−z0 | > r}

correspondingly.

Definition 2.5.3. Series on the right of (2.5.5) is called Laurent’s series.

Chapter 2. Basic Properties of Analytic Functions 83

Proof of Theorem 2.5.4. Let us consider z ∈ D and a smaller annulus D′ =

{z : r′ < |z − z0 | < R′ } with r < r′ < R′ < R, containing z, with the border
γ1 + γ2 , where γ2 = {z : |z − z0 | = R′ }, counter-clockwise oriented and
γ1 = {z : |z − z0 | = r′ }, clockwise oriented.

z
γ1 γ2
z0

Then according to Cauchy’s formula

Z Z Z
1 f (ζ) dζ 1 f (ζ) dζ 1 f (ζ) dζ
f (z) = = + .
2πi γ1 +γ2 ζ − z 2πi γ2 ζ − z 2πi γ1 ζ − z
Note that |z − z0 | < |ζ − z0 | for ζ ∈ γ2 and therefore there
∞
X (z − z0 )n
1 1
= =
ζ −z (ζ − z0 ) − (z − z0 ) n=0 (ζ − z0 )n+1

exactly like in the proof of Cauchy’s formula.

Meanwhile, |z − z0 | > |ζ − z0 | for ζ ∈ γ1 and therefore there
∞
1 1 X (ζ − z0 )n
=− =− n+1
.
ζ −z (z − z0 ) − (ζ − z0 ) n=0
(z − z0 )

Therefore, plugging into

Z Z
1 f (ζ) dζ 1 f (ζ) dζ
f (z) = +
2πi γ2 ζ −z 2πi γ1 ζ −z
we get
∞ Z
X 1 f (ζ) dζ
f (z) = n+1
× (z − z0 )n
n=0
2πi γ2
(ζ − z0 )
∞
X 1 Z
− f (ζ)(ζ − z0 )n dζ × (z − z0 )−n−1 .
n=0
2πi γ1
Chapter 2. Basic Properties of Analytic Functions 84

Observe that
∞ Z
X 1
- f (ζ)(ζ − z0 )n dζ × (z − z0 )−n−1
n=0
2πi γ1
−1 Z
X 1 f (ζ)
= m+1
dζ × (z − z0 )m
m=−∞
2πi -γ1 (ζ − z0 )

where we changed the sign in front and γ1 to −γ1 and plugged n = −m − 1.

Plugging m := n we get
∞ Z
X 1 f (ζ) dζ
f (z) = n+1
× (z − z0 )n
n=0
2πi γ2 (ζ − z0 )
−1 Z
X 1 f (ζ) dζ
+ n+1
× (z − z0 )n .
n=−∞
2πi −γ1 (ζ − z0 )

We can rewrite it as
∞
X
f (z) = an (z − z0 )n (2.5.5)
n=−∞
with Z
1 f (ζ) dζ
an = (2.5.6)
2πi γ (ζ − z0 )n+1

because due to Cauchy’s theorem integrals over γ, γ2 and −γ1 are equal.

−γ1 γ2
Chapter 2. Basic Properties of Analytic Functions 85

2.5.6 Isolated Singularities and Laurent’s Series

It is a special case of the previous: simply r = 0. So, we have
Corollary 2.5.5. Let z0 be an isolated singularity of an analytic function
f (z). Then f (z) could be decomposed into Laurent’s series converging in the
punctured disk {z : 0 < |z − z0 | < R} for some R > 0:
∞
X
f (z) = an (z − z0 )n . (2.5.5)
n=−∞

And the Laurent’s series characterize type of the isolated singularity:

Theorem 2.5.6. Let z0 be an isolated singularity; consider Laurent’s series
(2.5.5). Then
(i) z0 is a removable singularity if and only if an = 0 for all n < 0;
(ii) z0 is a pole of order m ≥ 1 if and only if an = 0 for all n < −m but
a−m ̸= 0;
(iii) z0 is an essential singularity if and only if there is an infinite number
of coefficients with n < 0 and an ̸= 0.
Proof. (i) We already have covered removable singularities. Recall that
f (z) has a removable singularity at z0 means that |f (z) ≤ M and then the
expression for an with n < 0 tends to 0 as r → 0.
(ii) We also have covered poles when f (z) → ∞ as z → z0 and proved
that (z − z0 )m f (z) has a removable singularity at z0 where m is an order
1
of zero of at z0 . Conversely, if Laurent’s series has a finite number of
f (z)
negative powers then we see that z0 is a pole.
(iii) So, what is left?––only essential singularities and Laurent’s series
with an infinite number of negative powers.

Example 2.5.6. z0 = 0 is an essential singularity of

∞ 0
1
X 1 1 X zn
(a) exp = = ;
z n=0
n! z n n=−∞ (−n)!
Chapter 2. Basic Properties of Analytic Functions 86

∞
1
X (−1)n
(b) cos = 2n
;
z n=0
(2n)!z
∞
1
X (−1)n
(c) sin = .
z n=0
(2n + 1)!z 2n+1

Example 2.5.7. (a) Let z0 be a pole of f (z). Then z0 will be a removable

singularity for (z − z0 )m f (z) where m is multiplicity of the pole.
(b) Let z0 be an essential singularity of f (z). Then z0 will be an essential
singularity for (z − z0 )m f (z).
Example 2.5.8. (a) z = 0 is an essential singularity for
1

f (z) = exp(z) + exp . It’s decomposition is
z
0 ∞
X zn X zn
+ .
n=−∞
(−n)! n=0 n!

1

= exp z + z1 .


(b) z = 0 is an essential singularity for f (z) = exp(z) exp
z
It’s decomposition is
∞ ∞ ∞
X zm X zn X
. = bp z p
m=0
m! n=0
n! p=−∞
with
X 1
bp = .
n≥0,m : n−m=p
n!m!

2.5.7 Singularities at ∞
Assume that ∞ ∈ D ⊂ C;b that means that ∞ has a neighbourhood contained
in D; in other words D ⊃ {z : |z| > r} for sufficiently large r. Like in the
case z0 ∈ C we have Definitions:
Definition 2.5.4. Let f (z) be an analytic function in {z : |z| > r}. Then
∞ is an isolated singularity of f (z).
Definition 2.5.5. (a) ∞ is a removable singularity if |f (z)| ≤ M for
z : |z| > r with sufficiently large M , r;
Chapter 2. Basic Properties of Analytic Functions 87

(b) ∞ is a pole if limz→∞ |f (z)| = ∞;

(c) ∞ is an essential singularity if neither of these two cases.

Example 2.5.9. The following singularities are not isolated singularities:

(a) z0 = ∞ and f (z) = cot(z) because cot(z) is singular at points zn = πn

with n ∈ Z \ {0}, so there is a sequence of singular points zn → z0 as
n → ∞;

√
(b) z0 = ∞ and f (z) = z because it cannot be defined as a single-valued
analytic function near ∞. It is called a branching point.

We do not cover these points in this section.

Since {z : |z| > r} is an annulus (with R = ∞) we can decompose f (z)
here into Laurent’s series:
∞
X
f (z) = an z n . (2.5.7)
n=−∞

On the other hand, if f (z) is analytic as |z| > r then g(w) := f w1 is

analytic in {w : 0 < |w| < r−1 } and type of singularity of g(w) at 0 coincides
with a type of singularity of f (z) at ∞.
Since
∞
X
g(w) = a−n wn
n=−∞

we conclude from Theorem 2.5.6:

Chapter 2. Basic Properties of Analytic Functions 88

Theorem 2.5.7. Let ∞ be an isolated singularity; consider Laurent’s series

(2.5.7). Then

(i) ∞ is a removable singularity if and only if an = 0 for all n > 0;

(ii) ∞ is a pole of order m ≥ 1 if and only if an = 0 for all n < m but

am ̸= 0;

(iii) ∞ is an essential singularity if and only if there is an infinite number

of coefficients with n > 0 and an ̸= 0.

Example 2.5.10. (a) Polynomial of degree m has a pole of multiplicity m

at ∞.

(b) ez , sin(z), cos(z) have infinite number of positive powers in their decom-
positions; and we again conclude that they have essential singularity
at infinity.

(c) Assume that all singularities of f (z) on C b are isolated and they
are poles. Then it can have only a finite number of singularities
z1 , . . . zN , which are poles of multiplicities m1 , . . . , mN . Then g(z) =
(z − z1 )m1 · · · (z − zn )mn f (z) is an entire analytic function and since
we assumed that ∞ is also a pole, we have |g(z) ≤ C(|z| + 1)M . Then
it is a polynomial P (z) and

P (z)
f (z) = Q(z) = (z − z1 )m1 · · · (z − zn )mn
Q(z)

is a rational function.
Conversely, any rational function has all singularities on C
b isolated,
and they are poles.

2.5.8 The Residue

Now we introduce a very important notion of the residue of analytic function
at the isolated singularity. In the next lecture we will prove the Residue
Theorem and apply to to many calculations.
Chapter 2. Basic Properties of Analytic Functions 89

Definition 2.5.6. (a) Let z0 ∈ C be an isolated singularity of an ana-

lytic function f (z). Decompose f (z) in its neighbourhood (that means, for
z : 0 < |z − z0 | < r with r > 0) into Laurent’s series
∞
X
f (z) = an (z − z0 )n (2.5.4)
n=−∞

Then coefficient a−1 is called the residue of f (z) at z0 and denoted as

Res(f ; z0 ):
Res(f ; z0 ) = a−1 . (2.5.8)

(b) Let ∞ be an isolated singularity of an analytic function f (z). De-

compose f (z) in its neighbourhood (that means, for z : |z − z0 | > r with
r > 0) into Laurent’s series
∞
X
f (z) = an z n (2.5.4)
n=−∞

Then coefficient a−1 with the opposite sign is called the residue of f (z) at
∞ and denoted as Res(f ; ∞):
Res(f ; ∞) = −a−1 . (2.5.9)
Remark 2.5.4. (a) Residue at non-isolated singularity is not defined.
(b) If z0 ̸= ∞ is a removable singularity of f (z), then Res(f, z0 ) = 0.
(c) On the contrary, even if ∞ is a removable singularity of f (z), it may
happen that Res(f, ∞) ̸= 0.
One can ask, why these definitions? Why we select coefficient a−1 among
others? And why we take an opposite sign at a−1 at ∞?
The answer to the first two questions follows from the equality:
Z
f (z) dz = 2πia−1 = 2πi Res(f ; z0 ) (2.5.10)
γ

where γ is a circle {z : |z| = r′ } with any sufficiently small r′ , counter-

clockwise oriented. Indeed, it follows from
(
2πi m = −1,
Z
(z − z0 )m dz = (2.5.11)
γ 0 m ̸= −1.
Chapter 2. Basic Properties of Analytic Functions 90

Exercise 2.5.1. Check it – it is easy!

The answer for the last question follows from the same first equality
(2.5.10) where γ is a circle {z : |z| = r′ } with any sufficiently large r′ .

γ
z0

(a) z0 ∈ C (b) z0 = ∞

However the counter-clockwise oriented circle is going around ∞ in the

wrong direction–a correct direction around ∞ is clockwise! Therefore, to have
the left-hand expression in (2.5.10) with properly (that means, clockwise)
oriented γ, we need to change the sign in the middle expression
Z
f (z) dz = −2πia−1 = 2πi Res(f ; ∞). (2.5.12)
γ

Therefore we arrive to the universal formula

Z
f (z) dz = 2πi Res(f ; z0 ). (2.5.13)
γ

for z0 ̸= ∞ and counter-clockwise oriented γ and for z0 = ∞ and clockwise

oriented γ.
Remark 2.5.5. Since γ in the definition of Res(f ; ∞) can be any closed curve
in the domain {z : |z| > R} with sufficiently large R, clock-wise oriented
and going exactly once around ∞, we conclude that
Res(f (z); ∞) = Res(f (z + z0 ); ∞) for any z0 ∈ C. (2.5.14)

2.5.9 Computation of Residues

Applications of the Residue Theorem (coming!) requires computation of
residues; so there are few recipes how to calculate residues. We consider
only z0 ∈ C. Recall that then the residue at removable singularity is 0.
Chapter 2. Basic Properties of Analytic Functions 91

Residue at Simple Pole. Let

g(z)
f (z) = , g(z0 ) ̸= 0, h(z0 ) = 0, h′ (z0 ) ̸= 0, (2.5.15)
h(z)
and g(z), h(z) analytic at z0 . Then
g(z0 )
Res(f ; z0 ) = ′ . (2.5.16)
h (z0 )

Indeed, f (z) has a simple pole at z0 and

g(z0 ) 1 + b1 (z − z0 ) + . . .
f (z) = ′
h (z0 )(z −
z0 ) 1 + c1 (z − z0 ) + . . .
g(z0 )  
= ′ 1 + d1 (z − z0 ) + . . .
h (z0 )(z − z0 )
g(z0 )
= ′ (z − z0 )−1 + a0 + a1 (z − z0 ) + . . .
h (z0 )

which implies (2.5.16).

To justify the first equality above observe that

1 + b1 (z − z0 ) + . . .
F (z) :=
1 + c1 (z − z0 ) + . . .

is analytic at z0 and F (z0 ) = 1. Here bk , ck , dk and ak denote some coefficients

we do not care about.
cos(z)
cos(z)
Example 2.5.11. (a) Res(cot(z); πn) = Res , πn = = 1.
sin(z) cos(z) z=πn
eiz
eiz e−1 i
(b) Res 2
; i = = =− .
z +1 (2z) z=i 2i 2e
cos(z)  cos(z) cosh(2) cosh(2)i
(c) Res 2 ; −2i = = = .
z +4 (2z) z=−2i −4i 4

Residue of (z − z0 )−m H(z). Let

H(z)
f (z) = (2.5.17)
(z − z0 )m
Chapter 2. Basic Properties of Analytic Functions 92

with H(z) analytic at z0 and m ≥ 1. Then

H (m−1) (z0 )
Res(f ; z0 ) = . (2.5.18)
(m − 1)!
Indeed,
∞
X
H(z) = bn (z − z0 )n ,
n=0
∞
X
f (z) = bn+m (z − z0 )n
n=−m

and the coefficient at (z − z0 )−1 is bm−1 , which implies (2.5.18).

 cos(z) π  − cos(z) 1
Example 2.5.12. (a) Res 3
; = z=π/4
=− √ .
(z − π/4) 4 2! 2 2
Here m = 3.
 sin(z) 
(b) Res ; i . Since m = 2, to apply (2.5.18) we need to have
(z 2 + 1)2
denominator (z − i)2 ; therefore we set H(z) = sin(z)(z + i)−2 . Then
H ′ (z) = cos(z)(z + i)−2 − 2 sin(z)(z + i)−3
cosh(1) i sinh(1) 1
=⇒ H ′ (i) = − −2 =−
4 −8i 4e
1
and the residue in question is − .
4e

Using Power Series. In more complicated cases of a multiple pole we

can use power series.
Hint. (A) It is convenient to move a singularity to 0: Res(f ; z0 ) = Res(g; 0)
with g(z) = f (z + z0 );
(B) If f (z) is an even function then Res(f ; 0) = 0 (also Res(f ; ∞) = 0).
Indeed, the Laurent’s decomposition would contain only even powers
of z.
Example 2.5.13. Calculate Res(tan2 (z); z = πn + π2 ), n ∈ Z. Using Hint (A)
π π
Res(tan2 (z); z = πn + ) = Res(tan2 (z + + πn); z = 0)
2 2
= Res(cot2 (z); z = 0).
Chapter 2. Basic Properties of Analytic Functions 93

Since cot(z) is odd, cot2 (z) is even and the required residue is 0 due to Hint
(B).
Example 2.5.14. Calculate Res(tan3 (z); z = πn + π2 ), n ∈ Z. Using Hint (A),
π π
Res(tan3 (z); z = πn + ) = Res(tan3 (z + + πn); z = 0)
2 2
= Res(cot3 (z); z = 0).
Then
z2

3 z2

3
3 cos3 (z) 1− 2
+ ... 1 1− 2
+ ...
cot (z) = 3 =
3 = 3
3
sin (z) z3 z 1− z2
z− 6
+ ... 6
+ ...
where . . . denote terms which do not contribute to term with z −1 in the
final decomposition (we need to get no more than z 2 in the numerator and
z 3 in denominator); therefore
3z 2 3z 2
cot3 (z) =z −3 1 − + . . . = z −3 − z −1 + . . .



+ ... 1 +
2 6
and the required residue is −1.

2.6 The Residue Theorem and its

Application to the Evaluation of
Definite Integrals
2.6.1 The Residue Theorem
Theorem 2.6.1 (Residue Theorem). Consider a bounded domain D and its
properly oriented boundary γ. Let f (z) be analytic in D with the exception of
a finite number of isolated singularities z1 , . . . , zN , which is also continuous
in D̄ = D ∪ γ.
Then
Z XN
f (z) dz = 2πi Res(f ; zk ). (2.6.1)
γ k=1

Remark 2.6.1. Remember, what is properly oriented? When going along γ,

domain is on your left.
Chapter 2. Basic Properties of Analytic Functions 94

Proof. Surround each point zk by a disk Dk (k = 1, . . . , N ) of a sufficiently

small radius, and remove these disks from D; the boundary of the domain
Q = D \ (D1 ∪ D2 ∪ . . . ∪ DN ) is γ − γ1 − . . . − γN .

z1

z3
z2

Since f is analytic in Q we can apply Cauchy’s Theorem:

Z
f (z) dz = 0
γ−γ1 −...−γN
Z N Z
X
=⇒ f (z) dz = f (z) dz
γ k=1 γk

due to the properties of line integral.

However, in virtue of the (2.5.1)
Z
f (z) dz = 2πi Res(f ; zk ), k = 1, . . . , N
γk

and we arrive to
Z N
X
f (z) dz = 2πi Res(f ; zk ). (2.6.1)
γ k=1

Definition 2.6.1. f (z) is a meromorphic function if it is analytic on

the complex plane C with the exception of the finite number of isolated
singularities at z1 , . . . , zN .
Corollary 2.6.2. Let f be a meromorphic function with singularities at
z1 , . . . , zN ∈ C. Then
N
X
Res(f ; zk ) + Res(f ; ∞) = 0. (2.6.2)
k=1
Chapter 2. Basic Properties of Analytic Functions 95

z1

z3
z2

Proof. Consider a domain D = {z : |z| > R} with a counter-clockwise

oriented boundary γ.
According to Theorem 2.6.1
N
X Z
2πi Res(f ; zk ) = f (z) dz = −2πi Res(f ; ∞)
k=1 γ

where the last equality is from the definition of the residue at ∞.

Therefore,
N
X
Res(f ; zk ) + Res(f ; ∞) = 0. (2.6.2)
k=1

2.6.2 Integrals of Rational Functions

P (z)
Consider a rational function f (z) = where P (z) and Q(z) are polyno-
Q(z)
mials of degrees m and n correspondingly. We want to calculate
Z ∞
P (x)
dx,
−∞ Q(x)

assuming that

(a) m ≤ n − 2,

(b) Q(x) does not have real roots.

Chapter 2. Basic Properties of Analytic Functions 96

If m ≥ n − 1 then the integral diverges at ∞ (but as m = n − 1 it could be

understood in the sense of the principal value–however we do not cover it).
If Q(x) has a real root x1 then integral diverges at x1 .

Theorem 2.6.3. Under these assumptions

Z ∞ m
P (x) X P (z)

dx = 2πi Res ; zk (2.6.3)
−∞ Q(x) k=1
Q(z)

where z1 , . . . , zk are all roots of Q(z) belonging to upper complex half-plane

U := {z ∈ C : Im(z) > 0}.

Proof. We want to apply Theorem 2.6.1 and for this we need to take a
+
bounded domain. We take a half-disk DR = {z : |z| < R, Im(z) > 0} which
contains all roots of Q(z) in the upper half-plane.

−R R

+
Then, according to Theorem 2.6.1, with ΓR the boundary of DR
Z m
P (z) X P (z)

dz = 2πi Res ; zk (2.6.4)
ΓR Q(z) k=1
Q(z)

where the left part equal to

Z R Z
P (x) P (z)
dx + dz
−R Q(x) γR Q(z)

because ΓR consists of the real segment [−R, R] and an upper arc γR .

What happens when R → ∞? Observe that

|Q(z)| ≥ 2ϵ0 |z|n − M |z|n−1 ≥ ϵ0 |z|n , |P (z)| ≤ M |z|m

Chapter 2. Basic Properties of Analytic Functions 97

for sufficiently large |z|. If you cannot prove it by yourself, look at pages
154–155 of the Textbook.
P (z)
Therefore, Q(z) ≤ M1 Rm−n on γR and
Z
P (z)
| dz| ≤ πR × M1 Rn−m → 0 as R → ∞
γR Q(z)

since m − n + 1 < 0. On the other hand,

P (x)
≤ M2 (|x| + 1)m−n ∀x ∈ R.
Q(x)

Indeed, it is so for |x| ≥ c and for |x| ≤ c it follows from the assumption
that Q(x)Rdoes not have real roots.
∞
Since −∞ M2 (|x| + 1)m−n dx < ∞ again because m − n + 1 < 0 we
conclude that
Z R Z ∞
P (x) P (x)
dx → dx as R → ∞
−R Q(x) −∞ Q(x)

and (2.6.4) implies

Z ∞ m
P (x) X P (z)

dx = 2πi Res ; zk . (3.6.3)
−∞ Q(x) k=1
Q(z)

Remark 2.6.2. Using lower half-disk D′ = {z : |z| < R, Im(z) < 0} one
could get
Z ∞ r
P (x) X P (z)

dx = −2πi Res ; zk (2.6.3)′
−∞ Q(x) k=m+1
Q(z)

where zk with k = m + 1, . . . , r are roots of Q(z) in the lower half-plane

(why minus? – orientation!) but due to Corollary 2.6.2 it is the same since
P
Res Q ; ∞) = 0 due to m ≤ n − 2.
Remark 2.6.3. Sure, we can calculate such integrals using decomposition
into primitive fractions as in Calculus I.
Chapter 2. Basic Properties of Analytic Functions 98

Z ∞
dx π
Example 2.6.1. = α > 0.
−∞ x2 +α 2 α
Indeed, there is just one root z = iα in the upper half-plane, and
1 1 1
Res( ; αi) = z=αi
=
z2 +α 2 2z 2αi
Z ∞
dx π
Example 2.6.2. = α > 0, β > 0.
−∞ (x2 + α2 )(x2 + β 2) αβ(α + β)

(a) Consider first α ≠ β. Then there are two simple roots z1 = αi and
z2 = βi in the upper half-plane and
 1  1 1
Res 2 2 2 2
; αi = = .
(z + α )(z + β ) 2z(z + β ) z=αi 2α(β − α2 )i
2 2 2

Similarly the second residue is calculated and their sum is

1 1 1
+ = .
2α(β 2 − α2 )i 2β(α2 − β 2 )i 2αβ(α + β)i

(b) As α = β one can consider a double root, and thus a double pole, but
also just plug α = β into the result above (not into each residue!).

2.6.3 Integrals over the Real Axis Involving

Trigonometric Functions
Theorem 2.6.4. Let P (x) and Q(x) are even real-valued polynomials of
degrees m and n correspondingly, m ≤ n − 2 and Q(x) have no real roots.
Then
Z ∞ m  P (z)eiz i
P (x) cos(x) h X
I := dx = Re 2πi Res ; zk (2.6.5)
−∞ Q(x) k=1
Q(z)
and
Z ∞ m  zP (z)eiz i
xP (x) sin(x) h X
J := dx = Im 2πi Res ; zk . (2.6.6)
−∞ Q(x) k=1
Q(z)

where z1 , . . . , zk are all roots of Q(z) belonging to upper complex half-plane

U := {z ∈ C : Im(z) > 0}.
Chapter 2. Basic Properties of Analytic Functions 99

Proof. Since cos(z) and sin(z) are unbounded in both the upper (and lower)
complex half-plane observe first that
Z ∞ P (x)eix 
I = Re dx (2.6.7)
−∞ Q(x)
and
Z ∞ xP (x)eix 
J = Im dx (2.6.8)
−∞ Q(x)

and eiz is bounded in the upper complex half-plane U .

Then the proof repeats the proof of Theorem 2.6.3. Namely, we consider
+
a half-disk DR = {z : |z| < R, Im(z) > 0} and calculate integrals without
taking real or imaginary parts using the Residue Theorem.

−R R

However, for J in the case of m = n − 2 we need some extra arguments.

First, we need to prove that J converges (albeit not absolutely). Indeed,
integrating by parts we see that
Z R x=R
Z R
f (x) sin(x) dx = −f (x) cos(x) + f ′ (x) cos(x) dx,
1 x=1 1

xP (x)
where f (x) = Q(x)
→ 0 as r → +∞ and since

|f ′ (x)| ≤ M (|x|2 + 1)m−n ≤ M (|x|2 + 1)−2

the integral in the right side converges absolutely. Integral from −∞ to −1

is considered in the same way.
We also need to estimate integral over arc γR . The Textbook does it on
pages 158–159 (Jordan’s Lemma) but we do instead a simpler thing: instead
of half-disk we consider a half-square.
Chapter 2. Basic Properties of Analytic Functions 100

−R + iR R + iR

[h!] −R R

Then integrals over vertical lines do not exceed

Z R
M Rm−n+1 e−y dy ≤ M Rm−n+1 ≤ M R−1
0

and integral over upper horizontal line does not exceed

Z R
M Rm−n+1 e−R dx ≤ 2M Rm−n+2 e−R ≤ 2M e−R
−R

and both of those tend to 0 as R → ∞.

Example 2.6.3. Calculate
Z ∞
cos(x) dx
I= , α > 0.
0 x2 + α 2
Solution. Since the integrand is an even function, we see that we can take
integral from −∞ to ∞ and then to halve it. Since Q(x) = x2 + α2 has just
one root αi in the upper complex half-plane, and this root is simple,

1 h  eiz i 1  eiz  πe−α

I= Re 2πi Res 2 , αi = × Re 2πi = .
2 z + α2 2 2z z=αi 2α

Example 2.6.4. Calculate

Z ∞
x sin(x) dx
J= , α > 0.
0 x2 + α 2
Chapter 2. Basic Properties of Analytic Functions 101

Solution. Since the integrand is an even function, we see that we can take
integral from −∞ to ∞ and then to halve it. Since Q(x) = x2 + a2 has just
one root αi in the upper complex half-plane, and this root is simple,
1 h  zeiz i 1  zeiz  πe−α
I = Im 2πi Res 2 , αi = × Im 2πi = .
2 z + α2 2 2z z=αi 2

Example 2.6.5. Calculate

Z ∞
sin(x)
I := dx. (2.6.9)
−∞ x
Solution. This and the following examples are not following from Theo-
rem 2.6.4 because Q(x) has a real root. We apply the modified method.
Observe first that
hZ −ε sin(x) dx Z R sin(x) dx i
I = lim lim+ +
R→∞ ε→0 −R x ε x
hZ −ε eix dx Z R eix dx i
= lim lim+ Im + .
R→∞ ε→0 −R x ε x
=:Jε,R

Now consider a curve Γε,R :

−R + iR R + iR

−R −ε ε R
iz
Due to Cauchy’s Theorem Γε,R e zdz = 0 (there are no singularities
R

inside). Exactly like in the proof of Theorem 2.6.4 it is easy to show that
integrals over both vertical lines and upper horizontal line tend to 0 as
R → ∞ and that
Z ∞ ix Z −1 ix
e dx e dx
and
1 x −∞ x
Chapter 2. Basic Properties of Analytic Functions 102

converge (but not absolutely). Therefore

Z eiz dz  Z dz Z (eiz − 1) dz 
I = − lim+ Im = − lim+ Im +
ε→0 γε z ε→0 γε z γε z
where γε is an arc of radius ε.
Observe that
Z Z 0
dz
= i dθ = −πi
γε z π
and
(eiz − 1) dz
Z
| | ≤ π max |eiz − 1| ≤ M ε → 0 as ε → 0+
γε z γε

and therefore I = Im(πi) = π.

Example 2.6.6.
∞
sin2 (x) dx
Z
I := . (2.6.10)
−∞ x2
Solution. Observe first that
hZ −ε (1 − cos(2x)) dx Z R (1 − cos(2x)) dx i
I = lim lim+ +
R→∞ ε→0 −R 2x2 ε 2x2
h −ε (1 − e2ix ) dx
Z Z R
(1 − e2ix ) dx i
= lim lim+ Re + .
R→∞ ε→0 −R 2x2 ε 2x2
=:Jε,R

Taking the same curve Γε,R as in the previous example, we see that due to
iz
Cauchy’s Theorem Γε,R e zdz = 0 (there are no singularities inside).
R

Again, integrals over both vertical lines and upper horizontal line tend
to 0 as R → ∞. Further,
Z ∞ Z −1
(1 − e2ix ) dx (1 − e2x ) dx
and
1 2x2 −∞ 2x2
converge absolutely. Therefore since 1 − e2iz = −2iz + (1 + 2iz − e2iz ),
Z (1 − e2iz ) dz 
I = − lim+ Re
ε→0 γε 2z 2
−i dz (1 − e2iz + 2iz) dz 
 Z Z
= − lim+ Re +
ε→0 γε z γε 2z 2
Chapter 2. Basic Properties of Analytic Functions 103

where γε is an arc of radius ε. Here the first integral equals −π and the
second tends to 0 as ε → 0 (as in the previous example). Finally, I = π.
Remark 2.6.4. To calculate
Z ∞
P (x) sin(x) dx
(2.6.11)
−∞ xQ(x)
and
∞
P (x) sin2 (x) dx
Z
(2.6.12)
−∞ x2 Q(x)
where P (x) and Q(x) are even real-valued polynomials of degrees m and
n correspondingly, m ≤ n, and Q(x) does not have real roots, we use the
same curve Γε,R and the same approach, albeit instead of Cauchy’s theorem
we use the Residue theorem.
Example 2.6.7.
Z
sin(x) dx
I := , α > 0, (2.6.13)
x(x2 + α2 )
and
sin2 (x) dx
Z
J := , α > 0. (2.6.14)
x2 (x2 + α2 )
Solution. Analysis of the previous two examples shows that the contribution
of 0 (or, more precisely, of −γε with ε → 0+ ) is απ2 (because there is an extra
1 1
factor z2 +α 2 z=0 = α2 ).

π   eiz  π  eiz 
I= + Im 2πi Res ; αi = + Im 2πi
α2 z(z 2 + α2 ) α2 2z 2 z=αi
π
= 2 (1 − e−α )
α
and
π   1 − e2iz  π  (1 − e2iz ) 
J= + Re 2πi Res ; αi = + Re 2πi
α2 2z(z 2 + α2 ) α2 4z 3 z=αi
−2α

π 1−e
= 2 1− .
α 2α
Chapter 2. Basic Properties of Analytic Functions 104

2.6.4 Integrals Involving ln(x) or Fractional Powers of x

In this subsection we will deal again mainly with examples to demonstrate
approaches than with general theorems. One can formulate theorems easily.
Consider
Z ∞ p
x P (x) dx
I= (2.6.15)
0 Q(x)

where P (x) and Q(x) are even real-valued polynomials of degrees m and n
correspondingly, and −1 < p < 1, p ̸= 0, m + p + 1 < n and Q(x) does not
vanish on the real line.
This integral is understood as (a garden variety) improper integral.
Example 2.6.8.
∞
xp dx
Z
I= , −1 < p < 1, α > 0. (2.6.16)
0 x2 + α 2
zp
Solution. Consider f (z) = 2 and a curve Γε,R which is already famil-
z + α2
iar:

−R −ε ε R

and select a branch z p which is xp as z = x > 0 and analytic inside Γε,R .

Then due to the Residue Theorem

z p dz  zp (αi)p
Z 
2 2
= 2πi Res 2 ; αi = 2πi
Γε,R z + α z + α2 2z z=αi
iπp
= παp−1 e 2 (2.6.17)
iπp
because ip = eZ2 on the selected branch.
R
xp dx
Let Iε,R = 2 2
denote a required integral, but from ε to R.
ε x +α
Chapter 2. Basic Properties of Analytic Functions 105

On the other hand, integral from −R to −ε is eiπp Iε,R because (−1)p =

eiπp on the selected branch. Therefore due to (2.6.17)
Z Z
iπp
iπp
. . . = παp−1 e 2


1+e Iε,R + ... + (2.6.18)
γR γε

where γR and γε denote large and small arc correspondingly. R

Observe that Iε,R → I as ε → 0+ and R → ∞. Further, γR . . . → 0 as
R
R → ∞ and γε → 0 as ε → 0+ (like in the previous lectures).
Then due to (2.6.18)
iπp
1 + eiπp I = παp−1 e 2

and, finally
iπ
παp−1 e 2 παp−1 παp−1 pπ

I= iπp
= iπp iπp = sec .
1+e e− 2 + e 2 2 2

Observe that the final answer is real and positive (as it should be).
Example 2.6.9.
Z ∞
ln(x) dx
I := , α > 0. (2.6.19)
0 x2 + α 2
Solution. We repeat arguments of Example 2.6.8, selecting the same domain
Log(z)
and curve, and f (z) = 2 where we select a branch Log(z) which is
z + α2
ln(x) as z = x > 0 and analytic inside Γε,R . What will change?
First, we have a different right-hand expression in (2.6.17):
 Log(z)  Log(z) π π iπ

2πi Res ; αi = 2πi = Log(αi) = ln(α) +
z2 + α 2 2z z=αi α α 2
because of the branch selection.
RR
Second, while Iε,R = ε ln(x) dx
x2 +α2
,
−ε Z −ε

ln(|x|) + πi dx
Z
Log(x) dx
=
−R x2 + α 2 x2 + α 2
R

−R Z R
ln(|x|) + πi dx
Z
dx
= = Iε,R + πi ,
ε x2 + α 2 2
ε x +α
2
Chapter 2. Basic Properties of Analytic Functions 106

where we changed x := −x passing from the first line to the second one.
R
Again, Iε,R → I as ε → 0+ and R → ∞, γR . . . → 0 as R → ∞ and
R
γε
. . . → 0 as ε → 0+ and
Z ∞
dx π iπ

2I + πi = ln(α) +
0 x2 + α 2 α 2

π ln(α)
and taking real part we get 2I = π ln(α) and I = while imaginary
R ∞ dx α π
2α
part will just bring us trivial 0 x2 +α2 = 2α .
The same trick would work in many cases, like with an extra factor xp
with −1 < p < 1.
Example 2.6.10.
Z ∞
x ln(x) dx
I := , α > 0, β > 0, α ̸= β. (2.6.20)
0 (x + α2 )(x2 + β 2 )
2

Solution. The same trick out of the box would not work here because without
ln(x) we have an odd function and
Z −ε Z R
x(ln(|x|) + πi) dx x(ln(|x|) + πi) dx
2 2 2 2
=− 2 2 2 2
−R (x + α )(x + β ) ε (x + α )(x + β )
RR
and adding ε (x2x+α
ln(|x|) dx
2 )(x2 +β 2 ) we get just integral without logarithmic factor.

So, to overcome this problem we add another logarithmic factor and

z Log2 (z)
consider f (z) = 2 .
(z + α2 )(z 2 + β 2 )
The “residue answer” will be then
 z Log2 (z) 
2πi Res , αi + . . .
(z 2 + α2 )(z 2 + β 2 )
z Log2 (z)
=2πi + ...
2z(z 2 + β 2 ) z=αi

2
2
ln(α) + πi 2
− ln(β) + πi 2
=πi
β 2 − α2



ln(α) − ln(β) ln(α) + ln(β) + πi
=πi .
α2 − β 2
Chapter 2. Basic Properties of Analytic Functions 107

On the other hand, the integral expression after we take the limit as R → ∞
and ε → 0+ is
Z ∞
2
x ln2 (x) dx
Z 0
x ln(|x|) + iπ dx
+
0 (x2 + α2 )(x2 + β 2 ) 2 2 2
−∞ (x + α )(x + β )
2
Z ∞ Z ∞
2
x ln2 (x) dx x ln(x) + iπ dx
= −
0 (x2 + α2 )(x2 + β 2 ) (x2 + α2 )(x2 + β 2 )
Z ∞
0
x 2 ln(x) + iπ dx
= − iπ .
0 (x2 + α2 )(x2 + β 2 )

This expression must be equal to the last expression on the previous slide.
We take only imaginary parts (taking real parts will bring a much easier
result)



ln(α) − ln(β) ln(α) + ln(β)
−2πI = π .
α2 − β 2
Taking α = β and calculating the limit we would automatically bring result
in this case as well:
ln(α)
−2πI = π .
α2

Example 2.6.11.
Z ∞
xp dx
I := , −1 < p < 1, p ̸= 0, α > 0. (2.6.21)
0 (x − a)2 + α2

Solution. If a ̸= 0, then (x − a)2 + α2 is neither even, nor odd and the

previous domain and curve do not work! We need something new, and this
is a keyhole domain.
Chapter 2. Basic Properties of Analytic Functions 108

Here the radius of the large arc is R, the radius of small arc is ε and the
width of the cut is 0!
zp
We take f (z) = where z p is the branch which on the upper
(z − a)2 + α2
side of the cut, when z = x + i0, x > 0, is equal to xp . Then on the lower
side of the cut, z = x − i0, x > 0, is equal to e2iπp xp .
One can see easily, that the integral over γR (the large arc) tends to 0 as
R → ∞, integral over γε (the small arc) tends to 0 as ε → 0+ , and integrals
over upper and lower sides of the cut tend to I and −e2iπp I correspondingly,
because while we integrate from ε to R on the upper side of the cut, we
integrate from R to ε on the lower side!
So the integral would be
(1 − e2iπp )I
in the end.
Now we have not one, but two poles z± = a ± αi. This is why we leave
keyhole domain as a last reserve.
Calculating residues
 zp  zp 1
Res 2 2
; a ± αi = =± (a ± αi)p .
(z − a) + α 2(z − a) z=a±αi 2αi
Therefore
π 
1 − e2iπp I = (a + αi)p − (a − αi)p


α
π 
= (a + αi)p − e2iπp (a − αi)p
α
where in the first line ζ p is a branch defined on 0 < arg(z) < 2π, while in
the second line ζ p is a branch defined on −π < arg(z) < π. You just need
to look at arguments of ζ and ζ p .
Then
π 
e−ipπ − eiπp I = e−ipπ (a + αi)p − eiπp (a − αi)p .


α
p
Taking imaginary part and observing that in these new settings ζ p = ζ we
get
2π  −iπp 
−2 sin(πp)I = Im e (a + αi)p .
α
Chapter 2. Basic Properties of Analytic Functions 109

Finally,
π  
I = − csc(πp) Im e−iπp (a + αi)p .
α

Remark 2.6.5. (a) For a = 0 this answer coincide with the answer obtained
in Example 1.
(b) One can consider examples with the factor xp lnq (x).
(c) For
Z ∞
ln(x) dx
I := , α > 0,
0 (x − a)2 + α2
Log2 (z)
one should take a keyhole domain (if a ̸= 0) and f (z) = 2 with
z + α2
Log(z) = ln(x) on the upper side of the cut and Log(z) = ln(x) + 2πi
on it’s lower side (think–why square).
(d) With keyhole domain one cannot consider examples with factors cos(x)
or sin(x) because eiz is bounded only in the upper half-plane.

2.6.5 Integrals of Trigonometric Functions over [0, 2π]

In this ssubsection we will deal again mainly with examples to demonstrate
approaches than with general theorems. One can formulate theorems easily.
Consider
Z 2π

P cos(θ), sin(θ) dθ
I :=
(2.6.22)
0 Q cos(θ), sin(θ)


where P (x, y) and Q(x, y) are real-valued polynomials and Q cos(θ), sin(θ) ̸=
0 for all θ ∈ [0, 2π].
Then one can rewrite it as
P z2 + 2z
1 z 1


, 2i − 2iz dz
Z
I := z 1 z 1

(2.6.23)
|z|=1 izQ 2 + 2z , 2i − 2iz

and use the Residue Theorem.

We start from some old example, we used Cauchy’s formula to solve:
Chapter 2. Basic Properties of Analytic Functions 110

Example 2.6.12.
Z 2π
dθ
I := , a > |b| > 0. (2.6.24)
0 a − b cos(θ)

Solution. Then
Z Z
dz 2i dz
I= z 1

=−
|z|=1 iz a − b( 2 + 2z ) |z|=1 bz 2 + b − 2az
Z
2 dz
=− .
bi |z|=1 (z − z1 )(z − z2 )
√ √
with z1 = 1b a − a2 − b2 , z2 = 1b a + a2 − b2 = z1−1 , |z1 | < 1 < |z2 |.

By the Residue Theorem

4π  1  4π 1 2π
I =− Res ; z1 = − =√ .
b (z − z1 )(z − z2 ) b z1 − z2 a2 − b 2

Example 2.6.13.
Z 2π
dθ
I := , a > |b| + |c|, |c| > 0. (2.6.25)
0 a2 − (b + c cos(θ))2

Solution. Then
Z
dz
I=
− [b + c( z2 + 2z
1 2


|z|=1 iz a2 )]
Z
4iz dz
= .
|z|=1 [2bz + c(z 2 + 1)]2 − 4a2 z 2

1
p

The denominator has two roots inside z1,2 = c
a±b− (a ± b)2 − c2 and
1
two roots outside z3,4 = z1,2 . Then
Z
4i z dz
I= 2 ,
c |z|=1 (z − z1 )(z − z2 )(z − z3 )(z − z4 )

which could be calculated using the Residue Theorem.

Chapter 2. Basic Properties of Analytic Functions 111

Remark 2.6.6. Example 7 on page 161 of the Textbook is superficial. Indeed

Z 2π Z π
dθ dθ
2
=2 2
0 a + c cos (θ) 0 a + c cos (θ)
Z π Z 2π
4dθ 2 dθ
= =
0 2a + c + c cos(2θ) 0 2a + c + c cos(θ)

where in the last equality we made a change θ := 2θ.

And the last integral is exactly what we considered in Example 2.6.12.
Chapter 3

Analytic Functions as Mappings

3.1 The Zeros of an Analytic Function

3.1.1 Introduction
The theme of this chapter is the geometry of analytic functions, in particular
the nature of their range. This theme manifests itself first in several theorems
in Section 1 that pinpoint the number of solutions to an equation of the
form f (z) = w0 . The answer is formulated in several different ways, all
connected closely with the behaviour of the function f on a closed curve.
This ability to count the number of solutions of f (z) = w0 is exploited
in Section 2 to prove that the range of a nonconstant analytic function is an
open set. This is a profound and elegant property of analytic functions that
leads in a natural way to the maximum modulus principle.
Even more important, we are led directly to the study of conformal
mapping, a tool of enormous beauty and power both in applications and
theory.
We mainly concentrate on Möbius map, also known as linear fractional
transformation.

3.1.2 Reminder: Definitions

Recall that

(a) z0 is a zero of analytic function f if f (z0 ) = 0.

112
Chapter 3. Analytic Functions as Mappings 113

(b) If f is not an identical 0, then there exists an integer m ≥ 1, such that

f (z0 ) = f ′ (z0 ) = . . . = f (m−1) (z0 ) = 0, f (m) (z0 ) ̸= 0.

We call m the order of zero z0 .

(c) Equivalently,

f (z) = (z − z0 )m g(z) with g(z0 ) ̸= 0 (3.1.1)

where g(z) is analytic.

(d) Zero is simple if m = 1, double if m = 2, triple if m = 3 and so on.

3.1.3 Zeroes are Isolated

Theorem 3.1.1. Suppose that f is analytic in a domain D. Suppose that
there are distinct points z1 , z2 , . . . ∈ D with f (zn ) = 0 for all n and that the
sequence {zn } converges to a point z0 ∈ D.
Then f (z) = 0 for all z ∈ D.

Remark 3.1.1. This theorem states that zeroes of analytic function are
isolated and cannot accumulate to a point in D.
However they can accumulate to the point on the boundary, including
∞. For example, zeroes of sin(z) are zn = πn and they accumulate to ∞,
and zeroes of sin z1 which is analytic in C \ {0} are accummulating to 0.

Proof of Theorem 3.1.1. Since f (zn ) = 0 and zn → z0 ∈ D we conclude

that f (z0 ) = 0. Let m be an order of zero at z0 , then (3.1.1) holds.
Then g(z) ̸= 0 for z : |z − z0 | < ε for some ε > 0 and therefore f (z) ̸= 0
z : 0 < |z − z0 | < ε. Contradiction!

3.1.4 Counting the Number of Zeroes

Now we can think about counting the number of zeroes of analytic function
f (z) in domain D. However, it is reasonable to count zeroes together with
their orders, for example, double zero is counted as 2, triple as 3 and so
on. . .
Indeed, counting numbers without their orders is unstable: if f (z) de-
pends continuously on a parameter, then zeroes can merge and split up
Chapter 3. Analytic Functions as Mappings 114

2
when this
√ parameter changes, like z + ε has two different simple zeroes
z1,2 = ± −ε as ε ̸= 0, and one double zero as ε = 0. More generally, it is
the same for roots of polynomials P (z) when coefficients change.

No Gimli’s statistics: “That still only counts as one!”

z−ε
We want also count poles as opposite to zeroes. For example has a
z+ε
simple pole z = −ε and a simple zero z = ε as ε ̸= 0 and has none as ε = 0
(remember that removable singularity is not a singularity at all!)
Recall that

(a) z0 is a pole of function f analytic in {z : 0 < |z − z0 < ε}, if z0 is a

1
zero of h(z) = .
f (z)
(b) The order of the pole n of f at z0 is the order of zero n of h(z) at z0 ; pole
is called simple, double, triple . . . as n = 1, 2, 3, . . . correspondingly.

(c) If z0 is a pole of order n in z0 then

f (z) = (z − z0 )−n g(z0 ) with g(z0 ) ̸= 0 (3.1.2)

where again g(z) is analytic (in the neighbourhood of z0 ).

Lemma 3.1.2. (i) Let f (z) have a zero of order m at z0 . Then

f′ 
Res ; z0 = m. (3.1.3)
f

(ii) Let f (z) have a pole of order n at z0 . Then

f′ 
Res ; z0 = −n. (3.1.4)
f
f′
Remark 3.1.2. (a) If z0 is a regular point, neither zero nor pole, then f
has no singularities at z0 and the corresponding residue is 0.

(b) If z0 is an essential singularity, then anything can happen, and we

assume that there are no essential singularities in D!
Chapter 3. Analytic Functions as Mappings 115

Proof of Lemma 3.1.2. (i) If f (z) has a zero of order m at z0 , then f (z) =
(z − z0 )m g(z),

f ′ (z) =m(z − z0 )m−1 g(z) + (z − z0 )m g ′ (z)

f′ g′
=⇒ =m(z − z0 )−1 +
f g
g′ f′ f′


with g
analytic at z0 . Then z0 is a simple pole of f
and Res f
; z0 =
m.

(ii) If f (z) has a pole of order n at z0 , then f (z) = (z − z0 )−n g(z),

f ′ (z) = − n(z − z0 )−n−1 g(z) + (z − z0 )−n g ′ (z)

f′ g′
=⇒ = −n(z − z0 )−1 +
f g
g′ f′ f′


with g
analytic at z0 . Then z0 is a simple pole of f
and Res f
; z0 =
−n.
Now we are able to proof the first major theorem of this Chapter (see
Theorem 1 of the textbook on page 173).

Theorem 3.1.3. Suppose that f is analytic in a domain D except for a

finite number of poles.
Let γ be a piecewise smooth properly oriented simple closed curve in D,
which does not pass through any pole or zero of f and whose inside lies in
D. Then
Z ′
1 f (z)
dz = N (f ; γ) − P (f ; γ) (3.1.5)
2πi γ f (z)

where

- N (f ; γ) is a number of zeros of f inside γ;

- P (f ; γ) is a number of poles of f inside γ.

All zeros and poles are counted with multiplicities.

Chapter 3. Analytic Functions as Mappings 116

Proof. Let z1 , . . . , zp be zeroes inside γ and w1 , . . . , wq be poles inside γ.

f′
Then has simple poles at z1 , . . . , zp ; w1 , . . . , wq and all other points of
f
inside γ are regular.
Due to the Residue Theorem

p q
f′ f′  X f′
Z
1 X 
dz = Res ; zj + Res ; wk
2πi γ f j=1
f k=1
f
p q
X X
= mj + (−nk ) = N (f ; γ) − P (f ; γ)
j=1 k=1

where equality between the first and the second lines follows from Lemma 3.1.2.

3.1.5 Infinity
What about infinity? Recall that

(a) f (z) has a zero of order m at ∞ if f (z) = z −m g(z) with g(z) having
removable singularity at ∞ and g(∞) ̸= 0. Then

f′ g′ f′ 
= −mz −1 + =⇒ Res ;∞ = m
f g f
since at ∞ residue is calculated with the opposite sign!

(b) f (z) has a pole of order n at ∞ if f (z) = z n g(z) with g(z) having
removable singularity at ∞ and g(∞) ̸= 0. Then

f′ −1 g′ f′ 
= nz + =⇒ Res ; ∞ = −n
f g f
since at ∞ residue is calculated with the opposite sign!

Therefore,
Remark 3.1.3. (a) Statement of Lemma 3.1.2 holds for ∞ as well!
Chapter 3. Analytic Functions as Mappings 117

(b) Further, Statement of Theorem 3.1.3 holds not only for inside γ but
also for outside γ (in which case one must assume that outside γ f
is analytic except for a finite number of poles, and ∞ is also either a
pole or zero, and it is included then in calculating N (f ; γ) and P (f ; γ).
Also γ must be properly oriented, so going along it leaves outside on
the left!

3.1.6 The Argument Principle

Theorem 3.1.3 states that
Z ′
1 f (z)
dz = N (f ; γ) − P (f ; γ) (3.1.5)
2πi γ f (z)

where γ is a a properly oriented closed curve, and

- N (f ; γ) is a number of zeros of f inside γ;

- P (f ; γ) is a number of poles of f inside γ

(and we always count multiplicities of zeros and poles).

f′
′
Observe that = log(f ) and therefore for a simple counter-clockwise
f
and “one-piece” closed curve γ the left-hand expression in (3.1.5) is equal to
Z ′ Z
f (z)
′
dz = log(f ) dz
γ f (z) γ
= log(f (γ(t1 )) − log(f (γ(t0 ))

where γ is parametrized by t ∈ [t0 , t1 ].

Obviously, the answer is 0 if log(f ) is a single-valued function, which
happens, when f has neither 0 nor poles inside γ, but it is not the case
otherwise.
In general it is not so, and since log(f ) = ln |f | + i arg(f ), and ln |f | is a
single valued function, the left-hand expression of (3.1.5) is equal tp

i arg(f (γ(t1 )) − i arg(f (γ(t0 ))

and we arrive to the following Theorem (see Theorem 2 of the Textbook,

page 175)
Chapter 3. Analytic Functions as Mappings 118

Theorem 3.1.4 (The Argument Principle). Suppose f is analytic on a

domain D except for isolated poles. Let γ be a piecewise smooth positively
oriented simple closed curve in D whose inside lies in D and which does not
pass through any zeros or poles of f .
Then,
" #
1 Change of arg(f )(z)
= N (f ; γ) − P (f ; γ) (3.1.6)
2π as z traverses γ

where γ is a a properly oriented closed curve, and

- N (f ; γ) is a number of zeros of f inside γ;
- P (f ; γ) is a number of poles of f inside γ.

Example 3.1.1. (a) f (z) = z (simple zero at 0) then, when z goes once
around 0 and its argument changes from 0 to 2π, arg(f (z)) changes
with the same speed from 0 to 2π;
(b) f (z) = z 2 (double zero at 0) then, when z goes once around 0 and its
argument changes from 0 to 2π, arg(f (z)) changes with the double
speed from 0 to 4π;
(c) f (z) = z 3 (triple zero at 0) then, when z goes once around 0 and
its argument changes from 0 to 2π, arg(f (z)) changes with the triple
speed from 0 to 6π;
(d) f (z) = z −1 (simple pole at 0) then, when z goes once around 0 and
its argument changes from 0 to 2π, arg(f (z)) changes with the same
speed but in the opposite direction from 0 to −2π;
(e) f (z) = z −2 double pole at 0) then, when z goes once around 0 and its
argument changes from 0 to 2π, arg(f (z)) changes with the double
speed but in the opposite direction from 0 to −4π;
(f) f (z) = z −3 triple pole at 0) then, when z goes once around 0 and
its argument changes from 0 to 2π, arg(f (z)) changes with the triple
speed but in the opposite direction from 0 to −6π.

(a) z (b) z 2 (c) z 3 (d) z −1 (e) z −2 (f) z −3

Chapter 3. Analytic Functions as Mappings 119

Example 3.1.2. Find the number of zeros of the function f (z) = z 3 − 2z 2 + 4

in the first quadrant.
Solution. We examine f (z) on the contour shown in figure below, R is very
large.

iR

(a) On the horizontal segment, from 0 to R, z = x and f (x) = x3 − 2x2 + 4

is real and greater than 2. Indeed, f ′ (x) = 0 =⇒ 3x2 − 4x = 0 =⇒
x = 0 or x = 43 with both values ≥ 2 and f (x) → +∞ as x → +∞.
Therefore its argument stays 0 here.
(b) On the
arc z = Reit (with t changing from 0 to π2 ) f (z) = R3 e3it 1 +
O(R−1 ) and therefore argument changes from 0 to ≈ 3π 2
.
(c) On the vertical segment z = iy (with y changing from R to 0) f (z) =
−y 3 i + 2y 2 + 4 and it belongs to the fourth quadrant {w : Re(w) >
0, Im(w) < 0} and it changes from ≈ 3π 2
as z = Ri to 2π as z = 0.

Consequently, as z traverses the contour, arg(f (z)) increases by exactly

2π, so f (z) = z 3 − 2z 2 + 4 has precisely one zero in the first quadrant.
Remark 3.1.4. (a) In above calculations it is important that f (z) belongs
to a specific quadrant (or other part of the plane), which does not
allow it to go around 0, so its increment is uniquely defined.
(b) We may calculate arg(f (z)) with a small error, but since its increment
is 2πn, n ∈ Z, the answer is precise.
Example 3.1.3. Let λ ≫ 1. Use the argument principle to show that the
equation

z + e−z = λ

has exactly one solution in the right half-plane {z : Re(z) > 0}.
Chapter 3. Analytic Functions as Mappings 120

Solution. We examine f (z) on the contour shown in figure below, R is very

large (depending on λ).

iR

−iR

(a) On the vertical segment z = iy,

h(z) = iy + e−iy − λ = (cos(y) − λ) + i(y − sin(y))
has a negative real part (so it belongs to the left complex half-plane
{w : Re(w) < 0}); when y = R we have arg(h(iy)) ≈ arg(iy) = π2 and
when y = −R we have arg(h(iy)) ≈ arg(iy) = π2 .
(b) On the other hand, on the arc, z = Reit and
h(Reit ) = Reit 1 + O(R−1 ) );

when t = − π2 we have z = −R and arg(h(z)) ≈ 3π 2

, when t changes
slightly, so does arg(h(z)) but then h(z) stays in the right complex
half-plane until t almost reaches π2 , and then arg(h(R)) ≈ 5π2
.

Consequently, as z traverses the contour, arg(f (z)) increases by exactly

2π, so h(z) = z − e−z − λ has precisely one zero in the first quadrant.

3.1.7 Rouché’s Theorem

Theorem 3.1.5 (Rouché’s Theorem). Suppose f and g are analytic on an
open set containing an piecewise smooth simple closed curve γ and its inside.
Also suppose that
|f (z) + g(z)| < |f (z)| ∀z ∈ γ. (3.1.7)
Then f and g have an equal number of zeros inside γ, counting multi-
plicities.
Chapter 3. Analytic Functions as Mappings 121

Proof. Note that the assumption (3.1.7) (strict inequality! ) ensures that
neither f nor g is zero on γ. We may assume further that all the common
zeros of f and g inside γ have been canceled; this affects neither the
assumptions nor the conclusion.
g
Let h = ; then
f
|h(z) + 1| < 1 ∀z ∈ γ

so the range of h on γ lies in the open disk of radius 1 centered at the point
−1.
In particular, arg(h(z)) has no net change as z traverses γ because
w = f (z) cannot go around 0. By the Argument Principle, the number of
zeros of h inside γ equals the number of poles of h inside γ.
But the number of zeros of h is just the number of zeros of g, and the
number of poles of h just the number of zeros of f .
Remark 3.1.5. (a) Suppose that f and g satisfy the hypotheses of Theo-
rem 3.1.5 except that

|f (z) − g(z)| < |f (z)| ∀z ∈ γ. (3.1.8)

It is then still true that f and g have equally many zeros inside γ. The
reasoning is elementary: according to Theorem 3.1.5, the assumption
(3.1.8) implies that f and −g have equally many zeros inside γ.
However, a point is a zero of −g exactly when it is a zero of g.
(b) So, we need to construct a a reference function g(z), satisfying (3.1.7)
or (3.1.8) and such that we can easily calculate the number of its zeros.
Example 3.1.4. Show that all the zeros of

p(z) = 3z 3 − 2z 2 + 2iz − 8

lie in the annulus {z : 1 < |z| < 2}.

Solution. (a) On the circle {z : |z| = 1}

|f (z) + 8| ≤ 3 + 2 + 2 = 7 < 8,

so p(z) and f (z) = 8 have the same number of zeros within {z : |z| = 1}, by
Rouché’s Theorem; thus, p(z) does not vanish in the disk {z : |z| = 1}.
Chapter 3. Analytic Functions as Mappings 122

(b) Furthermore, on {z : |z| = 2},

|p(z) − 3z 3 | ≤ 20 < 24 = |3z 3 |
so p(z) and g(z) = 3z 3 have an equal number of zeros within the circle
{z : |z| = 2}; since g(z) has 3 zeros here (more precisely, one triple zero
z = 0), we conclude that p(z) has three of its zeros within {z : |z| = 2}.
Consequently, all the zeros of p(z) lie in the annulus {z : 1 < |z| < 2}.

Example 3.1.5. Consider

P (z) = z 3 + 2z − 3 − i
and, using the argument theorem and Rouchè’s theorem calculate the number
of its roots in each of the following domains:

(a) {z : |z − 1| < 1}; (c) {z : |z| > 2}.

(b) {z : |z − 1| > 1, |z| < 2},

Solution. (a) As |z| = 2 consider Q(z) = z 3 ; then |Q(z)| = 8,

√
|P (z) − Q(z)| = |2z − 3 − i| ≤ 4 + |3 + i| = 4 + 10 < 8;
therefore P (z) has as many roots in {z : |z| < 2} as Q(z) has, which is 3
(we count orders).
(b) Then there are no roots in {z : |z| > 2} due to Principal Theorem of
Algebra.
(c) Consider z = w + 1 with |w| = 1; then
P (w + 1) = w3 + 3w2 + 5w − i
and we set Q(w) = 5w; then |Q(w)| = 5,
|P (w) − Q(w)| = |w3 + 3w2 − i| < 5.
Indeed, |P (w) − Q(w)| ≤ |w + 3|
√ + 1 < 5 except w = 1, and for w = 1 we
3 2
have |w + 3w − i| = |4 − i| = 17 < 5.
Therefore P (w + 1) has as many roots in {w : |w| < 1} as Q(w) has,
which is 1 (we count orders).
Finally, in the domain {z : |z − 1| > 1, |z| < 2}, there are 3 − 1 = 2 roots.
Chapter 3. Analytic Functions as Mappings 123

0
2
1

Example 3.1.6. Show that the equation

ez = e2 z

has a real root in the unit disk {z : |z| < 1}.

Are there non-real roots?
Solution. We are looking for zeros of f (z) = e2 z − ez . Compare it with
g(z) = e2 z. On the boundary |g(z)| = e2 and
|g(z) − f (z)| = |ez | ≤ eRe(z) ≤ e < e2 .
Therefore f (z) has as many roots in the unit disk as g(z), which has one
(at 1).
So, f (z) has exactly 1 root there: ez = e2 z =⇒ ez = e2 z, ez̄ = e2 z and
therefore z is also a root; since there is just one root, z = z and z is real.
Alternatively, consider a real function f (x) = ex − e2 x, x ∈ (0, 1). Then
f (0) = 1 > 0 and f (1) = e − e2 < 0 and since f is a continuous function,
there is x ∈ (0, 1) s.t. f (x) = 0.
G(z)
Corollary 3.1.6. (i) Assume that f (z) = H(z)
where G and H are poly-
nomials. Then

N (f ) = P (f ) (3.1.9)

where N (f ) and P (f ) are number of zeroes and poles of f on the

extended complex plane Cb (including ∞ and counting orders).

(ii) In particular, if f (z) = G(z) is a polynomial, then the number of roots,

counting multiplicities is equal to the order of polynomial G.
Chapter 3. Analytic Functions as Mappings 124

Proof. Indeed, G has just one pole, namely, ∞ and the order of it is equal
to the order of polynomial G.
Remark 3.1.6. The last statement is known as a Fundamental Theorem of
Algebra (of polynomials).

3.2 Maximum Modulus and Mean Value

3.2.1 Maximum Modulus
Suppose that f (z) is an analytic function on a domain D. We saw in the
beginning of Chapter 2 that if the range of f lies in a circle or on a straight
line, then f is necessarily constant. We shall see now that a great deal more
is true: either f is constant on D or the range of f is an open set.
Suppose that f is not identically constant, and let w0 = f (z0 ) be an
arbitrary point in the range of f . Then since f is not identically constant,
function f (z) − w0 is not identically zero, and so f (z) − w0 has a zero of
order m ≥ 1 at z0 .
Choose r > 0 so small that f (z) − w0 has no zero in the punctured disk
{z : 0 < |z − z0 | ≤ r}; this is possible because the zeros of a non-constant
analytic function are isolated (see Section 1).
Let δ > 0 be the minimum value of |f (z) − w0 | for all z : |z − z0 | = r, and
let w be any point with |w − w0 | < δ . Then on the circle {z : |z − z0 | = r},

|(f (z) − w) − (f (z) − w0 )| = |w − w0 | < δ ≤ |f (z) − w0 |. (3.2.1)

Therefore, Rouché’s Theorem implies that the two functions f (z) − w

and f (z) − w0 have an equal number of zeros in the disk {z : |z − z0 | ≤ r}.
However, f (z) − w0 has exactly m zeros, and hence so does f (z) − w.
This shows that each point w0 in the range of f lies at the center of a small
disk, which is also within the range of f . Thus, we have established the
following:

Theorem 3.2.1 (Theorem 1 in the Section 3.2 of the Textbook). Suppose

that f (z) is a non-constant analytic function on a domain D. Then the
range of f (z), as z varies over D, is an open set.

We can draw an immediate conclusion from the work that preceded the
statement of Theorem 3.2.1.
Chapter 3. Analytic Functions as Mappings 125

Theorem 3.2.2. Suppose that f (z) is a non-constant analytic function on

a domain D and that f (z) − f (z0 ) has a zero of order m at z0 .
Then f is m-to-1 near z0 , that means that preimage of w ̸= w0 contains
m distinct points (if we look only at points within the disk {z : |z − z0 | < r};
in particular, if f ′ (z0 ) = 0, then f is not one-to-one in any disk containing
z0 ).

Theorem 3.2.3 (The Maximum-Modulus Principle). 1) If f (z) is a non-

constant analytic function on a domain D, then |f (z)| can have no local
maximum on D.

Proof. Assume that z0 ∈ D is a point of local maximum: |f (z0 )| ≥ |f (z)|

for all z : |z − z0 | < r.
Then f (z0 ) lies on the boundary of the open set W := {f (z) : |z−z0 | < r}.
This is in contradiction to the fact that W is an open set containing f (z0 ).
The maximum-modulus principle (Theorem 3.2.3) has ramifications for
the real and imaginary parts of an analytic function. Suppose that f is a
non-constant analytic function on a domain D; let u = Re(f ) and g = ef .
Then both g and g1 are analytic on D, so neither of the functions |g| = eu
and | g1 | = e−u has a local maximum in D.
Therefore, u has no local maxima and no local minima in D: Further, if
we apply the same arguments to if (z) instead of f (z) then we would have
the same conclusion for v = − Im(f ) and we arrive to the following

Theorem 3.2.4. Let f is analytic and non-constant on a domain D.

(i) Re(f ) has no local maxima and no local minima on D.

(ii) Im(f ) has no local maxima and no local minima on D.

Remark 3.2.1. This is a special case of more general theorem, proven in

PDE: Let u be a harmonic function in domain D ⊂ Rn . Then, either u is
constant in D or f has no local maxima and no local minima in D.
Let us continue with maximum-modulus principle a bit further. Let D
be a bounded domain and let B be the boundary of D. Then the closure of
D, D ∪ B is both closed and bounded. We know from Calculus II that a
1)
Corollary 1 of theTextbook
Chapter 3. Analytic Functions as Mappings 126

continuous real-valued function on a compact set (that means closed and

bounded set) actually attains its maximum value.
Let us apply this to each of the three functions |f |, Re(f ), and − Re(f ),
where f itself is analytic on D and continuous on D ∪ B. We know that
each of these three functions is continuous on D ∪ B and so must attain its
maximum somewhere on D ∪ B.
If f is non-constant, then the maximum can not be attained on D itself.
Hence, the maximum must be attained on B, the boundary of D. Quite
obviously, the same conclusion holds if, in fact, f is constant on D. Thus, in
all cases, |f |, Re(f ), and − Re(f ) attain their maximum values on B, the
boundary of D:

Corollary 3.2.5. If f is analytic on a bounded domain D and continuous

on D ∪ B, where B is the boundary of D, then each of |f |, Re(f ), and
− Re(f ) attains its maximum value on B (and it is also true for Im(f ) and
− Im(f )).

Corollary 3.2.6. Let Re(f ) = 0 everywhere on B. Then Re(f ) = 0 in D,

so f itself is constant on all of D.

Remark 3.2.2. (a) This is a very special case of the uniqueness theorem
in PDE: let u be a harmonic function in the bounded domain D ⊂ Rn ,
and u = 0 on its boundary. Then u = 0 in D. Moreover, if u and v
are two harmonic functions in D; then u − v is also harmonic in D
and if u = v everywhere on B, then u = v everywhere in D.

(b) Assumption that D is being bounded is not a superfluous here. For

instance, f (z) = iz is analytic and non-constant on the whole plane,
Re(f ) = 0 on the real axis, and yet f is not constant on the upper
half-plane {z : Im(z) > 0}, whose boundary is the real axis.
Example 3.2.1.

Let u be a harmonic function in the disk

D = {(x, y) : x2 + y 2 ≤ 9}, equal to

| sin(θ)|
g(θ) =
2 + cos(θ)

on its boundary, where θ is a polar angle. Find

max u and min u

D D

and the points, where they are achieved.

Chapter 3. Analytic Functions as Mappings 127

Solution. Since g (and thus u also) is symmetric with respect to x we can

consider only points with θ : 0 ≤ θ ≤ π.
Finding maxima and minima of g: on the upper half-circle with 0 ≤ θ ≤ π

sin(θ) 2 cos(θ) + 1
g(θ) = =⇒ g ′ (θ) = =0
2 + cos(θ) (2 + cos(θ))2
1 2π 1
=⇒ cos(θ) = − =⇒ θ = =⇒ g(θ) = √ .
2 3 3
However we need to explore the ends θ = 0 and θ = π, where g(θ) = 0.
So

1
(a) maxD (u) = √ achieved as θ = 2π 3
and
3
θ = 4π
3
(don’t forget
√
about lower half circle),
3 3 3
or points − 2 ± i 2 .

(b) minD (u) = 0 achieved as θ = 0 and θ = π,

or points ±3.

Example 3.2.2.

Let

f (z) = z 2
and
D = {z : |z − 1| ≤ 1}.
Find
max Im(f (z)) and min Im(f (z))
z∈D z∈D

and point(s), where it is achieved.

Solution. Since maxima and minima are achieved on the boundary, we
consider z = a + cos(t) + i sin(t) with a = 1. Then

Im f (z) = 2 a + cos(t) sin(t) =: 2ϕ(t)
Chapter 3. Analytic Functions as Mappings 128

and
ϕ′ (t) = a cos(t) + cos2 (t) − sin2 (t) = 2 cos2 (t) + a cos(t) − 1 = 0
1 √ 
=⇒ cos(t) = −a ± a2 + 8 .
4
We get two values
√ √
1 3 3 3
cos(t1 ) = =⇒ x1,2 = , y1,2 = ± =⇒ Im(z1,2 ) = ± ,
2 2 2 2
and
cos(t3 ) = −1 =⇒ x3 = 0, y3 = 0 =⇒ Im(z3 ) = 0.
√
3 1 √ 
Therefore, maxD f (z) = , achieved at 3 + 3i and
√ 2 2
3 1 √ 
minD f (z) = − , achieved at 3 − 3i .
2 2
Example 3.2.3.

Let
f (z) = ze−z
and
D = {z : | Im(z)| < Re(z), |z| > 1}.
Find
max |f (z)|
z∈D

and point(s), where it is achieved.

Solution. We cannot apply maximum modulus principle directly to the

unbounded domain D, therefore we consider a truncated domain DR :=
{z ∈ D : |z| < R}, observing that f (z) → 0 as |z| → ∞ while z ∈ D.
Therefore maxz∈D |f (z)| is achieved on the boundary of D.
Consider boundary of D, consisting of two rays {z : z = (1 ± i)t, t > √12 }
and an arc {z : z = eiθ , − π4 < θ < π4 }.
−t∓it
√ −t
Then on the rays
√ −1f (z) = (1 ± i)te and |f (z)| = ϕ(t) := 2te and
√ ′
maxt≥1/ 2 ϕ(t) = 2e achieved as t = 1 (one can check that ϕ (t) > 0 in
the corners t = √12 and they are not the points of the maximum).
Chapter 3. Analytic Functions as Mappings 129

Meanwhile, on the arc |f (z)| = e− cos(θ) and the maximum achieved in

the corners.

Therefore
√
max |f (z)| = 2e−1
z∈D

achieved as z = (1 ± i).

Example 3.2.4.

Let
1
f (z) =
1 + z2
and
D = {z : | Im(z)| ≤ Re(z)}.
Find
max Im(f (z)) and min Im(f (z))
z∈D z∈D

and point(s), where it is achieved.

Solution. Observe that f (z) → 0 as D ∋ z → ∞ and therefore maximum
and minimum are achieved at z ∈ D (unless they are 0). By maximum
principle both are achieved on the boundary.
Consider rays y = ±x, x > 0, there
1 1 (1 ∓ 2ix2 )
f (z) = = =
(1 ± i)2 x2 + 1 ±2ix2 + 1 4x4 + 1
2x2
=⇒ Im(f (z)) = ∓ =: ∓φ(x).
4x4 + 1
Then
4x(1 − 4x4 ) 1
φ′ (x) = = 0 =⇒ x = √ .
(4x4 + 1)2 2
Chapter 3. Analytic Functions as Mappings 130

We need to consider only x = √1 . Finally, φ(x) = 0 as x = 0. So,

2

1 1
max Im(f (z)) = √ and min Im(f (z)) = − √
z∈D 2 z∈D 2

achieved at z = √1 (1 − i) and z = √1 (1 + i) respectively.

2 2

3.2.2 Schwarz’s Lemma

One of the best-known applications of the maximum-modulus principle is
the following result, known for historical reasons as Schwarz’s Lemma:

Theorem 3.2.7 (Schwarz’s Lemma; Theorem 2 in the Textbook). Let f (z)

be analytic in the unit disk D = {z : |z| < 1}. Assume that f (0) = 0 and
that |f (z)| ≤ 1 ∀z ∈ D. Then

(i) The following inequality holds:

|f (z)| ≤ |z| ∀z ∈ D. (3.2.2)

(ii) Equality can hold for some z ̸= 0 only if f (z) = λz with a constant
λ : |λ| = 1.

Proof. (i) Since f (0) = 0, we know that g(z) = f (z)/z is also analytic
in D. For |z| = r, we have |g(z)| = |f r(z) ≤ 1r .
By the maximum-modulus principle, the inequality |g(z)| ≤ 1/r is true
for |z| < r as well. Since r can be arbitrarily close to 1, we must conclude
that |g(z)| ≤ 1 if |z| < 1; thus, |f (z)| ≤ |z| in D.

(ii) Furthermore, if |f (z0 )| = |z0 | for some z0 ̸= 0, then 1 = |g(z0 )|;

consequently, |g(z)| has an interior maximum, which is possible only if g(z)
is a constant λ with |λ| = 1. Then f (z) = λz.

Remark 3.2.3. Schwarz’s Lemma does not hold for functions of real variable.
For example, f (x) = x22x+1 satisfies |u(x)| ≤ 1 for all x ∈ R and u is smooth,
u(0) = 0 but u(x) > |x| for x ∈ (−1, 1).
Chapter 3. Analytic Functions as Mappings 131

3.2.3 Mean Value Theorem

Theorem 3.2.8 (Mean-Value Theorem). Let f (z) be analytic function in the
disk {z : |z − z0 | < r} and continuous in the closed disk {z : |z − z0 | <≤ r}.
Then
Z 2π
1
f (z0 ) = f (z0 + reit ) dt. (3.2.3)
2π 0

where γ = {z : |z − z0 | = r}.

Proof. Cauchy’s Formula gives us

Z
1 f (ζ) dζ
f (z0 ) =
2πi γ ζ − z0

when γ is a circle and z0 is inside γ. Plugging ζ = z0 + reit with 0 ≤ t ≤ 2π

and dζ = ireit dt, we prove (3.2.3).
Taking a real part of (3.2.3) we arrive to

Theorem 3.2.9 (Mean-Value Theorem for harmonic functions). Let u(x, y)

be harmonic function in the disk {z : |z − z0 | < r} and continuous in the
closed disk {z : |z − z0 | ≤ r}. Then
Z 2π
1
u(z0 ) = u(z0 + reit ) dt. (3.2.4)
2π 0

where γ = {z : |z − z0 | = r}.

Example 3.2.5.

Let u be a harmonic function in the disk

D = {(x, y) : x2 + y 2 ≤ 9}, equal to

| sin(θ)|
g(θ) =
2 + cos(θ)

on its boundary, where θ is a polar angle. Find u(0).

Chapter 3. Analytic Functions as Mappings 132

Solution.
2π π
| sin(θ)| dθ | sin(θ)| dθ
Z Z
1 1
u(0) = =
2π 0 2 + cos(θ) π 0 2 + cos(θ)
1 θ=π 1
= − ln(2 + cos(θ)) θ=0
= ln(3).
π π

3.3 Linear Fractional Transformations

3.3.1 Introduction
In these couple of lectures we consider linear fractional transformations
(a.k.a. Möbius transformations) which are one of the most basic examples
of conformal transformations, which we consider after this. Here the in-
troduction of the extended complex plane C b := C ∪ {∞} will be especially
fruitfull.
b →C
Definition 3.3.1. Linear fractional transformation is a map T : C b
defined by
az + b
T (z) := (3.3.1)
cz + d
with a, b, c, d ∈ C, satisfying (ad − bc) ̸= 0.

Remark 3.3.1. Assumption ∆ := ad − bc = ̸ 0 is essential; otherwise (a, b)]

would be proportional to (b, c) and T (z) would be a constant, equal to the
coefficient of proportionality.

Theorem 3.3.1. (i) Composition of two linear fractional transformations

is a linear fractional transformation as well.

(ii) Inverse to a linear fractional transformation is also a linear fractional

transformation.

(iii) Identical transformation is also a linear fractional transformation.

Chapter 3. Analytic Functions as Mappings 133

a1 z + b 1 a2 w + b 2
Proof. (i) Let w = T1 (z) = and T2 (w) = . Then
c1 z + d 1 c2 w + d 2

a1 z + b 1
a2 + b2
c1 z + d1 a2 (a1 z + b1 ) + b2 (c1 z + d1 )
(T2 ◦ T1 )(z) = =
a1 z + b 1 c2 (a1 z + b1 ) + d2 (c1 z + d1 )
c2 + d2
c1 z + d 1
(a2 a1 + b2 c1 )z + (a2 b1 + b2 d1 ) a3 z + b 3
= = (3.3.2)
(c2 a1 + d2 c1 ) + (c2 b1 + d2 d1 ) c3 z + d3

is a linear fractional transformation.

az + b
(ii) Let w = T (z) = . Then
cz + d

w(cz + d) = (az + b) =⇒ (cw − a)z = (−wd + b)

dw − b
=⇒ z = T −1 (w) = (3.3.3)
a − cw
is a linear fractional transformation.

(iii) Finally, T (z) = z is a linear fractional transformation with

a = 1, b = 0, c = 0, d = 1.

Remark 3.3.2. Observe that to linear fractional transformation! T with

a b
coefficients a, b, c, d one can assign a 2 × 2-matrix M [T ] := .
c d

(a) Then (3.3.2) shows that to T3 := T2 ◦ T1 is assigned matrix

!
a2 a1 + b2 c1 a2 b1 + b2 d1
M [T3 ] = = M [T2 ]M [T1 ].
c 2 a1 + d 2 c 1 c 2 b 1 + d 2 d 1

(b) Further, (3.3.3) shows that to T −1 is assigned matrix

!
d −b
= ∆M [T ]−1 with ∆ = ad − bc.
−c a
Chapter 3. Analytic Functions as Mappings 134

!
1 0
(c) Finally, to T (z) = z is assigned matrix I = .
0 1
(d) However, the same transform T corresponds to the whole bunch of
quadruplets {a, b, c, d}, different by a common factor λ ̸= 0. Requiring that

∆ := ad − bc = 1 (3.3.4)

would ensure that to each transform T correspond exactly two such quadru-
plets, different by factor −1, and thus two matrices M [T ] with determinant
1, different by a factor −1.
(e) Under this condition to T −1 is assigned matrix M [T ]−1 .
(f) On the other hand, inverse map M 7→ T (where M is any 2×2-matrix
with determinant 1) is single valued and according to Part 1
(g) T [M2 M1 ] = T [M2 ] ◦ T [M1 ],
(h) T [M −1 ] = (T [M ])−1 ,
(i) and T [±I] = id is an identical map z → z.

We can say few mathematical jargon words showing how learned we are,
but we are not in Math Specialist class.
Theorem 3.3.2. A linear fractional transformation
a−z
T (z) = λ with a, λ ∈ C, |a| < 1 and |λ| = 1 (3.3.5)
1 − az
maps a unit disk D := {z : |z| < 1} onto itself.
Proof. Observe that if T is defined by (3.3.5) then
|a − z|2 |a|2 − 2 Re(az) + |z|2
|T (z)|2 = =
1 − az|2 1 − 2 Re(az) + |a|2 |z|2
and |T (z)| < 1 if and only if

1 + |a|2 |z|2 > |a|2 + |z|2 ⇐⇒ (1 − |a|2 )(1 − |z|2 ) > 0

which is true iff |z| < 1 that means z ∈ D. Therefore, it maps D onto
itself.
Chapter 3. Analytic Functions as Mappings 135

Theorem 3.3.3. (i) Composition of two transformations of the type

(3.3.5) is also a transformation of this type.
(ii) Inverse of transformation (3.3.5) is also a transformation of this type.
(iii) Identity transformation is also a transformation of type (3.3.5).
z − a1 w − a2
Proof. (i) Let w = T1 (z) = λ1 , T2 (w) = λ2 ; then
1 − a1 z 1 − a2 w
z−a1
λ1 1−a1z
− a2 λ1 (z − a1 ) − a2 (1 − a1 z)
(T2 ◦ T1 )(z) = λ2 z−a1 = λ2
1− λ1 1−a 1z
a2 (1 − a1 z) − λ1 (z − a1 )a2
z(λ1 + a2 a1 ) − (λ1 a1 + a2 ) z − a3
= λ2 = λ3
(1 + λ1 a1 a2 ) − (a1 + λ1 a1 a2 )z 1 − a3 z
λ1 + a2 a1 λ 1 a1 + a2
with λ3 = λ2 , a3 = (Check calculations! )
1 + λ1 a1 a2 1 + λ1 a1 a2
One can check easily that |λ3 | = 1 and |a3 | < 1:

|λ1 + a2 a1 |2 |λ1 |2 + 2 Re(λ1 a2 a1 ) + |a2 |2 |a1 |2

|λ3 |2 = |λ2 |2 = =1
|1 + λ1 a1 a2 |2 1 + 2 Re(λ1 a1 a2 ) + |λ1 |2 |a1 |2 |a2 |2

because |λ1 | = |λ2 | = 1.

|λ1 a1 + a2 |2 |λ1 |2 |a1 |2 + 2 Re(λ1 a1 a2 ) + |a2 |2

|a3 |2 = =
|1 + λ1 a1 a2 |2 1 + 2 Re(λ1 a1 a2 |) + |λ1 |2 |a1 |2 |a2 |2
|a1 |2 + 2 Re(λ1 a1 a2 ) + |a2 |2
= <1
1 + 2 Re(λ1 a1 a2 |) + |a1 |2 |a2 |2

because |a1 |2 + |a2 |2 < 1 + |a1 |2 |a2 |2 (since |a1 | < 1, |a2 | < 1).
z−a w − a2
(ii) If w = λ then z = λ2 with λ2 = λ and a3 = −λa,
1 − az 1 − a2 w
|λ2 | = 1 and |a2 | < 1 (Check it! ).
(iii) Finally w = z is given by (3.3.5) with λ = 1, a = 0.

Theorem 3.3.4. Let |z1 | < 1, |z2 | = 1 and |w1 | < 1, |w2 | = 1. Then
exists exactly one transformation L of type (3.3.5) such that L(z1 ) = a1 and
L(z2 ) = w2 .
Chapter 3. Analytic Functions as Mappings 136

Proof. Let us construct transformation T of the type (3.3.5) such that

T (z2 ) = 0 and T (z2 ) = 1. It is uniquely given by
  
1 − z 1 z2 z − z1
T (z) = . (3.3.6)
z2 − z1 1 − z1z

Then uniquely L = S −1 ◦ T with S(w) given by the same formula (3.3.5)

with w, w1 and w2 instead of z, z1 and z2 . Finally
     
1 − z 1 z2 z − z1 1 − w 1 w2 w − w1
= . (3.3.7)
z2 − z1 1 − z1z w2 − w1 1 − w1 w

Theorem 3.3.5. Any linear fractional transformation T , mapping D onto

D is transformation of the type (3.3.5).
Proof. There are z1 : |z1 | < 1 such that T (z1 ) = 0 and z2 : |z2 | = 1 such that
T (z2 ) = 1. Let S be a transformation of type (3.3.5) such that S(z1 ) = 0
and S(z2 ) = 1. Consider linear fractional transformation L = T ◦ S −1 , it
az + b
maps 0 onto 0, and therefore L(z) = with b = 0; it can be rewritten
cz + d
z
as L(z) = .
c1 z + d 1
If it maps D onto D then |z| = 1 =⇒ |c1 z + d1 | = 1 which is possible
if either |c1 | = 1, d1 = 0 or c1 = 0, |d1 | = 1. In the former case L(z) = c−1 1
(constant, which is impossible), in the latter L(z) = d−1 1 z.
−1
Finally, L(1) = 1 =⇒ d1 = 1 =⇒ L = T ◦ S = id =⇒ T = S of
type (3.3.5).
Remark 3.3.3. In the next Section 3.4, this proof will be given not just for
linear fractional transformations, but for more general conformal maps.

3.3.2 Fixed Points and Triples to Triples

Definition 3.3.2. A fixed point of the map z → T (z) is a point such that

T (z) = z. (3.3.8)

Theorem 3.3.6. A fractional linear transformation that is not identically

equal to z has at most two distinct fixed points.
Chapter 3. Analytic Functions as Mappings 137

az + b
Proof. z is a fixed point of fractional linear transformation T (z) =
cz + d
with c ̸= 0 if and only if z is a root of the quadratic equation

cz 2 + (d − a)z − b = 0

̸ 0 consider by
which has at most two distinct roots. Case c = 0, d =
yourself.
Corollary 3.3.7. Let T and S be two fractional linear transformations
that are equal at three distinct points: T (zj ) = S(zj ) for j = 1, 2, 3. Then
T (z) = S(z) for all z.
Proof. If T (zj ) = S(zj ) then (S −1 ◦ T )(zj ) = zj for j = 1, 2, 3 and since
S −1 ◦ T is also a fractional linear transformation we conclude that (S −1 ◦
T )(z) = z for all z, and T (z) = S(z) for all z.
Theorem 3.3.8. Let z1 , z2 and z3 be three distinct complex numbers, and
let w1 , w2 and w3 be also three distinct complex numbers. Then there exists
a unique linear fractional transformation L with L(zj ) = wj for j = 1, 2, 3.
Proof. Let
  
z − z1 z2 − z3
T (z) = . (3.3.9)
z − z3 z2 − z1
Then T (z1 ) = 0, T (z2 ) = 1 and T (z3 ) = ∞. Let us define S(w) by the
same formula (3.3.9) with w, w1 , w2 and w3 instead of z, z1 , z2 and z3 . Then
S(w1 ) = 0, S(w2 ) = 1 and S(w3 ) = ∞. Then L = S −1 ◦ T is a required
transformation. Due to Corollary 3.3.7 it is unique.
Finally,
     
z − z1 z2 − z3 w − w1 w2 − w3
= . (3.3.10)
z − z3 z2 − z1 w − w3 w2 − w1

Remark 3.3.4. We can extend these arguments to the case when one of
2 −z3
z1 , z2 , z3 or/and w1 , w2 , w3 is ∞. Indeed, if z1 = ∞, we can take T (z) = zz−z 3
;
z−z1 z−z1
if z2 = ∞, then T (z) = z−z3 ; and if z3 = ∞, then T (z) = z2 −z1 .
Example 3.3.1. Find a linear fractional transformation, that sends 0, 1, 2 to
−1, 0, 4 respectively.
Chapter 3. Analytic Functions as Mappings 138

Solution. Using (3.3.10) we have

z − 0
1 − 2
z
=
2−z 1−0 z−2
w + 1
0 − 4
w+1 8z − 8
= = −4 =⇒ w = .
w−4 0+1 w−4 −3z + 8

Example 3.3.2. Find a linear fractional transformation, mapping D onto D,

that sends 21 , 1 to 12 , −1 respectively.
Solution. Using ((7))3.3.7) we have

1 − 12 z − 12
  
2z − 1
1 z =
1− 2 1− 2 z−2
1  1
w−2

1+ 2 2w − 1 5z − 4
1 w =− =⇒ w = .
−1 − 2 1− 2 2−w 4z − 5

3.3.3 Lines and Circles

Theorem 3.3.9. A linear fractional transformation maps each circle onto
another circle or onto a straight line and each straight line onto another
straight line or a circle.
Proof. First, if T (z) = az + b with a ̸= 0 (so T is a linear transformation),
then T maps circles and straight lines to the circles and straight lines
respectively: the circle C = {z : |z−z0 | = r} is transformed to the circle C ′ =
{w : |w−w0 | = |a|r} with w0 = az0 +b and the straight line L = {z : Re(Az+
′
B) = 0}
is transformed to the straight line L = {w : Re (A/a)w + B −
(Ab/a) = 0}.
Thus, consider with c ̸= 0
az + b 1 h −∆ i
T (z) = = +a
cz + d c cz + d
and we see that T is a composition of several simple transformations: T =
1
W ◦ V ◦ U with U (z) = cz + d, V (w) = , and W (ζ) = 1c −∆ζ + a .


w
Chapter 3. Analytic Functions as Mappings 139

Since U and W are linear transformations, which map circles onto circles
and straight lines onto straight lines, it is sufficient to prove the declared
property for inversion map V . The equation

α(x2 + y 2 ) + βx + γy = δ

with real α, β, γ, δ where (α, β, γ) ̸= 0 describes either a circle (if α ̸= 0 and

β 2 + γ 2 + 4αδ > 0) or a straight line (if α = 0). Since
1 x − yi
= 2 =: u + iv = w
z x + y2
dividing
α(x2 + y 2 ) + βx + γy = δ
by x2 + y 2 we get
x y δ
α+β +γ 2 = 2 .
x2 +y 2 x +y 2 x + y2
x −y
Plugging u = x2 +y 2
, v= x2 +y 2
we get

α + βu − γv = δ(u2 + v 2 )

which also describes a circle (if δ ̸= 0) or a straight line (if δ = 0).

Remark 3.3.5. One can see easily that the inversion map transforms

(a) a straight line or a circle passing through the origin into a straight
line,

(a) z (b) w

(b) a straight line or a circle not passing through the origin into a circle.
Chapter 3. Analytic Functions as Mappings 140

(a) z (b) w

3.4 Conformal Mapping

3.4.1 Introduction
Conformal maps are functions that preserve angles between curves (more
precisely: between tangent lines to the curves in the points of intersection).
Let γ be the range of a smooth curve z(t), passing through the point
z0 = z(t0 ), a < t0 < b:

γ : z = z(t), a < t < b, z(t0 ) = z0 .

The curve γ has a tangent vector z ′ (t0 ) at z0 , and we suppose that z ′ (t0 ) ̸= 0.
Remark 3.4.1. If we consider another parametrization of the same curve γ,
z = z(t(s)), a′ < s < b′ , then we will get another tangent vector, but it will
be proportional with a positive coefficient if we assume that the direction in
which curve is passed is the same for both parameterizations, that is, t(s) is
increasing function of s.
What happens to the curve γ when we apply an analytic function f (z)
to it? It is transformed into a new curve Γ in the w-plane; Γ is given by

Γ : w(t) = f (z(t)), a < t < b,

Due to chain rule, the tangent vector to Γ at w0 = w(t0 ) = f (z0 ):

w′ (t0 ) = f ′ (z0 )z ′ (t0 ). (3.4.1)

Chapter 3. Analytic Functions as Mappings 141

In particular,

|w′ (t0 )| = |f ′ (z0 )| · |z ′ (t0 )|. (3.4.2)

and
arg(w′ (t0 )) = arg(f ′ (z0 )) + arg(z ′ (t0 )). (3.4.3)

Therefore under assumption

f ′ (z0 ) ̸= 0 (3.4.4)

we conclude that

(a) the tangent vector is scaled in length by a factor |f ′ (z0 )|, and

(b) the tangent vector is rotated by an angle ψ0 = arg(f ′ (z0 )).

Definition 3.4.1. Let γ1 and γ2 be two smooth oriented curves that intersect
at the point z0 = z1 (t0 ) = z2 (s0 ). Then the angle between γ1 and γ2 at z0
is the angle θ measured counter-clockwise from the tangent vector z1′ (t0 ) to
the tangent vector z2′ (s0 ), provided neither of these tangent vectors is zero:

θ = arg(z2′ (s0 )) − arg(z1′ (t0 )). (3.4.5)

γ2

z2′ (s0 )
z0 θ

γ1 z1′ (t0 )

Remark 3.4.2. This definition of the angle between two curves breaks down
if one of the vectors z1′ (t0 ) or z2′ (s0 ) is 0.
Chapter 3. Analytic Functions as Mappings 142

Definition 3.4.2. Suppose now that ϕ(z) is a function, perhaps not analytic,
defined in the disk {z : |z − z0 | < r and satisfying ϕ(z) ̸= ϕ(z0 ) in the
punctured disk {z : < |z − z0 | < r}.
Then map ϕ is conformal at z0 if, whenever two oriented curves γ1 and
γ2 meet at z0 , the angle from Γ1 = ϕ(γ1 ) to Γ2 = ϕ(γ2 ) is equal to the angle
from γ1 to γ2 .
Remark 3.4.3. Mercator projection in geography also preserves angles (but
stretches near polar domains more than near equatorial ones).
Theorem 3.4.1. If f is analytic near z0 and f ′ (z0 ) ̸= 0 then f is conformal
at z0 .
Proof. Indeed, if γ1 and γ2 intersect at z0 then Γ1 and Γ2 intersect at
w0 = f (z0 ) and equation (3.4.3)

arg(w′ (t0 ) = arg(f ′ (z0 )) + arg(z ′ (t0 )).

shows that the angle from Γ1 to Γ2 is equal to

arg(f ′ (z0 )) + arg(z2′ (s0 )) − arg(f ′ (z0 )) + arg(z1′ (t0 ))

   

= arg(z2′ (s0 )) − arg(z1′ (t0 )),

which is an angle from γ1 to γ2 .

Remark 3.4.4. (a) If f ′ (z0 ) = 0 then f is not conformal. Indeed, assume
that f (z)−f (z0 ) has 0 of multiplicity m ≥ 2 at z0 . Then f (z)−f (z0 ) = g(z)m
with g(z) also analytic, g(z0 ) = 0 and g ′ (z0 ) ̸= 0. Then g is conformal at
z0 but since arg(ζ m ) = m arg(ζ) we conclude that at w0 the angle from
Γ1 = f (γ1 ) to Γ2 = f (γ2 ) equals to the angle from γ1 to γ2 at z0 , multiplied
by m.
(b) However in the framework of 1 f is not one-to-one (it is m-to-1)
which implies Theorem 3.4.2 below.
(c) Since the level lines of Re(z) and Im(z) are orthogonal, we get another
proof that level lines of Re(f (z)) and Im(f (z)) for any conformal f (that is
analytic and f ′ ̸= 0) are also orthogonal.
(d) Note that z 7→ f (z) is anti-conformal : it preserves a magnitude of
the angle but changes it sign.
Chapter 3. Analytic Functions as Mappings 143

(e) One can prove the converse to Theorem 3.4.1: if f is conformal, then
it is analytic and f ′ ̸= 0.

(f) One can wonder, if there is something like conformal mappings in

higher dimensions. The answer is disappointing: there are but there are very
few of them; those are shifts, rotations, scalings x → λx, mirror reflections,
inversions x → |x|x2 and their compositions.
The more distant relatives are however plentiful.

Theorem 3.4.2. If f is analytic on a domain D and one-to-one, then f is

conformal at all points of domain D.

The following assertion is important but even the rigorous statement is

beyond our reach. You can find plenty counter-examples but they either are
somehow pathological, or somehow could be fixed. So it is not a theorem,
but more vague principle.

Extension of Conformal Map to the Boundary Principle. Let f be

a conformal map of domain G1 onto domain G2 . Then f maps the boundary
Γ1 of D1 onto boundary Γ2 of G2 .

3.4.2 Conformal Mappings of Unit Disk onto itself

Theorem 3.4.3. All conformal maps of D = {z : |z| < 1} onto itself are of
the form
z−a
T (z) = λ a, λ ∈ C : |λ| = 1, |a| < 1. (3.4.6)
1 − az
Proof. We already know that (3.4.6) gives us a conformal map of D onto
D and we need to prove that there are no other maps. Let S : D → D be
some conformal map. Since it is one-to-one correspondence, it maps 0 to
some point a. Consider map (3.4.6) with λ = 1. It maps D → D and a to 0.
Therefore L = T −1 ◦ S also maps D → D, and 0 to 0.
Since L maps its D → D, it also maps boundary Γ = {z : |z| = 1}
onto itself: |z| = 1 =⇒ |L(z)| = 1. Therefore, due to Schwartz’s Lemma
|L(z)| ≤ |z| for all z : |z| < 1 and, moreover, either

(a) L(z) = λz for some λ, or

Chapter 3. Analytic Functions as Mappings 144

(b) |L(z)| < |z| for for all z : 0 < |z| < 1.

But case (b) is impossible. Indeed, we can apply the same arguments to
inverse map L−1 and conculde that |L−1 (w)| ≤ |w| for all w : |w| < 1 and
plugging w = L(z) we conclude that |z| ≤ |L(z)| < |z| for all z : 0 < |z| < 1.
Contradiction.
Therefore
λz − a z−b
(T −1 ◦ S)(z) = λz =⇒ S(z) = T (λz) = =λ
1 − aλz 1 − bz

with b = λa is indeed map of type (3.4.6).

Corollary 3.4.4. Let z1 , w1 ∈ D and z2 , w2 ∈ Γ. Then there exists exactly

one conformal map T of D onto itself, such that T (z1 ) = w1 and T (z2 ) = w2 .

Proof. Indeed, any conformal map of D onto D is a map of type (3.4.6) and
for them this statement has been proven.

3.4.3 Examples
Example 3.4.1. w = z 2 , it maps the right half-plane {z : − π2 < arg(z) < π2 }
conformally onto plane with a cut {w : − π < arg(w) < π}. What happens
with level lines of z? {z = x + bi} become parabolas {w = (x2 − b2 ) + 2bxi}
and {z = a + yi} become also parabolas {w = (−y 2 + a2 ) + 2ayi}:

(a) z (b) w
Chapter 3. Analytic Functions as Mappings 145

(a) z (b) w

Example 3.4.2. w = ez ; it maps a stripe {z : | Im(z)| < π} onto a plane

with cut {w : − π < arg(w) < π}. Since z = x + yi maps onto w =
ex (cos(y) + i sin(y)) we conclude that y = b become rays and x = a became
circles:
√
Example 3.4.3. w = z, it maps the plane with a cut {z : −π < arg(z) < π}
conformally onto right half-plane {w : − π2 < arg(w) < π2 }, or a “double
plane” (Riemannian surface, we leave exact definitions to math specialist
class) {z : − 2π < arg(z) < 2π} conformally onto a plane with a cut
{w : − π < arg(w) < π}. Here we look the other way: w = u + iv and
z = u2 − v 2 + 2uvi, so we draw level lines of u2 − v 2 and 2uv, which are two
families of hyperbolas.

−1

−2
−2 −1 0 1 2
Chapter 3. Analytic Functions as Mappings 146

−1

−2
−2 −1 0 1 2
√
Example 3.4.4. w = z, it maps the plane with a cut {z : −π < arg(z) < π}
3

conformally onto sector {w : − π3 < arg(w) < π3 }, or a “tripple plane”

(Riemannian surface, we leave exact definitions to math specialist class)
{z : − 3π < arg(z) < 3π} conformally onto a plane with a cut {w : − π <
3 2
arg(w) <3.4.5.
π}. Here 1 look at level lines: w = u + iv and z = (u − 3uv ) +
we
Example
2
w = z
, it maps the punctured plane {z : |z| > 0} conformally
(3u − v 3punctured
ontov the )i, so we draw
planelevel
{w :lines
|w| of
3
> u0};−we
2
3uvknow (3u2the
andthat 3
v − vstraight
)i. lines
are mapped onto circles, in particular, lines {z = x + bi} onto circles
{|w +(2b)−1 i| < |2b|−1 } and {z = a+yi} onto circles {|w −(2a)−1 | < |2a|−1 }.

(a) z (b) w
Chapter 3. Analytic Functions as Mappings 147

q p L′

(a) z (b) w

 
w−p
Example 3.4.6. z = log w−q
̸ q, it is defined on the twice punctured
with p =
plane {w : |w − q| > 0, |w − q| > 0}. Consider inverse map; then we get
Circles of Appolonius. Indeed Re(z) = a means that |w−p| |w−q|
= ea and we get a
family of blue circles, while Im(z) = b means that arg(z − p) − arg(z − q) = b
and we get a family of arcs of red circles between p and q (see W2L1).
Example 3.4.7. w = cos(z). It maps conformally stripe {z : 0 < Re(z) < π}
onto plane with cuts C \ ((−∞, −1] ∪ [1, ∞)) (see W3L2):

(a) Vertical lines {z : x = Re(z) = const, y = Im(z)} are mapped to

w = u + iv with v = − sin(x) sinh(y), u = cos(x) cosh(y) which are
branches of hyperbolas satisfying
u2 v2
− = 1,
A2 B 2
with A = | cos(x)|, B = sin(x).

(b) Horizontal segments {z : 0 < x = Re(z) < π, y = Im(z) = const} are

mapped to w = u + iv with v = − sin(x) sinh(y), u = cos(x) cosh(y)
which are halves of ellipses satisfying
Chapter 3. Analytic Functions as Mappings 148

(a) z (b) w = cos(z)

u2 v 2
+ 2 =1
a2 b
with a = cosh(y), b = | sinh(y)|.

Example 3.4.8. w = cos(z). It maps conformally half-stripe {z : 0 < Re(z) <

2π, Im(z) > 0} onto plane with a cut C \ [−1, ∞).

(a) Vertical half-lines {z : x = Re(z) = const, y = Im(z)} are mapped to

w = u + iv with v = − sin(x) sinh(y), u = cos(x) cosh(y) which are
half-branches of the same hyperbolas as in Example 3.4.7.
(b) Horizontal segments {z : 0 < x = Re(z) < 2π, y = Im(z) = const} are
mapped to w = u + iv with v = − sin(x) sinh(y), u = cos(x) cosh(y)
which are the same ellipses as in Example 3.4.7.

Example 3.4.9 (Zhukovsky transform). w = 12 z + z1 . It maps D onto

C \ [−1, 1]. Indeed, circles z = reit with fixed r and t ∈ [0, 2pi)are mapped
onto confocal ellipses u = A cos(t), v = −B sin(t) with A = 12 (r + r−1 ),
Chapter 3. Analytic Functions as Mappings 149

(a) z (b) w = cos(z)

B = 12 (r−1 − r) and segments z = e−s eit with fixed t and s ∈ (−∞, 0]

are mapped onto confocal hyperbolae u = a cosh(s), v = −b sinh(s) with
a = cos(t), b = sin(t).

3.5 The Riemann Mapping Theorem and

Schwarz-Christoffel Transformations
3.5.1 The Riemann Mapping Theorem
The following theorem is one of the most important theorems in the theory
of conformal mappings:

Theorem 3.5.1 (The Riemann Mapping Theorem). Let G be a simply-

connected domain in the extended complex plane C b with the boundary, con-
taining at least 2 disjoint points. Then there exists a conformal map from G
onto D.

Remark 3.5.1. (a) The proof of this theorem is completely out of our
reach. Actually there are many proofs and none is constructive.

(b) If the boundary of simply-connected domain contains at least two

points, it contains continuum of them.

(c) We discuss which domains fit conditions of The Riemann Mapping

Theorem.
Chapter 3. Analytic Functions as Mappings 150

(a) z (b) w = (z + z −1 )/2

While The Riemann Mapping Theorem says how domains are mapped,
it does not discuss the mapping of their boundaries. This is covered by
Carathéodory’s theorem below. But first we need the following

Definition 3.5.1. Jordan curve is a simple (without self-intersections) closed

continuous curve (that means that there exists a one-to-one continuous map
from the unit circle {z : |z| = 1} onto Jordan curve.

Jordan curve has an important property, which looks obvious but in fact
is very non-trivial and deep:

Theorem 3.5.3 (Jordan Curve Theorem). Let γ be a Jordan curve in the

plane R2 . Then its complement, R2 \ γ, consists of exactly two connected
components. One of these components is bounded (the interior) and the
other is unbounded (the exterior), and the curve γ is the boundary of each
component.

Again, we cannot prove it in this class.

Chapter 3. Analytic Functions as Mappings 151

Theorem 3.5.4 (Carathéodory’s theorem). If f maps the open unit disk D

conformally onto a bounded domain G in C, then f has a continuous one-
to-one extension to the closed unit disk if and only if its boundary γ = ∂G
is a Jordan curve.

And again, we cannot prove it in this class.

3.5.2 Mirror Continuation

Theorem 3.5.5. Let f : D1 → D2 be a conformal map. Assume that their
boundaries Γ1 and γ2 respectively, contain straight segments or circular arcs
γ1 and γ2 correspondingly, and f : γ1 → γ2 .
Then f could be extended throughout γ1 by formula

f˜(z) = Mγ2 ◦ f ◦ Mγ1 (3.5.1)

where Mγ is defined on the next frame.

Proof. It follows from:

(a) f : γ1 → γ2 ;

(b) Mγ is anticonformal map, that is it preserves magnitude of angles, but

changes orientation.

γ
γ z
z
Mγ (z)
Mγ (z)

O
(a) (b)

Figure 3.1: On (a) z and Mγ (z) are on the equal distances from γ; if γ ⊂ R
then Mγ (z) = z. On (b) O is a center of the circle, R its radius and
R2
|OMγ (z)| = R2 |Oz|−1 ; if γ ⊂ {z : |z| = R} then Mγ (z) = .
z

Corollary 3.5.6. Let D1 = {z : r1 < |z| < R1 } and D2 = {w : r2 < |w| <
R2 } and f : D1 → D2 conformal one-to-one map of D1 onto D2 . Then
r1 r2
(i) = ;
R1 R2
Chapter 3. Analytic Functions as Mappings 152

µ r2 R2
(ii) Either f (z) = λ|z| or f (z) = with |λ| = = and |µ| = r1 R2 =
z r1 R1
r2 R1 .
Proof. Assume that f maps inner circle onto inner circle and outer circle
onto outer circle. Otherwise we replace f (z) by f (z)−1 , r2 by R2−1 and R2
by r2−1 .
Using mirror reflection we conclude that f maps conformally punctured
disk {z : 0 < |z| < R1 } onto punctured disk {w : 0 < |w| < R2 } and therefore
disk onto disk, and f (0) = 0. Then f (z) = λz and |λ| = R1−1 R2 :

(a) z (b) w

Since f maps inner circle onto inner circle, it should be also |λ| = r1−1 r2 ,
and it implies both (i) and (ii).

3.5.3 Examples
Example 3.5.1 (Upper half-plane to unit circle). (a) Consider the following
conformal map from H+ := {z : Im(z) > 0} onto D = {w : |w| < 1}.
According to Carathéodory’s theorem it’s boundary, {z : Im(z) = 0} should
be mapped onto {w : |w| = 1} and one of the way to achieve it is to take
z−i
w = f (z) =
z+i
we took “−” in the numerator and “+” in the denominator because otherwise
it would be singular at i ∈ H+ . And we can see that if z = x + yi with
y > 0, then |w| < 1 (check it!).
Further, since it is a linear fractional transformation, it must transform
lines {z : Im(z) = b} into either circles or straight lines, but those must be
circles, passing through 1 (when |z| → ∞, w → 1).
Chapter 3. Analytic Functions as Mappings 153

And indeed:
x + (y − 1)i [x + (y − 1)i][x − (y + 1)i]
w = f (z) = =
x + (y + 1)i x2 + (y + 1)2

(a) z (b) w

(b) Also, the general conformal map H+ → D is given by

z−a
w=λ , |λ| = 1, Im(a) > 0.
z−a
(c) Inverse map then
aw − λa
z= , |λ| = 1, Im(a) > 0.
w−λ
Example 3.5.2. General conformal map H+ → H+ is
c
w =a+ , a, b, c ∈ R, c > 0.
z−b
It does not include linear map w = cz + a with a, c ∈ R, c > 0 which also
maps H+ onto H+ .
Example 3.5.3. w = z β maps sector {z : | arg(z)| < α}, 0 ̸= β ∈ R onto
sector {w : | arg(w)| < βα}, α ≤ π, |β|α ≤ π. If β > 0 it maps 0 onto 0.
The general map from the first sector onto the second one does not
preserve 0. F.e. w = f (z) := z 2 maps Q = {z : Im(z) > 0, Re(z) > 0}
onto H+ , and the general conformal map Q → Q is f −1 ◦ T ◦ f with T from
Example 3.5.2.
Other examples were given before, see also pp 226–227 of the Textbook
Chapter 3. Analytic Functions as Mappings 154

3.5.4 Schwarz-Christoffel Transformations

A Schwarz-Christoffel transformation is an analytic conformal mapping of
the upper half-plane onto a polygon.
The key to understanding it is the examination of the behaviour at the
point x0 of the function f given by

f (z) = A(z − x0 )β + B

where x0 and β are real numbers, 0 < β < 2, and 0 ̸= A and B are complex
numbers.
Basically it is a composition of the linear function w = Aω + B which
shifts and rotates, ω = ζ β which maps upper half-plane onto sector {ω : 0 <
arg(ω) < απ} and a shift along real axis ζ = z − x0 .
Observe that f ′ (z) = βA(z − x0 )α with α = β − 1 =⇒ −1 < α < 1.
Consider

f ′ (z) = A(z − x1 )α1 · · · (z − xN )αN (3.5.2)

with x1 < x2 < . . . < xN and |αj | < 1 (so f (z) will be primitive of it). Then

x > xN =⇒ arg(f ′ (x)) = arg(A), (3.5.3)

xN > x > xN −1 =⇒ arg(f ′ (x)) = arg(A) + παN , (3.5.4)
xj+1 > x > xj =⇒ arg(f ′ (x)) = arg(A) + παj+1 + . . . + παN , (3.5.5)
x < x1 =⇒ arg(f ′ (x)) = arg(A) + πα1 + . . . + παN . (3.5.6)

Now we apply this to a mapping H+ onto polygon. Assume that it has

vertices w0 , w1 , . . . , wn listed counter-clockwise. And let θ0 , θ1 , . . . , θN be
exterior angles:
Note that

θ0 + θ1 + . . . + θN = 2π (3.5.7)

and for αj = −θj /π

−1 < αj < 1, α0 + α1 + . . . + αN = −2. (3.5.8)

Theorem 3.5.7 (Schwarz-Christoffel Theorem). Let P be a polygon in the

plane with vertices w0 . . . wN and corresponding exterior angles θ0 , θ1 , . . . ,
θN .
Chapter 3. Analytic Functions as Mappings 155

w0 θ4
w4
θ0
w2
θ2

θ1 θ3
w1 w3

Figure 3.2: Note directions of angles!

Let αj = −θj /π. Then there exist real numbers x1 < x2 < . . . < xN and
a constant A such that the function f (z) whose derivative is

f ′ (z) = A(z − x1 )α1 · · · (z − xN )αN , (3.5.2)

which gives a one-to-one analytic mapping of the upper half-plane H+ =

{z : Im(z) > 0} onto the polygon P ; f maps xj to wj for j = 1, . . . , N and

f (∞) = lim f (x) = w0 . (3.5.9)

x→±∞

Remark 3.5.2. (a) The problem however is to find a primitive of (3.5.2)

and then x1 , . . . , xN and A.

(b) Polygon in this content is something more general, than the bounded
polygon, leave alone bounded convex polygon, considered in geometry.
It can be unbounded and contain cuts.

(c) While there are plenty of examples in the Textbook, we consider few,
those where we can achieve a simple answer, may be even we already
know.
Example 3.5.4. Let P = {w : − a < Re(w) < a, Im(w) > 0}.
Chapter 3. Analytic Functions as Mappings 156

Note that there is no w0 (or rather w0 = ∞ but

we ignore it), θ1 = θ2 = π2 and α1 = α2 = − 21 .
In contrast to (3.5.8) α1 + α2 = −1.
Let us take x1,2 = ∓1. We can always rescale and
B
shift what we got. Then f ′ (z) = √ and
1 − z2
w = f (z) = B arcsin(z) + C.
And indeed, z = sin(w) maps strip {w : − π2 <
w1 Re(w) < π2 , Im(w) > 0} onto H+ (so B = 1,
w2
C = 0).
Chapter 3. Analytic Functions as Mappings 157

Example 3.5.5. Let P = C \ {w : − a < Re(w) < a, Im(w) > 0}.

Note that there is no w0 (or rather w0 = ∞
but we ignore it), and w1 , w2 are counter-
clockwise (remember, how we go around do-
main), and now θ1 = θ2 = − π2 and α1 = α2 =
1
2
.
Again, in contrast to (3.5.8) α1 + α2 = 1. Let
us take x1,2 = ∓1. We can always rescale and
shift what we got.
√
Then f ′ (z) = 2B 1 − z 2 and
w1 √
w2

w = f (z) = B z 1 − z 2 + arcsin(z) + C.

It is a nice exercise to prove that with B =

−1, C = 0 maps H+ onto P with a = π2 .
Example 3.5.6. P = H+ \ [0, i].

w2
w2

w3
w1 w3 w1

Here w1 and w3 are different points and θ1 = θ3 = π2 , θ2 = −π. To

understand why so, look on the picture on the right. Then α1 = α3 = − 12 ,
α2 = 1 and selecting x1,3 = ∓1, x2 = 0 (symmetry suggest that it is a good
Az √
idea) we get f ′ (z) = √ =⇒ f (z) = A z 2 − 1.
z2 − 1
And indeed, with A = 1 we get a required map.