Notes on complex analysis (ICTP)

FZF

1 Introduction
General overview of the notes:

1. Analytic functions and power series

2. Riemann Sphere and Möbius transformations

3. Complex integration

4. Bonus topics (Zeta function, etc.)

Notes will be at the level of [MA66].

Motivation
Complex analysis deals with analysis over the field of complex numbers C :=
{a + bi, a, b ∈ R; i2 = 1}. It is the local theory relevant for the study of complex
manifolds. For example, Riemann surfaces, i.e., spheres of genus zero, tori or genus
g = 1 and tori of g > 1. Taking a point around a point it is like taking a local
patch of Euclidean space.
This is important in maths and physics.

Example 1 String theory: our universe is 10-dimensional. Where spacetime is

4-dimensional, and the rest of the dimensions are called Calabi-Yau manifolds are
6-dimensional with real dimensions or rather complex 3-dimensional.

In mathematics we call analytic functions or holomorphic functions appear

often (Fourier analysis, Harmonic analysis, Number Theory, . . . )

Example 2 Riemann zeta function: This function is defined as follows

Z ∞ s−1
1 x
ζ(s) = dx, s ∈ C (3)
Γ(s) 0 ex−1

1
 R∞
where Γ(s) = 0
xs−1 e−x dx The zeta function may be defined as an analytic con-
tinuation of
∞
X 1
ζ(s) = (4)
n=1
n(s )
from {s ∈ C; Re(s) > 1} to all of C, where the zeta function is now an analytic
function related to the distribution of primes. The Riemann hypothesis says that
the zeros of the function lie on the region Re(s) = 1/2. This is, arguably, the most
difficult problem in mathematics.

1.1 Construction of C
Recall that in classical analysis we work over (R, +, ·), which is an ordered field,
i.e,

• for any x, y ∈ R precisely one of x < y, x > y, x = y, holds true.

• x, y > 0 then x + y > 0 and xy > 0

(R, +, ·) is however not algebraically closed.

Remark 5 Can (Rn , +, ·) be made into a field?

For n = 1, 2, 4, 8 is possible but for all other n is not possible!

Claim 6 If (R2 , +, ·) can be made into a field, then there must exist z ∈ (R2 , +, ·)
such that z ( 2) = −1

Proof. Choose a basis {1, e} of R2 . Then z = x · 1 + y · e, for x, y ∈ R. Now we

have z 2 = x2 ·1+2xy ·e+y 2 ·e2 . Where i·e = e and e2 = a·1+b·e for a, b ∈ R using
the vector space properties. Then z 2 = (x2 + ay 2 ) · 1 + (2xy + by 2 ) · e. We want to
2 2 2
( that (x, y) ∈ R exists such that z = −1. Then we must have 2xy +by = 0,
prove
y = 0, or
iff . y = 0 is not sufficient to get z 2 = −1. The other case, y ̸= 0 and
x = − 2b y
2 2
x = −b2
y. Then z 2 = ( −b
2
y) · 1 + (2( −b
2
)y 2 + by 2 ) · e = (( −b
2
y)) · 1. We claim that
b2 b2
a + 4 < 0 assuming
 y ̸
= 0. By
 contradiction we can assume a + 4
≥ 0 so we get
q q
2 2
two roots, z + y 2 a + b4 z − y 2 a + b4 = 0, since the field is has no zero
divisors, one of the terms in the parenthesis must be zero, which is a contradiction.
So the claim is proved. Now, by the latter, we set y = q 1 b2 and x = − 2b , then
−(a+ 4 )
 
2 2 −1
z 2 = (a + b4 )y 2 = (a + b4 ) a+b 2 = −1.
4

2
 Now pick z ∈ (R2 , +, ·), such that z 2 = −1, and call it i.1 .
The field we are using is

C := R[i]
:= {a + bi | a, b ∈ R}
= R[x]/(x2 + 1)
:= {a + bX ∈ R[X]} .

Remark 7 C is not an ordered field!

We cannot say that one number is larger than other number in general. Even
if the coefficients are larger it’s not possible to say it. The only case in which we
can say it is if we are within the R axis.

Remark 8 (Miracle!) C is algebraically closed! (very unlike R).

Theorem 9 (Fundamental theorem of algebra (FTA)) Any non-zero P ∈

C[x] (polynomials with complex coefficients) has at least one root in C. (Proof
later).

Summary
Geometric representation of complex numbers:

• C is a vector space of dimension 2.

• z = a + ib; a, b ∈ R, where a = Re(z) and b = Im(z).

• z̄ = a − ib is the complex conjugate.

• Memorise! i2 = −1, i3 = −i, i4 = 1, i5 = i

• |z|2 = z z̄ = a2 + b2 .

• |z| ≥ 0 being zero iff z = 0

• |z| is the Euclidean distance from z to 0.

• z+w =z+w

• zw = zw

• |zw| = |z||w|
1
Imaginary numbers were first introduced as “ideal” numbers by Gauss
1799.

3
 • If z ̸= 0 then 1
z
= z
|z|2

• Polar representation: z = r(cos(θ) + i sin(θ)), where r = |z| is the modulus

and θ = arg(z) is the argument of z. Note that the argument is not uniquely
defined (a headache indeed)!

• Euler’s formula: eiθ = cos(θ) + i sin(θ), where ez ≡ ∞ zk

P
k=0 k! , i.e.,

iθ (iθ)2 (iθ)3
e = 1 + iθ + + + ···
2! !3! !
(iθ)2 (iθ)3
= 1− + ··· + i θ − + ···
2! 3!
= cos(θ) + i sin(θ)

• Exponential form: z = reiθ , then if z1 = r1 eiθ1 and z2 = r2 eiθ2 , z1 z2 =

r1 r2 ei(θ1 +θ2 ) . NB. ei(θ+2πk) = eiθ for k ∈ Z. i.e., the complex exponential is
periodic.

2 Complex analysis: Differentiation and Holo-

morphic functions
Before we saw that C is a metric space with d(z, w) = |z − w|, isometric to R2 . It
has the following properties:
• Triangle inequality: |z + w| = |z| + |w|, and
• Topology: open disks {z ∈ C ; |z − w| < R} = D(w, R).
Definition 10 (Open subset) A ⊆ C is open if ∀a ∈ A has a dis D(a, ϵ) ⊆ A
for some ϵ > 0.
Definition 11 (Functions) f : C −→ C with C ∼ = R2 can be viewed as functions
f : R2 −→ R2 defined by z = x + iy 7→ (x, y) while f (z) = u(x, y) + iv(x, y) can be
divided into real and imaginary parts as well.
Definition 12 f : A −→ C, with A ⊆ C open. C a field means that we can
consider
f (z + h) − f (z)
h
for h ∈ C. We say that f is differentiable at a ∈ A if the limit
f (z + h) − f (z)
f ′ (a) := lim (13)
h→0 h
exists.

4
Definition 14 (Holomorphic function) f is holomorphic on A if f is dif-
ferentiable at each point a ∈ A.
z+z
Example 15 f (z) = 2
is not complex differentiable (even if it is such a well-
behaved function!).

Remark 16 All the usual rules for Differentiation apply for complex functions.

To know how a function is holomorphic we’ll use the all-important

Definition 17 (Cauchy-Riemann equations) Let z = x + iy with x, y ∈ R, let

f = u + iv with u, v : R2 −→ R, then we can approach the origin in particular in
the following cases:

• (x + k, y) → (x, y), then

f (z + k) − f (z) u(x + k, y) v(x + k, y)

:= +i (18)
k k k
If f is holomorphic then
 
′ ∂u ∂v
f (z) = +i (x, y). (19)
∂x ∂x

• (x, y + k) → (x, y), then

f (z + ik) − f (z) f (z + k) − f (z)

= −i (20)
ik k
or    
′ ∂u ∂v
f (z) = −i (x, y) + (x, y). (21)
∂y ∂y

as f is holomorphic we can approach it in any direction and the limit exists, but
it should be unique as well, i.e., we should have (19)=(21) giving
(
∂u ∂v
∂x
= ∂y
∂u ∂v
(22)
∂y
= − ∂x

called the Cauchy-Riemann equations.

Remark 23 Note that (holomorphic =⇒ CR-equations), but the converse is not

true, except for some reasonable conditions.

5
Theorem 24 (Goursat (partial converse: CR =⇒ holomorphic)) Let u, v :
A ⊆ R2 ∼ = C −→ R satisfy the CR equations (22) and assume that all partial
derivatives ∂u , ∂u , ∂v , ∂v are continuous, then f := u + iv is holomorphic on A.
∂x ∂y ∂x ∂y

Note that this theorem is not optimal, meaning there are more sophisticated or
complete theorems in that direction.
Proof. By calculus, these regularity conditions imply, up to first order terms,
∂u ∂u
u(x + a, y + b) − u(x, y) = a+ b + ϵ1
∂x ∂y
∂u ∂u
v(x + a, y + b) − v(x, y) = a+ b + ϵ2 (25)
∂x ∂y
ϵ1 ϵ2
where h
→ 0, h
→ 0 as h → 0 with h := a + ib. Using (22), with f = u + iv
 
∂u ∂v
f (z + a
|+{z ib}) − f (z) = ∂x + i ∂x (a + ib) + ϵ1 ϵ2 (26)
h

therefore
f (z + h) − f (z) ∂u ∂v
lim = +i (27)
h→0 h ∂x ∂y
and the limit exists. Therefore f is holomorphic on A.
Example 28 If the limits along the x or y axes exist, it doest not imply that the
2-variable limit exists. Take
x+y
(29)
x−y
which has a limit for every line, except for x = y. So, setting y = ax with a ̸= 1,
meaning the limit is
x(a + 1) a+1
= . (30)
x(1 − a) 1−a
Let’s give some examples of holomorphic functions
Example 31 f (z) = z is holomorphic. Let u = x, v = y, then, by (22)
(
∂u ∂v
∂x
= 1, ∂y =1
∂u ∂v
(32)
∂y
= 0, − ∂x = 0
f is holomorphic.
Example 33 f (z) = z. Where u = x, v = −y, then, by (22)
(
∂u ∂v
∂x
= 1, ∂y = −1
∂u ∂v
(34)
∂y
= 0, − ∂x = 0
therefore f is not holomorphic.

6
Alternative way to think of holomorphic functions
This way is a useful way depending on the conjugate. Think of z, z as independent
variables and consider f (z, z) a function of them independently. Then
(
z = x + iy, x = 12 (z + z)
(35)
z = x − iy, y = 2i1 (z − z)

and we apply the chain rule

1/2 −1/2i
z}|{ z}|{
∂f ∂x ∂f ∂y ∂f
= +
∂z ∂z ∂x ∂z ∂y
 
1 ∂f ∂f
= +i . (36)
2 ∂x ∂y

On the other hand if f = u + iv, we have

∂f ∂u dx ∂u dy ∂v dx ∂v dy
= + +i +i
∂z ∂x dz ∂y dz ∂x dz ∂y dz
1 ∂u 1 ∂u ∂v 1 1 ∂v
= +i +i −
2 ∂x 2 ∂y ∂x 2 2 ∂y
   
1 ∂u ∂v ∂u ∂v
= − +i + . (37)
2 ∂x ∂y ∂y ∂x

Now, (
∂u ∂v
∂f ∂x
− ∂y
=0
= 0 ⇐⇒ ∂u ∂v
⇐⇒ (22) (38)
∂z ∂y
+ ∂x
=0
which are the CR equations.
So to test the conditions required by the CR equations we can now use (38).
With this tool, functions like f (z) = |z| = zz are clearly not holomorphic, and we
can check it almost immediately. Also, we have the following.

Corollary 39 Any polynomial P (z) = nk=0 aK Z K is holomorphic in all of C.

P
These gives a bunch of examples which are easily verified to be or not to be holo-
morphic:

1. f (z) = sin(z) + |z|2 is non holomorphic.

2. f (z) = z is non holomorphic.

3. . . .

7
3 Power Series
Power series of the form +∞ n
P
n=0 an (z − a) are holomorphic. To see this we need to
develop some formalism and introduce some mathematical technology to prove it.

Definition 40 Let X be a metric space, then

1. A sequence xn ∈ X is Cauchy if for all ϵ > 0, ∃N ∈ N, such that d(xn , xm ) <

ϵ, ∀n, m ≥ N.

2. X is complete if every Cauchy sequence has a limit in X.

3. X is compact if every sequence has a convergent subsequence. In a topo-

logical way, it is compact if every open cover has a finite subcover. Finally
remember that in Rn (but not in general metric spaces) compactness means
that it is closed and bounded.

4. X is connected if U ⊆ X open and closed, then U = X or U = ∅

5. Let fn : X → Y is a sequence of functions between metric spaces. If fn→f

converges uniformly if ∀ϵ > 0, ∃N ∈ N such that d(fn (x), f (x)) < ϵ,
∀n ≥ N and ∀x ∈ X.

Recall, that

Theorem 41 If fn → f uniformly and fn are continuous, then the f is continu-

ous.

Then it’s easy to see

Example 42 Let fn : [0, 1] −→ [0, 1], defined by fn = xn , then

(
1 x=1
lim fn (x) = (43)
n→+∞ 0, x ̸= 0

, which is a discontinuous limit (see theorem above).

Definition 44 (Power Series) A power series is a series

+∞
X
an (z − a)n (45)
n=0

where a, a0 , a1 , . . . are given complex numbers and z ∈ C is a complex variable.

8
 z n may or not
P
Note that a power series does not always converge. For example
converge but in general it diverges for |z| ≥ 1.

Remark 46 For which z ∈ C does +∞ n

P
n=0 an (z − a) converge?

Theorem 47 Let

R := sup {r ≥ 0 ; a0 , a1 , . . . is a bounded sequence} . (48)

So, in particular R ≥ 0 or R = +∞. R is called the radius of convergence

of (45). Then
1. If |z − a| > R then (45) diverges.
2. If If |z −a| < R then (45) converges absolutely and uniformly on any compact
subset of the open disk D(a, R).

Warning! 1 If |z −a| = R, i.e., the boundary of R, the techniques of convergence

above tell nothing, and there is no general answer and we have to check case by
case.

Theorem 49 (Computing R) To compute R we use the following if the limit

exists:
1.
1
= lim sup |an |1/n (50)
R n→+∞
2.
|an |
R = lim (51)
n→+∞ |an+1 |

if the limit exists!

To prove part 2 of (47), i.e., the uniform convergence of a power series on compact
subsets of the disk we need the following theorem.
Theorem 52 (Weierstrass M -test) A ⊆ C open, and fn : A → C for n ∈ N.
Suppose that |fn (z)| ≤ M an where C ∋ M ≥ 0 and an ≥ 0, such that
X
an < ∞. (53)
n

Then X
fn (z) (54)
n≥0

converges absolutely and uniformly on A.

9
Proof. First note that
X X
|fn (z)| ≤ M an < ∞ (55)
n≥0 n
P
then fn (z) converges absolutely ∀z ∈ A. Recall that an absolutely convergent
complex series is convergent. Hence,
N
X
SN (z) := fn (z) → S(z), (56)
n=0

then
∞
X ∞
X ∞
X
|SN (z) − S(z)| = fn (z) ≤ |fn (z)| ≤ M an < ∞ (57)
n=N +1 n=N +1 n=N +1

where in the first inequality it was used that the sequence is convergent, and notice
that the last term between inequalities is independent of z ∈ A (which will lead
uniform convergence)! Therefore, ∀ϵ > 0, ∃N ′ > 0 such that N > N ′ , then
to P
M ∞ n=N +1 an < ϵ and hence |SN (z) − S(z)| < ϵ ∀z ∈ A.

Let n≥0 z n and fn (z) = z n and AP

P
Example 58 P = D(0, r) for 0 < r < 1. Then
|z n | < rn and rn < ∞. Then (52) implies that n≥0 z n converges uniformly on
1
D(0, r) to 1−z .

Theorem 59 (Radius of convergence theorem)

1. If |z − a| > R (then we are outside ofPthe disk) then by definition an (z − a)n

forms an unbounded sequence, thus, an (z − a)n diverges.

2. If |z − a| < R and let r < ρ < R, then if {an ρn }n∈N is a bounded sequence,
so ∃M > 0 such that |an ρn | ≤ M, ∀n ∈ N.

Proof. (2): Set fn (z) = an (z − a)n , we want to apply the M-test: if |z − a| < r
then
 n X  r n
n n n n r
|fn (z) = |an (z − a) || = |an | |z − a| ≤ an r = |an ρ | ≤M <∞
| {z } ρ n≥0
ρ
≤M | {z }
≤1

P
By the M-test n≥0 fn (z) converges uniformly on D(a, r) (this is what we mean
when we say that it converges for compact subsets).

10
 PN n
P∞ sum SN (z) = n=0 an (z − a)
Corollary 60 An immediate consequence is for the partial
converges uniformly, due to the M-test, to S(z) = n=0 an . which is continuous
as a function of a complex variable z, and applies in the radius of convergence.

Theorem 61

1.
1
= lim sup |an |1/n
R n→+∞

(with the convention that 1/0 = ∞ and 1/∞ = 0), and

2.
|an |
lim
n→+∞ |an+1 |

exists, then it equals R.

Proof. (idea of proof of (1): we want to prove that R ≤ 1/L and R ≥ 1/L)
(1) Let L = limn→+∞ sup |an |1/n . If r < 1/L, L < 1/r, then |an |1/n < 1/r
for all n large enough. Hence, |an rn ||r|n < 1 for all such n large enough. Then
1/L ≤ R.
1/n
n  that |an |
If r > 1/L, pick r >ρ > 1/L, such
n
> 1/ρ for all n large enough.
Hence, |an rn | = |an ρn | ρr > r
ρ
→ ∞ as n → ∞, thus {an rn }n∈N is not
bounded. Thus R ≤ 1/L.

Example 62 (Geometric series)

n n 1
P
n≥0 c z , R = |c|

Example 63 (Exponential series)

zn |an | (n+1)!
P
n≥0 n! then R = ∞ because |an+1 | = n!
=n+1→∞

Example 64 (Asymptotic series)

P n |an | n! 1
n≥0 z n! then R = 0 because |an+1 | = (n+1)!
= n+1
→0

Example 65
We don’t need to use these tests, for example, assume
1
f (z) = (66)
z
around z = 1 + i, then √
what is R? f is holomorphic except at 0. So that the radius
of convergence is R = 2, which is clear, because it is the radius before hitting the
problematic point.

11
3.0.1 Products of power series
P P
Let U = n≥0 un and V = n≥0 vn , if they converge absolutely, then so does
X
W = wn
n≥0

where n
X
wn := uk vn−k . (67)
k=0

For power series f (z) = n an z n , and g(z) = n bn z n . The radius of conver-

P P
gence of the product f g, Rf g is

Rf g ≥ min(Rf , Rg ). (68)

3.0.2 Power series are holomorphic

Hww do we rigorously differentiate ∞ n
P
n≥0 an (z − a) . The goal is to show that
they are infinitely differentiable with
f (n) (n)
= an (69)
n!
such that ∞
X
an (z − a)n
n≥0

is holomorphic in the disk D(0, R).

Theorem 70
Let ∞
X
f (z) = an (z − a)n
n≥0

and
R := lim sup an|1/n , (71)
n→+∞

then
1. f is holomorphic in D(a, R)
2. f is infinitely differentiable in D(a, R), with
∞
X
′
f (z) = nan (z − a)n−1 (72)
n=1

12
 3.
f (n) (a)
= an (73)
n!
4.
f (n) (z) (74)
has radius of convergence R.

Remark 75 The coefficients an are uniquely determined by f ′ . Here an = 0, ∀n

then f ≡ 0 in D(a, R).
Comparing to real analysis, take the following function:
( 1
e− x2 ; x ̸= 0
f (x) = (76)
0; x = 0.

with graph
2 y

1 2
y = e−1/x

x
−4 −2 2 4

−1

−2

which is very flat near the origin. It is a smooth function, i.e., it is a C ∞ real
valued function and by differentiating n times we get f (n) (0) = 0, ∀n. We can tell
that this function cannot be the restriction of a continuous function in C. Let us
1/t2
imagine x = it and have P∞ f (it) = e → ∞ as t → 0.
n
Proof. let f (z) = n≥0 an (z − a) with a ∈ C fixed. We want to show that
f ′ (z) exists and equals ∞ n−1
P
n=1 nan (z − a) within some R, then it will be infinitely
differentiable!
The radius of convergence of
X
nan (z − a)n−1 = lim sup |nan |1/n = lim sup n1/n |an |1/n = lim sup |an |1/n =: R.
(77)

13
Now, we fix also z ∈ C and consider as a function of h. Then
∞
1 X an
(f (z + h) − f (z)) = ((z − a + h)n − (z − a)n ) . (78)
h n=0
h
Now, we define
N
X an
SN (h) := ((z − a + h)n − (z − a)n ) (79)
n=0
h
n−1
we claim that |fn,h (z)| ≤ n|an |r if we are in a set where the following happens:
(
|z − a + h| < r
(80)
|z − a| < r
and if X
n|an |rn−1 < ∞. (81)
We, first have to prove the following.
an
((z − a + h)n − (z − a)n ) ≤ n|an |rn−1 < ∞. (82)
h
n n
−W
which we do by setting Z = z − a + h, W = z − a, h = Z − W . Then ZZ−W =
n−1 n−2 n−1 n−1
Z + Z W + ··· + W , then |· · · | ≤ n|an |r if |Z| < r and |W | < r.
Now, applying the M-test, we get
SN (h) → S(h) (83)
uniformly and since SN (h) are continuous, so is S(h).
In particular
X∞
SN (0) = lim SN (h) = nan (z − a)n−1 (84)
h→0
n=1
Then (
f (z+h)−f (z)
h ̸= 0
SN (h) = P∞ h n−1 (85)
n=1 nan (z − a) , h = 0.
which by continuity
∞
f (z + h) − f = (z) X
′
f (z) = lim = lim S(h) = S(0) = nan (z − a)n−1 . (86)
h→0 h h→0
n=1

We continue in this way and prove that

∞
X
′′
f (z) = n(n − 1)an (z − a)n−2 , (87)
n=2

etc.

14
4 Elementary Analytic/Holomorphic functions:
trigonometric, multi-valued, logarithms, branch-
cuts.
For any z ∈ C, let
+∞ n
z
X z
e = (88)
n=0
n!
which has a radius of convergence R = ∞.
Remark 89 A function of this type is called an Entire function: i.e., defined
and holomorphic in the entire complex plane. We’ll see this in more detail further
on. In particular if we have a function defined by power series and has R = ∞
then it is holomorphic and defined in the whole complex plane.
The following are also entire functions
∞
X −1n
sin(z) = z 2n+1 , R=∞ (90)
n=0
(2n + 1)!
∞
X (−1)n 2n
cos(z) = z , R = ∞. (91)
n=0
(2n)!

There are many similarities and many properties hold from real analysis.
Example 92 the complex exponential function has the following properties (mostly
different than the real counterpart):
1. ez ew = ez+w which is proved by using the convolution property (67).
1z
2. ez · e−z = 1, then ez is never zero on C and e−z = e

3. |ez |2 = ez+z

4. |ez | = eRe(z) , (ez = eRe(z) ei Im(z) , then |ez | = eRe(z) ei Im(z) = eRe(z) ).
| {z }
=1

5. e2πi = 1, meaning that ez is 2πi-periodic, i.e., ez+2πik = ez , ∀k ∈ Z. (This is

different to the real version of the exponential function, which is one-to-one,
etc.)
iz −iz eiz +e−iz
6. sin(z) = e −e
2i
and cos(z) = 2
which are 2π-periodic (just as their
real counterparts).

15
 7. In particular, ∄ limz→∞ ez
Remember that these functions may be expressed as power series, which are dif-
ferentiable within their radius of convergence. In the case of the exponential and
sine and cosine complex functions, they have R = ∞, thus, they are infinitely
differentiable, or holomorphic everywhere. Now we want to look at their inverses,
which are both interesting and annoying.

The complex logarithm

The complex logarithm is a multi-valued function defined as follows.
Definition 93 Suppose w ̸= 0. We say that z = log(w) iff w = ez . Say z = x+iy,
and w = ex eiy , then |w| = ex , then log(w) is well defined (it is the usual logarithm
up till now!).
In general, however w = reiθ , r > 0, θ ∈ R, a logarithm (the inverse of the
complex exponential) is given by

log w := log |w| + iθ (94)

or
log w = log |w| + i arg(w). (95)
Then it is easy to see elog |w|+iθ = |w|eiθ , however, we also have elog |w|+iθ+2πik =
|w|eiθ , ∀k ∈ Z because of the periodicity. This means that the logarithm depends
on the choice of the argument (modulo 2π), i.e.,

log : C∗ −→ C/2πiZ (96)

arg : C∗ −→ R/2πZ. (97)

In other words we have to make precise which logarithm will be used. In this
sense, it is usually convenient to take a standard choice for most of the cases.

Remark 98 It is not possible to define a logarithm in a domain that contains a

curve winding around the origin. Moreover we would like the logarithm, suitably
defined to be holomorphic and in particular continuous.
Say we take a unitary circle and start at ei0 , which is equal to ei2πk , f or k ∈
Z. We make a choice, say, k = 0, then by going around the circle we reach
eiπ/2 , eiπ , ei3π/2 , . . . , ei(2π−ϵ) . Also recall that log |w| = 0 (because |w| = 1) then the
only relevant part is i arg(w) from (95). Then arg = 0, iπ/2, iπ, i3π/2, . . . , i(2π −
ϵ), but at this last point the logarithm instead of going back to 0, it’ll go to 2πi.
This can be seen in the typical spiral (parking garage) graph of the complex loga-
rithm. This is the main point of the complex logarithm, which is related to complex

16
manifolds, that locally around any point it looks like C, but globally it’s something
different.
In other words if zt = eit for t ∈ R, naively speaking we want log zt = it but
zt=0 = z2πik , ∀k ∈ Z meaning it would not be single valued.

Even if we defined the logarithm within a specific interval (say, [0, 2π)), but this
does not solve the main problem, i.e., that it is not continuous in the same point,
because approaching from above and below gives a different value. Thus, we have
to remove something, like the negative real line, and which we will call a branch
cut. This cut will make it impossible for the logarithm to wind around the origin.

4.0.1 standard solution

We fix a branch (something that goes from the origin to infinity):

C \ (−∞, 0] −→ C (99)
w 7−→ Log w = log w + i arg(w) (100)

with the convention that arg(w) ∈ (−π, π), i.e., we remove all the negative real
axis, then define the logarithm in the other region:

Im(z)
Re(z)

This is a convention and is called the principal branch of the logarithm denoted
by Log, any other branch will be denoted by log. The argument is similarly
notated, i.e., Arg(w) for the principal branch, meaning we choose it such that

Arg(w) ∈ (−π, π) (101)

giving a single valued and continuous (in fact, holomorphic) function on C \

(−∞, 0].
Now for most purposes we’ll use Log, but we might need to cut something
different (like the positive real axis, or an oblique axis, etc.).

Remark 102 One advantage of the principal branch convention is that

Log 1 = 0. (103)

Which is not true for any branch cut!

17
4.0.2 Possible interpretation
ez is periodic, but can restrict to Im(z) ∈ (−π, π) for

{z ∈ C ; π < Im(z) < π} −→ C \ (−∞, 0]

z = x + iy 7−→ ez = ex eiy (104)
[C]
π Im(z)
−→
z Re(z) (105)
0 e

−π

which is bijective! therefore giving the principal branch.

Proposition 106

1. Log 1 = 0

2. Log zw = Log z+Log w ( mod 2πi) because Log zw = log(zw)+i Arg(zw) =

log |z| + log |w| + Arg(z) + Arg(w) ( mod 2πi). (to make it clearer take
θ1 ∈ (−π, π) and θ2 ∈ (−π, π), then eiθ1 eiθ2 = eiθ1 +θ2 , where it might hap-
pen that θ1 + θ2 ̸∈ (−π, π) for which we should add or substract such that it
becomes a multiple of 2πi).

4.0.3 Differentiability
Theorem 107 Let f : U −→ V and g : U −→ V , and U, V ⊆ C open, such that

1. g ◦ f (z) = z

2. f is continuous

3. g is analytic/holomorphic and g ′ (z) ̸= 0 on V

Then f is holomorphic on U and f ′ = 1

g ′ (z)
.

Example 108 Log z is holomorphic on C \ (−∞, 0] and

d 1
Log z = (109)
dt z
Proof. dtd ez = ez so if g(z) = ez then g ′ (z) ̸= 0on C, so that its inverse is analytic
on C \ branchcut and f ′ (z) = elog1 z = z1 .

18
4.0.4 The complex Square root
√
Definition 110 w = z iff z = w2 . The function C −→ C : z 7−→ z 2 is not

2
2
bijective, since rei(θ+π) = r2 ei(2θ+2π) = r2 ei2θ ei2π = reiθ , i.e., by adding a
multiple of π we get the same, then it is multivalued.
The standard solution means to restrict to Re(z) > 0,
{z ∈ Re(z) > 0} −→ C \ (−∞, 0]
z 7−→ z 2 (111)
[C] [C]

−→
2
(112)
0 (·) 0

which is bijective, i.e., has an inverse

√
: C \ (−∞, 0] −→ {z ∈ Re(z) > 0}
p
w 7−→ |w|ei Arg(w)/2 (113)
p 2
This is because |w|ei Arg(w)/2 = |w|ei Arg(w)/2 . This is the principal branch
of the square root, for Arg(w) ∈ (−π, π).
Remark 114 This means that when taking the square root, and go around two
laps we end up where we started; if we go around one lap, we have a new value.
We√can show that the square root is holomorphic on C\(−∞, 0] and the deriva-
tive dtd z = 2√1 z on that domain.

4.0.5 Powers
Let b ∈ R, then
z b := eb log z , z ̸= 0, b ∈ R (115)
Example 116 z 7→ z 5 is 2π/5-periodic. By restricting
{z ; −π/5 < Arg(z) < π/5} −→ C \ (−∞, 0] (117)
which is bijective and where z 7→ z 1/5 mapping
[C] [C]

−→
0 0

19
Derivation is clear by using the chain rule
d b d b
z = eb log z = eb log z = bz b−1 (118)
dz dz z

4.1 Inverse trigonometric functions

Definition 119
1 iw
e + e−iw .


w = arccos(z) ⇐⇒ z = cos w = (120)
2
Then everything is in terms of exponentials.
Let u = eiw such that u ̸= 0, then
 
1 1
z= u+ (121)
2 u
gives
u2 − 2zu + 1 = 0 (122)
giving √
u=z± z2 − 1 (123)
which is a complex square root. Then
1 √
w= log(z ± z 2 − 1) (124)
i
by taking the principal part
 
i Arg(z 2 −1)/2
p
2
Arccos(z) := −i Log z + |z − 1| e . (125)

This function is defined

[C]

-1 1
0

where the red lines represent the places where it is not defined.

20
5 Conformal Maps
We want to give a geometric interpretation of CR-equations (22).

Example 126 Let f = u + iv where u, v : R2 −→ R is a (real) differentiable

function viewed as a function f : R2 −→ R2 , i.e.,
   
x u(x, y)
7→ (127)
y v(x, y)

Then we consider the total derivative (Jacobian):

   ∂u ∂u      
′ x x ux uy x
f (x) : 7−→ ∂x
∂v
∂y
∂v ≡ (128)
y ∂x ∂y
y vx vy y

from R2 → R2 .
The CR-equations (22) imply that
   
ux uy ux uy
= =J (129)
vx vy −uy ux

The map     
x a b x
7→ (130)
y −b a y
is rotation and scaling. This is a local behaviour of f .

Roughly, a conformal map preserves angles and orientations locally.

Example 131 1. Scaling and rotations are conformal.

2. f (z) = Re(z) is not conformal because it doesn’t preserve angles.

3. f (z) = z is not conformal because it doesn’t preserve orientations. It does

preserve angles, it is called anti-conformal.

4. f (z) = z k for 2 < k ∈ N is conformal.

5. f (z) = ez is conformal.

6. Any non-constant holomorphic function is conformal (we’ll prove this later

on).

Definition 132 Let A ⊆ C an open subset, the map f : A −→ C is conformal

if

21
 1. As a map f : A → R2 if is real differentiable

2. f preserves angles of intersections between smooth curves2

 
ux uy
3. f preserves orientations, i.e., det ≡ det J > 0
vx vy

There exists an internal product ⟨α, β⟩ = cos θ∥α∥∥β∥, but we want to preserve

⟨α, β⟩
(133)
∥α∥∥β∥

under f . To do this we look at the tangent vectors of f (γ(t))and f (γ̂(t)) given by

the multiplication by J, so preserving angles means

⟨Jα, Jβ⟩ ⟨α, β⟩

= . (134)
∥Jα∥∥Jβ∥ ∥α∥∥β∥

Remark 135 A map that satisfies (1) and (2) in (132) but not (3) is called anti-
conformal.

Proposition 136 Any anti-conformal map is of the form z 7→ f (z) where f is

conformal.

Theorem 137 (Conformality, holomorphicity and CR equations) If f is confor-

mal in A ⊆ C open then the CR-equations (22) hold.
Conversely, if u, v : R2 −→ R satisfy the CR-equations and ux , uy , · · · exist
and are continuous, then f = u + iv is conformal (Caution: continuous functions
don’t satisfy these conditions!)
In other words, for regular “enough” functions: f is conformal iff it is holo-
morphic iff the CR-equations hold for f .
Fact: a holomorphic function such that f ′ (z) ̸= 0 ⇐⇒ f is conformal.

Proof. Suppose f is conformal. Then det J > 0. Consider the column vectors
 
ux
α= (138)
vx
 
uy
β= . (139)
vy
2
A smooth curve R ∋ t 7→ γ(t) = γ1 (t) + iγ2 (t) is smooth. N.B. a curve is a map, not its
image!
Angles between curves are angles between tangent vectors.

22
    
0 1
By definition multiplication by J preserves angles, so in particular ⊥
        1 0
1 0 1 1
then α = J ⊥J = β, similarly α + β = J ⊥J = α − β.
0 1 1 −1
Hence,
0 = ⟨α + β, α − β⟩ = ⟨α, α⟩ − ⟨β, β⟩ = ∥α∥2 − ∥β∥2 . (140)
thus ∥α∥ = ∥β∥, together with α ⊥ β, means that for a vector in R2 this means
that

23
References
[MA66] D Martin and LV Ahlfors, Complex analysis, McGraw-Hill, New York,
1966. 1

24