V5B1 – ADVANCED TOPICS IN COMPLEX ANALYSIS

WINTER SEMESTER 2024/25

WERN JUIN GABRIEL ONG

Preliminaries
These notes roughly correspond to the course V5B1 - Advanced Topics in
Complex Analysis taught by Prof. Ingo Lieb at the Universität Bonn in the
Winter 2024/25 semester. These notes are LATEX-ed after the fact with significant
alteration and are subject to misinterpretation and mistranscription. Use with cau-
tion. Any errors are undoubtedly my own and any virtues that could be ascribed
to these notes ought be attributed to the instructor and not the typist.

1
2 WERN JUIN GABRIEL ONG

Contents
Preliminaries 1
1. Lecture 1 – 11th October 2024 3
2. Lecture 2 – 15th October 2024 6
3. Lecture 3 – 17th October 2024 10
4. Lecture 4 – 22nd October 2024 13
5. Lecture 5 – 24th October 2024 17
6. Lecture 6 – 29th October 2024 19
7. Lecture 7 – 31st October 2024 21
8. Lecture 8 – 5th November 2024 26
9. Lecture 9 – 7th November 2024 29
10. Lecture 10 – 12th November 2024 34
11. Lecture 11 – 14th November 2024 38
12. Lecture 12 – 19th November 2024 42
13. Lecture 13 – 21st November 2024 45
14. Lecture 14 – 26th November 2024 48
15. Lecture 15 – 28th November 2024 51
16. Lecture 16 – 3rd December 2024 52
17. Lecture 17 – 5th December 2024 54
18. Lecture 18 – 10th December 2024 58
19. Lecture 19 – 12th December 2024 59
Appendix A. Basic Results in Complex Analysis 61
References 62
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 3

1. Lecture 1 – 11th October 2024

This course will cover complex analysis in one or more variables with a view
towards both number theory and algebraic geometry.
Recall the following definition.
Definition 1.1 (Affine Regular Quadric). An affine regular quadric is a set Definition 1.1
2 2 2
Q = {(u, v) ∈ C : v = f (u)} ⊆ C
Marginal notes
where f (u) is a univariate polynomial of degree 2 with distinct zeroes. will follow the
numbering from
We can apply a linear coordinate transformation to express the quadric Q in an lecture.
especially nice form.
Proposition 1.2. There exists a linear change of coordinates such that Q is given
by the set {(u, v) ∈ C2 : u2 + v 2 = 1} ⊆ C2 . Proposition 1.1
A quadric given in the above form is said to be in normal form.
Definition 1.3 (Normal Form of Quadric). An affine regular quadric Q ⊆ C2 is
said to be in normal form if it is given by {(u, v) ∈ C2 : u2 + v 2 = 1} ⊆ C2 .
From a first course in complex analysis, we have that cos2 (z) + sin2 (z) = 1 for
all z ∈ C and hence for an affine regular quadric in normal form Q, there is a map
h : C → Q by z 7→ (cos(z), sin(z)).
Proposition 1.4. Let Q be an affine regular quadric in normal form. The map
h : C → Q by z 7→ (cos(z), sin(z)) is holomorphic. Proposition 1.2
Proof. It suffices to see that the coordinate functions z 7→ sin(z), z 7→ cos(z) are
themselves holomorphic. ■
We can also find a rational parametrization. Choose a point (0, 1) ∈ Q. For
z ∈ C, we can consider the complex line v = zu + 1 which intersects Q at one
other point where v 2 = (zu + 1)2 . Since the quadric is in normal form, we have
v 2 = 1 − u2 so expanding we have 1 − u2 = z 2 u2 + 2zu + 1 which is a univariate
quadratic equation in u with z fixed as above. Rearranging the equation, we get
(1 + z 2 )u2 + 2zu = 0 and since we already know the intersection point (0, 1) ∈ Q we
−2z
consider the case when u ̸= 0 so (1 + z 2 )u + 2z = 0 and thus u = 1+z 2 . Substituting
1−z 2
for v, and making similar computations, we can see v = 1+z 2
. In fact, we have the
following result.
Proposition 1.5. LetQ be an affine
 regular quadric in normal form. The rational
−2t 1−t2
map C → Q by z 7→ 1+t2 , 1+t2 is injective and extends to a holomorphic map
b \ {±i} → Q.
C Proposition 1.3
Proof. The map is injective since a univariate equation of the form (1+z 2 )u2 +2zu =
0 identified in the preceding discussion has a unique nonzero root depending on z.
Furthermore, this function can be seen to be meromorphic as it is a rational function
and is hence holomorphic away from its poles which are ±i. ■
 4 WERN JUIN GABRIEL ONG

We can compactify the quadric by embedding it in projective space instead of C2 .

Definition 1.6 (Complex Projective Plane). The complex projective plane CP2 is
given by
(z0 , z1 , z2 ) ∈ C3 \ {(0, 0, 0)} / ∼


where (z0 , z1 , z2 ) ∼ (z0′ , z1′ , z2′ ) if and only if there exists λ ∈ C∗ such that z0 =
λz0′ , z1 = λz1′ , z2 = λz2′ .
We will write these equivalence classes as [z0 : z1 : z2 ].
There is a map ι0 : C2 → CP2 by (u, v) 7→ [1 : u : v] realizing CP2 as ι0 (C2 )∪{z0 =
0}, that is, by adjoining a CP1 at infinity. The closure of image Q b of Q under ι
2 2 2
consists of z0 = z1 + z2 where we have Q = ι0 (Q) ∪ {[0 : 1 : ±i]} which also extends
b
to a rational parametrization by [t0 : t1 ] 7→ [t20 + t21 : −2t0 t1 : t20 − t21 ] rendering the
diagram
C Q C2

CP1 Q
b CP2
commutative. We have the following.
Proposition 1.7. The map [t0 : t1 ] 7→ [t20 + t21 : −2t0 t1 : t20 − t21 ] defines a homeo-
Proposition 1.4 morphism CP1 → Q.
b
b the map admits an inverse [z0 : z1 : z2 ] 7→ [−z0 − z2 : z1 ].
Proof. On Q ■
This shows that Q
b is topologically a sphere.
We now consider the case of cubics.
Definition 1.2 Definition 1.8 (Affine Regular Cubic). An affine regular cubic is a set
K = {(u, v) ∈ C2 : v 2 = f (u)} ⊆ C2
where f (u) is a univariate polynomial of degree 3 with disctinct zeroes.
This too admits a normal form.
Proposition 1.9. There exists a linear change of coordinates such that K is given
by the set {(u, v) ∈ C2 : v 2 = 4u3 − g2 u − g3 } with 4u3 − g2 u − g3 having three
distinct roots.
Proof. Suppose f (u) = a3 u3+ a2 u2 + a1 u + a0 . The desired transformation is given
q 
a2
by u 7→ 3 14 · √1
3a + √
3 2
. ■
3 3 a3

For u1 , u2 , u3 the roots of 4u3 − g2 u − g3 as above, we can solve this depressed

cubic equation and observe that g2 and g3 are given by symmetric polynomials in
the u1 , u2 , u3 . In fact, u1 + u2 + u3 = 0, −4(u1 u2 + u1 u3 + u2 u3 ) = g2 , 4u1 u2 u3 = g3 .
We can then write the discriminant of this polynomial as
16(u1 − u2 )2 (u2 − u3 )2 (u3 − u1 )2 = g23 − 27g32 .
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 5

This defines an elliptic curve which is a well-studied object in number theory and
arithmetic geometry.
Analogously to the case of quadrics, we can pass to the projective closure K
b given
2 3 2 3
by the equation z0 z2 = 4z1 − g2 z0 z1 − g3 z0 which is a homogeneous polynomial
obtained as the one-point compactification of the regular affine cubic K at the point
[0 : 1 : 0].
This hints at a more general problem of finding rational parametrizations of in-
teresting subsets of Cn which as Galois theory suggests is not possible in general,
nor is it so in the case of cubics. Though the study of cubics will give rise to the
theory of elliptic functions.
As before, we can define a quartic.
Definition 1.10 (Affine Regular Quartic). An affine regular quartic is a set Definition 1.3
H = {(u, v) ∈ C2 : v 2 = f (u)} ⊆ C2
where f (u) is a univariate polynomial of degree 4 with disctinct zeroes.
Without loss of generality, we can take f (u) to be monic and of the form u 4i=2 (u−
Q

ai ) by translation. Letting x = u1 we have v 2 = x1 4i=2 ( x1 − ai ) and multiplying

Q

with x4 we can set x4 v 2 = (1 − a2 x)(1 − a3 x)(1 − a4 x) and again taking x4 v 2 = y 2

that y 2 : (1 − a2 x)(1 − a3 x)(1 − a4 x) which is a cubic in x with distinct zeroes that
we can put in the Weierstrass normal form for cubics as in Proposition 1.9 which
reduces the study of quartics to the study of cubics. The first part of the course will
focus on elliptic functions and their connections to number theory.
Returning to the discussion of the cubic above, we can note that the construction
of Proposition 1.5 on restriction to Q recovers the rational points on the circle and Proposition 2.1
by normalizing recover Pythagorean triples, those (a, b, c) ∈ N pairwise coprime such
that a2 + b2 = c2 . Definition 2.1
 6 WERN JUIN GABRIEL ONG

2. Lecture 2 – 15th October 2024

We recall some results on integration theory, first treating the case of rational
functions.
Proposition 2.1. The integral of a rational function can be expresssed as a the
Proposition 2.4 sum of a rational function and a linear combination of logarithms of linear forms.
Proof. Without loss of generality, by taking the partial fraction decomposition of a
f (z)
rational function, it suffices to consider integrals of the form (z−a) m where f (z) is a
1 d 1 1
polynomial in z. If m > 1 then note that 1−m dz (z−a)m−1 = (z−a)m so we can write
d 1 f (z) f (z) f ′ (z)
dz 1−m (z−a)m−1 = (z−a)m + (1−m)(z−a)m−1 reducing the integral to the integration of
f ′ (z) d 1 f (z) f (z)
(1−m)(z−a)m−1
+ dz 1−m (z−a)m−1 . Iterating this procedure, we get to the case of (z−a)
R 1
which can be treated via integration by parts and that z−a dz = log(z − a). ■
We now turn to the case of integrals related to cubics. Consider the integral
Z
1
p dz
4z 3 − g2 z − g3
which arises from the affine regular cubic in normal form as discussed in Proposi-
tion 1.9. We will soon be able to find a parameterization of the above integral to
solve it via the study of elliptic functions.
We begin with a more general treatment of periods of meromorphic functions.
Definition 2.2 (Period). Let f (z) be a meromorphic function. A complex number
Definition 1.1 ω ∈ C is a period of f if f (z + ω) = f (z) for all z.
Remark 2.3. It can be seen that 0 ∈ C is a period for all meromorphic functions.
To the end of showing that these periods form a discrete additive subgroup of C,
let us recall the following result concerning the classification of subgroups of discrete
subgroups of C.
Proposition 1.1 Lemma 2.4. If G is a discrete additive subgroup of C then G is one of the following:
• The trivial group 0.
• ω · Z for ω ∈ C.
• Ω = {a1 ω1 + a2 ω2 : a1 , a2 ∈ Z} with ωω21 ∈ C \ R.
Proof. If G is trivial, we are done. Otherwise, suppose G is non-trivial and consider
r > 0 such that the closed unit ball Br (0) of radius r contains a non-identity element.
Since Br (0) is compact and G is discrete the intersection Br (0) ∩ G consists of only
finitely many elements of G. Take ω1 nonzero in this intersection such that |ω1 | is
minimal. This shows Z · ω1 ≤ G and we are done if this is an equality.
If Z · ω1 < G then consider some ω2 ∈ G \ Z · ω1 with |ω2 | minimal. We first show
that ωω12 ∈
/ R. Suppose to the contrary that ωω21 ∈ R. Then there is an integer n such
that n < ωω12 < n + 1 where the inequalities are strict since ω2 ∈ G \ Z · ω1 . As such
|nω1 − ω2 | < |ω1 | contradicting minimality of |ω1 | showing Z · ω1 ⊕ Z · ω2 ≤ G.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 7

Note that since ω1 , ω2 are R-linearly independent, any complex number can be
written as an R-linear combination of ω1 , ω2 . For z ∈ C expressed as z = λ1 ω1 +λ2 ω2
with λ1 , λ2 ∈ R, take m1 m2 ∈ Z such that |λ1 − m1 |, |λ2 − m2 | ≤ 21 . In particular,
for z ∈ G possibly not in Z · ω1 ⊕ Z · ω2 we have that z ′ = z − m1 ω1 − m2 ω2 ∈ G as
well, but
1 1
|z ′ | = |(λ1 − m1 )ω1 + (λ2 − m2 )ω2 | < |ω1 | + |ω2 | ≤ |ω2 |
2 2
where the strictness of the middle inequality follows from R-linear independence
of ω1 , ω2 and the second inequality from 12 |ω1 | ≤ 12 |ω2 | by minimality of |ω1 | in
Br (0)∩G. Now noting that |ω2 | was minimal among G\Z·ω1 , we have that z ′ ∈ Z·ω1
so for z ′ = nω1 we can rewrite z = (n + m1 )ω1 + m2 ω2 showing z ∈ Z · ω1 ⊕ Z · ω2 ,
giving the claim. ■
We can now show that periods form an additive subgroup as follows.
Proposition 2.5. Let f (z) be a meromorphic function on C. If f is not constant,
then the periods of f form a discrete additive subgroup Ω of C.
Proof. Let ω, ω ′ be periods and −ω the additive inverse of ω in C. We can compute
f (z + (ω + ω ′ )) = f ((z + ω) + ω ′ ) = f (z + ω ′ ) = f (z)
f (z − ω) = f ((z + ω) − ω) = f (z)
showing that periods are closed under addition and inversion, and contain zero per
Remark 2.3. Associativity follows from associativity on the group additive group of
complex numbers, showing that the periods form a subgroup.
It remains to show discreteness. Let ω be a period and consider Br (ω) the open
ball of radius r with r chosen small enough that f is analytic on Br (ω). Suppose to
the contrary for each n ∈ N there exists a period ωn ∈ Br/n (ω). Then |ω − ωn | < nr
showing that the sequence ωn → ω as n → ∞. By the identity theorem Theorem A.1,
f is constant on Br (ω), a contradiction, as f is non-constant. ■
Elliptic functions are defined in terms of their period group.
Definition 2.6 (Elliptic Function). A meromorphic function f (z) is elliptic if its
period group contains a lattice.
Remark 2.7. Elliptic functions are often also known as doubly periodic functions.
Remark 2.8. Defining elliptic functions in terms of their period group containing a
lattice allows constant functions to be elliptic – since constant functions have period
group C.
Elliptic functions are determined by their values on their open period parallelo-
gram since for Ω a lattice, we can define an equivalence relation z ∼ z ′ if z − z ′ ∈ Ω.
Definition 2.9 (Period Parallelogram). Let Ω = {a1 ω1 + a2 ω2 : a1 , a2 ∈ Z} ⊆ C be
a lattice. The period parallelogram is given by
PΩ = {t1 ω1 + t2 ω2 : 0 ≤ t1 , t2 < 1} ⊆ C.
 8 WERN JUIN GABRIEL ONG

Under the equivalence relation described above, the quotient space C/Ω is an
additive group where each point of the period parallelogram is a representative of
the quotient. Moreover, this can be seen to be a complex torus using the standard
cut-and-paste diagram for a torus in topology which is a compact topological space,
and in fact a compact Riemann surface. As such, any Ω-periodic meromorphic
function is determined by its values on the torus.
We now consider a general property of lattices before returning to a discussion of
periodic functions.

Proposition 2.10. Let Ω = {a1 ω1 + a2 ω2 : a1 , a2 ∈ Z} ⊆ C be a lattice. Then

P 1
Proposition 1.2 ω∈Ω\{0} ω k is absolutely convergent for k > 2.

Proof. Denote Pℓ the parallelogram given by the lattice points of the convex hull
of ±ℓω1 , ±ℓω2 and ∂Pℓ its boundary. We have that |∂Pℓ | = 8ℓ and for C =
maxω∈∂P1 |ω| and that (ℓC)k ≤ |ω|k for ω ∈ ∂Pℓ . We can compute
∞ X
X 1 X 1
=
|ω|k |ω|k
ω\{0} ℓ=1 ω∈∂Pℓ
∞ ∞
X 8ℓ 8 X 1
≤ =
(ℓC)k Ck ℓk−1
ℓ=1 ℓ=1

which is only convergent if k > 2. ■

Let us return to a discussion of ellitpic functions.

Proposition 2.11. Let Ω = {a1 ω1 + a2 ω2 : a1 , a2 ∈ Z} ⊆ C be a lattice. Then the

elliptic functions with respect to Ω form a field that is closed under differentiation.

Proof. Double periodicity is preserved under sums, products, differences, and quo-
tients, and the operations distribute in the expected way. Preservation under dif-
ferentiation follows from the chain rule. ■

However, holomorphic elliptic functions are uninteresting.

Proposition 2.12. Let f be an elliptic function with respect to a lattice Ω. If f is

Proposition 2.2 holomorphic, then f is constant.

Proof. The closure of the open period parallelogram PΩ is compact on which |f (z)|
admits a maximum. By periodicity, f (z) is a bounded holomorphic function on C,
from which the claim follows by Liouville’s theorem A.2. ■

Generalizing our discussion to elliptic functions with poles, we can show the
following.

Proposition 2.13.PLet f be a nonconstant elliptic function and z1 , . . . , zm the poles

Proposition 2.3 of f in PΩ . Then m
i=1 reszi f = 0.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 9

Proof. Without loss of generality, we can take the poles to lie in the interior of the
1
R
open period parallelogram. By Cauchy’s residue theorem, we have 2πi ∂PΩ (z) =
f
Pm
i=1 reszi f . Denoting [a, b] the oriented straight line path from a to b we compute
Z Z Z
1
f (z)dz = f (z)dz + f (z)dz
2πi ∂PΩ [0,ω1 ] [ω1 ,ω1 +ω2 ]
Z Z
+ f (z)dz + f (z)dz
[ω1 +ω2 ,ω2 ] [ω2 ,0]
Z Z !
= f (z)dz − f (z)dz +
[0,ω1 ] [ω2 ,ω1 +ω2 ]
Z Z !
f (z)dz − f (z)dz
[ω1 ,ω1 +ω2 ] [0,ω2 ]

where both of the summands vanish by periodicity of f yielding the claim. ■

Remark 2.14. Proposition 2.13 implies that an elliptic function has at least two
poles when counting with multiplicity.
Given a non-constant – and hence meromorphic – elliptic function, it can be
shown that its image is all of C.
b

Proposition 2.15. Let f be a non-constant elliptic function. Then f assumes every

value in C
b equally often, counting with multiplicity.
b Note that the integral 1
R f ′ (z)
Proof. Let λ ∈ C. 2πi ∂PΩ f (z)−λ dz computes the difference
between the number of times f assumes the value λ and f assumes the value ∞
which is zero since the integrand is an elliptic function by Proposition 2.11 and zero
by Proposition 2.13. ■
 10 WERN JUIN GABRIEL ONG

3. Lecture 3 – 17th October 2024

We consider some additional properties of elliptic functions.
Proposition 3.1. Let Ω be a lattice with period parallelogram PΩ and f an elliptic
Proposition 2.5 function with respect to Ω. Let a1 , . . . , an and an+1 , . . . , aℓ be the zeroes and poles
of f , respectively, of orders m1 , . . . , mn and mn+1 , . . . , mℓ , respectively. Then
n ℓ
! !
X X
(3.1) mi ai − mi ai ∈ Ω.
i=1 i=n+1

Proof. Without loss of generality, we can take these zeroes and poles to lie in the
′ (z)
interior of the period parallelogram. Consider the function g(z) = z · ff (z) which
has simple poles at both the poles and zeroes of f with Rresidues mi ai and −mi ai ,
1
respectively. In particular, the sum of (3.1) is given by 2πi ∂PΩ g(z)dz. Decomposing
this as an integral over segments as in Proposition 2.13, we have
Z Z Z Z Z
g(z)dz = g(z)dz + g(z)dz + g(z)dz + g(z)dz.
∂PΩ [0,ω1 ] [ω1 ,ω1 +ω2 ] [ω1 +ω2 ,ω2 ] [ω2 ,0]

Considering the integral over the oriented segments [0, ω1 ] and [ω1 + ω2 , ω2 ] we can
write this integral
Z Z Z Z
g(z)dz + g(z)dz = g(z)dz − g(z)dz
[0,ω1 ] [ω1 +ω2 ,ω2 ] [0,ω1 ] [ω2 ,ω1 +ω2 ]
zf ′ (z) (z + ω2 )f ′ (z + ω2 )
Z Z
= dz − dz
[0,ω1 ] f (z) [0,ω1 ] f (z + ω2 )
f ′ (z)
Z
= ω2 dz
[0,ω1 ] f (z)
where we know that f (z) = f (z + ω2 ), f ′ (z) = f ′ (z + ω2 ) with f ′ (z) elliptic by
Proposition 2.11. In particular, this an integer multiple of ω2 . Arguing similarly,
we can see that
f ′ (z)
Z Z Z
g(z)dz + g(z)dz = ω1 dz
[ω1 ,ω1 +ω2 ] [ω2 ,0] [0,ω2 ] f (z)

which is also an integer multiple of ω1 , giving the claim. ■

We return to a consideration of the field of elliptic functions more generally.
Proposition 3.1 Proposition 3.2. Let Ω be a lattice. The function
X 1
fk (z) =
(z − ω)k
ω∈Ω

is a nonconstant elliptic function for k ≥ 3.

Proof. The function is elliptic by inspection so it remains to show that the function
is locally uniformly convergent. For this, fix r > 0 and consider the disc B2r = {z ∈
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 11

C : |z| < 2r} and note that |Ω ∩ B2r | < ∞ by discreteness of Ω. For any z with
|z| < r and ω with |ω| > 2r we have |ω|
2 ≤ |ω| − |z| ≤ |ω − z| so we have
X 1 k
X 1
k
≤ 2
|z − ω| |ω|k
|ω|≥2r ω∈Ω\{0}

where the latter is convergent by Proposition 2.10, giving the claim. ■

This allows us to produce elliptic functions of orders at least 3. A natural question
arises if there are elliptic functions of lower orders. By Proposition 2.13 it is clear
that elliptic functions of order 1 are not possible. Though, as it turns out, elliptic
functions of order 2 will play an important role in the theory.
We can rearrange the equation f3 (z) as
1 X 1
(3.2) f3 (z) − 3 =
z (z − ω)3
ω∈Ω\{0}

R z P has poles 1at all Ω \ {0} with residue zero. As such,

and we note that this function
we can form the integral 0 ω∈Ω\{0} (w−ω)3 dw which by convergence of the sum is
given by
X Z z 1 1 X

1 1

dw = − −
0 (w − ω)3 2 (z − ω)2 ω 2
ω∈Ω\{0} ω∈Ω\{0}
Rz 1 1
and similarly 0 w3 dw = − 2z 2 so we have by (3.2) that
Z  
1 1 X 1 1
f3 (z)dz = − 2 − − .
2z 2 (z − ω)2 ω 2
ω∈Ω\{0}

This yields the Weierstrass ℘-function.

Definition 3.3 (Weierstrass ℘-Function). Let Ω be a lattice. The Weierstrass ℘-
function of Ω is given by
1 X  1 1

℘(z) = 2 + −
z (z − ω)2 ω 2
ω∈Ω\{0}

Moreover, this function has the expected properties.

Proposition 3.4. Let Ω be a lattice. Then:
(i) ℘(z) is elliptic of order 2 and
(ii) ℘′ (z) is elliptic of order 3.
Proof of (i). ℘(z) is of order 2 by construction, with double poles at the lattice
points. We now show it is elliptic. Noting that ℘′ (z + ω) − ℘(z) = 0 we have that
℘(z + ω) − ℘(z) = Cω where Cω is a constant depending on ω ∈ Ω. For the basis
ω1 , ω2 of Ω we can consider for z = − ω2i that ℘( ω2i ) − ℘(− ω2i ) = Cωi but ℘ is even
so Cωi = 0 for i ∈ {1, 2} showing that it is elliptic. ■
Proof of (ii). This follows from the discussion above, for the Weierstrass ℘-function
arises as an integral of f3 (z) which is an elliptic function of order 3. ■
 12 WERN JUIN GABRIEL ONG

The Weierstrass ℘-function is extremely important to the study of elliptic func-

tions, since every elliptic function can be written as a rational function in ℘, ℘′ . We
first prove the following preparatory lemma.
Lemma 3.5. Let Ω be P a lattice and f (z) elliptic with respect to Ω with poles, if
any, in Ω. Then f (z) = ni=0 ai ℘(z)i for ai ∈ C.
Proof. If f (z) is a constant, we are done. Otherwise, we can take the Laurent series
expansion of f (z) around the origin which is of the form az−2n
2n + . . . since f is even.
a−2n
Now note that f (z) − ℘(z) n is even and elliptic with a pole of order at most 2n − 2.

Thus, repeating this process finitely many times, we eventually arrive at a constant
function where we can arrive at the desired claim by multiplying the expressions by
℘(z)2n to clear denominators. ■
Using the above lemma, we can in fact show that it suffices to use a rational
function in ℘(z) multiplied by ℘′ (z).
Proposition 3.6. Let Ω be a lattice and f (z) elliptic with respect to Ω. Then f (z)
Proposition 3.2 can be written as the product of ℘′ (z) and a rational function in ℘(z).
Proof. Note that if f (z) is an elliptic function of odd order, then f (z)/℘(z) is an
elliptic function of even order. So to prove the claim, it suffices to treat the case of
f (z) an elliptic function of even order.
Let f (z) be even with poles a1 , . . . , an in PΩ \ {0}. Noting that ℘(z) − ℘(ai )
vanishes for all ai , the function (℘(z) − ℘(ai ))mi f (z) willQhave pole nowhere in
PΩ \ {0} for ni sufficiently large. Then observing that f (z) ni=1 (℘(z) − ℘(ai ))mi is
an even elliptic function with poles only at lattice points. Thus by Lemma 3.5, it
can be written as a polynomial in Q℘(z) allowing us to rewrite f (z) as the quotient of
this polynomial by the product ni=1 (℘(z) − ℘(ai ))mi , that is, as a rational function
in ℘(z). ■
Remark 3.7. The expression of an ellitpic function in terms of ℘(z), ℘′ (z) is not
unique.
Consider the elliptic function ℘′ (z)2 . This is an even elliptic function of order 6
with poles on Ω. By Lemma 3.5, we can write this as a degree 3 polynomial in ℘(z),
say a0 + a1 ℘(z) + a2 ℘(z)2 + a3 ℘(z)3 with ai ∈ C. The coefficients ai , however, are
in fact highly structured and can be deduced from the lattice.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 13

4. Lecture 4 – 22nd October 2024

Following our previous dicussion, we seek to express ℘′ (z)2 the elliptic function
of order 6 in terms of a degree 3 polynomial in ℘(z) with coefficients determined by
the lattice. In particular, we have the following theorem.
Theorem 4.1. Let Ω be a lattice. Then the Weierstrass ℘-function of Ω satisfies Theorem 3.4
the nonlinear differential equation
(4.1) ℘′ (z)2 = 4℘(z)3 − g2 (Ω)℘(z) − g3 (Ω)
where g2 (Ω) = 60 ω∈Ω\{0} ω14 and g3 (Ω) = 140 ω∈Ω\{0} ω16 .
P P

Proof. We take Laurent expansions of ℘(z), ℘′ (z)2 , ℘(z)3 to observe

1
℘(z) = 2 + c2 z 2 + c4 z 4 + . . .
z
′ 4
2
℘ (z) = 6 − 8c2 z + 4c4 z 3 + . . .
z
1 1
℘(z)3 = 6 + 3c2 2 + 3c4 + . . .
z z
where we compute
1
℘′ (z) − 4℘(z)3 = −20c2 2 − 28c4 + . . .
z
and iterating the process once more to use ℘(z) write the pole of order 2 that
℘′ (z)2 − 4℘(z)3 + 20c2 ℘(z) = 28c4 + . . .
where we now note that the function on the right is a holomorphic elliptic function
and hence 28c4 itself. Rearranging, we yield ℘′ (z)2 = 4℘(z)3 − 20c2 ℘(z) − 28c4 . To
1 P 1 1
find c2 , c4 in terms of the lattice, we write h(z) = ℘(z)− z 2 = ω∈Ω\{0} (z−ω 2 ) − ω2
1
so taking the derivative we get h(m) (z) = (−1)m (m − 1)! ω∈Ω\{0} (z−ω)
P
m+2 . Now
(2m)
P 1
noting that (2m)!c2m = h (0) we conclude that c2m = (2m + 1) ω∈Ω\{0} ω2m+2
giving the claim. ■
We deduce the following as an immediate corollary.
Corollary 4.2. Let Ω be a lattice with Weierstrass function ℘(z). Then
(i) 2℘′ (z)℘′′ (z) = 12℘(z)2 ℘′ (z) − g2 (Ω).
(ii) ℘′′ (z) = 12℘(z)℘′ (z).
Proof. This follows from a direct computation of the derivative of the function (4.1)
in Theorem 4.1. ■
The field of elliptic functions with respect to a lattice Ω can be explicitly described
as follows.
Theorem 4.3. Let Ω be a lattice. Then the field of elliptic functions K(Ω) with
respect to Ω is isomorphic to
C(X)[Y ]/(Y 2 − 4X 3 + g2 (Ω)X + g3 (Ω)).
 14 WERN JUIN GABRIEL ONG

Proof. By Proposition 3.6, the map C(X)[Y ] by X 7→ ℘(z), Y 7→ ℘′ (z) is surjective

with kernel given by the relation imposed by the differential equation of Theorem 4.1.
■
In particular, this algebraically realizes the field of elliptic functions as a degree
2 extension of the field of rational functions on C.
By Proposition 2.13, we know that elliptic functions satisfy strong constraints
on their zeroes and poles. However, given a set of zeroes and poles alongside their
multiplicities, it is unclear if the inverse problem can be solved – constructing an
elliptic function with those zeroes and poles. This, however, turns out to be possible,
and will necessitate the development of the theory of elliptic modular functions.
We begin with the following lemma, which produces an example of such a function.
Lemma 4.4. Let Ω be a lattice. The function
1 X  1 1 z

f (z) = + + +
z z − ω ω ω2
ω∈Ω\{0}

is convergent with derivative −℘(z).

Rz
Proof. This function arises as the integral 0 ℘(w)dw and can be seen to be conver-
gent using the termwise estimate
1 1 z |z|2 2|z|2
+ + 2 ≤  ≤
z−ω ω ω |ω 3 | 1 − |z| |ω|3
|ω|

with the rightmost term convergent by Proposition 2.10. ■

This is the Weierstrass ζ function.
Definition 4.1 Definition 4.5 (Weierstrass ζ-Function). Let Ω be a lattice. The Weierstrass ζ-
function of Ω is given by
1 X  1 1 z

ζ(z) = + + + .
z z − ω ω ω2
ω∈Ω\{0}

Continuing along this line, we consider the the primitive of log ζ(z).
Lemma 4.6. Let Ω be a lattice. The function
z2
 
Y  z z
f (z) = z 1− exp +
ω ω 2ω 2
ω∈Ω\{0}

has derivative log ζ(z).

 2

Proof. The logarithm of the product is given by ω∈Ω\{0} log 1 − ωz + ωz + ωz 2 .
P

′ (z)
Rz  
f (z)
1
Note that this is equal to 0 (ζ(w) − w )dw = log z and thus ζ(z) − z1 = σσ(z) − z1
as desired. ■
This is the Weierstrass σ-function.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 15

Definition 4.1 Definition 4.7 (Weierstrass σ-Function). Let Ω be a lattice. The Weierstrass σ-
function of Ω is given by
z2
 
Y  z z
σ(z) = z 1− exp + .
ω ω 2ω 2
ω∈Ω\{0}

The following property of the ζ and σ functions will be used to construct an

elliptic function given a compatible set of zeroes and poles with multiplicities.
Proposition 4.8. Let Ω be a lattice and ζ(z), σ(z) the Weierstrass ζ and σ functions
with respect to Ω. Then:
(i) ζ(z + ωi ) = ζ(z) + ηi for ωi a generating element of the lattice and ηi some
constant depending on ωi .
(ii) σ(z + ωi ) = σ(z) exp(ηi z + ci ) for ωi a generating element of the lattice, ηi
the constant depending on ωi identified above, and ci some other constant
depending on ωi .
Proof of (i). We know the derivative of ζ(z+ωi )−ζ(z) is given by −℘(z+ωi )+℘(z) =
0 so integrating, we have ζ(z + ωi ) = ζ(z) + ηi for some constant ωi . ■
σ ′ (z)
Proof of (ii). From σ(z) = ζ(z) we have
σ ′ (z + ωi ) σ ′ (z)
= ζ(z + ωi ) = ζ(z) + ηi = + ηi
σ(z + ω) σ(z)
so integrating, we get σ(z + ωi ) = σ(z) exp(ηi z + ci ), as desired. ■
We can now prove the desired result.
Theorem 4.9. Let Ω be a lattice, a1 , . . . , an , an+1 , . . . , aℓ points in the interior of the
fundamental period parallelogram PΩ labeled by natural numbers m1 , . . . , mn , mn+1 , . . . , mℓ
such that
n ℓ
! !
X X
mi ai − mi ai = 0.
i=1 i=n+1
Then there exists an elliptic function f (z) with respect to Ω with zeroes at a1 , . . . , an
and poles at an+1 , . . . , aℓ of multiplicities m1 , . . . , mn and mn+1 , . . . , mℓ , respec-
tively.
Proof. The function Qn
σ(z − ai )mi
f (z) = Qℓ i=1
− ai )mi
i=n+1 σ(z
suffifces. By construction of σ(z), this function is meromorphic with the desired
zeroes and poles. To see that it is periodic, we note
f (z + ωj ) = f (z) exp(Aj )
where
n
X ℓ
X
Ai = mi (ηj (z − ai ) + cj ) − mi (ηj (z − ai ) + cj )
i=1 i=n+1
16 WERN JUIN GABRIEL ONG

which is zero modulo Ω by Proposition 4.8 (ii). ■

We conclude with the following remark.
Remark 4.10. A lattice can be normalized to have Ω = P ⟨1, τ ⟩ with τ having positive
imaginary part. Can define another function Θ(z, τ ) = n∈Z exp(πiτ n2 + 2πinz +
πin) which is known as the Θ-function of the lattice Ω. Further exposition of this
function can be found in a standard graduate text on complex analysis, but we will
not discuss these functions in the course.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 17

5. Lecture 5 – 24th October 2024

The differential equation of Theorem 4.1 suggests a close link between the theory
of elliptic functions and cubic curves. In fact, we will show that the Weierstrass
℘-function and its derivative determine a map from C/Ω → C2 with image a regular
affine cubic Definition 1.8. We now set up some requisite results.
Proposition 5.1. Let z1 , z2 ∈ C/Ω and ℘(z1 ) ̸= ℘(z2 ). Then Theorem 5.1
1 ℘′ (z1 ) − ℘′ (z2 ) 2
 
℘(z1 + z2 ) = −℘(z1 ) − ℘(z2 ) + .
4 ℘(z1 ) − ℘(z2 )
Proof. Consider the function f (z) = ℘′ (z) − a℘(z) − b for a, b ∈ C such that f (z1 ) =
0 = f (z2 ). In this case, we have ℘′ (z1 ) − a℘(z1 ) − b = ℘′ (z2 ) − a℘(z2 ) − b which
′ (z )−℘′ (z )
yields ℘℘(z 1 2
1 )−℘(z2 )
on solving for a.
Note also that f (z) is an elliptic function of order 3 since it is a C-linear combi-
nation of elliptic functions with ℘′ (z) of order 3 so by Proposition 2.13 there is some
z3 ∈ PΩ which is a zero of f satisfying z1 + z2 + z3 = 0 up to the quotient of Ω. As
such, f (−z1 − z2 ) = 0 which by substitution yields ℘′ (−z1 − z2 ) = a℘(−z1 − z2 ) + b
and thus −℘′ (z1 +z2 ) = a℘(z1 +z2 )+b by ℘′ (z) an odd function as in Proposition 3.4.
Now observe that the points (℘(z1 ), ℘′ (z1 )), (℘(z2 ), ℘′ (z2 )), (℘(z3 ), −℘′ (z3 )) lie on
the complex line v = au + b. But by the differential equation in Theorem 4.1,
℘(z1 ), ℘(z2 ), ℘(z3 ) are the zeroes of the cubic polynomial and hence satisfy the sym-
metric polynomial identity a2 = 4(℘(z1 ) + ℘(z2 ) + ℘(z3 )). So
1 ℘′ (z1 ) − ℘′ (z2 )
 
℘(z3 ) = ℘(z1 + z2 ) = −℘(z1 ) − ℘(z2 ) +
4 ℘(z1 ) − ℘(z2 )
as required. ■
We deduce the following as a corollary.
Corollary 5.2. If z ∈ C/Ω is such that 2z ∈
/ Ω then Theorem 5.1
 ′′ 2 2
1 12℘2 (z) − g2

1 ℘ (z)
℘(2z) = −2℘(z) + = −2℘(z) + .
4 ℘′ (z) 4 2℘′ (z)
Proof. The first follows from passing to the limit as z2 → z1 and the second from
the first by writing ℘′′ (z) in terms of the derivative of (4.1). ■
The construction of the embedding proceeds as follows.
Proposition 5.3. Let Ω be a lattice. Then C/Ω → C2 by z 7→ (℘(z), ℘′ (z)) is a
holomorphic parametrization of the regular affine plane cubic
E = {(u, v) ∈ C2 : v 2 = 4u3 − g2 (Ω)u − g3 (Ω)}.
Proof. The function is holomorphic on the quotient since the Weierstrass function
and its derivative have poles at the lattice points. The map is surjective since
for any (u, v) ∈ E there is z ∈ C/Ω with ℘(z) = u by Proposition 2.15 and
℘′ (z) = ±v satiesfies the equation for E by (4.1). The map is also injective since if
(℘(z1 ), ℘′ (z1 )) = (℘(z2 ), ℘′ (z2 )) we have z2 − z1 ∈ Ω since ℘(z1 ) = ℘′ (z2 ) excludes
18 WERN JUIN GABRIEL ONG

z1 + z2 ∈ Ω from Proposition 5.1 so z2 − z1 ∈ Ω, that is, they are identified in the

quotient, showing the map is injective. ■
This map to an affine regular plane cubic can be extended to a regular projective
cubic.
Proposition 5.4. Let Ω be a lattice and
E = {(u, v) ∈ C2 : v 2 = 4u3 − g2 (Ω)u − g3 (Ω)} ⊆ C2
the image of z 7→ (℘(z), ℘′ (z)). Then the closure E ⊆ CP2 has the structure of an
Abelian group with identity element the image of a lattice point.
Proof. E is given by the projective closure of E which takes elements of Ω to the
point [0 : 0 : 1], extending the group structure on the quotient C/Ω. ■
Remark 5.5. Explicitly, the sum of points p, p′ on E for an embedding φ : C → CP2
by z 7→ [1 : ℘(z) : ℘′ (z)] is given by φ(φ−1 (p) + φ−1 (p′ )).
As an aside, we show that cubics do not admit a rational parametrization.
Proposition 5.6. A regular projective cubic E does not admit a rational parametriza-
tion.
Proof. A rational parametrization would necessarily factor as a holomorphic map
b → C → E, but any map C
C b → C cannot be holomorphic. ■
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 19

6. Lecture 6 – 29th October 2024

Recall the construction of the regular affine cubic using the Weierstrass ℘-functions
Proposition 5.3. A natural question of the solution to the inverse problem arises. In
particular: given an affine regular cubic E ⊆ C2 , is there a lattice Ω such that the
Weierstrass functions of Ω parametrize E?
This can be shown using elliptic modular functions as hinted at in Section 4
preceding Lemma 4.4 as further developed in [Ahl79]. We address the easier question
of conditions of lattices that result in K(Ω), K(Ω′ ) being isomorphic. Theorem 4.3
connects this question to the former. We have the following preparatory lemmata.
Lemma 6.1. Let Ω be a lattice. If f (z) ∈ K(Ω) an elliptic function then f (z)
satisifes an algebraic differential equation.
Proof. For a lattice Ω, K(Ω) is a function field of dimension 1, that is, of transcen-
dence degree over C is 1. This implies that for any f, g ∈ K(Ω) there is an algebraic
equation F (x, y) such that F (f, g) = 0 so applying this to an elliptic function f (z)
and its elliptic derivative f ′ (z), we have the claim. ■
This extends to an algebraic relation between the elliptic functions f (z), f (z + ω)
and the constant f (ω) to yield for ω ∈ Ω an algebraic relation betewen f (z) and
f (z + ω).
Let us now consider an alternative way to define elliptic functions.
Proposition 6.2. Let q ∈ C× with 0 < |q| < 1. The set of meromorphic functions
f (z) on C× satisfying f (z) = f (qz) for all z ∈ C× forms a field Kq .
Proof. By inspection, the set of such functions is closed under addition and multi-
plication, and multiplicative inversion. ■
Let Aq,r be the annulus {z ∈ C× : r < |z| < r
|q| }.

Lemma 6.3. Let q ∈ C× with 0 < |q| < 1 and f (z) meromorphic on C× such that
f (z) = f (qz) for all z ∈ C× . Then f assumes all values in C
b already in the annulus
Aq,r . Furthermore, if f is holomorphic on Aq,r then f is constant.
Proof. If it does not assume all values of C
b then it is bounded on the closure of the
annulus and hence constant by Liouville’s theorem Theorem A.2. ■
We can thus form the torus T = C× /⟨q⟩ where ⟨q⟩ is the subgroup of C× generated
by multiples of q. This is a torus with fundamental domain represented by the
annulus, identifying the inner boundary circle with the outer.
Lemma 6.4. Let f be holomorphic on C× such that there is a constant c ̸= 0
satisfying f (qz) = c · f (z). Then c = q k for some k ∈ Z and f (z) = f (1) · z k .
P∞ n
Proof. Consider the Laurent series expansion of f (z) given by n=−N an z for
N ∈ N. We can compare coefficients
X∞ X∞
n
can z = an q n z n
n=−N n=−N
 20 WERN JUIN GABRIEL ONG

which are necessarily equal for all n, and which holds if and only if an = 0 for all
but one n = k at which we have cak z k = ak q k z k with the coefficent ak recovered as
f (1), giving the claim. ■
The subsequent lemma allows us to produce a class of meromorphic functions on
C× that are invariant under q-multiplication.
Lemma 6.5. Let q ∈ C× with 0 < |q| < 1. Then the function
∞
! ∞  !
Y Y q n
p(z) = (1 − q n z) 1−
z
n=1 n=0

is holomorphic on C× , and satisfies p(qz) = 1

− qz p(z).
Proof. By inspection, the only pole of the function is at 0, and the second statement
follows from a direct computation. ■
We can now show the main result.
Theorem 6.6. Let a1 , . . . , an and an+1 , . . . , aℓ be lists of distinct elements in the
annulus
r
Aq,r = {z ∈ C× : r < |z| < }
|q|
and m1 , . . . , mn , mn+1 , . . . , mℓ a list of positive integers. The following are equiva-
Theorem 6.2 lent:
(a) There exists a function f ∈ Kq with zeroes at a1 , . . . , an with multiplici-
ties m1 , . . . , mn and poles at an+1 , . . . , aℓ with multiplicities mn+1 , . . . , mℓ ,
respectively.
(b) ℓ = 2n and a1 . . . an q k = an+1 . . . aℓ for some integer k.
Proof. We make some preliminary observations. Note p has simple zeroes at each
positive q-power by inspection. And by passage to the inverse of a function, can
assume that counted with multiplicity there are more zeroes than poles. Setting
p(z/a1 )...p(a/am )
h(z) = p(z/a n+1 )...p(z/aℓ )
, we have that h(z) is meromorphic on C× as it is obtained
as a quotient of holomorphic functions, and we can set λ = aan+1 1 ...an
...aℓ .
(a)⇒(b) Let f ∈ Kq be given with prescribed zeroes and poles and set F (z) =
f (z)/h(z). The zeroes of h are some q-power multiplied by a so they cancel in
the annulus gainst zeroes of f , that is, F is holomorphic in the annulus. Now
F (qz) = λ1 F (z) implying F is holomorphic on C× . Applying Lemma 6.4, we get
1 k
λ q which gives the claim.  
p(z/a1 )...p(a/am )
(b)⇒(a) The function f (z) = c · z k · p(z/an+1 )...p(z/aℓ ) has the desired properties.
■
Returning to the discussion of elliptic functions, the construction above gives a
map Kq → K(Ω) by f (z) 7→ f (exp(2πiz)) where the image of p(z) is the Weierstrass
σ function up to a correction factor. We additionally remark that elliptic functions
are used to parametrize parabolas and ellipses, and are at times used in physical
mathematics.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 21

7. Lecture 7 – 31st October 2024

We turn to the study of analytic number theory, and in particular the prime
number theorem. Recall Riemann’s zeta function.
Definition 7.1 (Riemann Zeta Function). The Riemann zeta function is defined as
the infinite sum
∞
X 1
ζ(s) = .
ns
n=1

Remark 7.2. ζ(s) converges for Re(s) > 1: n−s = e−s log(n) here taking the real
logarithm, so |n−s | = n−Re(s) and the series converges – in fact, uniformly – on
Re(s) > 1.
We now define the Mellin transform which will be a crucial construction, which
will be justified by the following lemma.
Lemma 7.3. Let f : [1, ∞) → R be a locally integrable function
R∞ such that |f (x)| ≤
C · x for some constant C and k ∈ N. Then the integral 1 f (x)x−s−1 dx exists for
k

Re(s) > k.
Proof. For s such that Re(s) > k we have that |f (x)|x−s−1 ≤ C · xk−s−1 where
Re(k − s − 1) < 0 so the function is bounded above by C, giving the claim. ■
Remark 7.4. Recall that a locally integrable function is a function where the
integral over any finite interval exists.
Definition 7.5 (Mellin Transform). Let f : [1, ∞) → R be a locally integrable
function such that |f (x)| ≤ C · xk for some constant C and k ∈ N. The Mellin
transform of f is given by
Z ∞
Mf (s) = s f (x)x−s−1 dx.
1
The Mellin transform for the function x and the Gauss bracket will play a key
role in estimating the distribution of primes.
1
Lemma 7.6. The Mellin transform of f (x) = x is given by 1 + s−1 .

Proof. We compute
Z ∞ Z ∞
−s−1
s x·x dx = s x−s dx
1 1
s
= .
s−1
■
We now define the Gauss bracket and compute its Mellin transform.
Definition 7.7 (Gauss Bracket). The Gauss bracket [x] is given by the largest
integer at most x.
 22 WERN JUIN GABRIEL ONG

Lemma 7.8. The Mellin transform of the Gauss bracket f (x) = [x] is given by the
Riemann zeta function ζ(s).
Proof. We compute
Z ∞ ∞
X Z n+1
s [x]x−s−1 dx = s [x]x−s−1 dx
1 n=1 n

X∞ Z n+1
= ns x−s−1 dx
n=1 n
∞
x−s n+1
X  
= ns |
−s x=n
n=1
X∞
= n(n−s − (n + 1)−s )
n=1

note the series covnerges for Re(s) > 2 where we decompose the sum
∞
X ∞
X ∞
X
n(n−s − (n + 1)−s ) = n1−s − (n + 1 − 1)(n + 1)−s
n=1 n=1 n=1
X∞ X∞ ∞
X
= n 1−s
− n 1−s
+ n−s
n=1 n=2 n=2
= 1 + (ζ(s) − 1)
= ζ(s)
as desired. ■

From this, we can deduce the Mellin transform M[x]−x (s).

Proposition 7.9. The Mellin transform of the difference [x] − x is given by
1
M[x]−x (s) = ζ(s) − −1
s−1
on Re(s) > 1 and is meromorphic in that halfspace with pole at s = 1 with residue
Theorem 1 1.
Proof. For Re(s) > 2 the statement follows from Lemmas 7.6 and 7.8 and noting
s 1
that s−1 = s−1 + 1, and extends to Re(s) > 1 by the identity theorem Theorem A.1
as the functions agree in the right halfspace Re(s) > 2 but are holomorphic for
1
Re(s) > 1 away from the pole at 1. The function s−1 is the sole contributor to the
residue at 1 and has residue 1 by inspection. ■

The Riemann zeta function admits an alternative description via infinite products.
Theorem 7.10. The Riemann zeta function ζ(s) is equal to the infinite product
Q 1
Theorem 2 p prime 1−p−s for Re(s) > 1.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 23

Proof. The convergence of the sum ∞ 1

P
n=1 ns implies the convergence of the product
P 1
since it contains p prime ps is a subsequence. Now let Pk ⊆ P be the finite set of
the first k primes and consider the restricted product p∈Pk 1−p1 −s which is equal to
Q
Q P∞ 1
Pk m=0 (pm )s . Moreover since the product and sum are absoltuely convergent,
P 1
we can rewrite this as n∈Sk ns where S is the set of natural numbers with all
factors in Pk . Passing to the limit as k → ∞, we get the claim. ■

Remark 7.11. This already implies that there are infinitely many primes, since if
not we would have a finite product equalling a divergent harmonic series, a contra-
diction.

We also show the following structure result for zeroes of the Riemann zeta function
of real part 1.

Theorem 7.12. The Riemann zeta function ζ(s) has no zeroes with Re(s) = 1. Theorem 3

Proof. Suppose to the contrary there is a zero s with Re(s) = 1, say of the form
1 + ti. Consider the function

F (s) = ζ(s)3 ζ(s + it)4 ζ(s + 2it).

Observe that if ζ(1 + it) = 0 then at the pole of ζ(s) at s = 1 is canceled by the zero
of order four in the second factor. The function F (s) thus vanishes at s = 1 and is
holomorphic in a neighborhood around that point. Passing to logarithms, we have
that lims→1 log |F (s)| = −∞.
Now consider the product expression of Theorem 7.10 where we compute
 
X
log |ζ(s)| = Re  log(1 − p−s )−1 
p prime
 
∞
X X 1 
= Re 
n · pns
p prime n=1
∞
!
X an
= Re
ns
n=1

where the coefficient an is given by

(
1
r n = pr
an =
0 n ̸= pr
24 WERN JUIN GABRIEL ONG

where on substituting to our sum we have for s > 1 real

log |F (s)| = 3 log |ζ(s)| + 4 log |ζ(s + ti)| + log |ζ(s + 2ti)|
∞
X
an Re 3n−s + 4n−s−ti + n−s−2ti


=
n=1
∞
X an
= (3 + 4cos(t · log(n)) + cos(2t · log(n)))
ns
n=1

an
where we note that ns ≥ 0 and

3 + 4 cos(t · log(n)) + cos(2t · log(n)) = 3 + 4 cos(t · log(n)) + 2 cos2 (t · log(n)) − 1

= 2(1 + cos(t · log(n)))2

which is also positive. In particular lims→1 log |F (s)| =

̸ −∞, a contradiction. ■

We now introduce some language relating to the prime number theorem.

Definition 7.13 (Prime Counting Function). The prime counting function π(x) is
the number of primes at most x.

Definition 7.14 (Theta Function). The theta function is given by

X
ϑ(x) = log(p).
p prime,p≤x

The prime number theorem concerns the asymptotics of π(x), though it is often
easier to consider ϑ(x). These are linked by the following result.

Proposition 7.15. The existence of the following limits are equivalent

π(x)·log(x)
(a) limx→∞ x
ϑ(x)
(b) limx→∞ x
and they are equal if they exist.
 
Proof. ϑ(x)
x ≤ π(x) log(x)
x Observe
X
ϑ(x) = log(p)
p prime, p≤x
X log(x)
≤ · log(p)
log(p)
p prime, p≤x
X
= log(x) 1 = π(x) log(x).
p prime, p≤x
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 25

 
π(x) log(x) ϑ(x)
x ≤ x Conversely, we can estimate for 1 < y < x
X
π(x) = π(y) + 1
p prime, y<p≤x
1 X
≤ π(y) + log(p)
log(y)
p prime, y<p≤x
1
y+ ϑ(x).
log(y)
We now have
π(x) log(x) y log(x) log(x) ϑ(x)
≤ + .
x x log(y) x
x
Taking y = (log(x))2
we yield
π(x) log(x) 1 log(x) ϑ(x)
≤ +
x log(x) log(x) − 2 log(log(x)) x
where we have
1
lim =0
x→∞ log(x)
log(x)
lim =1
x→∞ log(x) − 2 log(log(x))

giving the first claim.

The second claim follows immediately from the inequalities. ■
26 WERN JUIN GABRIEL ONG

8. Lecture 8 – 5th November 2024

We continue towards our proof of the prime number theorem by estimating the
ratio ϑ(x)
x .
ϑ(x)
Lemma 8.1. The function is bounded on [1, ∞).
x

Proof. Consider the binomial coefficient 2k

k . Let k be a postive integer. If k < p <
2k


2k, then p| k since writing the factorial as
 
2k (2k) . . . (k + 1)
=
k k!
we have that p is in the numerator, but not canceled out by the denominator as p
is prime. In particular, this holds for each p for k < p < 2k so we have
 
Y 2k
p≤ ≤ 22k
k
k<p<2k

where the second inequality follows from 2k 2k 2k

P

r=0 r = 2 . Passing to the logarithm,
we have X
log(p) ≤ 2k · log(2).
k<p≤2k

Taking k = 2ℓ−1 for some ℓ, we obtain

X
log(p) ≤ 2ℓ+1 log(2)
p≤2ℓ

by summing inequalities of the above form. Noting for any x ∈ R≥0 an ℓ such that
2ℓ−1 < x ≤ 2ℓ we have
X
ϑ(x) ≤ log(p)
p≤2ℓ

≤ 2ℓ+1 log(2)
≤ 4x log(2)
ϑ(x)
giving x ≤ 4 log(2) hence the claim. ■

Recalling the construction of the Mellin transform from Definition 7.5, we show
the following.
Lemma 8.2. The Mellin transform Mϑ(x) (s) of the Theta function ϑ(x) is given
by
X log(p) ζ ′ (s) X log(p)
Mϑ(x) (s) = s
= − −
p ζ(s) p (ps − 1)
s
p prime p prime

for Re(s) > 1.

 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 27

Proof. We compute
Z ∞
Mϑ(x) (s) = s ϑ(x)x−s−1 dx
1
∞
X Z n+1
= s ϑ(x)x−s−1 dx ϑ constant on interval
n=1 n

X∞ Z n+1
= sϑ(n) x−s−1 dx ϑ(x) = ϑ(n) on interval [n, n + 1]
n=1 n
∞
x−s n+1
X  
= sϑ(n) |
−s s=n
n=1
X∞
ϑ(n) n−s − (n + 1)−s


=
n=1
where we note this is convergent for Re(s) > 1. Taking Re(s) > 2 we can decompose
the sum
X∞ X∞
= ϑ(n)n−s − ϑ(n)(n + 1)−s
n=1 n=1
X∞ X∞
= ϑ(n)n−s − ϑ(n − 1)n−s changing index of summation in 2nd sum
n=1 n=2
X∞
= (ϑ(n) − ϑ(n − 1))n−s
n=1
but noting (
log(n) n prime
ϑ(n) − ϑ(n − 1) =
0 otherwise
P log(p)
the above expression simplifies to p prime ps . Since this also converges for
Re(s) > 1, the identity theorem Theorem A.1 allows us to extend this function
to agree with Mϑ(x) (s) on this halfspace.
For the second equality, we consider the logarithmic derivative of ζ(s) as an Euler
product in Theorem 7.10 to observe
ζ ′ (s) X p−s log(p)
− =
ζ(s) 1 − p−s
p prime
X log(p)
=
ps − 1
p prime
with
X log(p) X log(p) X log(p)
− =
ps − 1 ps (ps − 1) ps
p prime p prime p prime
giving the claim. ■
28 WERN JUIN GABRIEL ONG

In summary, we have the following.

Proposition 8.3. The Mellin transform Mϑ(x)−x (s) of the function ϑ(x) − x is
given by
ζ ′ (s) 1 X log(p)
Mϑ(x)−x (s) = − −1− −
ζ(s) s−1 p (ps − 1)
s
p prime
for Re(s) ≥ 1.
Proof. The equality follows from Lemma 7.6 and the second equality of Lemma 8.2.
The domain of defnition can be extended to the closed halfspace Re(s) ≥ 1 since
the summation term of (8.3) converges for Re(s) > 12 and the pole of ζ(s) at s = 1
of order 1 is negated by the sum with the Mellin transform of x, so the function
admits a holomorphic extension to an open neighborhood of the closed halfspace
Re(s) ≥ 1. ■
To conclude the proof of the prime number theorem, we will provide an estimate
of this function by extending the domain of definition for the Mellin transform to
the closed halfspace Re(s) ≥ 1.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 29

9. Lecture 9 – 7th November 2024

We conclude the proof of the prime number theorem by considering a sequence
of estimation results.

R ∞ 9.1.−sLet f : [1, ∞) → R be a bounded locally integrable function and

Proposition
F (s) = 1 f (x)x dx. If F admits a holomorphic extension to a neighborhood of
the closed halfspace Re(s) ≥ 1 then Theorem 8
Z µ
f (x)
F (1) = lim dx.
µ→∞ 1 x
1
R F (s)−Fµ (s)
Proof. By Cauchy’s integral formula, we have F (a) − Fµ (a) = 2πi γ s−a ds
R µ f (x)
where Fµ (s) = 1 x dx. We consider the integral over the contour γ

s=1

γ1

δ R

γ2

traversed counterclockwise where δ > 0 sufficiently small that the complement γ2 of

the semicircular arc γ1 lies within the domain of holomorphy of F . So we have
F (s) − Fµ (1)
Z
1
F (1) − Fµ (1) = ds
2πi γ s−1
where we decompose the integral
F (s) − Fµ (1) F (s) − Fµ (s)
Z Z Z Z
Fµ (s) F (s)
(9.1) ds = ds − ds + ds.
γ s − 1 γ1 s − 1 γ2 s − 1 γ2 − 1
s

For the first two summands of (9.1), we consider the function

(s − 1)2 + R2 s−1
hµ,R (s) = µ
R2
which satisfies hµ,R (1) = 1 and consider instead
(F (s) − Fµ (s))hµ,R (s)
Z Z
Fµ (s)hµ,R (s)
ds, ds.
γ1 s − 1 γ2 s−1
30 WERN JUIN GABRIEL ONG

Now for the first summand of (9.1), we have

Z ∞
|F (s) − Fµ (s)| = f (x)x−s dx
µ
Z ∞
≤C x−Re(s) dx f (x) ≤ C
µ
C
= µ1−Re(s)
Re(s) − 1
2C
= µ1−s
(s − 1) + (s − 1)
so for Re(s) > 1 and |s − 1| = R we get
F (s) − Fµ (s) 2C
≤ µ1−s
s−1 (s − 1)2 + R2
and thus
(F (s) − Fµ (s))hµ,R (s) 2C 1−s (s − 1)2 + R2 s−1
≤ µ · µ
s−1 (s − 1)2 + R2 R2
showing
(F (s) − Fµ (s))hµ,R (s) 2C
≤ 2.
s−1 R
For the second summand of (9.1) we have on γ2
Z µ
|Fµ (s)| ≤ C x−Re(s) dx f (x) ≤ C
1
C
= (µ1−Re(s) − 1)
1 − Re(s)
C
≤ µ1−Re(s)
1 − Re(s)
and applying the same argument as before for Re(s) ≤ 1 and |s − 1| ≤ R we get
Fµ (s) 2C
≤ 2.
s−1 R
Finally, for the third summand of (9.1), we further decompose the contour γ2 into
γ2′ consisting of the segment in Re(s) ≤ 1 − δ ′ for 0 < δ ′ < δ and γ2+ , γ2− the arc
segments [1 + Ri, γ2′ (0)] and [γ2′ (1), 1 − Ri] so that the concatenation of γ2+ , γ2′ , γ2−1
is all of γ2 as previously defined. F (s) is bounded on γ2 so
 
s−1 1
F (s) + ≤ A(R, δ)
R2 s−1
for some constant A(R, δ) depending only on R and δ and on γ2′ we get the estimate
 
1 s−1 1 R · A(R, δ) RA −δ′ .
F (s) 2
+ µs−1 ds ≤ · sup µRe(s)−1 = µ
2πi γ2
′ R s−1 2 s∈γ2′ 2
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 31

′
The segments γ2+ , γ2− have length at most πδ2 so since |µRe(s)−1 | ≤ 1 we have
 
1 s−1 1 1 δ1 A
F (s) + µs−1 ds ≤ · πδ ′ A(R, δ) = .
2πi γ2′ R2 s−1 2π 2
Now for ε > 0 we can take R = 1ε and δ sufficiently small such that each of
the first two summands of (9.1) is less than ε by taking µ large and for the third
summand taking δ ′ small such that the integral over γ2′ and the sum of integrals
on γ2+ , γ2− are both at most ε by taking µ large and δ ′ small. In sum we have that
|F (1) − Fµ (1)| ≤ 4ε bounded by a function decreasing in µ, giving the claim. ■
The prime number theorem is a consequence of Proposition 9.1 and the following
lemma.
Lemma 9.2. Let f : [1, ∞) → R be a nondecreasing function such that
Z µ 
f (x) 1
lim − dx
µ→∞ 1 x2 x
f (x)
exists. Then limx→∞ x = 1.
Proof. Suppose to the contrary that the limit is not 1, then for η > 0 there is
an increasing sequence {xn }n≥0 such that | f (x n)
xn − 1| ≥ η in which case we can
assume, by rearranging the equation, that f (xn ) ≥ (1 + η)xn for all n by passing to
1+η
a subsequence. Taking ρ = 1+η/2 we have for xn ≤ x ≤ ρxn that
 η 
1+ x ≤ (1 + η)xn ≤ f (xn ) ≤ f (x)
2
with the last inequality by f increasing. So integrating on the interval [xn , ρxn ] we
get Z ρxn  
η log(ρ) f (x) 1
0< ≤ − dx
2 xn x2 x
a contradiction to existence fo the improper integra. ■
The above suffices to prove the prime number theorem.
Theorem 9.3 (Prime Number Theorem). Let π(x) be the prime counting function.
Then
π(x) · log(x)
lim = 1.
x→∞ x
Proof. By Proposition 7.15 it suffices to show limx→∞ ϑ(x) ϑ(x)
x = 1. Let f (x) = x − 1
and
Z ∞  Z ∞
ϑ(x)
F (s) = s − 1 x−s dx = s (ϑ(x) − x)x−s−1 dx = Mϑ(x)−x (s)
1 x 1
which admits a holomorphic extension to a neighborhood of the closed halfspace
Re(s) ≥ 1 by Proposition 8.3. Further noting ϑ(x) x is bounded by Lemma 8.1 and
is locally integrable, Proposition 9.1 applies and we have existence of the improper
integral Fµ (1) and the conclusion follows by Lemma 9.2. ■
32 WERN JUIN GABRIEL ONG

We now begin a discussion of complex analysis in several variables.

Consider Cn which is an n-dimensional complex vector space and thus a 2n-
dimensional real vector space.
Definition 9.4 (Orientation). An orientation on Cn is a choice of coordinates on
Cn as a 2n-dimensional real vector space.
Remark 9.5. We will use the orientation
(a1 , b1 , a2 , b2 , . . . , an , bn ) ∈ R2n
for (z1 , . . . , zn ) ∈ Cn with zj = aj + bj i. This behaves well with respect to taking
products of complex vector spaces.
Cn is a normed vector space.
Definition 9.6 (Maximum Norm). The maximum norm on Cn is given by ∥z∥max =
max1≤j≤n {|zj |C } where z = (z1 , . . . , zn ) ∈ Cn and | · |C is the standard norm on C.
Definition
qP 9.7 (Euclidean Norm). The Euclidean norm on Cn is given by ∥z∥ =
n n
j=1 |zj |C where z = (z1 , . . . , zn ) ∈ C and | · |C is the standard norm on C.

Remark 9.8. We will often just denote the standard norm | · |C on C as | · |.

Open discs plays a central role in univariate complex analysis. Its analogue in
complex analysis of several variables is the polydisc.
Definition 9.9 (Polydisc). Let τ ∈ Cn and r = (r1 , . . . , rn ) ∈ Rn>0 . The polydisc
Dr (τ ) of polyradius r is given by the set
{z ∈ Cn : |τj − zj | < rj , ∀1 ≤ j ≤ n}.
Remark 9.10. We can also define the polydisc Dr (τ ) of radius r ∈ R>0 as
{z ∈ Cn : |τj − zj | < r, ∀1 ≤ j ≤ n}.
We also have an analogue of the open ball.
Definition 9.11 (Multivariate Ball). Let τ ∈ Cn and r ∈ R>0 the ball Br (τ ) is
given by the set
{z ∈ Cn : ∥τ − z∥ < r}.
Remark 9.12. The open balls form a basis for the analytic topology on Cn .
Let us define the multivariate analogue of holomorphic functions via real and
complex differentiable functions.
Definition 9.13 (Real Differentiable). Let U ⊆ Cn be an open set and f : U → C
a function. f is real differentiable at τ ∈ U if there exist continuous functions
∆1 , . . . , ∆n , E1 , . . . , En continuous at τ such that
Xn n
X
f (z) − f (τ ) = ∆j (z)(zj − τj ) + Ej (z)(z j − τ j )
j=1 j=1
for all z ∈ U .
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 33

Definition 9.14 (Complex Differentiable). Let U ⊆ Cn be an open set and f :

U → C a function. f is complex differentiable at τ ∈ U if there exist functions
∆1 , . . . , ∆n : U → C continuous at z such that
n
X
f (z) − f (τ ) = ∆j (z)(zj − τj )
j=1
for all z ∈ U .
Definition 9.15 (Holomorphic Function). Let U ⊆ Cn be an open set and f : U →
C a function. f is holomorphic on U if f is complex differentiable at all z ∈ U .
The functions ∆j , ηj play a special role.
Definition 9.16 (Wirtinger Derivatives). Let U ⊆ Cn be an open set and f : U → C
a function. The Wirtinger derivatives of f at τ are given by ∆j (z), ηj (z) where
n
X n
X
f (z) − f (a) = ∆j (z)(zj − τj ) + Ej (z)(z j − τ j )
j=1 j=1
with ∆1 , . . . , ∆n , E1 , . . . , En : U → C continuous for all z ∈ U .
34 WERN JUIN GABRIEL ONG

10. Lecture 10 – 12th November 2024

Continuing our discussion of real and complex differentiation, we note that Wirtinger
derivatives are closely related to the multivariable Cauchy-Riemann equations.
Proposition 10.1. Let U ⊆ Cn be an open set and f : U → C a real differentiable
function of the form
n
X n
X
f (z) − f (τ ) = ∆j (z)(zj − τj ) + Ej (z)(z j − τ j ).
j=1 j=1

Then
1

∂zj f (z) = ∆j (z) = ∂xj f (z) − i · ∂yj f (z)
2
and
1

∂zj f (z) = Ej (z) = ∂xj f (z) + i · ∂yj f (z) .
2
Proof. Suppsose z1 = τ1 , . . . , zj−1 = τj−1 , zj+1 = τj+1 , . . . , zn = τn so f ′ (z) =
limzj →τj f (z)−f (τ )
zj −τj = ∆j (τ ), the second equality in the first line follows from the first
by applying the chain rule, and the second set of equalities follows by an analogous
computation. ■
As such, we can show the following theorem.
Theorem 10.2. Let U ⊆ Cn be an open set and f : U → C a function. f is
complex holomorphic if and only if it is real differentiable and satisfies the system
of partial differential equations
∂z1 f (z) = · · · = ∂zn f (z) = 0.
Proof. Proposition 10.1 expresses ∂zj f (z) in terms of Ej (z) which vanish identically
for a holomorphic function by Definition 9.14. ■
Remark 10.3. Evidently a holomoprhic function is holomorphic in each variable.
That is, for f : U → C holomorphic, the function f (τ1 , . . . , τj−1 , zj , τj+1 , . . . , τn ) is
a univariate holomorphic function in zj . Furthermore, while in real analysis there
exist functions that are differentiable in each variable but are not even continuous,
differentiability in each variable implies global differentiability in complex analysis
by Hartogs’ theorem. This is highly subtle and is beyond the scope of the course,
and an account can be found in the text of Hörmander [H90].
We can show that holomorphic functions on a domain behave well algebraically
and form a ring.
Proposition 10.4. Let U ⊆ Cn be open. The set of holomorphic functions OU is
a C-algebra that contains C[z1 , . . . , zn ].
Proof. Elements of C[z1 , . . . , zn ] are holomorphic on Cn and hence on U , and the
constants are holomorphic on U . Complex differentiable functions are preserved
sums and products, and thus so too are holomorphic functions. ■
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 35

We now state and prove the Cauchy integral formula for functions of several
complex variables. To do so, we will define integrals over the distinguished boundary
of a polydisc.
Definition 10.5 (Distinguished Boundary of Polydisc). Let Dr (τ ) be a polydisc of
polyradius r around τ . The distinguished boundary Tr (τ ) of Dr (τ ) is given by
{z ∈ Cn : |zj − τj | = rj }.
Remark 10.6. The distinguished boundary is the product of n copies of the topo-
logical circle S 1 , that is, is of real dimension n.
We can now turn to a discussion of the multivariate Cauchy integral formula.
Theorem 10.7 (Multivariate Cauchy Integral). Let U ⊆ Cn be an open set and
f : U → C a holomorphic function. If Dr (τ ) ⊆ U is a polydisc with distinguished
boundary Tr (τ ) then for all α ∈ U
Z
1 f (z)
f (α) = n
dz.
(2πi) Tr (τ ) (z1 − α1 ) . . . (zn − αn )
Proof. The n = 1 case is the univariate Cauchy integral formula. We proceed by
induction on n, supposing it holds for the case k. Consider the case k + 1 with a
function f (z1 , . . . , zk+1 ). Let g(z1 ) = f (z1 , α2 , . . . , αk+1 ) be a univariate function.
By Cauchy’s integral formula in one dimension, we have
Z
g(z1 )
f (α) = f (α1 , . . . , αk+1 ) = dw.
Tr1 (τ1 ) z1 − α1
But by the induction hypothesis, for fixed w, we have
Z
f (z1 , z2 , . . . , zn )
f (z1 , α2 , . . . , αk+1 ) =
T(r ,...,r ) ((τ2 ,...,τk+1 ))
(z 2 − α2 ) . . . (zn − αn )
2 k+1

and combining the two integrals yields the claim. ■

We can alternatively phrase Theorem 10.7 in terms of Cauchy kernels.
Definition 10.8 (Cauchy Kernel). Let U ⊆ Cn be an open set, f : U → C a
continuous function, and Dr (τ ) ⊆ U is a polydisc with distinguished boundary
Tr (τ ). Then the Cauchy kernel is given by
Z
1 f (w)
Cf (z) = n
dw1 . . . dwn .
(2πi) Tr (τ ) (w1 − z1 ) . . . (wn − zn )
Remark 10.9. By Theorem 10.7, the Cauchy kernel agrees with f if f is holomor-
phic.
The Cauchy kernel allows us to deduce that holomorphic functions of several
variables are infinitely differentiable.
Proposition 10.10. Let U ⊆ Cn be an open set containing a polydisc Dr (τ ) with
distinguished boundary Tr (τ ) and f : U → C a holomorphic function. Then f
admits holomorphic partial derivatives of all orders. Theorem 1.2
 36 WERN JUIN GABRIEL ONG

Proof. Differentiating the Cauchy kernel under the integral sign, we have
Z
ν ν1 ! . . . νn ! f (w)
∂z Cf (z) = dw . . . dwn
(2πi) n
Tr (τ ) (w1 − z1 )
ν 1 +1 . . . (wn − zn )νn +1 1
giving the claim. ■
We will now consider analogues of key results in complex analysis in the multi-
variate setting. This discussion will require the following lemma which often allows
us to reduce to the single variable case.
Lemma 10.11. Let α, β ∈ Cn and λ : C 7→ Cn by t 7→ α + tβ. If f is holomorphic
on U ⊆ Cn open then f ◦ λ is holomorphic on λ−1 (U ).
Proof. We compute
n
d X
(f ◦ λ)(t) = ∂zj (λ(t))βj
dt
j=1
which is a βj -weighted sum of holomorphic functions and hence holomorphic on
λ−1 (U ) = f |L . ■
We can now prove the multivariate analogues of the identity theorem and maxi-
mum principle.
Theorem 10.12 (Multivariate Identity). Let f be holomorphic on a domain G and
Theorem 1.5 identically zero on a nonempty open subset U of G. Then f ≡ 0 on G.
Proof. Let L be a line the image of λ through G and consider f ◦ λ the induced
holomorphic function of one variable. By the identity theorem in one variable Theo-
rem A.1, f is identically zero on all of L. Writing G as the union of all lines passing
through it gives the claim. ■
Theorem 10.13 (Multivariate Maximum Modulus). Let f be holomophric on a
Theorem 1.6 domain G and |f | has local maximum at τ ∈ G. Then f (z) = f (τ ) is constant.
Proof. Arguing as before, let L be a line the image of λ through G and consider
f ◦ λ the induced holomorphic function of one variable. By the maximum maximum
modulus principle in one variable, f is constant on all of L. Writing G as the union
of all lines passing through it gives the claim. ■
We now turn to a consequence of Theorem 10.7 and some surprising consequences.
Proposition 10.14. Let Dr (τ ) ⊆ Cn for n ≥ 2. If f is holomorphic in a punctured
neighborhood of the origin in Dr (τ ), then f is holomorphic on Dr (τ ).
Proof. By translation, it suffices to consider the case of a function f holomorphic on
D1 (0) \ D1/2 (0) in Cn , here considering polydiscs of fixed radius. By the univariate
Cauchy integral formula we have
Z
1 f (z)
f (α) = dz
(2πi)n Tr (τ ) (z1 − α1 ) . . . (zn − αn )
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 37

but fixing any αj the Cauchy integral formula in one variable holds but αj is arbi-
trary so the function extends in each variable to all of D1 (0) and so it does over-
all. ■
This implies that holomorphic functions have zeroes and thus poles along a subset
of Cn of codimension at least 1.
Corollary 10.15. Let f be a holomorphic function on U ⊆ Cn open. Then {z ∈
Cn : f (z) = 0} and {z ∈ Cn : f (z) = ∞} are non-isolated.
Proof. The poles of f are non-isolated by Proposition 10.14 and thus so too zeroes
are non-isolated under passage to 1/f (z). ■
With the language of holomorphic functions in hand, we can discuss holomorphic
maps.
Definition 10.16 (Holomorphic Map). Let U ⊆ Cn , V ⊆ Cm be open sets and
f = (f1 , . . . , fm ) : U → V be a continuous function. f is a holomorphic map if each
component function fk : U → C is holomorphic.
The chain rule generalizes to the multivariate setting.
Proposition 10.17 (Multivariate Chain Rule). Let U ⊆ Cn , V ⊆ Cm be open sets
and f = (f1 , . . . , fm ) : U → V and g : V → C holomorphic functions. Then
m
X ∂g ∂f
∂zj (g ◦ f ) = .
∂wk ∂zj
k=1

Proof. This is immediate from the chain rule and the Cauchy-Riemann equations.
■
38 WERN JUIN GABRIEL ONG

11. Lecture 11 – 14th November 2024

Recall the analytic continuation result of Proposition 10.14. We will build up to
showing a more general analogue that allows us to extend a holomorphic function
on the complement of a compact set U \ K ⊆ Cn to U ⊆ Cn known as Hartogs’
Kugelsatz. Hartogs’ original proof relied on Cauchy’s integral formula paired with
intricate geometric arguments. We provide a simplified proof that uses more tech-
nology. In particular, we continue our discussion of holomorphic maps.
Definition 11.1 (Holomorphic Jacobian). Let U ⊆ Cn , V ⊆ Cm be open sets and
f : U → V a holomorphic map. The holomorphic Jacobian is defined to be the
matrix
 ∂f1 ∂f1 
∂z1 (τ ) . . . ∂zn (τ )
JfHol (z) =  ... .. ..  ∈ Cm×n

. . 
∂fm ∂fm
∂z1 (τ ) ... ∂zn (τ )
and τ ∈ U .
We consider some examples.
Example 11.2. Let U ⊆ Cn be open and idU : U → U be the holomorphic identity
Hol is the identity matrix.
map. The holomorphic Jacobian Jid U

Formation of the Jacobian is a linear operation.

Proposition 11.3. Let U ⊆ Cn , V ⊆ Cm , W ⊆ Ck be open sets and f : U → V, g :
V → W holomorphic maps. Then Jg◦fHol = J Hol × J Hol .
g f

Proof. We compute, denoting the coordinates on V by w1 , . . . , wm , we have

 ∂g1 ∂g1   ∂f1 ∂f1 
∂w1 (f (τ )) . . . ∂wm (f (τ )) ∂z1 (τ ) . . . ∂zn (τ )
JgHol (f (τ )) × JfHol (τ ) = 
 .. .. ..   .. .. .. 
. . .  . . . 
∂gk ∂gk ∂fm ∂fm
∂w1 (f (τ )) . . . ∂wm (f (τ )) ∂z1 (τ ) . . . ∂zn (τ )
Pm ∂g1 ∂fr P m ∂g1 ∂fr 
r=1 ∂wr (f (τ )) · ∂z1 (τ ) . . . r=1 ∂wr (f (τ )) · ∂zn (τ )
=
 .. .. .. 
. . . 
Pm ∂gk ∂fr Pm ∂gk ∂fr
r=1 ∂wr (f (τ )) · ∂z1 (τ ) . . . r=1 ∂wr (f (τ )) · ∂zn (τ )
 
∂(g1 ◦f )
(τ ) . . . ∂(g∂z1 n◦f ) (τ )
 ∂z1. .. .. 
= .. . .


∂(gk ◦f ) ∂(gk ◦f )
∂z1 (τ ) . . . ∂zn (τ )
= JfHol
◦g (τ )

as desired. ■

We can define biholomorphisms which will be the appropriate notion of invertible

maps in complex analysis of several variables.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 39

Definition 11.4 (Biholomorphism). Let U ⊆ Cn , V ⊆ Cm and f : U → V a

Definition 2.1 holomorphic map. f is a biholomorphism if f −1 exists and is holomorphic.
A complex function f : U → V can be written as u + iv where u, v : R2n → R2m .
We can use this to produce the real Jacobian of the complex-valued function f .
Definition 11.5 (Real Jacobian). Let U ⊆ Cn , V ⊆ Cm and f : U → V a differen-
tiable map. For f = u + iv with u, v : R2n → R2m , the real Jacobian of f is defined
as  
Ju,x (τ ) Ju,y (τ )
R
Jf (τ ) = ∈ C2m×2n .
Jv,x (τ ) Jv,y (τ )
We can also define the complex Jacobian using Wirtinger derivatives and their
conjugates.
Definition 11.6 (Complex Jacobian). Let U ⊆ Cn , V ⊆ Cm and f : U → V a
differentiable map. The complex Jacobian of f is defined as
 
Jf,z (τ ) Jf ,z (τ )
JfC (τ ) = ∈ C2m×2n .
Jf,z (τ ) Jf ,z (τ )
The holomorphic, real, and complex Jacobians are related in the following way.
We omit the linear-algebraic proofs.
Proposition 11.7. Let U ⊆ Cn , V ⊆ Cm and f : U → V a differentiable map.
Then rank(JfR (τ )) = rank(JfC (τ )) and det(JfR ) = det(JfC ).
If f is holomorphic and in particular a biholomorphism, we can say more.
Proposition 11.8. Let U ⊆ Cn , V ⊆ Cm and f : U → V a biholomorphism. Then
det(JfC ) = | det(JfHol )|. Proposition 2.2

We will show that a holomorphic map is a biholomorphism when it is bijective

on sets. This is not true for differentiable maps. We set out some preparations for
the theorem.
Proposition 11.9. Let U ⊆ Cn , V ⊆ Cm and f : U → V a biholomorphism such
that JfHol (z) is invertible for all z. Then f is biholomorphic. Proposition 2.4

Proof. Note that JfR (z) is everywhere regular showing the existence of a real dif-
ferentiable inverse f −1 by the implicit function theorem. It remains to show f −1
is
 holomorphic.
 We can compute the complex Jacobian of f −1 which we write as
A B
and since f is holomorphic we have that its complex Jacobian is block
C D
 
M 0
diagonal , we have the product
0 M
  
M 0 A B
= id
0 M C D
showing B, C = 0 and thus f −1 is holomorphic. ■
 40 WERN JUIN GABRIEL ONG

We make some further recollections from analysis and differential topology.

Theorem 11.10 (Implicit Function). Let U ⊆ Cn , V ⊆ Cm and f : U × V → Cm
Hol (0) is regular
a holomorphic map. If f (0, 0) = 0 and the holomorphic Jacobian Jf,w
in the variables w on V then there exist open subsets U1 ⊆ U, V1 ⊆ V and a
Proposition 2.5 holomorphic map ϕ : U1 → V1 such that f (z, w) = 0 if and only if w = ϕ(z).
Theorem 11.11 (Submanifold). Let A ⊆ Cn , a ∈ A, 1 ≤ k ≤ n. The following are
Proposition 2.6 equivalent:
(a) There is a neighborhood U of a, a neighborhood V of 0 ∈ Ck , and a holo-
morphic homeomorphism ϕ : V → Cn such that ϕ(V ) = A ∩ U and JϕHol is
of rank k.
(b) There exists a neighborhood U of a and a holomorphic function f : U →
Cn−k such that U ∩ A = {z ∈ U : f (z) = 0} and JfHol is of rank n − k.
(c) There exists a neighborhood U of a and W ⊆ Cm of 0 and a holomorphic
map ϕ : U → W such that ϕ(A ∩ U ) = {w ∈ W : wk+1 = · · · = wm = 0}.
In analogy to differential topology, we have the following definition.
Definition 11.12 (Locally Analytic Submanifold). Let A ⊆ Cn , a ∈ A, 1 ≤ k ≤ n.
A is a locally analytic submanifold of dimension k if one and thus all of the following
Definition 2.2 conditions hold:
(a) There is a neighborhood U of a, a neighborhood V of 0 ∈ Ck , and a holo-
morphic homeomorphism ϕ : V → Cn such that ϕ(V ) = A ∩ U and JϕHol is
of rank k.
(b) There exists a neighborhood U of a and a holomorphic function f : U →
Cn−k such that U ∩ A = {z ∈ U : f (z) = 0} and JfHol is of rank n − k.
(c) There exists a neighborhood U of a and W ⊆ Cm of 0 and a holomorphic
map ϕ : U → W such that ϕ(A ∩ U ) = {w ∈ W : wk+1 = · · · = wm = 0}.
This definition globalizes.
Definition 11.13 (Analytic Submanifold). Let A ⊆ Cn , a ∈ A, 1 ≤ k ≤ n. A is an
analytic manifold of dimension k if it is a locally analytic submanifold of dimension
Definition 2.3 k for all a ∈ A.
We will require the subsequent lemma to prove the desired result.
Lemma 11.14. Let U ⊆ Cn open, f : U → C a holomorphic function, and M =
{z ∈ U : f (z) = 0} ⊆ U . If M is nonempty, there is a point a ∈ M such that M is
Lemma 2.7 a locally analytic manifold of dimension n − 1.
Proof. We proceed by cases. Let a ∈ M and suppose ∇f (a) ̸= 0. Then Theo-
rem 11.11 (a) applies showing it is a locally analytic submanifold. Otherwise, we
can take sufficiently many derivatives such that the derivative is nonzero – all deriva-
tives being zero implies the function is identically zero. Now let Λ = {r ∈ N : Dα f ≡
0, |α| ≤ r}. This set is finite since the function is zero if all derivatives vanish. Now
choose λ ∈ Λ maximal so that there is a point a ∈ M and Dβ f with |β| = λ and
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 41

∇(Dβ (f ))(a) ̸= 0. So M0 = {a ∈ M : Dβ f (a) = 0} is a locally analytic submanifold

of dimension n − 1. We show that M = M0 close to a. By Theorem 11.11, we
can take new coordinates on M0 such that M0 = {zn = 0} and a = 0 and consider
a holomorphic homeomorphism ϕ : V → Cn for V an open neighborhood of a by
ϕ(z ′ , zn ) where ϕ(0, zn ) ̸= 0 for zn ̸= 0 which has a zero of order k in zn = 0.
By continuity of zeroes for ramilies of holomorphic functions, ϕ(z ′ , zn ) must have k
zeroes for z ′ close to 0. But M ⊆ M0 implies that all these zeroes are in M0 so M
satisfies zn = 0 and is thus a locally analytic manifold of dimension n − 1. ■
42 WERN JUIN GABRIEL ONG

12. Lecture 12 – 19th November 2024

We complete the proof of Osgood’s theorem alluded to in Section 11.
Theorem 12.1 (Osgood). Let U ⊆ Cn , V ⊆ Cm and f : U → V a bijective
holomorphic map. Then f is a biholomorphism.
Proof. We proceed by induction on dimension of the source. If n = 1 then the open
mapping theorem implies that f is an open map so f −1 is defined, continuous, and
bijective. V (f ′ ) is discrete if it is nonempty and the Riemann extension theorem
implies that the inverse can be extended to a holomorphic function.
Assume the statement holds for maps from open subsets of Cn−1 to V ⊆ Cm . We
show that JfHol (z) is either regular or nonzero. Let τ be a point where JfHol (τ ) is
not the zero matrix. Up to permuting the coordinates in both source and target,
we can assume that the lower right corner of the matrix is nonzero. Using the
implicit function theorem Theorem 11.10, we can introduce new coordinates w1 =
z1 , . . . , wn−1 = zn−1 and wn = fm (z1 , . . . , zn ) so that f = (g1 , . . . , gm−1 , wn ) where
the gj are functions of w1 , . . . , wn−1 . We can now consider
fe = (g1 (w1 , . . . , wn−1 , 0), . . . , gm−1 (w1 , . . . , wm−1 , 0))
which is bijective and by the inductive hypothesis has J Hol (0) regular with nonzero
fe
determinant. This gives us an expression of the Jacobian determinant of f as a
block diagonal with J Hol
e (0) in the upper left corner and 1 in the lower right corner.
f
But these determinants agree, so JfHol (0) is nonzero.
Now consider h(z) = det(JfHol (z)) which is a holomorphic function in U . Assume
V (f ) ⊆ M is nonempty where M = V (JfHol (z)). By Lemma 11.14, there is τ suc
that M is an n − 1 dimensional local analytic submanifold so JfHol (z) = 0 for all
z ∈ M ∩ Uτ where Uτ is a sufficiently small neighborhood of τ . So f is constant on
M ∩ Uτ and thus not injective, a contradiction. ■
We now recall the definition and some elementary properties of differential forms.
Proofs can be found in [Lee13] for the real theory and [Lee24] for the complex theory.
In
Definition 12.2 (Tangent Vector). Let x0 ∈ Rn , S(x0 ) the set of functions in a
neighborhood of x0 , and D(x0 ) the subset of differentiable functions. A tangent
vector is a map D : D(x0 ) → R is a tangent vector if D(1) = 0 and D(gf ) = 0 for
all g ∈ S(x0 ), f ∈ D(x0 ) such that f (x0 ) = g(x0 ) = 0.
These maps naturally form a vector space.
Proposition 12.3. Let x0 ∈ Rn . If D is a tangent vector and f, g ∈ D(x0 ) then
D(f g) = g(x0 )D(f ) + f (x0 )D(g). Furthermore, the tangent vectors form a vector
space with basis ∂x∂ 1 , . . . , ∂x∂n .
This leads to the following definitions.
Definition 12.4 (Tangent Space). Let x0 ∈ Rn . The tangent space Tx0 Rn is the
R-vector space of tangent vectors at x0 .
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 43

Definition 12.5 (Cotangent Space). Let x0 ∈ Rn . The cotangent space (Tx0 Rn )∨

is the dual of the tangent space.
Since the tangent space is itself a vector space, we can define linear forms on it.
Definition 12.6 (Total Differential). Let x0 ∈ Rn and f ∈ D(x0 ) a differentiable
function at x0 . The total differential d is the map Tx0 Rn → R by D 7→ D(f ).
Remark 12.7. Total differentials of coordinate functions dxj form a basis of (Tx0 Rn )∨
dual to the basis ∂x∂ j of Tx0 Rn . As such, each cotangent vector v can be written
as ni=1 ai (x)dxi where ai (x) are functions on Rn . A cotangent vector is continu-
P
ous/differentiable/integrable if all the constituent ai s are so.
Definition 12.8 (Pfaffian Form). Let M ⊆ Rn be an open subset. A Pfaffian form
is the data of a cotangent vector Dx0 ∈ (Tx0 M )∨ for all x0 ∈ M .
The Pfaffian forms naturally form a module E 1 (M ) over the
Vp ring of functions
0
E (M ). Taking p
the exterior algebra, we can define E (M ) = 1
E (M ) and write
E(M ) = np=0 E p (M ).
L

n
Vp 1 Let M ⊆ R be open. A differential p-form
Definition 12.9 (Differential p-Form).
p
on M is an element of E (M ) = E (M ).
Note a p-form can be written as
X
fi1 ,...,ip · dxi1 ∧ dxi2 ∧ · · · ∧ dxip
1≤i1 <i2 <···<ip ≤n

with fi1 ,...,ip ∈ E 0 (M ) so we can define the differential of a p-form as the p + 1-form
X
dfi1 ,...,ip · dxi1 ∧ dxi2 ∧ · · · ∧ dxip .
1≤i1 <i2 <···<ip ≤n

These transform with respect to maps Rn → Rm as follows.

Proposition 12.10. Let U ⊆ Rn , V ⊆ Rm be open and f : U → V be a continuous
map. Let g ∈ E p (V )
X
g= gi1 ,...,ip · dxi1 ∧ dxi2 ∧ · · · ∧ dxip
0≤i1 <i2 <···<ip ≤m

and X
g◦f = (gi1 ,...,ip ◦ f ) · dgi1 ∧ dgi2 ∧ · · · ∧ dgip .
0≤i1 <i2 <···<ip ≤m
If f and the gi1 ,...,ip are differentiable then d(g) ◦ f = d(g ◦ f ).
Of special concern to us will be differential n-forms in n-dimensional objects.
Definition 12.11 (Integration over Differential Form). Let f be an n-form on a
measurable subset M of Rn . Then
Z Z
f= f1,...,n (x)dx
M M
where f = f1,...,n (x) · dx1 ∧ dx2 ∧ · · · ∧ dxn .
44 WERN JUIN GABRIEL ONG

In the complex setting the constructions and results carry over verbatim. How-
ever, instead of merely treating differentiable functions, we also consider holomor-
phic functions and thus holomorphic differential forms. These are defined as follows.
Definition 12.12 (Holomorphic Differential Form). Let U ⊆ Cn be open. A dif-
ferential 1-form f on U is holomorphic if ∂zj f · dzj = 0 for all j and
n
X n
X
f= ∂zj f · dzj + ∂zj dzj .
j=1 j=1
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 45

13. Lecture 13 – 21st November 2024

We consider power series in several complex variables.
Definition 13.1 (Multivariate Complex Power Series). The multivariate complex
power series ring is the ring C[z1 , . . . , zn ] with f ∈ C[[z1 , . . . , zn ]] of the form
X
f= cα z α .
α∈Zn
≥0

Remark 13.2. Here we use multi-index notation, where cα is a constant depending

on α = (α1 , . . . , αn ) and z α = z1α1 . . . znαn .
Remark 13.3. We can also write f as an infinite sum of homogeneous polynomials
of fixed degree.
Convergence is defined as follows.
P
Definition 13.4 (Series Convergence). Let α∈Zn cα is an infinite sum and τ ∈
≥0
Cn . The sum converges to τ if for all ε > 0, there exists a finite subset J ⊆ Zn≥0
such that
X
τ− cα < ε
α∈J0
for all J0 ⊆ J. Definition 3.1
Definition 13.5 (Formal Power Series Convergence). Let f ∈ C[[z1 , . .P . , zn ]]. f is a
convergent power series if there exists τ ∈ Cn \ {0} such that the series α∈Zn cα τ α
≥0
is convergent in the sense of Definition 13.4.
Definition 13.6 (Ring of Convergent Power Series). The ring of convergent power
series is the subring C{z1 , . . . , zn } ⊆ C[[z1 , . . . , zn ]] consisting of convergent power
series. Definition 3.2
We can provide a sufficient criterion for convergence.
Proposition 13.7. Let f (z) = α∈Zn cα z α be a formal complex power series and
P
≥0
τ ∈ Cn \ {0} such that |cα τ α | is bounded. Then f (z) is uniformly convergent for all
z ∈ B|τ | (0). Proposition 3.1
Proof. For any z ∈ B|τ | (0), we have |z| = λ|τ | for some 0 < λ < 1. Thus by
boundedness of |cα z α | ≤ |cα τ α | · λ|α| and the claim follows from α∈Zn λ|α| being
P
≥0
convergent. ■
More generally, we can write holomorphic functions as a convergent power series
in the following way.
Theorem 13.8. Let f be a holomorphic function on a polydisc Dr (τ ). Then
X ∂αf
f (z) = (z − τ )α .
n
α!
α∈Z≥0

for z ∈ Dr (τ ). Theorem 3.3

 46 WERN JUIN GABRIEL ONG

1 P (z−τ )α
Proof. We have that z 7→ w−z can be written as a geometric series α∈Zn (w−τ )α+1
≥0
which is uniformly convergent for w on the boundary of a proper sub-polydisc by
Proposition 13.7. Now applying Cauchy’s integral formula Theorem 10.7, we can
interchange integration and summation to yield
Z !
X 1 f (w)
f (z) = n α+1
dw (z − τ )α
n
(2πi) ∂Tr′ (τ ) (w − τ )
α∈Z≥0

but the integral is precisely the partial derivative, which was the claim. ■
We now build up to the definition of the structure sheaf on Cn .
Definition 13.9 (Germ of Functions). Let z ∈ Cn . The ring of germs Oz at z
is given by equivalence classes of functions f holomorphic in a neighborhood of z
under the pointwise operations where f ∼ g if and only if there exists an open set
V ⊆ Uf ∩Ug of the neighborhoods Uf , Ug on which f, g are holomorphic, respectively,
Definition 3.3 on which f |V = g|V .
We can define the structure sheaf as the disjoint union of these rings.
Definition
` 13.10 (Structure Sheaf). The structure sheaf O on Cn is given by
Definition 3.4 & 3.6 z∈Cn Oz .
There is an evident map p : O → Cn by (f, z) 7→ z.
Definition 13.11 (Continuous Section of Structure Sheaf). Let U ⊆ C. A contin-
uous section of O over U is a continuous function σ : U → O such that σ(z) ∈ Oz
Definiton 3.5 and there exists a holomorphic function f on U such that σ(z) = fz for all z ∈ U .
Remark 13.12. The definition above allows us to endow O with the structure of
a topological space with basis elements given by σ(U ) for U ⊆ Cn
Let us return to more general considerations of holomorphic functions.
Proposition 13.13. Let U ⊆ Cn be open, K ⊆ C an annulus around the origin
of internal and external radii r, R, respectively, and f : U × K → C a holomorphic
function. There exist unique holomorphic functions f0 on U × DR (0) and f∞ on
Proposition 4.1 U × (C \ Dr (0)) such that f = f0 + f∞ and limw→∞ f∞ (z, w) = 0.
Proof. We can write
Z Z
1 f (z, u) 1 f (z, u)
f (z, w) = du − du
2πi |u|=R′ u−w 2πi |u|=r′ u−w
for r′ < |w| < R′ which are holomorphic in z and w in |w| < R′ and |w| > r′ ,
respectively. We can take
(
f0 (z, w) |w| < R
fe(z) =
−f∞ (z, w) |w| > r
which is an entire function that goes to 0 as w → ∞. Thus fe is an entire function
limiting to 0 as w → ∞ showing uniqueness. ■
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 47

We can also count the number of zeroes with respect to w.

Proposition 13.14. Let U ⊆ Cn be open, K ⊆ C an annulus around the origin
of internal and external radii r, R, respectively, and f : U × K → C a holomorphic
function. If f has no zeroes in K then the number of zeroes of f (z, w) as a function
of w is independent of z.
Proof. The number of zeroes of f (z, w) as a function of w is given by the integral
Z
1 fw (z, w)
dw
2πi ∂Dr (0) f (z, w)
which is continuous in z and integer-valued, and thus constant. ■
We conclude with a statement highly reminiscent of the Weierstrass preparation
lemma, which we will soon prove.
Proposition 13.15. Let U ⊆ Cn be open, K ⊆ C an annulus around the origin
of internal and external radii r, R, respectively, and f : U × K → C a holomorphic
function with k zeroes in w. If f has no zeroes in K then there exists a nonzero
function c on U and a holomorphic function h on U × K such that
wk eh(z,w) = c(z)f (z, w).
f (z,w)
Proof. For each z ∈ U , the function w 7→ wk
= g(z, w) admits a holomorphic
gw (z,w)
logarithm with respect to w. Thus has a holomorphic logarithm with respect
g(z,w)
R w gu (z,u)
and we have h(z, w) = u g(z,u) du on U ×K which is holomorphic on K and depends
holomorphically on z, w.
We can take the derivative with respect to w given by
g(z, w)eh(z,w) gg(z,w)
w (z,w)
− eh(z,w) gw (z, w) eh(z,w)
=
g(z, w)2 g(z, w)
and thus for any z ∈ U there is a constant c(z) such that eh(z,w) = c(z)g(z, w) which
on substituting g(z, w) = f (z,w)
wk
yields the claim. ■
 48 WERN JUIN GABRIEL ONG

14. Lecture 14 – 26th November 2024

We now show the Weierstrass preparation theorem.
Theorem 14.1 (Weierstrass Preparation). Let U ⊆ Cn be open and DR (0) = D ⊆
C a disc. If f : U × D → C is holomorphic of and of order k with respect to the
coordinate w on D then there are functions e(z, w) holomorphic nonvanishing on
U × D and ω(z, w) holomorphic on U × D of the form wk + ak−1 (z)wk−1 + · · · +
a1 (z)w + a0 (z) such that f (z, w) = e(z, w) · ω(z, w).
Proof. Take r such that f |U ×K on the annulus K = K(r, R) satisfies the hypothe-
ses for Proposition 13.15. As such, taking h(z, w), c(z) such that wk eh(z,w) =
c(z) · f (z, w). Now applying the Laurent decomposition Proposition 13.13 we get
h(z, w) = h0 (z, w) + h∞ (z, w) with h0 , h∞ holomorphic on |w| < R, |w| > r, respec-
tively, and limw→∞ h∞ (z, w) = 0.
We have
wk eh0 (z,w)+h∞ (z,w) wk eh0 (z,w) h∞ (z,w)
f (z, w) = = e .
c(z) c(z)
eh0 (z,w)
Taking e(z, w) = c(z) , we have

f (z, w) = e(z, w)eh∞ (z,w) .

Moreover, since limw→∞ h∞ (z, w) = 0, limw→∞ eh∞ (z,w) = 1 for fixed z. We can
thus write
∞
X
eh∞ (z,w) = 1 + am (z)w−m
m=1
and decomposing the sum we have
k ∞
!
X X
−m
k h∞ (z,w)
w e =w k
am (z)w + wk · am (z)w−m
m=1 m=k+1
= ω(z, w) + R∞ (z, w)
P 
k −m of the desired form. To prove
where the summand ω(z, w) = wk a
m=1 m (z)w
the result, it suffices to show that R∞ (z, w) is of the desired form.
Now we have
f (z, w)
0= − ω(z, w) − R∞ (z, w) = R0 (z, w) − R∞ (z, w)
e(z, w)
where by uniqueness of Laurent decompositions on the annulus as ω(z, w) has the
same zeroes as f . ■
We now consider some special types of functions in the convergent power series
ring C{z1 , . . . , zn }.
Definition 14.2 (Regular of Fixed Order). Let f ∈ C{z1 , . . . , zn }. f is zn -regular
Definition 4.1 of order k if f (0, zn ) is not identically zero and has a zero of order k at zn = 0.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 49

The Weierstrass theorem shows that zeroes of holomorphic functions are highly
structured.
Lemma 14.3. If f a holomorphic function on an open neighborhood U around
the origin and vanishing at the origin in Cn+1 , then there exists a linear change of
coordinates T such that f ◦ T is zn regular of some order. Lemma 4.5
Proof. Consider a complex line L on which f is not identically zero. Setting this
line L as the w-coordinate, we get that f vanishes at w = 0. ■
Example 14.4. f (z1 , z2 ) is neither z1 nor z2 -regular. But under the transformation
z1 = u1 , z2 = u1 + u2 , we have that f (u1 , u2 ) = u21 + u1 u2 is u1 -regular of order 2.
Lemma 14.5. Let f be a zn -regular function of order k. There are constants
0 < r′ , 0 < r < R such that f converges on Dr′ (0) × DR (0) ⊆ Cn−1 × C, has no
zeroes outside the disc Dr′ (0) × Dr (0) ⊆ Cn−1 × C, and f (z ′ , zn ) is of order k with
respect to zn . Lemma 4.6
Proof. This is a direct application of Theorem 14.1 as the function f (0, zn ) has an
order k zero at the origin. ■
More generally, we can define Weierstrass polynomials as follows.
Definition 14.6 (Weierstrass Polynomial). A Weierstrass polynomial of order k is
a function in (z1 , . . . , zn−1 , zn ) = (z ′ , zn ) given by
ω(z ′ , zn ) = znk + ak−1 (z ′ )znk−1 + · · · + a1 (z ′ )zn + a0 (z ′ )
where ak (0) = 0.
We will require the following result about zn -regular functions of order k in the
following exposition.
Theorem 14.7 (Weierstrass Formal). Let f be a zn -regular function of order k.
For each g ∈ C{z1 , . . . , zn }, there exists a unique decomposition g = qf + r where
q ∈ C{z1 , . . . , zn } and r ∈ C{z1 , . . . , zn−1 }[zn ] of degree smaller than k in zn . Theorem 4.8
Proof. By the Weierstrass preparation theorem, we can write f = eω for ω a
Weierstrass polynomial of order k. Up to setting g = g/e, we can take f to be
a Weierstrass polynomial. Now consider g/ω and let r, R, r′ be such that g/ω
is holomorphic on Dr′ (0) × K(r, R), the product of a disc and annulus. By the
Laurent decomposition proposition 13.13, we can write ωg = q + h where q, h are
holomorphic in Dr′ (0) × DR (0) and Dr′ (0) × K(r, R), respectively. Now recall-
ing ω is given k
P∞ as zn + ak−1 (z ′ )znk−1 + · · · + a1 (z ′ )zn + a0 (z ′ ), we have that for Once again denot-
h(z , zn ) = m=1 cm (z )zn−m , the product ωh is given by
′ ′ ing z ′ for the coordi-
∞ nates z1 , . . . , zn−1 .
X
′ ′
ωh = b1 (z )znk−1 + · · · + b0 (z ) + bm (z ′ )zn−m .
m=1
P∞
Writing r(z ′ , zn ) = b1 (z ′ )znk−1 + · · · + b0 (z ′ ), R(z ′ , zn ) = m=1 bm (z
′ )z −m ,
we have
n
g = qω + r(z ′ , zn ) + R(z ′ , zn ) and it remains to show the latter summand vanishes.
50 WERN JUIN GABRIEL ONG

Rewriting this 0 = −g + qω + r(z ′ , zn ) + R(z ′ , zn ), we have that −g + qω + r

is holomorphic on Dr′ (0) × DR (0) and R(z ′ , zn ) is holomorphic outside Dr (0) and
hence the above is the Laurent decomposition of the zero function showing R(z ′ , zn )
vanishes. Uniqueness of q, r follows from uniqueness of the Laurent decomposition.
■
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 51

15. Lecture 15 – 28th November 2024

We have the following strengthening of the Weierstrass preparation theorem in
the cases where the function f lies in C{z1 , . . . , zn−1 }[zn ].
Proposition 15.1. Let U ⊆ Cn be open and DR (0) = D ⊆ C a disc and f ∈
C{z1 , . . . , zn−1 }[zn ]. If f : U × D → C is holomorphic of and of order k with respect
to the coordinate w on D then there are functions e(z, w) ∈ C{z1 , . . . , zn−1 }[zn ]
holomorphic nonvanishing on U × D and ω(z, w) holomorphic on U × D of the form
wk + ak−1 (z)wk−1 + · · · + a1 (z)w + a0 (z) such that f (z, w) = e(z, w) · ω(z, w).
Proof. By Theorems 14.1 and 14.7 we have f = eω = qω + r so r = 0, e = q with
uniqueness and f is of the desired form. ■
We now consider some algebraic properties of the ring of convergent power series.
Theorem 15.2. The ring of convergent power series C{z1 , . . . , zn } is Noetherian.
Proof. We proceed by induction on the number of variables n. This holds for n = 0
as C is a field and hence Noetherian.
Now suppose C{z1 , . . . , zn−1 } is Noetherian and thus C{z1 , . . . , zn−1 }[zn ] is Noe-
therian. Consider a ⊆ C{z1 , . . . , zn−1 }[zn ] a nonzero ideal and f ∈ a. By Lemma 14.5,
we can change coordinates such that f is zn -regular. Now for g ∈ C{z1 , . . . , zn },
Theorem 14.7 implies that we can write g = qf + r with r ∈ C{z1 , . . . , zn−1 }[zn ] of
degree less than k. This yields a homomorphism of C{z1 , . . . , zn−1 }-modules with
kernel f given by C{z1 , . . . , zn } 7→ C{z1 , . . . , zn−1 }⊕k by g 7→ (a0 (z ′ ), . . . , ak−1 (z ′ ))
where r = ak−1 (z ′ )znk−1 + · · · + a0 (z ′ ). Let a be the image of a under this map. Now
noting that C{z1 , . . . , zn−1 }⊕k is Noetherian as a module a is generated by finitely
many elements g1 , . . . , gs , the images of g1 , . . . , gs ∈ a. But this shows that g is a
linear combination of g1 , . . . , gs , f for g ∈ a so this forms a finite generating set of
the ideal showing that the ring is Noetherian. ■
We can also show that the ring of convergent power series C{z1 , . . . , zn } is fac-
torial, that is, a unique factorization domain. Doing so will require the following
proposition.
Proposition 15.3. Let U ⊆ Cn open containing the origin and f, g ∈ O(U ). If
f0 , g0 are relatively prime in O0 then there is a neighborhood V of the origin such
that fz , gz are relatively prime for all z ∈ V .
In turn:
Theorem 15.4. The ring of convergent power series C{z1 , . . . , zn } is a unique fac-
torization domain.
We now discuss meromorphic functions.
Definition 15.5 (Sheaf of Meromorphic Functions). The sheaf of meromorphic
n
`
functions K on C is given by z∈Cn Kz where Kz is the ring of germs of meromor-
phic functions at z.
Sections of K give rise to meromorphic functions in the same way as Defini-
tion 13.11.
 52 WERN JUIN GABRIEL ONG

16. Lecture 16 – 3rd December 2024

Recall that meromorphic functions are constructed locally, so a meromorphic
function on an open set U ⊆ Cn as a section of the sheaf of Definition 15.5 is
given by the data of triples {(Ui , fi , gi )}i∈I where {Ui }i∈I form an open cover of
U , fi , gi ∈ O(Ui ) with gi not identically zero and satisfying the conditions that
fi,z
( fgii )z = gi,z ∈ Kz for z ∈ Ui and fi gj = fj gi on Uij = Ui ∩Uj . In particular, the value
of a meromorphic function h is defined at z if its representative fgzz has nonvanishing
denominator. This in turn allows us to define the polar set of a meromorphic
function.
Definition 16.1 (Polar Set). Let U ⊆ Cn be open and h ∈ K(U ). The polar set
Mh of h in U is the set of points of U on which h is undefined.
Remark 16.2. If h is already holomorphic, then its polar set Mh is empty.
We now define the notion of an analytic hypersurface and show the polar set of
a single meromorphic function is an analytic hypersurface.
Definition 16.3 (Analytic Hypersurface). Let U ⊆ Cn be open. An analytic hy-
persurface is the vanishing locus of a holomorphic function f ∈ O(U ) in U .
We can now show the desired result.
Proposition 16.4. Let U ⊆ Cn be open and h ∈ K(U ) be a meromorphic function.
The polar Mh is an analytic hypersurface in U .
Proof. Without loss of generality, let V ⊆ U be sufficiently small that h = fg with
f, g ∈ O(V ). We have that Mh ∩ V = {g = 0} since if fz , gz are relatively prime
for all z ∈ U and a ∈ {g = 0} with h holomorphic at a implies ha ga = fa , a
contradiction to relatively primiality. ■
We now discuss a number of extension and convexity properties which is well-
known in the univariate setting using the tools of Hadamard factorization and
Blaschke products. In particular, we seek to characterize the domains of existence
of holomorphic functions on these domains, and to produce holomorphic and mero-
morphic functions with prescribed properties.
Definition 16.5 (Analytic Set). Let U ⊆ Cn be open. M ⊆ U is an analytic set if
there exists a presentation of M as the vanishing locus of finitely many holomorphic
functions.
Remark 16.6. In particular, M will be closed and for all z ∈ M there is a neigh-
borhood V and f1 , . . . , fk ∈ O(V ) such that V ∩ M = {f1 = · · · = fk = 0}.
One of the most elementary examples of an extension proerty is the extension
over analytic sets.
Theorem 16.7 (First Riemann Extension). Let U ⊆ Cn be open, M ⊊ U an
analytic set, and h holomorphic on U \ M and locally bounded on U . Then h
Proposition 1.1 extends to a holomorphic function on all of U .
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 53

Proof. Without loss of generality suppose 0 ∈ M and h = fg with {g = 0} ⊆ M

so g is a Weierstrass polynomial ω(z ′ , zn ). Now consider the univariate function
zn 7→ h(z ′ , zn ) for fixed z ′ . ω(z ′ , zn ) has finitely many zeroes so zn 7→ h(z ′ , zn ) is
bounded at these zeros and thus extends holomorphically over these zeroes given by
Cauchy integrals that vary holomorphically in z ′ . ■
We now introduce the Bochner-Martinelli kernel, which will play a key role in
what follows.
Definition 16.8 (Bochner-Martinelli Kernel). The Bochner-Martinelli kernel is
given by Definition 1.2
n
(n − 1)! X 1
β(w, z) = dw1 ∧ dw1 ∧ · · · ∧ dwj−1 ∧ dwj ∧ dwj+1 ∧ · · · ∧ dwn ∧ dwn .
(2πi)n j=1
∥wj − zj ∥2n

Remark 16.9. That is, in the expression of Definition 16.8, the wedge-factor dwj
is omitted, making the Bochner-Martinelli kernel a form of type (n, n − 1, 0, 0) in
the dw, dw, dz, dz’s, respectively.
We prove an elementary property of the Bochner-Martinelli kernel.
Lemma 16.10. Let U ⊆ Cn be a bounded open set with smooth boundary and τ ∈
U and f continuously differentiable on ∂U . Then d(f (w)β(w, z)) = ∂f (w) ∧ β(w, z).
Proof. Note that f (w)β(w, τ ) contain the holomorphic forms dwj so we can compute
d(f (w)β(w, τ )) = ∂(f (w)β(w, τ ))
= ∂f (w) ∧ β(w, τ )
n
wj − τj
 
(n − 1)! X
+ f (w) ∂ w j
dw1 ∧ dw1 ∧ · · · ∧ dwn ∧ dwn
(2πi)n j=1 ∥w − τ ∥2n

but
n n 
|wj − τj |2
  
X wj − τj X 1
∂wj = − n
∥w − τ ∥2n ∥w − z∥2n ∥w − τ ∥2n+2
j=1 j=1
=0
giving the claim. ■
 54 WERN JUIN GABRIEL ONG

17. Lecture 17 – 5th December 2024

We prove the Bochner-Martinelli integral formula.

Theorem 17.1 (Bochner-Martinelli Integral Formula). Let G ⊆ Cn be a bounded

domain with smooth boundary. If f is a continuously differentiable function on the
boundary ∂G then
Z Z
f (τ ) = f (w)β(w, τ ) − ∂f (w) ∧ β(w, τ )
∂G G

Theorem 1.2 for all τ ∈ G.

Proof. For fixed τ ∈ G and r > 0 sufficiently small that Br (τ ) ⊊ G. We have by

Stokes’ theorem and Lemma 16.10
Z Z Z
f (w)β(w, τ ) − f (w)B(w, z) = d(f (w)β(w, τ ))
∂G ∂Br (τ ) G\Br (τ )
Z
= ∂f (w) ∧ β(w, τ ).
G\Br (τ )

Note further that

Z Z
lim ∂f (w) ∧ β(w, τ ) = ∂f (w) ∧ β(w, τ )
r→0 G\Br (τ ) G

so we can rearrange the equations above to observe

Z Z Z
f (w)β(w, τ ) = f (τ ) β(w, τ ) − (f (w) − f (τ ))β(w, τ ).
∂Br (τ ) ∂Br (τ ) ∂Br (τ )
R
To prove the result, we show ∂Br (τ ) β(w, τ ) = 1 and that the second summand of
the expression above vanishes.
For the first claim, we apply Stokes’ theorem once more, and recall that the
n
volume of Br (τ ) is πn! r2n and compute
n
(n − 1)! X wj − zj
Z Z
β(w, τ ) = dw1 ∧ dw1 ∧ · · · ∧ dwj−1 ∧ dwj ∧ dwj+1 ∧ · · · ∧ dwn ∧ dwn
∂Br (τ ) ∂Br (τ ) (2πi) j=1 ∥w − τ ∥2n
n

n
(n − 1)!
Z X
= (wj − zj )dw1 ∧ dw1 ∧ · · · ∧ dwj−1 ∧ dwj ∧ dwj+1 ∧ · · · ∧ dwn ∧ dwn
r2n (2πi)n ∂Br (τ ) j=1
 
n
(n − 1)!
Z X
= 2n d (wj − zj )dw1 ∧ dw1 ∧ · · · ∧ dwj−1 ∧ dwj ∧ dwj+1 ∧ · · · ∧ dwn ∧ dwn 
r (2πi)n Br (τ ) j=1

(n − 1)!
Z
= n · dw1 ∧ dw1 ∧ · · · ∧ dwj ∧ dwj+1
r2n (2πi)n Br (τ )
(n − 1)!
Z
= n(2i)n dV
r2n (2πi)n Br (τ )
=1
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 55

For the second claim, the computation above yields

!
(n − 1)!
Z
n · dw1 ∧ dw1 ∧ · · · ∧ dwn ∧ dwn
r2n (2πi)n Br (τ )

(n − 1)!
Z X
+ 2n ∂wj f (w)(wj − τj ) · dw1 ∧ dw1 ∧ · · · ∧ dwn ∧ dwn .
r (2πi)n Br (τ ) j=1

By boundedness of U , |f (w) − f (τ )| ≤ C∥w − τ ∥ and |∂wj f (w)(wj − zj )| ≤ C∥w −

z∥ for some large constant C. So for all w ∈ ∂Br (τ ), the quantities |f (w) −
f (τ )|, |∂wj f (w)(wj − zj )| are bounded above by Cr so
!
(n − 1)!
Z Z
(f (w) − f (τ ))β(w, τ ) ≤ 2n 2 n · 2n CrdV = 2Cr
∂Br (τ ) r (2π)n Br (τ )

which vanishes as r → 0 proving the claim. ■

Remark 17.2. Note that the Bochner-Martinelli kernel is not a holomorphic form.
The Bochner-Martinelli integral formula allows us to prove Hartogs’ so-called
“Kugelsatz,” allowing us to extend holomorphic functions over compact domains.
For this, we will require the following lemmata, only one of which we prove.
Lemma 17.3. Let G ⊆ Cn be a domain and β(w, z) the Bochner-Martinelli kernel
on G and f continuously differentiable on ∂G. Then ∂ z β(w, z) = −∂ w β1 (w, z) where Lemma 1.5
1 n
 
n−2 1
β1 (w, z) = (n − 1)β(w, z) ∧ (∂ w β) ∧ ∂z β · .
2πi ∥w − z∥2n
Proof. See [Ran86, §4, Prop. 4.9]. ■
Lemma 17.4. Let G ⊆ Cn and f holomorphic on ∂G. Then
Z
F (z) = f (w)B(w, z)
∂G
is holomorphic in z for z ∈
/ ∂G. Lemma 1.6

Proof. We compute
Z
∂ z F (z) = f (w)∂ z B(w, z)
∂G
Z
=− f (w)∂ w β1 (w, z)
Z∂G
=− ∂ w (f (w)β1 (w, z))
∂G
Z
=− dw (f (w)β1 (w, z))
∂G
=0
as desired. ■
In turn:
56 WERN JUIN GABRIEL ONG

Theorem 17.5 (Hartogs’ Kugelsatz). Let U ⊆ Cn be an open, K ⊆ U compact,

and U \ K connected. If n ≥ 2 and f is holomorphic on U \ K then f extends
holomorphically to U .
Proof. Let G ⊊ U be a domain and G0 a subdomain of G fully containing K. By
hypothesis f is holomorphic in G\G0 and by the Bochner-Martinelli integral formula
Theorem 17.1, we have
Z Z
f (z) = f (w)β(w, z) − f (w)β(w, z).
∂G ∂G0
Denoting the integrals F1 (z), F2 (z), respectively, F1 is holomorphic for all z ∈
/ ∂G by
Lemma 17.4 and F2 holomorphic for all z ∈ / G0 . In particular, F1 is also holomorphic
on all of G and thus holomorphic in K and F2 (z) → 0 as |z| → ∞.
Since n ≥ 2, there exist hyperplanes that do not meet G0 on which F2 is zero,
showing F2 is zero by the identity theorem Theorem 10.12 and F1 is a holomorphic
extension of f since it coincides with f on G \ G0 once again using the identity
theorem. ■
We now return to a consideration of the multivariate Cauchy-Riemann equations,
in particular considering their role in producing holomorphic functions correspond-
ing to fixed differential form data.
The motivating problem is the existence of holomorphic functions satisfying Cousin-
I distributions.
Definition 17.6 (Cousin-I Distribution). Let U ⊆ Cn be open. A Cousin-I dis-
tribution on U is given by {(Ui , fi )}i∈I where {Ui }i∈I form an open cover of U ,
fi ∈ K(Ui ) holomorphic such that fi − fj is holomorphic on Uij = Ui ∩ Uj .
A solution to the Cousin-I problem is a meromorphic function compatible with
the data of a Cousin-I distribution in the following sense.
Definition 17.7 (Solution to Cousin-I Problem). Let U ⊆ Cn be open and {(Ui , fi )}i∈I
be a Cousin-I distribution on U . A solution to the Cousin-I problem is a meromor-
phic function f ∈ K(U ) such that f |Ui − fi is holomorphic for all i ∈ I.
Solutions to problems of this type are constructed by finidng a smooth solution
and “altering” it to a holomorphic solution by solving a ∂-equation.
We discuss a sequence of reductions for this problem, first recalling the following
notions from the theory of smooth manifolds.
Definition 17.8 (Partition of Unity). Let U ⊆ Rn be open and {Ui }i∈I a locally
finite open cover of U such that each Ui is relatively compact in U . A partition
of unity {ψi }i∈I subordinate to the coverP
{Ui }i∈I is the data of smooth functions
ψi : Ui → R such that supp(ψi ) ⊆ Ui and i∈I ψi (x) = 1 for all x ∈ U .
Partitions of unity exist for open sets of Cn .
Proposition 17.9. Let U ⊆ Rn be open and {Ui }i∈I a locally finite open cover of
U such that each Ui is relatively compact in U . There exists a partition of unity
{ψi }i∈I subordinate to the cover {Ui }i∈I .
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 57

A proof of Proposition 17.9 can be found in [Lee13, §2]. Note in particular that
holomorphic phenomena are in particular smooth which justifies their use in what
follows.
Proposition 17.10. Let U ⊆ Cn be open and {(Ui , fi )}i∈I , {(Vj , gj )}j∈J be Cousin-
I distributions such that their union is a Cousin-I distribution. If f ∈ K(U ) is a
solution to the Cousin-I distribution {(Ui , fi )}i∈I then f is a solution to the Cousin-I
distribution {(Vj , gj )}j∈J as well.
Proof. Let f ∈ K(U ) be a solution to the Cousin-I problem for the distribution
{(Ui , fi )}i∈I and {ψj }j∈J be a partition of unity subordinate to the cover {Vj }j∈J .
We want to show that f |Vj − gj ∈ O(Vj ) for all j ∈ J. For this, we note that
Vj admits a cover by {Ui ∩ Vj }i∈I on which fi |Ui ∩Vj − gj |Ui ∩Vj is holomorphic so
fi ψj − gj is holomorphic on all Vj and similarly f − fi |Ui ∩Vj is holomorphic for all
Ui ∩Vj and all i ∈ I with fixed j ∈ J which extends holomorphically to f |Vj −fi ψj on
Vj But their sum is f |Vj − gj and holomorphic as it is the finite sum of holomorphic
functions by local finiteness of the cover. ■
This motivates the following definition.
Definition 17.11 (Equivalent Cousin-I Distributions). Let U ⊆ Cn be open and
{(Ui , fi )}i∈I , {(Vj , gj )}j∈J be Cousin-I distributions. {(Ui , fi )}i∈I , {(Vj , gj )}j∈J are
equivalent Cousin-I distributions if their union is a Cousin-I distribution.
This reduces solving the Cousin-I problem to the following.
Proposition 17.12. Let U ⊆ Cn be open and f ∈ C ∞ (U ), f = nj=1 fj dzj , and
P

there exists u a smooth function such that ∂u = f and ∂f = 0. Then the Cousin-I
problem admits a solution. Proposition 1
P
Proof. Let {(Ui , fi )}i∈I be a Cousin-I distribution as above and set gi = j∈I ϕj fji
which is defined on Ui . Note that on Uik the difference gi − gk is given by fik as
expected and that ∂gi − ∂gj = 0. We can define function f as in the hypothesis of
the theorem by taking f |Ui = ∂gi which statisfies the hypothesis by inspection and
admits a solution u to the Cousin-I problem. ■
58 WERN JUIN GABRIEL ONG

18. Lecture 18 – 10th December 2024

We first show that problems of the type in Proposition 17.12 can be solved. These
are known as distributions of type (0, 1) since it is a linear combination of degree 1
antiholomorphic forms.
TheoremP18.1. Let G ⊆ Cn be a bounded domain with smooth boundary for n ≥ 2
n
and f = j=1 fj dzj such that ∂f = 0. Then
Z
u(z) = − f (w) ∧ B(w, z)
G
is smooth and satisfies ∂u = f .
Proof. By boundedness of G, the integral defining u exists and is given by a C-linear
combination of
fj (w)(wj − zj )
Z Z
fj (w)(z + y)yj
uj (z) = 2n
dV (w) = dV (y)
G ∥w − z∥ G ∥y∥2n
with the second equality following by translation and denoting y = w − z. Now
letting Lz be a differential operator with constant coefficients, we have
Lw fj (w)(wj − zj )
Z Z
Lz fj (z + y)yj
Lz u = 2n
dV (w) = dV (w).
G ∥y∥ G ∥w − z∥2n
R
Taking u(z) = − G f (w) ∧ B(w, z), we have u(z) as a C-linear combination of the
uj ’s and thus u is smooth. It remains to show that ∂u = f for which it suffices that
∂zj u = fj . But we get
Z
∂ z1 u = − ∂w1 f ∧ B(w, z)
G
Z
=− (∂w1 f1 · dw1 + · · · + ∂w1 f1 · dwn ) ∧ B(w, z)
G
Z
= − ∂f1 ∧ B(w, z)
= f1 (z)
with the final equality following from Theorem 17.1. Analogous computations hold
for ∂zj u. Moreover, u is holomorphic outside a large ball so u → 0 as ∥z∥ → ∞ so
u vanishes identically away from the origin, showing compact support of u. ■
This result in fact implies Theorem 17.5 and shows that Theorem 18.1 which does
not hold for n = 1.
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 59

19. Lecture 19 – 12th December 2024

We want to consider the ∂-equation more generally.
Theorem 19.1. Let G ⊆ Cn be a bounded domain with smooth boundary, f ∈
C 1 (G), and Z
1 f (w)
u(z) = dw ∧ dz.
2πi G w − z
Then ∂u = f on G. Theorem 3.2
Proof. Let τ ∈ G and Dr (τ ) ⊆ G. Consider an indicator function χ : G → R such
that (
1 z ∈ Dr (τ )
χ(z) =
0 z∈ /U
for U an open neighborhood of Dr (τ ). We then compute
(1 − χ(w))f (w)
Z Z
1 χ(w)f (w) 1
u(z) = dw ∧ dw + dw ∧ dw.
2πi C w − z 2πi G w−z
by Theorem 17.1. In particular the second summand is holomorphic in z. Letting
teh first summand be u1 (z) and z ∈ Dr (τ ) we have ∂u = ∂u1 = χ(z)f (z) = f (z) on
G. ■
We now show that each form of type (0, q) arises as the antiholomorphic differ-
ential of a form of type (0, q − 1). This is the statement of the Dolbeault lemma.
We proceed using the following lemma.
Lemma 19.2. Let Dr (τ ) be a polydisc and f a smooth (0, q)-form on Dr (τ ) such
that ∂f = 0. Then there exists a smooth (0, q − 1)-form u on Dr′ (τ ) with r′ < r
having relatively compact closure such that ∂u = f .
Proof. Let Dr′′ (τ ) ⊆ Dr′ (τ ) and Dr′ (τ ) ⊆ Dr (τ ) be polydiscs with relatively com-
pact closure. Define Ek to be the set of smooth (0, s)-forms
 
 X 
Ek = f = fJ dz J : fJ = 0 if l ∈ J, l > k .
 
|J|=s

We induct over k.
Note if f ∈ E0 then f is identically 0 and each form lies in En . For the base case,
observe that if f is a (0, 1)-form on Dr′ (τ ) with ∂f = 0 wtih f ∈ E0 we are done.
Now suppose the assumption holds for Ek−1 where we can write
X
f= fJ dz J = dzk ∧ g + h
|J|=s

with g, h ∈ Ek−1 . We can compute to observe 0 = ∂f = −dzk = −dz k ∧ ∂g + ∂h

where in particular dzk ∧ ∂g = ∂h. Now let
X
g= gT dz T .
|T |=q−1
 60 WERN JUIN GABRIEL ONG

We have that gT are holomorphic in zk+1 , . . . , zn since if ∂zl gT ̸= 0 or some l > k

then dzl is one of the summands of ∂g and thus dz K ∧∂g has dz K ∧dzl as a summand
but ∂h does not, giving the claim. Now using Theorem 19.1, we can set geT to be
the (0, 1)-forms for the problem ∂egT = gT which are holomorphic in zk+1 , . . . , zn by
the preceding discussion. We then set
X
ge = geT dz T
|T |=s−1

and compute
n
X X
∂e
g= ∂zl geT · dzl ∧ dz T
|T |=s−1 l=1

= h1 + ∂zk ∧ g
with g, h ∈ Ek−1 so taking u such that ∂u = h − h1 we have f = ∂g + ∂u as
desired. ■
We can now show the Dolbeault lemma in general.
Theorem 19.3 (Dolbeault Lemma). Let Dr (τ ) ⊆ Cn be a polydisc, f a smooth
(0, q)-form on Dr (τ ) such that ∂f = 0. Then there exists a smooth (0, q − 1)-form
Theorem 3.3 u on Dr (τ ) such that ∂u = f .
Proof. Let Dr0 (τ ) ⊆ Dr1 (τ ) ⊆ . . . be a sequence of polydiscs each Drt (τ ) having
compact closure in Drt+1 (τ ) and Dr (τ ) = ∞
S
t=0 rt ). We proceed by induction on
D (τ
the index t.
By Lemma 19.2, there exist functions u0 smooth on Dr0 (τ ) satisfying ∂u0 = f on
Dr0 (τ ). Now suppose that there are functions u0 , . . . , ul which are (0, q − 1)-forms
on Drλ (τ ) for 0 ≤ λ ≤ l such that ∂uλ = f on Drλ (τ ) and uλ = uλ−1 on Drλ−2 (τ ).
We can then construct u′l+1 on Drl+1 (τ ) such that ∂u′l+1 = f, ∂(u′l+1 − ul ) = 0 on
Drl (τ ).
We can thus find h on Drl−1 (τ ) such that ∂h = u′l+1 +ul with h smooth on Drl+1 (τ )
and taking ul+1 = u′l+1 − ∂h we have ∂ul+1 = f . We can then take u = liml→∞ ul
which is a solution on all of Dr (τ ). The case q = 1 follows by estimating using the
Taylor development and bounding the growth differences on polydiscs. ■
 ADVANCED TOPICS IN COMPLEX ANALYSIS – BONN, WINTER 2024/25 61

Appendix A. Basic Results in Complex Analysis

In this appendix, we collect some basic results of complex analysis, largely fol-
lowing the text of Stein and Shakarchi [SS03].
Theorem A.1 (Identity). Let D ⊆ C be a domain and f, g holomorphic functions
on D. If the set {z ∈ D : f (z) = g(z)} contains an accumulation point, then
f (z) = g(z) for all z ∈ D.
Theorem A.2 (Liouville). If f (z) is a holomorphic function such that |f (z)| ≤ M
for some M ∈ R≥0 then f is constant.
62 WERN JUIN GABRIEL ONG

References
[Ahl79] Lars Ahlfors. Complex analysis. An introduction to the theory of analytic functions
of one complex variable. English. 3rd edition, reprint of the 1979 edition. Vol. 385.
AMS Chelsea Publ. Providence, RI: American Mathematical Society (AMS), 2021.
isbn: 978-1-4704-6767-8.
[H90] Lars Hörmander. An introduction to complex analysis in several variables. English.
3rd revised ed. Vol. 7. North-Holland Math. Libr. Amsterdam etc.: North-Holland,
1990. isbn: 0-444-88446-7.
[Lee24] John M. Lee. Introduction to complex manifolds. English. Vol. 244. Grad. Stud.
Math. Providence, RI: American Mathematical Society (AMS), 2024. isbn: 978-
1-4704-7695-3; 978-1-4704-7782-0; 978-1-4704-7781-3. doi: 10.1090/gsm/244.
[Lee13] John M. Lee. Introduction to smooth manifolds. English. 2nd revised ed. Vol. 218.
Grad. Texts Math. New York, NY: Springer, 2013. isbn: 978-1-4419-9981-8; 978-
1-4419-9982-5. doi: 10.1007/978-1-4419-9982-5.
[Ran86] R. Michael Range. Holomorphic functions and integral representations in several
complex variables. English. Vol. 108. Grad. Texts Math. Springer, Cham, 1986.
[SS03] Elias M. Stein and Rami Shakarchi. Complex analysis. English. Vol. 2. Princeton
Lect. Anal. Princeton, NJ: Princeton University Press, 2003. isbn: 0-691-11385-8.
Universität Bonn, Bonn, D-53113
Email address: [email protected]
URL: https://wgabrielong.github.io/