Lecture Notes

Algebraic Geometry C

Course held by Notes written by

Prof. Marco Franciosi Francesco Paolo Maiale

Department of Mathematics
University of Pisa
June 13, 2017

1
 Disclaimer
These notes came out of the Algebraic Geometry C course, held by Professor Marco Franciosi in
the second semester of the academic year 2016/2017.
They include all the topics that were discussed in class; I added some remarks, simple proof, etc..
for my convenience.
I have used them to study for the exam; hence they have been reviewed patiently and carefully.
Unfortunately, there may still be many mistakes and oversights; to report them, send me an email
to frank094 (at) hotmail (dot) it.

2
Contents

1 Introduction 5
1.1 Motivating Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Introduction to Riemann Surfaces 8

2.1 Main Definitions and Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Projective Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.1 Multiplicities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.2 Resolution of the Singularities . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 First Examples Riemann Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

3 Functions and Maps 20

3.1 Functions on Riemann Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Holomorphic Maps Between Riemann Surfaces . . . . . . . . . . . . . . . . . . . . . 22
3.3 Global Properties of Holomorphic Maps . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4 The Degree of a Holomorphic Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.5 Hurwitz’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4 More Examples of Riemann Surfaces 29

4.1 Application of Hurwitz’s Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.2 Automorphism of Riemann Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.3 Group Actions on Riemann Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

5 Differential Forms 38
5.1 Holomorphic 1-forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
5.2 Integration of a 1-form along paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

6 Sheaf Theory 44
6.1 Definitions and First Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
6.2 Exact Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
6.3 Čech Cohomology of Sheaves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.4 Sheaves of OX -modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
6.5 GAGA Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.6 Invertible OX -module Sheaves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
6.7 Operations on Sheaves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3
7 Divisors 55
7.1 Divisors on Riemann Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
7.2 Invertible Sheaf of OX -modules associated to a divisor D . . . . . . . . . . . . . . . 56
7.3 Linear Systems of Divisors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
7.4 Divisors and Maps to Projective Space . . . . . . . . . . . . . . . . . . . . . . . . . . 60
7.5 Inverse Image of Divisors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.6 Canonical Divisor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.7 Riemann-Hurwitz Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

8 The Riemann-Roch Theorem and Serre Duality 69

8.1 Rough Estimate of h0 (X, OX [D]) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
8.2 Riemann-Roch Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
8.3 Serre Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
8.3.1 Mittag-Leffler Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
8.3.2 Proof of Serre Duality Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . 76
8.4 The Equality of the Three Genera . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
8.5 Analytic Interpretation (Hodge) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

9 Applications of Riemann-Roch Theorem 83

9.1 Very Ample Divisors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
9.2 Algebraic Curves and Riemann Surfaces . . . . . . . . . . . . . . . . . . . . . . . . . 88
9.2.1 Equivalence Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
9.3 Existence of Globally Defined Meromorphic Functions . . . . . . . . . . . . . . . . . 96
9.4 Low Degree Divisors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
9.5 Canonical Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
9.6 Riemann-Roch: Geometric Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
9.7 Clifford Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

10 Abel’s Theorem 109

10.1 Picard Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
10.2 Jacobian of X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
10.3 Abel-Jacobi Map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
10.4 Abel-Jacobi Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

4
Chapter 1

Introduction

1.1 Motivating Examples

21
Let f : R → R be the function defined by the formula f (x) := x2 + x3 . We want to evaluate the
integral between zero and minus one of f , that is,
Z 0
I= f (x) dx.
−1

Set x := t2 − 1; the differential is given by dx = 2t dt, and, if we change variables in the integral
above, then it turns out that Z 0
I= 2t2 (t2 − 1) dt,
−1
and this can be easily computed by standard means. This substitution does not come out of nowhere;
indeed, we can consider the curve in R2 defined by C : y 2 = x3 + x2 , and take the parametrization
given by a beam of lines originating from the point (0, 0).
More precisely, let us consider the family of lines
1,6

rt : y = tx,
1,2

intersecting C in the origin. Clearly, it is singular point (of

0,8
order 2) in the intersection; thus by Bezout’s theorem there
0,4
exists a unique p ∈ C such that rt ∩ C = {0, p}, and the order
of p is 1. Consequently, we have
-2,4 -2 -1,6 -1,2 -0,8 -0,4 0 0,4 0,8 1,2 1,6 2 2,4
( (
y 2 = x3 + x2 x = t2 − 1
=⇒
-0,4
y = tx y = t (t2 − 1),
-0,8
i.e., the substitution used above: x = t2 − 1.
-1,2 In other words, we used the same method (rational
parametrization through lines) as in rational, trigonometric
-1,6
integral formulas, derived in the first year of calculus.
Figure 1.1: y 2 = x3 + x2 Remark 1.1. The curve C is rational, i.e., it is birational to
A1 (or to P1 if we consider t ∈ R ∪ {∞}). More precisely, the
subset Γ := {(x, f (x)) : x ∈ R} ⊂ R2 can be parametrized by
rational functions of t.

5
Definition 1.1 (Abelian Integrals). We say that
Z
R (w, z(w)) dw (1.1)
γ

is an abelian integral if γ : [0, 1] → C is a path, R is a rational function and w and z(w) satisfy a
polynomial relation P (w, z) = 0.
Example 1.1. Let us consider the relation z = f (w) = w2 . For every z̃ 6= 0, ∞ there exist
w1 , w2 ∈ C such that f (w1 ) = f (w2 ) = z̃.
More precisely, outside of z̃ = 0, ∞, we have a double cover of C, thus
it makes sense to consider two copies of the complex plane C1 ∼
1,6
= C and
C2 ∼= C, such
1,2
that C i corresponds to {wi }.
However f −1 (0) = {0} and there exists an action of monodromy in a
0,8

neighborhood of 0, i.e., any closed path of base point z̃ switch the two
0,4

roots (w1 7→ w2 , w2 7→ w1 ). Thus, the key idea is to modify (slightly)

-2,4 -2 -1,6 -1,2 -0,8 -0,4 0 0,4 0,8 1,2 1,6 2 2,4

C1 and C2 in such a way as to, by monodromy, pass from w1 to w2 .

-0,4

Consider the half line (from 0 to ∞) given by

-0,8

-1,2 {Re(w) ≥ 0, Im(w) = 0} ⊂ C,

-1,6

and cut both C1 and C2 along it. At this point we can enlarge both cuts,
Figure 1.2: z(w) = w 2 and glue together along the corresponding edges, obtaining a surface
homeomorphic to the sphere (see Figure 1.3).
Example 1.2. Let us consider the polynomial relation z 2 = f (w) = (w2 − 1) (w2 − 4). For every
w 6= ±1, ±2 there exist C1 ∼ = C and C2 ∼
= C, copies of the complex plane, for the possible values of
z. Consider the half lines (from ±1 to ±2) given by

{Re(w) ∈ [−2, −1], Im(w) = 0} ⊂ C,

{Re(w) ∈ [1, 2], Im(w) = 0} ⊂ C.

Cut both C1 and C2 along them. At this point, we enlarge both the cuts and paste together the
corresponding edges, obtaining (this time we are concerned about what happens at ∞) Figure 1.4.

6
Figure 1.3: Topological moves of Example 1.1

Figure 1.4: Topological moves of Example 1.2

7
Chapter 2

Introduction to Riemann Surfaces

In this chapter, we introduce the notion of Riemann surface, and we profoundly analyze the funda-
mental example: smooth algebraic projective curves.
In the final part, we discuss two examples - both of which will appear many times in the remainder
of this course: the Riemann sphere, and the complex torus.

2.1 Main Definitions and Basic Properties

In this section, we give the definition of Riemann surface. The reader should pay attention to the
fact that a Riemann surface does not need to be connected, but we will ask for it in the definition
because we will mostly be dealing with connected compact Riemann surfaces.
Definition 2.1 (Riemann Surface). Let X be a topological manifold. We say that X is a Riemann
surface if the following properties are satisfied:
(a) dimR X = 2.
(b) X is Hausdorff, second-countable (i.e., there exists a countable basis) and connected.
(c) X has a complex structure. Namely, there exists an atlas

U = {ϕi : Ui → Vi ⊆ C}i∈I

such that ϕ : Ui → Vi is a homeomorphism of open sets, and the transition maps ϕi, j :=
ϕj ◦ ϕ−1
i are biholomorphic functions. In particular, the map

ϕi, j : ϕi (Ui ∩ Uj ) ⊂ C → ϕj (Ui ∩ Uj ) ⊂ C

is holomorphic with respect to the complex variable z.

Definition 2.2 (Biholomorphic). Let f : U ⊂ Cn → V ⊂ Cn be a complex function. We say that
f is biholomorphic if f is holomorphic, bijective and its inverse f −1 : V → U is also holomorphic.
Remark 2.1. A Riemann surface X is always orientable. In fact, the Jacobian of the transition
maps is always strictly positive (since ϕi, j is holomorphic), hence the atlas is orientated. The
interested reader may find a complete proof of this fact here.
Theorem 2.3 (Structure). Let X be a compact Riemann surface. Then X is either homeomorphic
to a sphere (g(X) = 0), a torus (g(X) = 1) or a n-torus (g(X) = n).

8
 Figure 2.1: Structure Theorem for Compact Riemann Surfaces.

2.2 Projective Curves

Let P2 (C) be the complex projective space of dimension 2, and let [z1 : z2 : z0 ] be the coordinates
so that {z0 = 0} is the line at infinity. An algebraic curve is defined as

X := {F (z1 , z2 , z0 ) = 0} ,

where F is a homogeneous1 polynomial of degree d. Recall that

(a) X is irreducible ⇐⇒ F is irreducible;

p
(b) X is reduced ⇐⇒ I(F ) = I(F ),
where I(F ) is the ideal associated to F (in this case, simply I(F ) = (F )).

N.B. From now on we will assume that X is an irreducible and reduced (projective) algebraic
curve.

Definition 2.4 (Singular points ). A point p ∈ X is singular for X if

∂F
F (p) = (p) = 0, ∀ i = 0, 1, 2.
∂ zi
1 For every λ ∈ C, it turns out that F (λ z1 , λ z2 , λ z0 ) = λd F (z1 , z2 , z0 ).

9
Definition 2.5 (Smooth). The (projective) algebraic curve X ⊂ P2 (C) is smooth if and only if no
point p ∈ X is singular.
Remark 2.2. The projective complex plane P2 (C) admits a standard atlas U := {(Ui , ϕi )}i=0, 1, 2 ,
which is defined by setting
 
zj zk
Ui := {zi 6= 0} and ϕi : Ui −→ C2 , [z1 : z2 : z0 ] 7−→ , .
zi zi
Then U induces, by restriction, an atlas on X, which is given by
 
X := (Ui ∩ X, ϕi X ) i=0, 1, 2 .

Proposition 2.6. The projective algebraic curve X is smooth if and only if the affine algebraic
curve Xi := X ∩ Ui ⊂ Ui ∼
= C2 is a smooth for every i = 0, 1, 2.
Theorem 2.7 (Implicit function). Let F ∈ C[z1 , z2 ] be any polynomial, and denote by X := {F =
0} ⊂ C2 the associated algebraic variety.
Let p ∈ X be a point such that ∂z2 F (p) 6= 0. There are a neighborhood Up of p and a holomorphic
function G : U → V such that
X ∩ U = {(z1 , G(z1 )) | z1 ∈ V } .
Corollary 2.8. Let p ∈ X be a smooth point. There exists a neighborhood Up of p (which is exactly
the one given by Theorem 2.7), such that X has a local complex structure, that is,
'
U ∩ X = {(z1 , G(z1 ))} −→ {z1 | z1 ∈ V } .
Theorem 2.9. Let X ⊂ P2 (C) be a smooth algebraic curve of degree d. Then X is a compact
Riemann surface, and its genus is equal to
(d − 1)(d − 2)
g(X) = .
2
Proof. We will prove this result later, but the reader which is already interested in a simple proof
of this fact, may jump and take at look at this paper.

2.2.1 Multiplicities
Recall. Let X ⊆ P2 (C) be any irreducible and reduced algebraic curve, and let us denote it by
X := {F (z1 , z2 , z0 ) = 0},
where F is an homogeneous irreducible polynomial of degree d such that the associated ideal (F ) is
radical. Let p ∈ X be a singular point of X, that is, a point where F and all its derivatives vanish:
∂F ∂F ∂F
F (p) = (p) = (p) = (p) = 0.
∂ z1 ∂ z2 ∂ z0
Definition 2.10 (Multiplicity). The multiplicity of a point p in X is the least integer k among all
the multiplicities of p with respect to the intersection between X and the lines passing through p.
More precisely, we define
moltp (X) := min {moltp (r, X) : r line passing through p} .
It is straightforward to prove that a point p ∈ X is smooth if and only if its multiplicity moltp (X)
is equal to 1. In particular, any singular point has multiplicity greater or equal than 2.

10
Proposition 2.11. Assume that p := (0, 0, 1) belongs to X. The dehomogenization of F with
respect to the coordinate z0 (i.e., in X0 = X ∩ U0 ) is given by
X
F (z1 , z2 , 1) = Fk (z1 , z2 ),
k≥m

where the Fk are homogeneous polynomials of degree k. Then the multiplicity of X at p is the
minimum degree of the dehomogenization, that is,

moltp (X) = m.

Definition 2.12 (Ordinary point). Let p ∈ X be any singular point, and suppose that its multi-
plicity is equal to m. We say that p is an ordinary multiple point if, locally,
m
Y
F = Hj ,
j=1

where the Hj are linear forms such that Hj = Hi ⇐⇒ i = j.

Example 2.1.
(1) An ordinary double point is locally (in a neighborhood of p := (0, 0, 1)) given by the equation
z1 · z2 = 0 (see e.g. Figure 2.2, left).

(2) A non-ordinary double point is locally given by an equation of the form z13 = z22 , and it
corresponds to a singular cuspid kind of point (see e.g. Figure 2.2, right).

Figure 2.2: Examples of singular points: left ordinary, right non-ordinary.

11
2.2.2 Resolution of the Singularities
Introduction. Let X be a singular projective algebraic curve. The primary goal of this subsection
is to give to the reader two methods which are useful to resolve the singularities of X, that is, to
find a smooth algebraic curve X
e and a surjective map

e − X.
Φ:X

Topological Approach. Let p ∈ X be an ordinary multiple point, and let U be a neighborhood

of p that does not contain any other singular point of X. By definition, the polynomial F is locally
(i.e., in U) given by the product of m linear forms:
m
Y
F = Hj .
j=1

If we denote by ∆ the Poincaré disk, that is,

∆ := {z ∈ C | |z| < 1 } and ∆∗ := ∆ \ {0},

then the algebraic curve - without the singular point p - is locally homeomorphic to the union of m
copies of ∆∗ ; more precisely, it turns out that

(U ∩ {Hj = 0}) \ {p} ∼

= ∆∗ , ∀ j ∈ {1, . . . , m}.

Consequently, the algebraic curve X is locally homeomorphic to the wedge of m disks (∆) of center
p, that is, there is a homeomorphism
m
U ∩X ∼
^
= ∆.
j=1

If we remove the singular point p, the wedge is clearly homeomorphic to the disjoint union of the m
disks deprived of their centers, i.e.,
m
(U ∩ X) \ {p} ∼
G
= ∆∗ .
j=1

At this point, the resolution of the singularity p is entirely straightforward: we add a center to each
disk ∆∗ . Formally, we define the smooth manifold
e := (X \ {p}) ∪ {q1 , . . . , qm } ,
X

and we prove that it is homeomorphic to a disjoint union of balls, i.e.,

m m
e∼ (∆ ∪ {qj }) ∼
G G
∗
X = = ∆.
j=1 j=1

In the general case, we give a brief sketch of what the ideas behind are. Take any p ∈ X and any
chart which sends p to the origin of C in such a way that there exists (locally) a function f : X → C
with the additional property that
f : f −1 (∆∗ ) → ∆∗
is a covering of order m.
On the other hand, we know that the connected coverings of ∆∗ ∼= S 1 are all and only of the
m
form z 7→ z . Thus we only need to add m points over the point (0, 0).

12
 Figure 2.3: Idea behind the topological approach.

Blowup Approach. Let p = (0, 0) ∈ C2 . The blowup of C2 at the point p is defined as follows:

Blp C2 := {(z1 , z2 ; [a : b]) | z1 b = z2 a } ⊆ C2 × P1 (C).

There exists a map

π : Blp C2 −→ C2

such that π −1 (p) ∼

= P1 (C). We denote by E the fiber π −1 (p), and, from now on, we will refer to it
as the exceptional line (since it contains the directions of the lines passing through p).
Consequently, the complement of E in the blowup is homeomorphic to the complement of the
fiber, that is,
Blp C2 \ E ∼ = π −1 C2 \ {(0, 0)} .

Assume that p = (0, 0) ∈ X0 is a singular point of the affine algebraic curve X0 := {F (z1 , z2 ) = 0} ⊆
C2 . Then F is a sum of homogeneous polynomials of order ≥ m, that is,
X
F (z1 , z2 , 1) = Fk (z1 , z2 ),
k≥m

and we may assume, without loss of generality, that {z1 = 0} is not tangent to X0 (i.e., a 6= 0).
If we set v := b/a, then we can define the strict transform of F with respect to the coordinates
(z1 , v) as follows:
Fe(z1 , v) := F (z1 , z1 · v) · z1−m . (2.1)

13
In a neighborhood U of p the map
n o
X f0 ∩ U − X0 ∩ U
f0 := Fe = 0 =⇒ X

is surjective, and it is straightforward to prove that u is a slope coefficient of a tangent line at p to

X0 if and only if it belongs to E.

Remark 2.3. If X is a singular algebraic curve, then there is no guarantee that X e will be smooth
after a single application of the blowup approach.
The next result states that a finite sequence of blowups is enough to obtain a smooth algebraic
curve, and in Example 2.4 we describe a case where two steps are necessary.
Theorem 2.13. Let X be a singular algebraic curve. There exists a finite sequence of blowups
e n − X
X e n−1 − . . . − X
e 1 − X

e n is a compact Riemann surface (thus a smooth algebraic curve).

such that X
Idea. The proof of this result is divided into three steps. The reader may try to prove it as an
exercise.

Step 1. There are only finitely many singular points in X. Hint: apply Bezout’s theorem applied
to the polynomials (F, F 0 ), where F 0 is the usual derivative of F .

Step 2. Local resolution of the singularities.

Step 3. For every singular point p ∈ X, the quantity moltp (X) eventually decreases to 0 as the
second step is repeated.

Definition 2.14 (Infinitely Near). Let p ∈ X be a singular point. A point q is infinitely near to
p ∈ X, and we denote it by q ∈ E(p), if q belongs to the exceptional line E.
Theorem 2.15. Let X ⊂ P2 (C) be an irreducible and reduced algebraic curve.
(a) If X is smooth, then X is a compact Riemann surface of genus

(d − 1)(d − 2)
g(X) = .
2

(b) If X is singular, then there exist a compact Riemann surface Xe and a birational morphism
π : X − X (that is, an isomorphism outside of singular points) such that
e
  (d − 1)(d − 2) X
g X
e = − δp ,
2
p∈Sing(X)

where δp is equal to
mp (mp − 1) X mq (mq − 1)
δp = + .
2 2
q∈E(p)

14
2.2.3 Examples
In this brief subsection, we describe the resolution of singularities in very few simple cases, and we
come back to the first example of the course.
Example 2.2. Let X be, locally, the algebraic curve defined by the polynomial

F (z1 , z2 ) := z12 − z22 = 0.

The reader may check that p = (0, 0) is singular, and its multiplicity is equal to 2. If we set
z2 := v · z1 , then the strict transform of F is given by

Fe(z1 , v) = z12 − v 2 z12 z1−2 = 1 − v 2 .

 

Consequently, the intersection E ∩ {Fe(z1 , v) = 0} is made of two points (v = ±1), corresponding to

the singular cross of lines in X (see Figure 2.4, left).
Example 2.3. Let X be, locally, the algebraic curve defined by the polynomial

F (z1 , z2 ) := z13 − z22 = 0.

The point p = (0, 0) is singular, and its multiplicity is also equal to 2. If we set z2 := v · z1 , then
the strict transform of F is given by

Fe(z1 , v) = z13 − v 2 z12 z1−2 = z1 − v 2 .

 

Clearly the point q ∈ E(p) is smooth, but it is tangent to the exceptional line E, thus the contribution
of δp is nonzero (see Figure 2.4, center).
Example 2.4. Let X be, locally, the algebraic curve defined by the polynomial

F (z1 , z2 ) := z14 − z22 = 0.

The point p = (0, 0) is singular, and its multiplicity is also 2. If we set z2 := v · z1 , then the strict
transform of F is given by

Fe(z1 , v) = z14 − v 2 z12 z1−2 = z12 − v 2 .

 

The algebraic curve X e 1 is singular and, actually, it is the same algebraic curve of Example 2.2. If
we set z1 := h · v, then the second strict transform of F is

Fe(2) (h, v) = 1 − h2 ,

e 2 − X
thus X e 1 − X is the resolution of singularities X (see Figure 2.4, right).

Example 2.5 (Global). Let X ⊂ C2 be the affine curve defined by the equation

z22 − (z12 − 1)(z12 − 4) = 0.

Its projective closure is obtained via homogenization of the equation, that is,

X := {F (z1 , z2 , z0 ) = 0} ⊂ P2 (C), F (z1 , z2 , z0 ) = (z12 − z02 )(z12 − 4 z02 ) − z02 z22 .

It is a simple exercise to prove that the only singular point is p∞ = (0, 1, 0) and that it belongs to
the line at infinity.

15
 Figure 2.4: From left to right: Examples 2.2, 2.3 and 2.4

Consequently, about p∞ it makes sense to use the chart with coordinates (z1 , 1, z0 ) in such a
way that, at least locally, the algebraic curve is given by the equation
∼p∞ 
X = z02 − (z12 − z02 )(z12 − 4 z02 ) = 0 = z02 = (z 0 )4 .


As a consequence, around p∞ the situation is similar to the one studied in Example 2.4. In particular,
e whose genus is given by2
there is a compact Riemann surface X,
 
g Xe = 3 − δp = 1,
∞

and a surjective map X e − X. We conclude that X e ∼= T, that is, X

e is homeomorphic to the
complex torus (see Example 1.2).
The result is coherent with the fact that the algebraic curve X may be obtained from a 3-torus
by gluing together the two extremal holes and throttling them in such a way to get a 1-torus with
a weird point p∞ .
Remark 2.4. The genus that we have introduced in this subsection is called arithmetic genus of a
planar algebraic curve, and it is equal to the leading coefficient of the Hilbert polynomial (associated
to the local coordinate ring).
This notion of genus is equivalent to the topological one if X is a smooth algebraic curve, where
h1 (X, Z)
gtop (X) := , h1 := dim H 1 (X, Z) .
2
2 Indeed, it is straightforward to prove that mp∞ = 2, mq∞ = 2 and mqg = 1.
∞

16
Recall that, equivalently, we have
χtop (X) + 2
gtop (X) := ,
2
where χtop (X) is the topological Euler characteristic.

2.3 First Examples Riemann Surfaces

Recall that a Riemann surface X is a Hausdorff, second-countable, connected manifold endowed
with a complex structure.

Riemann Sphere C∞ . Let S 2 := {(x1 , x2 , x3 ) ∈ R3 : x21 + x22 + x23 = 1} be the two-dimensional

sphere, and let us consider the atlas given by the stereographic projections:
x1 x2
ϕ0 : U0 := S 2 − {N } → C, ϕ0 (x1 , x2 , x3 ) := +ı ,
1 − x3 1 − x3
x1 x2
ϕ1 : U1 := S 2 − {S} → C, ϕ1 (x1 , x2 , x3 ) := +ı .
1 + x3 1 + x3
If we set A0 := {(U0 , ϕ0 ), (U1 , ϕ1 )}, then one can easily check that it is not a holomorphic atlas
since the transition map
x y
ϕ0, 1 = ϕ1 ◦ ϕ−1
0 (x, y) = +ı 2
x2 + y 2 x + y2
does not satisfy the Riemann-Cauchy equations3 . Therefore, we try to modify one of the charts
above, e.g. we take the conjugate of ϕ1 :
x1 x2
ϕ1 : U1 := S 2 − {S} → C, ϕ1 (x1 , x2 , x3 ) := −ı .
1 + x3 1 + x3
The transition map becomes
x y 1
ϕ0, 1 = ϕ1 ◦ ϕ−1
0 (x, y) = −ı 2 = ,
x2 + y 2 x + y2 z
and it is an easy exercise to check that it is holomorphic, that is,

A := {(U0 , ϕ0 ), (U1 , ϕ1 )}

is a holomorphic atlas of C∞ . In particular, the atlas A induces a holomorphic structure on the

Riemann sphere C∞ and, as a consequence, it turns out that C∞ is biholomorphic to P1 (C).
In fact, if we consider the complex projective space with coordinates [z0 : z1 ], together with the
atlas
U0 := {z0 6= 0} , U1 := {z1 6= 0} ,
then the coordinate in U0 is z = zz01 , the coordinate in U1 is w = z0
z1 and the transition map above is
the change of coordinates from U0 to U1
1
z ∈ U0 7→ w = ∈ U1 .
z
3A function f = u + ı v : A ⊆ C −→ C is holomorphic if and only if
∂x u = ∂ y v and ∂y u = −∂x v.

17
Complex Tori T. Let Λ ⊂ C be a lattice of the form Z ω1 + Z ω2 , where {ω1 , ω2 } are linearly
independent over R, that is,
ω1
τ := ∈ C \ R.
ω2
The complex torus is defined as the quotient space T = CΛ which is topologically homeomorphic
to the product S 1 × S 1 .
Let π : C → CΛ be the standard projection. The topology on T is the quotient topology, thus
we only need to equip it with a complex structure.
For any point z ∈ C \ Λ there is a neighborhood Uz ⊂ C such that Uz ∩ Λ = ∅. If we set

δ := min {d(ξ1 , ξ2 ) : ξ1 , ξ2 ∈ Λ} ,

then, given z ∈ C \ Λ and p = π(z) ∈ T, the key idea is to find a neighborhood of p, starting from
the image of Uz via the map π. Indeed, the set

∆(z, ) := {ξ ∈ C : |ξ − z| < ,  < δ} ,

is strictly contained in Uz for a suitable choice of  > 0, thus we can find a local holomorphic
structure at each point p by using the covering
n o
v(p, ), π −1 v(p, )

F := ,
p∈T

where v(p, ) is homeomorphic to ∆(z, ) via π.

For any p0 , p1 ∈ T there are charts ϕ0 : U0 := v(p0 , ) → ∆(z0 , ) and ϕ1 : U1 := v(p1 , ) →
∆(z1 , ); thus the transition map is given by the composition

ϕ1 ◦ ϕ−1
0 : C → C.

If ϕ0 (U0 ) and ϕ1 (U1 ) belong to the same fundamental parallelogram of C \ Λ, then there is nothing
to prove since we can define the transition map as the identity on the intersection.
On the other hand, if ϕ0 (U0 ) and ϕ1 (U1 ) belong to different fundamentals parallelograms, then
it turns out that
π ϕ1 ◦ ϕ−1


0 (z) = π(z), ∀z ∈ ϕ0 (U0 ) ∩ ϕ1 (U1 ).
Consequently, the function η(z) := ϕ1 ◦ ϕ−1 0 (z) − z is continuous and with values in Λ, a discrete
set, thus it needs to be locally constant. In particular, the transition map is locally given by

ϕ1 ◦ ϕ−1
0 = z + c,

which is clearly holomorphic. Therefore the atlas F induces a complex structure on T.

Remark 2.5. If T1 and T2 are two complex tori, then we will prove in Section 4.2 that T1 is not
necessarily biholomorphic to T2 .
On the other hand, the first example shows that every Riemann surface, whose genus is g = 0,
is biholomorphic to P1 (C).

18
Figure 2.5: Complex Tori

19
Chapter 3

Functions and Maps

3.1 Functions on Riemann Surfaces

Definition 3.1 (Holomorphic function). Let X be a Riemann surface. A function f : X → C is
holomorphic at p ∈ X if there exists a chart around p
∼
ϕ : Up −−−−−→ ∆ ⊆ C

such that the composition f ◦ ϕ−1 : ∆ −→ C is holomorphic at ϕ(p) (or, equivalently, if it is

holomorphic in an appropriate open subset of ∆ containing ϕ(p)).
Example 3.1. Let X = C∞ be the Riemann sphere and let p∞ be the point at infinity. The reader
may easily prove as an exercise that
 
1
f (z) is holomorphic at p∞ ⇐⇒ f is holomorphic at 0.
z

Hence, if f (z) = p(z)/q(z) is a holomorphic function and a quotient of polynomials, then the degree
of p needs to be less or equal than the degree of q. Notice that this is not a sufficient condition.
Example 3.2. Let X ⊆ P2 (C) be a smooth algebraic curve and let p = [z1 : z2 : z0 ] ∈ X such that
z0 6= 0. Then z1 /z0 and z2 /z0 are locally holomorphic functions in P2 (C), that is,
z1  z2 
and
z0 X z0 X
are holomorphic functions at p.
Notation. Let X be a Riemann surface and let U ⊆ X be an open subset. We denote the set of
all the holomorphic functions from U to C by

OX (U ) := {f : U → C | f holomorphic } .

Definition 3.2 (Singularities Type). Let f : U \ {p} → C be a holomorphic function and let
ϕ : Up −→ ∆ be a chart around p. We say that the singularity at p is
(1) removable if ϕ(p) is a removable singularity for f ◦ ϕ−1 : ∆∗ ⊂ C → C;
(2) a pole if ϕ(p) is a pole for f ◦ ϕ−1 : ∆∗ ⊂ C → C;
(3) essential if ϕ(p) is an essential singularity for f ◦ ϕ−1 : ∆∗ ⊂ C → C.

20
Definition 3.3 (Meromorphic function). Let X be a Riemann surface. The function f : X → C is
meromorphic at p ∈ X if
(a) there exists an open neighborhood U ⊆ X of p such that f is holomorphic in U \ {p};
(b) p is either a removable singularity, or a pole.
Proposition 3.4 (Characterization). Let X be a Riemann surface, and let f : X → C be a mero-
morphic function. Then f locally is the sum of a Laurent series
X
f (z) = cn z n ,
n≥k

and vice versa.

Definition 3.5 (Order). Let f : X → C be a meromorphic function. The order of f at p ∈ X is
the minimum integer k in the Laurent series expansion with nonzero coefficient, that is,

ordp (f ) = min {k : ck 6= 0} .

Remark 3.1. Any holomorphic function is also a harmonic function (by Riemann-Cauchy); hence
holomorphic functions satisfy the maximum principle.
Theorem 3.6 (Maximum Modulus). Let f : U ⊆ X → C be a holomorphic function defined on any
open connected subset U of X. If there exists p ∈ U such that

|f (x)| ≤ |f (p)|, ∀ x ∈ U,

then f is constant.
An important consequence of this theorem is that introducing only the holomorphic functions
on compact Riemann surfaces would be a great restriction.
Corollary 3.7. Let f : X → C be a holomorphic function and let X be a compact Riemann surface.
Then the function f is constant.
Theorem 3.8. Let f : P1 (C) → C be a meromorphic function. Then f is a rational function, that
is there exist p, q homogeneous polynomial of the same degree such that
p(z)
f (z) = .
q(z)
Proof. Let us set U0 := {z0 6= 0} ⊆ P1 (C), and recall that U0 ∼
= C with coordinate z = z1 /z0 .
Consider the dehomogenization
fe(z) := f (z, 1),
and let n o
{λj }j∈J := zeros and poles of fe in U0 and ej := ordfe(λj ).
The function Y ej
R(z)
e := (z − λj )
j∈J

comes with the same poles and zeros of fe in U0 , so we may extend it to a function R defined on the
whole projective space P1 (C) as follows:
Y e
R(z) := z0n (bj z1 − aj z0 ) j ,
j∈J

21
 P
where λj = [aj : bj ] and n = − j ej . In conclusion, we notice that the function

f (z)
g(z) :=
R(z)

has no zeros or poles in U0 , hence we only need to check the infinity point p∞ = [1 : 0].

Useful Trick. If p∞ is not a pole for g, then g has no poles, and it is hence constant. If on the
other hand, g has a pole in p∞ , the reciprocal 1/g is holomorphic and consequently constant.

3.2 Holomorphic Maps Between Riemann Surfaces

In this section, X and Y will denote Riemann surfaces unless stated otherwise.
Definition 3.9 (Holomorphic Map). A mapping F : X −→ Y is holomorphic at p ∈ X if and
only if there exist charts ϕ : Up → ∆1 on X and ψ : VF (p) → ∆2 on Y such that the composition
ψ ◦ F ◦ ϕ−1 is holomorphic at ϕ(p).
It is particularly useful to visualize the definition above through the following commutative
diagram:
F
Up VF (p)
ϕ ψ
f
∆1 ⊆ C ∆2 ⊆ C

Definition 3.10. A mapping F : X → Y is holomorphic if and only if it is holomorphic at each

point p ∈ X.
Lemma 3.11. Let F : X → Y be a map between Riemann surfaces.
(a) The identity idX : X → X is holomorphic.
(b) The composition between a holomorphic map and a holomorphic function is a holomorphic
function.
More precisely, if F : X → Y and g : W ⊆ Y → C are holomorphic and W is an open subset
of Y , the composition g ◦ F is a holomorphic function on F −1 (W ).
(c) The composition between holomorphic maps is still a holomorphic map.
More precisely, if F : X → Y and G : W ⊆ Y → Z are holomorphic maps and W is an open
subset of Y , the composition G ◦ F is a holomorphic map from F −1 (W ) to Z.
(d) The composition between a holomorphic map and a meromorphic function is a meromorphic
function.
More precisely, let F : X → Y be a holomorphic map, let g : W ⊆ Y → C be a meromorphic
function and let W ⊆ Y be an open subset. If F (X) is not contained in the set of poles of g,
then g ◦ F is a meromorphic function on F −1 (W ).
(e) The composition between a holomorphic map and a meromorphic map is a meromorphic map.
More precisely, let F : X → Y be a holomorphic map, let G : W ⊆ Y → C be a meromorphic
map and let W ⊆ Y be an open subset. If F (X) is not contained in the set of poles of G, then
G ◦ F is a meromorphic function from F −1 (W ) to Z.

22
Proof. These are all trivial facts, thus the proof is left to the reader as a simple exercise.
Let F : X → Y be a nonconstant holomorphic map between Riemann surfaces. For every subset
W ⊆ Y , F induces a C-algebra homomorphism
g7→g◦F
F ∗ : OY (W ) −−−−−→ OX F −1 (W ) .

Similarly, F induces a C-algebra homomorphism between meromorphic functions on W and mero-

morphic function on F −1 (W ) via composition:
g7→g◦F
F ∗ : MY (W ) −−−−−→ MX F −1 (W ) .

∗
If F : X → Y and G : Y → Z are holomorphic maps, then it is trivial to prove that the operator
reverses the composition order, i.e.,
∗
(G ◦ F ) = F ∗ ◦ G∗ .

Corollary 3.12. Riemann surfaces equipped with holomorphic mappings form a category.
Proposition 3.13 (Open Mapping Theorem). Let F : X → Y be a nonconstant holomorphic map
between Riemann surfaces. Then F is open.
Corollary 3.14. Let F : X → Y be a nonconstant holomorphic map. Assume that
(a) X is a connected and compact Riemann surface;
(b) Y is a connected Riemann surface.
Then the map F is surjective, and Y is compact.
Proof. From the Open Mapping Theorem 3.13 it follows that F is an open map. Consequently the
image of X, F (X), is open in Y .
On the other hand, X is compact and hence F (X) is a compact subset of a Hausdorff space Y ,
which means that F (X) is closed. Finally, since Y is connected and F is nonconstant, we infer that
F (X) = Y .

3.3 Global Properties of Holomorphic Maps

Let f be a holomorphic function defined on a Riemann surface X. The complex plane C =: Y is a
Riemann surface; hence we may always identify f with the holomorphic map f : X → Y .

Meromorphic Map Identification. Let f be a meromorphic function defined on X. By def-

inition, f is holomorphic away from its poles, and thus it assumes as values complex numbers.
Therefore, it is natural to define a map F : X → C∞ by setting
(
f (x) if x is not a pole of f ,
F (x) := (3.1)
∞ if x is a pole of f .

Theorem 3.15. There exists a 1 − 1 correspondence between

 
( )  Holomorphic maps 
Meromorphic functions f  
←→ F : X → C∞
defined on X  
which are not identically ∞
 

23
Sketch of the Proof. First, we observe that the function defined by (3.1) is holomorphic at every
point of X. The proof of this simple fact is left to the reader as an exercise.
Observe also that, since C∞ ∼ = P1 (C), holomorphic maps F : X → C∞ are in correspondence
with holomorphic maps F : X → P1 (C), hence it suffices to prove that there is a 1-1 correspondence
 
( )  Holomorphic maps 
Meromorphic functions f 
1

←→ F : X → P (C) .
defined on X  
which are not identically ∞
 

Let p ∈ X be any point. Locally - in a neighborhood Up 3 p - the function f is the ratio of two
holomorphic functions, i.e.
g(x)
f (x) = ∀ x ∈ Up ⊂ X.
h(x)
The corresponding map to P1 (C), in this neighborhood of p, is given by

Up 3 x 7→ [g(x) : h(x)] ∈ P1 (C).

A meromorphic function is not globally the ratio of holomorphic functions; thus this representation
is possible only locally, in a neighborhood of each point.

Normal Form [2]. In this paragraph, we want to briefly introduce the so-called normal form of
a holomorphic map between Riemann surfaces.
Proposition 3.16. Let F : X → Y be a nonconstant holomorphic map, and let p ∈ X be a point
of the domain. There exists a unique integer m ≥ 1 which satisfies the following property: for every
chart ψ : UF (p) ⊂ Y → ∆0 centered1 at F (p), there exists a chart ϕ : Up ⊂ X → ∆ centered at p
such that
ψ ◦ F ◦ ϕ−1 (z) = z m .
Definition 3.17 (Multiplicity). The multiplicity of F at p, denoted by multp F , is the unique integer
m such that there are local coordinates near p and F (p) with F having the form z 7→ z m .

There is an easy way to compute the multiplicity that does not require finding charts realizing
the normal form. Take any local coordinates z near p and w near F (p), and let

z0 ←→ p and w0 ←→ F (p).

There exists a holomorphic function h such that w = h(z) in such a way that w0 = h(z0 ); hence the
multiplicity multp F of F at p is one more than the order of vanishing of the derivative h0 (z0 ) of h
at z0 , that is,  
dh
multp F = 1 + ordz0 .
dz
In particular, the multiplicity is the exponent of the lowest strictly positive term of the power series
for h. Namely, we have that
∞
X
h(z) = h(z0 ) + ci (z − z0 )i =⇒ multp F = min {i ∈ Z | ci 6= 0 } .
i=m
1A chart ϕ is centered at a point q ∈ X if ϕ(q) = 0.

24
 Figure 3.1: Idea of Normal Form and Multiplicity

3.4 The Degree of a Holomorphic Map

In this section, we introduce the notion of degree of a holomorphic map, and we set the ground for
the main result of this chapter: the Hurwitz formula.
Theorem 3.18. Let F : X → Y be a nonconstant holomorphic map between connected and compact
Riemann surfaces. For each y ∈ Y , the quantity
X
dy (F ) := multp F
p∈F −1 (y)

is constant, independent of y ∈ Y .
Proof. The idea of the proof is to show that the map y 7→ dy (F ) is a locally constant function from
Y to Z. Since Y is connected, a locally constant function must be constant.

Step 1. Let y ∈ Y and let F −1 (y) = {x1 , . . . , xn } be the fiber. Set moltxj (F ) := mj to be the
multiplicity at xj , for each j = 1, . . . , n.
By Proposition 3.16 (normal form) there are neighborhoods Ui of xi such that Ui ∩ Uj = ∅ for
i 6= j, and F U sends zi to wi = zimi .

i

Step 2. The thesis is equivalent to the existence of a neighborhood U of y with the additional
property that, for any y 0 ∈ U ,
n
[
F −1 (y 0 ) ⊂ Uj .
j=1

We argue by contradiction. Suppose that there exists a sequence of points (pk )k∈N ⊂ X such that
n
[
pk ∈
/ Uj ,
j=1

but F (pk ) converges to y. Since X is compact and F is continuous, there exists a subsequence
h→+∞
(pkh )h∈N such that pkh −−−−−→ x̄ and F (x̄) = y.

25
 Hence x̄ must be equal to xj for some j ∈ {1, . . . , n}, but this is absurd since no point of the
sequence (pk )k∈N lies in the neighborhoods Ui of the xi ’s.
Definition 3.19 (Degree). Let F : X → Y be a nonconstant holomorphic map between connected
and compact Riemann surfaces. The degree of F , denoted by deg F , is the quantity dy (F ) computed
at any possible y ∈ Y .
Corollary 3.20. Let F : X → Y be a holomorphic map of connected compact Riemann surfaces.
The map F is locally biholomorphic to ψ ◦ F ◦ ϕ−1 (sending z to z) and deg F = 1 if and only if
X∼=Y.
Notation. Let F : X → Y be a mapping of Riemann surfaces.
(a) A point p ∈ X is called a ramification point if moltp (F ) ≥ 2.
(b) A point q ∈ Y is called a branch point if it is the image of a ramification point.
(c) The ramification index in p ∈ X is defined as moltp (F ) − 1.

3.5 Hurwitz’s Formula

Introduction. Let X be a compact connected Riemann surface of genus g. The Euler-Poincaré
topological characteristic is defined as

χtop := b0 − b1 + b2 ,

where bi := dim (Hi (X, R)) = rank (Hi (X, Z)) is the i-th Betti’s number. In a similar fashion, if X
is a manifold, then
X n
dimR (X) = n =⇒ χtop = (−1)j · bj (X).
j=0

Lemma 3.21. Let X be a compact connected Riemann surface of genus g(X). Then

b0 = b2 = # connected components = 1,

while H1 (X, Z) = Ab (π1 (X)). In particular, the following identity holds:

χtop = 2 − 2 g(X). (3.2)

Proposition 3.22. The Euler-Poincaré characteristic does not depend on the triangulation of X,
that is,
χtop = v − e + f,
where v is the number of vertexes, e is the number of edges and f is the number of faces.
Proof. A sketch of the argument may be found in [2, Page 51].
Theorem 3.23 (Hurwitz’s Formula). Let F : X → Y be a nonconstant holomorphic map between
compact Riemann surfaces of genus g(X) and g(Y ) respectively. Then
X
2 (g(X) − 1) = 2 deg F · (g(Y ) − 1) + [multp F − 1] (3.3)
p∈X

Proof. The Riemann surface X is compact, thus the set of ramification point is finite and the sum
on the right-hand side is finite.

26
Step 1. Let us take any triangulation τ of Y , such that each branch point of F is a vertex. Denote
by v the number of vertexes, e the number of edges and t the number of triangles (faces).
Assume that, if q ∈ Y is a branch point and T 3 q a triangle, then T is contained in a neighbor-
hood Uq of q such that
mq
G
F : Uj → Uq
j=1

is in normal form. Lift this triangulation to X via the map F , i.e. τ 0 = F −1 (τ ), and notice that any
ramification point is a vertex of a triangle.

Step 2. Since there are no ramification point over the general point of any triangle, each one lifts
to deg(F ) triangles in X. Let q ∈ Y be any point; then
 −1  X X
F (q) = 1 = deg F + [1 − multp F ] .
p∈F −1 (q) p∈F −1 (q)

The number of edges of τ 0 is e0 = deg F · e, the number of triangles is t0 = deg F · t and the number
of vertexes is  
X X
v0 = deg F + (1 − multp F ) =
q∈v(Y ) p∈F −1 (q)

X X
= deg F · v − [multp F − 1] =
q∈v(Y ) p∈F −1 (q)

X
= deg F · v − [multp F − 1] .
p∈v(X)

Therefore we have that

2 g(X) − 2 = −v 0 + e0 − t0 =
X
= −deg F · v + [multp F − 1] + deg F · e − deg F · t =
p∈v(X)

X
= 2 deg F · (g(Y ) − 1) + [multp F − 1] ,
p∈X

since every ramification point in X is, actually, contained in the set v(X) of vertexes of X by
construction.
Remark 3.2. Let X be a compact Riemann surface. Then there are only finitely many points
p ∈ X with multiplicity greater or equal than 2.
Remark 3.3. The Hurwitz’s formula (3.3) gives us more information than the value of the genus.
In fact, if we divide it by 2, it turns out that
1 X
g(X) = deg F · (g(Y ) − 1) + [multp F − 1] + 1,
2
p∈X

and thus

27
 (a) g(X) ≥ g(Y );
P
(b) the sum p∈X (multp F − 1) is even.

Example 3.3. Let F : P1 (C) → P1 (C) be a function induced by a homogeneous polynomial p of

degree d. More precisely, if z = z1 /z0 is the coordinate associated to the chart U0 := {z0 6= 0} ∼
= C,
then the restriction of F to U0 is given by

Fe : U0 ∼
= C −→ C, z 7−→ p(z).

We are interested in finding the ramification points and computing the multiplicities (to check the
validity of the Hurwitz’s formula).

Step 1. The ramification points, away from infinity, are the discrete set given by

{ramification points of F } ∩ U0 = {z ∈ C | p0 (z) = 0 } ,

and hence there are d − 1 (not necessarily distinct) ramification points.

Step 2. On the other hand, at the infinity point p∞ , we simply pass to the second chart U1 :=
{z1 6= 0} ∼
= C, with coordinate w = z0 /z1 , and we notice that

F (w) = wd .

Therefore, we can infer that

mult∞ (F ) = d − 1,
1
and, recalling that g(P (C)) = 0, the Hurwitz’s formula (3.3) yields to

−2 = d · (−2) + R ⇐⇒ R = 2 (d − 1),

which is coherent with the computation above.

28
Chapter 4

More Examples of Riemann

Surfaces

In the first part of this chapter, we show a simple application of the Hurwitz’s formula: we compute
the genus for a smooth algebraic curve X ⊂ P2 (C), finally proving what we have mentioned several
times, that is,
(d − 1)(d − 2)
g(X) = .
2
Next, we study the group of automorphisms for compact Riemann surfaces of genus 0 and 1; in the
final part, we investigate the action of finite groups G, and we prove an estimate on |G| which follows
from the Hurwitz’s formula.

4.1 Application of Hurwitz’s Formula

Genus of Algebraic Curves. Let X = {f (z1 , z2 , z0 ) = 0} ⊂ P2 (C) be an algebraic curve,
defined by a homogeneous polynomial f of degree equal to d.
Assume that X is smooth (so that X is a Riemann surface, by Dini’s theorem1 ) and assume also
that, up to a change of coordinates, the following properties are satisfied:
(a) p = [1 : 0 : 0] ∈
/ X;
(b) {z2 6= 0} is not tangent at any point of X.

Let us consider the projective line L = {z0 = 0} = ∼ P1 (C), and let us denote by π : X → L the
associated projection. As usual, we can work in the chart U0 = {z0 6= 0} with coordinates z = z1 /z0
and w = z2 /z0 so that
f
X ∩ U0 L ∩ U0
∈

(z, w) z

First, we observe that by assumption (a), the degree of the map associated to f is exactly equal to
d on the intersection X ∩ {z0 6= 0}.
1 See Theorem 2.7.

29
 
On the other hand, at infinity there are no ramifications, and thus R = R U0 . More precisely,
the ramification points of X ∩ U0 may be explicitly found as a solution of the system

fe(z, w) = 0
 ∂ fe
∂ w (z, w) = 0,

or, equivalently, of the system 

f (z1 , z2 , z0 ) = 0
 ∂f
∂ z0 (z1 , z2 , z0 ) = 0.
Hence by Bezout theorem the sum of the multiplicities of the ramification points is equal to the
product of the degrees, that is,
R = d · (d − 1).
Finally, from the Hurwitz’s formula (3.3), it turns out that

(d − 1)(d − 2)
2 g(X) − 2 = −2 d + R =⇒ g(X) = ,
2
as we suggested many times in the previous chapters.
The reader may consult [2, pp. 144-145] for a different approach to the problem, which still
results in a simple application of the Hurwitz’s formula.

4.2 Automorphism of Riemann Surfaces

Genus 0 Automorphisms. Let X be a compact Riemann surface of genus zero. If f : P1 (C) →
P1 (C) is an automorphism, then its degree is necessarily deg f = 1, and thus

f (z0 , z1 ) = (a z1 + b z0 , c z1 + d z0 ).

A necessary condition for f to be an automorphism is that the two linear polynomials have no zero
in common, that is,  
a b
det = ad − bc 6= 0.
c d
In the affine setting (e.g. in U0 = {z0 6= 0}), it turns out that, with respect to the coordinate
z = z1 /z0 , the mapping is given by

az + b
f¯ : C → C, z 7−→ .
cz + d
In particular, the group of automorphism of P1 (C) can be completely characterized as

P GL(2; C) = GL(2; C)C∗ ∼

= Aut P1 (C) ,

and the isomorphism is explicitly given by

 
GL(2; C) ∗ 3 A = a b 7−→ f (z , z ) = (a z + b z , c z + d z ) ∈ Aut P1 (C)
.
0 1 1 0 1 0
C c d

30
Genus 1 Automorphisms. In this paragraph, we characterize the holomorphic mappings f be-
tween complex tori and give a criterion to decide if f is an isomorphism or not.
Proposition 4.1.
(1) Let X be a compact Riemann surface of genus g(X) = 1.
Then X is isomorphic to a complex torus CΛ, with Λ = Z ω1 + Z ω2 lattice generated by
R-linearly independent elements ω1 and ω2 .
(2) Let ω1 , ω2 ∈ C be two elements such that
ω1
τ := ∈ C \ R,
ω2

and denote by Λ the associated lattice. Then the quotient CΛ is a compact Riemann surface
of genus one.
We are now ready to prove the main theorem about holomorphic maps between complex tori. In
particular, by the end of the section, we shall show that there are non-isomorphic complex tori (and
hence non-isomorphic Riemann surfaces of genus one).

Theorem 4.2 ([2]). Let X = CΛ and let Y = CΓ be compact Riemann surfaces of genus 1. A
holomorphic map f : X → Y is induced by a function G : C → C defined by G : z 7→ γ z + α, where
α and γ are fixed complex numbers. Moreover, the following properties hold true:
(a) If 0 7→ 0, then α = 0 and f is a group homomorphism.
(b) The mapping f is an isomorphism if and only if γ · Λ = Γ.
Proof. By composing f with a suitable translation on Y we may always assume that f (0) = 0.

Step 1. Since g(X) = g(Y ) = 1, the Hurwitz’s formula (3.3) proves that f is an unramified map.
In particular, it is a topological covering, and hence so is the composition f ◦ πX : C → Y .
Since the domain is simply connected, this must be isomorphic - as a covering - to the universal
covering p : C → Y . Therefore there is a map G : C → C and a commutative diagram
G
C C

πX πY

f
X Y

Step 2. The map G is induced on the universal coverings by the commutativity of the diagram
above and sends 0 to a lattice point; we may assume in fact that G(0) = 0, since composing with
translation by a lattice point does not affect the projection map π. Moreover f is a well-defined
map of quotients, thus
f (X) ⊆ Y =⇒ G(Λ) ⊆ Γ,
that is,
G(z + `) ≡Γ G(z), ∀ ` ∈ Λ.
Therefore there exists a lattice point ω(z, `) ∈ Γ such that

ω(z, `) = G(z + `) − G(z) ∈ Γ.

31
But Γ is a discrete subset and C is connected, thus we infer that ω(z, `) is locally constant in the
variable z. In particular, it turns out that

∂z [G(z + `) − G(z)] = 0, ∀ ` ∈ Λ,

and thus G0 is invariant for Λ (i.e., up to translations for elements of the lattice). As a consequence,
G0 is uniquely determined by its values on a fundamental parallelogram PΛ .

Step 3. Therefore G0 : C → C is a holomorphic function, whose value is determined by G0 PΛ ,
and thus it is bounded (since P is compact). By Liouville’s Theorem2 , there exists γ ∈ C such that
G0 (z) = γ and this concludes the proof of the point (a).

Step 4. Finally, if γ · Λ = Γ, then γ −1 · Γ = Λ, and so the map H(z) = γ −1 (z − α) induces a

holomorphic map from Y to X which is an inverse for G.
Remark 4.1. The degree of f is also given by the index of γ · Λ in Γ, that is,
 
deg f = ΓγΛ .
 

In particular, if f is an isomorphism, its degree is equal to 1 and γ · Λ = Γ.

Proposition 4.3. Let X = CΛ be a compact Riemann surface of genus 1. The holomorphic map

f : X → X, z 7−→ γ · z

is an automorphism of X - sending 0 to 0 - if and only if either

(1) Λ is a squared lattice, and γ is a 4th -root of unity; or
(2) Λ is a hexagonal lattice, and γ is a 6th -root of unity; or
(3) Λ is neither squared or hexagonal, and γ = ±1.

Proof. By Theorem 4.2 a necessary condition for f : X → X to be an automorphism (sending 0 to

0) it that γ · Λ = Λ, and thus kγk = 1. If γ = ±1 there is nothing else to prove.
Assume that γ ∈ / R and let ` ∈ Λ \ {0} be an element of minimal length. Then so is γ · ` and it
belongs to Λ. Clearly γ · (γ`) ∈ Λ, thus there exists m, n ∈ Z such that

γ 2 ` = n ` + m γ`,

therefore γ is a root (of norm equal to 1) of the polynomial

p(λ) = λ2 − m λ + n, m, n ∈ Z.

The proof is now complete since

p(λ) = λ2 ± 1 Λ is a square,

p(λ) = λ2 ± λ ± 1 Λ is a hexagonal.

2 See Corollary 3.7.

32
Corollary 4.4. Let X be a compact Riemann surface of genus 1. Then X ∼
= CΛ, where Λ = h1, τ i
and τ = ξ + ı η is a complex number such that η > 0.
e = hω1 , ω2 i, then one can consider the isomorphism of lattices given by
Proof. If Λ

e −→ Λ, ω2
G:Λ ω1 7→ 1, ω2 7→ .
ω1
If the imaginary part is not positive, then we may consider the isomorphism of lattices given by

e −→ Λ, ω1
H:Λ ω1 7→ ω2 7→ 1.
ω2

Corollary 4.5. Let Λ = h1, τ i and let Λ0 = h1, τ 0 i be two lattices defined over C. Set X = CΛ
and X 0 = CΛ0 . Then X ∼
= X 0 if and only if there exists
 
a b
∈ SL(2, Z),
c d

such that
a + b τ0
τ= .
c + d τ0
Proof. By Theorem 4.2 a sufficient condition for X to be isomorphic to X 0 is the existence of a
complex number γ ∈ C such that γ · Λ = Λ0 .
Equivalently, we only need to prove that there is γ such that hγ, γ τ i generates the lattice Λ0 .
The inclusion ⊆ is satisfied if there are integers a, b, c, d such that

γ = c τ 0 + d, γ τ = a τ 0 + b.

Eliminating γ from these equations gives a relation between τ and τ 0 , that is,
a + b τ0
τ= .
c + d τ0
Finally, for γ and γ τ to generate Λ0 , the determinant of the matrix (i.e., ad − bc) must be equal
to ±1. But it is easy to see that it is exactly equal to 1, since both τ and τ 0 lie in the upper
half-plane.

Conclusion. We proved that the group of automorphisms of a compact Riemann surface X fixing
0, denoted by Aut0 (X), is isomorphic to

Aut0 (X) ∼
= Z4 if Λ is square;

Aut0 (X) ∼
= Z6 if Λ is hexagonal;

Aut0 (X) ∼
= Z2 otherwise.

This simple result yields to a surprising fact: the complex torus defined using a square lattice is
not isomorphic to a complex torus defined using a hexagonal lattice.
Thus there are non-isomorphic complex tori, i.e., if g ≥ 1 then there exist surfaces of the same
genus which are not isomorphic (this does not happen for g = 0 since the only surface of genus zero
is the projective space).

33
4.3 Group Actions on Riemann Surfaces
Finite Group Actions. In this section, G will denote a finite group and X a Riemann surface.
In the last paragraph, we will briefly talk about the case of G infinite.
Definition 4.6 (Action). An action of a group G on a set X is a map µ : G × X → X, denoted by
µ(g, p) := g · p, which satisfies the following properties:
(1) (g h) · p = g · (h · p) for any g, h ∈ G and p ∈ X;

(2) e · p = p for p ∈ X, where e ∈ G is the identity.

The reader who is already familiar with the basic definitions may skip this paragraph. The orbit
of a point p ∈ X is the set G · p := {g · p | g ∈ G }. If A ⊂ X is any subset, we denote by G · A the
set of orbits of points in A, that is, [
G·A= G · p.
p∈A

The stabilizer of a point p ∈ X is the set of the elements of the group G not moving p, i.e.,
Gp = {g ∈ G | g · p = p }.
Theorem 4.7 (Class Formula). Let G be a finite group acting on a set X. For any p ∈ X it turns
out that
|G · p| · |Gp | = |G| . (4.1)
Definition 4.8 (Effective Action). Let G be a finite group acting on a set X. The action is said to
be effective if the associated kernel is trivial.

More precisely, the kernel K associated to an action is the intersection of all stabilizer subgroups
\
K= Gp .
p∈X

Therefore it is a normal subgroup of G, and thus the quotient group GK acts on X with trivial
kernel and identical orbits to the action of G. In particular, we can always assume without loss of
generality that an action is effective.
Definition 4.9 (Holomorphic Action). Let G be a finite group acting on a set X. The action is
said to be holomorphic if for every g ∈ G, the bijection

φg : X 3 p 7−→ g · p ∈ X

is a holomorphic map from X to itself, i.e., φg belongs to Aut(X).

Remark 4.2. The quotient space associated to an action, denoted by XG , is the set of the orbits.
Recall that the topology on XG is easily defined via the natural projection, i.e.,

U ⊂ XG is open ⇐⇒ π −1 (U ) is open in X.

Recall also that π is an open map when the action is continuous (or, even better, holomorphic).

34
Stabilizer Subgroups. In this short paragraph, we list some facts about the stabilizer subgroup
of a finite group G, acting holomorphically and effectively on a Riemann surface.
Proposition 4.10. Let G be a finite group acting holomorphically and effectively on a Riemann
surface X, and let p ∈ X be a fixed point.
(1) The stabilizer subgroup Gp is a finite cyclic group.
(2) If G is not finite, then Gp is still a cyclic group if it is finite.
(3) The points of X with nontrivial stabilizers are a discrete subset.

The Quotient Riemann Surface. In order to put a complex structure on the quotient surface
X , we must find complex charts.
G
Proposition 4.11. Let G be a finite group acting holomorphically and effectively on a Riemann
surface X, and let p ∈ X be a fixed point. Then there exists an open neighborhood U of p such that:

(a) U is invariant under the action of the stabilizer subgroup Gp ;

(b) U ∩ (g · U ) = ∅ for every g ∈
/ Gp ;

(c) the map UGp → XG , which sends a point q ∈ U to its orbit [q], is a homeomorphism onto
an open subset of the quotient XG ;
(d) no point of U , except p, is fixed by an element of Gp .
Sketch of the Proof. Suppose that G \ Gp = {g1 , . . . , gn } are the elements of G not fixing p. A Rie-
mann surface is, in particular, Hausdorff, thus, for each i = 1, . . . , n, we can find open neighborhoods
Vi of p and Wi of gi · p such that Vi ∩ Wi = ∅.
In particular, gi−1

· Wi is an open neighborhood of p as i ranges in {1, . . . , n}. Let us consider
Ri = Vi ∩ gi−1 · Wi , and let us set

\ n
[
U := g · R, where R = Ri .
g∈Gp i=1

This is exactly the sought open neighborhood of p, and it is now easy to check that it satisfies the
properties (a)-(d).
Theorem 4.12. Let G be a finite group acting holomorphically and effectively on a Riemann surface
X. The quotient XG is a Riemann surface, whose complex charts are given by Proposition 4.11.
Moreover, π : X → XG is a holomorphic map, whose degree is equal to |G|, such that multp (π) =
|Gp | for any point p ∈ X.

Ramification of the Quotient Map. Let G be a finite group acting holomorphically and effec-
tively on a Riemann surface X, and denote by Y = XG the quotient space.
Suppose that y ∈ Y is a branch point, and let x1 , . . . , xs be the points of X lying above y, i.e.,
π −1 (y) = {x1 , . . . , xs }. Clearly the xi ’s are all in the same orbit by definition, thus they all have
conjugate stabilizer subgroups, and each one of them is of the same order r.
Moreover, the number s of points in this orbit is the index of the stabilizer, and so is equal to
|G|/r. This argument proves the following lemma:

35
Lemma 4.13. Let G be a finite group acting holomorphically and effectively on a Riemann surface
X, and let Y = XG be the quotient.
For every branch point y ∈ Y , there is an integer r ≥ 2 such that π −1 (y) consists of exactly |G|/r
points of X, and at each of these preimage points π has multiplicity exactly equal to r.

Corollary 4.14. Let G be a finite group acting holomorphically and effectively on a Riemann surface
X, and let Y = XG be the quotient. Suppose that there are k branch points y1 , . . . , yk ∈ Y such
that, for each i = 1, . . . , n, π has multiplicity ri at the |G|/ri points above yi . Then
k
X |G|
2 g(X) − 2 = |G| (2 g(Y ) − 2) + (ri − 1) =
i=1
ri

" k  #
X 1
= |G| 2 g(Y ) − 2 + 1− .
i=1
ri

Lemma 4.15. Let r1 , . . . , rk be given integers such that ri ≥ 2 for each i. Let
k  
X 1
R := 1− .
i=1
ri

Then it turns out that

(a) 

 k =1 any r1 ;



R < 2 ⇐⇒ (k, {ri }) = k =2 any r1 , r2 ;


 k =3 {ri } = {2, 2, r3 }; or

k =3 {ri } = {2, 3, 3}, {2, 3, 4} or {2, 3, 5}.


(b) (
k=3 {ri } = {2, 3, 6}, {2, 4, 4} or {3, 3, 3}; or
R = 2 ⇐⇒ (k, {ri }) =
k=4 {ri } = {2, 2, 2, 2}.

1
(c) If R > 2 then R ≥ 2 + 42 .

Hurwitz’s Theorems on Automorphism For compact Riemann surfaces of genus bigger or

equal to 2, Corollary 4.14 leads to a bound on the order of the group G acting holomorphically and
effectively.
Theorem 4.16 (Hurwitz’s Theorem). Let G be a finite group acting holomorphically and effectively
on a compact Riemann surface X of genus g(X) ≥ 2. Then

|G| ≤ 84 · (g(X) − 1) .

Proof. By Corollary 4.14 it turns out that

h   i
2 g(X) − 2 = |G| 2 g XG − 2 + R , (4.2)

where R is defined as in the Lemma above.

36
    
(1) Suppose that g XG ≥ 1. If there is no ramification, i.e., R = 0, then g XG ≥ 2 (since
g(X) − 2 > 0), and this implies immediately that

|G| ≤ g(X) − 1.
 
If the ramification is nonzero, i.e., R 6= 0, then R ≥ 1/2. Therefore 2 g XG − 2 + R ≥ 1/2,
and from (4.2) it follows that
|G| ≤ 4 · (g(X) − 1) .
 
(2) Assume then that g XG = 0. Then (4.2) reduces to

2 g(X) − 2 = |G| [R − 2] ,

which forces R > 2. Therefore Lemma 4.15 implies that R − 2 ≥ 1/42, i.e,

|G| ≤ 84 · (g − 1),

as claimed.

In fact, the group of all automorphisms of a compact Riemann surface of genus at least two is a
finite group. It implies that for such a Riemann surface, we have

|Aut(X)| ≤ 84 · (g(X) − 1) ,

since the full group of the automorphisms certainly acts holomorphically and effectively on X; we
shall prove this later on in the course.

37
Chapter 5

Differential Forms

The main result of this chapter is the residue theorem for compact Riemann surfaces, which will be
extremely useful in some of the most significant results of this course (e.g., Serre duality theorem).

5.1 Holomorphic 1-forms

Definition 5.1 (Holomorphic 1-Forms). Let V be an open subset of C. A holomorphic 1-form (in
the coordinate z) on V is an expression ω of the form

ω = f (z) dz,

where f : V → C is a holomorphic function.

Let ω1 = f (z) dz, ω2 = g(w) dw be holomorphic 1-forms, respectively in the coordinates z and
w, defined on open subsets V1 , V2 ⊂ C.
Definition 5.2 (Transformation). Let T : V2 → V1 be a holomorphic map such that z = T (w). We
say that ω1 transforms to ω2 under T if and only if

g(w) = f (T (w)) · T 0 (w), ∀ w ∈ V2 .

Remark 5.1. If T is an invertible transformation and S is its inverse, then ω1 transforms to ω2

under T if and only if ω2 transforms to ω1 under S.
Definition 5.3 (Meromorphic 1-Forms). Let V be an open subset of C. A meromorphic 1-form (in
the coordinate z) on V is an expression ω of the form

ω = f (z) dz,

where f : V → C is a meromorphic function.

Let ω1 = f (z) dz, ω2 = g(w) dw be meromorphic 1-forms, respectively in the coordinates z and
w, defined on open subsets V1 , V2 ⊂ C.

Definition 5.4 (Transformation). Let T : V2 → V1 be a holomorphic map such that z = T (w). We

say that ω1 transforms to ω2 under T if and only if

g(w) = f (T (w)) · T 0 (w), ∀ w ∈ V2 .

38
Differential forms on Riemann surfaces. In this paragraph, we extend, in a natural way, the
definition of holomorphic 1-forms on Riemann surfaces.
We denote by X a Riemann surface, and we let A = {ϕα : Uα → Vα ⊆ C}α∈I be a complex atlas
associated to X.
Definition 5.5 (Holomorphic Form, [2]). A holomorphic 1-form on X is a collection of holomorphic
1-forms {ωφ }, one for each chart φ : U → V in the coordinate of the codomain V , such that if two
charts have overlapping domains, then the associated holomorphic 1-form ωφ1 transforms to ωφ2
under the change of coordinate T := φ1 ◦ φ−1
2 .

On the other hand, to define a holomorphic 1-form on a Riemann surface one does not need to
give a holomorphic 1-form on each chart, but only the charts of some atlas.
Lemma 5.6. Let {ωα } be a given collection of 1-forms, one for each chart of the atlas A, which
transform to each other on their overlapping domains. Then there exists a unique holomorphic
1-form on X extending this collection on any of the charts of X.
Definition 5.7 (Meromorphic Form, [2]). A meromorphic 1-form on X is a collection of meromor-
phic 1-forms {ωφ }, one for each chart φ : U → V in the coordinate of the codomain V , such that
if two charts have overlapping domains, then the associated holomorphic 1-form ωφ1 transforms to
ωφ2 under the change of coordinate T := φ1 ◦ φ−1
2 .

Order. Let p ∈ X be a point and let ω be a meromorphic 1-form, defined in a neighborhood

U ⊂ X of a point p.
Let z be a local coordinate centered at p, in such a way that ω = f (z) dz for some function
f : U → C, meromorphic at the point z = 0.
Definition 5.8. The order of ω at p, denoted by ordp (ω), is the order of the function f at the
origin z = 0, i.e.,
ordp (ω) = ord0 (f ).
It is a simple exercise to prove that this definition does not depend on the particular local
representation f of ω, nor on the neighborhood U of p.

C ∞ Forms. Let X be a Riemann surface and let p ∈ X be a point. If we take a chart

∼
ϕ : Up −
→V ⊆C

centered at p, with local coordinate z, then a straightforward computation yields to the following
result:    
∂ 1 ∂ ∂ ∂ 1 ∂ ∂
= −ı , = +ı .
∂z 2 ∂x ∂y ∂ z̄ 2 ∂x ∂y
In particular, a function f : V ⊂ C → C is holomorphic at p if and only if
∂f
(p) = 0.
∂ z̄
Therefore the differential of a C ∞ function f : Up ⊂ X → C is given by
∂f ∂f
df = dz + dz̄.
∂z ∂ z̄
A 1-form ω of class C ∞ is an expression of the form

ω = g1 (z, z̄) dz + g2 (z, z̄) dz̄,

39
and it is holomorphic if and only if g1 is a holomorphic function not depending on z̄, and g2 is
identically equal to 0, i.e., a holomorphic 1-form of class C ∞ is an expression of the form

ω = g1 (z) dz.

In conclusion, if ω is a 1-form of class C ∞ , then its differential dω is a 2-form and it is given by the
formula  
∂ g2 ∂ g1
dω = − dz ∧ dz̄.
∂z ∂ z̄

5.2 Integration of a 1-form along paths

Path Integration. Let X be a Riemann surface and let
∼
ϕ:U −
→V ⊆C

be any chart. If γ : [a, b] → X is a piece-wise differentiable path such that γ ([a, b]) ⊂ U , then the
composition ϕ ◦ γ : [a, b] → V ⊂ C is also a path, sending t to z(t).
If we identify U ' V , then the 1-form can be locally written in the form ω = g1 (z, z̄) dz +
g2 (z, z̄) dz̄ and thus we can define the integral along γ as follows:
Z Z b
ω := [g1 (z(t), z̄(t)) · z 0 (t) + g2 (z(t), z̄(t)) · z̄ 0 (t)] dt. (5.1)
γ a

If γ : [a, b] → X is a generic path, then γ ([a, b]) is a compact set in X and thus there exist a
finite number of charts ϕ1 : U1 → V1 , . . . , ϕn : Un → Vn such that
n
[
γ([a, b]) ⊆ Ui .
i=1

If we let γi := γ Ui , we can define the integral of ω along γ as
Z n Z
X ai
ω := [g1, i (z(t), z̄(t)) z 0 (t) + g2, i (z(t), ·z̄(t)) · z̄ 0 (t)] dt. (5.2)
γ i=1 ai−1

Winding Number. Let γ : [a, b] → C be a closed path around the origin. The integral
Z
1 1
Iγ (0) := dz
2πı γ z

is called winding number of γ and, intuitively, it counts the number of complete rotations around
the origin.
More precisely, it depends only on the class of homotopy of γ, and it is easy to prove that, if γ
is homotopic to S 1 counterclockwise oriented, then Iγ (0) = 1.
Lemma 5.9. Let X be a Riemann surface, and let p ∈ X be any point. If ω is a meromorphic
1-form locally defined on a chart Up around p and γ is a simple path, contained in Up , not enclosing
any other pole of f , then Z
1
Resp (ω) = ω.
2πı γ

40
 Figure 5.1: Covering of the path γ

Recall that the residue of ω at a certain point p ∈ X is defined by looking at ω locally. More
precisely, if ω = f (z) dz in a neighborhood of p, then
X
f (z) = cn z n ,
n≥m

and the residue is exactly the coefficient of 1/z, i.e., c−1 .

Theorem 5.10 (Stokes). Let X be a Riemann surface and let D ⊂ X be a triangulable domain,
whose border is piece-wise differentiable. If ω is a C ∞ 1-form on X, then
Z ZZ
ω= dω. (5.3)
∂D D

Theorem 5.11 (Residues Theorem). Let X be a compact Riemann surface and let ω be a mero-
morphic 1-form on X. Then the sum of the residues is zero, i.e.,
X
Resp (ω) = 0.
p∈X

Proof. The set of poles of ω is a discrete subset of X, thus it is finite by compactness of X. As-
sume that p1 , . . . , pn are the poles of ω, and let γ1 , . . . , γn be simple paths enclosing only the
corresponding pole pi . Let Di be closed sets such that ∂ Di = γi , and let D = X \ ∪ni=1 Di ; then
Z
ω = 2 π ı Respi (ω), ∀ i = 1, . . . , n.
γi
Pn
Formally ∂ D = − i=1γi (see Figure 5.2), therefore
Xn Xn Z Z ZZ
2πı Respi (ω) = ω=− ω=− dω = 0,
i=1 i=1 γi ∂D D

41
since ω is holomorphic on D.
Corollary 5.12. Let X be a compact Riemann surface and let f : X → C a nonconstant meromor-
phic function. Then the sum of the orders is zero, i.e.,
X
ordp (f ) = 0.
p∈X

Proof. First, we observe that ordp (f ) = n if and only if - locally - it turns out that

f (z) = cn z n + O(z n+1 ).

Consider the logarithmic differential

1
ω= df,
f
and notice that the differential of f is (locally) defined by

df = f 0 (z) dz = n cn z n−1 dz + . . . .

In particular, multiplying by 1/f we find that

1 n 
df = + ... ,
f z
and thus the key identity  
1
ordp (f ) = Resp df ,
f
which concludes the proof since the sum of the residues is zero by Theorem 5.11.

42
Figure 5.2: Idea of the Residues Theorem proof

43
Chapter 6

Sheaf Theory

In this chapter, we want to introduce and develop the sheaf language to simplify the comprehension
of the next topics.
In particular, we present the cohomology groups (which will ease the discussion of the divisor
vector space L(D)), and we also prove the long exact sequence in cohomology theorem (which will
be used extensively in the next chapters.)
The reader may consult these notes for a more in-depth dissertation.

6.1 Definitions and First Properties

Definition 6.1 (Sheaf). Let X be a topological space. A sheaf F on X associates to each open set
U ⊂ X an abelian group F(U ), along with a restriction map ρUV : F(V ) → F(U ) for any open sets
U ⊂ V , satisfying the following conditions:

(1) Compatibility Conditions.

(1.1) F(∅) = 0.
(1.2) ρU
U = idU .
(1.3) If W ⊂ V ⊂ U , then ρU V U
W = ρW ◦ ρV .

(2) Locality Conditions.

(2.1) If U := {Ui }i is a covering of U , then

σ ∈ F(U ) : σ Ui = 0 ∀ i =⇒ σ ≡ 0.

(2.2) For any covering U := {Ui }i∈I of U and any collection {σi }i∈I of sections σi ∈ F(Ui ) it
turns out that, if  
σi Ui ∩Uj = σj Ui ∩Uj

for any i, j, then there exists σ ∈ F(U ) such that σ Ui = σi .
If F satisfies only the compatibility conditions, then we say that F is a presheaf.
Definition 6.2 (Stalks). Let F be a sheaf on a topological space X, and let p ∈ X. The stalk at
p, denoted by Fp , is the direct limit of all sections containing p.

44
 More precisely, suppose that U and V are two open subsets, both containing p, with two section
σU and σV ; define an equivalence relation
 
σU ∼ σV ⇐⇒ ∃ W ⊂ U ∩ V : σU W = σV W .

Then the stalk at p is defined by setting

 
G
Fp := lim F(U ) =  F(U ) / ∼ .
−→
U 3p U 3p

There is a group homomorphism ρU : F(U ) → Fp mapping a section σα to its equivalence class,

and the image is called the germ of σα .

Example 6.1. To get accustomed with the definitions, the reader may try to check that the following
are all sheaves.
(a) The (locally) constant sheaf U → 7 C(U ) = C, denoted by C, whose restriction maps are the
identities between the copies of C.

(b) The sheaf of holomorphic functions U 7→ OX (U ) := {f : U → C | f is holomorphic in U }. For

any p ∈ X the stalk is given by

OX, p := {f : X → C | f is meromorphic outside of p } .

(c) The sheaf of meromorphic functions U 7→ MX (U ) := {f : U → C | f is meromorphic in U }.

(d) The sheaf of the holomorphic differentials U 7→ Ω1X (U ) := {f (z) dz | f is holomorphic in U }.

Morphisms of Sheaves. Let E and F be sheaves on X. A morphism f : E → F is a collection

of group homomorphisms fU : E(U ) → F(U ) such that the following diagram is commutative

fU
E(U ) F(U )

ρV
U ρV
U

fV
E(V ) F(V )
 

that is, (fU (σ)) V = fV σ V .
Example 6.2 (Inclusion Maps). The first kind of morphisms between sheaves we study are the
inclusion maps. Indeed, they come up whenever, for any U ⊂ X, the group F(U ) is a subgroup of
the group G(U ).
(1) Constant sheaves: Z ⊂ R ⊂ C.
(2) Holomorphic/Meromorphic sheaves: C ⊂ OX ⊂ MX .
∗
(3) Nonzero holomorphic/meromorphic sheaves: OX ⊂ M∗X .

(4) Sheaves of functions with bounded poles: if D1 ≤ D2 are divisors on X, then OX [D1 ] ⊂ OX [D2 ].

45
Kernel. Suppose that φ : F → G is sheaf map between two group sheaves on X. Define a subsheaf
K ⊂ F, called the kernel of φ, by setting

K(U ) := ker(φU )

for any U ⊆ X, that is, the group associated to the open set U is exactly the kernel of the group
homomorphism φU : F(U ) → G(U ).
Proposition 6.3. Let φ : F → G be a sheaf map between two group sheaves on X.
(a) The kernel K is a sheaf.
(b) The cokernel (which is defined in the same way) is a presheaf, but it is generally not a sheaf.

Associated Sheaf. If F is a presheaf, then it is always possible to extend it to a sheaf Fe, which
is usually called associated sheaf.
For example, we define coker(f )(U ) to be a collection of sections σα ∈ G(Uα ) for an open covering
{Uα }α of U , such that for all α and β it turns out that
 
σα U ∩U − σβ U ∩U ∈ fUα ∩Uβ (F(Uα ∩ Uβ )) .
α β α β

The definition depends on the choice of an open covering. Thus we need to find a way to get rid of
this obstacle.
The idea, as we shall see also later in the course, is to use the direct limit. More precisely, we
identify two collections {(Uα , σα )} and {(Vβ , σβ )} if for all p ∈ Uα ∩ Vβ there exists an open set W
satisfying p ∈ W ⊂ Uα ∩ Vβ , such that
 
σα W − σβ W ∈ fW (F(W )) .

This identification yields an equivalence relation and correspondingly we define the coker sheaf as
the group of equivalence classes of the above sections.

6.2 Exact Sequences

In this section, we introduce a fundamental notion that is used vastly in algebra and geometry: short
exact sequences, and the long exact sequence in cohomology theorem.

Short Exact Sequences of Sheaves. We say that a sequence of sheaf maps

φ
0→
− K→
− F−
→G→
− 0

is a short exact sequence if the sheaf map φ is surjective and K is the kernel sheaf associated to φ.
Remark 6.1. There is an equivalent - and maybe more useful - definition of a short exact sequence
of sheaf maps, which relies on the notion of a short exact sequence of abelian groups. More precisely,
the sequence of sheaf maps
φ ψ
0→
− F− →G− →H→ − 0
is an exact short sequence if and only if
φp ψp
0→
− Fp −→ Gp −−→ Hp →
− 0

is an exact sequence of abelian groups, for every p ∈ X.

46
Example 6.3. Here is a brief list of short exact sequences.
(a) On a Riemann surface, the sequence
d=∂
0→ − OX −−−→ Ω1X →
− C→ − 0

is exact, since the kernel sheaf of the differential map is exactly the (locally) constant sheaf.
(b) The sequence
∗ e2π i ·
0→
− Z→
− OX −−−→ OX →
− 0
is exact, since the kernel sheaf of the exponential map is exactly the integer-valued (locally)
constant sheaf.

(c) The sequence

1, 0 ∂¯
− Ω1X →
0→ − EX − 2
→ EX →
− 0
is exact.
(d) The sequence
d
− C∞ − 1 2


− C→
0→ → ker d : EX → EX →
− 0
is exact, since the kernel of d in this setting is exactly the constant functions space.

Sheaves on Riemann surfaces. Let X be a compact Riemann surface and let p ∈ X. The
skyscraper sheaf centered at p is defined as
(
0 if x 6= p
(Cp )x =
C if x = p.

We can also define the sheaf of holomorphic functions such that p is a zero, that is,

JX, p = {f is holomorphic and f (p) = 0} ,

which can be easily denoted using the divisors. In fact, if we let f ∈ Jp , then it is easy to prove that
div(f ) ≥ p, and thus we may denote it by

JX, p = OX [−p].

Proposition 6.4. Let X be a compact Riemann surface. There exists an exact sequence of sheaf
maps
f 7→f (p)
0→
− OX [−p] →
− OX −−−−−→ Cp →
− 0.
Idea of the Proof. Notice that, if x 6= p, then

(Cp )x = 0 and (OX [−p])x ∼

= OX, x

and that
(OX [−p])p ∼
= {maximal ideals in OX, p } and OX, p ∼
MX = Cp .

47
6.3 Čech Cohomology of Sheaves
In this section will assume that every covering is locally finite. This assumption is by no means
necessary at this point, but it will come in handy (and, actually, necessary) soon.

Čech Cochains. Let F be a sheaf of abelian groups on a topological space X. Let U := {Ui }i
be an open covering of X, and fix an integer n ≥ 0. For every collection of indices (i0 , . . . , in ), we
denote the intersection of the corresponding open sets by

Ui0 , ..., in := Ui0 ∩ · · · ∩ Uin .

The deletion of the k-th index is denoted by the symbol iˆk , and it is clear by the definition that

Ui0 , ..., in ⊂ Ui0 , ...,ibk , ... in .

Definition 6.5. A Čech n-cochain for the sheaf F over the open cover U is a collection of sections
of F, one over each Ui0 , ..., in .

The space of Čech n-cochains for F over U is denoted by C q n (U, F). In particular, a Čech 0-
cochain is simply a collection of sections, that is, one gives a section of F over each open set in the
cover. Similarly, a 1-cochain is a collection of sections of F over every intersection of two open sets
of the cover; the typical notation for a 1-cochain is (fi, j ) ∈ F (Ui ∩ Uj ).
Remark 6.2. If φ : F → G is a sheaf map, then it induces map on the cochains space
q n (U, F) → C
φ:C q n (U, G)

for any open covering U, defined by

(fi0 , ..., in ) 7−→ (φ(fi0 , ..., in )) .

Čech Cochains Complexes. Define a co-boundary operator

q n (U, F) → C
d:C q n+1 (U, F)

by setting
d ((fi0 , ..., in )) := (gi0 , ..., in+1 ),
where
n+1
X  
gi0 , ..., in+1 = (−1)k ρ fi0 , ..., ibk , ..., in+1 .
k=0

In the above formula ρ denotes the restriction map for the sheaf F corresponding to the inclusion
Ui0 , ..., in ⊂ Ui0 , ...,ibk , ... in .
Any n-cochain c with d(c) = 0 is called a n-cocycle; the space of n-cocycles is denoted by
n
Z (U, F) and it is simply the kernel of d at the n-th level.
q
Any n-cochain c with c = d(c0 ) for some (n − 1)-cochain c0 is called a n-coboundary; the space of
n-coboundaries is denoted by B q n (U, F).
It is straightforward, but tedious, to prove that d ◦ d = 0. Thus we have a Čech cochain complex
0 q0 d q1 d q2 d
0−
→C (U, F) −
→C (U, F) −
→C (U, F) −
→ ....

48
Cohomology with respect to a Cover. The fact that d2 = 0 implies that for every n ∈ N
q n (U, F) ⊂ Zqn (U, F).
B

Definition 6.6 (Cohomology). The nth Čech cohomology group H

|n (U, F) of F with respect to
the open cover U is the quotient group
n
|n (U, F) = Zq (U, F)q n
H B (U, F).

There is a lot of work behind the following definition (see [2, pp. 295–297]), but the main point
is that we can define a Čech cohomology group independent of the cover U. In fact, one introduces
a refinement of U and proves that the cohomologies can be compared and they depend only on the
particular coverings.

Definition 6.7 (Čech Cohomology). Fix a sheaf F and a integer n ≥ 0. The nth Čech cohomology
group of F on X is the group
H n (X, F) = lim H
|n (U, F).
−→
U

Proposition 6.8. Let X be a complex manifold, paracompact and smooth. If R is the constant sheaf
on X, then
n
H n (X, R) = Hdr n
(X, R) = Hsing (X, R).
Remark 6.3. There is an isomorphism

H 0 (X, F) ∼
= F(X).

In fact, the 0-coboundary is given by {0}, while the 0-cycle is given by

Zq0 (X, F) = {{fγ }γ∈Γ | fα − fβ = 0 in Uα ∩ Uβ for any α, β ∈ Γ } .

Therefore fα = fβ in the intersection Uα ∩ Uβ easily implies that it is possible to extend both fα

and fβ to a function f in X.
In particular, if X is a compact Riemann surface, then the 0 cohomology group of the holomorphic
sheaf is given by
H 0 (X, OX ) = OX (X) = C,
since a holomorphic function defined on the whole compact X is bounded and thus constant.

6.4 Sheaves of OX -modules

Sheaves of OX -modules. Let X be a complex manifold (i.e., a manifold with a holomorphic
structure).

Definition 6.9 (Coherent). A sheaf F of OX -modules is coherent if and only if for any p ∈ X there
exist an open neighborhood U ⊂ X and an exact sequence
s r
OX (U ) →
− OX (U ) →
− F(U ) →
− 0,

i.e, if and only if it is finitely presented.

49
Example 6.4. Let X be a Riemann surface. Then the holomorphic 1-form sheaf Ω1X is invertible
and the isomorphism is given by

Ω1X (U ) 3 f (z) dz 7−→ f (z) ∈ OX (U ).

Example 6.5. If dimC X = n, then

O(X)n  Ω1X (U ), (f1 , . . . , fn ) 7−→ f1 dz1 + · · · + fn dzn ,

where z1 , . . . , zn are local coordinates. Clearly the sheaf is coherent but it is not invertible.

Fundamental Properties. In this paragraph we briefly discuss some of the most important
properties of the particular class of sheaves we just introduced.
Definition 6.10 (Support). Let F be a sheaf on X. The support of F is defined as the set of the
nontrivial stalks, that is,
spt (F) = {x ∈ X | Fx 6= 0 } .
Proposition 6.11. Let X be a compact complex manifold (with holomorphic structure) and let F
be a coherent sheaf over X.
(a) The p-th cohomology group H p (X, F) is a finite-dimensional C-vector space.

(b) The p-th cohomology group is zero for any p > dimC X.
Corollary 6.12. Let X be a compact Riemann surface and let F be a coherent sheaf over X. Then
the p-th cohomology group is given by
(
p = 0 if p ≥ 2
H (X, F)
6= 0 if p = 0, 1.

Theorem 6.13. Let X be a compact complex manifold (with holomorphic structure). If

ϕ ψ
0→
− F−
→G−
→H→
− 0

is an exact sequence of coherent OX -modules, then there is a long exact sequence in cohomology,
that is, there exists ∂ such that

H 0 (ϕ) H 0 (ψ) ∂
− H 0 (F) −−−−→ H 0 (G) −−−−→ H 0 (H) −
0→ → H 1 (F) →
− ...

is an exact long sequence.

Idea of the Proof. We can always choose a covering U := {Uα }α∈I such that, for any Uα , it turns
out that
ϕα ψα
0→
− F(Uα ) −−→ G(Uα ) −−→ H(Uα ) →− 0
is a short exact sequence. Consequently, the sequence

H 0 (ϕ) H 0 (ψ)
− H 0 (F) −−−−→ H 0 (G) −−−−→ H 0 (H)
0→

is exact, but the latter map may not be surjective (in general, it will not be). We want to define a
map ∂ : H 0 (H) → H 1 (F) that makes the sequence exact at H 0 (H).

50
 Let σ ∈ H 0 (H) and let us consider two open sets of the covering, e.g., Uα and Uβ . We already
know that there exists sα ∈ G(Uα ) such that sα 7→ σ Uα and there exists sβ ∈ G(Uβ ) such that

sβ 7→ σ  . Let us set δ̃(σ) := sα − sβ = gα, β .
Uβ
Clearly gα, β ∈ Z 1 (Uα ∩ Uβ ) ⊆ G (Uα ∩ Uβ ) and by the exactness of the local sequence, it turns
out that ψ(gα, β ) = 0 in H (Uα ∩ Uβ ) and therefore there exists fα, β ∈ F (Uα ∩ Uβ ) such that

ϕ(fα, β ) = gα, β .

We can set
∂(σ) := {fα, β }α, β ∈ H 1 (F)
and, by using the induction principle, we conclude the proof.
Corollary 6.14. Let X, F, G and H be as above. Then
dimC (spt(F ))
X
(−1)i dimC H i (F) .


χ(G) = χ(F) + χ(H), where χ(F) :=
i=0

6.5 GAGA Principle

Let X be a projective manifold, equipped with the Zariski topology, and let
alg
OX := {f : X → C | f algebraic } .

On the other side, let X be a compact holomorphic manifold, equipped with the Hausdorff topology,
and let
h
OX := {f : X → C | f holomorphic } .
Theorem 6.15. Let X be a smooth projective manifold. Then there exists an application
O
F 7−→ F h := F h
OX ,
alg
OX

alg
where F is a coherent sheaf of OX -modules and F h a coherent sheaf of OX
h
-modules, such that
i
HZar (X, F) ∼ i
= HHau (X, F h ).

6.6 Invertible OX -module Sheaves

Definition 6.16 (Invertible). A sheaf L of OX -modules is invertible if and only if there exists a
covering U := {Ui }i∈I of X such that:
∼
(a) For any i ∈ I, there is an isomorphism φi : L(Ui ) −
→ OX (Ui ).
(b) For any i, j ∈ I, there is an invertible function fi, j , defined on Ui ∩ Uj , such that φi = fi, j · φj .

Remark 6.4. Let L be an invertible sheaf of OX -modules and let Ui , Uj , Uk ∈ U. It follows from
the definition that, in the triple intersection, the following relation holds true:

fi, k = fi, j · fj, k .

51
Remark 6.5. Equivalently, a sheaf F of OX -modules  is invertible if and only if for every p ∈ X
there is an open neighborhood U of p, such that O U ∼

= F U as sheaves of O U -modules on the
space U .  
The invertible sheaves are locally free rank one O-modules. An isomorphism φU : O U → F U
is called a trivialization of F over U .

Remark 6.6. By definition, if U is an open neighborhood of p such that O U ∼

 
= F U , then there
is an isomorphism φU : O(U ) → F(U ).
Let us consider the complex projective line P1 (C) endowed with the usual atlas given by U0 =
{[1 : z1 ]}, with local coordinate z := z1 /z0 , and U1 = {[z0 : 1]}, with local coordinate w := z0 /z1 .

Example 6.6. For any fixed m ∈ Z, we consider the invertible sheaf OP1 [m], which is defined as
follows:
(a) for i = 1 and i = 2, there are isomorphisms OP1 [m](Ui ) ∼
= O(Ui ) ∼
= O(C);
(b) the transition map is given by  m
z1
f1, 0 = .
z0

We are interested in computing the 0-cohomology group of OP1 [m] for m = 0, m = 1 and m ∈ Z.

(1) If m = 0, then it is straightforward to prove that

H 0 P1 , OP1 = C.


OP1 [0] = OP1 and

(2) Let m = 1. By definition of the sheaf OP1 [1], it turns out that

∃ f (z) ∈ OP1 (U0 ) such that f is holomorphic in U0 ,

∃ g(z) ∈ OP1 (U1 ) such that g is holomorphic in U1 .

To compute the group H 0 P1 , OP1 we need to check how f and g glue in the intersection
U0 ∩ U1 . The assumption (b) on the transition map easily implies that
 
1
f (z) = z g , ∀ z ∈ U0 ∩ U1 .
z

If we use the Laurent develop, it turns out that

X X
f (z) = ai z i and g (z) = bi z −i ,
i≥0 i≥0

hence f (z) = z · g(1/z) in the intersection if and only if

b−1
f (z) = a0 + a1 z and g(z) = + b0 .
z
Therefore we can easily conclude that
( )
homogeneous polynomials
H 0 P1 , OP1 [1] ∼


= .
of degree 1 in z0 and z1

52
 (3) Let m > 0 be any positive integer. There is an isomorphism
(  )
 homogeneous polynomials
∼
= H 0 P1 , OP1 [m] .


p(z0 , z1 ) 

 of degree m in z0 and z1
which is defined by  p(z , z )
 z0m 1 in U0
0
p(z0 , z1 ) 7−→
 p(z0 , z1 ) in U1 .
z1m
The same result holds true for m < 0, but there are no polynomials of degree less than zero.
In particular, the H 0 cohomology group is the trivial one, that is,
H 0 P1 , OP1 [m] = 0, ∀ m ∈ Z− .

6.7 Operations on Sheaves

Tensor Product. Let L and F be invertible sheaves of OX -modules. Their tensor product is the
sheaf denoted by L ⊗ F which is defined, in terms of co-cycles, as follows:
(L, F) 3 (`i, j , fi, j ) 7−→ `i, j · fi, j ∈ L ⊗ F.
For example, if m, k ∈ Z, it is relatively easy to prove that the following isomorphism exists:
OP1 [m] ⊗ OP1 [k] ∼
= OP1 [m + k].

Inverse. Let L be an invertible sheaf. The inverse is denoted by L−1 and it is the unique sheaf
such that
L ⊗ L−1 = OX .
In particular if X is a Riemann surface or, more generally, a complete holomorphic manifold
 O
{invertible sheaves} , is a group.

For a more precise formulation of the above argument, the reader may jump directly to Section 10.1.

Line Bundle. There is a correspondence between invertible sheaf on a smooth manifold X and
line bundles, that is, ( )
Invertible sheaves ∼
←→ {Line bundles} .
on X smooth
If X is an invertible Riemann surface, then
∃ U := {Ui }i∈I covering such that
∼

{F → X Line bundle} ←→ Fi : F Ui → Ui × C which sends z to (z, fi (z))
and fi = gi, j · fj in the intersection Ui ∩ Uj .
A holomorphic section of F is simply the holomorphic mapping fi : Ui → C such that
fi = gi, j · fj in Ui ∩ Uj .
In particular, there is a correspondence
∼
F sheaf of the sections of F ←→ F → X Line bundles,
given by
fi (and gi, j ) 7−→ F → Ui × C : z 7→ (z, fi (z)) .

53
 Figure 6.1: Idea of the construction

Algebraic Curves. Let F be an invertible sheaf. Clearly it corresponds locally to φi = z ni which

is 0 if and only if z = 0 is a point of multiplicity ni (an analogous argument works for ∞).
If X ⊂ P2 (C) is a smooth algebraic curve of degree d, then we can always consider the vector
subspace made up of fixed degree smooth algebraic curves, that is,

W := {smooth algebraic curves of fixed degree h} ⊂ P2 (C).

By Bezout’s theorem1 , it follows that for any Y ∈ W , the intersection Y ∩ X is given by h · d points
with the right multiplicities (this assertion is not very precise, but we only want to give a rough idea
of this construction).
Let Y1 , Y2 ∈ W and consider the points pi ∈ X ∩ Y1 and qi ∈ Y2 ; since W is a vector subspace,
also the linear combinations of these two elements will individuate Y3 ∈ W , such that the points
ri ∈ X ∩ Y3 are linear combinations of the previous points (see the Figure 6.1).

1 Let X and Y be plane algebraic curves of degree n and m respectively. Then there are m · n points in the

intersection X ∩ Y , counted with the respective multiplicities, provided that X and Y have no common components.

54
Chapter 7

Divisors

7.1 Divisors on Riemann Surfaces

Definition 7.1 (Divisor). Let X be a Riemann surface. A divisor D on X is a discretely supported
function D : X → Z, that is, a formal sum
X
D= D(p) · p,
p∈X

where D(p) ∈ Z is equal to the multiplicity of D at p, and D(p) 6= 0 for only finitely many p ∈ X.

Divisors Group. Given D1 and D2 divisors on X, there is a sum operation which is defined by
setting X
D1 + D2 = [D1 (p) + D2 (p)] · p,
p∈X

and it is easy to prove that D1 + D2 is still a divisor on X. In particular, if we denote the divisors
on X by Div(X), it turns out that (Div(X), +) is a commutative group.
Definition 7.2 (Degree). Let X be a Riemann surface. The degree is the mapping
X X
deg : Div(X) → Z, D= D(p) · p 7−→ D(p).
p∈X p∈X

Principal Divisors. Let X be a compact Riemann surface and let f : X → C ∪ {∞} be a

meromorphic function. There is a mapping div : M (X; C ∪ {∞}) → Div(X) defined by setting
X
div(f ) := ordp (f ) · p.
p∈X

The divisor associated to a function is called principal and, by the Residues Theorem 5.11, it turns
out that the degree is always equal to 0, that is,
X
deg (div(f )) = ordp (f ) = 0.
p∈X

Example 7.1. Let X = P1 (C) and let f (z0 , z1 ) = z0 (z0 − z1 ) z1−2 . The principal divisor associated
to f is given by
div(f ) = 1 · [0 : 1] + 1 · [1 : 1] − 2 · [1 : 0],
coherently with the properties already discussed above.

55
Definition 7.3 (Poles/Zeros divisor of f ). Let X be a compact Riemann surface and let f : X →
C ∪ {∞} be a meromorphic function. The divisor of the zeros is defined as
X
div0 (f ) := ordp (f ) · p,
p : ordp (f )≥0

while the divisor of the poles is defined as

X
div∞ (f ) := (−ordp (f )) · P.
p : ordp (f )≤0

Definition 7.4 (Effective Divisor). Let X be a compact Riemann surface. A divisor D ∈ Div(X)
is effective if D(p) ≥ 0 for every p ∈ X.
Consequently, any divisor D ∈ Div(X) may be written as a difference between two effective
divisors, that is,
D = D0 − D∞ .
Finally, there is a partial order on the set of all divisors which is defined by

D1 ≥ D2 ⇐⇒ D1 − D2 ≥ 0 ⇐⇒ D1 (p) − D2 (p) ≥ 0 ∀ p ∈ X.

7.2 Invertible Sheaf of OX -modules associated to a divisor D

Let X be a compact Riemann surface. The sheaf of OX -modules associated to a divisor D ∈ Div(X)
is denoted by OX [D], and it is defined by setting

X ⊇ U 7−→ OX [D](U ) := {f : U → C | f is meromorphic and div(f ) ≥ −D } .

Proposition 7.5. Let X be a compact Riemann surface and let D ∈ Div(X). Then OX [D] is an
invertible sheaf.
Proof. Let spt (D) = {p1 , . . . , pn }. If we set U0 := X \ {p1 , . . . , pn }, then OX [D] (U0 ) consists of
meromorphic functions with no poles, that is, it is isomorphic to the group of holomorphic function
defined on U0 :
OX [D] (U0 ) ∼
= OX (U0 ).
For any i = 1, . . . , n we may choose a neighborhood Ui 3 pi such that Ui ∩ Uj = ∅ whenever i 6= j.
If the multiplicity
 of pi is equal to ni , then we may locally (in Ui ) choose ϕi = z ni in such a way
that D Ui = div(ϕi ) for any i = 1, . . . , n. Thus there is an isomorphism


∼ 1
OX (Ui ) −
→ OX [D](Ui ), f 7−→ · f.
ϕi
More precisely, there is an equivalence
 
∼ g
OX [D](Ui ) = {f : Ui → C | f is meromorphic and ordpi (f ) ≥ −ni } −
→ f= ,
ϕi
where g : Ui → C is holomorphic. Indeed, in a more general setting than X Riemann surface, it
turns out that the transition maps are given by
ϕj
fi, j = .
ϕi

56
Definition 7.6. Let X be a compact Riemann surface and let D ∈ Div(X). The 0th cohomology
group is the vector space of the global sections of OX [D], and it is denoted by L(D).
More precisely, it turns out that
L(D) := H 0 (X, OX [D]) = {f : X → C | f meromorphic and div(f ) ≥ −D } .
Example 7.2. Let X = P1 (C), p = [1, 0] and D = 1 · p. If we denote by z = z1 /z0 the local
coordinate in U0 = {[1 : z1 ]} ∼
= C, then p = 0 in U0 . By definition
OP1 [D] (U0 ) = {f : U0 → C | f meromorphic and ord0 (f ) ≥ −1 } ,
hence the following equality also holds true:
 
g(z)
OP1 [D] (U0 ) = | g : U0 → C holomorphic in U0 .
z
Assume that w is the local coordinate of U1 = {[z0 : 1]} ∼= C; it remains to study how the functions
behaves in the intersection U0 ∩ U1 . If we use the Laurent develop, it turns out that
X X X
f (z) = ai z i and f (w) = ai wi = ai z −i ,
i≥−1 i≥−1 i≥−1

hence f (z) = f (w) in the intersection if and only if

a−1
f (z) = + a0 and f (w) = a0 + a−1 w.
z
Therefore we can easily conclude that
na o
−1
OP1 [D] (U0 ) = + a0 and OP1 [D] (U1 ) = {a−1 w + a0 } ,
z
that is, there is an isomorphism (see Example 6.6)
OP1 [D] ∼
= OP1 [1].
We conclude this section with a brief discussion of L(D), as D ranges in the divisor group of a
Riemann surface X. First notice that if D1 ≤ D2 , then there is a natural inclusion L(D1 ) ⊆ L(D2 ).

Empty L(D). Recall that a meromorphic function f is holomorphic if and only if div(f ) ≥ 0;
therefore
L(0) = O(X) := {f : X → C | f holomorphic on X } .
In particular, if X is compact the only holomorphic functions on the whole X are the constants;
thus L(0) = O(X) ∼ = C.
Lemma 7.7. Let X be a compact Riemann surface. If D is a divisor on X with degree strictly less
than zero, then L(D) = {0}.
Proof. Let f ∈ L(D) be a nonzero function. The divisor
E := D + div(f )
is positive (E ≥ 0), by definition of L(D). Therefore deg(E) ≥ 0 and we conclude that there is a
contradiction by taking the degree of the defining formula of E:
deg(E) > 0 > deg(D) = deg(E).

57
Proposition 7.8. Let X = P1 (C) be the complex projective space and let D ∈ Div(X) be a positive
divisor such that deg(D) = d. Then

L(D) = H 0 P1 , OP1 [D] ∼

= {homogeneous polynomials in z0 , z1 of degree d} .
Proof [2]. Let us write the divisor as
n
X
D= ei · λ i + e∞ · ∞
i=1

with λi ∈ C distinct, such that e1 + · · · + en + e∞ = d ≥ 0, and let us consider the function

n
Y
fD (z) = (z − λi )−ei .
i=1

With the above notation, it turns out that the thesis is equivalent to proving that

L(D) = {fD (z) · g(z) | g(z) is a polynomial of degree at most deg(D) } .

Step 1. Fix a polynomial g(z) of degree e and notice that ∞ is a pole of g, whose degree is equal
to e. The divisor of fD (z) is exactly
n n
!
X X
−ei · λi + −ei · ∞,
i=1 i=1

therefore
div (fD (z) · g(z)) + D = div(g) + div(fD ) + D ≥
!
X
≥ ei + e∞ − e · ∞ = (deg(D) − e) · ∞,
i

which proves that e ≤ deg(D). This proves that the given space is a subspace of L(D).

h
Step 2. Vice versa, let us take any nonzero h(z) ∈ L(D) and let us set g := fD . We have

div (g) = div(h) − div(fD ) ≥ −D − div(fD ) ≥

!
X
≥ − ei − e∞ · ∞ = −deg(D) · ∞,
i

which shows that g can have no poles in the finite part C, and can have a pole of order at most
deg(D) at ∞. This forces g to be a polynomial of degree at most deg(D).

7.3 Linear Systems of Divisors

Linear Equivalence. In this paragraph, we introduce the notion of equivalence between divisors
(on a compact Riemann surface) and prove that it is an equivalence relation in the space Div(X).
Definition 7.9 (Linear Equivalence). Let X be a (compact) Riemann surface and let D1 , D2 ∈
Div(X). The divisors are said to be linearly equivalent if there exists a meromorphic function
f : X → C such that
div(f ) = D1 − D2 .

58
Notation. If D1 and D2 are equivalent divisors, we shall write D1 ∼ D2 (or D1 ≡ D2 ).
Proposition 7.10. Let X be a compact Riemann surface.
(1) ∼ is an equivalence relation in Div(X).
(2) D ∼ 0 if and only if there exists a meromorphic function f : X → C such that D = div(f ).

(3) If D1 ∼ D2 , then deg(D1 ) = deg(D2 ).

Remark 7.1. If X = P1 (C), then

OP1 [D] = OP1 [deg(D)]

immediately implies that D1 ∼ D2 if and only if deg(D1 ) = deg(D2 ).

Linear Systems. We are finally ready to introduce the notion of linear system associated with a
divisor D ∈ Div(X).
Definition 7.11 (Complete Linear System). Let X be a Riemann surface and let D ∈ Div(X) be
any divisor. The complete linear system of D, denoted by |D|, is the set of all nonnegative divisors
E ≥ 0 which are equivalent to D, i.e.

|D| = {E ∈ Div(X) | E ∼ D and E ≥ 0 } .

There is a geometric/algebraic structure to a complete linear system |D| which is related to the
vector space L(D). Let P(L(D)) be the projective space associated to the vector space L(D); we
may define a function
S : P (L(D)) → |D|
by sending the span of a function f ∈ L(D) to the divisor div(f ) + D. Note that this map is well
defined, since the divisor of a multiple λ · f is equal to the divisor of f .

Lemma 7.12. If X is a compact Riemann surface, the map S defined above is a 1-1 correspondence.
Proof. Suppose that there are functions f, g : X → C such that S(f ) = S(g). If we cancel the D’s,
it turns out that div(f ) = div(g) or, equivalently, that
 
f
div = 0.
g

The function f /g has no zeros or poles on X, thus (by compactness of X) it must be a identically
equal to a nonzero constant λ ∈ C, i.e., they are the same element in the domain of S.
Let E ∈ |D| be any divisor. By definition E ∼ D and E ≥ 0, therefore there exists f ∈ L(D)
such that
E = div(f ) + D,
which is equivalent to S(f ) = E, i.e., S is surjective.
Thus for a compact Riemann surface, complete linear systems have a natural projective space
structure.

59
Example 7.3. Let X = P1 (C), let p = [0 : 1] ∈ X and set D := 1 · p. We have already proved that
 
az + b
H 0 P1 , OP1 [D] ∼


=
z
in U0 = {z0 6= 0} with local coordinate z = z1 /z0 . The linear system |D| is given by the set of
positive divisors E, such that
E − D = div(f ).
The meromorphic function f has a unique zero of order 1, and hence

div(f ) = q − p q ∈ P1 (C),

implies that
|D| = E = 1 · q  q ∈ P1 (C) .
 

Proposition 7.13. Let X be a compact Riemann surface and let D ∈ Div(X). Then

deg(D) < 0 ⇐⇒ H 0 (X, OX [D]) = 0 ⇐⇒ |D| = ∅.

Proof. The former equivalence is easy and it has already been proved above. For any E ∈ |D|, it
turns out that
( (
E≥0 E≥0
=⇒ =⇒ absurd, E is positive.
E = D + div(f ) deg(E) = deg(D) + 0 < 0

7.4 Divisors and Maps to Projective Space

In this section, we shall be concerned with the possibility to embed a Riemann surface into a
projective space holomorphically.

Holomorphic Maps to Projective Space. The first step is to understand what is the meaning
of a ”holomorphic map to the complex projective space Pn ”.
Definition 7.14 (Holomorphic Map). Let X be a Riemann surface. A map ϕ : X → Pn is
holomorphic at the point p ∈ X if there are a neighborhood Up of p and holomorphic functions
σ0 , . . . , σn : Up → C, not all zero at p, such that - locally - ϕ has the form

ϕ(x) = [σ0 (x) : . . . : σn (x)] .

Observe that, if one of the σi ’s is nonzero at p, then it will be nonzero in a neighborhood of p;

thus the map given by the σi ’s is be well defined - at least locally.

Maps to Projective Space as Meromorphic Functions. On a compact Riemann surface, the

holomorphic maps are constant, and thus one cannot expect to use the same holomorphic function
σi at all points p ∈ X to define a holomorphic map.
Let X be a Riemann surface. Choose n + 1 meromorphic functions σ = (σ0 , . . . , σn ) on X, not
all identically to zero. Define ϕσ : X → Pn by setting

ϕσ (p) = [σ0 (p) : . . . : σn (p)] . (7.1)

A priori ϕσ is defined at p if

60
 (1) p is not a pole of any σi , and
(2) p is not a zero of every σi .
The reader may check by herself that ϕσ is holomorphic at all points p satisfying both condition (1)
and (2) (i.e., it is holomorphic at every definition point).
On the other hand, the function may also be defined on points which violate the first condition,
as a consequence of the next result.
Lemma 7.15. If the meromorphic functions σ0 , . . . , σn are not all identically zero at p, then the
map (7.1) given above can be extended to a holomorphic map defined at p.
Proof. Set
m := min ordp σi .
i=0, ..., n
By definition there is a neighborhood Up of p such that

σi  Up

has not other pole inside U , that is, using the local coordinate z it turns out that
gi (z) = z −m σi (z)
is a holomorphic map at all point of Up , for each i = 0, . . . , n. Therefore, it suffices to define the
function ϕσ at the point p as follows:
ϕσ (p) := [g0 (p) : . . . : gn (p)] = z −m [σ0 (p) : . . . : σn (p)] .

Example 7.4. Let X = P1 (C), p = [1 : 0] ∈ X, and let D = 2 · p ∈ Div(X). In the previous section
we have proved that the 0-th cohomology group may be identified as follows:
  
0 1

1 
H P , OP1 [D] = P (z)  deg(P ) ≤ 2 .

z2
It is easy to prove that in homogeneous coordinates we have the identity
c0 z02 + c1 z0 z1 + c2 z12
 
H 0 P1 , OP1 [D] =


,
z12
and hence a basis of the 0-th cohomology group is given by
 2  2
z0 z0 z1 z1
σ0 = , σ1 = 2 and σ2 = .
z1 z1 z1
As a consequence, Lemma 7.15 allows us to write the ϕσ : P1 → P2 as
[z0 : z1 ] 7−→ z02 : z0 z1 : z12 ,
 

which is the so-called Veronese embedding (see Figure 7.4). Clearly the image of this map inside P2
is given by
[y0 : y1 : y2 ] ∈ P2 (C)  y12 = y0 y2 ,
 

and hence   
  E = div(f ) + D, E ≥ 0 such that 
  
|D| = E ∈ Div(X) f has a pole of order 2 in p and .


  
  f has two zeros of order 1 (q1 and q2 ) 
More precisely, in P2 the divisor E = q1 + q2 corresponds to the intersection of the line ` with the
image of the map ϕσ .

61
 Figure 7.1: Veronese Embedding

Formally, it turns out that, if ϕ : X → Pn is holomorphic and H is a hyperplane of Pn , then the

intersection H ∩ ϕ(X) may be regarded as a divisor of X.
Definition 7.16 (Hyperplane Divisor). Let
( n
)
X
H = h := ai yi = 0 ⊂ Pn (C)
i=0
Pn
be a hyperplane. If q is a point of ϕ(X) ∩ H, and h0 = i=0 bi yi defines a hyperplane such that
h0 (q) 6= 0, then we define the pull-back of H as
 
 
X X h
ϕ∗ (H) =  ordp ◦ ϕ · p .
−1
h0
q∈H∩ϕ(X) p∈ϕ (q)

Remark 7.2. In the previous definition, we divide h by h0 since h is a holomorphic map and, as
we have already proved earlier, it would be constant on any compact Riemann surface.

Correspondence Linear Subsystems-Subspaces. In this paragraph, we want to prove that

the only restriction for ϕ|D| not being holomorphic, is that there exists a point p ∈ X such that
p ∈ spt E for every divisor E ∈ |D|.
Definition 7.17 (Base Point Free). Let V ⊂ |D| be a linear system on a Riemann surface X. A
point p ∈ X is a base point of V if and only if every divisor E ∈ V contains p, that is,

E ≥ p.

A linear system is base point free - b.p.f. from now on - if there are no base points.
Remark 7.3. A point p ∈ X is a base point for a linear system V ⊆ |D| if and only if

f (p) = 0, ∀ f ∈ V ⊆ H 0 (X, OX [D]) .

Equivalently, if σ0 , . . . , σn is a basis for V , then p is a base point if and only if

σi (p) = 0, ∀ i = 0, . . . , n.

62
 Let V ⊂ H 0 (X, OX [D]) be a linear subsystem. Suppose that V is b.p.f, and suppose also that
V is a projective space of dimension n + 1. It induces a morphism

ϕ : X → Pn (C) = P(V v ) x 7−→ (σ0 (x), . . . , σn (x)) ,

where V v is the geometric dual of V (i.e. the vector space of the hyperplanes of V ), and the σi are
meromorphic functions not identically equal to zero.
Theorem 7.18. There is a 1-1 correspondence
n
 
 holomorphic maps ϕ : X → P (C)
( )
0 
b.p.f. linear systems V ⊆ H (X, OX [D]) 
↔ with non degenerate image, .
of projective dimension n + 1, and D a d-divisor
such that ϕ∗ (H) is a d-divisor

 


The holomorphic map needs to have a non-degenerate image, and this assumption cannot be
relaxed. Indeed, if the image ϕ(X) is contained in a hyperplane of Pn (C), then the correspondence
fails to be 1-1.
The reader may prove this fact by herself; e.g., consider a map X → P2 and look at it as immersed
in a higher-dimension space X → P2 (C) ,→ P3 (C).
Proof. Let V be a b.p.f system as in the assumptions. The associated holomorphic map is the one
we have already constructed above, i.e.,

ϕ : X → Pn (C) = P(V v ) x 7−→ (σ0 (x), . . . , σn (x)) .

Suppose that y0 , . . . , yn are the coordinates of Pn , and let p ∈ ϕ(X) ∩ H be any point,
( n
) ( n
)
X X
H := h := ai yi = 0 and H0 := h0 := bi yi = 0
i=0 i=0

be hyperplanes such that p ∈

/ H0 . By definition it turns out that
 
∗
X h
ϕ (H) = ordp ◦ ϕ · p,
h0
p∈X

and this divisor has clearly degree equal to d.

The opposite arrow is a direct consequence of Lemma 7.19 which is stated and proved right
below.
Lemma 7.19. Let ϕ : X → Pn (C) be a holomorphic map and assume that it is b.p.f., i.e., for any
p ∈ X there exists i such that σi (p) 6= 0. Let
X
D := − min (ordp (σi )) · p
i
p∈X
Pn
and let H := {h := i=0 ai yi = 0} be a hyperplane. Then
n
!
X
div ai σi + D = ϕ∗ (H).
i=0

63
Proof. Let p ∈ X, let j be the index such that ordp (D) = −ordp (σj ), and let h0 := yj . It easily
follows that
n n
!
h X σi X
◦ϕ= ai =⇒ ordp (ϕ∗ (H)) = ordp ai σi −ordp (σj ),
h0 i=0
σ j i=0 =ordp (D)

therefore
{ϕ∗ (H) | H projective hyperplane } = {div(f ) + D | f ∈ hσ0 , . . . , σn i } .

Example 7.5. Let ϕ : P1 → P2 be the Veronese embedding, i.e.,

 2 
z0 z0
(z0 , z1 ) 7−→ , ,1 ,
z12 z1

locally in the chart U0 . The divisor is given by

−D = −2 · [1, 0] =⇒ D = 2 · p, p := [1, 0],

and notice that ϕ(p) = p∞ ∈ ϕ(P1 ) ⊂ P2 .

Let H be the hyperplane defined by the equation h := a0 y0 + a1 y1 + a2 y2 = 0 and let H0 be
the hyperplane defined by the equation h0 := y0 = 0. Then the pullback is given by

ϕ∗ (H) = div a0 z02 + a1 z0 z1 + a2 z22 − div z02 + 2 · p = div a0 z02 + a1 z0 z1 + a2 z22 ,

coherently with the fact that the Veronese embedding is globally given by

(z0 , z1 ) 7−→ z02 , z0 z1 , z12 .

7.5 Inverse Image of Divisors

Let X and Y be compact Riemann surfaces and let F : X → Y be a holomorphic function. The
pullback via F of q ∈ Y is defined by
X
F ∗ (q) = moltp (F ) · p.
p∈F −1 (q)

Definition 7.20 (Divisor Pullback). Let D be a divisor on Y of the form

X
D= n(q) · q.
q∈Y

The pullback of D via F is a divisor on X, defined by the formula

X
F ∗ (D) := n(q) · F ∗ (q).
q∈Y

Proposition 7.21. Let X and Y be compact Riemann surfaces and let F : X → Y be a holomorphic
function.
(1) The pull-back F ∗ : Div(Y ) → Div(X) is a group homomorphism.

64
 (2) If g : Y → C is a meromorphic function, then

F ∗ (div(g)) = div (F ∗ g) = div (g ◦ F ) .

(3) The degree is multiplicative, i.e.,

deg (F ∗ (D)) = deg(F ) · deg(D).

(4) The pull-back commutes with the holomorphic function sheaf associated to a divisor, that is,

F ∗ (OY [D]) = OX [F ∗ (D)] .

Theorem 7.22. Let X be a compact Riemann surface, and let ϕ : X → Pn (C) be a holomorphic
map. Assume that Y = ϕ(X) ⊆ Pn is a smooth algebraic curve of degree e := deg(Y ). Then

deg (ϕ∗ (H)) = deg(Y ) · deg (ϕ : X → Y ) .

Remark 7.4. In general, the image Y = ϕ(X) is an algebraic curve (not necessarily smooth) - and
a Riemann surface -, whose degree is defined by the formula
X
deg(Y ) = deg(Y ∩ H) = ordq (H).
q∈Y ∩H

Proof. Let us consider a point p ∈ X, and let us consider the hyperplanes

( n
) ( n
)
X X
H := h := ai yi = 0 and H0 := h0 := bi y i = 0
i=0 i=0

such that ϕ(p) ∈ H ∩ Y and ϕ(p) ∈

/ H0 . By definition, it turns out that
   
∗ h h
ordp (ϕ (H)) = ordp ◦ ϕ = moltp (ϕ) · ordϕ(p) ,
h0 h0
therefore  
X h
deg (ϕ∗ (H)) = moltp (ϕ) · ordϕ(p) =
h0
p∈X
 
X X
=  moltp (ϕ) · ordq (H) =
q∈Y p∈ϕ−1 (q)
X
= deg (ϕ : X → Y ) · ordq (H) .
q∈Y
| {z }
=e

7.6 Canonical Divisor

The sheaf Ω1X consists of all the holomorphic 1-forms defined on X. Recall that, locally, a 1-form
can be identified to a holomorphic function, i.e.,

Ω1X 3 ω 7−→ ω = f (z) dz.

65
In particular, the reader may check by herself that Ω1X is an invertible sheaf. Let U = {Ui }i∈I be a
∼
covering of X, and let ϕi : Ui −
→ Vi ⊂ C be a collection of charts such that
∼
→ Vi ∼

Ui − = ∆, ω Ui 7−→ fi (z) dz.

The transition maps are denoted, as usual, by ϕi, j := ϕi ◦ ϕ−1

j . It follows that, if the intersection
Ui ∩ Uj is nonempty, the 1-forms can be glued together by

ωj = (ωi ◦ ϕi, j ) · ϕ0i, j . (7.2)

Proposition 7.23. Let ω1 , ω2 ∈ Ω1X . There exists a unique meromorphic function g : X → C such
that ω2 = g ω1 . In particular, locally

ω2 = f2 (z) dz = g(z) f1 (z) dz,

where ωi = fi (z) dz.

Proof. Let U = {Ui }i∈I be a (connected) covering of X. For each i it turns out that
(1) (2) ∼
ω1 = fi (z) dz and ω2 = fi (z) dz in Ui −
→ Vi .
(1) (2)
Since fi (z) and fi (z) are both holomorphic at every point of the open set Vi , the function
(2)
fi (z)
hi (z) := (1)
fi (z)

is meromorphic at every point of Vi . In the intersection Ui ∩ Uj , it follows from (7.2) that

 
(2)
fi ◦ ϕi, j · ϕ0i, j (z) f
(2)
hi (z) =   = i(1) ◦ ϕi, j (z),
(1)
fi ◦ ϕi, j · ϕ0i, j (z) fi

that is, the function

g(z) = hi ◦ ϕi (z) for x ∈ Vi and x = ϕ(z),
is a well-defined meromorphic function on the whole surface X.
Definition 7.24 (1-Forms Divisor). Let X be a Riemann surface and let U = {Ui }i∈I be a covering
of X. The canonical divisor associated to the holomorphic 1-form ω ∈ Ω1X , denoted by div(ω), is
locally defined by  X
div(ω) Ui = ordp (f ) · p,
p∈Ui

where f is the function such that ω = f (z) dz in Ui .

Proposition 7.25. Let X be a Riemann surface and let ω1 , ω2 ∈ Ω1X be holomorphic 1-forms.
There exists a meromorphic function g such that

div(ω1 ) = div(ω1 ) + div(g),

that is, ω1 ∼ ω2 .

Proof. A simple corollary of Proposition 7.23.

66
Definition 7.26 (Canonical Divisor). A divisor KX ∈ Div(X) is a canonical divisor if it is the
divisor of a holomorphic 1-form, that is,
∃ ω ∈ Ω1X : KX = div(ω).
0
Remark 7.5. The canonical divisor is not unique, but, for any KX = div(ω 0 ) and any KX
00
=
00 0 00
div(ω ), it turns out that KX ∼ KX .
Remark 7.6. There is an isomorphism OX [KX ] ∼
= Ω1X as invertible sheaves, since the co-cycles are
the same.
Example 7.6. Let us consider the sheaf Ω1P1 , and suppose that z is the local coordinate of U0 and
w is the local coordinate of U1 . It turns out that
1 1
w= =⇒ dw = − 2 dz in U0 ∩ U1 ,
z z
therefore  
1 1
f0 (z) dz = f1 (w) dw ⇐⇒ f0 (z) = − 2 f1 dz.
z z
In particular, the co-cycle is given by
 −2
1 z1
f0, 1 (z) = − 2 = − ,
z z0
and this immediately implies that Ω1P1 ∼
= OP1 (−2).

Example 7.7. Let X = CΛ be a Riemann surface of genus 1. Then ω = 1 · dz is a holomorphic

1-form, but it is zero at the quotient. In particular, KX = 0 and thus Ω1X ∼
= OX [KX ] = OX .

7.7 Riemann-Hurwitz Theorem

In this final section, we want to state and prove the Riemann-Hurwitz theorem, which links the
canonical divisor of two compact connected Riemann surfaces via a morphism.
Theorem 7.27 (Riemann-Hurwitz). Let X and Y be compact connected Riemann surfaces, and let
π : X → Y be a morphism. Then
KX ∼ π ∗ (KY ) + R,
where R is the ramification divisor, and it is defined as
X
R= (ordp (π) − 1) · p.
p∈X

Proof. Recall that there is an isomorphism between invertible sheaves Ω1Y ∼

= OY [KY ] since the
co-cycles coincide.

Step 1. Let p ∈ X, and let q ∈ Y be a point in the image (i.e. π(p) = q). Let Wq be a neighborhood
of q in Y such that
ωY = f (w) dw,
where KY = div(ωY ). By Proposition 3.16, there exists a neighborhood Up of p in X such that
z 7−→ w = z m ,

π U : Up −→ Wq ⊆ Y,
p

where p ←→ 0 and q ←→ 0.

67
Step 2. By definition, we have that

π ∗ (f (w) dw) = f (z m ) · m z m−1 dz,

where, intuitively, the first term corresponds to the differential ωX , and the second term corresponds
to the ramification at p.
Therefore, by taking the divisors of both left-hand side and right-hand side, we have that
  
div ωX Up = div (f (z m )) = π ∗ (div(f (w))) + div z m−1 ,

and this concludes the proof, since

X
KX = div (ωX ) = π ∗ (div(ωY )) + (ordp (π) − 1) · p.
p∈X

Corollary 7.28. Let X be a connected Riemann surface of genus g, and let KX be a canonical
divisor of X. Then deg(KX ) = 2g − 2 = −χtop (X).
Proof. Let π : X → P1 be a morphism of degree d. Such a map always exists, but the result is
highly nontrivial1 . It follows from the Riemann-Hurwitz Theorem 7.27 that
X
KX = π ∗ (KP1 ) + R =⇒ deg(KX ) = d · deg(KP1 ) + (ordp (π) − 1) . (7.3)
| {z }
=−2d p∈X

On the other hand, the Hurwitz Theorem 3.23 gives us the identity
X
χtop (X) = d · χtop (P1 ) − (ordp (π) − 1) , (7.4)
p∈X

hence, if we combine (7.3) and (7.4) together, then we can conclude that

deg(KX ) = 2g − 2 = −χtop (X).

Remark 7.7. As a consequence of the previous result, it turns out that there are three big families
of algebraic curves, i.e.,
g=0 deg(KX ) < 0
g=1 deg(KX ) = 0
g≥2 deg(KX ) > 0.

1 In Section 9.3 we have proved that there is a meromorphic function f : X −→ C, and we know that this can be

identified with a holomorphic map F : X −→ P1 (C).

68
Chapter 8

The Riemann-Roch Theorem and

Serre Duality

Let X be a Riemann surface, and let be given a divisor D ∈ Div(X). The primary goals of Chapter
8 and of Chapter 9 are the following:
1) Find an isomorphism for the 0th cohomology group H 0 (X, OX [D]), or, at least, an estimate
of the dimension h0 (X, OX [D]).
2) Study the map ϕ|D| . More precisely, we would like to know if |D| is a b.p.f. linear system (i.e.,
if ϕ|D| is a morphism) and, in that case, if ϕ|D| is injective (or, even better, an embedding).

8.1 Rough Estimate of h0 (X, OX [D])

Remark 8.1. Let X be a compact Riemann surface, and let p ∈ X. There exists a short exact
sequence of sheaves, given by
0→− Ip →
− OX →− Cp → − 0,
where Ip is the sheaf of the ideals of the function vanishing at p, OX is the holomorphic function
sheaf and Cp is the skyscraper sheaf.
Proposition 8.1. Let X be a compact Riemann surface, and let p ∈ X. There is an isomorphism
of sheaves
Ip ∼
= OX [−p].
Proof. Notice that, locally, p is the divisor of a function ϕ : U → ∆ ⊂ C that sends p to 0. The
reader may fill in the details of the proof as an exercise, following one of the two possibilities below:

(1) Prove that OX [−p] is locally generated by ϕ, as a consequence of the fact that the sheaf OX [p]
is (locally) generated by the function ϕ1 .

(2) Prove that for any subset U ⊆ X

f ∈ OX [−p](U ) ⇐⇒ f = ϕ · g,

where g is an holomorphic function defined on U .

69
Proposition 8.2. Let D ∈ Div(X) be a divisor, and let p ∈ X be a point. Then there is a short
exact sequence of sheaf maps

0→
− OX [D − p] →
− OX [D] →
− Cp →
− 0.

Proof. It suffices to prove the exactness in each position.

Left Exactness. There is a natural inclusion

OX [D − p] ,→ OX [D],

as a consequence of the fact that

div f + D − p ≥ 0 =⇒ div f + (D − p) + p ≥ 0.
| {z }
=D

Middle/Right Exactness. Let Uy ⊆ X be an open neighborhood of a point y 6= p ∈ X, and

assume that p ∈
/ Uy . In this case it is straightforward to prove that

OX [D − p](Uy ) ∼
= OX [D](Uy )

is an isomorphism, coherently with the fact that

Cp (Uy ) = 0.

Let Up ⊆ X be an open neighborhood of p. If we set m := ordp D0 , then it turns out that there is a
meromorphic function
f (z) = z −(m+1) h(z) locally in Up ,
for some h non-vanishing at p, such that
 
div f Up + (D − p) Up = −p.

It follows that the function f generates the cokernel, and thus

OX [D] ∼ m+1 ∼
}=C∼
OX [D − p = C · {z = Cp ,

since C · {z m+1 } is supported at p.

Remark 8.2. More generally, there is a functor

F : (Div(X), +) −→ ({invertible sheaves}, ⊗) ,

defined by

D 7→ OX [D] and D1 + D2 7→ OX [D1 ] ⊗ OX [D2 ] ∼

= OX [D1 + D2 ].

It is not hard to prove that this functor is exact, since

(0 →
− OX [−p] →
− OX →
− Cp →
− 0) ⊗OX OX [D]

is isomorphic to
0→
− OX [D − p] →
− OX [D] →
− Cp →
− 0.
This functor is analyzed more in-depth in Section 10.1 using the language of the Picard group.

70
Remark 8.3. Recall that

H 0 (X, OX [D]) = 0,
deg D < 0 =⇒
|D| = ∅.

Proposition 8.3. Let X be a compact connected Riemann surface, and let D be a divisor of positive
degree d ≥ 0. Then the following estimate holds true:

h0 (X, OX [D]) ≤ d + 1. (8.1)

Proof. We first distinguish between divisor with empty linear system and divisor with a nontrivial
linear system, and then we proceed by induction on the degree of D.

Case |D| = ∅. If |D| = ∅, then deg D = 0 (since it is positive by assumption) and thus we infer by
the previous remark that
h0 (X, OX [D]) = 0.

Case |D| = 6 ∅, Base Step. Let D be a divisor of degree 0, and let E ∈ |D| be an effective divisor
linearly equivalent to D.
Clearly E is the null divisor 0 (the coefficients are positive, and they sum to zero;) thus |D| = {0}
and the dimension satisfies the estimate (8.1) as expected:

h0 (X, OX [D]) = 1.

Case |D| =6 ∅, Inductive Step. Let D be a divisor of degree d, let E ∈ |D| be an effective divisor,
and take p ∈ spt(E). By Proposition 8.2 there is a short exact sequence

0→
− OX [E − p] →
− OX [E] →
− Cp →
− 0,

which induces a long sequence in cohomology (see Theorem 6.13):

− H 0 (X, OX [E − p]) →
0→ − H 0 (X, OX [E]) →
− H 0 (X, Cp ) →
− ...

The map H 0 (X, OX [E]) →− H 0 (X, Cp ) is not necessarily surjective, therefore we can only infer an
inequality between the dimensions, i.e.,

h0 (X, OX [E]) ≤ h0 (X, OX [E − p]) + h0 (X, Cp ) .

In conclusion, recall that there is an isomorphism H 0 (X, Cp ) ∼

= C, and apply the inductive hypoth-
esis to obtain the sought inequality:

h0 (X, OX [E]) ≤ (d − 1) + 1 + 1 = d + 1.

Skyscraper Sheaf. Let X be a Riemann surface, let p ∈ X be a point and take any natural
number n ∈ N. Take D ∈ Div(X) and set ∆ := n · p; by Proposition 8.2 there is a short exact
sequence
0→− OX [D − ∆] →
− OX [D] → − O∆ →− 0.

71
The sheaf O∆ is the skyscraper sheaf, supported in {p}, and such that

0
 if q 6= p
(O∆ )q =
OX, pMn ∼
 = Cn if q = p,
X, p

where OX, p is the stalk of the holomorphic function sheaf at p, and MX, p is the maximal ideal at
p. More precisely, in the local coordinate we have p = 0 and div(z) = p, which in turn implies that

OX, p n = a0 + · · · + an−1 xn−1 .

(x )

Remark 8.4. The above argument can be easily generalized to a linear combination of points.
(a) The isomorphism H 0 (X, O∆ ) ∼
= Cn proves that

h0 (X, O∆ ) = n.

(b) By Proposition 6.11 it turns out that

H 1 (X, O∆ ) = 0,

since the dimension of the support is zero - i.e., strictly less than the index of the cohomology
group.
(c) If ∆ = n1 · p1 + · · · + nk · pk , then O∆ is a skyscraper sheaf supported in {p1 , . . . , pk }. In a
similar fashion, the reader may prove that
k
X
h0 (X, O∆ ) = ni and h1 (X, O∆ ) = 0.
i=1

8.2 Riemann-Roch Formula

The major result we want to achieve in this section is the Riemann-Roch formula for compact
connected Riemann surfaces.
Remark 8.5. Recall that the dimension of X over C is, by definition, equal to 1. Therefore Betti’s
numbers are all zero except for h0 and h1 , that is,

hj (X, OX [D]) = 0 ∀ j ≥ 2, ∀ D ∈ Div(X).

Definition 8.4 (Holomorphic Euler Characteristic). The Euler (holomorphic) characteristic of a

surface X with respect to a divisor D ∈ Div(X) is defined by

χ (X, OX [D]) = h0 (X, OX [D]) − h1 (X, OX [D]) .

Definition 8.5 (Arithmetic Genus). The arithmetic genus of a Riemann surface X is defined by

pa (X) := 1 − χ (X, OX ) .

Remark 8.6. If X is a compact connected Riemann surface, then h0 (X, OX ) = 1 and hence

pa (X) = h1 (X, OX ) .

72
Theorem 8.6 (Riemann-Roch). Let X be a connected compact Riemann surface, and let D ∈
Div(X) be any divisor. Then it turns out that

χ (X, OX [D]) = deg D + χ (X, OX ) , (8.2)

or, equivalently, that

h0 (X, OX [D]) − h1 (X, OX [D]) = deg D + 1 − pa (X). (8.3)

Proof. We first assume that the divisor D is effective and we derive (8.2) by induction on the degree
of d = deg D; only then we solve the general case.

Effective Divisor, Base Step. If deg D = 0, then D = 0 (since the coefficients are positive and
their sum is equal to zero). It follows that

OX [D] ∼
= OX =⇒ (8.2).

Effective Divisor, Inductive Step. Suppose that deg D = d > 0, and let p ∈ spt(D) be a point.
By Proposition 8.2 there is a short exact sequence

0→
− OX [D − p] →
− OX [D] →
− Cp →
− 0,

which induces an identity on the Euler characteristics, that is,

χ (X, OX [D]) = χ (X, OX [D − p]) + χ (X, Cp ) .

As we have already observed, the Euler characteristic of Cp is given by the difference between
h0 (X, Cp ) and h1 (X, Cp ); since the dimension of the support is less than 1, it turns out that

χ (X, Cp ) = h0 (X, Cp ) − h1 (X, Cp ) = 1 − 0 = 1.

Using the induction hypothesis, we immediately obtain the thesis for an effective divisor:

χ (X, OX [D]) = χ (X, OX ) + deg(D) − deg(p) + 1 = χ (X, OX ) + deg(D).

General Divisor. Let D be any divisor, and let

D = D+ − D−

be the decomposition of D in the positive part and the negative part (both of which are effective).
As usual, by Proposition 8.2 there is a short exact sequence

− OX [D+ − D− ] →
0→ − OX [D+ ] →
− CD − →
− 0,

which induces an identity of Euler characteristics, that is,

χ (X, OX [D]) = χ X, OX [D+ ] − χ (X, CD− ) =

= χ (X, OX ) + deg D+ − deg D− =

= χ (X, OX ) + deg D,

and this concludes the proof.

73
8.3 Serre Duality
The primary goal of this section is to use every tool we have introduced so far to prove the notorious
Serre Duality Theorem, which will come to handy to justify different results in the following sections.
Theorem 8.7 (Serre). Let X be a compact connected Riemann surface, let KX be a canonical
divisor and let D be any divisor on X. Then there is an isomorphism

H 1 (X, OX [D]) ∼
v
= H 0 (X, OX [KX − D]) ,
v
where denotes the dual vector space.

8.3.1 Mittag-Leffler Problem

Let X be a compact Riemann surfaces. Let p1 , . . . , ps ∈ X be given points, and suppose that for
any i = 1, . . . , s there is a polar polynomial, that is,
−1
in Upi ∼
X
hi (z) = ak z k , = ∆ neighborhood of pi with local coordinate z.
k=−ni

In this section, we investigate the Mittag-Leffler problem, that is, we want to determine if there
exists a function meromorphic on X such that:
(1) The function f : X → C is holomorphic outside of the finite set {p1 , . . . , ps }.
(2) The principal part of f in Upi is given by the polar polynomial hi .
A meromorphic function f : X → C satisfying these properties exists locally, but the problem is to
find one globally defined. The answer, as we shall be able to prove soon, depends on

H 0 (X, OX [D]) and H 1 (X, OX [D]) ,

where the divisor is simply defined by

s
X
D := ni · pi .
i=1

Laurent Tails. Let X be a Riemann surface, let p ∈ X be a point, and let Up 3 p be an open
neighborhood with coordinate zp . A Laurent tail with respect to p is a function of the form
kp
X
rp (zp ) = ai zpi , (8.4)
i=−np

where ai ∈ C are complex coefficients.

Definition 8.8 (Laurent Tail Divisor). A Laurent tail divisor on X is a finite formal sum
X
rp (zp ) · p,
p∈X

where rp (−) is a Laurent polynomial in the local coordinate zp , that is, a Laurent series of the form
(8.4) with a finite number of terms.

74
Notation. Let X be a Riemann surface. We denote by TX the set of all the Laurent tail divisors
defined on X.
Definition 8.9 (Laurent Tail Sheaf). Let
X
D= D(p) · p ∈ Div(X).
p∈X

The Laurent tail divisor sheaf associated to D is defined by setting

  
X  

U 7−→ TX [D](U ) := rp (−) · p  ∀ p ∈ U : kp < −D(p) , (8.5)
 
p∈X 

where kp is the maximal order of the function rp , as defined in (8.4).

The reader may check by herself that (8.5) actually defines a sheaf. For every divisor D ∈ Div(X),
there is a truncation map
tD : TX (U ) −→ TX [D](U ),
which is defined by X X
rp (−) · p 7−→ tD (rp )(−) · p,
p∈X p∈X

where
−D(p)−1
X
tD (rp )(zp ) = ai zpi .
i=−np

Meromorphic Field. Let us consider the field

M := {field of meromorphic function on X} .

The constant presheaf may be also defined by setting

MX (U ) := {f : U → M | f continuous and M has the discrete topology } ,

in such a way that

U connected =⇒ MX (U ) ∼
= M,
and the restriction maps are the identity maps. If we denote by MX the associated sheaf, then one
can prove that
(a) H 0 (X, MX ) ∼
= M, and
(b) H 1 (X, MX ) = 0.
In particular, for every divisor D ∈ Div(X) there exists a homomorphism of sheaves

αD : MX −→ TX [D],

which can be easily defined locally as

−D(p)−1
X X
p ∈ Up =⇒ f (zp ) = ai zpi 7−→ rp (zp ) = ai z i .
i≥−np i=−np

75
By definition, the kernel of αD is isomorphic to OX [D]; hence there is a short exact sequence of
sheaf maps
α
0→− OX [D] →− MX −−D → TX [D] →
− 0
inducing a long exact sequence in cohomology, that is,
α
− H 0 (X, OX [D]) →
0→ − H 0 (X, MX ) −−D
→ H 0 (X, TX [D]) →
− H 1 (OX [D]) →
− 0.

By Remark 6.3 we can infer that

H 0 (X, MX ) ∼
= M(X) and H 0 (X, TX [D]) ∼
= TX [D](X),

and thus it follows that

L(D) := H 0 (X, OX [D]) ∼

= ker(αD ) and H 1 (X, OX [D]) ∼
= coker(αD ).

By definition, there is an isomorphism

OX [Kx − D] ∼
= Ω1X [−D],

from which it follows that

H 0 (X, OX [Kx − D]) ∼
= H 0 X, Ω1X [−D] =

= {ω = f (z) dz | f meromorphic and ordp (f ) ≥ D(p) } .

8.3.2 Proof of Serre Duality Theorem

Road Map. In this section, we finally demonstrate the Serre Theorem 8.7 based on what we have
proved so far. The road map of the proof is the following:
(1) There exists a pairing

Res(·, ·) : H 0 X, Ω1X [−D] × TX [D](X) → C.

(2) The map defined above pass to the quotient. More precisely, it turns out that

Res(·, T ) ≡ 0 ∀ T ∈ Im(αD ),

and hence there exists a pairing

Res(·, ·) : H 0 X, Ω1X [−D] × coker(αD ) → C.

(3) The pairing defined in the previous step is non degenerate.

Proof of Theorem 8.7. The argument is rather involved. Hence we divide it into many different
steps.

Step 1. Let

Res(·, ·) : H 0 X, Ω1X [−D] × TX [D](X) → C,

(ω, T ) 7→ Res(ω, T )

be the map defined by X

Res(ω, T ) := Resp (Tp ω).
p∈X

76
More precisely, if Up is an open neighborhood of a point p ∈ X with local coordinate zp , then it
turns out that
−D(p)−1
X X
i
ai zpi ,


ω= ci zp dzp and Tp =
i≥D(p) i=−sp

and thus X
Resp (Tp ω) = a−i−1 · ci
i≥D(p)

is exactly equal to the coefficient of zp−1 .

Step 2. In this step, the primary goal is to prove that the map defined above descends to the
quotient in the second variable, that is, to

TX [D](X)
Im(αD ).

Let f ∈ M be a meromorphic function, let p ∈ X be any point, and let zp be the associated local
coordinate; then
−D(p)−1
X X
i
f (zp ) = ai zp 7−→ αD (f )(zp ) = ai zpi .
i≥−np i≥−np

The residue at p is thus given by

X
Resp (f · ω) = a−i−1 · ci = Resp (αD (f ) · ω)
i≥D(p)

since the terms whose index is j ≥ −D(p) of αD (f ) do not give any contribution to the sum above.
The Residue Theorem 5.11 immediately implies that
X
Resp (f · ω) = 0 =⇒ Res (αD (f ) · ω) = 0,
p∈X

which is exactly what we wanted to prove.

Step 3. In this step, we want to prove that the functional

v
Res : H 0 (X, OX [KX − D]) → H 1 (X, OX [D]) , ω 7→ Res(ω, −) (8.6)

is an isomorphism (i.e., the pairing is non degenerate).

(a) Linear. The linearity of the map (8.6) follows easily from the properties of the residue.
P
(b) Injective. Let D = p∈X D(p) · p, and let ω be such that
X
Res (T, ω) = 0, ∀T = Tp · p.
p∈X

Let p ∈ X be a point, let zp be the local coordinate, and let k = ordp (ω) (in particular,
−1 − k < −D(p)). It follows that

zp−1−k · p ∈ TX [D](X),

77


and also that, if we set ω := i≥k ci zpi dzp , with the lowest coefficient ck different from 0,
P
then one can easily check that
X
Res(ω, z −1−k · p) = Resp (zp−1−k · ci zpi dzpi = ck ,
i≥k

which is not zero. This contradiction shows that Res(ω, −) cannot be the identically zero map
on H 1 (X, OX [D]), unless ω = 0.
(c) Surjective. Recall that

= TX [D](X)Im(αD ).
H 1 (X, OX [D]) ∼

Let us consider a functional Φ : TX [D](X) → C vanishing on the image of αD , that is, assume
that 
Φ Im(α ) ≡ 0.
D


We want to construct a differential form ω ∈ H 0 X, Ω1X [−D] such that

Φ(−) = Res(ω, −).

The proof of this property is a consequence of two technical lemmas; hence we interrupt the
argument for a few pages and resume it when we are ready to conclude.

Truncation Maps. Let D1 , D2 ∈ Div(X) be two divisors, and assume that D1 ≤ D2 . There
exists a truncation map
D1
tD2
: TX [D1 ](X) −→ TX [D2 ](X),
which is defined by
−D1 (p)−1 −D2 (p)−1
X X
ai z i 7−→ ai z i .
i≥−np i≥−np
0 0
Let D ∼ D be linearly equivalent divisors (i.e., D = D − div(f )). Let p ∈ X be a point, and let
rp ∈ TX [D](X) be the Laurent tail given by
−D(p)−1
X
rp (zp ) = ai zpi .
i≥−np

There is a unique integer h such that f (zp ) = zph (i.e., it is a map of order h at p, and zp is the local
coordinate at p), and thus
−D(p)−1
X
(f · rp ) (zp ) = ai z i+h and deg(f · rp ) < −D(p) + ordp f = −D0 (p).
i≥−np

We conclude that there exists an isomorphism

∼
µf : TX [D](X) −−−−−→ TX [D − div f ](X),

which is defined by X X
rp (−) · p 7−→ (f · rp ) (−) · p.
p∈X p∈X

78
To prove that µf is an actual isomorphism, it is enough to check that the map

µ f1 : TX [D − div(f )](X) −→ TX [D](X)

is the inverse.
Remark 8.7.
(a) It may be useful to rewrite the isomorphism as
∼
µf : TX [D + div f ](X) −−−−−→ TX [D](X).

(b) Let Φ : TX [D](X) −→ C be a linear functional. If Φ vanishes on Im αD , then the composition

Φ ◦ µf vanishes on the whole image of αD+div f .
Lemma 8.10 ([2]). Let Φ1 and Φ2 be two linear functionals defined on H 1 (X, OX [A]) for some
divisor A ∈ Div(X). There is a positive divisor C and nonzero meromorphic functions f1 , f2 ∈
H 0 (X, OX [C]) such that
A−C−div f1
Φ1 ◦ tA ◦ µf1 = Φ2 ◦ tA−C−div
A
f2
◦ µf2

as functionals on H 1 (X, OX [A − C]). In other words, the two maps on TX [A−C](X) in the diagram
t
TX [A − C − div f1 ](X) TX [A](X)
µf1
Φ1

TX [A − C](X) C
µf2 Φ2

t
TX [A − C − div f2 ](X) TX [A](X)

are equal for some C and some f1 , f2 ∈ H 0 (X, OX [C]) \ {0}.

Proof. We argue by contradiction. Suppose that no such divisor C and functions fi exist. Then for
every positive divisor C it turns out that the C-linear map

v
H 0 (X, OX [C]) × H 0 (X, OX [C]) → H 1 (X, OX [A − C])

defined by sending a pair (f1 , f2 ) to

A−C−div(f1 ) A−C−div f2
Φ 1 ◦ tA ◦ µf1 − Φ2 ◦ tA ◦ µf2

is injective. In particular, for every such C we must have

h1 (X, OX [A − C]) ≥ 2 · h0 (X, OX [C]) , (8.7)

and, as a consequence of the Riemann-Roch Theorem 8.6, we can also infer that

h1 (X, OX [A − C]) = h0 (X, OX [A − C]) + g(X) − 1 − deg(A − C). (8.8)

The divisor C is positive; hence

h0 (X, OX [A − C]) ≤ h0 (X, OX [A]) and deg(A − C) ≤ deg(A).

79
It follows from (8.7) and the Riemann-Roch Theorem 8.6 that

h1 (X, OX [A − C]) ≥ 2 · h0 (X, OX [C]) ≥ 2 [deg(C) + 1 − g(X)] = 2 deg(C) + K1 ,

where K1 is a constant, and it follows from (8.8) that

h1 (X, OX [A − C]) ≤ deg(C) + h0 (X, OX [A]) + g(X) − 1 − deg(A) = deg(C) + K2 ,

where K2 is another constant. These growth rate are clearly incompatible for deg(C) sufficiently
big, and this gives the sought contradiction.


Lemma 8.11 ([2]). Let D1 ∈ Div(X) be a divisor, and let ω ∈ H 0 X, Ω1X [−D1 ] be a differential
form. Suppose that there is another divisor D2 ≥ D1 such that the residue map

Res(ω, −) : TX [D1 ](X) → C

vanishes on the kernel  

Ker tD 1
D2 : T X [D 1 ](X) −→ T [D 2 ](X) .


Then ω belong to H 0 X, Ω1X [−D2 ] .


Proof. We argue by contradiction. If ω ∈ / H 0 X, Ω1X [−D2 ] , then there exists a point p ∈ X with
k = ordp (ω) < D2 (p). Let us consider the Laurent tail divisor

Z = zp−k−1 · p.

Then Z ∈ ker(tD
D2 ), but the residue map does not vanish; this contradiction proves the lemma.
1

Proof of Theorem 8.7, Part II. We are now ready to finish the proof of the Serre duality theorem.
(d) Surjective, Part II. Let Φ : H 1 (X, OX [D]) → C be a functional, which we consider as a
functional on TX [D](X), vanishing on αD (MX ).
Let ω be a holomorphic 1-form, and let K = div(ω) be a canonical divisor so that

ω ∈ H 0 (X, OX [K]) = H 0 X, Ω1X .

Let A ∈ Div(X) be a divisor such that A ≤ D and A ≤ K, so that ω ∈ H 0 X, Ω1X [−A] . Let
us set ΦA := Φ ◦ tA
D : TX [A](X) → C. By Lemma 8.10 it turns out that there exists a divisor
C ≥ 0 and f1 , f2 meromorphic functions such that

ΦA ◦ tA−C−div
A
f1
◦ µf1 = Res(ω, −) ◦ tA−C−div
A
f2
◦ µf2 . (8.9)
A−C−div f2
In the right-hand side of (8.9) we have the map Res(ω, −) ◦ tA , which is nothing else
than the residue map Res(ω, −) acting on TX [A − C − div f2 ](X); on the other hand, the
composition Res(ω, −) ◦ µf2 is exactly equal to

Res(f2 · ω, −) : TX [A − C](X) → C,

and hence the identity (8.9) becomes

ΦA ◦ tA−C−div
A
f1
◦ µf1 = Res(f2 · ω, −).

Composing with µ1/f1 it turns out that

 
A−C−div f1 f2
ΦA ◦ t A = Res · ω, −
f1

80
 as functionals on TX [A − C − div f1 ](X).


We observe that (f2 /f1 ) ω belongs to H 0 X, Ω1X [C + div f1 − A] , and also that
 
f2 
A−C−div f1

Res · ω, − ≡ 0 on Ker tA .
f1


By Lemma 8.11 we have that (f2 /f1 ) ω ∈ H 0 X, Ω1X [−A] , and hence
 
f2
Res · ω, − = ΦA .
f1

By definition, the map ΦA is the composition between Φ and tA D ; hence the residue map above
vanishes on the kernel of Ker(tA
D ), which, in turn, implies that
 
f2 f2
· ω ∈ H 0 X, Ω1X [−D] =⇒ Φ = Res · ω, − : H 1 (X, OX [D]) −→ C,


f1 f1
and this completes the proof of the theorem.

8.4 The Equality of the Three Genera

Corollary 8.12. Let X be a compact connected Riemann surface and let KX be a canonical divisor.
There is an isomorphism
H 1 (X, OX ) ∼
v
= H 0 (X, OX [KX ]) .
Definition 8.13 (Geometric Genus). The geometric genus of a Riemann surface X is defined by

pg (X) := h0 (X, OX [KX ]) .

Corollary 8.14. Let X be a compact connected Riemann surface. The three notions of genus are
equivalent, that is,
pa (X) = pg (X) = g(X).
Proof. The Serre Duality Theorem 8.7 immediately implies that

pa (X) = h1 (X, OX ) = h0 (X, OX [KX ]) = pg (X),

therefore it remains to prove that one of them also coincides with the number of holes g(X).
The Riemann-Hurwitz formula 7.27 asserts that deg(KX ) = 2 (g(X) − 1), while the Riemann-
Roch Theorem 8.6 asserts that

h0 (X, OX [D]) − h1 (X, OX [D]) = deg D + 1 − pa (X).

The Serre Duality Theorem 8.7 implies

2g − 1 = 2 pa (X) − h1 (X, OX [KX ]) ,

hence it suffices to prove that h1 (X, OX [KX ]) = 1. But this is, once again, a simple consequence
of the Serre duality:
h1 (X, OX [KX ]) = h0 (X, OX ) = 1.

81
8.5 Analytic Interpretation (Hodge)
Let Ω1X be the sheaf of holomorphic 1-form. There exists a short exact sequence of sheaves
f 7→df
0→ − OX −−−−→ Ω1X →
− C→ − 0,

where the middle map is locally defined as follows:

f (z) 7−→ df (z) := f 0 (z) dz.

The long exact sequence in cohomology is thus given by

− H 0 (X, C) →
− H 0 (X, OX ) →
− H 0 X, Ω1X →− H 1 (X, C) →


0→ − ...

− H 1 (X, OX ) →
− H 1 X, Ω1X →− H 2 (X, C) →


... → − 0.

Clearly H i (X, C) ∼
= H i (X, R) ⊗R C implies that

H 0 (X, C) ∼


 = C,


H 1 (X, C) ∼
= C2g ,


H (X, C) ∼

 2
= C,

while
∼ ∼ 0

0 0
H (X, OX ) = C =⇒ H (X, OX ) = H (X, C) ,




H 0 X, Ω1X = H 0 (X, OX [KX ]) ,


H X, Ω1X = H 1 (X, OX [KX ]) ∼ = H 0 (X, OX ) ∼
=C∼

 1

= H 2 (X, C) .
We infer that there is a short exact sequence

− H 0 X, Ω1X → − H 1 (X, C) →
− H 1 (X, OX ) →


0→ − 0,

and it induces an equality on the dimensions of the cohomology groups, i.e.,

pa (X) + pg (X) = 2 g.

82
Chapter 9

Applications of Riemann-Roch
Theorem

In this chapter, we fully exploit the Serre duality theorem and the Riemann-Roch theorem to show
major results about both high-degree and low-degree divisors.
Moreover, we prove that, if X is a compact Riemann surface and D ∈ Div(X) is a divisor such
that deg D ≥ 2 g(X) + 1, then the analytic manifold

ϕ|D| (X) := Y ⊆ Pn

is also an algebraic curve.

In the last sections, we investigate the canonical map, and we set the ground for the notorious
Clifford theorem.

9.1 Very Ample Divisors

Definition 9.1 (Very Ample Divisor). Let X be a holomorphic manifold. A divisor D ∈ Div(X) is
said very ample if the associated map
v

ϕ|D| : X → Pn (C) = P H 0 (X, OX [D])

is an embedding.
Remark 9.1. In particular, a divisor D ∈ Div(X) is very ample if and only if
1) the linear system |D| is b.p.f.;
2) the morphism ϕ|D| is injective;


3) the differential d ϕ|D| p is injective at all points p ∈ X.

Definition 9.2 (Ample). Let X be a holomorphic manifold. A divisor D ∈ Div(X) is ample if

there exists a natural number k ∈ N such that the divisor k · D is very ample.
Remark 9.2. Let D be a b.p.f. divisor. The morphism ϕ|D| : X → Pn (C) is defined by

p 7−→ (σ0 (p), . . . , σn (p)) ,

where σ0 , . . . , σn is a basis for H 0 (X, OX [D]).

83
 1) The morphism ϕ|D| is injective if and only if, for any p, q ∈ X, there is a global section
f ∈ H 0 (X, OX [D]) such that f (p) 6= f (q).
Equivalently, ϕ|D| is injective if and only if
(a) there is f ∈ H 0 (X, OX [D]) such that f (p) = 0 and f (q) 6= 0;
(b) there exists E ∈ |D| such that p ∈ spt(E) and q ∈ / spt(E).


2) The differential d ϕ|D| p sends the tangent space TP X to the tangent Tϕ|D| (P ) Pn , thus it is
injective if and only if there exists f ∈ H 0 (X, OX [D]) such that ordp (f − f (p)) = 1.
Theorem 9.3 (Very Ample - Numerical Criterion). Let X be a compact connected Riemann surface,
and let D ∈ Div(X) be a divisor.
(a) The linear system |D| is b.p.f. if and only if, for any p ∈ X, it turns out that

h0 (X, OX [D − p]) = h0 (X, OX [D]) − 1.

(b) The divisor D is very ample if and only if, for any p, q ∈ X (eventually p = q), it turns out
that
h0 (X, OX [D − p − q]) = h0 (X, OX [D]) − 2.

Proof.
(a) Let p ∈ X. By Proposition 8.2 there is a short exact sequence of sheaf maps

0→
− OX [D − p] →
− OX [D] →
− Cp →
− 0,

which induces a long exact sequence in cohomology (see Theorem 6.13):

f 7→f (p)
− H 0 (X, OX [D − p]) →
0→ − H 0 (X, OX [D]) −−−−−→ H 0 (X, Cp ) →
− ....

Observe that there exists a global section f ∈ H 0 (X, OX [D]) such that f (p) 6= 0 if and only
if H 0 (X, OX [D]) − Cp , which is equivalent to the identity
 
f (p)
H 0 (X, OX [D − p]) = Ker H 0 (X, OX [D]) −−−→ H 0 (X, Cp ) .

The dimension of H 0 (X, Cp ) is equal to 1, thus we can infer that

h0 (X, OX [D − p]) = h0 (X, OX [D]) − 1,

which is exactly what we wanted to prove.

(b) Assume that |D| is a b.p.f. linear system.

Step 1. Suppose that there are two points p, q ∈ X such that ϕ|D| (p) = ϕ|D| (q), that is,
ϕ|D| is not injective. It follows that, as vector spaces,

H 0 (X, OX [D − p]) = H 0 (X, OX [D − q]) ,

which in turn implies that

h0 (X, OX [D − p]) = h0 (X, OX [D − q]) = h0 (X, OX [D]) − 1.

84
On the other hand, the assumption allows us to infer that

H 0 (X, OX [D − p]) = H 0 (X, OX [D − p − q]) = H 0 (X, OX [D − q]) ,

as vector spaces, and hence

h0 (X, OX [D − p − q]) = h0 (X, OX [D]) − 1 6= h0 (X, OX [D]) − 2.

Vice versa, suppose that the formula does not hold true, i.e., there are p, q ∈ X such that

h0 (X, OX [D − p − q]) = h0 (X, OX [D]) − 1.

By Proposition 8.2 there is a short exact sequence

0→
− OX [D − p − q] →
− OX [D − p] →
− Cq →
− 0,

which induces a long exact sequence in cohomology (see Theorem 6.13):

− H 0 (X, OX [D − p − q]) →
0→ − H 0 (X, OX [D − p]) →
− H 0 (X, Cq ) →
− ....

The assumption on the dimension proves that

H 0 (X, OX [D − p − q]) ∼
= H 0 (X, OX [D − p]) ,

is an isomorphism, and hence

 
f 7→f (q)
H 0 (X, OX [D − p]) = Ker H 0 (X, OX [D − p]) −−−−−→ H 0 (X, Cq ) .

In particular, for any f ∈ H 0 (X, OX [D]) it follows that

f (p) = 0 f (q) = 0,

since
f (p) = 0 =⇒ f ∈ H 0 (X, OX [D − p]) =⇒ f (q) = 0.

Step 2. We now want to prove that d ϕ|D| p
is injective if and only if

h0 (X, OX [D − 2 · p]) = h0 (X, OX [D]) − 2.

First, we observe that

H 0 (X, OX [D])  MX, pM2

d ϕ|D| p
: Tp X → Tϕ|D| (p) Pn is injective ⇐⇒
X, p
⇐⇒

⇐⇒

∃ f ∈ H 0 (X, OX [D]) : ordp (f − f (p)) = 1.

If we set ∆ = 2 · p, then one can easily prove that


0 0
H (X, OX [D]) − H (X, Cp )


⇐⇒ H 0 (X, OX [D]) − H 0 (X, O∆ ) .
H 0 (X, OX [D]) − MX, pM2


X, p

85
 Indeed, locally p corresponds to z = 0 and ∆ corresponds to z 2 = 0; hence O∆ is isomorphic
to the quotient C[[z]]z 2 . Then we have that

H 0 (X, OX [D]) − H 0 (X, O∆ )

if and only if

− H 0 (X, OX [D − 2 · p]) →
0→ − H 0 (X, OX [D]) →
− H 0 (X, O∆ ) →
− 0

is a short exact sequence; thus

h0 (X, OX [D − 2 · p]) = h0 (X, OX [D]) − 2,

and this is exactly what we wanted to prove.

Theorem 9.4 (High-Degree Divisors). Let X be a compact connected Riemann surface of genus
g(X), and let D ∈ Div(X) be a divisor of degree d.

(a) If d ≥ 2 g(X) − 1, then

H 1 (X, OX [D]) = {0}.

(b) If d ≥ 2 g(X), then |D| is a b.p.f. linear system.

(c) If d ≥ 2 g(X) + 1, then D is very ample.
Proof.
(a) Let KX be a canonical divisor. By Corollary 7.28 it turns out that

deg (KX − D) = deg (KX ) − deg (D) < 0,

and hence the Serre Duality Theorem 8.7 implies that

H 1 (X, OX [D]) ∼
v
= H 0 (X, OX [KX − D]) = {0}.

(b) Let p ∈ X be any point. By Proposition 8.2 there is a short exact sequence of sheaf maps

0→
− OX [D − p] →
− OX [D] →
− Cp →
− 0,

inducing a long exact sequence in cohomology (see Theorem 6.13):

− H 0 (X, OX [D − p]) →
0→ − H 0 (X, OX [D]) →
− H 0 (X, Cp ) →
− ...

− H 1 (X, OX [D − p]) →
... → − H 1 (X, OX [D]) →
− 0.

By assumption, the degree of the divisor D − p is greater of equal than 2 g(X) − 1, and hence
by (a) it follows that

H 1 (X, OX [D]) = 0 and H 1 (X, OX [D − p]) = 0.

Therefore H 0 (X, OX [D])  H 0 (X, Cp ) is surjective, and this is enough to conclude that the
linear system |D| is b.p.f., as a consequence of the numerical criterion (i.e., Theorem 9.3).

86
 (c) Let p, q ∈ X be points, and let us set ∆ := p + q. There is a short exact sequence, deriving
from Proposition 8.2,

0→
− OX [D − p − q] →
− OX [D] →
− C∆ →
− 0,

which induces a long sequence in cohomology (see Theorem 6.13):

− H 0 (X, OX [D − p − q]) →
0→ − H 0 (X, OX [D]) →
− H 0 (X, O∆ ) →
− ...

− H 1 (X, OX [D − p − q]) →
... → − H 1 (X, OX [D]) →
− 0.

By assumption, the divisor D − p − q has degree greater or equal than 2 g(X) − 1, and thus it
follows from (a) that
H 1 (X, OX [D − p − q]) = 0.
Therefore H 0 (X, OX [D])  H 0 (X, O∆ ) is surjective and, since H 0 (X, O∆ ) has dimension
2, it follows that
h0 (X, OX [D]) − 2 = H 0 (X, OX [D − p − q]) .
In conclusion, the numerical criterion (Theorem 9.3) implies that D is a very ample divisor,
which is exactly what we wanted to prove.

Corollary 9.5. Let X be a compact connected Riemann surface. If D ∈ Div(X) is an effective

divisor, then D is ample.

Corollary 9.6. Let X be a compact connected Riemann surface.

(1) If g(X) = 0, then X ∼
= P1 (C).
(2) If g(X) = 1, then X ∼
= {F3 = 0} ⊆ P2 is a cubic plane curve.
Proof.

(1) Let p ∈ X be a point, and let us consider the divisor D := 1·p. Clearly deg D ≥ 2 g(X)+1 = 1,
thus D is very ample and h1 (X, OX [D]) = 0. By Riemann-Roch 8.6 it follows that

h0 (X, OX [D]) = deg D + 1 − g(X) = 2,

and hence ϕ|D| : X → P1 is an embedding of degree equal to 1; since X is compact and

connected, we infer that ϕ|D| is the sought isomorphism.
(2) Let p ∈ X be a point, and let us consider the divisor D := 3·p. Clearly deg D ≥ 2 g(X)+1 = 3,
thus D is very ample and h1 (X, OX [D]) = 3. By Riemann-Roch 8.6 it follows that

h0 (X, OX [D]) = deg D + 1 − g(X) = 3,

therefore ϕ|D| : X → P2 is an embedding, and D is a divisor relative to a section of an

hyperplane of dimension 3, that is, ϕ|D| is cubic (see Theorem 9.7).

87
9.2 Algebraic Curves and Riemann Surfaces
In this section, the primary goal is to prove that there exists a 1-1 correspondence of categories
( ) ( )
Riemann surfaces ∼ Smooth algebraic
←−−−→ .
compact and connected projective curves

Algebraic Curves. In this paragraph, the primary goal is to prove the following statement: If X
is a compact Riemann surface and D ∈ Div(X) is a divisor of degree deg D ≥ 2 g(X) + 1, then the
analytic manifold
ϕ|D| (X) := Y ⊆ PN (C)
is also an algebraic manifold. More precisely, we will sketch the proof of the following theorem:
Theorem 9.7. Let X be a compact connected Riemann surface, and let D ∈ Div(X) be a divisor
of degree deg D ≥ 2 g(X) + 1. Then

ϕ|D| (X) := Y ⊆ PN (C)

is an algebraic curve, that is,

Y = V(g1 , . . . , gr ),
where g1 , . . . , gr ∈ C[x0 , . . . , xN ] are homogeneous polynomials.
Remark 9.3. There is a more general theorem, proved by Wei-Liang Chow, which asserts that
Y ⊂ PN (C) analytic manifold is an algebraic manifold (see [1]).
Proof. The argument is rather involved. Hence we divide the proof into many steps, and we state
and prove everything we need in this environment.

Step 1. Let us consider the graded algebra

O
R(D) := H 0 (X, OX [n · D]) , (9.1)
n≥0

along with the maps

H 0 (X, OX [n · D]) ⊗ H 0 (X, OX [m · D]) −→ H 0 (X, OX [(n + m) · D])

defined by sending the tensor product s⊗t to s·t, as n and m range in the set of all natural numbers.
Example 9.1 (Projective Space). Let X = P1 (C) with coordinates [x0 , x1 ], and let D := [0 : 1] be
a divisor. We have already proved that

H 0 (X, OX [D]) = hx0 , x1 i ,

H 0 (X, OX [2 · D]) ∼
= x20 , x0 x1 , x21 ,



hence
H 0 (X, OX [D]) ⊗ H 0 (X, OX [2 · D]) → H 0 (X, OX [3 · D])
is a well-defined map, which sends p ⊗ q to p · q, where p and q are homogeneous polynomials of
degree respectively one and two (coherently with the definition of degree for polynomials).

88
Step 2. Now, we state a result concerning the graded algebra R(D) which will be essential in this
proof; the reader may refer to this paper.

Theorem 9.8 (Castelnuovo-Mumford). Let X be a compact Riemann surface, and let D ∈ Div(X)
be a divisor such that deg D ≥ 2 g(X) + 1. Then the graded algebra R(D), defined in (9.1), is
generated in degree one, that is, for every n ∈ N there is a surjective map
⊗n
H 0 (X, OX [D]) − H 0 (X, OX [n · D]) .

We observe that, for any n ∈ N, the left-hand side may be written as the direct sum between the
symmetric and the antisymmetric part, i.e.,
⊗n
H 0 (X, OX [D]) = Λn H 0 (X, OX [D]) ⊕ Symn H 0 (X, OX [D]) .

The product · is commutative; hence

Λn H 0 (X, OX [D]) 3 t ⊗ s − s ⊗ t 7−→ t · s − s · t = 0 ∈ H 0 (X, OX [n · D]) ,

which means that for every n ∈ N

⊗n
H 0 (X, OX [D])  H 0 (X, OX [n · D]) ⇐⇒ Symn H 0 (X, OX [D])  H 0 (X, OX [n · D]) .

Step 3. There is a natural identification


v 
H 0 (X, OX [D]) = hx0 , . . . , xN i ∼= CN +1 ,

from which it follows that for every n ∈ N there is an isomorphism

Symn H 0 (X, OX [D]) ∼


= C[x0 , . . . , xN ]n ,
where C[x0 , . . . , xN ]n is the set of all homogeneous polynomials in the variables x0 , . . . , xN of degree
n. If we set
Y := ϕ|D| (X) ⊆ PN (C),
where x0 , . . . , xN are the coordinates of PN (C), then the surjective map
O O
Symn H 0 (X, OX [D]) − H 0 (X, OX [n · D]) ,

n≥0 n≥0

which exists as a consequence of Theorem 9.8, induces a different surjective map - by composing
with the isomorphism above -, that is,
M
α : C[x0 , . . . , xN ] = C[x0 , . . . , xN ]n − R(D).
n≥0

Step 4. Let us consider the ideal

I := Ker(α),
and let us denote by Z the algebraic variety associated with I, that is, set
Z := V(I).
Every polynomial p ∈ I vanishes on Y , as a consequence of the fact that the support of the graded
algebra R(D) is entirely contained in Y . More precisely, the following inclusion holds
Y ⊆ V (I) = Z,
and therefore the thesis is equivalent to showing that the opposite containment also holds.

89
Remark 9.4 (Irreducibility).
(1) The graded algebra R(D) is an integral domain since X is a connected surface; hence X is an
irreducible surface.
On the other hand, the surface Y is the image via embedding of X, and it is thus irreducible
as well.

(2) The reader may check, as a simple exercise, that the ideal I is prime; consequently, the algebraic
variety Z is irreducible.
Remark 9.5. As a consequence of the previous Remark, it is enough to prove that the dimension
dimC (Z) is equal to 1 to infer that Y = Z.

Step 5. In this final step, we briefly introduce the concept of Hilbert polynomial, and we state a
major result - due to Hilbert and Serre - concerning the relation between the dimension of Z over
C and the behavior of the polynomial at infinity.
Definition 9.9 (Hilbert Polynomial). Let Z ⊆ PN (C) be an algebraic variety. The Hilbert polyno-
mial associated to Z is the polynomial such that

t  0 =⇒ pZ (t) = dimC S(Z)t ,

where S(Z)t := C[x0 , . . . , xN ]tI , that is, the t-degree part of S(Z).
Theorem 9.10 (Hilbert-Serre). Let Z ⊆ PN (C) be a 1-dimensional algebraic variety. For every t,
the Hilbert polynomial is given by
pZ (t) = a1 t + a0 .
In particular, it turns out that
deg(pZ ) = dimC (Z).

In the previous steps we have proved that there is a commutative diagram

C[x0 , . . . , xN ]t S(Z)t
'

'

Symt H 0 (X, OX [D])

H 0 (X, OX [t · D])

hence
pZ (t) = dimC S(Z)t = h0 (X, OX [t · D]) .
Finally, the Riemann-Roch Theorem 8.6 allows us to find the Hilbert polynomial, which is given by

h0 (X, OX [t · D]) = t · deg D + 1 − g(X) =⇒ dimC Z = 1,

and this concludes the proof of Theorem 9.7.

90
Equivalence of Categories. In this final paragraph, we state and prove three relevant results
which will allow us to demonstrate the equivalence theorem mentioned at the beginning.
Proposition 9.11. Let Y ⊆ PN (C) be an algebraic curve of degree d, and assume N ≥ 4. Then
there exists a point O ∈ PN (C) such that the canonical projection, centered at O, given by

πO : PN (C) → PN −1 (C)

has the additional property that

πO (Y ) ∼
= Y.
Proof. The projection centered at O has the property πO (Y ) ∼ = Y if and only if

(a) πO Y is injective, if and only if O ∈
/ {secant lines to Y };


(b) d πO Y is injective, if and only if O ∈ / {tangent lines to Y }.

Secant. The set of all the secant lines to Y is given by

sec(Y ) := {Spanhp, qi | p 6= q ∈ Y },

and hence there exists a morphism

Φ : P1 × (Y × Y \ ∆Y ) −→ sec(Y ),

where ∆Y := {(p, p) | p ∈ Y } is the diagonal of Y , which is defined by

([λ0 : λ1 ], p, q) 7−→ (λ0 p + λ1 q) .

Tangent. The set of all the tangent lines to Y may be locally identified by the isomorphism

tan(Y ) ∼
= {(p, [λ0 : λ1 ]) | p ∈ Y, [λ0 : λ1 ] ∈ P1 (C) },

which is defined by
λ0
Ψ : Y × P1 (C) → tan(Y ), (p, [λ0 : λ1 ]) 7−→ p + v · ,
λ1
where v is a tangent vector.

Dimensional Argument. The previous points allow us to infer that

dim (sec(Y )) ≤ 2 · dim(Y ) + 1 = 3,

dim (tan(Y )) ≤ dim(Y ) + 1 = 2,

which, in turn, imply the following estimate on the dimension:

dim (sec(Y ) ∪ tan(Y )) ≤ 3.

In conclusion, the assumption N ≥ 4 is sufficient to infer that such a point O - satisfying (a) and
(b) - must exist.

91
 Figure 9.1: Idea of Proposition 9.11

Corollary 9.12. Let X be a compact connected Riemann surface. There exists an isomorphism
∼
Φ : X −−−−−→ Y ⊆ P3 (C),
where Y is an algebraic curve.
Proof. Let D ∈ Div(X) be a divisor such that deg D ≥ 2 g(X) + 1. The morphism
ϕ|D| : X ,→ Y0 ⊆ PN (C)
is an embedding; if we compose it with a sequence π1 , . . . , πN −3 of adequate projections (whose
existence is a consequence of Proposition 9.11), then we obtain that
ϕ|D| πN −3 ◦···◦π1
X −−−→ Y0 −−−−−−−−→ Y ⊆ P3 (C)
is an isomorphism, since it is composition of isomorphisms.
Proposition 9.13. Let Y be an algebraic curve of degree d in P3 (C). There exists a point O ∈ P3 (C)
such that the projection centered at O
πO : P3 (C) → P2 (C)
has the property that πO (Y ) ⊆ P2 (C) is an algebraic curve of degree d, with a finite number of simple
knots.

Equivalence of Categories. We are now ready to state and prove the main result of the section.
Theorem 9.14. There is a 1-1 correspondence
( ) ( )
Riemann surfaces ∼ Smooth algebraic
←−−−→ .
compact and connected projective curves

92
 Figure 9.2: Idea of the proof of Proposition 9.13

Proof. Let X be a compact connected Riemann surface, and let p ∈ X be a point. The divisor

D = (2 g(X) + 1) · p ∈ Div(X)

induces an embedding
∼
→ Y ⊆ PN (C),
ϕ|D| : X −
and this is exactly what we wanted to prove.
Vice versa, let Y ⊆ PN (C) be an algebraic curve. A finite number of applications of Proposition
9.11 yields to a projection π
e such that

e(Y ) := Ye ⊆ P3 (C)
π

has the property that, for every p ∈ Ye , there is an affine neighborhood Up 3 p such that
 
∂ gi
Ye ∩ Up = V (g1 , g2 ) ∩ Up and rank = 2.
∂ xj

The conclusion follows immediately if one applies the maximal rank theorem1 .

Alternative Approach. Let π : PN (C) → P2 (C) be the projection, and let Ye be the algebraic
curve with a finite number of simple knots (see Proposition 9.13).
The construction of the Riemann surfaces follows from the blowup method introduced in Sub-
section 2.2.2, but it is quite a lot harder.
1 Maximal Rank Theorem: If F : M → N has maximal rank near a point p ∈ M , then there exist a neighborhood
∼ ∼
U of p and V of F (p), and there are diffeomorphisms u : Tp M −→ U and v : TF (p) N −→ V such that F (U ) ⊆ V and

dFp = v −1 ◦ F ◦ u.

93
9.2.1 Equivalence Theorem
In this subsection, we want to give a different proof of Theorem 9.14 that allows us, via a third
category, to be more precise about the category morphisms.
Theorem 9.15 (Chow). Let Y ⊆ PN (C) be an analytic manifold.
(1) Y is an algebraic variety, that is, Y = V(IY ).
(2) Any meromorphic function on Y is rational.
(3) Any holomorphic map f : Y → Y 0 is given by rational functions.
More precisely, if we let

C[Y ] = C[x0 , . . . , xN ]IY and C(Y ) = Frac (C[Y ]) ,

then the field of all the meromorphic functions f : Y → C, denoted by M(Y ), is isomorphic
to C(Y ).
Remark 9.6. If Y is an algebraic curve, i.e. dimC Y = 1, then the field

M(Y ) ∼
= C(Y )

has transcendental degree one over C.

Theorem 9.16. There is a 1-1 correspondence between the following three categories:
( ) ( )
Riemann surfaces Smooth algebraic
compact and connected projective curves

 
 Field of the form C(X) 
 
with transcendental degree .
 
one over C
 

Proof. We divide the argument into three steps.

Step 1.1. Let us consider the functor

 
 Field of the form C(X) 
( ) 
Riemann surfaces 
Φ: −→ with transcendental degree
compact and connected  
one over C
 

defined by
Φ(X) := M(X),
that is, it sends a compact connected Riemann surfaces to the field of meromorphic functions defined
on X, and also
Hom(X, Y ) 3 f 7−→ f ∗ ∈ Hom (M(Y ), M(X)) ,
where f is a surjective (i.e., nonconstant) morphism, and

f ∗ (h) := h ◦ f.

94
This functor is essentially surjective, as a consequence of the following result which we will not prove.

Theorem 9.17. Let M be a field with transcendental degree one over C. Then M is
isomorphic to the field of quotient of

C[x, y] ,
(F )

where F is an irreducible polynomial.

In particular, every M in the codomain induces an affine algebraic curve given by

e = V(F ) ⊆ C2 .
X

The Riemann surface X such that Φ(X) = M can be easily defined starting from X: e take the
2
projectivization in P (C), and then resolve the singularities (see Subsection 2.2.2).

Step 1.2. The functor Φ is fully faithful2 . Indeed, let us set X1 = V(F ) and X2 = V(G), and let
us consider the morphism
ϕ : M1 −→ M2 .
We may always consider for the maps
∼
αi : C(Xi ) −−−−−→ Mi , i = 1, 2,

sending the projections πx and πy respectively to fi and gi . Let

Ψ : C2 → C2 , Ψ := α2−1 ◦ ϕ(f2 ), α2−1 ◦ ϕ(g2 ) =: (R1 (x, y), T1 (x, y))

be a function such that the following diagram is commutative:

ϕ
M1 M2
α1 α2

Ψ∗
C(X1 ) C(X2 )

Therefore Ψ induces a map between Riemann surfaces

Ψ : X2 → X1 with Ψ∗ = α2−1 ◦ ϕ ◦ α1 ,

and, actually, it turns out that

0 = G(f2 , g2 ) =⇒ 0 = α2−1 ◦ ϕ ◦ G(f2 , g2 )

which, in turn, implies that

0 = G α2−1 ◦ ϕ(f2 ), α2−1 ◦ ϕ(g2 ) = G(R1 , T1 ) ∈ M2 ∼

= C(X2 ).
2 Definition. The map
∼
ΦX, Y : Hom(X, Y ) −−−−−→ Hom (Φ(X), Φ(Y ))
is an isomorphism for every X, Y objects.

95
Step 2. In this paragraph, we give the idea behind the correspondence
 
( )  Field of the form C(X) 
Smooth algebraic ∼
 
←−−−→ with transcendental degree .
projective curves  
one over C
 

Let X ⊆ PN (C) be a smooth algebraic curve; by Proposition 9.13 it turns out that it is birational
to an algebraic curve X 0 ⊆ P2 (C). On the other hand
∼ e
Y ⊆ P2 (C) −
→ Y ⊆ C2
is also a birational correspondence, and hence
∼ e
→ Y ⊆ C2
X−
is a birational map, as it is the composition of birational maps. We conclude by noticing that
   
C Ye ∼ = Frac C[x, y](F ) .

Step 3. The third equivalence was already proved in Theorem 9.14. The morphisms are easily
defined using the commutativity of the diagram.

9.3 Existence of Globally Defined Meromorphic Functions

Let X be a compact Riemann surface, let p ∈ X be a point, and let
D := n · p, n ≥ 2 g(X) − 1
be a simple divisor. By Proposition 8.2 there is a short exact sequence
0→
− OX →
− OX [D] →
− OD →
− 0,
which induces a long exact sequence in cohomology (see Theorem 6.13):
− H 0 (X, OX ) →
0→ − H 0 (X, OX [D]) →
− H 0 (X, OD ) →
− ...

− H 1 (X, OX ) →
... → − H 1 (X, OX [D]) →
− 0.
As a consequence of Theorem 9.4, it turns out that
n ≥ 2 g(X) − 1 =⇒ H 1 (OX [D]) = 0,
and hence
− H 0 (X, OX ) →
0→ − H 0 (X, OX [D]) →
− H 0 (X, OD ) →
− H 1 (X, OX ) →
− 0.
is an exact sequence. It follows that
h0 (OX [D]) = n − g(X) + 1 ≥ g(X),
thus the second term of the sequence above cannot be trivial, that is,
H 0 (OX [D]) 6= 0.
In particular, there exists a global meromorphic function f such that div∞ (f ) ≤ n · p, that is, f has
a pole of order at most n at p.

96
9.4 Low Degree Divisors
In this section, we denote by X a compact connected Riemann surface (since we require both of the
assumptions to hold true in most of the results we will be presenting).

Recall. We have proved that the following result holds true for divisors of (relatively) high degree:
Theorem 9.18 (High-Degree Divisors). Let X be a compact connected Riemann surface of genus
g(X), and let D ∈ Div(X) be a divisor of degree d.
(a) If d ≥ 2 g(X) − 1, then
H 1 (X, OX [D]) = {0}.

(b) If d ≥ 2 g(X), then |D| is a b.p.f. linear system.

(c) If f ≥ 2 g(X) + 1, then D is very ample.

In this section, we shall be mainly concerned with the properties of small order divisors (precisely:
divisors of order 0, 1 or 2.) We start the discussion with two simple remarks.
Remark 9.7. Let D ∈ Div(X) be a divisor of degree deg D = 0. Then

h0 (X, OX [D]) = 1 ⇐⇒ D ∼ 0,

or, equivalently,
h0 (X, OX [D]) = 0 ⇐⇒ D 6∼ 0.
Proof. We have proved in Lemma 7.12 that - assuming X is a compact Riemann surface - there is
a 1-1 correspondence
P H 0 (X, OX [D]) ∼


= |D|.
In particular, it turns out that

h0 (X, OX [D]) = 1 ⇐⇒ dim |D| = 0 ⇐⇒ |D| = {0},

that is, if and only if D ∼ 0. In a similar fashion, one could prove the equivalent formulation, i.e.,

h0 (X, OX [D]) = 0 ⇐⇒ dim |D| = −1 ⇐⇒ |D| = ∅.

Remark 9.8. Let D ∈ Div(X) be a divisor of degree 2 g(X) − 2. The Serre Duality Theorem 8.7,
together with the remark above, proves that

h1 (X, OX [D]) = 1 ⇐⇒ D ∼ KX ,

or, equivalently,
h1 (X, OX [D]) = 0 ⇐⇒ D 6∼ KX .

97
Low Degree Divisors. Recall that we have proved in Proposition 8.3 that there is a rough
estimate on the dimension of the 0th cohomology group if D is a divisor of positive degree:

h0 (X, OX [D]) ≤ deg D + 1. (9.2)

Proposition 9.19. Let X be a compact connected Riemann surface, and let D ∈ Div(X) be a
divisor of degree equal to 1. Then

h0 (X, OX [D]) = 2, X = P1 (C),

h0 (X, OX [D]) ≥ 2 ⇐⇒
∼
→ P1 (C) is the isomorphism.
and ϕ|D| : X −

Proof. The estimate (9.2) proves that

h0 (X, OX [D]) ≥ 2 ⇐⇒ h0 (X, OX [D]) = 2 .

deg D = 1 =⇒

The idea is to show that

ϕ|D| : X → P H 0 (X, OX [D])
v
∼
= P1 (C)
is a morphism (i.e., the linear divisor system |D| is b.p.f.) of degree 1. For any p ∈ X there is a
short exact sequence

− H 0 (X, OX [D − p]) →
0→ − H 0 (X, OX [D]) →
− H 0 (X, Cp )

from which it follows that


deg(D − p) = 0
 

 h0 (X, OX [D − p]) = 1
h0 (X, OX [D]) = 2 =⇒
H 0 (X, OX [D]) − Cp .

 
h0 (X, C ) = 1

p

By Theorem 9.3 it turns out that |D| is b.p.f., and the morphism ϕ|D| has degree equal to deg D = 13 ,
which is exactly what we wanted to prove.

Alternative Conclusion. For every p ∈ X

h0 (X, OX [D − p]) = 1 =⇒ D ∼ p,

and this, in turn, implies that

OX [D] ∼
= OX [p] ∀ p ∈ X.
More precisely, the points are all equivalent; since this is a property that characterizes P1 (C), we
infer that X is the complex projective line4 .
Definition 9.20 (Hyperelliptic). A compact connected Riemann surface X of genus g(X) ≥ 2 is
hyperelliptic if there exists a divisor D ∈ Div(X) such that

deg D = 2 and h0 (X, OX [D]) = 2.

3 The reader should pay attention that this is not always true; see [2, pp 164-165]. In this case it is true since P1 (C)

is a Riemann surface.
4 It is easy to prove that p, q ∈ P1 (C) are always equivalent, e.g. consider the function

p 0 z1 − p 1 z0
f (z) = .
q0 z1 − q1 z0

98
Remark 9.9. Equivalently, a Riemann surface X, satisfying the same assumptions as above, is
hyperelliptic if there exists a divisor D ∈ Div(X) such that:
(1) The linear divisor system |D| is b.p.f. (as a consequence of Proposition 9.19).
·2
→ P1 (C) has degree 2. In the remainder of the section, we shall
(2) The morphism ϕ|D| : X −
1
denote it by g2 .

Figure 9.3: The point w is called Weierstrass ramification point. By Riemann-Hurwitz there are
only 2 g(X) + 2.

99
Proposition 9.21. Let X be a hyperelliptic Riemann surface. The morphism g21 : X → P1 (C) is
unique.
Proof. We argue by contradiction.

·2
Step 1. Let D1 , D2 ∈ Div(X), and let ϕ|Di | : X −→ P1 (C) be the associated morphisms. By
assumption
deg(Di ) = 2 and h0 (X, OX [Di ]) = 2,
hence there are divisors in |Di | (which will still be denoted by Di ) such that

D1 = p + q and D2 = p + r.

Step 2. Let L := p + q + r be the minimal divisor containing both; we claim that

h0 (X, OX [L]) = 3.

Suppose that h0 (X, OX [L]) < 3. Then it is necessarily equal to 2, and we can easily derive a
contradiction looking at the exact sequences below:

− H 0 (X, OX [p + q]) →
0→ − H 0 (X, OX [L]) →
− H 0 (X, Cr ) ,

− H 0 (X, OX [p + r]) →
0→ − H 0 (X, OX [L]) →
− H 0 (X, Cq ) .

The middle terms are the same, hence there are exact sequences

− H 0 (X, OX [p + q]) →
0→ − H 0 (X, OX [L]) →
− H 0 (X, Cq ) ,

− H 0 (X, OX [p + r]) →
0→ − H 0 (X, OX [L]) →
− H 0 (X, Cr ) ,

and by the assumption on the dimension it turns out that

H 0 (X, OX [p + q]) = H 0 (X, OX [L]) = H 0 (X, OX [p + r])

which is absurd.

Step 3. Let s, t ∈ X be points, and let us set ∆ := s + t. There is an exact sequence

− H 0 (X, OX [L − s − t]) →
0→ − H 0 (X, OX [L]) →
− H 0 (X, C∆ ) ,

and we immediately observe that dim C∆ = 2, and the divisor L − s − t has degree equal to 1. Thus,
by Proposition 9.19, it turns out that

dim H 0 (X, OX [L − s − t]) ≤ 1,

which, in turn, implies that

H 0 (X, OX [L])  H 0 (X, C∆ ) .

100
Step 4. The numerical criterion (see Theorem 9.3) implies that the divisor L is very ample. We
have already proved that
deg(L) = h0 (X, OX [L]) = 3,
hence the morphism ϕ|L| : X ,→ P2 (C) is an embedding, whose image ϕ|L| (X) is an algebraic curve
of degree equal to 3, in contradiction with the fact that g(X) ≥ 2 since

2 · (2 − 1)
g algebraic curve of degree 3 in P2 = = 1.
2

Remark 9.10. Let X be a compact connected Riemann surface of genus g(X) = 2. A canonical
divisor KX has the additional properties

deg(KX ) = h0 (X, OX [KX ]) = 2,

and hence X is always hyperelliptic.

9.5 Canonical Map

In this section, the primary goal is to study the canonical map, that is, the map associated to the
canonical divisor KX .
More precisely, we will prove that, if X is a compact connected Riemann surface of genus g(X) ≥ 2
which is not hyperelliptic, then ϕ|KX | is an embedding.
Theorem 9.22. Let X be a compact connected Riemann surface of genus g(X) ≥ 2. Then the
canonical divisor KX is b.p.f., i.e., the map

ϕ|KX | : X → Pg(X)−1 (C)

is a morphism.
Proof. Fix p ∈ X. There is a short exact sequence (see Proposition 8.2) given by

0→
− OX [KX − p] →
− OX [KX ] →
− Cp →
− 0

which induces a long sequence in cohomology (see Theorem 6.13):

− H 0 (X, OX [KX − p]) →

0→ − H 0 (X, OX [KX ]) →
− H 0 (X, Cp ) →
− ...

− H 1 (X, OX [KX − p]) →

... → − H 1 (X, OX [KX ]) →
− 0.

It follows from the Serre Duality Theorem 8.7 that

H 1 (X, OX [KX ]) ∼
= H 0 (X, OX ) ∼
v
= C,

H 1 (X, OX [KX − p]) ∼

= H 0 (X, OX [p]) ∼
v
= C,

where the latter isomorphism is a consequence of the fact that deg p = 1 is an effective divisor, but
X is not isomorphic to P1 (C) since the genus is strictly greater than zero (see Proposition 9.19). As
a consequence, there is an isomorphism
∼ ∼
H 1 (X, OX [KX − p]) − → H 1 (X, OX [KX ]) ,
→C−

101
which, in turn, implies that
H 0 (X, OX [KX − p]) − H 0 (X, Cp )
is a surjective map.
By arbitrariness of p ∈ X, we conclude that the thesis holds true as a consequence of the numerical
criterion (see Theorem 9.3).

Equivalent Approach. By Riemann-Roch 8.6 it turns out that

h0 (X, OX [KX − p]) − h1 (X, OX [KX − p]) = deg(KX − p) +1 − g(X),
| {z } | {z }
=1 =2 g(X)−3

which, in turn, implies that

h0 (X, OX [KX − p]) = g(X) − 1 = h0 (X, OX [KX ]) − 1.
Since this equality holds for every p ∈ X, the numerical criterion allows us again to infer that the
thesis holds true.
Theorem 9.23. Let X be a compact connected Riemann surface of genus g(X) ≥ 2. The canonical
divisor KX is very ample if and only if X is not hyperelliptic.
Proof. First, we observe that Theorem 9.22 asserts that the linear divisor system |KX | is b.p.f. under
these assumptions. Let p, q ∈ X, set ∆ := p + q, and consider the exact sequence in cohomology
given by
− H 0 (X, OX [KX − p − q]) →
0→ − H 0 (X, OX [KX ]) →
− H 0 (X, C∆ ) →
− ...

− H 1 (X, OX [KX − p − q]) →

... → − H 1 (X, OX [KX ]) →
− 0.
By Serre Duality Theorem 8.7 it turns out that
H 1 (X, OX [KX ]) ∼
= H 0 (X, OX ) ∼
v
= C,

H 1 (X, OX [KX − p − q]) ∼

v
= H 0 (X, OX [p + q]) .
In conclusion, it follows from Theorem 9.3 that
D is very ample ⇐⇒ H 0 (X, OX [KX ]) − H 0 (X, C∆ ) ,
and hence it is enough to observe that
H 0 (X, OX [KX ]) − H 0 (X, C∆ ) ⇐⇒ h0 (X, OX [p + q]) = 1
for every p, q ∈ X; or, equivalently,
H 0 (X, OX [KX ]) 6− H 0 (X, C∆ ) ⇐⇒ ∃ p, q ∈ X : h0 (X, OX [p + q]) = 2
which means X is hyperelliptic by definition.
Theorem 9.24. Let X be a hyperelliptic Riemann surface. The morphism ϕ|KX | : X → Pg(X)−1
can be factorized as follows:
g1 νg−1
2
ϕ|KX | : X −→ P1 (C) −−−→ Pg(X)−1 (C),
where νg−1 is the Veronese embedding of degree g(X) − 1.

102
Proof. The morphism ϕ|KX | sends X into an algebraic curve Y ⊆ Pg(X)−1 (C), but it is not 1-1 as a
consequence of the previous characterization. Hence deg ϕ|KX | ≥ 2 and, if we set d := deg Y , then

deg KX = 2 g(X) − 2 = d · deg ϕ|KX | =⇒ d ≤ g(X) − 1.

The reader should convince herself that it suffices to prove that

Y = νg−1 P1 (C) ,

that is, Y is a normal rational curve5 , to conclude the proof.

Step 1. Let π : Ye → Y be the resolution of the singularities of Y (see Subsection 2.2.2), and let
H ⊆ Y be a divisor hyperplane.
The linear system |π ∗ (H)| is the (g(X) − 1)-dimensional space associated to the pullback divisor
∗
π (H), and hence
ϕ|π∗ (H)| : Ye −→ Pg(X)−1 (C)
is a morphism such that
   
deg ϕ|π∗ (H)| Ye ⊆ Pg(X)−1 = g(X) − 1.

Step 2. The Riemann-Roch Theorem 8.6 proves that Ye ∼ = P1 (C), and also
 
H 0 Ye , OYe [π ∗ H] = H 0 P1 (C), OP1 (C) .

It follows that
 
β
ϕ|π∗ (H)| : P1 (C) −→ Pg(X)−1 (C), x 7−→ x 7−→ xα
0 x1 ,
α+β=g(X)−1

that is, ϕ|π∗ (H)| is the Veronese embedding; consequently, we infer that

d = g(X) − 1 and deg ϕ|KX | = 2.

Moreover, for any p, q ∈ X it turns out that

ϕ|KX | (p) = ϕ|KX | (q) ⇐⇒ h0 (X, OX [p + q]) = 2 ⇐⇒ |p + q| = g21 ,

that is,
ϕ|KX | (p) = ϕ|KX | (q) ⇐⇒ g21 (p) = g21 (q),
and this proves that
g1 νg−1
2
ϕ|KX | : X −→ P1 (C) −−−→ Pg(X)−1 (C).

9.6 Riemann-Roch: Geometric Form

In this section, we denote by X a nonhyperelliptic Riemann surface, by ϕ the canonical map ϕ|KX | ,
and we identify X with its image via ϕ (which is an embedding).
5A smooth, rational curve of degree n in the projective space Pn (C).

103
Geometric form. Let D = p1 + · · · + pd ∈ Div(X) be a divisor of degree d. We may always think
of {p1 , . . . , pd } as a set of points in Pg(X)−1 (C); in particular, it makes sense to define

Span(D) := Span (p1 , . . . , pd ) ⊆ Pg(X)−1 (C).

Theorem 9.25 (Riemann Roch, Geometric Form). The projective dimension of D is deg D − 1
minus the dimension of its span, that is,

dim |D| = deg D − 1 − dim Span(D). (9.3)

Proof. The short exact sequence of sheaf maps

0→
− OX [KX − D] →
− OX [KX ] →
− CD →
− 0

induces a long exact sequence in cohomology, that is,

− H 0 (X, OX [KX − D]) →

0→ − H 0 (X, OX [KX ]) →
− H 0 (X, CD ) .

Clearly  
 hyperplanes H of Pg(X)−1 (C) such that
H 0 (X, OX [KX − D]) =
H vanishes on D
 

 
 hyperplanes H of Pg(X)−1 (C) such that
= ,
H vanishes on Span(D)
 

that is,
dim Span(D) + dim |KX − D| = g(X) − 2. (9.4)
v

Since Pg(X)−1 is canonically isomorphic to P H 0 (X, OX [KX ]) we infer that

h1 (X, OX [D]) = h0 (X, OX [KX − D]) = dim |KX − D| + 1,

and, if we substitute it into the identity (9.4), we obtain

h1 (X, OX [D]) = g(X) − 1 − dim Span(D).

By Riemann-Roch 8.6 it follows that

h0 (X, OX [D]) = h1 (X, OX [D]) + deg D + 1 − g(X) =

= deg D − dim Span(D),

and thus (9.3) is proved.

9.7 Clifford Theorem

Recall. Let D be a divisor on a compact connected Riemann surface X of genus g(X). The linear
system of divisors |D| is isomorphic (see Lemma 7.12) to

P H 0 (X, OX [D]) ,

104
and, in particular, it turns out that

dim |D| = h0 (X, OX [D]) − 1.

If deg D ≥ 2 g(X) − 1, then we already know that

h1 (X, OX [D]) = 0 and dim |D| = deg D − g(X).

In this section, we state and prove the so-called Clifford theorem, which concerns divisors of smaller
degree; precisely, it gives a bound on the dimension of |D| when deg D ≤ 2 g(X) − 2.
Lemma 9.26. Let X be a compact connected Riemann surface, and let D ∈ Div(X) be a divisor.
Then
dim |D| ≥ k ⇐⇒ ∀ {p1 , . . . , pk } ⊂ X, ∃ D0 ∈ |D| : spt (D0 ) = {p1 , . . . , pk }.
Proof. We argue by induction. The base step (k = 0) is trivially true, hence we only focus on the
inductive step k =⇒ k + 1.

Inductive step: ” ⇐= ”. Assume that for any collection of k + 1 points of X there exists a
divisor D0 ∈ |D| such that
spt (D0 ) ⊇ {p1 , . . . , pk+1 }.
By inductive assumption, this is enough to infer that - at least - the projective dimension of |D| is
≥ k. Let us pick pk+1 ∈ Basis (Span |D|), and let us consider the divisor D1 := D − pk+1 . Clearly

∀ {p1 , . . . , pk } ⊂ X, ∃ D10 ∈ |D1 | : spt (D10 ) = {p1 , . . . , pk },

thus by inductive assumption it turns out that dim |D1 | ≥ k. On the other hand, we chose pk+1 in
such a way that |D − pk+1 | $ |D|; hence

dim |D| > dim |D1 | ≥ k =⇒ dim |D| ≥ k + 1,

which is exactly what we wanted to prove.

Inductive step: ” =⇒ ”. Assume that dim |D| ≥ k + 1. As remarked in the introduction of the
section, it implies that h0 (X, OX [D]) ≥ k + 2. Let p1 , . . . , pk+1 ∈ X be a given collection of points,
and let us set
∆ := p1 + · · · + pk+1 .
By Proposition 8.2 there is a short exact sequence

0→
− OX [D − ∆] →
− OX [D] →
− C∆ →
− 0,

which induces a long exact sequence in cohomology

− H 0 (X, OX [D − ∆]) →
0→ − H 0 (X, OX [D]) →
− H 0 (X, C∆ ) →
− ...

and we immediately notice that, by assumption, the middle term has dimension ≥ k + 2, while the
last has dimension equal to k + 1. In particular, there exists a section s ∈ H 0 (X, OX [D]) which
vanishes on ∆, i.e.,
s(pi ) = 0, ∀ i = 1, . . . , k + 1.
The proof is now concluded, but it is worth underlining that the right arrow does not need any
inductive assumption since we have never used it in our argument.

105
Corollary 9.27. Let D1 , D2 ∈ Div(X). Then
dim |D1 | + dim |D2 | ≤ dim |D1 + D2 | .
Proof. Let us set di := dim |Di | for i = 1, 2. Given a collection of d1 + d2 points
{p1 , . . . , pd1 , q1 , . . . , qd2 } ⊂ X,
it follows from Lemma 9.26 that there are D10 ∈ |D1 | and D20 ∈ |D2 | divisors such that
spt (D10 ) ⊇ {p1 , . . . , pd1 } and spt (D20 ) ⊇ {q1 , . . . , qd2 }.
On the other hand, the divisor D10 + D20 belongs to |D1 + D2 |, and thus by Lemma 9.26 we infer that
dim |D1 + D2 | ≥ d1 + d2 = dim |D1 | + dim |D2 |.

Remark 9.11. Equivalently, the corollary asserts that the image of the map
µ : H 0 (X, OX [D1 ]) × H 0 (X, OX [D2 ]) → H 0 (X, OX [D1 + D2 ])
has dimension
dim Im(µ) ≥ h0 (X, OX [D1 ]) + h0 (X, OX [D2 ]) − 1.
Theorem 9.28 (Clifford Theorem). Let X be a compact connected Riemann surface of genus g(X),
and let D ∈ Div(X) be an effective divisor of degree deg D ≤ 2 g(X) − 2. Then
1
dim |D| ≤ deg D,
2
and the equality holds if and only if either
(1) D = 0, D = KX ; or
(2) X is hyperelliptic and |D| = r · |E|, where |E| = g21 and r is the number of the couples formed
by hyperelliptic divisors.
Proof. The argument is rather involved; hence we divide the proof into four main steps.

Step 1: Inequality. Assume that h1 (X, OX [D]) = 0. By Riemann-Roch 8.6 it turns out that
1
h0 (X, OX [D]) = deg D − g(X) + 1 ≤ deg D + 1,
2
where the red inequality follows from the assumption on the degree of D.
If h1 (X, OX [D]) 6= 0, then we can also reduce to the Riemann-Roch formula using a simple
trick. Indeed, by duality it turns out that
h1 (X, OX [D]) 6= 0 ⇐⇒ h0 (X, OX [KX − D]) 6= 0,
and hence it suffices to consider the linear systems |D| and |KX − D|. The Riemann-Roch 8.6,
together with Serre Duality Theorem 8.7, implies that
dim |D| − dim |KX − D| = deg D − g(X) + 1, (9.5)
while the previous Corollary 9.27 implies that
dim |D| + dim |KX − D| ≤ dim |KX | = g(X) − 1. (9.6)
We conclude the first part of the proof combining (9.5) and (9.6) to obtain the sought inequality:
1
dim |D| ≤ deg D,
2

106
Step 2: Equality ” ⇐= ”. This implication is trivial since for D = 0 or D = KX the equality
holds as a straightforward application of the Serre Duality Theorem 8.7.
On the other hand, if X is a hyperelliptic surface and |D| = r ·g21 , then by definition deg D = 2·r,
and hence it is enough to notice that r = dim |D|.

Step 3: Equality ” =⇒ ”. Let D 6= 0, KX be any divisor such that

deg D = 2 · dim |D|.
We argue by induction on the projective dimension of |D|. If dim |D| = 1, then the degree of D is
equal to 2 and thus there is nothing to be proved.
Assume that deg D ≥ 4. Let us consider a divisor E ∈ |KX − D|, and let us pick two points
p, q ∈ X such that p ∈ spt(E) and q ∈ / spt(E). The dimension of |D| is greater or equal than 2,
hence by Lemma 9.26 there exists D0 ∈ |D| such that
{p, q} ⊆ spt(D0 ).
e := D0 ∩ E, where the intersection between divisors is to be intended as follows:
Set D
D(s)
e = min{D0 (s), E(s)}, ∀ s ∈ X.

By construction q ∈ e < deg D; similarly p ∈ spt(E) implies that

/ spt(E), thus deg D

p ∈ spt(E) ∩ spt(D0 ) =⇒ deg D

e > 0.

Let us consider the short exact sequence

ψ ϕ
0→
− OX [D]
e −→ OX [D] ⊕ OX [E] −
→ OX [D + E − D]
e →− 0,
where ψ = (ı1 , ı2 ) is given by the pair of inclusions

ı1 : OX [D]
e ,→ OX [D],

ı2 : OX [D] ,→ OX [E],
and ϕ = r1 − r2 is the difference between the maps
h i
r1 : OX [D] −→ OX D + (E − D)
e ,

h i
r2 : OX [E] −→ OX D + (E − D)
e .

The long exact sequence in cohomology gives us the inequality

h0 (X, OX [D] ⊕ OX [E]) = h0 (X, OX [D]) + h0 (X, OX [E]) ≤
   h i
≤ h0 X, OX [D]
e + h0 X, OX D + E − D
e ,

while the fact that E ∼ KX − D implies that

   h i
h0 (X, OX [D]) + h0 (X, OX [E]) ≤ h0 X, OX [D]
e + h0 X, OX KX − D
e ,

that is,    
dim |D| + dim |KX − D| ≤ dim D  + dim KX − D
e  .
 e 

At this point, we observe that:

107
 (1) The left-hand side may be computer explicitly, i.e.,
dim |D| + dim |KX − D| = g(X) − 1.
Indeed, from (9.5) we infer that
1 1
dim |D| = deg D =⇒ dim |KX − D| = g(X) − 1 − deg D,
2 2
and this implies that
1 1
dim |D| + dim |KX − D| = g(X) − 1 − deg D + deg D = g(X) − 1.
2 2
(2) The right-hand side, by Corollary 9.27, satisfies the following inequality:
   
dim D  + dim KX − D e  ≤ g(X) − 1.
 e 

Therefore the projective dimension of the linear system associated with D e is equal to half of the
degree, and by inductive assumption, it turns out that X is hyperelliptic and
 
D = r̄ · g21 .
 e

It remains to prove that D itself is a multiple of g21 . Let E be the hyperelliptic divisor of X, that is,
the divisor such that |E| = g21 , and let us set s := dim |D|; we want to prove that
D + (g(X) − 1 − s) · E = KX .
First, we notice that
dim |(g(X) − 1 − s) · E| = g(X) − 1 − s,
since it can be obtained by repeating g(X) − 1 − s the linear system g21 ; hence by Corollary 9.27 it
turns out that
dim |D| + dim |(g(X) − 1 − s) · E| = g(X) − 1 =⇒

=⇒ h0 (X, OX [D + (g(X) − 1 − s) · E]) ≥ g(X).

On the other hand, a straightforward computation proves that
 
1
deg D + deg (g(X) − 1 − s) · E = deg D + 2 · g(X) − 1 − deg D = 2 g(X) − 2,
2
and thus the Clifford inequality allows us to infer that
dim |D + (g(X) − 1 − s) · E| ≤ g(X) − 1 =⇒

=⇒ h0 (X, OX [D + (g(X) − 1 − s) · E]) ≤ g(X).

In conclusion, the claim is proved since KX is the unique divisor - up to the equivalence relation -
satisfying the two properties
deg KX = 2 g(X) − 2 and h0 (X, OX [KX ]) = g(X).
In particular, the divisor D is equal to KX − (g(X) − 1 − s) · E and X is hyperelliptic; hence
KX = (g(X) − 1) · E =⇒ D = s · E = s · g21 ,
which is exactly what we wanted to prove.

108
Chapter 10

Abel’s Theorem

In this final chapter, we first investigate the relation between the divisor group (Div(X), +) with
the Picard group (Pic(X), ⊗), where X is a compact Riemann surface.
Successively, we introduce the Jacobian manifold associated with a compact Riemann surface X,
and we prove that there is an isomorphism

Pic0 (X) ∼
= Jac(X).
In particular, we derive this result from the Abel theorem (simply stated) and the Jacobi inversion
theorem (entirely proved).

10.1 Picard Group

In this section, we denote by OX and MX the set of all the holomorphic (respectively, meromorphic)
∗
functions f : X → C - where X is a compact Riemann surface -. Moreover, we denote by OX and
∗
MX respectively, the nonzero elements.
Proposition 10.1. Let X be a Riemann surface. There is a canonical isomorphism
 ∗ 
ϕ : (Div(X), +) −−−−−→ H 0 X, MXO∗ ,
∼
X

which is also a group homomorphism.

∗
Proof. In the literature, the sheaf MXO∗ is called the divisor sheaf of X.
X

Step 1. Let  ∗ 
σ ∈ H 0 X, MXO∗
X
be a global section, and let U = {Uα }α∈I be an open covering of X satisfying the usual properties,
that is, 
σ Uα = fα , fα : Uα → C meromorphic function
such that
fα ∗
fα 6= 0 and ∈ OX (Uα ∩ Uβ ) .
fβ
In particular, for any p ∈ Uα ∩ Uβ it turns out that

ordp fα = ordp fβ ,

109
and hence the divisor X
D= ordp fαp · p
p∈X

is well-defined, where Uαp is an element of U that contains p.

P
Step 2. Let D = p∈X n(p) · p be a divisor of X. There exists an open covering {Uα }α∈I of
X with the property that for p there is a neighborhood Up and there is a holomorphic function
gp, α ∈ OX (Uα ) such that  

div gp, α  = p.
Up ∩Uα

We define Y n(p)
fα = (gα, p ) ∈ M∗X (Uα ),
p∈X
∗
and it is easy to prove that the collection {fα }α defines a global section of MXO∗ .
X

Picard Group. Recall that the invertible sheaves on X form a group, called the Picard group of
X, with the tensor product; more precisely, we have the isomorphism

(Pic(X), ⊗) ∼
= H 1 (X, O∗ ) .

The short exact sequence

∗
0→ ∗
− OX → − MXO∗ →
− M∗X → − 0
X
induces a long exact sequence in cohomology, i.e.,
 ∗ 
∗
− H 0 (X, OX
0→ )→ − H 0 X, MXO∗ →
− H 0 (X, M∗X ) → − ...
X

 ∗ 
∗
− H 1 (X, OX
... → )→ − H 1 X, MXO∗ →
− H 1 (X, M∗X ) → − 0.
X

By Proposition 10.11 - which is proved afterwards - we have

H 1 (X, M∗X ) = H 1 (X, MX ) = 0.

Hence the isomorphisms above, along with the definitions, prove that the long exact sequence reduces
to the shorter exact sequence given by

− C∗ →
0→ − H 0 (X, M∗X ) →
− Div(X) →
− Pic(X) →
− 0,

and hence
Div(X) − Pic(X).

Equivalent Assertion. Given an invertible sheaf L, there is a covering {Ui }i∈I of X and there
are maps 
ψi : L Ui → OX (Ui ), σ 7→ fi · σ,

and hence we can take D U = div(fi ).
i

110
Proposition 10.2. Let ∼ be the equivalence relation on Div(X) given by
D1 ∼ D2 ⇐⇒ D1 − D2 = div(g).
The map ϕ : Div(X) −→ Pic(X) passes to the quotient, and it induces an isomorphism

e : Div(X)∼ −→ Pic(X)
ϕ

D 7−→ OX [D],
which is also a group homomorphism.
Proof. Let {Ui }i∈I be a covering of X, and let fi be the local equation of D, that is,

div fi = D Ui .
1
The map ψi : OX (Ui ) −→ OX [D](Ui ) sends 1 to fi for every i ∈ I, and hence we have a nice local
explicit expression for Ψ.

Injective. We claim that

ϕ(D) = 0 ⇐⇒ D = div(h) ⇐⇒ D ∼ 0.
If D = div h, then the sheaf OX [D] is globally defined by h1 , and hence it is isomorphic to OX , that
is, ϕ(D) = 0.
Vice versa, if ϕ(D) = 0, then the map OX [D] −→ OX is an isomorphism and it defined by the
product with a function h, i.e.,
Ψ : OX [D] → OX , σ 7−→ h · σ.
1
In particular, the divisor D is equal to div h by definition, and this is enough to prove that the claim
holds.

Group Homomorphism. We want to prove that

?
ϕ(D1 + D2 ) = OX [D1 ] ⊗ OX [D2 ].
But this is an easy consequence of the fact that there is an isomorphism given by
∼
OX [D1 ] ⊗ OX [D2 ] −
→ OX [D1 + D2 ], σ1 ⊗ σ2 7−→ σ1 · σ2 .

Structure of Pic(X). The Picard group of X may also be seen as the countable disjoint union of
”Picard strips”, that is, G
Pic(X) = Picd (X),
d∈Z
where
Picd (X) = {L ∈ Pic(X) | deg L = d } .
The degree of an invertible sheaf may be defined by the surjective map Div(X)  Pic(X) simply
considering the image of Divd (X). In a similar fashion, one may define it as
deg L := χ (X, L) − χ (X, OX ) ,
coherently with the fact that
deg D = d =⇒ deg OX [D] = d,
as a straightforward consequence of the Riemann-Roch Theorem 8.6

111
Remark 10.1. For every divisor d ∈ Z there is an isomorphism
∼
Picd (X) −−−−−→ Pic0 (X), L 7−→ L ⊗ OX [−d · p],

which can also be seen via ϕ as

∼
Divd (X) −−−−−→ Div0 (X), D 7−→ D − d · p.

10.2 Jacobian of X
Let X be a compact Riemann surface of genus g(X) := g, and let us consider the symplectic basis1
of the first homology group H1 (X, Z) given by the closed paths a1 , . . . , ag , b1 , . . . , bg (see Figure
10.1), satisfying the intersection conditions

ai · bi = 1 ∀ i = 1, . . . , g,
(10.1)
a · b = 0 otherwise.
j k

Figure 10.1: The generators of the π1 (Tg ).

Remark 10.2. Recall that, for every divisor D ∈ Div(X),

ΩX
1 [D] = OX [KX + D].

By Serre Duality Theorem 8.7 it turns out that

H 0 X, Ω1X ∼= H 0 (X, OX [KX ]) =⇒ H 0 X, Ω1X ∼= H 1 (X, OX ) .

1 Definition. A symplectic basis is a basis e1 , f1 , . . . , en , fn of a vector space endowed with a nondegenerate

alternating bilinear form satisfying (10.1).

112
Definition 10.3 (Period). A period is a linear functional
Z
H 0 X, Ω1X 3 ω 7−→


ω ∈ C,
[C]

where [C] is a class of H1 (X, Z).

Periods Matrix. Fix a basis {ω1 , . . . , ωg } of H 0 X, Ω1X , and let us consider the symplectic
basis {a1 , . . . , ag , b1 , . . . , bg } of H1 (X, Z). The matrix
R R R R 
a
ω1 . . . . . . ag 1
ω b1 1
ω ... ... bg 1
ω
 1. .. .. .. .. .. 
 . . .
 . . . . 

 . .. .. ..  ∈ M (C, g × 2g) ,
 . .. ..
R . . R . R . . R . 


a1 g
ω ... ... ag g
ω b1 g
ω ... ... bg g
ω

is called the periods matrix, and it has the form (A |B ). From now on, we shall denote by
Z 
Ai = ωj ,
ai j=1, ..., g

Z 
Bi = ωj ,
bi j=1, ..., g

the columns of the these two matrices.

Lemma 10.4.

(1) The matrices A, B ∈ M (C, g × g) are invertible.

(2) The column vectors Ai , Bi are a real basis of Cg , i.e., they are 2g R-linear vectors.
(3) The transpose commutes, that is,
AT B = B T A.

Definition 10.5 (Jacobian). The Jacobian of X is the quotient between the dual space of Ω1X and
the lattice induced by the matrix of periods, that is,

v
H 0 X, Ω1X
Jac(X) := R R  .
ω , bi ωj
ai j i, j=1, ..., g

Remark 10.3. The Jacobian of X is also equal to the quotient

v
H 0 (X, OX [KX ])
Jac(X) = ,
ı (H 1 (X, Z))

v
where ı : H 1 (X, Z) → H 0 X, Ω1X is the map sending a class of equivalence [C] to the functional
Z
H 0 X, Ω1X 3 ω 7−→


ω ∈ C.
[C]

113
Remark 10.4. Let us consider Λ - the lattice in Cg generated by a1 , . . . , ag , b1 , . . . , bg -, that is,
 
X g X g 
Λ := mi · Ai + nj · Bj | mi , nj ∈ Z .
 
i=1 j=1

g
The Jacobian of X is isomorphic to the quotient C Λ via the inclusion maps, and it is hence a
complex (compact) torus.

10.3 Abel-Jacobi Map

Let us fix a point p0 ∈ X. The Abel-Jacobi map is defined by
Z p 
ω
 p0 1 
 
 . 
 . 
Ap0 : X −→ Jac(X) p 7−→  .  (mod Λ) ,
 
Z p 
 
ωg
p0

where Z p Z
= ,
p0 γ
for any path γ starting from p0 and ending in p.
Remark 10.5. The Abel-Jacobi map is well-defined. Indeed, if γ and γ 0 are two paths between p0
and p, then Z Z Z
ω− ω = ω ∈ Λ,
γ γ0 η

where η is a closed path with base point p0 (see Figure 10.2).

The map Ap0 : X → Jac(X) may be extended by linearity to Div(X), and from now on we shall
denote by A0 the restriction of Ap0 to Div0 (X).
Remark 10.6. The map A0 does not depend on the base point p0 . Indeed, if we consider a divisor
n
X n
X
D= pi − qi ∈ Div0 (X),
i=1 i=1

then Z  Z  Z Z 
ω1 ω1 ω1 − ω1
 αi   βi   αi βi

     
 .   .   .. 
  .
A0 (pi − qi ) =  .. − . = . (mod Λ) .
   

     
Z  Z  Z Z 
     
ωg ωg ωg − ωg
αi βi αi βi
The reader may jump to Figure 10.3 to have a better understating of what is going on here. For
every i the closed path ηi := αi − γi − βi belongs to H1 (X, Z), and hence
Z Z T
ω1 , . . . , ωg = (0, . . . , 0)T .
ηi ηi

114
 Figure 10.2: Well-definition of the Abel-Jacobi map

As a consequence, we obtain
Z 
ω1
 γi 
 
 . 
A0 (pi − qi ) =  .. (mod Λ) ,
 

 
Z 
 
ωg
γi

which implies that A0 : Div0 (X) → Jac(X) does not depend on the base point p0 .

10.4 Abel-Jacobi Theorems

In this final section, we state and partially prove two fundamental results concerning the Abel-Jacobi
map A0 , and we conclude by showing the main consequences they lead us to.
Theorem 10.6 (Abel’s Theorem). Let X be a compact Riemann surface, and let D ∈ Div0 (X).
Then A0 (D) = 0 if and only if there exists a meromorphic function f : X → C such that D = div(f ),
i.e., D ∼ 0.
Theorem 10.7 (Jacobi Inversion Theorem). Let X be a compact Riemann surface, and fix p0 ∈ X.
For any λ ∈ Jac(X), there are p1 , . . . , pg ∈ X points such that
g
!
X
A0 (pi − p0 ) = λ.
i=1
Pg
Moreover, if λ is a generic point, then the divisor D = i=1 pi is unique.

115
 Figure 10.3: Independence of A0 from p0 .

Corollary 10.8. There are isomorphisms

0
= Div (X)∼ ∼
Pic0 (X) ∼ = Jac(X).

Abel-Jacobi Map in Positive Degree. Fix p0 ∈ X, and let us consider the set of all positive
divisor of degree d, that is,
n o
Divd+ (X) := D ∈ Divd (X) | D ≥ 0 .

The map Ad is defined by taking the restriction of Ap0 to Divd+ (X) and, in particular, it is defined
by setting  d Z 
X pi
 ω1 

 i=1 p 0


 .. 
d .
Ad : Div+ (X) −→ Jac(X), D 7−→  (mod Λ) .
 

 
 
X d Z pi 
ωg
 
i=1 p0

116
The Abel-Jacobi map induces a commutative diagram given by
Ad ∼
Divd+ (X) Jac(X) Picd (X)

τ t

A0 ∼
Div0 (X) Jac(X) Pic0 (X)

where
τ (D) = D − d · p0 and t(L) = L ⊗ OX [−d · p0 ].
Observe that X is canonically isomorphic to Div1+ (X),
hence the Abel-Jacobi map induces
 p 
Z
ω
 p0 1 
 
 . 
 . 
A1 : X −→ Jac(X), D 7−→  .  (mod Λ) .
 
Z p 
 
ωg
p0

Proposition 10.9. Let X be a compact connected Riemann surface of genus g(X) ≥ 1. Then the
Abel-Jacobi map A1 : X → Jac(X) is injective.
Proof. We argue by contradiction. Let p, q ∈ X be points such that A1 (p) = A1 (q), and observe
that the divisor p − q has degree zero. Therefore

A1 (p) − A1 (q) = A0 (p − q) = 0,

and by Abel Theorem it follows that p − q = div(f ), where f is a function with a zero and a pole.
The resulting morphism f : X → P1 (C) has degree one, and thus X ∼ = P1 (C), which is absurd since
the genii are different by assumption.
Remark 10.7. Let λ ∈ Jac(X) be a point such that

λ = Ap0 (D)

for some divisor D ∈ Divd+ (X). Then

A−1
p0 (λ) = |D|

since Ap0 (D) = Ap0 (D0 ) if and only if D0 ∼ D (see the proof of Proposition 10.9).
g(X)
Proposition 10.10. Let X be a compact connected Riemann surface, and let D ∈ Div+ (X) be a
generic divisor. Then h1 (X, OX [D]) = 0.
Proof. By Jacobi Theorem it turns out that for any λ ∈ Jac(X) generic point there exists a unique
g(X)
divisor D ∈ Div+ (X) such that Ap0 (D) = λ. The previous remark, on the other hand, asserts
that
A−1
p0 (λ) = |D|,

and hence |D| = {D} is given by the divisor itself only. Hence h0 (X, OX [D]) = 1, and by the
Riemann-Roch Theorem it follows that

h1 (X, OX [D]) = h0 (X, OX [D]) − deg D − 1 + g(X) = g(X) + 1 − g(X) − 1 = 0.

117
Proposition 10.11. Let X be a compact Riemann surface, and let M be the field of all the mero-
morphic functions defined on X, and let MX be the constant sheaf on X. Then

h1 (X, MX ) = 0.

Proof. Let D ∈ Div(X) be a divisor such that h1 (X, OX [D]) = 0 (which exists by Proposition
10.10). There is a long exact sequence in cohomology given by

− H 0 (X, OX [D]) →
0→ − H 0 (X, MX ) →
− H 0 (X, τX [D]) →
− ...

− H 1 (X, OX [D]) →
... → − H 1 (X, MX ) →
− H 1 (X, τX [D]) .

The sheaf τX [D] is supported on a finite set of points, hence the dimension h1 (X, τX [D])) is equal
to zero; by construction we also have that h1 (X, OX [D]) = 0, thus the thesis holds true.

Conclusion. In this last paragraph, we only give a proof of the Jacobi inversion theorem. The
Abel theorem requires a lot of work, and we will give it for granted (the reader may consult [2, pp.
250-263] for a lengthy proof).

Proof of Jacobi Theorem. Let us denote by X (g) the symmetric product of g(X) copies of X, that
is, the quotient
X (g) = X × · · · × Xσg .

First, we observe that dimC X = 1 is a particular case because it allows us to conclude that X (g) is
a g-manifold2 (i.e., X (g) is not singular, as it could have been in higher dimension). Let
 g Z 
X pi
 
p1  ω1 
i=1 p0
 
 
..
   
(g) (g)
.
A : X −→ Jac(X),  .  7−→ 
.
 . 
 (mod Λ) .
   
 g Z 
X pi 
pg  ωg 
i=1 p0

be the Abel-Jacobi map naturally defined on X (g) . It is enough to prove that for any generic point
(p1 , . . . , pg ) ∈ X (g) , the Abel-Jacobi map A(g) is a local isomorphism.


Step 1. Let {ω1 , . . . , ωg } be a basis for H 0 X, Ω1X , and take p1 , . . . , pg ∈ X distinct points such
that
ωi = hi dzi ,
where zi is the local coordinate relative to pi and (z1 , . . . , zg ) local coordinates of X (g) , and also
satisfying the following property: the g × g complex matrix

(hj (pi ))i, j=1, ..., g

is upper triangular, and it is not degenerate.

2 The reader may refer to this post for a proof of this fact.

118
Step 2. In a neighborhood of (p1 , . . . , pg ) ∈ X (g) it turns out that
 g Z 
X pi +zi
 ω1 

 i=1 p 0


 .. 
(g) .
A (p1 , . . . , pg ) =  (mod Λ) ,
 

 
 g Z 
X pi +zi 
 ωg 
i=1 p0

and hence
g Z
 
X pi +zi
 h1 dz1  h (p )
 p  1 1
 i=1 0  
..

∂ (g) ∂     . 
A (p1 , . . . , pg ) = .  =  .. 
 
.
∂ zi ∂ zi 

  
 g Z 
X pi +zi 
 h dz  hg (pg )
g g
i=1 p0

The matrix associated with the differential dA(g) is thus given by

(hj (pi ))i, j=1, ..., g ,

which is a maximal rank matrix by construction (recall that we chose the pi ’s in such a way to have
this property), and hence A(g) is a local isomorphism.

Step 4. The surface X is irreducible and compact. Hence the map A(g) is proper, the manifold
Jac(X) is irreducible and A(g) is also a dominant map3 , that is, A(g) is surjective.

Step 5. The fiber of A(g) is isomorphic to the linear system |p1 + · · · + pg |, and this is isomorphic
to the projectivization of the global sections, i.e.,

|p1 + · · · + pg | ∼
= P H 0 (X, OX [p1 + · · · + pg ]) .

Since A(g) is a local isomorphism, it turns out that the dimension of the fiber is zero and hence there
exists one and only one divisor p1 + · · · + pg (if a projective subspace contains two points, then it
contains the whole line between them).

3 The reader may jump here for an overview of dominant morphisms between schemes.
Definition 10.12. A morphism f : X → S of schemes is dominant if the image of f is a dense subset of S.
Lemma 10.13. Let f : X → S be a morphism of schemes. If every generic point of every irreducible component of
S is in Im f , then f is dominant.

119
Index

g21 , 100 degree, 55

effective, 56
Abel’s Theorem, 115 group, 55
Abel-Jacobi map, 114 hyperplane, 62
on divisors, 114 invertible sheaf, 56
on divisors of degree d, 116 Laurent tail, 74
on points, 114 linear equivalence, 58
Abelian Integrals, 6 linear system, 59
Algebraic curve of degree 0, 97
Classification theorem, 10 of high degree, 86
Resolution theorem, 14 of low degree, 97
Smoothness, 10 poles/zeros, 56
Arithmetic genus, 72 principal, 55
Automorphisms pullback, 64
of genus 0 surfaces, 30 sheaf, 109
of genus 1 surfaces, 30 very ample, 83
b.p.f., 62 Euler-Poincaré characteristic
Betti’s number, 26 holomorphic, 72
Biholomorphic functions, 8 topological, 26
Blowup, 13
Exceptional line, 13 GAGA Principle, 51
Branch point, 26 Geometric genus, 81
Group action, 34
Canonical map, 101 class formula, 34
for hyperelliptic curve, 102 effective, 34
for nonhyperelliptic curve, 102 holomorphic, 34
Cech
coboundary, 48 Hodge analytic interpretation, 82
Cochains, 48 Holomorphic
cocycle, 48 1-form, 38
cohomology, 49 1-form on Riemann surface, 39
complexes, 48 degree, 26
Chow Theorem, 94 function, 20
Clifford Theorem, 104 map, 22
Complex Atlas, 8 map to projective space, 60
normal form, 24
Dehomogenization, 11 normal form, multiplicity, 24
Divisor, 55 singularity types, 20
1-form, 66 Homogenization, 15
ample, 83 Hurwitz’s Formula, 26
canonical, 67

120
Hurwitz’s Theorem, 36 stalks, 44
Hyperelliptic curve, 98 support, 50
tensor product, 53
Jacobi Theorem, 115 Singular Points
Jacobian, 113 Infinitley near, 14
multiplicities, 10
Laurent tails, 74 of a curve, 9
Logarithmic differential, 42 ordinary, 11
Stokes Theorem, 41
Meromorphic
Strict transform, 13
1-form, 38
Structure Theorem, 8
1-form on Riemann surface, 39
1-form order, 39 Weierstrass ramification point, 99
function, 21 Winding number, 40
order, 21
Mittag-Leffler problem, 74
Morphism, 23

Numerical Criterion, 84

Path integration, 40
Period, 113
Periods Matrix, 113
Poincaré disk, 12

Ramification point, 26
Reduced and Irreducible, Curves, 9
Residues Theorem, 41
Riemann Surface, 8
Complex Torus, 18
quotient, 35
Riemann Sphere, 17
Riemann-Hurwitz Theorem, 67
Riemann-Roch Theorem, 73
Geometric Form, 103

Serre Duality Theorem, 76

Sheaf, 44
and algebraic curves, 54
associated sheaf, 46
coherent, 49
inverse, 53
invertible, 51
kernel, 46
Laurent tails, 75
line bundles equivalence, 53
long exact sequence, 50
morphism, 45
on Riemann surfaces, 47
short exact sequence, 46
skyscraper, 71

121
Bibliography

[1] S. S. Chern and V. V. Shokurov. The collected papers of Wei-Liang Chow. World Scientific
Publisching Co., River Edge, NJ, 2002.

[2] Miranda R. Algebraic Curves and Riemann Surfaces. Graduate Studies in Mathematics. Amer-
ican Mathematical Society, April 1, 1995.

122