The Riemann-Roch Theorem

by

Carmen Anthony Bruni

A project
presented to the University of Waterloo
in fulfillment of the
project requirement for the degree of
Master of Mathematics
in
Pure Mathematics

Waterloo, Ontario, Canada, 2010

c Carmen Anthony Bruni 2010

 Declaration

I hereby declare that I am the sole author of this project. This is a true copy of the project,
including any required final revisions, as accepted by my examiners.

I understand that my project may be made electronically available to the public.

ii
 Abstract

In this paper, I present varied topics in algebraic geometry with a motivation towards the
Riemann-Roch theorem. I start by introducing basic notions in algebraic geometry. Then
I proceed to the topic of divisors, specifically Weil divisors, Cartier divisors and examples
of both. Linear systems which are also associated with divisors are introduced in the next
chapter. These systems are the primary motivation for the Riemann-Roch theorem. Next, I
introduce sheaves, a mathematical object that encompasses a lot of the useful features of the
ring of regular functions and generalizes it. Cohomology plays a crucial role in the final steps
before the Riemann-Roch theorem which encompasses all the previously developed tools. I
then finish by describing some of the applications of the Riemann-Roch theorem to other
problems in algebraic geometry.

iii
 Acknowledgements

I would like to thank all the people who made this project possible. I would like to thank
Professor David McKinnon for his support and help to make this project a reality. I would
also like to thank all my friends who offered a hand with the creation of this project. This
project was funded in part by an Ontario Graduate Scholarship and a President’s Scholarship
from the University of Waterloo.

iv
 Dedication

To my family and friends.

v
Contents

1 Introduction 1

2 Introductory Algebraic Geometry 2

2.1 Affine Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Projective Geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Varieties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

3 Weil Divisors 8

4 Cartier Divisors 12

5 Linear Systems 14

6 Sheaves 17
6.1 Correspondence Between Sheaves and Cartier Divisors . . . . . . . . . . . . 25
6.2 Connection between Weil Divisors and Cartier Divisors . . . . . . . . . . . . 26
6.3 Canonical Class . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

7 Čech Cohomology 31
7.1 Theorems of Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

8 Riemann Roch Theorem 37

9 Applications of Riemann Roch 39

References 43

vi
Chapter 1

Introduction

Algebraic geometry is the study of geometric principles using algebraic ideas. Through
this approach, many new results can be derived and many innovative ideas have been de-
veloped by using correspondences between algebra and geometry. Ideas such as divisors,
sheaves, and sheaf cohomology can all be found to have underlying geometric roots despite
being very algebraic concepts.

In this paper, I wish to describe one very important connection between algebra and
geometry called the Riemann-Roch theorem. This useful theorem of arithmetic geometry
gives us an idea of how topological, algebraic and geometric ideas can all come together
to form a beautiful succinct formula. It encompasses many of the ideas from divisors and
sheaves and results from very straightforward applications of these intricate concepts.

I will begin by introducing some of the basic terminology from algebraic geometry that
will serve as a foundation for subsequent chapters. I will then discuss Weil divisors followed
by Cartier divisors and their connections. Linear systems are the next natural step towards
the Riemann-Roch theorem and are indeed one of the more motivating structures towards
the inception of this theorem. Next, I will introduce the powerful language of sheaves and
discuss how divisors bring about special kinds of sheaves. I will also introduce the canonical
class associated with varieties. I then delve into Čech cohomology which I will use strictly
as a device geared towards the proof of the Riemann-Roch theorem. As such, I will not fully
develop the theory but rather explain the notation and terminology and then reference many
of the major results that I will use to prove the Riemann-Roch theorem. Equipped with
our armamentarium of mathematical tools, I will present the Riemann-Roch theorem in its
entirety followed by some of its many applications to algebraic geometry.

1
Chapter 2

Introductory Algebraic Geometry

In this chapter, I will go through some of the preliminary notions that are fundamental in
the study of algebraic geometry. In general, our primary focus will be with smooth projective
varieties but whenever possible, I will present theorems in their most general form. I start
with reviewing the notions of a variety and then proceed into projective varieties. I will
define what it means to take the dimension of a variety and I will talk about the local ring
of a variety X with respect to a subvariety Y . This turns out to be a discrete valuation ring
whenever the codimension of Y is 1 and in the next chapter, I will define a valuation on it
to help describe properties with divisors.

Throughout this paper, let k be an algebraically closed field.

2.1 Affine Geometry

Definition 2.1. Let An := {(a1 , .., an ) | ai ∈ k}. A subset X ⊆ An is an algebraic set if it is
the zero set of a set of polynomials S defined over k[x1 , .., xn ]. Denote the zero set by V (S).

Definition 2.2. Let X ⊆ An be a non-empty algebraic set. Suppose that X is such that if
X = X1 ∪ X2 for algebraic sets X1 and X2 then one of X1 or X2 equals X. Then we call X
an irreducible algebraic set.

Definition 2.3. An affine algebraic variety is an irreducible algebraic set. We will system-
atically identify algebraic varieties over k with their k-rational points.

Definition 2.4. The ideal of an algebraic set is I(X) := {f ∈ k[x1 , .., xn ] | f (p) =
0 for all p ∈ X}. Note that if X ⊆ Y for algebraic sets X, Y , then I(Y ) ⊆ I(X).

2
Definition 2.5. The coordinate ring of an affine variety X, denoted by Γ(X), is

Γ(X) := k[x1 , .., xn ]/I(X)

Proposition 2.6. An affine algebraic set X is a variety if and only if I(X) is a prime ideal
in Γ(X).

Proof. ⇒ Suppose X is a variety and let f, g ∈ k[x1 , .., xn ] with f g ∈ I(X). Then < f g >⊆
I(X). Moreover, X = V (I(X)) ⊆ V (f g) = V (f )∪V (g). Hence X = (X ∩V (f ))∪(X ∩V (g)).
Without loss of generality, the irreducibility of X implies that X = X ∩ V (f ) ⊆ V (f ) and
thus f ∈ I(X) proving that I(X) is prime.

⇐ Suppose that I(X) is prime and suppose X = X1 ∪ X2 . This reveals that I(X) =
I(X1 ) ∩ I(X2 ). If I(X) = I(X1 ) then X1 = X. Otherwise, there exists an f ∈ I(X1 )\I(X).
However, for any g ∈ I(X2 ), it is the case that f g ∈ I(X1 ) ∩ I(X2 ) = I(X). The primality
of I(X) yields that g ∈ I(X) (since f ∈
/ I(X)). Thus I(X2 ) ⊆ I(X) ⊆ I(X2 ) and hence
I(X) = I(X2 ) in other words X = X2 . Hence X is irreducible as required. 

2.2 Projective Geometry

Definition 2.7. Define the projective n-space Pn to be

Pn = An+1 \{0}/ ∼

where ∼ is an equivalence relation defined by

(x0 , .., xn ) ∼ (y0 , .., yn )

⇔ (x0 , .., xn ) = λ(y0 , .., yn )

for some λ ∈ k ∗ .

Remark. In a similar manner to the affine case, we can define a projective algebraic set, the
ideal of a projective algebraic set, and an irreducible projective algebraic set.

Definition 2.8. Let X ⊆ Pn . Then X is a projective variety if X = {x ∈ Pn | f (x) =

0 for all f ∈ S} where S is a set of homogeneous polynomials of degree d defined over
k[x0 , .., xn ]. Note that although it does not make sense to evaluate functions at projective
points, testing to see whether a function vanishes at a projective point is a well defined notion.
This follows by noting for x := [x0 : .. : xn ] ∈ Pn and s ∈ k, one sees that f (sx) = sd f (x)

3
and so if x is a zero point, then f (sx) = f (x) = 0. This is necessary since points in Pn are
only defined up to multiplication by a nonzero scalar. Notice that functions in the fraction
f
field, say g
are well defined so long as the degree of the numerator equals the degree of the
denominator. This occurs because

f (sx) sdeg(f ) f (x) f (x)

= deg(g) = sdeg(f )−deg(g)
g(sx) s g(x) g(x)

Thus, the evaluation map is well defined if and only if deg(f ) = deg(g).

2.3 Varieties
From here on out, unless otherwise stated, I will use the word variety to mean either an
affine variety or a projective variety.

Proposition 2.9. Every algebraic set (affine or projective) is the finite union of varieties.

Proof. Let X be an algebraic set. If X is irreducible then we are done so suppose it is not
irreducible. Then X = X1 ∪ X10 with neither equal to X. If both X1 and X10 are irreducible
then we are done. Otherwise, suppose without loss of generality that X1 cannot be written
as the finite union of varieties. Then continuing on in this fashion, one gets a chain of sets
X ) X1 ) X2 ) ... that does not terminate. Consider the ideals of each set. By Hilbert’s
Nullstellensatz, we have a one to one correspondence between Xi and I(Xi ). In particular,
I(Xi ) = I(Xi+1 ) if and only if Xi = Xi+1 . This shows us that no two of the I(Xi ) are
equivalent. Taking the ideals gives us

I(X) ) I(X1 ) ) I(X2 ) ) ...

which is a non stabilizing strictly ascending chain of ideals contradicting the fact that the
ring is Noetherian. Thus we get termination and so each algebraic set is a finite union of
varieties. 

Remark 2.10. Notice that this can be extended to a unique decomposition of a variety if
we insist that no irreducible variety is contained inside another in the decomposition.

Definition 2.11. Let X be a variety with X equal to the zero set of S, a set of polynomials
over k[x1 , .., xn ]. Then, X is a smooth variety if at every point p ∈ An (respectively Pn−1 if
∂f ∂f
the variety is projective) ∇f (p) := ( ∂x1
(p), ..., ∂xn
(p)) 6= 0 for some f ∈ S.

4
Definition 2.12. One can equip An or Pn with a topology by making every algebraic set
closed. This is called the Zariski topology.

Definition 2.13. Let X be a variety (a subset of either An or Pn−1 ) and x ∈ X. A function

f from X to k is regular at x if there is an open neighbourhood of x, say x ∈ U ⊆ X, and
p(y)
two polynomials p, q ∈ k[x1 , .., xn ] such that f (y) = q(y)
for all y ∈ U and q(y) 6= 0. In the
projective case, we also require that deg(p) = deg(q). This function is said to be regular on
X if it is regular at every point. The ring of regular functions is denoted by OX (X). The
ring of regular functions on an open subset of X is denoted by OX (U). If the base variety is
clear, then the subscript is often omitted.

Remark 2.14. Note that if f is regular on X, it may be that over different open sets,
different polynomials p and q are used.

Definition 2.15. Let X be a variety and Y a subvariety of X. The local ring of X along Y ,
denoted by OY,X , is the set of pairs (U, f ) where U ⊆ X is open, U ∩ Y 6= ∅ and f ∈ OX (U)
under the identification that (U1 , f1 ) = (U2 , f2 ) if f1 = f2 on U1 ∩ U2 . Note that this is a
local ring with the maximal ideal consisting of pairs (U, f ) such that f |U ∩Y = 0. This ideal
is written as

MY,X := {f ∈ OY,X | f (x) = 0 for all x ∈ Y }

Definition 2.16. The function field of a variety X is OX,X , the local ring of X along X.
This is denoted by k(X) and the non-zero elements by k(X)∗ . In particular, if X is projective
g
and f = h
∈ k(X)∗ , then deg(g) = deg(h) by (2.8) and thus deg(f ) := deg(g) − deg(h) = 0.
Elements of the function field are called rational functions.

Remark 2.17. We could also have defined the function field of an affine variety to be the
fraction field of the coordinate ring Γ(X) (which is just OX (X)). In this manner, we could
extend the definition to the projective case by restricting our variety to an affine open piece.
This definition is well defined (that is, independent of the open affine piece we choose) as
open sets are dense in the Zariski topology. In fact, this is precisely how we will define
rational functions on an open set.

Definition 2.18. A map φ : X → Y between two varieties X and Y is a morphism if it is

continuous and for every open set U ⊆ Y and every regular function f on U, we have that
f ◦ φ is regular on φ−1 (U). We say that X and Y are isomorphic if there exists two regular

5
functions f : X → Y and g : Y → X such that g ◦ f and f ◦ g are the identity maps on X
and Y respectively.

Definition 2.19. Let X be a variety. Then we define an open affine subset of X to be an

open subset of X that is isomorphic to an open subset of An .

Definition 2.20. Let X be an affine variety and U an open subset of X. Then a rational
function on U is an element of the field of fractions of the ring of regular functions on U. This
is well defined since the ring of regular functions is indeed an integral domain (see [2, p.82]).
More generally, if X is a variety, U is any open subset of X, and V is an open affine subset
(which exists by [2, p. 25]) of X, then define a rational function on U to be an element on
the fraction field of the ring of regular functions on U ∩ V. This is well defined as open sets
are dense in the Zariski topology.

Definition 2.21. Let X be a variety. Then the dimension of X is the transcendence degree
of the function field k(X) over k and is denoted by dim X. If Y is a subvariety of X, then
the codimension of Y in X is dim X − dim Y .

Proposition 2.22. Let X be a variety and Y a subvariety of X. If X 6= Y then dim Y <

dim X

Proof. See [3, p. 55]. 

I would like to finish this section with an important idea from ring theory. These upcoming
propositions help to give light to the local rings and their structure. Discrete valuation rings
are the precise structure that allows us to connect divisors and functions in a natural manner.

Definition 2.23. A discrete valuation ring is a principal ideal domain that is not a field and
contains a unique maximal ideal.

Note. A discrete valuation ring can also be described as a Noetherian ring that is not a field
and with a unique maximal ideal that is principal.

Definition 2.24. In the case of a discrete valuation ring, if the maximal ideal is M = (t),
then t is a uniformizer for M .

Proposition 2.25. Let R be a discrete valuation ring with maximal ideal M = (t). For
every r ∈ R, if r 6= 0 then there is a unique n ∈ N and u ∈ R∗ such that r = utn .

Proof. If r is a unit, then use n = 0. Otherwise, r ∈ (t) (since it is maximal) so there exists
a r1 ∈ R such that r = r1 t. Now r1 ∈ R so either r1 is a unit in which case the process

6
stops or r1 ∈ (t) and so there exists an r2 ∈ R such that r1 = r2 t. Continuing this process,
eventually it is the case that rn ∈ R∗ for otherwise, (r1 ) ⊆ (r2 ) ⊆ ... is a chain of ideals that
does not stabilize contradicting the fact that a discrete valuation is Noetherian. Thus, this
process stops for some rn . Then r = r1 t = r2 t2 = ... = rn tn with rn ∈ R∗ as claimed. 
With all of this in place, I conclude with a theorem that will be used in the next section.
Its proof however required a bit more machinery and is omitted for succinctness.

Proposition 2.26. Let X be a smooth variety and Y a subvariety of codimension 1. Then

OY,X is a discrete valuation ring.

Proof. See [2, p. 174]. 

7
Chapter 3

Weil Divisors

Divisors are a very useful tool for studying algebraic geometry. They reveal a large amount
of information about the variety in question. One can retrieve information about their zeros,
poles, and structure of functions defined over the variety through the use of divisors. To
begin, I start by looking at Weil divisors and how one can associate functions to divisors. I
then prove some important properties of divisors as well as give a couple of examples.

Definition 3.1. Let X be an algebraic variety. A Weil divisor is a finite formal sum of
codimension one subvarieties of X. This is denote by D := ki=1 nYi Yi where each of the Yi
P

are subvarieties of X of codimension one.

There is a natural group structure on the set of all Weil Divisors of an algebraic variety
where addition is defined variety wise on two divisors, that is,
k
X k
X k
X
mYi Yi + n Yi Y i = (mYi + nYi )Yi
i=1 i=1 i=1

Let Div (X) = {D|D is a Weil Divisor}. This forms a free abelian group of codimension one
subvarieties of X.

Definition 3.2. The degree of a divisor to be the sum of the nYi ’s. This is denoted by
deg(D).

Definition 3.3. A divisor D is effective (denoted D ≥ 0) if every term in the divisor is

non-negative. That is, nYi ≥ 0 for all i. One can induce a partial order on divisors by setting
D1 ≥ D2 ⇔ D1 − D2 ≥ 0.

Definition 3.4. The support of a divisor is the union of all Yi with nYi 6= 0.

8
Example 3.5. Let X be the zero set of f , where f is defined by f (x, y, z) = xy − z 2 . Notice
that X is a smooth projective plane curve. Divisors on X can be defined as formal sums of
points of X. For example D := 3[1 : 1 : 1] − [0 : 1 : 0] is a divisor on X. This divisor is not
effective as the coefficient of [0 : 1 : 0] is negative. Its support is {[1 : 1 : 1], [0 : 1 : 0]}

Next I define the order of a function. One can use the order of a function to help define
a divisor canonically associated to functions.

Definition 3.6. Let X be a smooth algebraic variety and Y a subvariety of X of codimension

1. Recall from proposition (2.26) that OY,X is a discrete valuation ring. We define the
valuation on this ring as follows. The order of a function f ∈ OY,X , denoted ord Y (f ), to be
the value of n from (2.25). That is, if M = (t) is the maximal ideal for OY,X and f = utn
∗
for some unit u ∈ OY,X then ord Y (f ) = n.

Note that ord Y : OY,X \{0} → N (alternatively, one could define ord Y (0) = ∞ and just
extend the range). In particular, there is a natural extension to the fraction field k(X)∗ by
f
defining the order of g
∈ k(X)∗ to be ord Y ( fg ) = ord Y (f ) − ord Y (g) thus giving a function
from k(X)∗ to Z.

Proposition 3.7. The order function has the following properties for all f, g ∈ k(X)∗

(i) ord Y (f g) = ord Y (f ) + ord Y (g)

(ii) Only finitely many Y have the property that ord Y (f ) 6= 0

(iii) ord Y (f ) ≥ 0 ⇔ f ∈ OY,X

∗
(iv) ord Y (f ) = 0 ⇔ f ∈ OY,X

(v) If X is projective, then the following are equivalent

1. ord Y (f ) ≥ 0 for all Y

2. ord Y (f ) = 0 for all Y

3. f ∈ k ∗ (f is a non-zero constant function)

Proof. (i) Let f = u1 tn1 and g = u2 tn2 from (2.25). Note that f g = u1 tn1 u2 tn2 =
(u1 u2 )tn1 +n2 and so ord Y (f g) = ord Y (f ) + ord Y (g). 

9
 p(x1 ,..,xm )
(ii) Let f = q(x1 ,..,xn )
. If ord Y (f ) 6= 0, then Y ⊆ X ∩ V (pq). So either X ⊆ V (pq) in
which case X ⊆ V (p) and f is the zero function (a contradiction) or dim(X ∩ V (pq)) =
dim(X) − 1. Using the decomposition of a variety (2.10) as well as proposition (2.22),
we have that each of the components must be of dimension at most dim(X) − 1. The
uniqueness tells us that Y has to be contained in one of the finitely many irreducible
components of X ∩ V (pq). Since Y also has dimension dim(X) − 1, the proposition
(2.22) tells us that Y has to be equal to one of the finitely many irreducible components.


∗
(iii) If ord Y (f ) =: n ≥ 0 then by (2.25), write f = utn for some unit u ∈ OY,X and where
the maximal ideal is (t). By closure of OY,X , it must be that f ∈ OY,X . If f ∈ OY,X ,
then f = utn by (2.25) and so ord Y (f ) = n ≥ 0. 

∗
(iv) If ord Y (f ) = 0, then one can write by (2.25) f = ut0 = u for some unit u ∈ OY,X . So
∗ ∗
f is a unit and thus f ∈ OY,X . Conversely, if f ∈ OY,X then f is a unit so f = f t0 and
thus ord Y (f ) = 0. 

(v) (1) ⇒ (2) Let f = hg . Since X is projective, deg(g) = deg(h) by ( 2.16). If ord Y (f ) ≥
g
0 for all Y , then h
= f = utn by (2.25). Now, since u is a unit, one must have
u1
that u = u2
with deg(u1 ) = deg(u2 ). Cross multiplying and taking the degrees gives
deg(g) + deg(u1 ) = deg(h) + deg(u2 ) + n and thus n = 0. Hence ord Y (f ) = 0 for all Y .

(2) ⇒ (1) Immediate.

∗
(2) ⇒ (3) By (iv), f ∈ OY,X . Since this must hold for all Y , it is clear that f cannot
have any zeroes or poles for if it did there would be a Y where the function f would
not be a unit. Thus, f is a function without zeroes and poles and hence, f ∈ k ∗ .

(3) ⇒ (2) Immediate. 

Definition 3.8. Let X be a variety and f ∈ k(X)∗ . The divisor of f is

X
div (f ) = ord Y (f )Y ∈ Div (X)
Y

This is well defined by the previous proposition.

Definition 3.9. A divisor is said to be principal if it is the divisor of a function.

10
Proposition 3.10. Every divisor over An is principal.
Pk
Proof. Let D := i=1 nYi Yi ∈ Div (An ) and suppose Yi is defined by fi ∈ k[x1 , ..., xn ].
Consider the function
k
Y
f := fini
i=i

This is a well defined function with the property that div (f ) = D as required. 

Definition 3.11. Let X be an algebraic variety and let D, D0 ∈ Div (X). Then D and D0
are said to be linearly equivalent (denoted D ∼ D0 ) if D = D0 + div (f ) for some f ∈ k(X)∗ .
This defines an equivalence relation on Div (X).

Definition 3.12. The divisor class group of X (denoted Cl (X)) is the group of divisor
classes modulo linear equivalence.

Example 3.13. Let f (x, y, z) = y 2 z − x3 − xz 2 and suppose X = V (f ). Note that X is a

smooth projective plane curve. Let D1 := 2[0 : 0 : 1] + [0 : 1 : 0] and D2 := 3[0 : 1 : 0].
Notice that D1 ∼ D2 as div ( xz ) = 2[0 : 0 : 1] − 2[0 : 1 : 0] = D1 − D2 . In fact, D1 = div (x)
and D2 = div (z).

11
Chapter 4

Cartier Divisors

With Weil divisors, subvarieties of codimension one were used to tell us properties about
the variety. These can be defined locally as the zeroes and poles of a single function. Cartier
divisors, give a way to reverse this process, that is I will define divisors using this local
property and ensure that functions mesh appropriately in a global sense. Similar to Weil
divisors, I will introduce the concept and its important properties. I will then describe how
Weil and Cartier divisors are related as well as explain an important example called the
canonical divisor.

Definition 4.1. Let X be an algebraic variety and I an indexing set. A Cartier divisor is
an equivalence class of collection of pairs [{(Ui , fi )}i∈I ] such that
S
(i) i∈I Ui = X with each Ui open

(ii) fi ∈ k(X)∗

(iii) fi fj−1 ∈ O(Ui ∩ Uj )∗

(iv) The equivalence relation ∼ is defined by {(Ui , fi )}i∈I ∼ {(Vj , gj )}j∈J ⇔ fi gj−1 ∈ O(Ui ∩
Vj )∗ for every i ∈ I and j ∈ J

As done with Weil divisors, Cartier divisors form a group denoted CaDiv (X) by defining
the sum of two divisors as

{(Ui , fi )}i∈I + {(Vj , gj )}j∈J = {(Ui ∩ Vj , fi gj )}i∈I,j∈J

Note that indeed fi gj (fk gl )−1 ∈ O((Ui ∩ Vj ) ∩ (Uk ∩ Vl ))∗ as fi fk−1 ∈ O(Ui ∩ Uk )∗ and
gj gl−1 ∈ O(Vj ∩ Vl )∗ both hold by property (iii). Thus, our function is defined on the
quadruple intersection and is invertible there.

12
Definition 4.2. The group of Cartier divisor classes modulo linear equivalence is called the
Picard group of X and is denoted by Pic (X).

Definition 4.3. The support of a Cartier divisor is defined to be the union of the set of
zeroes and poles of all of the fi .

Definition 4.4. An effective Cartier divisor is a Cartier divisor such that every fi ∈ O(Ui ).
In other words, each fi has no poles.

Definition 4.5. Let X be a variety and f ∈ k(X)∗ . Then associate to f a Cartier divisor
called the principal Cartier divisor by div (f ) = {(X, f )}
It turns out that there is a natural connection between Weil divisors and Cartier divisors
in a smooth setting. Due to the nature of the proof, it is easier to justify this claim in a later
chapter after developing more tools. For now I will state the claim as it ties in the previous
two chapters very nicely.

Theorem 4.6. Let X be a smooth variety. Then there exists a φ such that

φ : CaDiv (X) → Div (X)

defines an isomorphism between Cartier divisors and Weil divisors. Further, this map extends
naturally to an isomorphism between Pic (X) and Cl (X).

13
Chapter 5

Linear Systems

Linear systems will turn out to be a motivating factor for coming up with the Riemann-
Roch theorem. This reveals another way of looking at divisors algebraically through the view
point of linear algebra.

Definition 5.1. Let X be a variety and let D ∈ Div(X). To D, we associate a vector space
L(D) over k by

L(D) = {f ∈ k(X)∗ |D + div(f ) ≥ 0} ∪ {0}

It is a matter of checking the definition to verify that L(D) is indeed a k-vector space. We
denote the vector space dimension of L(D) by l(D). Further we associate a projective variety,
denote by P(L(D)), to L(D) via the following definition.

P(L(D)) := L(D)/k ∗ = L(D)/ ∼

where ∼ is the equivalence relation defined by f ∼ g ⇔ f = λg where λ ∈ k ∗ .

Definition 5.2. The set of all effective divisors linearly equivalent to D is called the complete
linear system and is denoted by |D|.

Remark 5.3. Consider the following map

Φ : P(L(D)) → |D| = {D0 | D0 ≥ 0, D0 ∼ D}

f (mod k ∗ ) 7→ D + div (f )

Notice that this map is one to one for if D + div (f ) = D + div (g) then indeed f = λg and
so f = g in P(L(D)). This map is also onto as any element of |D| is D + div (f ) for some
function f . Hence, this map is a bijection. Using this Φ map allows us to define a linear
system as follows.

14
Definition 5.4. A linear system on a variety X is a set of effective divisors all linearly
equivalent to a divisor D and parametrized by a linear subvariety (one generated by linear
forms) of P(L(D)). Being parameterized by a linear subvariety means that when we view
our linear system under the Φ map (in particular its inverse), we get a linear subvariety of
P(L(D)).

Remark 5.5. It turns out that one can also define a linear system L as a subset of |D| such
that the following set is a k-vector subspace of k(X).

V (L) := {f ∈ k(X)∗ | D + div (f ) ∈ L} ∪ {0}

This can be seen by examining the map Φ defined in (5.3). Using this correspondence, we
note that P(V (D)) will be a linear subvariety of P(L(D)) if and only if L is a linear system.

Remark 5.6. Another important fact that will be used is that when X is projective, this
dimension is finite and in fact dim |D| = l(D) − 1 (see [3, p. 55]).

Proposition 5.7. Let X be a variety and let D, D0 ∈ Div(X). Then,

(i) k ⊆ L(D) if and only if D ≥ 0

(ii) If D ≤ D0 then L(D) ⊆ L(D0 )

(iii) If D0 = D + div(g) (with g 6= 0) then the map θ : L(D0 ) → L(D) defined by θ(f ) = gf
is an isomorphism of k-vector spaces. In particular, the dimension `(D) depends only
on the class of D in Pic(X).

Proof. (i) If k ⊆ L(D), then taking some c ∈ k one sees that D + div(c) ≥ 0 or in other
words, D ≥ 0. Moreover, if D ≥ 0, then D + div(c) = D + 0 ≥ 0 for all c ∈ k. Thus
k ⊆ L(D). 

(ii) Let f ∈ L(D). This yields D0 + div(f ) ≥ D + div(f ) ≥ 0 and thus f ∈ L(D0 ). 

(iii) I claim that θ is a linear map. If f, f 0 ∈ L(D0 ) and c ∈ k then

θ(f + cf 0 ) = g(f + cf 0 ) = gf + cgf 0 = θ(f ) + cθ(f 0 )

Therefore the map is linear. It is one to one since if f ∈ ker(θ), then 0 = θ(f ) = gf
and since g 6= 0 it must be that f = 0. Moreover, this function is surjective for if

15
 f0
f 0 ∈ L(D), then consider f = g
. This f lies in L(D0 ) since

D0 + div (f ) = D0 + div (f 0 ) − div (g) = D + div (g) + div (f 0 ) − div (g) = D + div (f 0 ) ≥ 0
(5.1)
0
where the last inequality holds since f 0 ∈ L(D). Moreover, θ(f ) = g fg = f 0 and so θ is
surjective. Hence, it is an isomorphism as k-vector spaces. 

Proposition 5.8. Let X be a projective variety and let D ∈ Div(X). If deg(D) < 0 then
l(D) = 0.

Proof. Let f ∈ L(D) and suppose that f is a non-zero function. Then D + div (f ) is effective
and by (2.16) it is the case that deg(div (f )) = 0 and hence that 0 ≤ deg(D + div (f )) =
deg(D). This is a contradiction. Hence f is the zero function. Thus, L(D) = {0} and hence
l(D) = 0.

16
Chapter 6

Sheaves

Sheaves is the next step on the road to Riemann-Roch. The concept of a sheaf dates back
about twenty years before its use in algebraic geometry in 1954 by Serre. I will give many
examples of sheaves throughout this section and prove several of the important properties
that will be needed in the proof of the Riemann-Roch theorem. I will discuss exact sequences
as they will be used in the next section where cohomology is introduced.

Definition 6.1. Let X be a topological space. A presheaf F on X consists of the following:

(i) For each U ⊆ X with U open, a set F(U)

(ii) For all open subsets W ⊆ V ⊆ U ⊆ X , a map

ρU ,V : F(U) → F(V)

with ρU ,U = idF (U ) and ρU ,W = ρV,W ◦ ρU ,V . These maps are sometimes called restriction
maps.

If F(U) has additional structure such as being a group, ring, or module, then the presheaf
is called a presheaf of groups, rings or modules respectively. In this case, the ρ maps defined
above must also preserve this structure.

Definition 6.2. Let X be a topological space. A presheaf F is a sheaf if for any open U ⊆ X
S
and any open cover i∈I Ui , the following is true,

(i) Let f, g ∈ F(U) with ρU ,Ui (f ) = ρU ,Ui (g) for all i ∈ I. Then f = g.

(ii) Given fi ∈ F(Ui ) such that ρUi ,Ui ∩Uj (fi ) = ρUi ,Ui ∩Uj (fj ), then there exists a (unique)
f ∈ F(U) with ρU ,Ui (f ) = fi for all i ∈ I.

17
Definition 6.3. Let F be a sheaf on a topological space X. The set F(X) (treating X as
an open subset of itself) is the set of global sections of F.

Definition 6.4. Let X be a variety. An OX -module is a sheaf F on X such that

(i) For each open set U ⊆ X, F(U) is an OX -module over the ring OX (U).

(ii) For each pair of open sets V ⊆ U ⊆ X, the map ρU ,V : F(U) → F(V) is an OX (U)
module homomorphism.

Example 6.5. (i) Direct Product of Sheaves: Let Fλ for λ ∈ Λ be sheaves on a topological
space X. The sheaf F on X defined by
Y
F(U) = Fλ (U)
λ∈Λ
Y
ρU ,V = ρλU ,V
λ∈Λ
Q
is a sheaf called the direct product of sheaves. It is denoted by F = λ∈Λ Fλ .

For the following examples, let X be a variety equipped with the Zariski topology.
Then the following, equipped with the natural restrictions ρU ,V , define sheaves of OX -
modules.

(ii) Skyscraper Sheaf : Let p ∈ X. Define for an open set U ⊆ X


k if p ∈ U

K(p) (U) =
0 if p ∈

/U

The sheaf K(p) is called the skyscraper sheaf.

(iii) Sheaf of Rational Functions: Define

KX (U) = { rational functions defined on U}

The sheaf KX is called the sheaf of rational functions.

(iv) Sheaf of Regular Functions: Define

OX (U) = { regular functions defined on U}

The sheaf OX is called the sheaf of regular functions.

18
 (v) Sheaf of a Divisor : Let D := {(Ui , fi ) | i ∈ I} be a Cartier divisor. Define as a
sub-OX -module of KX

1
LD (Ui ) = OX (Ui )
fi
= {f ∈ KX (Ui ) | D + div (f ) ≥ 0}

The equality is immediate becuase this sheaf must generate only functions f where
D + div (f ) ≥ 0 as it is a submodule of KX . Notice since Ui form an open cover, this
determines LD . The second definition shows immediately that LD depends only on
the equivalence class of D. Also, from the second definition, we can see that L−D =
fi OX (Ui ).

Definition 6.6. Let F, G be two (pre)sheaves. A morphism of (pre)sheaves φ : F → G is

a collection of homomorphisms φU for each open U ⊆ X that preserves the structure of the
sheaf (for example, if F(U) was a OX -module, then these homomorphisms are required to
be OX -module homomorphisms) and such that the following diagram is commutative,

U φ
F(U) −−− → G(U)
 
ρ ρ̂
y U ,V y U ,V
V φ
F(V) −−−→ G(V)

Definition 6.7. Two (pre)sheaves F and G are isomorphic if there is a morphism {φU } of
(pre)sheaves such that for every open set U ⊆ X, we have that φ|U is an isomorphism in the
appropriate setting (for example, as topological spaces, groups, rings, or modules).

Definition 6.8. Let F be a sheaf on a topological space X and let x ∈ X. Define an

F
equivalence relation ∼ on F := x∈U F(U) as follows. Let f ∈ F(U) and g ∈ F(V). Then
f ∼ g if ρU ,U ∩V (f ) = ρV,U ∩V (g). The stalk at x, denoted Fx is

Fx := lim F(U) := F/ ∼
−→
x∈U

The construction above is called the direct limit of the F(U) over all open sets containing x.
The elements of a stalk are called germs. Notice that if the sheaf is a sheaf of groups, rings,
or modules, then the stalk is a group, ring, or module respectively.

Example 6.9. If one takes the sheaf of rings of regular functions OX , then the stalk at a
point x is simply the ring of functions regular at x, namely Ox,X .

19
Note. Sheaf homomorphisms induce homomorphisms at the level of stalks. These will pre-
serve the structure (say for group, ring or module homomorphisms).

Theorem 6.10. Let φ : F → G be a morphism of sheaves defined over a topological space

X. Then φ is an isomorphism if and only if the induced maps on the stalks, denoted φx , are
isomorphisms for each x ∈ X.

Proof. The proof, while not difficult, is technical. See ([2, p. 63]).
One key idea still unexplored is the ability to turn a presheaf into a sheaf. This will
become important for defining properly the tensor product of two sheaves. It turns out that
there is an elegant way to perform this construction called the sheafification of a presheaf.

Definition 6.11. Let F be a presheaf. Then the sheafification of F is a sheaf F + together

with a morphism θ : F → F + such that it satisfies the following universal property. For any
sheaf G and any morphism φ : F → G, there is a unique morphism ψ : F + → G such that
the following diagram commutes:
θ
F > F+

φ ψ
∨
>

G
Proposition 6.12. The sheafification of a presheaf F always exists and is unique up to
isomorphism.

Proof. We construct the sheaf F + as follows. Let U be an open set and F + be the set of
F
functions s : U → p∈U Fp such that for each p ∈ U, we have that s(p) ∈ Fp and there is
a neighbourhood V ⊆ U with p ∈ V and an element t ∈ F(V) such that for all q ∈ V the
germ tq = s(q). From here, it is easy to verify that F + equipped with the natural restrictions
forms a sheaf. This induces a natural morphism θ : F → F + that has the universal property.
Moreover, the uniqueness is a result of the universal property. 

Example 6.13. (i) Inverse Image Sheaf : Let f : X → Y be a continuous function from
two topological spaces X and Y . For and sheaf G on Y , define the inverse image sheaf
to be the sheafification of the presheaf

U 7→ lim G(V)
V ⊇f (U )

where the limit is the usual colimit taken over any open set V that contains f (U). This
is denoted by f −1 G.

20
 (ii) Restriction Sheaf : Let Z be a subset of a topological space X and view Z as a topo-
logical subspace when equiped with the induced topology. If i : Z → X is the inclusion
map, and F is a sheaf on X, then we call i−1 F the restriction of F to Z and denote it
by F |Z .

(iii) Image and Kernel Sheaf : Let F and G be sheaves defined over a topological space X.
Let φ : F → G be a morphism of sheaves. Then we define the kernel sheaf to be the
sheaf

U 7→ ker(φU )

(note that this is a presheaf which is also a sheaf so taking the sheafification is redun-
dant). Also, we define the image sheaf to be the sheafification of the presheaf

U 7→ im(φU )

(iv) Homomorphism Sheaf : Let F and G be two OX -modules. For each open set U ⊆ X,
define

HomOX (F, G)(U) := {OX |U -homomorphisms from F |U to G |U }

:= homOX |U (F |U , G |U )

With the natural restrictions, HomOX defines a sheaf.

f g
Definition 6.14. A sequence of sheaves F → G → H is exact if im(f ) = ker(g).

Example 6.15. Let D be a divisor on a projective curve X (so dim X = 1) and let P ∈ X.
There is a natural injection from LD to LD+P directly from the definition. Next, define a
map Φ from LD+P to K(P ) as follows.

0 if P ∈
/U

Φ |U (f ) =
f (P ) if P ∈ U


This map is clearly onto. As such there is an exact sequence

0 → LD → LD+P → K(P ) → 0

Such exact sequences are referred to as short exact sequences.

21
 r
Definition 6.16. Consider OX ⊕ ... ⊕ OX = OX . This is a free OX -module of rank r. Let
| {z }
r
F be an OX -module on X. Then F is locally free if for every x ∈ X, there exists an open
neighbourhood x ∈ U ⊆ X such that F|U ∼ = OX r
|U . A locally free sheaf of rank 1 is also
called an invertible sheaf.

Example 6.17. Let D = {(Ui , fi )}i∈I be a Cartier divisor. Notice that LD is an invertible
sheaf. We use the isomorphism defined by

OX |Ui → L(D) |Ui

1 7→ fi−1

Notice that this can be defined for all open sets since Ui is an open cover so U ∩ Ui 6= ∅ for
some i. For this set U, map 1 to fi−1 . Since D is a Cartier divisor, this is well defined on the
overlap.

Example 6.18. (i) Tensor Product of Sheaves: Define the tensor product of two OX -
modules, F and G, denoted F ⊗OX G, to be the sheafification of the presheaf

U 7→ F(U) ⊗OX (U ) G(U)

When there is no confusion, the tensor product is often denoted by F ⊗ G.

(ii) Exterior Product of Sheaves: Let F be an OX -module. Define the nth exterior product
of F to be the sheafification of the presheaf
n
^
U 7→ F(U)

This along with the natural restrictions forms a presheaf whose sheafification is called
the nth exterior product of two sheaves.

(iii) Dual Sheaf : Let F be a locally free sheaf of finite rank. Define F̆ := HomOX (F, OX )
to be the dual sheaf of F.

Theorem 6.19. Let F and G be locally free OX modules of finite rank r and s respectively.
Then HomO (F, G) ∼
X = F̆ ⊗ G.

Proof. To show this isomorphism, we will use the universal property of sheafification. Let
F̆ ⊗pre G denote the presheaf of F̆ ⊗ G. By the universal property, it suffices to find an

22
isomorphism from this presheaf to HomOX (F, G). Notice that for each s ∈ G(U), there is a
natural map Ψs : OX |U → G |U defined by multiplying elements by s |U . Define

ΦU : homOX |U (F |U , OX |U ) ⊗OX |U G(U) → homOX |U (F |U , G |U )

(φ, s) 7→ (ψs ◦ φ)

Now by (6.10), we have that the collection {ΦU } is an isomorphism if and only if the induced
maps on the stalks are an isomorphism. As F is locally free, the stalks themselves are free
(that is are isomorphic to Ox,X ). At this level, the morphisms are exactly the canonical
isomorphisms of Ox,X modules. This shows that the induced morphisms are isomorphisms
at the level of stalks and hence showing the original isomorphism. 

Proposition 6.20. Let F be a locally free OX module of rank 1. Then,

HomOX (F, F) ∼
= OX

Proof. Let Φ ∈ HomOX (F, F)(U) for some small enough open set U ∈ X. Note that Φ(f ) =
f Φ(1), where 1 is the identity element of OX |U , and so Φ is determined by Φ(1). There are
F |U possible elements all resulting in a homomorphism. However, since F is a locally free
sheaf of rank 1, we have that F |U ∼
= OX |U . Hence HomOX (F, F)(U) ∼
= F |U ∼
= OX |U . This
gives the required sheaf isomorphism. 

Corollary 6.21. Let X be a variety. Then HomOX (OX , OX ) ∼

= OX

Proof. Immediate from the previous proposition. 

Corollary 6.22. Let F be an invertible sheaf on a variety X. Then F̆ is invertible and

F ⊗ F̆ ∼
= OX .

Proof. Immediate from (6.19) and (6.20).

Remark 6.23. Notice that the set of invertible sheaves forms a group with the operation of
tensor product and OX as the identity element. In particular, the sheaf satisfying F ⊗G = OX
is unique since if two sheaves say G, H satisfy

F ⊗G ∼
= OX ∼
=F ⊗H

Then tensoring both sides by F̆ gives us G ∼

= H.

23
Example 6.24. Let D := {(Ui , fi )|i ∈ I} be a Cartier divisor. Note that LD ⊗ L−D ∼
= OX .
We use the universal property to show this. Denote the presheaf of LD ⊗L−D by LD ⊗pre L−D .
Then we define a morphism φU on an open set U ⊆ X by defining it on its intersection of
the open cover via

(LD ⊗pre L−D )(U ∩ Ui ) → OX (U ∩ Ui )

f ⊗ g 7→ f g

This map is well defined on the overlaps by the properties of the Cartier divisor D and thus
forms a map from (LD ⊗pre L−D )(U) to OX (U). This map defines an isomorphism of rings. It
1
is clear that the map is onto for if f ∈ OX (U ∩ Ui ), then fi
⊗ fi f 7→ f . This is one to one for
1 1 1
any element of the tensor product can be described as fi
⊗fi h by noting fi
g ⊗fi h = fi
⊗fi gh.
It is clear from the definition that {φU } is a morphism of sheaves and hence is an isomorphism
of sheaves. Note in particular that this means that L−D ∼ = L˘D by the uniqueness discussed
in the previous remark.

Theorem 6.25. Let D1 , D2 be Cartier divisors on a variety X. Then LD1 −D2 ∼

= LD1 ⊗ L˘D2

Proof. Let D1 = {(Ui , fi )}i∈I and D2 = {(Vj , gj )}j∈J so D1 − D2 = {(Ui ∩ Vj , fi gj−1 )}i∈I,j∈J .
By the above remark (6.23), it suffices to show that LD −D ∼ = LD ⊗ L−D . To show this
1 2 1 2

isomorphism, we will use the universal property of sheafification. Let LD1 ⊗pre L−D2 denote
the presheaf of LD1 ⊗ L−D2 . By the universal property, it suffices to find an isomorphism
from this presheaf to LD1 −D2 . Define a ring homomorphism ΦU for any open set U ⊆ X by
defining it on the intersections of the open cover Ui ∩ Vj via,

(LD1 ⊗pre L−D2 )(U ∩ Ui ∩ Vj ) → LD1 −D2 (U ∩ Ui ∩ Vj )

f ⊗ g 7→ f g

Similar to (6.24), we can see that we have defined a map on all of U by considering it over
1
every set in the open cover. The function f g is in LD1 −D2 (U ∩ Ui ∩ Vj ) since f = h
fi 1
and
1 1
g= h
gj 2
and so f g = hh
fi g j 1 2
∈ LD1 −D2 (U ∩ Ui ∩ Vj ). On the overlaps, the maps are well
defined by the properties of Cartier divisors. Just as before, we can show that this map is
indeed an isomorphism at the level of rings. More importantly, {ΦU } defines an isomorphism
of sheaves showing the two sheaves are isomorphic. 

Corollary 6.26. L0 ∼
= OX

24
Proof. Let D be a Cartier divisor on X. Then the above shows that

L0 = LD−D ∼
= L˘D ⊗ LD ∼
= HomOX (LD , LD ) ∼
= OX

where the last two equalities holds by (6.25) and since LD is locally free of rank 1 (6.20)
applies. 

6.1 Correspondence Between Sheaves and Cartier Di-

visors
We would like to reinterpret Cartier divisors in the language of sheaves. In order to do
this, we shall restrict our attention to sheaves over an abelian group. This allows us to define
the quotient sheaf.

Definition 6.27. Let F be a sheaf. A sheaf G is said to be a subsheaf of F if for every open
set U ⊆ X, G(U) is a subgroup of F(U) and the restriction maps of G are the maps induced
by F.

Definition 6.28. Let G be a subsheaf of F and let φ : G → F be the natural inclusion map.
Define the quotient sheaf F/G to be the sheafification of the presheaf U 7→ F(U)/G(U).

Definition 6.29. A Cartier divisor on a variety X is a global section of the quotient sheaf
∗ ∗
KX /OX where the ∗ denotes the set of invertible elements of their respective sheaves.

Remark 6.30. This definition completely coincides with our previous definition of a Cartier
divisor. If D is a Cartier divisor according to our previous definition, then it is clear from
that definition that we get a global section of the presheaf of our quotient sheaf. Using the
sheafification map gives us one for the quotient sheaf. Now, consider an element of the global
∗ ∗ ∗ ∗
section of KX /OX . Denote the presheaf by F and F + := KX /OX , the sheafification of the
presheaf. This element is a map s(x) = sx where sx ∈ Fx and obeys an additional condition.
This condition states that there is an open covering Ui of X and sections fi ∈ F(Ui ) with
the property that whenever x ∈ Ui , then sx is the germ of fi at x. When we look at what
F is, it is immediate that we have defined the conditions for the first definition of a Cartier
divisor.

It turns out that there is a nice correspondence between Cartier divisors and invertible
sheaves which I now describe.

25
Theorem 6.31. The map D 7→ LD gives a one to one correspondence between Cartier
divisors on X and invertible subsheaves on KX .

Proof. Notice that LD is an invertible sheaf by (6.17) and hence the map takes divisors into
invertible sheaves. Next, start with an invertible subsheaf of KX say F and take fi on the
open set Ui to be the inverse of a local generator of the sheaf on Ui . Doing this for an open
1
cover will give us a Cartier divisor say D. Applying the map gives us LD = O (Ui )
fi x
= F(Ui )
true by choice of fi as the inverse of a local generator. This gives the required one to one
correspondence. 

6.2 Connection between Weil Divisors and Cartier Di-

visors
With the language of sheaves behind us, I can now prove the correspondence between
Weil divisors and Cartier divisors on smooth varieties. Before I do this I need one more
definition.

Definition 6.32. Let X be a variety and Y a subvariety. Let IY (U) be the ideal in the ring
OX (U) consisting of the regular functions which vanish at all points of Y ∩ U. The presheaf
U 7→ IY (U) is a sheaf called the sheaf of ideals IY of Y . Notice that this is a subsheaf of
OX . We can actually define a sheaf of ideals to be a subsheaf of OX , say I, such that I(U)
is an ideal of OX (U).

Theorem 6.33. Let X be a smooth variety. Then there exists a φ such that

φ : CaDiv (X) → Div (X)

defines an isomorphism between Cartier divisors and Weil divisors. Further, this map extends
naturally to an isomorphism between Pic (X) and Cl (X).

Proof. Let Y be an irreducible subvariety of codimension one of X and let D = {(Ui , fi )}i∈I
be a Cartier divisor. The goal is to define a notion of order for D along Y . So pick a Ui
so that Ui ∩ Y 6= ∅. Define ord Y (D) = ord Y (fi ) (here, smoothness is necessary to ensure
that the notion of an order is well defined). This notion is well defined for if some other
Uj was chosen so that Uj ∩ Y 6= ∅, then by the definiton of a Cartier divisor we have that
fi fj−1 ∈ O(Ui ∩ Uj )∗ . This means that this function is invertible. So by (3.7) it is the case

26
that ord Y (fi fj−1 ) = 0 and thus ord Y (fi ) = ord Y (fj ). Now, map the Cartier divisor D to
P
the Weil divisor ord Y (D). The construction also clearly takes principal Cartier divisors
to principal Weil divisors and effective Cartier divisors to effective Weil divisors.
P
For the other direction, let D = nYi Yi be a Weil divisor. Next, pick an x ∈ X and
consider the local ring Ox,X . Notice that by 2.17 each Yi corresponds to an ideal sheaf
Pi = I(Yi ) in Γ(X). Then Pi Ox,X is principal by [2, p.7]. So let fx be a generator for
Pi Ox,X . Then on some open neighbourhood Ux of x, it is true that ÷(fx ) = D|Ux . It is
clear that D0 := (Ux , fx ) is a Cartier divisor. Moreover, it is well defined since if fx and gx
are defined on the same open set Ux then fx gx−1 ∈ O(Ux )∗ and by definition define the same
Cartier divisor.

It is easy to show that this construction gives an isomorphism. Moreover, this isomor-
phism can be viewed in the natural way to give an isomorphism from Pic (X) to Cl (X) as
it sends principal divisors to principal divisors and likewise for effective divisors. This gives
the two isomorphisms as needed in the proof. 

6.3 Canonical Class

Now, we can define the canonical class. The canonical class is the set of all canonical
divisors of a variety. I will describe how this class is constructed in this section. First, we
use machinery from module theory to describe differentials so that later we can extend these
results to sheaves. To do this, we also need to describe how to create a sheaf from a module.
We then describe the sheaf of relative differential forms by piecing together affine pieces and
gluing them together.

For this section, let A be a commutative ring with identity, B an A-algebra, and M a
B-module.

Definition 6.34. A A-derivation of B into M is a map d : B → M such that for all a ∈ A

and b, b0 ∈ B ,

(i) d(b + b0 ) = db + db0

(ii) d(bb0 ) = bdb0 + b0 db

(iii) da = 0

27
Definition 6.35. The module of relative forms of B over A is a B-module ΩB/A together
with an A-derivation d : B → ΩB/A that satisfies the following universal property. For
any B-module M , and for any A-derivation d0 : B → M , there exists a unique B-module
homomorphism f : ΩB/A → M such that the following diagram commutes.

d
B > ΩB/A
f
d0 > ∨
M
With this in place, we now wish to decribe the sheaf of relative differentials. Essentially
what we want is to take the previous construction and use it in the setting of sheaves.

Definition 6.36. Let X be a topological space. Let F and G be sheaves of rings with
f : F → G a morphism of sheaves. This allows us to consider G as an F-module. Define
Ωpre
G/F to be the presheaf defined by

U 7→ ΩG(U )/F (U )

with restrictions maps taken using (for V ⊆ U an open set)

G(U) → G(V) → Ωpre

G/F (V)

where the maps are defined by the usual restrictions. Notice that this map is a F(U) deriva-
tion and so it factors through Ωpre pre
G/F (U) giving the restriction maps for ΩG/F . Now that we

have a presheaf, we sheafify it to get ΩG/F , the sheaf of relative differentials.

Definition 6.37. Let f : X → Y be a morphism of varieties. Using the inverse image sheaf,
we can retrieve a morphism

f −1 OY → OX

which turns OX into a f −1 OY module. We define ΩX/Y to be ΩOX /f −1 OY .

Example 6.38. Let’s discuss an important example of the sheaf of differentials. Let X
be a smooth variety and for a fixed p ∈ X, consider ΩX/{p} . Notice that there is only one
morphism from X to {p}, say f . In this case, notice that f −1 (O{p} ) is a sheaf where all of
the stalks are just k. This is true because the inverse image sheaf in this case is a colimit
where there is only one non-empty open set, namely {p} and so the presheaf construction for
the inverse image sheaf simply sends an open set U of X to O{p} ({p}) = k. By construction,
this fact still holds true for the sheafification. Hence the stalk of f −1 (O{p} ) is a direct limit

28
taken where every element of the direct limit is just k and so the stalk is simply k. Notice in
particular that our result doesn’t depend on {p} nor on the morphism (as there is only one
to choose).
Vdim(X)
Definition 6.39. Let X be a smooth variety. The canonical sheaf is ωX = i=1 ΩX/{p} ,
the top exterior product of ΩX/{p} for any p ∈ X. Notice that since X is smooth, our sheaf
ΩX/{p} is a locally free sheaf of rank dim(X) and hence ωX is an invertible sheaf (see [2, p.
177]).

One of the most important properties of the canonical sheaf is that it is an invertible sheaf.
So by (6.31) there is an associated divisor to ωX . This divisor will be called a canonical divisor
and denoted by K. Notice that a canonical divisor is unique up to linear equivalence. This
definition of the canonical class uses a lot of the language of sheaves. From a different view
point, we can define local parameters and local coordinates. Then, we will define the divisor
associated to ωX .

Definition 6.40. Local Parameters. Let x be a smooth point on a variety X with dim(X) =
n. Then t1 , .., tn ∈ Ox,X are called local parameters at x if the ti ∈ Mx and if they give a
basis of Mx /M2x

Definition 6.41. Local Coordinates. Let x be a smooth point on a variety X with dim(X) =
n. Then t1 , .., tn ∈ Ox,X are called local coordinates on X if ti − ti (x) give local parameters
at all x in X.

Example 6.42. Over An , x1 , .., xn is a set of local coordinates for any point x ∈ An .

Let’s decribe the alternate viewpoint for canonical divisors. For any open set U (suffi-
ciently small) choose rational functions f1 , .., fn such that they form a system of local coordi-
nates everywhere on U. Thus, df1 ∧ ... ∧ dfn ∈ ωX (U) is a nonzero differential form. Moreover,
for any α ∈ ωX , one can write α = fU (df1 ∧ ... ∧ dfn ) for some rational function fU . This is
possible since the set of differential forms of degree n form a 1 dimensional linear space over
OX (U) spanned by df1 ∧ ... ∧ dfn (see [6, p. 88] for more details). Do this for an open cover
of X and taking all open sets gives a divisor,

div(α) = {(U, fU )}

Note that this is well defined. If α0 ∈ ωX (U) then note α0 = gα where g ∈ k(X)∗ and so

div (α0 ) = div (α) + div (g)

29
and thus α0 ∼ α as divisors.

In either language, we get a well defined divisor class that describes the same object. In
the original attempt at a result similar to the Riemann-Roch theorem, it was this piece of
the puzzle that was lacking from the final formula. We conclude this section by giving an
example of a computation of a canonical class.

Example 6.43. The Canonical Class of Pn .

x0 x1
Let d xn+1 ∧ d xn+1 ... ∧ d xxn+1
n
be a (non-zero) differential form on Pn . Let

Ui = {x ∈ Pn |xi = 1}

Note that on Un+1 there are no zeroes or poles for the differential form as 1dx0 ∧ ... ∧ dxn and
the constant function 1 has no poles on the open affine set Un+1 . For all other 1 ≤ i ≤ n,

x0
d xn+1 ∧ ... ∧ d xxn+1
i−1 1
∧ d xn+1 ∧ d xxn+1
i+1
... ∧ d xxn+1
n

−x0 −xi−1
1
=( xn+1 dx0 + x2n+1
dxn+1 ) 1
∧ ... ∧ ( xn+1 dxi−1 + x2n+1
dxn+1 ) ∧ ( x−1
2 dxn+1 )
n+1

1 −xi+1 1 −xn
∧ ( xn+1 dxi+1 + x2n+1
dxn+1 )... ∧ ( xn+1 dxn + x2n+1
dxn+1 )
1
=( xi−1 (dx0 ∧ dx1 ∧ ... ∧ dxi−1 )) ∧ ( x−1
2
1
dxn+1 ) ∧ ( xn−i (dxi+1 ∧ ... ∧ dxn ))
n+1 n+1 n+1

= x−1
n+1 (dx0 ∧ dx1 ∧ ... ∧ dxi−1 ∧ dxn+1 ∧ dxi+1 ∧ .. ∧ dxn )
n+1
n−i+1
= (−1)
n+1
xn+1
(dx0 ∧ dx1 ∧ ... ∧ dxi−1 ∧ dxi+1 ∧ ... ∧ dxn+1 )

Thus, as a Cartier divisor, the canonical class of Pn is

n−i+1
D := {(Ui , (−1)
xn+1
)|1 ≤ i ≤ n} ∪ {(Un+1 , 1)}
n+1

or as a Weil divisor,

D := −(n + 1){xn+1 = 0}

30
Chapter 7

Čech Cohomology

Thus far, I have introduced many topics in algebraic geometry. Unfortunately this alone
is not enough machinery to prove Riemann-Roch. In this chapter, I introduce the theory
of cohomology over a topological space. As I am using cohomology mainly as a tool, I will
not prove many of the results, some of which have proofs that far exceed the capacity of a
paper of this size. I do give references however to the proofs should the reader be interested
in pursuing the results.

Throughout this section, unless otherwise stated, let X be a topological space, F a sheaf
of abelian groups on X and let U = {Ui }i∈I be an open cover for X.

Definition 7.1. The nth cochain group is defined by

Y
C n (U, F) = F(Ui0 ∩ ... ∩ Uin )
(i0 ,..,in )∈I n+1

Each cochain group consists of elements f(i0 ,..,in ) ∈ F(Ui0 ∩ ... ∩ Uin ) and addition define on
this group is defined componentwise.

Definition 7.2. Define coboundary operators from the nth cochain group to the (n + 1)st
cochain group as follows.

δn :C n (U, F) → C n+1 (U, F)

n+1
X
f(i0 ,..,in ) (−1)k ρV,W (f(i0 ,..,iˆk ,...,in+1 ) ) =: g(i0 ,..,in+1 )
k=0

where V = Ui0 ∩ ... ∩ Ûik ∩ ... ∩ Uin+1 and W = Ui0 ∩ ... ∩ Uin+1 .

31
Remark 7.3. It is often clear which δn is being used and hence the subscript n is often
omitted.

Definition 7.4. The nth Čech cohomology groups of an open cover U of X is defined as
follows. First H 0 (U, F) := F(X), the group of global sections. Then define for n ≥ 1

n-cocycles Z n (U, F) = ker(δn )

n-coboundaries B n (U, F) = im(δn−1 )

It can be shown that the nth coboundary group lies in the nth cocycle group (by showing
that δn ◦ δn = 0) and thus the quotient group can be formed, namely

Hn (U, F) := Z n (U, F)/B n (U, F)

This is called the nth Čech cohomology group of F with respect to an open cover U.

Note. In the case of the first cohomology group, one denotes elements of Z 1 (U, F) by (fij ).
It is clear by the definition that an element (fij ) is in Z 1 (U, F) if and only if fij = fij + fjk
for all indices i, j, k. This implies that fii = 0 and also that fij = −fji . Moreover, an element
(fij ) is in B 1 (U, F) provided that there is an element (gi ) ∈ C 0 (U, F) with fij = gj −gi = δ(gi )
on Ui ∩ Uj .
The goal now is to extend this notion to one that is independent of the chosen open cover.
It turns out that the most useful way to do this is by using a direct limit construction. In
order to accomplish this, I will introduce a few more definitions.

Definition 7.5. V = {Vj }j∈J is said to be finer than U = {Ui }i∈I if there is a function
τ : J → I such that Vj ⊆ Uτ (j) . Denote this by V < U.

Definition 7.6. If V < U, define τVU : C n (U, F) → C n (V, F) by

τVU ((fi0 ,...,in )) = (fτ (j0 ),...,τ (jn ) |Vj0 ∩...∩Vjn )

where τ (jk ) = ik for all k ∈ {1, .., n}. These maps induce maps on the cohomology groups
going from H n (U, F) to H n (V, F). They are independent of choice of τ and if W < V < U
V
then τW ◦ τVU = τW
U
.

H n (U, F) as follows. Let h1 ∈

F
Definition 7.7. Define an equivalence relation ∼ on U

H n (U, F) and h2 ∈ H n (U0 , F). Then

h1 ∼ h2 ⇔ there exists a V with refinement map τ finer

0
than U and U0 such that τVU (h1 ) = τVU (h2 )

32
Definition 7.8. The nth Čech cohomology group of X is then defined to be
G
H n (X, F) := H n (U, F)/ ∼= lim H n (U, F)
−→
U U

Note that in the n = 0 case, H 0 (X, F) := F(X).

In the case of H 1 , one can actually prove a strong result.

Theorem 7.9. For any sheaf F on X, H 1 (X, F) = 0 ⇔ H 1 (U, F) = 0 for every open cover
U of X.

Proof. See [4, p. 296].

Definition 7.10. A set of functions φi defined on a variety X is called an integer partition

of unity for an open cover {Ui } of X if

(i) The functions only take integer values.

(ii) Every point of X lies in only finitely many of the support sets of the φi (that is, the
set of points where the function is non-zero).
P
(iii) i φi (p) = 1 for all p ∈ X

(iv) The support of each φi is contained in Ui .

Lemma 7.11. Integer valued partitions of unity always exist on a variety X.

Proof. Totally order the open covering {Ui } and set


1 if p ∈ Ui and p ∈ S
/ j<i Uj

φi (p) :=
0 otherwise


These functions give the desired results. 

Proposition 7.12. Let X be a variety and let F = K(p) be the skyscraper sheaf for some
p ∈ X. Then H 1 (X, F) = 0.

Proof. By (7.9), it suffices to show that H 1 (U, F) = 0 where U := {Ui } is an arbitrary open
covering of X. Let φi be an integer valued partition of unity, which exists by (7.11). The
goal is to show that every 1-cochain (fij ) is a coboundary. Define for fixed i and j,

φj fij if p lies in the support of p

ψi,j :=
0

otherwise

33
 P
Next, set gi := − k ψi,k where k ranges over the entire indexing set. Notice that
X X X X
gj − gi = − ψj,k + ψi,k = − φk fjk + φk fik
k k k k
X X
= φk (fik − fjk ) = φk fij = fij
k k

Thus, δ(gi ) = (fij ) is a coboundary and hence the first cohomology group is trivial. 

7.1 Theorems of Cohomology

Here I present theorems that are important in the theory of Čech cohomology. These
theorems will allow for the Euler characteristic of a long exact sequence of cohomology groups
to be defined which will become one of the main ingredients of the proof of the Riemann-Roch
theorem.

Theorem 7.13. (Long Exact Sequence of Cohomology) Let X be a variety and let 0 → F →
G → H → 0 be a short exact sequence of sheaves. Then, there exists morphisms making the
following a long exact sequence:

0 → H 0 (X, F) → H 0 (X, G) → H 0 (X, H) →

→ H 1 (X, F) → H 1 (X, G) → H 1 (X, H) → ...

Proof. See [4, p. 299]. 

Theorem 7.14. (Cartan-Serre) Let X be a smooth projective variety and let F be a locally
free sheaf. Then H n (X, F) is finite dimensional.

Proof. See [5]. 

Theorem 7.15. (Grothendieck’s Vanishing Theorem) Let X be a variety of dimension n.

Then for all i > n and all sheaves of abelian groups F on X, H i (X, F) = 0.

Proof. See [2, p. 208]. 

Theorem 7.16. (Serre Duality) Let X be a smooth projective space, K a canonical divisor,
r := dim X and F a locally free sheaf. Then for each 0 ≤ i ≤ n, we have that,

H i (X, F) ∼
= H n−i (X, L(K) ⊗ F)

Proof. See [2, p. 244]. 

34
Definition 7.17. Let X be a projective variety over a field k and let F be a locally free
sheaf on X. The Euler Characteristic of F is defined by

X
χ(F) = (−1)i dim H i (X, F)
i=0

Notice that by Cartan-Serre, and Grothendieck’s vanishing theorem that this is well
defined.

Proposition 7.18. Let X be a smooth projective variety and let F, G, H be a locally free
sheaf on X. Suppose that

0→F →G→H→0

is a short exact sequence. Then, χ(G) = χ(F) + χ(H).

Proof. By (7.13), this sequence induces a long exact sequence

f1 f2 f3
0 → H 0 (X, F) → H 0 (X, G) → H 0 (X, H) →
f4 f5 f6
→ H 1 (X, F) → H 1 (X, G) → H 1 (X, H) → ...

This sequence terminates for some n by (7.15). Notice that im(f1 ) = {0} = ker(fm ) where
m = 3n + 1 and that im(fi ) = ker(fi+1 ) for i ∈ {1, ..., m − 1} holding by exactness. The
rank-nullity theorem gives

dim H 0 (X, F) = dim ker(f2 ) + dim im(f2 ) = dim im(f1 ) + dim ker(f3 ) = dim ker(f3 )
dim H 0 (X, G) = dim ker(f3 ) + dim im(f3 ) = dim ker(f3 ) + dim ker(f4 )
dim H 0 (X, H) = dim ker(f4 ) + dim im(f4 ) = dim ker(f4 ) + dim ker(f5 )
..
.
dim H n (X, G) = dim ker(fm−2 ) + dim im(fm−2 ) = dim ker(fm−2 ) + dim ker(fm−1 )
dim H n (X, H) = dim ker(fm−1 ) + dim im(fm−1 ) = dim ker(fm−1 ) + dim ker(fm )
= dim ker(fm−1 )

Alternately subtracting and adding rows yields

dim H 0 (X, F) − dim H 0 (X, G) + dim H 0 (X, H) − dim H 1 (X, F) + dim H 1 (X, G)
− dim H 1 (X, H) + dim H 2 (X, F) − ... + (−1)n dim H n (X, F)
+ (−1)n+1 dim H n (X, G) + (−1)n dim H n (X, H) = 0

35
A quick check shows that the left hand side of the above equation reduces to

χ(F) − χ(G) + χ(H) = 0

This gives the result. 

36
Chapter 8

Riemann Roch Theorem

One of the fundamental problems that people were interested in was to count the number
of functions with a specific number of zeros and poles. This number of course is often infinite
and so mathematicians focused on computing the dimension of the space of meromorphic
functions with predefined zeroes and poles. This is the value of l(D) where the divisor D
reflects the zeroes and poles as described in the problem. It is this problem (sometimes called
the Riemann-Roch problem) that the Riemann-Roch theorem solves.

The Riemann-Roch theorem was a dual effort of both Bernhard Riemann and Gustav
Roch in the mid 1800s. Riemann showed that l(D) ≥ deg(D) − dim H 1 (X, OX ) + 1 in all
cases. He could not however identify when this equality held. It was Roch that provided
the error term of l(K − D) thus proving an equality. This theorem will apply many of the
mechanics developed in the previous chapters.

Definition 8.1. The genus of a smooth curve is the value dim H 1 (X, OX ). Denote it by g.

Theorem 8.2. (Riemann Roch Theorem) Let X be a smooth projective curve. Then for all
D ∈ Div (X),

l(D) − l(K − D) = deg(D) − g + 1

where K is a canonical divisor of X.

Proof. Note that l(D) = dim H 0 (X, LD ) and

l(K − D) = dim H 0 (X, L(K − D)) = dim H 0 (X, L(K) ⊗ L˘D )

37
where L˘D denotes the dual sheaf. Since X is a projective variety, the Serre Duality theorem
(7.16) can be invoked which says that H 0 (X, L(K) ⊗ L˘D ) and H 1 (X, LD ) are dual and thus

dim H 0 (X, L(K) ⊗ L˘D ) = dim H 1 (X, LD )

So it suffices to show that

dim H 0 (X, LD ) − dim H 1 (X, LD ) = 1 − g + deg(D)

Proceed by mathematical induction. For D = 0, note that LD = OX (6.26). Moreover,

H 0 (X, OX ) = OX (X) = k which holds since X is a projective variety. Thus dim H 0 (X, LD ) =
1. Moreover, by definition,

dim H 1 (X, LD ) = dim H 1 (X, OX ) = g

and hence

dim H 0 (X, LD ) − dim H 1 (X, LD ) = 1 − g = 1 − g + 0 = 1 − g + deg(D)

proving the base case. Next, suppose the claim is true for any divisor D and show the claim
is true for D + P (this is inductive as any divisor here is just a formal finite sum of points).
First recall by (6.15) that

0 → LD → LD+P → K(P ) → 0

is an exact sequence. By (7.13), this induces a long exact sequence

f1 f2 f3
0 → H 0 (X, LD ) → H 0 (X, LD+P ) → k →
f4 f5 f6
... → H 1 (X, LD ) → H 1 (X, LD+P ) → 0

which holds since dim H 0 (X, K(P ) ) = dim k = 1 and since H 1 (X, K(P ) ) = 0 by (7.12). Next,
note that im(f1 ) = {0} = ker(f6 ) and im(fi ) = ker(fi+1 ) for i ∈ {1..5}. Using the rank-nullity
theorem exactly as in (7.18),

dim H 0 (X, LD ) − dim H 0 (X, LD+P ) + 1 − dim H 1 (X, LD ) + dim H 1 (X, LD+P ) = 0

Rearranging and using the fact that deg(D + P ) − deg(D) = 1 yields

dim H 0 (X, LD ) − dim H 1 (X, LD ) − deg(D)

= dim H 0 (X, LD+P ) − dim H 1 (X, LD+P ) − deg(D + P )

The induction hypothesis says that dim H 0 (X, LD ) − dim H 1 (X, LD ) = 1 − g + deg(D) so
plugging and rearranging yields

dim H 0 (X, LD+P ) − dim H 1 (X, LD+P ) = 1 − g + deg(D + P )

This completes the induction and finishes the proof. 

38
Chapter 9

Applications of Riemann Roch

Armed with the Riemann-Roch theorem I can now discuss some of the many applications
of this theorem to algebraic geometry. The Riemann-Roch theorem answers many of the
underlying questions in regard to linear systems and gives insight to linking topology and
algebraic geometry.

Corollary 9.1. Let C be a smooth projective curve of genus g and let K be a canonical
divisor on C. Then l(K) = g and deg(K) = 2g − 2.

Proof. First, apply the Riemann-Roch theorem (8.2) with the divisor D = 0. Notice that
l(0) = 1 and thus 1 − l(K) = l(D) − l(K − D) = deg(D) − g + 1 = −g + 1. Isolating gives
the first result. Next, for the second result, set D = K and note

g − 1 = l(K) − l(0) = l(D) − l(K − D) = deg(K) − g + 1

Isolating gives the result. 

Corollary 9.2. Let X be a smooth projective curve of genus g and let D ∈ Div (X). Then
if deg(D) ≥ 2g − 1 then l(D) = deg(D) − g + 1.

Proof. Let K be a canonical divisor on X. Corollary (9.1), gives deg(K) = 2g − 2 and

deg(D) ≥ 2g − 1 so deg(K − D) < 0. Therefore, by (5.8), l(K − D) = 0. Then an application
of Riemann-Roch (8.2), yields l(D) = deg(D) − g + 1. 

Proposition 9.3. If D, D’ are two divisors on a variety X, then there is a well-defined map:

µ : L(D) ⊗ L(D0 ) → L(D + D0 )

f ⊗ f 0 7→ f f 0

39
Proof. It suffices to check that if f ∈ L(D) and f 0 ∈ L(D0 ) then f f 0 ∈ L(D + D0 ). Note that
D + div(f ) ≥ 0 and D0 + div(f 0 ) ≥ 0. Using this information, it is easy to see that

D + D0 + div(f ) + div(f 0 ) ≥ 0

Hence the map is well defined. 

Remark 9.4. This map, in general, may not be injective nor surjective. Let X = P so that
divisors over X are simply points. Now, take D = 2[1 : 0] and D0 = 3[1 : 0]. Then by
Riemann-Roch (8.2) and since the canonical divisor of X has degree 2 by (6.43) (and hence
l(K − D) = 0) gives l(D) = 2 and l(D0 ) = 3 so dim(L(D) ⊗ L(D0 )) = 2 · 3 = 6. However,
also by Riemann Roch, l(D + D0 ) = l(5[1 : 0]) = 5 and thus, this map cannot be injective as
you have a 6 dimensional space mapping into a 5 dimensional space.

For a non-surjective example, let X = P. Suppose D, D0 are divisors such that D and
D0 are points (of multiplicity one each) and D + D0 = 0. Then, L(D0 ) = {0} as no rational
function can be added (as divisors) to D0 to get something effective. So the left hand side of
the map is simply the trivial vector space. However, D + D0 = 0 and L(0) 6= {0} thus the
map cannot be surjective.

Remark 9.5. It is imperative to notice the difference between this situation and the case
with sheaves. With sheaves LD ⊗LD0 ∼
= LD+D0 is always true. The difference is with regard to
the tensor. In the case of sheaves, the tensor is over OX where in the case of the proposition,
the tensor is taken over the base field k.

Corollary 9.6. (Plücker’s Formula) Let X be a smooth projective curve of degree d. Then
(d−1)(d−2)
its genus is g = 2
.

Proof. To solve this, I shall construct a differential form whose divisor has degree d(d − 3).
Let p(X, Y, Z) = 0 be the degree d defining equation for the curve. After a potential change
of coordinates, assume that Z = 0 intersects the curve in n distinct points P1 , .., Pn and that
none lie on the line Y = 0. Moreover, without loss of generality, assume that the function
Z
v= Y
gives local parameters at each point so that ord Pi (v) = 1 for each i ∈ {1, ..., n}.

Let U := X\{P1 , ..., Pn }. In affine coordinates (at Z = 1), where (x, y) = ( X , Y ), notice
Z Z

that U is defined by p(x, y, 1) = 0. Moreover, near each Pi by the above comments, one may
take (u, v) = ( X , Z ) as coordinates near each Pi . So take the differential form
Y Y
dx
ω :=
pY (x, y, 1)

40
where the subscript denotes the partial derivative with respect to that variable. Notice that
p(X, Y, Z) = 0 and so,

pX (X, Y, Z)dX + pY (X, Y, Z)dY + pZ (X, Y, Z)dZ = 0

and hence on the affine piece Z = 1, one observes that

pX (x, y, 1)dX + pY (x, y, 1)dY = 0

Note also that on Z = 1 gives

X ZdX − XdZ
dx = d( )= = dX
Z Z2
Similarly, dY = dy and thus, combining the above yields that
dx dy
pX (x, y, 1)dx + pY (x, y, 1)dy = 0 ⇒ ω = =−
pY (x, y, 1) pX (x, y, 1)
valid since no Pi lies on Y = 0. Since X is smooth, notice that X cannot have a point so
that both pY (x, y, 1) and pX (x, y, 1) are both 0. Hence ω has no poles on U. Moreover, since
x and y are local parameters at every point of U, one sees that ω has no zeros on U. Hence
x
div (ω) = 0 on U. Next, note that u = y
and v = y1 . Combining these two facts yields

v d−3 dv ( y1 )d−3 d( y1 ) ( 1 )d−3 ( −dy

deg(pX ) y y2
) −dy
= x 1 = y = =ω
pX (u, 1, v) pX ( y , 1, y ) pX (x, y, 1) pX (x, y, 1)

This shows that ord Pi (ω) = d − 3 for each i. Hence

d
X
div (ω) = (d − 3) Pi
i=1
Xd
⇒ deg(div (ω)) = (d − 3) deg(Pi ) = (d − 3)d
i=1

This culminates to reveal deg(K) = d(d − 3). Now, (9.1) yields that deg(K) = 2g − 2 and so
combining the results reveals that
1 (d − 1)(d − 2)
g = (d2 − 3d + 2) =
2 2
as required. 

Corollary 9.7. (Clifford’s Theorem) Let X be a smooth projective curve of genus g and let
D ∈ Div (X) with K a canonical divisor on X. Suppose that both l(D) and l(K − D) are
nonzero. Then 2l(D) ≤ deg(D) + 2.

41
Proof. By Riemann-Roch (8.2), note that l(D) − l(K − D) = deg(D) − g + 1 so it suffices
to show that l(D) + l(K − D) ≤ g + 1 as then summing the two results gives the desired
conclusion. Next consider the map

|K − D| × |D| → |K|
(E, F ) 7→ E + F

It is clear that this map is onto; take any divisor G in |K| then note that (G − D, D) 7→ G.
Thus dim |K − D| + dim |D| ≤ dim |K|. By (5.6), dim |D| = l(D) − 1 for any divisor D.
Thus, l(K − D) − 1 + l(D) − 1 ≤ l(K) − 1 ⇒ l(D) + l(K − D) ≤ g + 1. This completes the
claim. 

42
References

[1] William Fulton. Algebraic Curves. Advanced Book Classics. Addison-Wesley Publishing
Company Advanced Book Program, Redwood City, CA, 1989. An introduction to alge-
braic geometry, Notes written with the collaboration of Richard Weiss, Reprint of 1969
original.

[2] Robin Hartshorne. Algebraic Geometry. Springer-Verlag, New York, 1977. Graduate
Texts in Mathematics, No. 52.

[3] Marc Hindry and Joseph H. Silverman. Diophantine Geometry, volume 201 of Graduate
Texts in Mathematics. Springer-Verlag, New York, 2000. An introduction.

[4] Rick Miranda. Algebraic Curves and Riemann Surfaces, volume 5 of Graduate Studies in
Mathematics. American Mathematical Society, Providence, RI, 1995.

[5] S. P. Novikov. The Cartan-Serre theorem and inner homologies. Uspehi Mat. Nauk, 21(5
(131)):217–232, 1966.

[6] Kenji Ueno. Algebraic geometry. 3, volume 218 of Translations of Mathematical Mono-
graphs. American Mathematical Society, Providence, RI, 2003. Further study of schemes,
Translated from the 1998 Japanese original by Goro Kato, Iwanami Series in Modern
Mathematics.

43