Algebraic Geometry: The Riemann-Roch Theorem for

Curves and Surfaces

Laurens Walleghem

July 1, 2019

Contents
1 Introduction 2

2 Riemann-Roch theorem for curves 2

2.1 Arithmetic genus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Divisors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.3 Invertible sheaves and linear systems . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Differential forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.5 Riemann-Roch theorem for curves . . . . . . . . . . . . . . . . . . . . . . . . 10

3 Riemann-Roch theorem for surfaces 11

3.1 Intersection multiplicity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Cohomology and superabundance . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.3 Riemann-Roch theorem for surfaces . . . . . . . . . . . . . . . . . . . . . . . 13

1
1 Introduction
The Riemann-Roch theorem is interesting in the sense that it combines algebraic geometry,
topology and complex analysis. It formulates a connection between the genus of a curve or
surface and the rational functions defined on a curve or surface. It can be used for example
in the classification of nonsingular projective curves.

In this work we use the definitions and conventions as in [Har77]. The Riemann-Roch
theorem can be given in terms of general schemes and sheaves, but we will only be concerned
with the ‘classical’ projective varieties as defined in part I of [Har77]. Our main concern here
will be to give examples of classical varieties for which the Riemann-Roch theorem holds,
and what the different elements of the theorem are. More specifically, we will look at the
Riemann-Roch theorem for classical curves and surfaces. Some results or definitions will still
involve schemes or sheaves, but in that case we show what the concepts and results imply for
classical projective varieties. Unless stated otherwise, we will always assume a variety to be
projective, irreducible and nonsingular as defined in the classical sense in section I of [Har77].
Thus by a curve or surface we mean a projective nonsingular curve or surface. If we talk
about a point, a closed point is meant, not a generic point as can appear in schemes, unless
stated otherwise. Thus, even though it is not always necessary in our results and definitions,
we will assume varieties to be defined over an algebraically closed field k.

2 Riemann-Roch theorem for curves

2.1 Arithmetic genus
Definition 2.1.1. Let X be a projective variety variety. The degree of a projective variety
and the Hilbert polynomial were defined in I.7 in [Har77]. We can define the arithmetic
genus pa (X) of a variety X ([Har77], I. ) using the Hilbert polynomial PX by the formula
pa (X) = (−1)dim X (PX (0) − 1).
Remark 2.1.2. It can be proved that the arithmetic genus is an invariant under isomorphism
([Har77], Ex III.5.3). Actually, it is even a biration invariantfor most varieties, for example
for curves and surfaces ([Har77], V. 5.6.1).
Example 2.1.3. From the proof of Propostion I.7.6 in [Har77] we know that the Hilbert
polynomial of Pn is given by PPn (z) = z+n


n . Thus we have that
 
n
pa (Pn ) = (−1)n (PPn (0) − 1) = (−1)n ( − 1) = 0.
n
Example 2.1.4. For a hypersurface H ⊂ Pn of degree d, we have from Propostion I.7.6 in
[Har77] that the Hilbert polynomial is given by
   
z+n z−d+n
PH (z) = − .
n n
Hence the arithmetic genus is given by
        
n−1 n n−d n n−d d−1
pa (H) = (−1) − − 1 = (−1) = .
n n n n

2
Thus in P2 , a curve C of degree d has arithmetic genus equal to
1
pa (C) = (d − 1)(d − 2).
2
Remark 2.1.5. There is an interesting fact about curves in Pn over algebraically closed
fields k. A Riemann surace is a complex manifold of complex dimension 1. It can also be
seen as a non-singular projective curve in P2 over C. But as a real manifold, such a Riemann
surface has dimension 2, making it a real surface. For such orientable surfaces we have a
notion for the genus as the number of handles of the surface, and the notions of genus and
arithmetic genus are the same for such Riemann surfaces.

2.2 Divisors
Let us first explain some intuition behind divisors on curves. Let C be a nonsingular projective
curve in P2 which isPnot a line, and let L be a line in P2 . Then L∩C consists of a finite
P number
of points, L ∩ C = i ni Pi where ni are the multiplicities. We then say that D = ni Pi is a
divisor. As L varies, we get other intersection points and thus other divisors. If we let L vary
over all lines in P2 , we could reconstruct C from all these divisors. These divisors are said to
form a linear system. Now let D = L ∩ C and D0 = L0 ∩ C be two divisors, where L and L0
are defined by the equations f = 0 respectively f 0 = 0. Then g = f /f 0 is a rational function
on P2 which has ‘poles’ at the points of D0 and zeros at the points of D. The divisors D and
D0 are then said to be linearly equivalent, and the existence of such a rational function g
which has poles at D0 and zeros at D will be taken as the definition of linear equivalence of
divisors. The exact definition of a divisor is the following, as given in II.6 in [Har77].
Definition 2.2.1. Let X be projective non-singular variety. A prime divisor on X is a
closed subvariety of codimension one. Let Div X be the free abelian group generated by the
prime divisors; anP element in Div X is then called a (Weil) divisor. So we can write a Weil
divisor D as D = i ni Yi for integers ni , where we sum P over all the prime divisors Yi on X
and only finitely many ni are non-zero. A divisor D = i ni Yi is said to be effective if all
ni ≥ 0.
For a prime divisor Y on X and a generic point η we know that the local ring Oη,X is a
discrete valuation ring with quotient field equal to the function field K on X. Denote the
corresponding valuation by vY and let K ∗ denote the multiplicative group of K. Then for f
a non-zero rational function we have that vY (f ) is an integer, and we say that f has a zero
of order vY (f ) along Y if vY (f ) > 0, and we say that f has a pole of order vY (f ) along Y if
vY (f ) < 0.
Definition 2.2.2. For a function f ∈ K ∗ we define the divisor (f ) of f by
X
(f ) = vY (f ) · Y.
Y

It can be seen that this sum over all prime divisors is actually a finite sum. A principal
divisor is a divisor which is equal to the divisor of a function.
Definition 2.2.3. We define an equivalence relation on the divisors on X by saying that D
and D0 are linearly equivalent if their difference is a principal divisor. The equivalence class
of a divisor is then called a divisor class and the set of all equivalence classes is then called
the divisor class group and is denoted by Cl(X).

3
Remark 2.2.4. ([Sha13a], p. 148) If X be a classical projective curve, then the valuation
is the following. Let g be a polynomial on an open set of X and take an arbitrary point
p ∈ X. As the ring Op is a local ring, there is a unique maximal ideal m which is principal by
Theorem 6.2A in [Har77], so we have m =< m > for some element m ∈ Op . Then vP (g) is
defined to be the greatest natural natural number l such that g ∈ ml . For a rational function,
we define then vP (g1 /g2 )vP (g1 ) − vP (g2 ). The fact that the valuation looks like this can
be seen from the fact that the ring Op can be recovered as all rational functions f in the
function field K which satisfy vP (f ) ≥ 0 and because of the fact that a prime divisor P is
completely determined by the valuation vP .
Remark 2.2.5. For a higher dimensional classical projective variety X we can do something
analogous. Let Y be a prime divisor and let f ∈ K be a rational function. Then vY (f ) is
equal to the length of the module OX (Y )/(f ), where OX (Y ) denote the regular functions on
X ∩ Y = Y . The length is always finite. The rest of the procedure is exactly the same as in
the previous remark.
DefinitionP2.2.6. Let X be a nonsingular projective curve, then a divisor on X is a finite
sum PD = i ni Pi of points Pi on X. We then define the degree of the divisor D to be
D = i ni . Thus the degree of a principal divisor is zero, deg(g) = 0 (where g is a rational
function), so that the degrees of any two divisors in the same divisor class are the same and
we can talk about the degree of a divisor class.
Remark 2.2.7. On X = Pn we can also P define the degree of a divisor on X. A divisor
D on X is given by a finite sum D = i ni Yi of hypersurfaces Yi ⊂ Pn . The degree of a
hypersurface Yi can be defined using the Hilbert polynomial as in Section I.7 of [Har77], and
there one also proves that if Yi is a hypersurface whose ideal is generated by a homogeneous
degree di polynomial, then deg Yi = di . The degree of the divisor D is then defined as
X
deg D = ni deg Yi .
i

Let H be the hyperplane x0 = 0. If D is a divisor of degree D, then D ∼ dH, and for any
rational function we have deg(f ) = 0. Moreover, the degree funcion deg : Cl → Z is an
isomorphism of groups. For a proof of this, see II.6.4 in [Har77].
Remark 2.2.8. The degree of a principal divisor on a projective nonsingular curve X is
always equal to zero. This implies that the morphism deg : ClX → Z is surjective ([Har77],
Cor. II.6.10).
Example 2.2.9. Let X = P1 . A prime divisor is then a point of P1 , and an effective divisor
is a finite sum of points times a multiplicity. Any divisor can be written as D = d1 − d2 for
two effective divisors d1 , d2 . The valuation vP (f ) of a rational function f ∈ K ∗ at a point of
P1 is then just the maximal integer l such that f ∈ (ml ) where m = (m) is the maximal ideal
of Op . So for example, at P = [0, 1] we have that

x x2
vP ( ) = 1, vP ( ) = 2,
y y2
x − 2y y
vP ( ) = 0, vP ( ) = −1,
y x
y x
so the function x has a pole at P , the function y has a zero of order 1 at P and the function
2
x
y2 has a zero of order 2 at P . The divisor class group Cl(P1 ) is isomorphic to Z by the

4
isomorphism
deg : Cl(P1 ) → Z : D → deg(D).
As the degree deg is independent of the representant of an equivalence class in Cl(P1 ), this
function deg is well-defined,
PN and obviously it is a surjective morphism.
PN To prove that it is
injective, let D = i=1 ni Pi be a divisor with deg(D) = 0, that is, i=1 ni = 0. Let now Pi
have coordinates [ai : bi ] in P1 . Consider the function
N
Y
g= (bi x − ai y)ni ,
i=1

as deg(D) = 0, this is a rational function and thus an element of the function field. Also, by
the expression for the valuation vP above it is clear that
  
ni if P = Pi ,
vP (bi x − ai y)ni =
0 if P 6= Pi ,
so that we have (g) = D. Hence deg(D) = (g) = 0, so that deg is indeed injective. The same
technique can be applied for Pn to see that Cl Pn = Z ([Har77], Prop. II.6.4).
Example 2.2.10. ([Har77], Example II.6.6.1) Consider the non-singular quadric surface Q
in P3 given by the equation xy = zw. As we know, Q is isomorphic with P1 × P1 (Exercise
I.2.15 in [Har77]). (Note that P1 × P1 6= P2 .) We claim that
Cl Q ∼
= Z ⊕ Z.
Denote the projections of Q onto P1 by p1 and p2 . Let us have a look at the homomorphisms
p∗1 and p∗2 defined by
X X
p∗1 : Cl P1 → Cl Q : D = ni P i → ni p−1
1 (Pi ),

and analogously we can define p∗2 . Let us now prove that p∗1 is injective (the proof that p∗2
is injective is analagous). Consider the closed subvariety Y = P × P1 of Q for a projective
point P ∈ P1 . Then Q − Y = A1 × P1 and the composition map
p∗
Cl P1 →
1
Cl Q → Cl(A1 × P1 ),
where the last map is the restriction map, is an isomorphism (see for example the proof of
II.6.6 in [Har77]). By Prop II.6.5 in [Har77], we have an exact sequence
Z → Cl Q → Cl(A1 × P1 ),
where the first map is obtained by sending 1 to Y . As Cl P1 = Z (see the previous example),
we get an exact sequence
p∗
0 → Cl P1 →
1
Cl Q → Cl(A1 × P1 ),
which means that
Cl Q ∼
= Cl P1 ⊕ Cl (A1 × P1 ) = Z ⊕ Z,
as Cl (A1 × P1 ) ∼
= Cl P1 = Z by Proposition II.6.6 in [Har77]. A divisor on the quadric surface
Q is then said to be of type (a, b) ∈ Z ⊕ Z if it belongs to the divisor class corresponding to
(a, b).

5
2.3 Invertible sheaves and linear systems
Associated to divisors, we have invertible sheaves on the ringed space. We will see that we can
see the latter as finite-dimensional vector spaces, whose dimension is an important concept
in the Riemann-Roch theorem. Under the right conditions, we will have an isomorphism
between the class group and the group of invertible sheaves. [Per08], p 142
Definition 2.3.1. Let (X, OX ) be a ringed space. A sheaf F on X of OX -modules
 is said
to be locally free of rank n if X can be covered by open sets U for which F U = OX (U )n .
An invertible sheaf on a ringed space is a locally free OX -module of rank 1.
Remark 2.3.2. Recall that for our purposes in this paper, we can assume X to be a classical
variety, and OX is then a sheaf which gives at every open set U of X the ring OX (U )
consisting of all regular functions on U ⊂ X. A sheaf of modules is said to be locally free if
there exists a cover of X by open sets U such that F|U is a free OX (U )-module, that is, if
F|U = OX (U )nU for some nU ∈ N, for all open sets U in the cover.
Remark 2.3.3. The space of invertible sheaves on a ringed space X inherits a group structure
by the tensor product ⊗ as group operation; as OX ⊗OX OX = OX , we have that if F and G
are invertible sheaves, then F ⊗OX G is again an invertible sheaf. The neutral element is OX
and the inverse of an invertible sheaf L is given by the dual sheaf Hom(L, OX ), as

L ⊗OX Hom(L, OX ) ∼
= Hom(L, L) = OX .

(This is proposition 6.12 in [Har77].) The group of all invertible sheaves on X up to

isomorphism, is called the Picard group and denoted by PicX.
Lemma 2.3.4. ([Har77], Cor. II.6.16) If X is a nonsingular projective variety, then the
map
Cl X → Pic X : D 7→ L(D)
is an isomorphism of groups.
Definition 2.3.5. ([Per08], Sec. VIII.c) On a projective variety X with function field K,
for a divisor D we have an associated invertible sheaf on X

L(D) = { f ∈ K | D + (f ) ≥ 0 } .
P
The condition D + (f ) ≥ 0 means vY (f ) ≥ −nY,D where D = Y nY,D Y . It turns out that
this is a vector space over k and that L(D) and L(D0 ) are isomorphic as k-vector space if
D ∼ D0 . We define l(D) = dimk L(D). Sometimes L(D) is also denoted by Γ(X, L(D)).
Example 2.3.6. Consider the projective space X = P1 . Let P be the point [1, 0]. For the
divisor D = P we have that L(D) = { f ∈ K | P + (f ) ≥ 0 }, so L(D) consists of all the
rational functions f = g/y for degree 1 polynomials g ∈ k[x, y]. Thus as a k-vector space, a
basis for L(D) is y/y = 1, x/y so that l(D) = 2. If we would take the divisor D0 = 2P , then
L(D0 ) consists of all the rational functions f = g/y 2 for degree 2 polynomials g ∈ k[x, y].
Thus as a k-vector space, a basis is x2 /y 2 , y 2 /y 2 = 1, xy/y 2 = x/y so that l(D0 ) = 3.
Example 2.3.7. Let us generalise the previous example. Take X = Pn and let D be a
divisor of degree d, then D ∼ dH where H is the hyperplane x0 = 0 (see Remark 2.2.7). Thus
L(D) consists of all rational functions f /xd0 where f ∈ k[x0 , x1 , · · · , xn ] is a homogeneous
degree d polynomial, so that L(D) is spanned as a k-vector space by rational functions of the
form m/xd0 , where m is a monomial of degree d. Thus l(D) = dimk L(D) = n+d


n .

6
Definition 2.3.8. ([Har77], Def. p. 157) Let X be a nonsingular projective variety, and let
D0 be a divisor on X. The set containing all effecive divisors linearly equivalent to D0 is
the called the complete linear system associated to D0 and is denoted by |D0 |. There is a
1 : 1 correspondence between |D0 | and the set (Γ(X, L(D0 )) − 0)/k ∗ by sending a non-zero
s ∈ Γ(X, L(D0 )) to the class containing the divisor of zeros of (s) (see also Prop II.7.7 in
[Har77]). By the correspondence above, a linear sstem has the structure of closed points of a
projective space, and has dimension dim |D0 | = dimk Γ(X, L(D0 )) − 1.
Example 2.3.9. ([Har77], Example II.7.8.3) Let X = Pn , then the complete linear system
|D| consisting of all effective divisors linearly
equivalent to the divisor D = dY0 (here Y0 is
the hypersurface x0 = 0) has dimension n+d n − 1, as the vector space Γ(X, L(D0 )) consists
of all rational functions f
/xd0 for homogeneous degree d polynomials f ∈ k[x0 , x1 , · · · , xn ],
which has dimension n+d n .

2.4 Differential forms

This part is closer to differential geometry and hence is more intuitive. For a variety X, we
will define the sheaf ΩX which can be thought of as the cotangent space, and we will define
the associated canonical sheaf ωX . Basic Algebraic Geometry I, p. 190
Definition 2.4.1. Let B be an A-algebra and M a B-module. An A-derivation of B into
M is a map d : B → M such that

(1)d(b + b0 ) = db + db0 ,
(2)d(bb0 ) = b0 db + bdb0 ,
(3)da = 0,

for all a ∈ A, b, b0 ∈ B. The module of relative differential forms of B over A is defined by a

B-module ΩB/A and an A-derivation d : B → ΩB/A which satisfies the universal property
that for every an derivation d0 of B over A there exists a unique B-module homomorphism f
such that d0 = f ◦ d.
Remark 2.4.2. Let f : B ⊗A B → B be the diagonal homomorphism for which f (bb0 ) = b⊗b0
and define I = ker(f ). Then (I/I 2 , d) is a module of relative differential forms, where
d : B → I/I 2 satisfies db = 1 ⊗ b − b ⊗ 1. This is the inspiration of the general definition of
relative differentials on schemes. definition.

Definition 2.4.3. ([Har77], Remark II.8.9.2) Let X be a variety. Then the sheaf of relative
differentials of X over k is defined as follows. We cover X by affine open subsets V , then
we define the sheaf of relative differentials of V over k by ΩV /k = Ω̃B/k . The sheaf of
relative differentials of X over k, denoted by ΩX/k is then obtained by glueing the sheaves
Ω̃V /k together, with derivation d : OX → ΩX/k obtained by glueing together the derivations
d : V → ΩV /k . Often we will use the notation ΩX instead of ΩX/k .

Remark 2.4.4. Here the notation Ω̃V /k is the notation from the definition on p. 110 in
[Har77]. The space ΩV /k is a module over the affine coordinate ring of V and Ω̃V /k is then
the sheaf of modules associated to ΩV /k . By Proposition II.5.1 in [Har77] it is then an
OV -module.

7
Remark 2.4.5. More generally, for a morphism f : X → Y between general schemes, the
sheaf of relative differentials, can be defined in terms of the sheaf I/I 2 for an appriopriate
sheaf of ideals I. Notice the similarity with Remark 2.4.2.
Remark 2.4.6. For any nonsingular variety X of dimension n the sheaf ΩX is a locally free
sheaf of rank n ([Har77], Theorem II.8.15).
Remark 2.4.7. On affine n-space X = An we have coordinates x1 , · · · , xn . Then ΩX is a
free OX -module of constant rank n and its generators are the global sections dx1 , · · · , dxn .
The differential associated to a function f ∈ k[x1 , · · · , xn ] is then defined by
X ∂f
df = dxi .
i
∂x i

Definition 2.4.8. ([Har77], Def p. 180) Let X be a nonsingular variety of dimension n,

then we define the tangent sheaf of X by TX = Hom
Vn OX (ΩX , OX ). It is locally free of rank n.
The canonical sheaf of X is defined to be ωX = ΩX .
Remark 2.4.9. Notice that ωX an invertible sheaf on X, so for every open U ⊂ X it can
be generated as an OX (U ) module by one element αU ∈ ωX (U ).
Remark 2.4.10. The geometric genus of a projective nonsingular variety X is defined to
be pg = dimk Γ(X, ωX ), which turns out to be a birational invariant ([Har77], Th. II.8.19).
For projective nonsingular curves, it is the same as the arithmetic genus pa defined earlier
([Har77], III.7.12.2).
On a n-dimensional variety, we can also associate a divisor class to differential n-forms,
the canonical divisor class.
Definition 2.4.11. ([Sha13a], p. 204-205) Let X be a nonsingular variety of dimension n
and let α be a n-form, that is, an element of ωX . If (Ui )i is an affine cover of X, then we can
write on every Ui the n-form α as

αUi = gi dx1 ∧ · · · ∧ dxn ,
for some function gi on Ui . Now if Ui ∩ Uj is nonempty, then the functions gi and gj are
related by a Jacobian of coordinate transformations. As the Jacobian is regular and non-zero
on Ui ∩ Uj , we have that gi /gj is regular and non-zero on Ui ∩ Uj , which means that the
divisor (gi /gj ) is zero on Ui , meaning that we can associate a divisor (gi ) to gi and thus to
α. Such a divisor is called a canonical divisor. As every two n-forms on X are related by a
rational function f on X, all canonical divisors are linearly equivalent, so that they form a
single class of divisors, called the canonical class.
Example 2.4.12. Let X = P1 and let U0 = {(x, 1) ∈ P1 } U1 = {(1, y) ∈ P1 }. Then
U0 , U1 is an open cover of X. We can write a one-form on U0 , as f (x)dx and on U1 as
g(y)dy = −g(1/x)1/x2 dx. Thus on U0 ∩ U1 , where we have y = 1/x we must have that
1
f (x)dx = g(y)dy = −g(1/x) dx.
x2
Thus we must have that x2 f (x) + g(1/x) = 0. A solution to this is
1
f (x) = , g(y) = 1.
x2
The canonical class is thus equal to [−2(0, 1)] (here (0, 1) = (0, t) is a projective point in P1 .

8
Example 2.4.13. Let us generalise the previous example a bit. Consider the projective
n-space X = Pn with projective coordinates (x0 , x1 , · · · , xn ) and consider the affine cover
Ui = { (x0 , · · · , xn ) | xi 6= 0 } for i = 0, · · · , n. Define on the affine subset subset U0 the
n-form α by
1
α= dy1 ∧ · · · ∧ dyn
y1 y2 · · · yn
in terms of the affine coordinates yi = xx0i on U0 , and define α analogously on U1 , · · · , Un .
Then this one-form has simple poles at x0 = 0, x1 = 0, · · · , xn = 0. Thus the divisor
corresponding to this one-form is the divisor −Y0 − Y1 − · · · − Yn , where Yi is the hypersurface
xi = 0 in Pn . This divisor is linearly equivalent to the divisor −(n + 1)Y0 as their difference
is the divisor of the rational function x0 xx1n···xn . Thus the canonical class of Pn is the class
0
[−(n + 1)Y0 ].
Example 2.4.14. Consider the quadric surface Q in P3 , given by the equation xy = zw.
We know that Q is isomorphic to P1 × P1 and from the previous example we know that the
canonical class of P1 is [−2P ] for a projective point P in P1 . Also, the canonical sheaf of a
product of two nonsingular varieties X and Y is given by the tensor product of the canonical
sheaves ([Har77], Exercise II.8.3): ωX×Y = p∗ ωX ⊗ q ∗ ωY , where p and q are the projections
on X respectively Y , and p∗ , q ∗ are then as in Example 2.2.10. Thus if α1 , α2 are elements of
ωP1 and are given by α1 = f1 (y1 )dy1 , α2 = f2 (y2 )dy2 on an open affine set, then
f1 (y1 )g1 (y2 )dy1 ∧ dy2
is an 2-form in ωP1 ×P1 , given on an open affine set. Thus a canonical divisor of P1 × P1 is
then given by
K = −2(P × P1 ) − 2(P1 × P 0 )
for projective points P, P 0 ∈ P1 . Thus the canonical divisor K of Q is of the type (−2, −2).
Example 2.4.15. Consider the conic C in P2 given by the 2-uple embedding
ϕ : P1 → P2 : (x, y) → (x2 , y 2 , xy).
The equation determining this conic is thus x22 = x0 x1 in P2 . From Ex I.3.4 in [Har77] we
know that the conic C is isomorphic with P1 . In the example above, we have seen that a
canonical divisor is then equal to −2(0, y), for the projective point (0, y) ∈ P1 , corresponding
to the divisor −2(0, y 2 , 0) on the conic C. Thus [−2(0, 1, 0)] is the canonical class on C in P2 .
Example 2.4.16. Consider the Veronese surface V in P5 . It can be seen as the 2-uple
embedding of P2 in P5 ,
ψ : P2 → P5 : (x0 , x1 , x2 ) 7→ (x20 , x0 x1 , x0 x2 , x21 , x1 x2 , x22 )
and it turns out that by this embedding P2 is an isomorphism onto its image ([Har77], Ex
I.3.4). This means that
Cl V ∼
= Cl P2 ∼= Z.
Thus a divisor is totally determined by its degree. The canonical class on V is then the
canonical class on P2 , which was equal to [K] = [−3L] for a line L in P2 . So if we take
for example
 the line x0 = 0in P2 , this corresponds by the 2-uple isomorphism to the conic
C = (0, 0, 0, x21 , x1 x2 , x22 )  x1 , x2 ∈ k, (x1 , x2 ) 6= (0, 0) on the Veronese surface V ⊂ P5 .
Thus the canonical class on the Veronese surface is [−3C] where C is the conic on the Veronese
surface as described above.

9
2.5 Riemann-Roch theorem for curves
Now we finally have established all the necessary concepts, we can look at the Riemann-Roch
theorem of curves. Recall that for a curve we had that the arithmetic genus pa and the
geometric genus pg were equal, so that we can speak about the genus g of a curve.

Theorem 2.5.1. (Riemann-Roch for curves) Let X be a curve of degree d of genus g,

let D be a divisor on X and denote the canononical divisor by K. Then

l(D) − l(K − D) = deg D + 1 − g. (1)

Example 2.5.2. ([Har77], example 1.3.3) On a projective nonsingular curve X the geometric
genus and the arithmetic genus coincide, so that we have

pa (X) = g = pg (X) = dimk Γ(X, ΩX ) = dimk Γ(X, L(K)) = l(K).

Moreover, note that l(0) = 1, as L(0) consists of all rational functions f without poles, so
L(0) ∼
= k. Using this in the Riemann-Roch theorem for curves with D = K, we get that

deg K = l(K) − l(0) + g − 1 = 2g − 2.

Example 2.5.3. ([Har77], Example IV.1.3.5) By Corollary I.6.11 in [Har77] we know that
every abstract curve is birationally equivalent to a nonsingular projective curve, and by
Corollary I.6.12 in [Har77] the categories of nonsingular projective curves and dominant
morphisms and of quasi-projective curves and dominant rational maps are equivalent. As
we are only working with nonsingular projective curves here, a curve X is rational (that is,
birational to P1 ) if and only if X is isomorphic with P1 . We will show that actually a curve
(nonsingular and projective) is rational if and only if it has genus zero. We already know
from Example 2.1.3 that pa (P1 ) = 0. So suppose that X is a curve with genus 0. We know
from the example above that deg K = 2g − 2 = −2. Let P, Q be two distinct points in X
and define the divisor D by D = P − Q, so that deg(D) = 0 and deg(K − D) = −2, which
implies that l(K − D) = 0. Thus applying Riemann-Roch gives that

l(D) = 1.

Thus the complete linear system |D| is nonempty, and D is thus linearly equivalent to
some effective divisor, thus deg D ≥ 0. Now as D has degree zero, this means that D
must be equivalent to an effective divisor with degree zero. There is only one such effective
divisor, namely the zero diviosr. Hence we have D ∼ 0, so that P ∼ Q. Thus there is a
non-zero rational function f ∈ K such that (f ) = P − Q. Now we have the inclusion of
fields i : k(f ) → K, and we will use to construct a birational morphism φ : X → P1 as in
Example II.6.10.1 of [Har77], so that indeed X will be a rational curve. To do so, we use
the proofs of Corollary I.6.12, Proposition I.6.8, Theorem I.4.4 and Theorem I.3.5 in [Har77].
Let i : k(f ) → K be the inclusion homomorphism. By Corollary I.6.12, we can associate
to every function field a curve, and a homomorphism between the function fields induces
a rational map of the corresponding curves. In that way we get a map φ : U ⊂ X → P1 ,
where U ⊂ X is an open set. Finally, using Proposition I.6.8, we can (uniquely) extend the
morphism φ : U → P1 to get a morphism φ : X → P1 . It turns out that this morphism is a
degree 1 morphism, (see Example II.6.10.1), which means that the [k(f ) : K] = 1, and thus
by the correspondence of Corollary I.6.12 we can conclude that X and P1 are birational.

10
Example 2.5.4. Let us apply Riemann-Roch to a specific curve. Suppose we have a
nonsingular projective curve C in P2 of degree 3. Then the genus of the curve is equal to 1
by Example 2.1.4). Thus by Example 2.5.2, we have that deg K = 2g − 2 = 0. Let D now be
a divisor on C of degree d > 0. Then deg(K − D) = −d, and thus l(K − D) = 0. Applying
Riemann-Roch, we then get that
l(D) = deg D = d.

3 Riemann-Roch theorem for surfaces

3.1 Intersection multiplicity
In I.7 of [Har77] we have already seen some intersection theory, but we need to expand this
theory just a bit in order to formulate the Riemann-Roch theorem for surfaces. In this section
we always assume X to be a projective nonsingular surface over an algebraically closed field
k, unless stated otherwise. We say that two curves meet transversally at the the point P if
the local polynomials f, g of C and D at P generates the maxmimal ideal mP of OP , X.
Theorem 3.1.1. ([Har77], Th V.1.1)) There is a unique function (called the intersection
pairing) Div × Div → Z : (C, D) → C.D which only depends on the linear equivalence classes,
is symmetric and additive and gives C.D = #(C ∩ D) if C and D are nonsingular curves
meeting transversally.
A useful way to calculate the intersection multiplicity is by using the following propostion.
Propostion 3.1.2. ([Har77], Prop. V.1.4) Let C and D be curves on X having no common
irreducible compononent. Let P be an arbitary point in X and suppose that C and D are
determined by the local equatios f and g at P . Then
X
C.D = (C.D)P ,
P ∈C∩D

where the intersection multiplicity (C.D)P is defined as the dimension of OP,X /(f, g) as a
k-vector space.
2 2
Example P 3.1.3. Let Y be a curve of degree d in P and let L be a line in P . Then
(L.Y ) = P (L.Y )P = d (Exercise 5.4 in [Har77]).
Remark 3.1.4. For a divisor D on a variety X, we name the number K 2 = K.K the
self-intersection number.
Example 3.1.5. ([Har77], Example V.1.4.2) Take X = P2 , as PicX = Z we can take the
class h of a line as a generator. Let h0 be another line. As any two lines are equivalent, we
have
h2 = h.h = h.h0 = 1,
as any two distinct lines meet transversally in one point. (Note how we exploit the properties
of the intersection pairing.) Let C and D now be curves in X of degrees n respecively m.
By Prop. II.6.4 in [Har77], we know that then C ∼ nH and D ∼ mh. As the intersection
pairing is additive, we then have that
C.D = (nh).(mh) = nmh2 = nm.

11
Example 3.1.6. ([Har77], Example V.1.4.3) Let Q be the nonsingular quadric surface in P3
given by the equation xy = zw. As we have seen in Example 2.2.10, we know that Cl Q ∼ = Z⊕Z,
and as X satisfies the conditions of Lemma 2.3.4, we have that Pic Q = Cl Q ∼ = Z ⊕ Z.
To generate this group, we can take lines l of type (1, 0) and lines m of type (0, 1). From
Example 2.2.10 we see that l then corresponds to a line Pl × P1 ⊂ P1 × P1 ∼ = Q and that m
corresponds to a line Pm × P1 , where Pl and Pm are projective points in P1 . Such two lines l
and m meet each other transversally in one point, so l.m = 1. Two distinct lines l and l0 in
the same class (1, 0) obviously do not meet each other as they correspond to the lines Pl × P1
respectively Pl0 × P1 for two distinct projective points Pl , Pl0 ∈ P1 , so we have l2 = l.l0 = 0,
and similarly, m2 = 0. Thus for any two curves C of type (a, b) and D of type (a0 , b0 ) on X,
we have
C.D = (a(1, 0) + b(0, 1)).(a0 (1, 0) + b0 (0, 1)) = ab0 + a0 b.
Example 3.1.7. ([Har77], Example V.1.4.4) We have seen that, on a variety X, there is only
one linear equivalence class of divisors corresponding to the canoncial sheaf ωX and we denote
a divisor in this linear equivalence class by K. The self-intersection K 2 = K.K of a canonical
divisor K then only depends on X. Take for example X = P2 . From Example 2.4.13 we have
that K = −3h for a hyperplane h, so that K 2 = 9. For the quadric surface Q in P3 , given by
the equation xy = zw, we know from Example 2.4.14 that the canonical divisor K has type
(−2, −2), so that by Example 2.2.10 we have that K 2 = (−2)(−2) + (−2)(−2) = 8.

3.2 Cohomology and superabundance

Now that we have introduced the intersection number C.D we need a final notion before
stating the Riemann-Roch theorem for surfaces, that of superabundance. To state it, we need
cohomology on a variety, so let us say a word about the basics of cohomology on a variety.
For a general scheme, the cohomology can be defined using right derived functors, but for
varieties we will discuss, these cohomologies coincide with the Čech cohomology, which is
often easier to work with. We will introduce the cohomology in this way here, based on III.4
in [Har77].
Definition 3.2.1. Let U = (Ui )i∈I be an open cover of a variety X and let F be sheaf of
abelian groups on X. For all p ≥ 0 we define
Y
C p (U, F) = F(Ui0 ∩ · · · ∩ Uip ).
i0 <···<ip

Next we define maps d : C p → C p+1 by

p+1
X
(−1)k αi0 ,··· ,iˆk ,··· ,ip+1 Ui

(dα)i0 ,··· ,ip = ∩···∩Uip+1
,
0
k=0

where iˆk means omit ik . As we have d2 = 0, we have indeed defined a complex C (U, F) of
abelian groups. The Čech cohomology groups of F with respect to the covering U is then
defined to be
Ȟ p (U, F) = hp (C (U, F)) = ker dp /imdp−1 ,
where dp is the map dp : C p (U, F) → C p+1 (U, F) in the complex.

12
Theorem 3.2.2. ([Har77], Th. III.4.5) For a noetherian separated scheme X, an open
affine cover U of X and a quasi-coherent sheaf F on X, the Čech Cohomology coincides with
the general cohomology on schemes as defined in III.1 of [Har77].
Remark 3.2.3. The cohomology groups Ȟ p (U, F) will be denoted by H p (X, F) if the
conditions of the above theorem are satisfied.
Remark 3.2.4. Notice that we always have the equality Ȟ 0 (U, F) ∼ = Γ(X, F) = F(X). An
element s is in H 0 (X, F) if and only if it is in the kernel of the first morphism d : C 0 (U, F) →
C 1 (U, F). But the latter the just means that we can write 0 = ds = si − sj on Ui ∩ Uj (for
all i, j), so that s defines a global section on X ([Per08], Prop. VII 2.3).
Having introduced cohomology, we can define the superabundance of a divisor on a surface.
Definition 3.2.5. Let D be a divisor on X. The superabundance s(D) is defined as

s(D) = dim H 1 (X, L(D)).

3.3 Riemann-Roch theorem for surfaces

Having introduced the superabundance and the intersection pairing, we finally are ready to
state the Riemann-Roch theorem for surfaces.
Theorem 3.3.1. (Riemann-Roch for surfaces) Let D be an arbitrary divisor on a surface
X, and denote the canonical divisor by K. Then
1
l(D) − s(D) + l(K − D) = D.(D − K) + 1 + pa (X).
2
Example 3.3.2. THIS EXAMPLE HAS TO BE CHECKED! Consider the non-singular
quadric surface Q given by xy = zw in P3 . This surface is isomophic to P1 × P1 (see
Exercise I.2.15 in
[Har77]). As this is a degree 2 surface in P3 , we know from Example 2.1.4
that pa (Q) = 13 = 0. From Example 2.2.10 we know that the divisor class group of Q is
isomorphic to Z2 , that is Cl Q ∼= Z ⊕ Z, and by Example 2.4.14 we know that the canonical
divisor K is of type (−2, −2). In Example 3.1.6 is explained how the intersection pairing of
the quadric surface works. Using this information, we can already work out the right part of
the Riemann-Roch equation for surfaces. Let D be a divisor on Q of type (a, b), then
1 1 1
D.(D − K) + 1 + pa (Q) = (a, b).(a + 2, b + 2) + 1 = (a(b + 2) + (a + 2)b) + 1
2 2 2
= ab + a + b + 1.

Now let us investigate the left side of the Riemann-Roch equation. Let D be a divisor of
type (a, b), thus given by the union la ∪ mb for the lines l = P1 × P, m = Q × P1 for some
points P, Q ∈ P1 . Then L(D) consists of all rational functions f on P1 × P1 such that
(f ) ≥ −al − bm. Let the coordinates on P1 × P1 be given by (x0 , x1 ) × (y0 , y1 ). Thus the space
L(D) is generated by all rational functions of the form F/xa0 · G/y b

0 for
F (x0 , x1 ), G(y0 , y1 )
a+1 b+1
homogeneous of degree a and b, thus l(D) = dimk L(D) = 1 1 = (a + 1)(b + 1) if
a ≥ 0, b ≥ 0 and l(D) = 0 if a < 0 or b < 0. Then K − D is of type (−2 − a, −2 − b) and
thus l(K − D) = (−1 − a)(−1 − b) = (a + 1)(b + 1) if a ≤ −1, b ≤ −1 and l(K − D) = 0 if

13
a > −1 or b > −1. Thus we get that the superabundance s(D) of a divisor D of type (a, b)
with a ≥ 0, b ≥ 0 is equal to
1
s(D) = − D.(D − K) − 1 − pa (Q) + l(D) + l(K − D)
2
= −ab − a − b − 1 − 0 + (a + 1)(b + 1) + 0 = 0.
Similarly, if a < −1, b < −1 we have that
1
s(D) = − D.(D − K) − 1 − pa (Q) + l(D) + l(K − D)
2
= −ab − a − b − 1 − 0 + 0 + (a + 1)(b + 1) = 0.
and if (a, b) are none of the previous cases, then we get
1
s(D) = − D.(D − K) − 1 − pa (Q) + l(D) + l(K − D)
2
= −ab − a − b − 1 + 0 = −ab − a − b − 1.
Example 3.3.3. Consider the Veronese surface V in P5 . It can be seen as the 2-uple
embedding of P2 in P5 ,
ψ : P2 → P5 : (x0 , x1 , x2 ) 7→ (x20 , x0 x1 , x0 x2 , x21 , x1 x2 , x22 )
and it turns out that this embedding is an isomorphism from P2 onto its image. From
Example 2.1.3 we know that pa (P2 ) = 0, so that pa (V ) = 0 as the arithmetic genus is
invariant under isomorphism. In Example 2.4.16 we have seen that Cl V ∼ = Z, and that
a canonical divisor is K = −3C where C is a conic on the Veronese surface in P5 , which
corresponds on P2 to K = −3L where L is a line, so that the degree of K is −3. We can
take a line h in P2 (corresponding under the 2-uple embedding to a conic Ch on the Veronese
surface) as a generator for Cl V , and as any two lines in P2 have an intersection, we have
that h2 = h.h = h.h0 = 1, where h0 is a line different from h. If D and D0 are divisors of
degrees m and n, then D = mh and D0 = nh, so that D.D0 = mnh2 = mn. As explained
above in Example 2.3.7, a divisor D of degree d on P2 has
 
2+d 1
dimk L(D) = = (d + 1)(d + 2).
2 2
Thus the right hand side of the Riemann-Roch equation for the Veronese surface becomes
1 1 1 1
D.(D − K) + 1 + pa (V ) = d.(d + 3) + 1 = d(d + 3) + 1 = (d + 1)(d + 2).
2 2 2 2
2+d


For the left hand side, we have that l(D) = 2 if d ≥ 0 and l(D) = 0 if d < 0, and as K − D
has degree −3 − d we have that l(K − D) = −1−d = d+2



2 2 if d ≤ −3 and l(K − D) = 0 if
d > −3. Notice that if −3 < d < 0, thus for d = −1, d = −2 we also have that the right hand
side of the Riemann-Roch equation cancels. Thus the left hand side is
 
2+d 1
l(D) − s(D) + l(K − D) = − s(D) = (d + 2)(d + 1) − s(D).
2 2
Thus the Riemann-Roch theorem tells us that
1 1
(d + 1)(d + 2) − s(D) = (d + 1)(d + 2),
2 2

14
so that the superabundance of a divisor D of degree d on the Veronese surface is given by

s(D) = 0.

This could be checked by calculating the Čech cohomology of H 1 (P2 , L(D). For any divisor
D on Y = P2 , the intvertible sheaf L(D) is isomorphic to OY (d) for some integer d. But by
III.5.1 in [Har77], we then see that H 1 (Y, L(D)) = H 1 (Y, OY (d)) = 0 for any integer d, so
that indeed s(D) = dim H 1 (Y, L(D)) = 0.

References
[Har77] R. Hartshorne. Algebraic Geometry. Graduate Texts in Mathematics, No. 52.
Springer-Verlag, New York-Heidelberg, 1977.
[Per08] D. Perrin. Algebraic Geometry: An Introduction. Unversitext, Springer-Verlag
London Limited, 2008.
[Sha13a] I. R. Shafarevich. Basic Algebraic Geometry 1, Varieties in Projective Space.
Springer-Verlag Berlin Heidelberg, Third Edition, 2013.
[Sha13b] I. R. Shafarevich. Basic Algebraic Geometry 2, Schemes and Complex Manifolds.
Springer-Verlag Berlin Heidelberg, Third Edition, 2013.

15