LECTURE NOTES FOR ALGEBRAIC GEOMETRY

LEONARDO CONSTANTIN MIHALCEA

Contents
1. Notation 3
2. What is algebraic geometry ? 3
3. Affine algebraic varieties 6
3.1. From algebra to geometry 6
3.2. From geometry to algebra 8
3.3. Hilbert’s Nullstellensatz: statement and consequences 9
3.4. Proof of the Nullstellensatz 10
3.5. Irreducibility 12
3.6. Morphisms (part I) 14
3.7. Exercises 16
4. Projective algebraic varieties 18
4.1. Projective spaces and subspaces 18
4.2. Covering by affine spaces and Zariski topology 18
4.3. Projective algebraic varieties 19
4.4. Ideals of projective varieties 20
4.5. Cones and projective Nullstellensatz 21
4.6. Projective coordinate ring 21
4.7. A word about morphisms of projective varieties 22
4.8. Exercises 23
5. Sheaves, ringed spaces, and affine algebraic varieties 24
5.1. Sheaves 24
5.2. Ringed spaces 25
5.3. Affine algebraic varieties as ringed spaces 26
5.4. Affine algebraic varieties, revisited 30
5.5. The equivalence between morphisms and coordinate rings 31
6. Algebraic varieties 33
6.1. Definition and basic properties 33
6.2. Local rings 36
7. Projective algebraic varieties, revisited 39
7.1. Graded rings and modules 39
7.2. The sheaf of regular functions on a projective variety 39
7.3. From affine to projective and back: homogenization and dehomogenization 40
1
2 LEONARDO CONSTANTIN MIHALCEA

8. Morphisms (II) 44
8.1. Morphisms to affine varieties 44
8.2. Morphisms to projective varieties 44
8.3. Images of morphisms 47
9. Products 49
9.1. Categorical product 49
9.2. Products of varieties 50
9.3. Separated varieties and the Haussdorff axiom 52
10. Dimension 55
10.1. Topology 55
10.2. Krull dimension and dimension of algebraic varieties 55
10.3. Expected dimension of systems of equations 57
10.4. Projective versions 59
11. The fibres of a morphism 60
12. Sheaves of modules 62
12.1. Definitions 62
12.2. Quasi-coherent sheaves on affine varieties 63
12.3. Quasi-coherent sheaves on projective varieties 66
13. Hilbert polynomials and Bézout’s theorem 69
13.1. The Hilbert function and the Hilbert polynomial 69
13.2. Examples of Hilbert polynomials; arithmetic genus 70
13.3. Bézout’s theorem 72
13.4. Degree of the union of two projective varieties 73
14. Start of semester 2: goals 75
15. Schemes 76
15.1. Spectrum of a ring 76
15.2. The structure sheaf of Spec(R) 79
15.3. Preschemes and morphisms of preschemes 82
16. Products of preschemes 87
16.1. Examples of fibre products 87
16.2. Preschemes and algebraic varieties 90
17. Proj(R) and projective schemes 91
17.1. Proj(R) and its structure sheaf 91
17.2. Projective subschemes 92
18. More properties of schemes 94
18.1. Integral schemes; open and closed embeddings 94
18.2. Separated morphisms; schemes 95
18.3. Proper morphisms 97
19. Relative differentials 99
19.1. (Quasi-)coherent sheaves 99
19.2. The module of relative differentials 100
19.3. The sheaf of relative differentials 102
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 3

20. Locally free sheaves and vector bundles 108

20.1. Definition and equivalence to locally free sheaves and vector bundles 108
20.2. Example: Line bundles on Pn 110
20.3. Example: Line bundles from hypersurfaces in Pn (Cartier divisors I) 111
21. Cartier divisors 112
22. Rational equivalence and the Chow group 115
22.1. Rational equivalence 115
23. Proper push-forward and flat pull-back 117
23.1. Proper push-forward 117
23.2. Fundamental class of a subscheme 118
23.3. Flat pull-back 118
23.4. An exact sequence 120
24. Chern classes of line bundles 121
24.1. Cartier divisors, Weil divisors, pseudodivisors 121
24.2. Intersections by divisors and the first Chern class 123
25. Chern classes of vector bundles 125
25.1. Chern roots and the splitting principle 126
25.2. Some answers to enumerative geometry questions 129
26. Appendix: Results from commutative algebra 130
26.1. Rings and modules of fractions; localization 130
26.2. Primary and irreducible ideals 131
26.3. Topology 131
26.4. Limits, stalks, and localization 131
References 132

1. Notation
Throughout the notes: N = {0, 1, 2, . . .} (the set of natural numbers); Z = N ∪ −N
(the set of integers); Q = {a/b : a, b ∈ Z, b ̸= 0} (the set of rational numbers); R is
the set of real numbers, and C := R + iR is the set of complex numbers; here i2 = −1.
All fields are commutative. For k a field, k[x1 , . . . , xn ] denotes the polynomial ring with
coefficients in k.

2. What is algebraic geometry ?

The strict definition of the algebraic geometry is the study of solutions of polynomial
equations. But very rarely equations are explicitly written in a problem one may solve.
More often, we know that some set of equations exist, and based on this we obtain
information about the given questions. Here are some examples:
4 LEONARDO CONSTANTIN MIHALCEA

• (Bèzout’s theorem) Consider two plane curves, given by polynomials of degree a

and b. How many intersection points can they have ? (Answer: One expects ab
points, but there may be fewer.)
• (Enumerative Geometry) Consider a cubic surface in space (i.e., given by a poly-
nomial of degree 3). Does it contain any lines ? If yes, how many ? (Answer:
for a sufficiently general cubic polynomial, there are 27 lines.)
• (Rationality) When can we expect to have a parametrization ? (E.g., for curves
this says that all points are of the form x = x(t), y = y(t), where t is the
parameter, and x(t), y(t) are rational functions in t).
As an example, x(t) = t2 ; y(t) = t3 is a parametrization of the plane curve
given by equation x3 − y 2 = 0. However, the curve x3 − y 2 = 1 has no such
parametrization.
• (Dimension count) What is the ‘dimension’ of the solution set given by d polyno-
mials in the n-space ? One may expect that this is n−d (each equation decreases
the degree of freedom by 1). However, this may not be the case. For instance,
consider the space curve given parametrically by

C = {(x, y, z) = (t3 , t4 , t5 ) : t ∈ C} ⊂ C3 .

This may also be described by equations


3
x = yz

C = y 2 = xz
z 2 = x 2 y

.

Note that there are three equations, so we would expect only two equations.
However, if we leave out the last equation, the z-axis satisfies the first two;
similarly we cannot remove any of the other two equations.
Further, one may actually show that the common locus of

3
x = yz

y 2 = xz
z 2 = x2 y + ε


for small ε ∈ C is actually 0 dimensional.

• (Index theorems) Cut an algebraic object by sufficiently many hyperplane sec-
tions, until we get to a curve. What can we say about this curve ? (Is it connected
? Can we write a parametrization ? What is its genus ?)
• (Changing the field of definition) What happens if we look for solutions of a
system of equation over a specific field ? For instance, consider the plane curve
given by equation:
x2 + y 2 = 1.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 5

Can we find complex solutions ? Real solutions ? Rational solutions ? (i.e.

solutions over Q). What if we change the equation to x3 + y 3 = 1 ? (This is
essentially Fermat’s Last Theorem.)
• (Borel-Weil-Bott theorem) Consider an (irreducible) representation V of the
group SL2 (C) (2 × 2 matrices with complex coefficients and determinant 1).
Is there a way to naturally obtain V as functions on a space where SL2 acts ?
(Yes, V is the global sections - a.k.a. generalized functions - of a certain bundle
on the complex projective line P1 .)
6 LEONARDO CONSTANTIN MIHALCEA

3. Affine algebraic varieties

In this section we define affine varieties and the Zariski topology. Then we use the
Hilbert’s Nullstellensatz to give a dictionary between affine varieties and ideals in poly-
nomial rings.
3.1. From algebra to geometry. Let k be a field. The (n-dimensional) affine space
is defined by
An := Ank = {(a1 , . . . , an ) : ai ∈ k}.
Point-wise it is the same as the k-vector space k n , but unlike in the vector space situation,
0 will not play a special role. The most basic object in algebraic geometry is the affine
(algebraic) variety, defined next.
Definition 3.1. Let S be a nonempty subset of k[x1 , . . . , xn ]. The affine (algebraic)
variety V (S) is defined by
V (S) = {a = (a1 , . . . , an ) ∈ An : P (a) = 0 ∀P ∈ S}.
If S = {P1 , . . . , Pk } then we denote V (P1 , . . . , Pk ) := V ({P1 , . . . , Pk }).
One may think of V (S) as the common solutions of the system of polynomial equations
with elements from S. Some examples of affine algebraic varieties:
• ∅ = V (1) and An = V (0);
• The solutions of a system of linear system is an algebraic variety;
• If n = 1, then the only possibilities for affine algebraic varieties are: ∅, A1 , and a
finite collection of points (cf. Lemma 3.1).
• If n = 2 then in addition we may also have curves, of the form V (P ), P ∈ k[x, y].
For instance:
– A parabola in the plane: V (y − x2 );
– A cusp V (x2 − y 3 );
• For n = 3 we may have surfaces such as V (x2 + y 2 + z 2 − 1) (a sphere).
Lemma 3.1. Let X ⊂ k be an algebraic variety. Then either X = k or X consists of
finitely many points.
Proof. This follows because any nonzero univariate polynomial has finitely many roots.
□
Proposition 3.2. Let V and W be affine varieties. Then V ∪ W and V ∩ W are also
affine varieties.
Proof. Let V = V (S1 ) and W = V (S2 ) for some S1 , S2 ⊂ k[x1 , . . . , xn ] (not necessarily
finite). We claim that
V ∩ W = V (S1 ∪ S2 ),
V ∪ W = V (f g : f ∈ S1 , g ∈ S2 ).
The first equality is immediate from definition. For the second, observe that the inclusion
⊂ is clear. We need to prove the opposite inclusion. It is equivalent to proving that if
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 7

a∈/ V ∪ W , then a ∈/ V (f g : f ∈ S1 , g ∈ S2 ). If a ∈
/ V ∪ W then there exist f ∈ S1
and g ∈ S2 such that f (a) ̸= 0 and g(a) ̸= 0. Then f (a)g(a) = (f g)(a) ̸= 0, and we are
done. □
Example 3.3. (An example of a space which is not an algebraic variety.) Let
C := {(x, sin(x)) : x ∈ R} ⊂ R2 .
If this were an affine variety, then C is the common zero locus of some polynomials
Pi (x, y) ∈ k[x, y]. But then C ∩ {(x, 0) : x ∈ R} is also an algebraic variety, with
equations Pi (x, y) = 0 and y = 0 (by Proposition 3.2). Observe that this intersection
consists of the infinite discrete set {(kπ, 0); k ∈ Z}. On the other side, the zero locus
of the system Pi (x, 0) = 0 is either ∅, the whole x-axis, or finitely many points on the
x-axis. This is a contradiction.
(A shorter way to say this is that C ∩ R is an affine variety in R, identified to the
x-axis. But we know the affine varieties in R from Lemma 3.1, and C ∩ R is not of that
form.)
Consider the function
V : { subsets of k[x1 , . . . , xn ]} → { subsets of An }; S 7→ V (S).
We study some properties of V .
Proposition 3.4. (1) The function V is decreasing, i.e. for S1 ⊂ S2 , V (S1 ) ⊃ V (S2 ).
(2) Let ⟨S⟩ be the ideal generated by S. Then V (S) = V (⟨S⟩).
(3) Let S be a nonempty subset in k[x1 , . . . , xn ]. There exist finitely many elements
P1 , . . . , Ps ∈ S such that V (S) = V (P1 , . . T
. , Ps ). S
(4) For sets Si ∈ k[x1 , . . . , xn ] (i ∈ I), V (Si ) = V ( Si ).
(5) For finitely many sets S1 , . . . , Sp ⊂ k[x1 , . . . , xn ],
V (S1 ) ∪ V (S2 ) ∪ . . . ∪ V (Sp )
is an affine variety.
Proof. Part (1) follows by definition: if we impose more equations, the resulting variety
will be smaller.
For part (2), S ⊂ ⟨S⟩, therefore V (⟨S⟩) ⊂ P V (S). Conversely, take x ∈ V (S) and
f ∈ ⟨S⟩. Then f is a finite combination f = ai Pi for some Pi ∈ S (by definition),
therefore f (x) = 0, proving the reverse inclusion.
Part (3) utilizes the Hilbert Basis Theorem Theorem 26.1: the ideal ⟨S⟩ = ⟨P1 , . . . , Ps ⟩
for some Pi ∈ S, therefore
V (S) = V (⟨S⟩) = V (⟨P1 , . . . , Ps ⟩) = V (P1 , . . . , Ps ).
S T
S part (4), observe that since V (Si ) ⊃ V ( Si ) (by
For Spart(1)), it follows that V (Si ) ⊃
V ( Si ). We prove the reverse inclusion. If x ∈ V ( Si ), then for any i ∈ I, and any
P ∈ Si , P (x) = 0. Then x ∈ V (Si ), and since i is arbitrary, it is also in the intersection
T
V (Si ).
8 LEONARDO CONSTANTIN MIHALCEA

Part (5) follows by induction from Proposition 3.2. □

Corollary 3.5. Any finite set is an affine variety.
Proof. One point is an algebraic variety. Then apply Proposition 3.4, part (5). □
Definition 3.2. Consider the collection of sets:
{∅, An } ∪ {V (S) : S ⊂ k[x1 , . . . , xn ]}
Proposition 3.4 implies that these sets satisfy the properties of closed sets of a topology
on An , called the Zariski topology.
If X ⊂ An is any subset, we will consider it as a topological space with induced
topology.
Remark 3.1. The Zariski topology is very different from the classical ‘Euclidean’ topol-
ogy. For instance:
• The Zariski topology is not Hausdorff. In fact, if k is infinite, then any two open
sets in An intersect. (This will come up later, in relation to irreducibility.)
• The Zariski topology does not behave well with respect to products. For instance
A2 = A1 × A1 but the Zariski topology on A2 is not the product topology. (Exer-
cise.)
A basis of open sets for the Zariski topology is given by distinguished open sets:
for f ∈ k[x1 , . . . , xn ], define
Df := An \ V (f ) = {a = (a1 , . . . , an ) : f (a) ̸= 0}.
3.2. From geometry to algebra. We defined the map V sending
S ⊂ k[x1 , . . . , xn ] 7→ V (S) ⊂ An .
We now define a map in the opposite direction. For Σ ⊂ An (subset) define
I(Σ) = {P ∈ k[x1 , . . . , xn ] : P (x) = 0 ∀x ∈ Σ}.
Clearly I(Σ) is an ideal in k[x1 , . . . , xn ]. Further, the ideals arising in this way must be
radical: indeed if f n ∈ I(Σ) then f (x)n = 0 for all x ∈ Σ, thus f (x) ≡ 0 on Σ.
There we defined a map
I : { subsets of An } → { radical ideals in k[x1 , . . . , xn ]}; Σ ⊂ An 7→ I(Σ).
We would like to study I and the relation between I and V . This will lead to a ‘dic-
tionary’ between geometry (affine varieties) and algebra (radical ideals in a polynomial
ring), which sits at the foundation of Algebraic Geometry.
Example 3.6. (a) I(0) consists of all polynomials which vanish at x = 0. Any such
polynomial must satisfy P (0) = 0, giving that I(0) = ⟨x1 , . . . , xn ⟩.
(b) If k is infinite, then I(An ) = (0). (Exercise.)
(c) Let k = Fp , the field with p elements. Then for any x ∈ A1k , xp − x = 0, but
obviously xp − x is not the zero polynomial.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 9

The next proposition gives the first properties of the map I.

Proposition 3.7. (a) I is decreasing, i.e. if Σ1 ⊂ Σ2 , then I(Σ1 ) ⊃ I(Σ2 ).
(b) When restricted to affine varieties, V ◦ I = id, i.e. for any affine variety X ⊂ An ,
V (I(X)) = X.
Proof. Part (a) follows from definition. For (b), again from definition, X ⊂ V (I(X)). To
prove the reverse inclusion, let X = V (S), for some S ⊂ k[x1 , . . . , xn ]. Then ⟨S⟩ ⊂ I(X),
therefore V (I(X)) ⊂ V (S) = X. □
Corollary 3.8. When restricted to algebraic varieties, the function Σ 7→ I(Σ) is injec-
tive.
Proof. This follows from Proposition 3.7. □
Remark 3.2. In general. the map I 7→ V (I) is not injective. For instance, take k = R.
Then
V (⟨x2 + 1⟩) = V (1) = ∅.
The issue here is that R is not algebraically closed; we will see that I is bijective for k
algebraically closed.
3.3. Hilbert’s Nullstellensatz: statement and √ consequences. Recall that the rad-
n
ical of an ideal I√⊂ k[x1 , . . . , xn ] is defined by I = {f ∈ I : ∃n ∈ N, f ∈ I}. An ideal
is radical if I = I.
The main result of this section is the following:
Theorem 3.1 (Hilbert’s Nullstellensatz). Let k be algebraically closed. Then for any
ideal I ⊂ k[x1 , . . . , xn ],
√
I(V (I)) = I.
Before we prove this theorem, we list some consequences.
Corollary 3.9. The functions
V : { radical ideals in k[x1 , . . . , xn ]} → { affine varieties in An }; J 7→ V (J);
I : { affine varieties in An } → { radical ideals in k[x1 , . . . , xn ]}; Σ 7→ I(Σ).
are inverse to each other, i.e. for any radical ideal J, I(V (J)) = J and for any affine
variety Σ , V (I(Σ)) = Σ. This gives a one-to-one, inclusion reversing, correspondence,
between radical ideals in k[x1 , . . . , xn ] and affine algebraic varieties in An .
Proof. We already proved in Proposition 3.7 that V (I(Σ)) = Σ. The fact that I(V (J)) =
J for radical ideals J is a consequence of the Nullstellensatz (Theorem 3.1). □
Remark 3.3. One may wonder if there is a more general correspondence, where the
ideals are not necessarily radical. That leads to theory of (Noetherian) affine schemes.
10 LEONARDO CONSTANTIN MIHALCEA

Remark 3.4. Once we know about morphisms of affine varieties, we will see that the
equivalence above may be upgraded to an equivalence of contravariant functors between
the category of
{ finitely generated k-algebras } = {k[x1 , . . . , xn ]/I}
to
{ (reduced) affine varieties with morphisms} = {Spec(k[x1 , . . . , xn ]/I)}.
Corollary 3.10. There is a one-to-one correspondence between maximal ideals in k[x1 , . . . , xn ]
and points in An . Any maximal ideal is of the form ⟨x1 − a1 , . . . , xn − an ⟩ for some
(a1 , . . . , an ) ∈ An , and the correspondence is given by
⟨x1 − a1 , . . . , xn − an ⟩ ↔ (a1 , . . . , an ).
Proof. By Corollary 3.9, the maximal ideals corresponds to minimal, non-empty, affine
varieties, i.e. single points. These ideals are precisely of the form stated. □
Example 3.11. Let n = 1, and let I ⊂ k[x] be an ideal. Since any ideal in k[x] is
principal, I = ⟨P (x)⟩. Because k is algebraically closed
P (x) = (x − a1 )n1 · . . . · (x − as )ns ,
for some ni ∈ N. Then the Nullstellensatz says that
p
I(V (P )) = I({a1 , . . . , as }) = ⟨(x − a1 ) · . . . · (x − as )⟩ = ⟨P ⟩.
Example 3.12. Take n = 2. If P (x, y) ∈ k[x, y] is a non-constant irreducible polynomial
then I(V (P )) = ⟨P ⟩ and V (P ) is a curve. (We will see this curve is irreducible.)
3.4. Proof of the Nullstellensatz. We start by proving a weaker version of Nullstel-
lensatz.
Theorem 3.2 (Weak Nullstellensatz). Let k be algebraically closed and let I be an ideal
such that I ̸= k[x1 , . . . , xn ]. Then V (I) ̸= ∅. Further, if I is maximal, then V (I) is a
single point in An .
The theorem has the following interpretation. Consider a system of polynomial equa-
tions 


 P1 (x1 , . . . , xn ) = 0
P2 (x1 , . . . , xn ) = 0

..


 .

P (x , . . . , x ) = 0.
s 1 n
P
If there exist a1 , . . . , as ∈ k[x1 , . . . , xn ] such that ai Pi = 1, then obviously the system
has no solution. The theorem says that the converse is also true.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 11

Proof of Theorem 3.2. Since I ̸= k[x1 , . . . , xn ], we may find a maximal ideal J which
contains I. Since V (I) ⊃ V (J), we may assume I is maximal to start with. Define:
K := k[x1 , . . . , xn ]/I.
This is a k-algebra and since I is maximal K ⊃ k is also a field.
Lemma 3.13. K = k.
Proof. (In the case k is uncountable.) We will prove this in the case when k is uncount-
able. The main steps for the general proof are outlined in Remark 3.5 below.
If the extension K ⊃ k is algebraic, we are done, since k is algebraically closed.
Therefore we assume that there exist t ∈ K which is transcendent over k. Then we have
field extensions
P (t)
k ⊊ k(t) = { : P, Q ∈ k[x]} ⊂ K.
Q(t)
We claim that the elements
1
S := { : a ∈ k} ⊂ k(t)
t−a
are linearly independent over k. Indeed, assume that there exists a non-trivial finite
combination X ci
= 0; ci ∈ k.
t − ai
We may assume that ai ’s are distinct (otherwise collect the like terms). Consider a
coefficient ci0 , and multiply both sides by (t−a1t−a)·...·(t−as )
i
∈ k[t] to obtain
0

(3.1) ci0 (t − a1 ) · . . . · (t − ai0 −1 )(t − ai0 +1 ) · . . . (t − as ) = (t − ai0 )H(t),

where H(t) ∈ k[t]. Since t is transcendental, there is a well defined evaluation k-algebra
homomorphism ψ : k[t] → k sending t 7→ ci0 . Applying ψ to both sides gives ci0 = 0.
This proves the linear independence claim.
On the other side, by the definition of K, it is a finitely generated k-algebra, therefore
as a vector space it must have a countable basis. Therefore it cannot contain the linearly
independent set S, which contains uncountably many elements. This implies that there
are no elements t ∈ K which are transcendental over k. This finishes the proof of the
Lemma. □
Since k = k[x1 , . . . , xn ]/I, it follows that xi + I = ai for 1 ≤ i ≤ n, and ai ∈ k.
Equivalently, xi − ai ∈ I, implying that ⟨x1 − a1 , . . . , xn − an ⟩ ⊂ I. Since both ideals are
maximal, this forces equality, and finishes the proof of the theorem. □
Remark 3.5. (A high-brow proof of Lemma 3.13, for arbitrary k = k.) By Noether
normalization Lemma (see Theorem 26.2 below), if trdegk K/k = N there exist alge-
braically independent elements y1 , . . . , yN such that k ⊂ k[y1 , . . . , yN ] ⊂ K and K/k ⊂
k[y1 , . . . , yN ] is an integral extension.
12 LEONARDO CONSTANTIN MIHALCEA

Now we use another result in commutative algebra, stating that for an integral exten-
sion of integral domains A ⊂ B, if B is a field, then A is also a field; see Lemma 26.1.
We apply this to the integral extension k[y1 , . . . , yN ] ⊂ K; then k[y1 , . . . , yN ] is a field.
This forces N = 0, thus K/k is an integral extension, and since k is algebraically closed,
k = K.
√
Proof of the strong
√ Nullstellensatz Theorem 3.1. We need to show that J = I(V (J)).
The inclusions J ⊂ I(V (J)) follows from definition.
To prove the converse, let J = ⟨f1 , . . . , fs ⟩ for some fi ∈ k[x1 , . . . , xn ]. We need to
prove the following. Let g ∈ k[x1 , . . . , xn ]; if g(a) = 0 for any a such that f1 (a) = . . . =
fs (a), then g N ∈ J for some number N . Define the ideal:
B := ⟨f1 , . . . , fs , 1 − gxn+1 ⟩ ⊂ k[x1 , . . . , xn , xn+1 ].
We first show that we must have that B = k[x1 , . . . , xn , xn+1 ]. Indeed, if B ̸= ⟨1⟩,
then there exists a maximal ideal m ⊃ B. By the Weak Nullstellensatz Theorem 3.2, it
follows that
m = ⟨x1 − a1 , . . . , xn − an , xn+1 − an+1 ⟩.
Then
f1 (a1 , . . . , an ) = . . . = fs (a1 , . . . , an ) = 1 − an+1 g(a1 , . . . , an ) = 0.
But the first s equalities imply that g(a1 , . . . , an ) = 0, which is impossible.
We now consider the situation when B = k[x1 , . . . , xn , xn+1 ] = ⟨1⟩. Then there exist
hi ∈ k[x1 , . . . , xn+1 ] such that
X
hs+1 (x, xn+1 )(1 − xn+1 g(x)) + hi (x, xn+1 )fi (x) = 1.
1
This is an identity in k[x1 , . . . , xn , xn+1 ]. We specialize to xn+1 = g(x)
to obtain
X 1
hi (x, )fi (x) = 1.
g(x)
After clearing denominators (which are simply powers of g), we obtain an identity of the
form X
h′i (x)fi (x)g(x)ni = g(x)N .
This shows that g N ∈ ⟨f1 , . . . , fs ⟩ = J, and we are done. □
3.5. Irreducibility. Let k be an arbitrary field. In this section we explore the first
properties of affine varieties, in relation to the algebra/geometry dictionary. We start
with the following definition.
Definition 3.3 (Coordinate ring). Let X ⊂ An be an affine variety and I := I(X). The
ring
Γ(X) := k[x1 , . . . , xn ]/I
is called the coordinate ring of X.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 13

In a sense which will become precise later, Γ(X) is the ring of (algebraic) regular
functions on X. Some examples (to avoid arithmetic issues, here we assume k is of
characteristic 0):
• Γ(An ) = k[x1 , . . . , xn ];
• If C1 = V (y − x2 ) ⊂ A2 is a parabola, Γ(C1 ) = k[x, y]/⟨y − x2 ⟩;
• If C2 is the union of x and y axes in A2 , then Γ(C2 ) = k[x, y]/⟨xy⟩.
The parabola and the union of lines are somehow different: the parabola is made of
‘one part’, the other is a union of two parts. We want to make this precise.
Definition 3.4. A topological space X is called reducible if it can be written as
X = X1 ∪ X2 where X1 , X2 ⊊ X are closed subsets. If X is not reducible, it is called
irreducible.
• The real affine line A1 = R with Euclidean topology is reducible: R = (−∞, 0] ∪
[0, ∞);
• The affine line A1k over any infinite field with Zariski topology is irreducible.
(Because the non-trivial closed sets are just finite sets of points.)
• The union of lines C2 from above is obviously reducible, but one can show that
the parabola is irreducible.
Proposition 3.14. Let X be an irreducible space. Then the following hold:
(1) If ∅ =
̸ U ⊂ X is an open subset, then U is dense in X.
(2) Any two nonempty subsets in X intersect.
(3) Any non-empty subset U ⊂ X (with induced topology) is irreducible.
Proof. Write X = U ∪ (X \ U ). Since X is irreducible one of the two closed subsets must
be X, forcing X = U .
We leave (2) and (3) as exercises. □
Proposition 3.15. Let X ⊂ An be an affine variety (with Zariski topology) and let
I(X). Then the following are equivalent:
(1) X is irreducible;
(2) The ideal I(X) is a prime ideal;
(3) The coordinate ring of X, Γ(X) is an integral domain.
Proof. Obviously (2) ⇔ (3), so we will prove (1) ⇔ (2).
”⇒” Assume that X is irreducible, and take f, g ∈ k[x1 , . . . , xn ] such that f g ∈ I(X).
Then ⟨f g⟩ ⊂ I(X), therefore X = V (I(X)) ⊂ V (f g) = V (f ) ∪ V (g). It follows that
X = (X ∩ V (f )) ∪ (X ∩ V (g)).
By irreducibility it follows that either X ⊂ V (f ) or X ⊂ V (g), i.e. f ∈ I(X) or
g ∈ I(X).
”⇐” Assume that I(X) is prime, and write X = X1 ∪ X2 where X1 , X2 are closed.
First observe that
(3.2) I(X1 ∪ X2 ) = I(X1 ) ∩ I(X2 )
14 LEONARDO CONSTANTIN MIHALCEA

If X ̸= Xi , by injectivity of the map I (Corollary 3.8), it follows that there exists

Fi ∈ I(Xi ) \ I(X). Then
F1 F2 ∈ I(X1 ) ∩ I(X2 ) = I(X1 ∪ X2 ) = I(X),
and by hypothesis this implies that either F1 or F2 is in I(X), a contradiction. □
As a consequence of Proposition 3.15, the parabola V (y − x2 ) ⊂ A2 is irreducible,
since ⟨y − x2 ⟩ is a prime ideal.
Any affine variety is a finite union of irreducible components:
Theorem 3.3. Let X be an affine variety. Then X can be written as
(3.3) X = X1 ∪ . . . ∪ Xs
where each Xi ⊂ X is closed and irreducible affine variety. If we require that Xi ⊈ Xj
for any pair i, j then the decomposition (3.3) is unique up to permuting factors.
Proof. Exercise. □
The (unique) varieties Xi from the decomposition (3.3) are called irreducible com-
ponents and the decomposition itself is the decomposition into irreducible components.
In commutative algebra, the analogue of this decomposition is called the primary de-
composition of an ideal. For explicit ideals I, there are algorithms to obtain this, and
there is software which will find the primary decomposition.

3.6. Morphisms (part I).

Definition 3.5. Let X ⊂ Ap , Y ⊂ As be two affine varieties and F = (f1 , . . . , fs ) :
X → Y a map between them, where fi : X → k. We say that F is a is a morphism, if
fi ∈ Γ(X). We say that F is an isomorphism if there exists a morphism G : Y → X
such that F ◦ G = idY and G ◦ F = idX .
In other words, morphisms F : X → Y are restrictions of morphisms F : Ap → As
such that F (X) ⊂ Y . This may sound rather restrictive; it is, in other categories, but
it turns out to be fine in the algebraic category.
An equivalent way to state this definition is to require that fi ’s are polynomials in
k[x1 , . . . , xs ] such that for any Q ∈ I(Y ) ⊂ k[x1 , . . . , xs ],
(3.4)
Q(f1 (x1 , . . . , xp ), . . . , fs (x1 , . . . , xp )) = 0 ∀(x1 , . . . , xp ) ∈ X ⇐⇒ Q(f1 , . . . , fs ) ∈ I(X).
This definition implies that F is continuous in Zariski topology.
For a morphism F = (f1 , . . . , fs ) : X ⊂ Ap → Y ⊂ As we may define a k-algebra
homomorphism F ∗ : Γ(Y ) → Γ(X) by
F ∗ (xi ) = fi , ∀1 ≤ i ≤ s,
where the bars mean that we’re taking images inside the corresponding quotient rings.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 15

Observe that this morphism is well defined. For that, it suffices to show that F ∗ (Q) ∈
I(X) for any Q ∈ I(Y ). Utilizing that F ∗ is an algebra homomorphism, and Equa-
tion (3.4), we obtain
F ∗ (Q) = Q(f1 (x), . . . , fs (x)) ∈ I(X),
as needed. We denote by Hom(X, Y ) the set of morphisms from X to Y .
Given a k-algebra homomorphisms φ : Γ(Y ) → Γ(X) we may define a morphism
Fφ : X → Y by
Fφ (x) = (f1 (x), . . . , fs (x))
where fi (x) are any lifts to k[x1 , . . . , xp ] of φ(xi ). Clearly, Fφ∗ = φ.
Lemma 3.16. The assignments F 7→ F ∗ and φ → Fφ are compatible with composition,
i.e.
(F ◦ G)∗ = G∗ ◦ F ∗ ; Fφ◦ψ = Fψ ◦ Fφ .
Proposition 3.17. Let X ⊂ Ap , Y ⊂ As be two affine varieties. There is a one-to-
one correspondence between morphisms F = (f1 , . . . , fs ) ∈ Hom(X, Y ) and k-algebra
homomorphisms Homk−alg (Γ(Y ), Γ(X)) given by F 7→ F ∗ .
Proof. One needs to check that for any F ∈ Hom(X, Y ), FF ∗ = F , and that for any
φ ∈ Homk−alg (Γ(Y ), Γ(X), Fφ∗ = φ. We already checked the second equality. For the
first, take x ∈ X, and assume that F = (f1 , . . . , fs ). Then F ∗ (xi ) = fi and
FF ∗ (x) = (F ∗ (x1 ), . . . , F ∗ (xs )) = (f1 (x), . . . , fs (x)) = F (x).
□
Remark 3.6. For k algebraically closed, Proposition 3.17 says that the contravariant
functor Γ from the category of affine varieties and morphisms to finitely gener-
ated, reduced, k-algebras and k-algebra homomorphisms, sending X 7→ Γ(X)
is an equivalence of categories.
We now record some algebraic properties of morphisms which correspond to geometric
information. To start, Proposition 3.17 implies that:
Corollary 3.18. A morphism F : X → Y is an isomorphism if and only if F ∗ : Γ(Y ) →
Γ(X) is a k-algebra isomorphism.
In examples below we take char k = 0.
Example 3.19 (Add picture). Consider the parabola C := V (y − x2 ) ⊂ A2 . There
is a morphism F : A1 → C defined by F (t) = (t, t2 ). This is an isomorphism, with
inverse G(x, y) = x. In fact, F induces the isomorphism F ∗ : k[x, y]/⟨y − x2 ⟩ → k[t],
F (x) = t, F (y) = t2 .
Example 3.20 (Add picture). Consider the cusp C = V (y 2 − x3 ) and F : A1 → C
defined by F (t) = (t2 , t3 ). This is a bijective morphism which is not an isomorphism.
16 LEONARDO CONSTANTIN MIHALCEA

The bijectivity statement is clear. To prove it is not an isomorphism, observe that

F : Γ(C) → k[t] is given by F (x) = t2 , F (y) = t3 . Therefore the image of F ∗ is the
∗

subring k[t2 , t3 ] ⊊ k[t], showing that F ∗ is not surjective.

In fact, one may show that the rings Γ(C) and k[t] are not isomorphic. Indeed, k[t]
is integrally closed, as it is a PID, but Γ(C) is not. To see this, first observe that F ∗
above induces a ring isomorphism Γ(C) ≃ k[t2 , t3 ]. The fraction field of Γ(C) is k(t)1.
Now t ∈ k(t) is a solution of the monic polynomial X − t2 ∈ k[t2 , t3 ][X], showing that t
is integral over k[t2 , t3 ]; but t is not in k[t2 , t3 ].
Geometrically, the cusp has a singularity at (0, 0), while A1 is non-singular. So they
cannot be isomorphic in any category where a reasonable notion of smoothness is defined.
A morphism F : X → Y is called dominant if F (X) is dense in Y , i.e. F (X) = Y .
Proposition 3.21. Let X, Y be two affine varieties, and F : X → Y .
(a) F is dominant if and only if F ∗ : Γ(Y ) → Γ(X) is injective.
(b) If X is irreducible and F is dominant then Y is irreducible.

Proof. (a) Assume that F = (f1 , . . . , fs ). For any P (x) ∈ Γ(Y ), F ∗ (P (x)) = P (f1 , . . . , fs ).
Then P (x) ∈ ker(F ∗ ) if and only if P (f1 , . . . , fs ) ∈ I(X), i.e. for all x ∈ X, F (x) ∈
V (P ) ∩ Y .
If F ∗ is not injective then P may be chosen outside I(Y ), therefore the open set
DP = Y \ V (P ) is non-empty, and disjoint from the image. Conversely, if the image is
not dense then we can find an open set U ⊂ Y such that Im(F ) ⊂ Y \ U . The latter is
a nonempty closed subset not equal to Y , therefore there exists P ∈ I(Y \ U ) \ I(Y ).
Then 0 ̸= P ∈ ker(F ∗ ).
To prove (b), observe that by (a) Γ(Y ) is a subalgebra of Γ(X). Since X is irreducible,
Γ(X) is a domain, therefore so must be Γ(Y ). □

3.7. Exercises. 1. Prove that the topology on A2 = A1 ×A1 is not the product topology.

2. Prove that the distinguished open sets Df form a basis of open sets for the Zariski
topology.

3. (a) Prove that if k is infinite, then I(An ) = (0). Find a counterexample for this in
a finite field different from k = Fp from Example 3.6 above.
(b) Fix a ∈ An . Find I(a).
(c) Let k = Fp (the field with p-elements. Find I(A1k ). Is it prime ?

4. A topological space X is called Noetherian if every descending chain of closed sub-

sets X ⊋ X1 ⊋ X2 ⊋ . . . is finite. Prove that every affine algebraic variety is Noetherian.

1This means that C and A1 are birational.

 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 17

5. Prove Theorem 3.3. (One proof relies on the fact that affine varieties are Noether-
ian.)
18 LEONARDO CONSTANTIN MIHALCEA

4. Projective algebraic varieties

In the development of algebraic geometry, it quickly became apparent that projective
varieties are much better behaved than the affine varieties. This is (partially) due to
the fact that compact spaces behave better than non-compact ones. We start with the
most basic projective variety.
4.1. Projective spaces and subspaces. Let V be a vector space over the field k.
n ∈ Z>0 . The projective space P(V ) is defined by
P(V ) = {ℓ ⊂ V : ℓ ⊂ V vector subspace , dim ℓ = 1}.
If we identify V ≃ k n+1 (i.e. we choose a basis for V ), then P(V ) is denoted by Pn . In
this case Pn may also be regarded as the set of equivalence classes [x0 : . . . : xn ] where
(x0 , . . . , xn ) ∈ k n+1 \ 0 and (x0 , . . . , xn ) ≃ (λx0 , . . . , λxn ) for λ ∈ k ∗ . The coordinates
x0 , . . . , xn are called projective coordinates. If dim V = n + 1, then we define the
dimension of P(V ) to be dim P(V ) = n.
A linear projective (sub)space of P(V ) is a subset of the form P(W ) where W ⊂ V
is a vector subspace. If dim W = 1 then P(W ) is a point; if dim W = 2 then P(W ) is
called a projective line; if dim W = 3 then P(W ) is a projective plane, and so on.
Lemma 4.1. Let P(V1 ), P(V2 ) ⊂ P(V ) be projective subspaces. Then P(V1 ) ∩ P(V2 ) =
P(V1 ∩ V2 ). In paricular, if dim P(V1 ) + dim P(V2 ) ≥ dim P(V ), then P(V1 ) ∩ P(V2 ) is
nonempty.
Proof. By definition, P(V1 ) ∩ P(V2 ) is the set of lines ℓ ⊂ V1 ∩ V2 , i.e. P(V1 ∩ V2 ). If the
dimension condition is satisfied then dim V1 + dim V2 ≥ dim V + 1, giving dim V1 ∩ V2 ≥
1. □
A projective change of coordinates is any transformation u : P(V ) → P(V )
given by a linear transformation u ∈ GL(V ). For such transformations it is clear that
u(P(W )) = P(u(W )), thus it sends a projective subspace to one of the same dimension.
By definition, the projective space is the set of lines in a vector space V . The set
∗
P (V ) of hyperplanes H ⊂ V is also a projective space, called the dual projective
space. Indeed, the inclusion H ⊂ V determines a line in V ∗ as the kernel of the
projection ker(V ∗ → H ∗ ). This shows that the dual projective space is just the usual
projective space P(V ∗ ).
4.2. Covering by affine spaces and Zariski topology. Take V = k n+1 . For 0 ≤
i ≤ n, define
Ui := {[x0 : . . . : xi : . . . xn ] : xi ̸= 0}.
Lemma 4.2. (a) The sets Ui , for 0 ≤ i ≤ n, cover Pn , i.e. Pn = ni=0 Ui .
S
(b) There are bijections φi : Ui → An defined by
x0 xi−1 xi−1 xn
[x0 : . . . : xi : . . . xn ] 7→ ( , . . . , , , . . . , ).
xi xi xi xi
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 19

The inverse is given by

(x1 , . . . , xn ) 7→ [x1 : . . . : 1 : . . . : xn ].
Proof. Immediate from definitions. □

We may define the Zariski topology on Pn by identifying Ui ≃ An , then require that

a set U ⊂ Pn is open iff U ∩ Ui is open for each i. In particular, each Ui will be open in
Pn .

4.3. Projective algebraic varieties. A non-zero polynomial F ∈ k[x0 , . . . , xn ] is ho-

mogeneous if
F (λx0 , . . . , λxn ) = λd F (x0 , . . . , xn ), ∀λ ∈ k ∗ .
If F is homogeneous, the integer d is the degree of F . Any non-zero polynomial F may
be written (uniquely) as F = F0 + . . . + Fd where Fi ’s are homogeneous. We start with
the following simple lemma.
Lemma 4.3. Let F ∈ k[x0 , . . . , xn ] be a polynomial with F = F0 + . . . + Fd the decom-
position into irreducible components. Let x̄ = [x0 : . . . : xn ] ∈ Pn . Then F (x) = 0 for all
representatives of x̄ if and only if Fi (x′ ) = 0 for all i, and x′ is any representative of x̄.
Proof. The ⇐= direction is clear. For the other direction, observe that any λ ∈ k ∗ is a
root of the polynomial equation
0 = F (λx′ ) = F0 (x′ ) + λF1 (x′ ) + . . . + λd Fd (x′ ).
Since k is infinite, this forces that F0 (x′ ) = . . . = Fd (x′ ) = 0, as required. □

The Lemma suggests that if one wants to look at zeros of polynomials which are inde-
pendent of choices of projective coordinates, then it suffices to restrict to homogeneous
polynomials.
Definition 4.1. Let S ⊂ k[x0 , . . . , xn ] be a subset consisting of homogeneous polynomi-
als. Define
Vp (S) := {[x0 , . . . , xn ] : F (x0 : . . . : xn ) = 0, ∀F ∈ S}.
A projective algebraic variety is any subset of Pn of the form Vp (S).
The function Vp sending S 7→ Vp (S) has properties very similar to those from the
affine case:
• VP is decreasing;
• Vp (S) = Vp (⟨S⟩); in particular we may assume that S is finite;
• Vp (0) = Pn ;
• intersections of projective varieties are projective; finite unions of projective va-
rieties are projective
20 LEONARDO CONSTANTIN MIHALCEA

The last property allows us to (re)define Zariski topology on Pn , by requiring that Vp (S)
are closed subsets.
One difference is the behavior with respect to (homogeneous) maximal ideals. Let
m = ⟨x0 , . . . , xn ⟩ ⊂ k[x0 , . . . , xn ]. Then
Vp (m) = ∅.
Because of this, we will call ⟨x0 , . . . , xn ⟩ the irrelevant ideal. In the affine case, the
irrelevant ideal is maximal, and it corresponds to the point (0, . . . , 0) ∈ An+1 . One may
ask what are the ideals corresponding to points. Let [a0 , . . . , an ] ∈ Pn , and pick ai ̸= 0;
w.l.o.g. a0 ̸= 0. Then
a1 an a1 an
[a0 , . . . , an ] = [1 : : . . . : ] = Vp (x1 − x0 , . . . , xn − x0 ).
a0 a0 a0 a0
One may go from an affine variety to a projective variety, by the process called ho-
mogenization. If f (x1 , . . . , xn ) ∈ k[x1 , . . . , xn ], then we may define
x1 xn
F (x0 , . . . , xn ) = xdeg
0
f
f ( , . . . , ).
x0 x0
Note that this is a homogeneous polynomial. Then one may homogenize each of the
generators of the ideal of the affine variety.
To illustrate, consider the parabola P = V (x22 −x1 ) and the hyperbola C = V (x1 x2 −1).
Recall that P and C are not isomorphic when regarded as affine varieties (because their
coordinate rings are not isomorphic - see homework). The homogenizations of the two
equations are:
F1 := x22 − x1 x0 ; F2 = x1 x2 − x20 .
Since F1 , F2 are the same up to rearranging labels, Vp (F1 ) will be isomorphic to Vp (F2 ).
Therefore, the projective completion of the parabola is the same as the projectiviza-
tion of the hyperbola. But let’s see what are the points at infinity are added in each
case. We have
V (F1 ) ∩ (x0 = 0) = {[0 : 1 : 0]} ; V (F2 ) ∩ (x0 = 0) = {[0 : 1 : 0], [0 : 0 : 1]}.

4.4. Ideals of projective varieties. For any subset V ⊂ Pn , the ideal of V is

Ip (V ) = {F ∈ k[x0 , . . . , xn ] : F (x) = 0 ∀x ∈ V }.
Clearly Ip (V ) is a homogeneous ideal, i.e. it is generated by homogeneous polynomials.
As in the affine case, it is easy to check that the function Ip sending V 7→ Ip (V ) has
the following properties:
• Ip (V ) is a radical ideal;
• Ip is decreasing;
• Vp (Ip (V )) = V ; in particular Ip is injective;
• V is irreducible if and only if Ip (V ) is a prime ideal.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 21

4.5. Cones and projective Nullstellensatz. For a projective variety X ⊂ Pn there

is the associated cone C(X). If π : k n+1 \ 0 → Pn is the projection, then
C(X) := π −1 (X) ∪ {0}.
More generally, a subset X ⊂ An+1 is called a cone if
(x0 , . . . , xn ) ∈ X =⇒ λ(x0 , . . . , xn ) ∈ X, ∀λ ∈ k ∗ .
Lemma 4.4. (a) Let I ⊊ k[x0 , . . . , xn ] be a homogeneous ideal and X = Vp (I). Then
C(X) = V (I) in An+1 .
̸ X = Vp (Ip (X)) ⊂ Pn (i.e. the projective variety X is given by the ideal
(b) Let ∅ =
Ip (X)). Then I(C(X)) = Ip (X), i.e. the ideals of X and the cone over X coincide.
Proof. Exercise. □
Lemma 4.4 gives a one-to-one correspondence between projective varieties and cones
over projective varieties, in such a way that the ideals of the sets in question are pre-
served.
Theorem 4.1 (Projective Nullstellensatz). Let I ⊊ k[x0 , . . . , xn ] be a homogeneous ideal
and k be algebraically closed. √
(a) If Vp (I) ̸= ∅ then √
Ip (Vp (I)) = I. √
(b) If Vp (I) = ∅ then I ⊃ ⟨x0 , . . . , xn ⟩. That is, either I = (1) or I = ⟨x0 , . . . , xn ⟩.
Proof. Let X = Vp (I). If X ̸= ∅, by Lemma 4.4,
√
Ip (X) = I(C(X)) = I(V (I)) = I,
where the last equality is from the affine Nullstellensatz (Theorem 3.1). This proves (a).
For (b), if X = ∅ then since I is homogeneous, the variety V (I) ⊂ An+1 is either empty
or the single point (0, . . . , 0). If V (I) = ∅ then I = (1) by the affine Nullstellensatz. If
V (I) = (0, . . . , 0) then again by the affine Nullstellensatz,
√
I(V (I)) = I = ⟨x0 , . . . , xn ⟩.
□
4.6. Projective coordinate ring. If X = Vp (I) ⊂ Pn then in analogy to the affine
case one may associate the projective coordinate ring
Γp (X) = k[x0 , . . . , xn ]/I.
Since I is a homogeneous ideal, it follows that Γp (X) is a homogeneous ring. We may
choose the grading
(Γp (X))i = ρ((k[x0 , . . . , xn ])i ),
where ρ : k[x0 , . . . , xn ] → k[x0 , . . . , xn ]/I is the natural projection. (But there are many
other gradings, e.g. obtained by shifting this one.)
Another significant difference from the affine case is that the projective coordinate
ring depends on the ambient projective space Pn .
22 LEONARDO CONSTANTIN MIHALCEA

Example 4.5. Consider the map ν2 : P1 → P2 defined by

[x0 : x1 ] 7→ [x20 : x0 x1 : x22 ].
(This is a particular case of the Veronese embedding.) It is easy to see that ν2 is injective,
and we will see later that P1 ≃ ν2 (P1 ). Further,
ν2 (P1 ) = Vp (x21 − x0 x2 );
(a smooth conic in P2 ). But
Γp (P1 ) = k[u, v] ̸≃ k[x0 , x1 , x2 ]/⟨x21 − x0 x2 ⟩ = Γp (ν2 (C)).
Indeed, the left is a polynomial ring, thus a UFD, while the second ring is not (x21 has
two factorizations). Geometrically, the affine variety of the left hand side is A2 , but the
right hand side is a cone over a smooth conic, which is singular at the origin.
Definition 4.2. Let 0 ̸= F ∈ Γp (X) of degree > 0. The distinguished open set
DF+ ⊂ X is defined by
DF+ := {x ∈ X : F (x) ̸= 0}.
Clearly DF+ is a non-empty open set in X.
As an example, Dx+i = Ui ≃ An from §4.2.
Lemma 4.6. The distinguished open sets DF+ form a basis for the Zariski topology of
X ⊂ Pn . In fact, any open subset of X is a finite union of distinguished open sets DF+ .
Proof. Exercise. □

4.7. A word about morphisms of projective varieties. A natural way to define

a morphism φ : Pn → Pm is to specify an (m + 1)-tuple of homogeneous polynomials
F0 , . . . , Fm ∈ k[x0 , . . . , xn ] of the same degree. If in addition the only common zero of
the Fi ’s is (0, . . . , 0), then
(4.1) φ : Pn → Pm ; φ(x) = [F0 (x) : . . . : Fm (x)]
is a well defined map, and it is a morphism. Here are same examples:
The Veronese embedding νd : Pn → PN is defined by
(4.2) νd [x0 : . . . : xn ] = [. . . : xd00 · . . . · xdnn : . . .],
P
where di = d.
The Segre embedding νa,b : Pa × Pb → P(a+1)(b+1)−1 is defined by
(4.3) νa,b ([x0 : . . . : xa ], [y0 : . . . : yb ]) = [. . . : xi yj : . . .].
It is not hard to see that the images of both νd and νa,b are projective varieties. For
example, the image of
νd : P 1 → P n ; [x0 : x1 ] 7→ [xn0 : x0n−1 x1 : . . . : xn1 ],
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 23

is given by  
1 Z0 Z1 . . . Zn−1
νd (P ) = {[Z0 : . . . : Zn ] : rank ≤ 1}.
Z1 Z2 . . . Zn
A similar determinantal description may be given for the image of the Segre embedding.
TODO.
Further, both maps are injective, and in fact are isomorphisms onto their images (to
be proved later). As a consequence, the product of projective spaces is a projective
variety.
However, asking for all morphisms to be of this form is too strong. For example,
consider a potential map
φ : P2 → P1 ; [x0 : x1 : x2 ] 7→ [F0 (x) : F1 (x)],
where F0 , F1 ∈ k[x0 , x1 , x2 ] are homogeneous polynomials of the same degree. We will
see later that codim Vp (F0 ) ∩ Vp (F1 ) ≤ 2; in other words the common zero locus of F0 , F1
will be nonempty in P2 . A similar argument implies that there are no non-constant
morphisms P2 → P1 . In practice, we will write morphisms form any projective variety
φ : Z → Pm as in Equation (4.1) above; we will allow common zero loci, but then we
need to analyze whether one can use other relations (e.g. the equations of Z) to resolve
the indeterminacies.
In general, a morphism will be defined by local data, i.e. φ : Z → Pm is a morphism
iff for any affine covering Vi of Pn and any affine covering Ui of Z, the restrictions of φ
to φ−1 (Vi ) ∩ Uj → Ui give a morphism of affine varieties. More on this is discussed in
§8 below.
This leads us to the next section.
4.8. Exercises. 1. Prove Lemma 4.4.
24 LEONARDO CONSTANTIN MIHALCEA

5. Sheaves, ringed spaces, and affine algebraic varieties

In this section we introduce the main objects of the ‘classical’ algebraic geometry, in
their natural context. A variety will be a pair (X, OX ) of a topological space X and
a sheaf OX of regular functions on X, which is locally isomorphic to an affine variety.
The affine varieties and and projective varieties are examples of these, but so will be any
open subset of these. This generality is required to define basic objects such as regular
functions of projective varieties, and morphisms of these.
We start with a simple example, that of defining a global regular function on P1 .
Write P1 = U0 ∪ U1 , where
P i Ui = {[x0 : x i ] : xi ̸= 0}. A regular function on U0 is
x1
a polynomial P (t) = a t , where t = x0 . Similarly a regular function on U1 is a
P i j
polynomial Q(1/t) = bj (1/t ) ∈ k[1/t]. To define a global regular function, the two
functions must agree on the common domain U0 ∩ U1 , given by t ̸= 0. The condition
that P (t) = Q(1/t) forces that P (t) = Q(t) ≡ a0 , i.e. both are a constant.
This shows that to define a meaningful notion of regular functions on P1 we need to
take into consideration local data. This is achieved by the notion of a sheaf.
5.1. Sheaves.
Definition 5.1. Let X be a topological space. A presheaf F on X is given by the
following data:
• For each open set U ⊂ X, a set F(U );
• For every pair of open sets V ⊂ U in X, a restriction map rV,U : F(U ) → F(V ),
such that these satisfy the following compatibility conditions:
(1) If W ⊂ V ⊂ U are open sets, then rW U = rW V rV U ;
(2) rU U = idF (U ) .
Denote by f|V := rV U (f ). The presheaf F is a sheaf if in addition it satisfies the
following gluing property:
S
• (patching axiom) For any open set U such that U = Ui is covered by some
open sets Ui and for any fi ∈ F(Ui ) (i ∈ I) such that
(fi )|Ui ∩Uj = (fj )|Ui ∩Uj /,
there exists a unique f ∈ F(U ) such that f|Ui = fi .
An element f ∈ F(U ) is called a section over U .
Example 5.1. If X is a topological space, then the assignment
U 7→ {f : U → K : f continuous function on U },
defines a presheaf of K-valued functions, denoted by Fun(X; K).
More generally, if X, Y are topological spaces, then we may define a presheaf Fun(X; Y )
by
Fun(X, Y )(U ) = {f : U → Y : f continuous function on U }.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 25

The restriction map rV U isSgiven by the usual restriction of functions. Both are in fact
sheaves, as a map f : U = Ui → Y is continuous if and only if all f|Ui are continuous.
A morphism of presheaves φ : F → G is a collections of maps φU : F(U ) → G(U )
which commutes with restrictions:
U φ
F(U ) −−−→ G(U )
 
(5.1) rV U
y
rV U
y
V φ
F(V ) −−−→ G(V )
Example 5.2. (A presheaf which is not a sheaf.) The sheaf of constant K-valued func-
tions on a topological space X is a presheaf, but in general not a sheaf. Indeed, if X
is disconnected, then constant sections over the components may not glue to a global
section, unless all section values are the same.
Definition 5.2. (Stalks) Let F be a sheaf on X and x ∈ X. The collection F(U ), with
x ∈ U forms an inverse system. Define
Fx := lim F(U ).
←−
x∈U

This is called the stalk of F at x.

Example 5.3. If F is the sheaf of continuous functions f : U → R, then the stalk at x
consists of germs of continuous functions. Two functions f, g have the same germ at x
if f = g in a neighborhood of x.
Definition 5.3. (Sheafification) For simplicity, let’s assume that F is a presheaf of K-
valued functions on X. Then there exists a sheaf F + ⊃ F, called the sheafification of
F, defined by gluing local data, as follows:
F + (U ) = {f : U → K : ∀x ∈ U, ∃V ⊂ U open , x ∈ V, and g ∈ F(V ) s.t. f|V = g}.
This will be a sheaf, by definition. One may actually show that F + is the ‘best possible
sheaf obtained from F’, in the sense that if φ : F → G is a morphism to a sheaf G, then
it extends uniquely to a morphism φ+ : F + → G. (Exercise!)
5.2. Ringed spaces. Among the most important sheaves are the sheaves of regular
functions. These are sheaves of rings, or, more precisely, sheaves of k-algebras. This
means that the restriction maps rV U , and the maps φU from diagram 5.1, are homomor-
phisms of k-algebras.
Definition 5.4. A ringed space is a pair (X, OX ) where X is a topological space, and
OX is a sheaf of k-algebras. The sheaf OX is called the structure sheaf of X.
Assumption. From now on, we assume that k is algebraically closed, and that OX
is a sheaf of k-valued functions containing the constant functions.
26 LEONARDO CONSTANTIN MIHALCEA

Definition 5.5. (Morphisms of ringed spaces.) Let (X, OX ) and (Y, OY ) be two ringed
spaces. A morphism of ringed spaces φ : (X, OX ) → (Y, OY ) is given by a continuous
map φ : X → Y such that
∀ V ⊂ Y open, and ∀g ∈ OY (V ), g ◦ φ ∈ OX (φ−1 (V )).
For V ⊂ Y open, denote by
φ∗V : OY (V ) → OX (φ−1 (V )); g 7→ g ◦ φ.
This is a homomorphisms of k-algebras.
Remark 5.1. The assumption that OX is a sheaf of functions means that the algebra
homomorphisms φU commute with restriction homomorphisms, i.e. for any V ⊂ U , the
following diagram commutes:
φ∗U
OY (U ) −−−→ OX (φ−1 (U ))
 
rφ−1 (V )φ−1 (U )  r

V Uy
y
φ∗
V
OY (V ) −−−→ OX (φ−1 (V ))
Without this assumption, the commutation of the diagram will be part of the data of
what it means to be a morphism of ringed spaces.
Example 5.4. If we consider (X, OX ) to be a differential manifold endowed with the
sheaf of C ∞ functions on X, then the condition that φ : (X, OX ) → (Y, OY ) is a
morphism is equivalent to requiring that φ is a differentiable map.
5.3. Affine algebraic varieties as ringed spaces. Let X = V (I) ⊂ An be an affine
variety, where I is a radical ideal. We would like to define a sheaf of regular functions OX
so that (X, OX ) is a ringed space. We already know what the global regular functions
should be:
OX (X) = Γ(X) = k[x1 , . . . , xn ]/I.
In order to define the regular functions in OX (U ) for U ⊂ X arbitrary open set, the
strategy is to cover it by distinguished open sets,
U = Df1 ∪ . . . ∪ Dfp ,
define regular functions OX (Dfp ), then ‘glue’ these functions to give regular functions
on U . To make this precise we need a lemma.
Lemma 5.5. Let X be a topological space, U a basis of open sets in X, and K a set.
Assume we are given an assignment:
U 7→ F(U ) = {f : U → K : functions }, ∀U ∈ U
such that:
• (Restriction) If V ⊂ U are in U, and s ∈ F(U ), then s|V ∈ F(V ).
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 27

• (Gluing) If U ∈ U is covered by open sets Ui ∈ U, and s : U → K is a function

such that s|Ui ∈ F(Ui ), then s ∈ F(U ).
Then there exists a unique sheaf of K-valued functions F such that
F(U ) = F(U ), ∀U ∈ U.
S
Proof. Any open set U ⊂ X is covered by open sets Ui ∈ U. For U = i∈I Ui , define
F(U ) = {s : U → K : ∀i, s|Ui = si }.
S
It is not hard to check that this assignment is independent of the way we write U = Ui ,
and that this is a sheaf. □
Let X = V (I) be an algebraic variety, considered as a topological space under Zariski
topology. It has a basis of open sets
Df := X \ V (f ) = {x ∈ X : f (x) ̸= 0}
for f ∈ Γ(X). Lemma 5.5 allows us to define a sheaf by specifying its values on the sets
Df . Recall that Fun(X; k) defined by
Fun(X; k)(U ) = {f : U → k : continuous }
denotes the sheaf of continuous k-valued functions.
Theorem 5.1. Let X = V (I) ⊂ An be an affine algebraic variety. Then there exists a
unique sheaf of k-algebras on OX on X, called the structure sheaf of X, such that for
any 0 ̸= f ∈ Γ(X),
P
OX (Df ) := Γ(X)f = { : P ∈ Γ(X), n ≥ 0}.
fn
Further, there is an injective map of sheaves of k-algebras OX → Fun(X; k), sending
P P (x)
n
7→ , x ∈ Df .
f f (x)n
(For a brief review of definitions about localized rings, see §26.1. More details are
given in [AM69, Ch. 3].)
Proof. We will apply Lemma 5.5. To start we will show that for 0 ̸= f ∈ Γ(X) there is
a well defined ring homomorphism
P P (x)
OX (Df ) → {f : Df → k : continuous }, →
7 .
fn f (x)n
To show well-defineness, we need to show that for each 0 ̸= f ∈ Γ(X), the elements of
Γ(X)f are functions on Df . This follows because fPn11 = fPn22 as elements in Γ(X)f if and
only if
∃p ∈ N f p (P1 f n2 − P2 f n1 ) = 0 ∈ Γ(X).
28 LEONARDO CONSTANTIN MIHALCEA

Regarding f = f (x) as a k-valued function on Df , and since f (x) ̸= 0 on Df , this implies

that
(5.2) P1 (x)f (x)n2 − P2 (x)f (x)n1 = 0 ∀x ∈ Df ,
therefore
P1 (x) P2 (x)
= ∈ Fun(X; k)(Df ).
f (x)n1 f (x)n2
Next we show that this map is injective. Assume that
P1 (x) P2 (x)
n
= ∈ Fun(X; k)(Df ).
f (x) 1 f (x)n2
Then Equation (5.2) holds for all x ∈ Df , and regarding f, P1 , P2 as polynomials in
k[x1 , . . . , xn ], we obtain that
f (x)(P1 (x)f (x)n2 − P2 (x)f (x)n1 ) = 0 ∀x ∈ X.
Then the Nullstellensatz implies that
√
f (P1 f n2 − P2 f n1 ) ∈ I = I ⇐⇒ f (P1 f n2 − P2 f n1 ) = 0 ∈ Γ(X).
The last condition means precisely that
P1 P2
n
= n2 ∈ Γ(X)f .
f 1 f
Next we show that the restriction and gluing conditions from Lemma 5.5 hold.
Restriction condition. Let 0 ̸= f1 , f2 ∈ Γ(X) such that Df1 ⊂ Df2 . This is equivalent
to V (f2 ) ⊂ V (f1 ), giving
p
f1 ∈ I(V (f1 )) ⊂ I(V (f2 )) = ⟨f2 ⟩,
by Nullstellensatz. Then f1p = gf2 in Γ(X) for g ∈ Γ(X) and some positive integer p.
Clearly, g ∈ Df1 . For any P ∈ Γ(X), this gives the following identity in Γ(X):
P f1p = P gf2 .
Therefore, as functions on Df1 ,
P (x) P (x)g(x) P (x)g(x)
= = ,
f2 (x) g(x)f2 (x) f1 (x)p
showing that the restriction of fP2 to Df1 is an element of Γ(X)f1 .
Gluing condition. Let U = Df = pi=1 Dfi be an open set, where 0 ̸= f, fi ∈ Γ(X). It
S
follows that
(5.3) V (f ) = X \ Df = V (f1 ) ∩ . . . ∩ V (fp ) = V (f1 , . . . , fp ).
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 29

(See Proposition 3.4(iv).) Let s : U → k be a continuous function such that si := s|Dfi ∈

Γ(X)fi . We need to show that s ∈ Γ(X)f . Write
Pi
si = ∈ Γ(X)fi .
fini
The intersection Dfi ∩ Dfj = Dfi fj , and the equalities on the overlaps mean that
Pi (x) Pj (x)
= ∀x ∈ Dfi fj , ∀i, j.
n
fi (x) i fj (x)nj
Possibly by multiplying the numerators by some powers of fi ’s, we may assume that
ni = nj = N for all i, j. Then
Pi (x)fj (x)N − Pj (x)fi (x)N = 0 ∀x ∈ Dfi fj ,
implying that
fi (x)fj (x)(Pi (x)fj (x)N − Pj (x)fi (x)N ) = 0 ∀x ∈ X.
As before, the Nullstellensatz implies that
(5.4) (fi fj )(Pi fjN − Pj fiN ) = 0 ∈ Γ(X).
We are looking for P ∈ Γ(X) and m ∈ N such that over Dfi ,
P Pi P
si = m
⇐⇒ N = m .
f fi f
By Nullstellensatz and Equation (5.3) applied to the ideal V (f1 , . . . , fp ) = V (f11+N , . . . , fp1+N ),
there exists m ∈ N such that
X
fm = bi fi1+N .
P
Take P := Pi bi fi . Then for a fixed i0 we have
X X X
P fi1+N
0
= b (f f
i i i0 )P f N
i i0 = b (f f
i i i0 )P f
i0 i
N
= P f
i0 i 0 bi fi1+N = Pi0 fi0 f m

where the second equality follows from Equation (5.4). This implies that in Dfi0 ,
P (x) Pi0 (x)
m
=
f (x) fi0 (x)N
□
Remark 5.2. It is possible that Df1 = Df2 for different f1 , f2 ∈ Γ(X). Then
p p
⟨f1 ⟩ = ⟨f2 ⟩
and as in the beginning part of the proof above one checks that Γ(X)f1 = Γ(X)f2 .
30 LEONARDO CONSTANTIN MIHALCEA

Remark 5.3. Finding sections of OX (U ) for arbitrary U is in general much more subtle.
For instance, one can show that
OA2 (A2 \ 0) = OA2 (A2 ).
This corresponds to the classical fact from complex analysis that a multivariable analytic
function may be extended over sets of codimension ≥ 2.
5.4. Affine algebraic varieties, revisited. We are now in position to re-define, and
slightly extend, the notion of affine algebraic variety from the beginning of the notes.
Definition 5.6. An affine algebraic variety is a ringed space (X, OX ) isomorphic
(as ringed spaces) with (V, OV ), where V = V (I) ⊂ An is an ‘old’ affine variety together
with its structure sheaf.
A morphism of affine algebraic variety is a morphism of ringed spaces.
One advantage to viewing affine algebraic varieties as ringed spaces is that this is
an intrinsic notion, independent of coordinates. The next proposition shows that when
viewed this way, the set of ‘new’ affine varieties is strictly larger than the ‘old’ set of
affine varieties.
Proposition 5.6. Let V = V (I) ⊂ An be an affine algebraic variety, and let 0 ̸= f ∈
Γ(V ). Then (Df , (OV )|Df ) is an affine algebraic variety.
Proof. Consider a representative f ∈ k[x1 , . . . , xn ] for the element f ∈ Γ(X). Let
I = ⟨f1 , . . . , fp ⟩, and define an ideal J ⊂ k[x1 , . . . , xn , xn+1 ] by
J = ⟨f1 , . . . , fp , 1 − xn+1 f ⟩.
Define φ : Df → V (J) by
1
(x1 , . . . , xn ) 7→ (x1 , . . . , xn , ).
f (x1 , . . . , xn )
Observe that φ is a continuous bijection. (One way to see continuity is to restrict to
the basis of distinguished open sets.) The inverse is the projection π : V (J) → Df ,
π(x1 , . . . , xn , xn+1 ) = (x1 , . . . , xn ), which is obviously continuous; thus φ is a homeomor-
phism. In particular, as sets, φ(Df ) = V (J). We now need to show that J is a radical
ideal. To do that we define a ring homomorphism
1
φ∗ : k[x1 , . . . , xn , xn+1 ]/J → Γ(V )f ; xi 7→ xi (1 ≤ i ≤ n); xn+1 7→ .
f
This is induced from a similar ring homomorphism k[x1 , . . . , xn , xn+1 ]/J → k[x1 , . . . , xn ]f ,
and it is easy to see it is well defined and surjective. We prove it is injective. Let
P (x1 , . . . , xn , xn+1 ) ∈ k[x1 , . . . , xn+1 ] and write
X j
P (x1 , . . . , xn , xn+1 ) = xn+1 Pj (x1 , . . . , xn ).
j
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 31

Let x = (x1 , . . . , xn ) and let N be the largest power of the powers of xjn+1 . Assume that
φ∗ (P (x1 , . . . , xn , xn+1 ) = 0, i.e.
X f (x)N −j Pj (x)
N
= 0 ∈ Γ(X)f .
j f (x)
This means that X X
f (x)s f (x)N −j Pj (x) = ai (x)fi (x) ∈ I.
j

Regard this as an identity in k[x1 , . . . , xn+1 ]. After multiplying by xN +s

n+1 , and working
modulo the ideal J (where xN +s
n+1 f (x)
N −j+s
= xjn+1 ), this shows that the left hand side
is in J, as claimed.
Now observe that I is radical if and only if Γ(V ) has no nilpotent elements. Then
Γ(V )f has no nilpotent elements (cf. Lemma 26.4), showing that J is radical.
Finally, it remains to show that φ is an isomorphism of ringed spaces. This is left as
an exercise. (Idea: reduce to distinguished open sets.) □
Example 5.7. Let S := {p1 , . . . , pk }Q⊂ A1 be a finite set of points. Then A1 \ S is
affine. (Apply Proposition 5.6 to f = (x − pi ).)
Example 5.8. The group GLn (C) is an affine algebraic variety, as it may defined as
the distinguished open set Mn (C)det .
5.5. The equivalence between morphisms and coordinate rings. For an affine
algebraic variety (X, OX ), denote by
Γ(X) := OX (X)
the coordinate ring of X.
Let (X, OX ) and (Y, OY ) be affine algebraic varieties and φ : (X, OX ) → (Y, OY ) a
morphism. By definition, φ∗Y : Γ(Y ) → Γ(X) is a k-algebra homomorphism, sending a
regular element g ∈ Γ(Y ) to φ∗Y (g) = g ◦ φ.
Conversely, assume we are given a k-algebra homomorphism F ∗ : Γ(Y ) → Γ(X). We
show that this induces a morphism of algebraic varieties F # : (X, OX ) → (Y, OY ). First,
as in Lemma 3.16 we obtain a continuous map φ : X → Y such that φ∗ = F ∗ . Next, if
f ∈ Γ(Y ) with f := F ∗ (f ) ∈ Γ(X), then we have an induced homomorphism
∗ a F ∗ (a)
φ : Γ(Y )f → Γ(X)f ; 7→ n .
fn f
This induces homomorphisms OY (Df ) → OX (Df ). As in Lemma 5.5 this data extends
uniquely to data giving a morphism of ringed spaces.
This argument proves:
Proposition 5.9. Let X, Y be affine varieties. There is a one-to-one, direction revers-
ing, correspondence
Homk (X, Y ) ←→ Homk (Γ(Y ), Γ(X)),
32 LEONARDO CONSTANTIN MIHALCEA

between morphisms of affine algebraic varieties and k-algebra homomorphisms of coor-

dinate rings.
Example 5.10. Consider U = A2 \ 0. One may show that
OA2 (A2 \ 0) = OA2 (A2 ).
From this it follows that A2 \ 0 is not an affine variety. (If it were, the inclusion
A2 \ 0 ,→ A2 would be an isomorphism, but it is not.)
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 33

6. Algebraic varieties
6.1. Definition and basic properties.
Definition 6.1. A topological space X is quasi-compact if any open covering has a
finite subcover.
Note that we do not require that the space is Hausdorff: a quasi-compact space which
is Hausdorff is called compact.
Lemma 6.1. (a) Closed subsets of a quasi-compact space X are quasi-compact.
(b) Finite unions of quasi-compact sets are quasi-compact.
Proof. Same as for the ‘compact’ situation, thus omitted. □
Example 6.2. Affine algebraic varieties are quasi-compact but in general they are not
compact.
Lemma 6.3. Any affine algebraic variety X is quasi-compact. (But in general it is not
Hausdorff, thus not compact.)
Proof. Exercise. □
Lemma 6.4. Let X be an affine algebraic variety and U ⊂ X open. Then U is quasi-
compact.
Proof. We may assume X ⊂ An . Take X \ U = V (f1 , . . . , fp ) for fi ∈ Γ(X). Then
U = Df1 ∪ . . . ∪ Dfp . Each of Dfi is affine by Proposition 5.6, thus quasi-compact. Then
the claim follows from Lemma 6.1. □
Definition 6.2. An algebraic variety is a quasi-compact ringed space (X, OX ) which
is locally isomorphic to an affine variety, in the following sense:
For any x ∈ X there exists an open set U containing x such that (U, (OX )|U ) is an
affine variety.
If U ⊂ X is an open set which is isomorphic to an affine algebraic variety then U is
called an affine open set.
To simplify notation, when we speak about varieties, affine varieties, etc, instead of
writing (X, OX ) we will simply write X, and we omit the associated sheaf of regular
functions.
Lemma 6.5. Let X be an algebraic variety. Then the affine open sets form a basis of
open sets for the topology of X.
Proof. Let U ⊂ X be open. We need S to show that U may be written as S a union of
affine open sets. By definition X = Ui where each Ui is affine, thus U = (U ∩ Ui ).
It suffices to show that U ∩ Ui is a union of affine algebraic varieties. W.l.o.g. we may
34 LEONARDO CONSTANTIN MIHALCEA

assume that Ui ⊂ An is a (closed) subset. Then Ui \ (Ui ∩ U ) is again closed in An , thus

given by some equations f1 , . . . , fj ∈ Γ(Ui ). We deduce that
j
[
U ∩ Ui = Ui \ (V (f1 ) ∩ . . . ∩ V (fj )) = (Ui \ V (fj ))
i=1

Each Ui \ V (fj ) is a distinguished open set, thus affine by Proposition 5.6. □

Next two propositions show that open sets and closed sets of varieties are varieties.
Proposition 6.6. Let X be an algebraic variety and U ⊂ X is open. Then (U, (OX )|U )
is an algebraic variety.
Proof. Quasi-compactness is an exercise. The restriction of a sheaf to an open set is
always a sheaf, which finishes the proof. □
Example 6.7. Proposition 6.6 and Example 5.10 shows that A2 \ 0 is a variety which
is not an affine variety.
Proposition 6.8. Let X be an algebraic variety and Z ⊂ X be a closed subset. For
V ⊂ Z open, define
OZ (V )′ := {f : V → k : ∃U ⊂ X open and g : U → k, s.t. V = U ∩ Z, g|V = f }.
Define OZ (V ) := OZ+ (V )′ (the sheafification of OZ (V )′ ). Then (Z, OZ ) is an affine
variety.
Remark 6.1. In general OZ′ is only a presheaf: two restricted sections agreeing on an
overlap V1 ∩ V2 may not agree on the intersection U1 ∩ U2 of the corresponding open sets
Ui ⊂ X.
Before starting the proof, we recall the definition of the sheafification in the case at
hand. If V ⊂ Z open,
OZ (V ) := {f : V → k : ∀x ∈ V, ∃U ⊂ X open , x ∈ U, and g ∈ OX (U ) s.t. f|U ∩V = g|U ∩V }.
Proof. Since Z is closed, it is quasi-compact (Lemma 6.1). Clearly (Z, OZ ) is a ringed
space. It remains to show that if U ⊂ X is open affine then (U ∩ Z, (OZ )|(U ∩Z) is affine.
For this, we may assume that X ⊂ An is affine and Z ⊂ X is closed. We have two
sheaves defined on Z: the sheaf OZ above, and the usual sheaf of regular functions on
Z (regarded as a closed subset of X), which we temporarily denote by RZ . We need to
show that
OZ = RZ .
As in the proof of Lemma 6.5, one can show that the distinguished open set Df ∩ Z,
where f ∈ Γ(X), form a basis of open sets for the topology of Z. Then from Lemma 5.5
it suffices to show that the two sheaves agree on these distinguished open sets.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 35

To start, take f ∈ Γ(X) and f its image in Γ(Z) = RZ (Z). If f|Z = 0, there is nothing
to do, as Df = ∅. Otherwise,
a
RZ (Df ) = { r : a ∈ Γ(Z)} ⊂ OZ (Df ).
f
Conversely, take V ⊂ Z open and s ∈ OZ (V ). Since the distinguished open sets form a
basis for the topology of Z, it follows that locally s is given by restrictions of sections
si ∈ OX (Dgi ) = Γ(X)gi ,
where V = ∪(Z ∩ Dgi ) and gi ∈ Γ(X). The restrictions of si give sections si of OZ over
Dgi = Dgi ∩ Z. By definition, the restrictions si = (si )|Dgi ∩Z are in RZ (Dgi ), and since
both OZ and RZ are sheaves, it follows that
OZ (V ) ⊂ RZ (V ).
This finishes the proof. □
Remark 6.2. We shall see later that the proof above shows that there is a surjective
homomorphism of sheaves OX → OZ . The proof checks surjectivity on stalks of these
sheaves.
Example 6.9. Let Z ⊂ P2 given by equations x0 x1 = 0, x0 (x0 − x2 ) = 0. Then Z =
Z1 ∪ Z2 has two components, where Z1 ≃ P1 is given by x0 = 0 and Z2 = {[1 : 0 : 1]}.
Take any open set U ⊂ P2 such that Z ⊂ U . This is equivalent to requiring that
P2 \ U ⊂ P2 \ Z, in particular, (P2 \ U ) ∩ V (x0 ) = ∅. By Bézout this implies that P2 \ U
is a finite set, implying that
OP2 (P2 \ U ) = OP2 (P2 ) = k,
i.e. constant sections. This implies that the restriction
OP2 (P2 \ U ) → OZ (Z)
is not surjective. (Indeed, OZ contains more than constant sections, e.g. sections con-
stant on each of the components.)
Proposition 6.10 (Irreducible decomposition). Let X be an algebraic variety. Then
X = X1 ∪. . .∪Xp where Xi is a closed irreducible algebraic variety and no Xi is included
in another Xj .
Proof. Since X is quasi-compact, it can be covered by finitely many affine varieties
Ui . Each Ui has a decomposition by irreducible components (Theorem 3.3). After
taking closures, and possibly removing redundant components one obtains the claimed
decomposition. □
Definition 6.3. A closed subset of a variety with its induced sheaf of functions is called
a subvariety.
36 LEONARDO CONSTANTIN MIHALCEA

Remark 6.3. The notion of subvariety differs among authors. In English circles, a
subvariety is usually closed and irreducible, while in textbooks such as that of Perrin, it
is simply an intersection of an open and closed subset of a variety (with induced ringed
space structure).
6.2. Local rings. The sheaf of regular functions of an algebraic variety may be defined
as an intrinsic object. The main tool to do this is the notion of local ring, which will
turn to be closely related to the notion of stalks, and to rings of fractions.
Definition 6.4. Let (X, OX ) be an algebraic variety, endowed with its sheaf of regular
functions. Fix x ∈ X and consider the set of pairs (U, f ), where x ∈ U ⊂ X is open and
f ∈ OX (U ). Define the equivalence relation
(U, f ) ≃ (V, g) ⇐⇒ ∃x ∈ W ⊂ U ∩ V open s.t.f|W = g|W .
The equivalence class of (U, f ) is called the germ of f at x, denoted by fx .
The set of all germs fx is denoted by OX,x , and it is called the local ring of X at x.
From a topological point of view, the set OX,x is clearly ‘local’, as it depends only on
the neighborhood of x. Algebraically, the ‘local’ terminology is justified by the following
proposition.
Proposition 6.11. The set OX,x is a local k-algebra. More precisely, it is a k-algebra
with the unique maximal ideal
mx := {fx ∈ OX,x : f (x) = 0}.
Furthermore,
OX,x /mx ≃ k.
Proof. Clearly OX,x has a ring structure given by adding and multiplying germs:
(U, f ) + (V, g) = (U ∩ V, f + g); (U, f ) · (V, g) = (U ∩ V, f · g).
(One should check these operations are well defined; this is left as an exercise.) Note
that the 0 of this ring is represented by a germ of the function identically 0 on some
neighborhood of x. There is a surjective ring homomorphism:
ex : OX,x → k; (U, f ) 7→ f (x).
The kernel of ex is ker(ex ) = mx , and by the first isomorphism theorem
OX,x / ker(ex ) = OX,x /mx ≃ k.
This implies that mx is a maximal ideal.
Furthermore, any element in fx ∈ OX,x \ mx is invertible, with inverse (1/f )x . Indeed,
we may choose an open affine set x ∈ U such that f is defined on U (cf. Lemma 6.5).
The hypothesis that f (x) ̸= 0 means that x ∈ Df . Then the inverse 1/f ∈ OU (Df ) is a
regular function on Df , and it determines the germ (1/f )x . This implies that mx is the
unique maximal ideal, and finishes the proof. □
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 37

Recall from Theorem 3.2 that if k is algebraically closed, there is a one-to-one corre-
spondence between points x ∈ An and maximal ideals mx ⊂ k[X1 , . . . , Xn ] given by
x = (x1 , . . . , xn ) ←→ ⟨X1 − x1 , . . . , Xn − xn ⟩.
If J is any ideal in k[X1 , . . . , Xn ], the ideals in the quotient ring k[X1 , . . . , Xn ]/J are in
one-to-one correspondence to the ideals in k[X1 , . . . , Xn ] containing J. Combining the
two gives a one-to-one correspondence between points x ∈ V (J) and maximal ideals in
k[X1 , . . . , Xn ]/J given by
x ∈ V (J) ←→ mx /J ⊂ k[X1 , . . . , Xn ]/J.
Proposition 6.12. Let X be an algebraic variety, x ∈ X, and let x ∈ U ⊂ X be any
affine open set, with coordinate ring R := Γ(U ). Let mx ⊂ R be the maximal ideal
corresponding to x. Then the following hold:
(a) There is an isomorphism Rmx → OX,x , which is local, i.e. it sends the (unique)
maximal ideal in Rmx to mx ⊂ OX,x .
(b) There are one-to-one correspondences between the following three sets:
• irreducible (closed) subvarieties X containing x;
• prime ideals of Rmx ;
• prime ideals p ⊂ mx ⊂ R.
Proof. W.l.o.g. we may assume that U ⊂ An and that x = (x1 , . . . , xn ). Elements of Rmx
P
are (classes of) rational functions of the form Q where P, Q ∈ k[X1 , . . . , Xn ] and Q ∈
/ mx .
This last condition means that Q(x) ̸= 0; in particular, the germ Qx is invertible in OX,x .
Then we may define a ring homomorphism
P Px
φx : Rmx → OX,x ; 7→ .
Q Qx
I leave it as an exercise to check that φx is an isomorphism. Furthermore,
P
φ−1
x (mx ) = { ∈ Rmx : P (x) = 0} ⊂ mx Rmx .
Q
This shows that the isomorphism preserves the maximal ideals, therefore it is local.
To prove (b), note that the set of irreducible subvarieties of X containing x is in bijec-
tion to the set of irreducible subvarieties of U containing x. (This bijection is obtained
by intersecting with U , and the inverse by taking closure in X.) By Nullstellensatz, the
latter is in bijection to the set of prime ideals p in R = Γ(U ) such that p ⊂ mx . The
last equivalence follows from the description of ideals in rings of fractions; see [AM69,
Prop. 3.11]. □
Proposition 6.13. Let f : (X, OX ) → (Y, OY ) be a morphism of varieties. Let x ∈ X
and y = f (x). Then f induces local homomorphism of local rings f # : OY,y → OX,x , i.e.
a homomorphism such that (f # )−1 (mx ) = my .
38 LEONARDO CONSTANTIN MIHALCEA

Proof. Exercise. (The definition of f # follows from the diagram

f
f −1 (V ) / V
φ
# 
f # (φ)=φ◦f
k
where y ∈ V ⊂ Y is open.) □
Remark 6.4. Proposition 6.12 and Proposition 6.13 will be part of definitions of local
rings of schemes, and morphisms of schemes.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 39

7. Projective algebraic varieties, revisited

The goal of this section is to put a variety structure on zero loci of homogeneous
polynomials. The resulting structure will be called again a projective algebraic variety.
7.1. Graded rings and modules. A central notion in the study of projective varieties
is that of graded ring and graded module. We briefly recall these, following [AM69,
Ch. 10].
A graded ring R is a ring together with a family of subgroups (Rn )n≥0 of the additive
group R such that
L
• R= Rn ;
• Ri · Rj ⊂ Ri+j .
The hypotheses
L imply that R0 is a subring and that each Rn is an R0 -module. If
R+ := n>0 Rn then R+ is an ideal of R.
Example 7.1. (a) The polynomial ring k[x0 , . . . , xn ] is a graded ring, with grading given
by total degree.
(b) For J a homogeneous ideal, the quotient k[x0 , . . . , x − n]/J is a graded ring. If
F is a homogeneous polynomial F ∈ k[x0 , . . . , xn ], then the degree of the image of F in
k[x0 , . . . , x − n]/J is the minimal degree achieved by a representative of F .
A graded module M over a graded ring R is an M -module R together with a family
of subgroups (Mn )n≥0 such that
L
• M= Mn ;
• Ri · Mj ⊂ Mi+j .
Each Mn is an R0 -module. An element x ∈ M is called homogeneous if x ∈ Mn for
somePn; then n is the degree of x. Any y ∈ M can be written uniquely as a finite sum
y = n yn ; the yn ’s are called the homogeneous components of y.
For M, N graded R-modules, a homomorphism of graded R-modules is an R-module
homomorphism f : M → N such that f (Mn ) ⊂ Nn for all n ≥ 0.
If R is a graded ring and f ∈ R is a homogeneous element then we may define a
grading on the localized ring Rf by setting
r
deg k = deg(r) − k deg(f ),
f
whenever r is a homogeneous element. Denote by R(f ) the subring of elements in Rf of
degree 0.
7.2. The sheaf of regular functions on a projective variety. Let X := Vp (J) ⊂ Pn
be the zero locus of a homogeneous (radical) ideal J. The coordinate ring Γp (X) :=
k[x0 , . . . , xn ]/J is a graded ring. For a homogeneous polynomial F ∈ k[x0 , . . . , xn ] define
DF+ := {x ∈ X : F (x) ̸= 0}.
This is an open set in X, called (as before) a distinguished open set.
40 LEONARDO CONSTANTIN MIHALCEA

Definition 7.1. Let X := Vp (J) ⊂ Pn and F ∈ k[x0 , . . . , xn ] of degree > 0. There exists
a unique sheaf of k−valued functions on X, denoted by OX , such that
P
OX (DF+ ) := Γp (X)(F ) = { r : P homogeneous and deg P = r deg(F )}.
F
Proof. The proof is based again on Lemma 5.5, and it follows the outline of Proposi-
tion 5.6. The only observation to make is that elements in Γp (X)(F ) are indeed functions
on X, independent of choices of representatives of points in Pn . □
Remark 7.1. The condition that deg(F ) > 0 is needed. Indeed, if F is a constant,
let’s say F = 1, then Γ(X)(F ) contains only constants, so OX (DF+ ) consists of constant
sections. But this does not lead to a sheaf (e.g. X may be disconnected and there are
functions constant on each of the connected components).
Proposition 7.2. Let X := Vp (J) ⊂ Pn . Then the ringed space (X, OX ) is an algebraic
variety.
Any ringed space which is isomorphic with (Vp (J), OVp (J) ) is called a projective
variety.
Proof. We first reduce the statement to the situation when X = Pn , therefore we assume
for now that (Pn , OPn ) is an algebraic variety. Then any closed subset X of Pn is quasi-
compact.
Next we need to check that X is covered by affine open sets. Clearly
[n
Pn = Dx+i .
i=0

We will prove in Proposition 7.5 below that each (Dx+i , (OPn )|Dx+ ) is isomorphic to the
i
affine variety (An , OAn ). Then X ∩ Dx+i with restricted sheaf of functions is (isomorphic
to) a closed variety of An .
One subtlety is that one needs to check that the restriction of the sheaf OX to Dx+i
coincides with the sheaf of regular functions on X ∩ Dx+i when regarded as a subvariety
in Dx+i . This follows because the sheaf OX is the sheafification of the restriction sheaf of
regular functions OPn (as in Proposition 6.8). □
7.3. From affine to projective and back: homogenization and dehomogeniza-
tion. Consider the polynomial ring k[x0 , . . . , xn ] and fix an indeterminate, x0 for sim-
plicity. We will define two operations (depending on x0 ):
• (dehomogenization) For a homogeneous polynomial F ∈ k[x0 , . . . , xn ] the de-
homogenization is
F♭ := F (1, x1 , . . . , xn ) ∈ k[x1 , . . . , xn ].
This is a ring homomoprhism k[x0 , . . . , xn ] → k[x1 , . . . , xn ] (since the evaluation
map is a ring homomorphism). One may also apply this to ideals: if J = ⟨Pi :
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 41

i ∈ I⟩ ⊂ k[x0 , . . . , xn ] is a homogeneous ideal, its dehomogenization is

J♭ := ⟨(Pi )♭ : i ∈ I⟩ ⊂ k[x1 , . . . , xn ].
• (homogenization) Let f ∈ k[x1 , . . . , xn ] be a polynomial of total degree d,
written as f = fd + fd−1 + . . . + f0 and fi is homogeneous of degree i. The
homogenization of f is
x1 xn
f ♯ := xd0 f0 + xd−1 d
0 f 1 + . . . + f d = x0 f ( , . . . , ).
x0 x0
Observe that (f g)♯ = f ♯ g ♯ , but in general (f +g)♯ ̸= f ♯ +g ♯ ; in fact, the latter may
not even be homogeneous. Thus homogenization is not a ring homomorphism.
For an ideal I ⊂ k[x1 , . . . , xn ] the homogenization is defined by:
I ♯ := ⟨P ♯ : P ∈ I⟩.
Example 7.3. Let f = x21 − x22 − 1 ∈ k[x1 , x2 ]. Then f ♯ = x21 − x22 − x20 and (f ♯ )♭ = f .
Example 7.4. More generally, (f ♯ )♭ = f and any homogeneous element in F ∈ k[x0 , . . . , xn ]
may be written as F = xd0 f ♯ , for some f ∈ k[x1 , . . . , xn ], where xd0 is the largest power
of x0 dividing F .
Consider the set U0 = Dx+0 ⊂ Pn defined by x0 ̸= 0. There is a bijection
φ : An → U0 ; (x1 , . . . , xn ) 7→ [1 : x1 : . . . : xn ]
with inverse x
xn  1
φ−1 : U0 → An ; [x0 : . . . : xn ] 7→
. ,...,
x0 x0
The next proposition completes the proof of Proposition 7.2.
Proposition 7.5. The map φ : (An , OAn ) → (U0 , (OPn )|U0 ) is an isomorphism of (affine)
varieties.
Proof. We need to show that φ is a homeomorphism (in Zariski topology), and that
φ and φ−1 are morphisms of ringed spaces. Observe that as in the affine case, the
distinguished open sets form a basis for the respective topologies. Therefore all the
‘local’ statements will be proved when restricted to these sets.
First observe that if F ∈ k[x0 , . . . , xn ] is homogeneous, then
φ−1 (DF+ ∩ U0 ) = DF♭ ,
and for f ∈ k[x1 , . . . , xn ],
φ(Df ) = Df+♯ ∩ U0 .
Since
(f ♯ )♭ = f and (F♭ )♯ = F/(xd0 )
(for appropriate degree d), we deduce that φ is a homeomorphism. We now show that
φ and φ−1 are morphisms of ringed spaces, i.e. they send regular sections to regular
sections.
42 LEONARDO CONSTANTIN MIHALCEA

Using that DF+ ∩ Dx+0 = Dx+0 F , one may check (exercise!) that:
 P (x , . . . , x )  P (1, x1 , . . . , xn )
∗ 0 n
φ = .
(x0 F (x0 , . . . , xn ))r F (1, x1 , . . . , xn )r
P (x1 ,...,xn )
To define the map (φ−1 )∗ observe that any fraction of the form (f (x1 ,...,xn ))r
can be
P (x0 ,...,xn )
completed to a unique fraction of the form of degree 0, where P |P and r′ ≥ r.
(x0 f ♯ )r′
♯

(Multiply the denominator with appropriate power of x0 f ♯ , then the numerator with
appropriate power of x0 . Then
 P (x , . . . , x )  P (x0 , . . . , xn )
−1 ∗ 1 n
(φ ) = .
(f (x1 , . . . , xn ))r (x0 f ( xx10 , . . . , xxn0 ))r′
and that
deg P
 P (x , . . . , x ) 
−1 ∗ 1 n r deg(f )−deg(P ) x0 P ( xx10 , . . . , xxn0 )
(φ ) = x0 .
(f (x1 , . . . , xn ))r (xdeg
0
f
f ( xx1 , . . . , xxn ))r
0 0

This finishes the proof. □

Corollary 7.6. Pn is irreducible.
Proof. This follows because ni=0 Dx+i is open and dense in each Dx+i ≃ An , thus it must
T
be irreducible, and its closure is the whole Pn . □
Remark 7.2. Using the Veronese embedding we will show later that if X is a projective
variety and DF+ ⊊ X is a non-trivial distinguished open set, then DF+ is affine.
Remark 7.3. We will also show later that if X is an connected projective algebraic
variety then OX (X) = k, i.e. the only global regular functions are the constants.
We have seen that if J ⊂ k[x0 , . . . , xn ] is a homogeneous ideal, then V (J)∩U0 = V (J♭ ).
The next result gives a geometric interpretation of the homogenization of an ideal.
Proposition 7.7. (Homogenization yields projective closure.) Let I ≤ k[x1 , . . . , xn ],
X = V (I) ⊂ An and X ⊂ Pn its closure, where An is identified to the distinguished open
set U0 . Then the following hold.
(a) X = Vp (I ♯ );
(b) If I = ⟨f ⟩ is a nonzero principal ideal, then X = Vp (f ♯ ).
Proof. Obviously Vp (I ♯ ) is closed and it contains X, thus X ⊂ Vp (I ♯ ). We now show
that it is the smallest closed subset containing X. Let Y ⊂ Pn be any closed subset such
that Y ⊃ √ X. Write Y = Vp (J) for some homogeneous ideal J ⊂ k[x0 , . . . , xn ]. We claim
that J ⊂ I ♯ ; this will imply that
Vp (I ♯ ) ⊂ V (J) = Y
proving that Vp (I ♯ ) = X.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 43

To prove the claim, take F ∈ J and as in Example 7.4 write F = xd0 f ♯ for some
f ∈ k[x1 , . . . , xn ]. Since xd0 f ♯ = 0 on X ⊂ Y it follows that f = 0 on X = X ∩ U0 , i.e.
√
f ∈ I(X) = I(V (I)) = I
by the Nullstellensatz. Then f p ∈ I for some p ∈ Z≥0 and
(f p )♯ = (f ♯ )p ∈ I ♯
√ √ √
thus f ♯ ∈ I ♯ . It follows that F = xd0 f ♯ ∈ I ♯ , giving that J ⊂ I ♯ , as claimed.
To prove (b), we observe first that
⟨f ⟩♯ = ⟨{(f g)♯ = f ♯ g ♯ : g ∈ k[x1 , . . . , xn ]}⟩ = ⟨f ♯ ⟩.
Then by (a), V (f ) = V (f ♯ ). □
Example 7.8. (Homogenization does not commute with generation.) Let I = ⟨x1 , x2 −
x21 ⟩ ⊂ k[x1 , x2 ]. The associated variety is V (I) = {(0, 0)} and by Proposition 7.7,
V (I ♯ ) = {[1 : 0 : 0]} ⊂ P2 .
However, the ideal generated by homogenizations J = ⟨x1 , x2 x0 − x21 ⟩ has zero locus
̸ Vp (I ♯ ).
Vp (J) = {[1 : 0 : 0], [0 : 0 : 1]} =
Geometrically, V (I) is the transversal intersection of a conic Q with an affine line L;
this intersection contains one point, but Q ∩ L ⊊ Q ∩ L, since the latter contains a
second point at infinity.
44 LEONARDO CONSTANTIN MIHALCEA

8. Morphisms (II)
The goal of this section is to give some examples of morphisms to affine varieties, and
morphisms between projective varieties.
8.1. Morphisms to affine varieties. Let φ : (X, OX ) → (Y, OY ) be a map.
Proposition 8.1. The condition that φ is a morphism is local:
S
• φ is a morphism if and only ifSthere exists an open cover Y = Vi , and for each
i, an open cover φ−1 (Vi ) = Uij , such that the restriction φi : Uij → Vi is a
morphism.
Proof. Immediate from definition, using that sections over Uij glue to a section of
OX (φ−1 (Vi )); cf. Definition 5.5. □
Informally, this says that if we have a globally defined map φ : X → Y , the fact that it
is a morphism may be checked in local charts. (This is similar to what happens in other
situations, such as differentiable maps, or holomorphic functions.) One can simplify this
statement in case Y is affine.
Proposition 8.2. If (Y, OY ) is an affine variety, then it suffices to check the morphism
condition for global sections in Γ(Y ) = OY (Y ):
• φ is a morphism if and only if for all f ∈ Γ(Y ), φ∗ (f ) = f ◦ φ ∈ OX (X).
Proof. If X is affine, this follows from Proposition 3.17. In general, cover X by affine
varieties, and apply the previous proposition together with the affine situation. □
Example 8.3. The projection p : An+1 \ 0 → Pn is a morphism. (All components are
global sections on An+1 \ 0, with no common zeros.) Note that we cannot extend this
morphism over 0.
In the simplest case, when n = 1, this map is sending a point (x, y) ∈ A2 to the line
[x : y] ∈ P1 . One may take a modification A f2 → A2 which over (0, 0) has fibre all
possible slopes of lines through (0, 0); in other words, replace (0, 0) ∈ A2 by a P1 . This
will lead to the notion of the blow-up of a point, and to a morphism A f2 → P1 extending
A2 → P1 over (0, 0).
Remark 8.1. If φ : (X, OX ) → (Y, OY ) is a morphism and V ⊂ X, W ⊂ Y are
subvarieties such that φ(V ) ⊂ W , then the restriction φ : V → W is a morphism.
8.2. Morphisms to projective varieties. Many morphisms between projective vari-
eties arise from the following construction.
Consider b + 1 homogeneous polynomials of the same degree F0 , . . . , Fb ∈ k[x0 , . . . , xa ].
Then there is a well defined map:
φ : Pa \ Vp (F0 , . . . , Fb ) → Pb ; [x0 : . . . : xa ] 7→ [F0 (x) : . . . : Fb (x)].
Proposition 8.4. The above map φ is a morphism.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 45

Proof. Cover the image Pb with the standard affine open sets: Pb = Dx+i . The preimage
S

φ−1 (Dx+i ) = DF+i \ Vp (F0 , . . . , Fb )

is an open set in the distinguished open set DF+i . A global section on Dx+i is of the form
G
xri
where G ∈ k[x0 , . . . , xb ] is homogeneous of degree r. The pull-back via φ is
 G  G(F , . . . , F )
0 b
φ∗ r = r
,
xi Fi
which is clearly on OPa (DF+i ). □
Example 8.5 (Projections). Let p ∈ Pn be a point and let H ⊂ Pn be a hyperplane such
that p ∈
/ H. Then we may define the projection from p onto the hyperplane H to
be the map
πp : Pn \ {p} → H; q 7→ pq ∩ H.
For example one may take H be given by equation xn = 0, and p = [0 : . . . : 0 : 1]. Then
the line containing p and a point q = [x0 : . . . : xn ] is of the form
P(⟨(x0 , . . . , xn ), (0, . . . , 1)⟩) = {⟨ax0 , . . . , axn−1 , axn + b⟩ : [a : b] ∈ P1 },
and the projection is defined by
πp [x0 : . . . : xn ] = [x0 : . . . : xn−1 ].
(Note that not all xi ’s may be 0, as the points p, q are distinct.)
One may use elimination theory to show that if X ⊂ Pn is closed and p ∈
/ X, then
n−1
πp (X) ⊂ P is closed.
Example 8.6 (Conics). Let φ : P1 → P2 be defined by
φ[x : y] = [x2 : xy : y 2 ]
By Proposition 8.4 this is a well defined morphism. Its image is the conic
C := Im(φ) = Vp (Y 2 − XZ) ⊂ P2 ,
where X, Y, Z are the coordinates on P2 . It is easy to check that φ is injective; in fact,
φ is an isomorphism onto its image.
We construct the inverse using ‘stereographic projection’. Let U0 = {X ̸= 0} and
U1 = {Z ̸= 0}, two distinguished open sets in P2 . Observe that:
C = (C ∩ U0 ) ∪ {[0 : 0 : 1]} = (C ∩ U1 ) ∪ {[1 : 0 : 0]}.
Let P1 = V0 ∪ V1 , the usual decomposition into affine varieties. Then also observe that
φ−1 (C ∩ U0 ) = V0 ; φ−1 (C ∩ U1 ) = V1 .
Further,
φ[1 : s] = [1 : s : s2 ]; φ[s : 1] = [s2 : s : 1].
Define fi : C ∩ Ui → Vi (i = 0, 1) by
f0 [1 : s : s2 ] = [1 : s]; f1 [s2 : s : 1] = [s : 1].
46 LEONARDO CONSTANTIN MIHALCEA

In the chart U0 , s = Y /X, therefore using the equations of C,

[1 : s : s2 ] = [1 : Y /X : Y 2 /X 2 ] = [1 : Y /X : XZ/X 2 ] = [X : Y : Z].
In other words, in the chart C ∩ U0 , f0 is defined by
f0 [X : Y : Z] = [X : Y ].
A similar argument shows that in the chart C ∩ U1 , s = Y /Z, and f1 is given by
f1 [X : Y : Z] = [Y : Z].
Note that f0 is not defined at [0 : 0 : 1], while f1 is not defined at [1 : 0 : 0]. We need to
show that f0 and f1 glue, i.e. they agree on the overlap C ∩ U0 ∩ U1 = C ∩ {XZ ̸= 0} =
C ∩ {XY Z ̸= 0}. To this end, observe that for [X : Y : Z] ∈ C ∩ {XY Z ̸= 0},
[X : Y ] = [XY : Y 2 ] = [XY : XZ] = [Y : Z].
This proves that φ−1 exists as a morphism, and it is given by gluing of f0 and f1 .
Warning: Even though φ : P1 → C is an isomorphism, it does not induce an iso-
morphism of projective coordinate rings. In fact, it induces the morphism:
Γp (C) = k[X, Y, Z]/⟨Y 2 − XZ⟩ → k[x, y] = Γp (P1 ); X 7→ x2 ; Y 7→ xy; Z 7→ y 2 .
Not only that this is not an isomorphism (it is not surjective), but the two rings are
actually not abstractly isomorphic, e.g. because k[X, Y, Z]/⟨Y 2 − XZ⟩ is not a UFD.
Example 8.7 (Veronese embedding). One may generalize the previous example as fol-
lows. Fix d ∈ N and define
(8.1) νd : Pn → PN −1 ; νd [x0 : . . . : xn ] = [. . . : xd00 · . . . · xdnn : . . .],
P
where di = d. Here N equals the number of monomials of total degree d in n + 1
variables, i.e.  
n+d
N=
n
This is a morphism, and as in the previous example a local calculation shows it is actually
an isomorphism onto its image. This image equals Vp (J) ⊂ PN −1 , where
J = ker(νd∗ = k[Z0 : . . . : ZN −1 ] → k[x0 : . . . : xn ]); Zi 7→ xi00 · . . . · xinn ,
ij = d. The variety Vp (J) = νd (Pn ) is called the Veronese variety.
P
for
Example 8.8 (Veronese curves). If n = 1, νd : P1 → PN is an example of a determi-
nantal variety:
 
1 Z0 Z1 . . . Zn−1
νd (P ) = {[Z0 : . . . : Zn ] : rank ≤ 1}.
Z1 Z2 . . . Zn
For example, for d = 3, one recovers the twisted cubic ν3 (P1 ) ⊂ P3 :
 
U V Z
{[U : V : Z : T ] : rank ≤ 1} = Vp (U Z − V 2 , U T − V Z, V T − Z 2 ).
V Z T
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 47

Example 8.9 (Veronese surfaces). Take ν2 : P2 → P5 . The image of ν2 may be realized

as a determinantal variety as follows. Consider symmetric matrices
 
Z0 Z3 Z4
M = Z3 Z1
 Z5  .
Z4 Z5 Z2
Then the Veronese surface is given by the condition that rank M = 1. More generally,
(n+1)(n+2)
−1
the image of the Veronese variety ν2 (Pn ) ⊂ P 2 is given by the vanishing of the
2 × 2 minors of the (n + 1) × (n + 1) symmetric matrix with entry Zi−1,j−1 in position
(i, j). See [Harris].
Example 8.10 (Coordinate free Veronese). There is also a coordinate free description
of the Veronese embedding. If Pn = P(E) for some vector space E, and for a fixed degree
d,
νd : P(E) → P(Symd E); v → v d ,
where Symd (E) is the d-th symmetric power of E.
An important property of the Veronese embedding is that any hypersurface Vp (F ) ⊂
n
P where
P F isi0 homogeneous of degree d is the pull-back of a hyperplane via νd : if
F = ai0 ...in x0 · . . . · xinn , then
X
(8.2) Vp (F ) = νd−1 (HF ) where Hd = ai0 ...in Zi0 ...in .
This observation is utilized to prove the following projective analogue of Proposition 5.6.
Proposition 8.11. Let F ∈ k[x0 , . . . , xn ] be a homogeneous polynomial of degree d.
Then DF+ ⊂ Pn is an affine variety.
Furthermore, if X ⊂ Pn is a closed projective subvariety, then X ∩ DF+ is affine.
Proof. From Equation (8.2) it follows that DF+ = νd−1 (DH +
d
), where Hd is a linear form.
N −1
Up to a projective change of coordinates, Vp (Hd ) ⊂ P is a coordinate hyperplane,
+
thus DHd is affine. Since νd is an isomorphism over the image, it follows that DF+ is
isomorphic to a closed subset of an affine variety, thus again affine.
Same argument also proves the statement involving the closed subvariety X ⊂ Pn . □

8.3. Images of morphisms. Let f : X → Y be a morphism of varieties. Assume for

simplicity that X, Y are irreducible. (Note that if X is irreducible, then so is the image
f (X).) A natural question is what type of object is the image of a morphism ? For
example, is it always a subvariety of Y ? (Note: We will discuss preimages later on this
class, in relation to systems of parameters, and expected dimension.)
The first is the following result, whose proof is deferred.
Proposition 8.12. Let f : X → Y be a morphism of projective varieties. Then f is
closed, i.e. for any Z ⊂ X closed, f (Z) is closed.
48 LEONARDO CONSTANTIN MIHALCEA

There is a more general version of this, where f is only assumed to be proper; a

morphism of projective varieties is proper.
However, consider the morphism f : A2 → A2 , (x, y) 7→ (x, xy). The horizontal line
y = y0 is sent to the line spanned by (1, y0 ), and the image of f is
f (A2 ) = (A2 \ {y − axis}) ∪ {(0, 0)}.
This set is neither open, nor closed, around the origin. It is an example of a con-
structible set, defined next.
A locally closed set of a variety X is an intersection of an open set and a closed set
in a variety X. A constructible set is a finite union of locally closed sets. One may
check that equivalently, the class of constructible sets is the smallest class of objects
containing open sets, closed sets, and finite intersections and complements of such.
Theorem 8.1 (Chevalley). Let f : X → Y be a morphism. Then f (X) is constructible.
Furthermore, f sends constructible subsets to constructible subsets.
Proofs may be found in Mumford’s ‘Red Book’, pag. 51, or in Harris’ ‘Algebraic
Geometry’, pag. 39, in the particular case when Y = Pn .
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 49

9. Products
In this section we define the product of two varieties. We certainly want that Ar ×As =
A , but recall from a previous homework problem that the topology on Ar+s is not
r+s

the product topology. It turns out that the correct way to define products is to use the
categorical notion of product.

9.1. Categorical product. Let C be a category; this consists of a class of objects

Ob(C), morphisms F : X → Y for X, Y ∈ Ob(C) together with the abillity to compose
morphisms so we recover the usual properties of composition (identities, associativity,
etc). Let now X, Y, Z ∈ Ob(C) be three objects. Given two morphisms f : X → Z
and g : Y → Z, a product X ×Z Y is any object in C equipped with two morphisms
πX : X ×Z Y → X, πY : X ×Z Y → Z, such that the following universal mapping
property holds:
• For any object W and morphisms pX : W → X and pY : W → Y so that
f ◦ pX = g ◦ pY there exists a unique morphism p : W → X ×Z Y such that
πX ◦ p = pX and πY ◦ p = pY .

W
pX
p
$
X ×Z Y /& X
pY πX
πY f
 g 
Y / Z
The product X ×Z Y , if it exists, is unique up to an isomorphism commuting with the
projections.

Example 9.1. Take C = Sets the category of sets and maps between sets. Take A, B, C
sets and f : A → C, g : B → C. Then the product is

A ×C B = {(a, b) : f (a) = g(b)},

together with the projections to A and B respectively.

There is also the (dual) notion of categorical coproduct, which satisfies a similar
universality property, except that all arrows are reversed.

Example 9.2. Let C =k−Alg be the category of k-algebras (k a field), and k-algebra
homomorphisms. Let A, B, C, D be k-algebras and maps A → B, A → C, B → D
and C → D such that the resulting diagram commutes. (In other words, B, C, D are
A-algebras.) The morphisms B → D and C → D induce an A-bilinear k-algebra homo-
morphism B × C → D. The coproduct B ×A C is the algebra tensor product B ⊗A C.
50 LEONARDO CONSTANTIN MIHALCEA

(See e.g. [Lang, Ch. XVI, §6.)

DV ld
i iC

B ⊗O A C o CO
iB

Bo A
9.2. Products of varieties. The main theorem of this section is the following:
Theorem 9.1. Any two algebraic varieties X, Y have a product (as an algebraic variety).
Note that there is no choice for what should be the underlying set of X × Y : it’s just
the set theoretic product of X and Y . We need to realize this as a ringed space, locally
isomorphic to an affine variety. To do this, the idea is to construct first products of
affine varieties, then ‘glue’ these to get a variety structure.
Remark 9.1. One may also ask where is Z from X ×Z Y : in this case, Z is a point,
regarded as an (affine) algebraic variety. Any variety maps to pt, and X ×pt Y is denoted
simply by X × Y .
Proposition 9.3. Let X, Y be affine algebraic varieties with coordinate rings Γ(X) and
Γ(Y ). Then the following hold:
(1) There is a product X × Y which is an affine algebraic variety with coordinate
ring Γ(X) ⊗k Γ(Y );
(2) A basis for the topology of open sets is given by the distinguished open sets Df ,
where X
f= fi (x)gj (y); fi ∈ Γ(X), gj ∈ Γ(Y ).
(3) The local ring OX×Y,(x,y) is the localization of Γ(X) ⊗k Γ(Y ) at the maximal ideal
mx ⊗ Γ(Y ) + Γ(X) ⊗ my .
Proof. Let X = V (f1 , . . . , fn ) ⊂ Ar and Y = V (g1 , . . . , gm ) ⊂ As . Then the set X ×
Y ⊂ Ar+s is the zero locus of the polynomials fi , gj regarded in k[x1 , . . . , xr ; y1 , . . . , ys ].
Furthermore,
k[x1 , . . . , xr ; y1 , . . . , ys ]/⟨fi ; gj ⟩ ≃ k[x1 , . . . , xr ]/⟨fi ⟩ ⊗k k[y1 , . . . , ys ]/⟨gj ⟩ = Γ(X) ⊗k Γ(Y ).
(There is a natural bilinear map
k[x1 , . . . , xr ]/⟨fi ⟩ × k[y1 , . . . , ys ]/⟨gj ⟩ → k[x1 , . . . , xr ; y1 , . . . , ys ]/⟨fi ; gj ⟩,
which by the universal property induces a map from the tensor product. The latter
is an isomorphism when regarded as k-vector spaces, so it must be an isomorphism in
general.)
We now prove that X × Y is a categorical product. Obviously there are projections
πX : X × Y → X and πY : X × Y → Y , corresponding to maps Γ(X) → Γ(X × Y ),
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 51

f (x) 7→ f (x) ⊗ 1 and similarly for Y . Let now pX : Z → X and pY : Z → Y be any

morphism. Regarding X×Y as a product in Sets, we obtain a unique map p : Z → X×Y
which satisfies the required commutations. We need to show that p is a morphism. Since
X × Y is affine, by Proposition 8.2, it suffices to show that p∗ (f (x) ⊗ g(y)) ∈ Γ(Z). Since
Γ(X × Y ) is generated (as an algebra) by elements of the form f (x) ⊗ 1 and 1 ⊗ g(y),
it suffices to consider these elements. But then
p∗ (f (x) ⊗ 1) = p∗ πX
∗
(f (x)) = p∗X (f (x)) ∈ Γ(Z),
since pX : Z → X is a morphism. Same argument applies to 1 ⊗ g(y). This proves (1)
and (2) at the same time.
We now turn to (3). By Proposition 6.12 we know that OX×Y,(x0 ,y0 ) is the localization
of Γ(X) ⊗ Γ(Y ) at the maximal ideal m(x0 ,y0 ) of functions P vanishing at (x, y). Clearly
mx0 ⊗ Γ(Y ) + Γ(X) ⊗ my0 ⊂ m(x0 ,y0 ) .PConversely, let h = ai (x) ⊗ bj (y) ∈ m(x0 ,y0 ) such
that ai (x0 ) = αi , bj (y0 ) = βj . Then αi ⊗ βj = 0, and
X X X
ai (x) ⊗ bj (y) = ai (x) ⊗ bj (y) − αi ⊗ βj
X X
= (ai (x) − αi ) ⊗ bj (y) + αi ⊗ (gj (y) − βj );
the last quantity is in mx0 ⊗ Γ(Y ) + Γ(X) ⊗ my0 . This finishes the proof. □
Corollary 9.4. Ar × As = Ar+s .
Proof. This repeats the first part of the proof: the projection morphisms from Ar+s to
Ar and As induce a (unique) map f : Ar+s → Ar × As . At the level of coordinate rings
this gives an isomorphism k[x1 , . . . , xr ] ⊗ k[y1 , . . . , ys ] ≃ k[z1 , . . . , zr+s ], thus the desired
isomorphism. □
Theorem 9.2. Any two varieties X × Y have a product.
Proof. The idea is to cover X and Y by affine varieties, then ‘glue’ the result, to form
X × Y . First of all X × Y exists as a set. We wish give it a variety structure.
To start, we give X × Y a topology as follows: for any U ⊂ X and V ⊂ Y open
affine, consider the distinguished open sets DP fi gj ⊂ U × V , where fi ∈ Γ(U ), gj ∈
Γ(V ). We declare these to give a basis for the topology of P X × Y . (Observe that if
′ ′
U ⊂ X and V ⊂ Y are other open affines, and functions fi′ gj′ , then we may find
′′ ′ ′′ ′
U ⊂ U ∩ U and V ⊂ V ∩ V such that D(P fi gj )·(P fi′ gj′ ) is a distinguished set on
U ′′ × V ′′ ⊂ (U ∩ U ′ ) × (V × V ′ ).) Note that when restricted to products of affines U × V ,
this recovers the topology of that product. Furthemore, we may also define a ringed
space structure on X × Y by declaring that for DP fi gj ⊂ U × V ,
OX×Y (DP fi gj ) := (Γ(U ) ⊗ Γ(V ))P fj gj .
One may check that the conditions from Lemma 5.5 are satisfied, and this gives a ringed
space structure to X × Y . This space X × Y is quasi-compact, as X × Y may be written
as a finite union of quasi-compact spaces Ui × Vj , for Ui , Vj affine.
52 LEONARDO CONSTANTIN MIHALCEA

We now need to check that the resulting ringed space X ×Y is the categorical product
X × Y . Take any morphisms pX : Z → X and pY : Z → Y which satisfy the required
commutativity requirements. We need to show that the induced set-theoretic map p :
Z → X × Y is a morphism. By Proposition 8.1, it suffices to check that for each U ⊂ X,
V ⊂ Y affine, the restriction p : ZU,V := p−1 (U × V ) → U × V is a morphism. Since
U × V is affine, by Proposition 8.2 it suffices to check that for any s ∈ Γ(U × V ),
p∗U,V (s) ∈ OZ (ZU,V ). This follows as in the proof of Proposition 9.3 because Γ(U × V )
is generated by f ⊗ 1 and 1 ⊗ g for f ∈ Γ(U ), g ∈ Γ(V ), and since p∗U,V (f × 1) = p∗X (f ),
p∗U,V (1 × g) = p∗Y (g) are elements in OZ (ZU,V ). □
Remark 9.2. In addition, one may show the following:
(1) If U × X is open, then U × Y will be an open subset of X × Y .
(2) If Z ⊂ X is a closed subvariety, then Z × Y will be a closed subvariety of X × Y .
(For this, it suffices to show that for any U ⊂ X and V ⊂ Y open affine,
(Z ∩ U ) × Y is closed in U × V .)
For a, b ∈ N define νa,b : Pa × Pb → P(a+1)(b+1)−1 by
(9.1) νa,b ([x0 : . . . : xa ], [y0 : . . . : yb ]) = [. . . : xi yj : . . .].
Note that this is a well defoned map, and an analogous proof as the one from Proposi-
tion 8.4 this is a morphism, called the Segre embedding. (Note that here Pa × Pb is
already given a product variety structure.)
Lemma 9.5. (a) The image of νa,b is given by the determinants of 2 × 2 minors in the
matrix:
Z0,0 Z0,1 . . . Z0,b
 
Z1,0 Z1,1 . . . Z1,b 
 . .. . 
 .. . . . . .. 
Za,0 Za,1 . . . Za,b
where Zi,j is the coordinate corresponding to xi yj . (This is another example of a deter-
minantal variety.)
(b) Let Sa,b := νa,b (Pa × Pb ) ⊂ P(a+1)(b+1)−1 be the image of the Segre embedding. Then
νa,b : Pa × Pb → Sa,b is an isomorphism.
Proof. Exercise. □
Theorem 9.3. The product of two projective varieties is projective.
Proof. If X ⊂ Pa and Y ⊂ Pb are closed then so is X × Y ⊂ Pa × Pb (cf. Remark 9.2).
By Lemma 9.5, Pa × Pb is isomorphic to a projective variety, thus it is projective. □

9.3. Separated varieties and the Haussdorff axiom. The following is the analogue
of the Haussdorff condition in algebraic geometry.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 53

Definition 9.1. An algebraic variety X is called separated if for all varieties Z and
f
*
all morphisms Z 4 X the set {z ∈ Z : f (z) = g(z)} is a closed subset of Z.
g

Consider the diagonal map ∆ : X → X ×X given by x 7→ (x, x). This is the morphism
induced by the universality property, where p1 , p2 : X × X → X are the projections:
X
id
∆
# &
X ×X / X
p2
id
p1
  
X / pt
If X is separated, then by definition
∆(X) = {(a, a) ∈ X × X} = {(a, b) ∈ X × X : p1 (a, b) = p2 (a, b)}
is closed in X × X. The converse is also true:
Proposition 9.6. Let X be a variety. Then X is separated if and only if ∆(X) ⊂ X ×X
is closed.
Proof. We only need to show that if ∆(X) is closed, then X is separated. Let f, g : Z →
X be two morphisms. Then there is an induced morphism (f, g) : Z → X × X, and
{z ∈ Z : f (z) = g(z)} = (f × g)−1 (∆(X)).
The latter is a closed set by hypothesis. □
Example 9.7 (A non-separated variety: the line with 2 origins). The projective line P1
may be seen as the as gluing two copies U0 , U1 ≃ A1 along A1 \ {0} by the map x 7→ 1/x.
We have checked that P1 is an algebraic variety, and, for example, OP1 (P1 ) = k.
We may also glue along A1 \ {0} by x → x; this gives a different variety structure.
We call it X, and we may visualize it as ‘the line with 2 origins’. For example, Γ(X)
will contain non-constant global sections.
7 U0  o
≃

A1 >X
≃
' /
U1
The variety X is not separated; indeed, take i1 , i2 : A1 → X be the compositions. Then
{x ∈ A1 : i1 (x) = i2 (x)} = A1 \ {0},
54 LEONARDO CONSTANTIN MIHALCEA

which is not closed in A1 .

Proposition 9.8. The following hold:
(1) Closed subvarieties of separated varieties are separated.
(2) Products of separated varieties are separated.
(3) Let f : X → Y be a morphism with Y separated. Then the graph of f
Γf = {(x, y) : y = f (x)} ⊂ X × Y
is closed.
(4) Affine varieties are separated.
(5) Projective varieties are separated.
Proof. Part (1) follows from definitions, and part (2) from Remark 9.2.
For part (3), consider the morphism (f, id) : X × Y → Y × Y . Since Y is separated,
the diagonal ∆(Y ) is closed (cf. by Proposition 9.6). Then Γf = (f, id)−1 (∆(Y )) is also
closed. 2
From the construction of products of affine varieties, it follows that the diagonal is
closed, thus proving part (4).
For part (5), it suffices to assume that the projective variety X is a projective space,
X = Pn . Then for any x, y ∈ Pn , there exists a hyperplane H ⊂ Pn such that x, y ∈
Pn \ H. Since Pn \ H is affine, the conclusion that Pn is separated follows from the
Lemma 9.9 below. □
Lemma 9.9. Let X be an algebraic variety. Assume that for all x1 , x2 ∈ X there exists
an open affine U ⊂ X such that x1 , x2 ∈ U . Then X is separated.
Proof. Let f, g : Y → X be two morphisms, and let Z := {y ∈ Y : f (y) = g(y)}. Take
z ∈ Z, and x1 = f (z), x2 = g(z). By assumption there exists an open affine set V ⊂ X
such that x1 , x2 ∈ V . Let U := f −1 (U ) ∩ g −1 (V ) ⊂ Y . Then z ∈ U , and U is open. The
restrictions of f, g : U → V are again morphisms. Since V is affine, it is separated, thus
Z ∩ U is closed in U , in particular, Z ∩ U = Z ∩ U . Then z ∈ Z ∩ U , thus z ∈ Z, proving
that Z is closed. □
For further reading on separated varieties, close to approach from these notes, I rec-
ommend [Mum99, §6].

2As
an aside, observe that Γf = Image((id, f ) : X → X × Y ). Further, the restriction of the first
projection from X × Y to X gives an isomorphism Γf → X.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 55

10. Dimension
The purpose of this section is to introduce the notion of dimension for algebraic vari-
eties. We will then utilize it to study the expected dimensions of systems of polynomial
equations, and dimension of generic fibers of morphisms.

10.1. Topology. We start with the following definition.

Definition 10.1. Let X be a topological space. The dimension of X is the supremum
n such that there exists a chain
∅=
̸ X0 ⊊ X 1 ⊊ . . . ⊊ Xn ⊆ X
where each Xi is irreducible and closed in X.
If Y ⊂ X is a subspace, the codimension codimX Y := dim X − dim Y .
We will denote by dim X the dimension of X. In many proof we will use the following
properties of irreducible sets.
Lemma 10.1. Let X be a topological space, U ⊂ X an open, nonempty subset, and
Y ⊂ X. Then the following hold:
• If X is irreducible then U is dense, and any two nonempty open sets intersect.
• If Y is irreducible, then Y is irreducible, and Y ∩ U is irreducible.
Example 10.2. One may show that for An with the Euclidean topology, dim An = 0.
(Because the only irreducible closed sets are points.)
Proposition 10.3. Let Y ⊂ X. Then dim Y ≤ dim X. Further, if Y is closed, irre-
ducible and Y ̸= X, then dim Y < dim X.
Proof. Let F0 ⊊ F1 ⊊ . . . ⊊ Fk = Y be a chain of closed irreducible subset of Y . Then
each F i is irreducible, and we have a chain of closed subsets F 0 ⊂ F 1 ⊂ . . . ⊂ F k =
Y ⊂ X. Observe that Fi = Y ∩ F i , thus the chain is strictly increasing, and it gives
k ≤ dim X. If Y is closed an irreducible then to any chain as above we may add X,
proving that k < dim X. □
S
It is also easy to show that if X = Xi is a union of closed subsets Xi , then dim X =
sup dim Xi . One such union is obtained when Xi ’s are the irreducible components of X.

10.2. Krull dimension and dimension of algebraic varieties. We now turn to

algebra and algebraic varieties. Let R be a ring. Recall that the Krull dimension of
R is the maximum number n such that
p0 ⊊ p1 ⊊ . . . ⊊ pn ,
where each pi is a prime ideal in R. We will use the same notation dim R for the Krull di-
mension. A fundamental theorem in algebra calculates the dimension as a transcendence
degree.
56 LEONARDO CONSTANTIN MIHALCEA

Theorem 10.1. Let R be a finitely generated k-algebra which is an integral domain. Let
K be the fraction field of R. Then the Krull dimension of R is the transcendence degree
K/k.
Proof. See ADDREF. □
Example 10.4. The Krull dimension of k[x1 , . . . , xn ] equals to n. A chain of prime
ideals is given by 0 ⊊ ⟨x1 ⟩ ⊊ ⟨x1 , x2 ⟩ ⊊ . . . ⊊ ⟨x1 , . . . , xn ⟩.
Definition 10.2. Let X be an irreducible affine algebraic variety. Then the coordinate
ring Γ(X) is an integral domain, and the function field K(X) is defined to the fraction
field of Γ(X).
By Theorem 5.1 the function field of X coincides with the function field of any dis-
tinguished open set in X.
Proposition 10.5. Let X be an irreducible algebraic variety, and U ⊂ X a nonempty
open set. Then the following hold:
(1) If X is affine then
dim X = dim Γ(X) = trdegk K(X) < ∞.
(2) For an arbitrary (irreducible) X, dim X = dim U < ∞.
Proof. The equality dim X = dim Γ(X) follows from the Hilbert’s Nullstellensatz, which
gives an order reversing correspondence between closed irreducible subvarieties of X and
prime ideals in Γ(X). The second equality, and the finiteness conclusion, follow from
Theorem 10.1.
We prove the second part in several steps. First, assume that X is affine. Then we
can pick a distinguished open set Df ⊂ U ⊂ X. Since K(Df ) = K(X), it follows from
part (1) and Proposition 10.3 that
dim X = dim Df ≤ dim U ≤ dim X.
For arbitrary X, this also proves that any two open affine open sets U1 , U2 ⊂ X must
have the same dimension. Indeed, their intersection is non-empty, and one can find an
open set V ⊂ U1 ∩ U2 ; this satisfies:
dim U1 = dim V = dim U2 .
Let r := dim V for any ∅ ≠ V ⊂ X open, affine. Find an open affine V such that V ⊂ U .
Then
r = dim V ≤ dim U ≤ dim X,
and it suffices to show that dim X = r.
Take a chain of closed irreducible subsets in X: F0 ⊂ F1 ⊂ . . . ⊂ Fn = X, and take
V open affine such that V ∩ F0 ̸= ∅. Intersecting with V gives a chain F0 ∩ V ⊂ . . . ⊂
Fn ∩ V = V . Each Fi ∩ V is closed and irreducible in V (since Fi are irreducible), and
the sets Fi ∩ V are distinct and non-empty, because Fi = Fi ∩ V . This implies that
dim X ≤ dim V and we are done. □
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 57

Corollary 10.6. (a) dim An = dim Pn = n.

(b) Any variety of dimension 0 is finite.
Proof. The function field of An is k(x1 , . . . , xn ), which has transcendence degree n over
k. To calculate the dimension of Pn , we may take one of the distinguished open sets
Ui ≃ An , then apply Proposition 10.5.
For part (b), w.l.o.g. we may assume that X is irreducible, and we will show that
X consists of a single point. Take a nonempty open affine set U ⊂ X; then U is also
irreducible and dim U = 0. By the Nullstellensatz it follows that U corresponds to a
maximal ideal in a polynomial ring, thus it must be a point. Then X = U is also a
point, and we are done. □
10.3. Expected dimension of systems of equations. The main result of this section
is the following:
Theorem 10.2. Let X be an irreducible variety, U ⊂ X open, 0 ̸= g ∈ OX (U ) and Z
an irreducible component of {x ∈ U : g(x) = 0}. Then dim Z = dim X − 1.
Proof. Take U0 ⊂ U open affine such that Z ∩ U0 ̸= ∅. Then Z ∩ U0 is closed and
irreducible in U0 . By irreducibility and by the Nullstellensatz it corresponds to a prime
ideal p which contains the restriction f := g|U0 ∈ R := Γ(U0 ), and it is minimal with this
property. Since the Z ∩ U0 ̸= U0 , f ̸= 0. The result follows from Krull’s Hauptidealsatz
theorem, stated next. □
Theorem 10.3 (Krull’s principal ideal theorem / Krull’s Hauptidealsatz). Let R be
a finitely generated algebra over k which is an integral domain. Let f ∈ R and let
p containing f which is minimal among prime ideals containing f (this is called an
isolated prime for f ). If f ̸= 0 then
dim R/p = dim R − 1.
Proof. TODO: ADDREF. □
Definition 10.3. Let X be an irreducible variety and let Z ⊂ X be a closed subset. We
say that Z has pure dimension r if all its irreducible components have dimension r.
Similar for pure codimension r.
With this terminology, Theorem 10.2 states that the zero locus of g has pure codi-
mension 1.
Here’s a converse to Theorem 10.2.
Proposition 10.7. Let X be an irreducible variety and let Z ⊂ X irreducible, closed, of
codimension 1. Then for any open U ⊂ X such that Z ∩ U ̸= ∅, and for any f ∈ OX (U )
such that f ≡ 0 on Z, Z ∩ U is an irreducible component of V (f ).
Proof. Take W be an irreducible component of V (f ) such that W ⊃ Z ∩ U . Then
dim X > dim W ≥ dim Z ∩ U = dim X − 1.
Since Z ∩ U is closed in W , it follows that W = Z ∩ U . □
58 LEONARDO CONSTANTIN MIHALCEA

Example 10.8. The hypothesis that X is irreducible is essential. For instance, take
X = V (xy) ⊂ A2 ; this is the union of the coordinate axes. Now take
f = x(x + y + 1) ∈ Γ(X) = k[x, y]/⟨xy⟩.
Then V (f ) = V (x) ∪ {(−1, 0)}; this is not pure-dimensional.
Corollary 10.9. Let X be an irreducible variety and let Z be an irreducible component
of V (f1 , . . . , fr ), with each fi ∈ OX (X). Then codim Z ≤ r.
Proof. Induction on r. The case r = 1 follows from Theorem 10.2. Let Z ′ be an
irreducible component of V (f1 , . . . , fr−1 ). By induction, codimZ ′ ≤ r −1. By hypothesis
Z is an irreducible component of Z ′ ∩ V (fr ) ⊂ V (f1 , . . . , fr ). If fr does not vanish
identically on Z ′ , then dim Z = dim Z ′ − 1, as claimed. □
Proposition 10.10. Let X be an affine irreducible variety and let Z ⊂ X be a closed
subvariety of codimension r. Then there exist f1 , . . . , fr ∈ Γ(X) such that Z is a com-
ponent of V (f1 , . . . , fr ).
Proof. See [Mum99, Cor. I.7.4]. □
A subvariety Z ⊂ X of codimension r which is given by r equations is called a
complete intersection. Being a complete intersection is a rather strong constraint.
For example, if X = Pn and Y = Vp (F1 , . . . , Fr has codimension r then by Bèzout
theorem the degree of Y must be
r = deg(Y ) = deg(F1 ) · . . . · deg(Fr ).
This degree may often be calculated by other cohomological methods.
Example 10.11. The degree of the twisted cubic C := νe (P1 ) ⊂ P3 equals to 3. Since C
is not included in a hyperplane, it follows that I(C) cannot be generated by 2 equations.
Example 10.12. Consider the variety of matrices:
 
a b c
M1 := {A = : rk(A) ≤ 1} ⊂ A5 .
d e f
One may projectivize M1 , and P(M1 ) is the image of P2 × P1 ,→ P5 under the Segre
embedding. In particular, codim M1 = 2. One may show that deg(P(M1 )) = 3, which
implies that Ip (P(M1 )) cannot be generated by 2 equations. Then I(M1 ) (the cone over
P(M1 )) cannot be generated bby 2 equations.
These results show that a codimension 1 irreducible variety may be realized as a
component of the zero locus of a single equation. But does it equal to V (f ) ? This is
not always possible; the next result gives a sufficient condition for this to hold.
Proposition 10.13. Let X be an affine irreducible variety such that Γ(X) is an UFD.
let Z ⊂ X be closed of pure codimension 1. Then there exists f ∈ Γ(X) such that
Z = V (f ).
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 59

Proof. We start by observing that since R := Γ(X) is an UFD, any minimal nonzero
prime must be principal. (Proof: Let p ⊂ R be a minimal prime, and let f ∈ p. Since
R is an UFD, we may decompose f as a product of prime factors. The fact that p is
prime implies that it must contain one of the prime factors of f , say f ′ . But the ideal
⟨f ′ ⟩ is prime and ⟨f ′ ⟩ ⊂ p, thus ⟨f ′ ⟩ = p.)
Now the components of Z correspond to minimal prime ideals in R, generated by
some prime elements f1 , . . . , fr in R. Then Z = V (f1 · . . . · fr ) and we are done. □
10.4. Projective versions. All these results have projective versions, involving projec-
tive coordinate rings and homogeneous polynomials. For instance, the following are the
analogues of Theorem 10.2 and Proposition 10.13.
Theorem 10.4. Let X be an irreducible projective variety with projective coordinate ring
Γp (X) and F ∈ k[x0 , . . . , xn ] be a homogeneous, non-constant polynomial such that F
does not vanish identically on X. Then X ∩ Vp (F ) is nonempty and of pure codimension
1, unless dim X = 0.
Sketch of the proof. Except for the non-emptiness, the claim follows from Theorem 10.2.
To prove non-emptiness, assume that dim X ≥ 1 and let X ∗ ⊂ An+1 be the cone over
X. By definition, the cone is the affine variety given by the homogeneous ideal Ip (X).
One may prove that dim X ∗ = dim X + 1. (Exercise!) Then 0 ∈ X ∗ ∩ V (F ), and by
Theorem 10.2 dim X ∗ ∩ V (F ) ≥ 1. In particular X ∗ ∩ V (F ) contains other points except
for the origin, and since it is k ∗ -stable it must contain a full line through the origin. We
deduce that X ∩ Vp (F ) is non-empty. □
Proposition 10.14. Every closed subset of Pn of pure codimension 1 is a hypersurface,
i.e. it is given by the vanishing of a single homogeneous polynomial.
60 LEONARDO CONSTANTIN MIHALCEA

11. The fibres of a morphism

For the proofs of the results in this section we refer to [Mum99, S 8].
Let X, Y be two varieties and f : X → Y a morphism. The morphism f is called
dominant if f (X) is dense in Y , i.e. f (X) = Y .
The following is a generalization of Proposition 3.21.
Proposition 11.1. Let X be an algebraic variety, and let f : X → Y be a dominant
morphism. Then Z := f (X) is irreducible and and f induces an injective map f ∗ :
k(Z) ,→ k(X).
Recall the definition of ‘pure dimensional’ from Definition 10.3. The main result of
this section is the following:
Theorem 11.1. Let f : X → Y be a dominant morphism of irreducible algebraic
varieties. Then the following hold:
(a) Let W ⊂ Y be an irreducible subset and let Z ⊂ f −1 (W ) be an irreducible compo-
nent of f −1 (W ) that dominates W . Then
codimX Z ≤ codimY W .
(b) There exists an open set U ⊂ Y such that:
• U ⊂ f (X);
• For any irreducible W ⊂ Y such that W ∩ U ̸= ∅, and for any component
Z ⊂ f −1 (W ) such that Z ∩ f −1 (U ) ̸= ∅,
codimY W = codimX Z.
Among the most important consequences is the following corollary:
Corollary 11.2. Let f : X → Y be a dominant morphism of irreducible algebraic
varieties. Let r := dim X − dim Y . Then for any y ∈ Y , the fibre f −1 (y) either empty,
or each irreducible component has dimension ≥ r.
Further, there exists an open set U ⊂ Y such that for any y ∈ U , f −1 (y) is pure
dimensional of dimension r.
Proof. We apply Theorem 11.1 to the case when W = {y}. Observe that codimY W =
dim Y , therefore the condition codimY W ≥ codimX Z translates into dim Y ≥ dim X −
dim Z, i.e. dim Z ≥ dim X − dim Y = r. □
Example 11.3. Consider the projection p2 : X × Y → Y sending (x, y) 7→ y. Clearly
all fibres are isomorphic to X. In particular, this shows that
dim X × Y = dim X + dim Y .
Example 11.4. Consider the hyperbola C := V (xy − 1) ⊂ A2 . Since C is closed, The
second projection restricted to C is a morphism f : C → A1 . This is not surjective, but
it is dominant (the image is A1 \ {0}.) The fibre f −1 (y) is a single point unless y = 0,
when f −1 (0) = ∅. This illustrates Corollary 11.2.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 61

Example 11.5 (Ramified covers). Consider f : P2 → P2 defined by [x0 : x1 ] 7→ [x20 : x21 ].

By Proposition 8.4 this is a morphism. The fibers over z ̸= [0 : 1], [1 : 0] are two points,
while the fibres over [0 : 1], [1 : 0] are a single point. The latter is called the ramification
locus of f .
It is typical for a surjective map f : X → Y where X, Y are projective and dim X =
dim Y to have a ramification locus where the fibre changes behavior.
Example 11.6. (The blowup at a point.) Consider the morphism f : A2 → A2 defined
by (x, y) 7→ (x, xy). This is a surjective morphism. The fibre over (0, 0) is
f −1 (0, 0) = {(0, y)} ≃ A1
(thus positive dimensional), while the fibre over a point in the open set U = A2 \ {(0, 0)}
a single point. This gives another example illustrating Corollary 11.2.
Example 11.7. (Line - plane incidence) Recall that the dual projective plane (P2 )∨
parametrizes two-dimensional vector spaces in k 3 . Consider the incidence:
X := {(L ⊂ P ⊂ C3 ) : linear subspaces such that dim L = 1, dim P = 2} ⊂ P2 × (P2 )∗
This is also called the flag variety for C3 . There are projections p1 : X → P2 and
p2 : X → (P2 )∗ . The fibres are all isomorphic to P1 :
p−1 3
1 (L) = P(C /L); p−1
2 (P ) = P(P ).
We claim that X is a projective variety.
To start, observe that we may cover (let’s say) (P2 )∗ by affine open sets Ui such that
−1
p2 (Ui ) ≃ Ui × P1 . For instance, take U0 = {X + aY + bZ = 0} ⊂ (P2 )∗ where
(a, b) ∈ A2 . The projective line X + aY + bZ = 0 corresponds to a two-dimensional
vector space Va,b ⊂ k 3 , and one can find a basis v1 , v2 ⊂ Va.b . (Of course this basis
depends on a, b.) We identify Va,b with (a, b). The fibre
p−1
2 (a, b) = {([u : v], (a, b)) : [u : v] ∈ P(Va,b )}
The coordinate ([u : v], (a, b)) may be regarded as the line uv1 + vv2 ⊂ Va,b . This gives a
bijection
p−1 1
2 (U0 ) ≃ U0 × P .
Such sets cover X and they may be glued to give an structure of an algebraic variety on
X. This is closed in P2 × (P2 )∗ , because it is closed in any of these sets. For instance,
p−1 1 2 −1 2 2 ∗
2 (U0 ) ∩ X ≃ U0 × P ⊂ U0 × P = p2 (U0 ) ∩ (P × (P ) ).
Either from this description, or from Theorem 11.1, one deduces that dim X = 3.
62 LEONARDO CONSTANTIN MIHALCEA

12. Sheaves of modules

We have seen in Section 5 that sheaves sit at the foundation of the definition of
algebraic varieties. In particular, an algebraic variety (X, OX ) is equipped with a sheaf
of regular functions on it. However, a variety carries more structure on it than just the
regular functions. For instance:
• There is a correspondence between affine varieties and their coordinate rings. Is
there a geometric interpretation of modules over coordinate rings ?
• We would like to include homogeneous polynomials on the projective coordinate
rings as global sections of a certain sheaf;
• Inspired from differential topology, we would like to associate sheaves to (local
sections of) vector bundles;
• We want to be able to talk about extensions of (generalized) functions from a
subvariety to an ambient variety.

12.1. Definitions.
Definition 12.1. Let (X, OX ) be a ringed space. An OX -module is a sheaf F such that
for any U ⊂ X open, F(U ) is a OX (U )-module, and the restriction homomorphisms are
maps of modules, i.e. for any V ⊂ U ,

F(U )
res / F(V )
O O

OX (U )
res / OX (V )
Here OX (V ) and F(V ) become OX (U )-modules via the restriction maps.
A homomorphism of OX -modules φ : F → G is given by the data of OX (U )-
module homomorphisms φU : F(U ) → G(U ) for any open set U ⊂ X, compatible with
the restriction maps.
Definition 12.2. Let (X, OX ) be a ringed space and φ : F → G a homomorphism of
OX -modules.
• The kernel of φ, denoted by ker(φ), is the sheaf given by the assignment
U 7→ ker(φU ).
• The image of φ, denoted by Im(φ), is the sheafification of the presheaf
U 7→ Im(φU ).
• A sequence of sheaves and sheaf homomorphisms

0 / F
u / G
v / H / 0
is (short) exact if ker(u) = 0 (the zero sheaf ), Im(v) = H and Im(u) = ker(v).
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 63

Example 12.1 (The exponential sequence). The following sequence of sheaves of abelian
groups is short exact:

/ / e2πiz / /
0 Z OC OC∗ 0

Here Z be the sheaf given by locally constant Z-valued functions, OC is the (additive) sheaf
of holomorphic functions on C, and OC∗ is the (multiplicative) sheaf of non-vanishing
holomorphic functions. All these are sheaves of abelian groups.
Consider the non-vanishing holomorphic function f (z) = z on C∗ = C − 0. Using that
the complex logarithm is defined on any simply connected open subset of C∗ , we see that
z is in the image of the exponential map over any such domain. However, z is not in the
image of any single holomorphic function defined on the whole C∗ . (The fundamental
group of C∗ is Z.) This shows that sheafification is needed in the definition of the image.
Definition 12.3. Let f : X → Y be a morphism of algebraic varieties, and F a sheaf
on X. The direct image of F is the sheaf on Y , denoted by f∗ F, and defined by
V 7→ F(f −1 (V )).
Example 12.2. Let i : Y ,→ X be a closed subvariety and let FY be the sheaf of k-
valued functions on Y . There is a morphism of OX -modules r : OX → FY defined by
sending s ∈ OX (U ) to its restriction on U ∩ Y . By definition, the image of this sheaf
Im(r) = OY , regarded as a OX -module. (Again recall that the definition of OY required
the use of sheafification.)
The kernel of r is an ideal sheaf i.e. a sheaf IY on X where IY (U ) is the ideal of
sections on U ⊂ X vanishing of Y ∩ U . (In particular this is an OX -module.) We have
another short exact sequence
(12.1) 0 / IY / OX / i∗ OY ≃ OY / 0
Remark 12.1. As we shall see later, the sequence from Equation (12.1) may be also
be realized by gluing “local data”. if one assumes that Y, X are affine and Y ⊂ X is
closed, this is the sequence associated with the short exact sequence of Γ(X)-modules
0 → IY → Γ(X) → Γ(Y ) → 0 where IY is an ideal of Y in X.
Definition 12.4. Let F, G be two OX -modules. The tensor product F ⊗OX G is the
sheafification of the presheaf
U 7→ F(U ) ⊗OX (U ) G(U ).
12.2. Quasi-coherent sheaves on affine varieties. Let X ⊂ An be an affine algebraic
variety and let R := Γ(X) = OX (X) be its coordinate ring. For any sheaf F of OX -
modules there is an associated R-module M := F(X); we will also use the notation
Γ(F) = F(X). In other words, taking global sections gives a map:
(12.2) Γ : ( sheaves of OX -modules ) → (Γ(X) − modules)
64 LEONARDO CONSTANTIN MIHALCEA

The purpose of this section is to perform the inverse construction: associate a OX -module
for any R-module. Recall that for an R-module M and f ∈ R, the localization
m
Mf = M ⊗R Rf = { s : m ∈ M, s ∈ Z}/ ≃
f
m1 m2
The equivalence is given by f s1
= f s2
if and if only if there exists f s such that f s (f s2 m1 −
f s1 m2 ) = 0 in M .

Definition 12.5. Let M be an R-module. Define the OX -module M

f as the unique sheaf
given by
Mf(Df ) = Mf := M ⊗R Rf .

As usual, the fact M f defines uniquely a sheaf a OX follows because the distinguished
open sets Df form a basis of open sets for X.
√
Example 12.3. If M = R, then M f = OX . More generally, if I = I ⊂ R is an ideal,
then Ie = IY , the ideal sheaf of Y = V (I).

Remark 12.2. Observe that Γ(M

f) = M , giving an inverse of the map from Equa-
tion (12.2).
This is the main theorem of this section.
Theorem 12.1. Let X be an affine algebraic variety. The correspondence
(OX (X) − modules) −→ ( sheaves of OX -modules ); M 7→ M
f

is functorial with respect to maps of Γ(X)-modules, it is exact, and it commutes with

direct sums and tensor products.
Proof. As before, set R := Γ(X). We start with the proof of functoriality. Take φ :
M → N a map of R-modules. For any f ∈ R, this induces a map φf : Mf → Nf of
Rf -modules, defined in an obvious way:
m
φ(m)
φf = .
fs fs
f(Df ) → N
This determines a map M f→N
e (Df ), which extends to the required map M e
of OX -modules.
We now prove that the assignment preserves exactness. Consider a short exact se-
quence of R-modules
0 / M
u / N
v / P / 0
We need to prove that
/ u′ / v′ / /
0 M
f N
e Pe 0
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 65

is exact, where u′ , v ′ are the induced maps. To check exactness, it suffices to show
exactness when restricted to distinguished open sets. For this, it suffices to show that
the localized sequence
uf vf
0 / Mf / Nf / Pf / 0
is exact.3 Clearly vf is surjective, and vf ◦ uf = 0, therefore Im(uf ) ⊂ ker(vf ). We need
to prove the converse inclusion, and also that uf is injective.
Let fns ∈ ker(vf ). Then v(n)
fs
= 0 in Pf . Then there exists a nonnegative integer k such
that f v(n) = v(f n) = 0. It follows that f k m ∈ ker(v) = Im(u), so there exists m ∈ M
k k

such that u(m) = f k n. Then

n nf k u(m)
s
= k+s = k+s ∈ Im(uf ).
f f f
We now prove that uf is injective. Let fms ∈ ker(uf ). Then f k u(m) = 0 for some
nonnegative integer k. This implies that f k m ∈ ker(u), i.e. f k m = 0 in M . This implies
that fms = 0 in Mf , as claimed.
Finally, we need to show that the assignment preserves direct sums and tensor prod-
ucts. This follows from the standard facts that
(M ⊕ M ′ )f = Mf ⊕ Mf′ ; (M ⊗R M ′ )f = Mf ⊗Rf Mf′ .
□
Example 12.4. Let X ⊂ An be an affine variety, and let Y ⊂ X be a closed subvariety
such that Y = V (I). We have an exact sequence of Γ(X)-modules
0 / I / Γ(X) / Γ(Y ) = Γ(X)/I / 0
If one takes the associated sequence of OX -modules one recovers the sequence from Equa-
tion (12.1).
Definition 12.6. Let X be an affine variety. A OX -module F is called quasi-coherent
it F = M
f for some Γ(X)-module M . The OX -module F is called coherent if in addition
M is finitely generated over Γ(X).
Example 12.5 (A non quasi-coherent sheaf). Let X be affine and p ∈ X. Define a
presheaf F on X by (
OX (U ) if a ∈ /U
F(U ) =
0 if a ∈ U .
This is actually a sheaf. Observe that F(X) = 0, thus if F were quasi-coherent then F is
the zero sheaf. However, it is clear that F has non-zero local sections. (See [Hartshorne],
[Perrin].)
3Equivalently, Rf is a flat R-module.
66 LEONARDO CONSTANTIN MIHALCEA

Proposition 12.6. There is an equivalence of categories

ModR ≃ ModOX
where ModR is the category of R-modules and R-module homomorphisms and ModOX is
the category of OX -modules and OX -module homomorphisms.
Proof. This follows from definitions of the objects involved and Theorem 12.1. □
Definition 12.7. Let X be an algebraic variety and F a sheaf of OX modules. We
say that F is quasi-coherent (respectively coherent if there exists a covering by affine
S
varieties X = Ui such that for each i, OX (Ui ) = Ri and F|Ui ≃ M
fi for some Ri -module
Mi (respectively a finitely generated Ri -module).
If in addition the modules Mi are free over Ri of the same rank, then F is called
locally free.
Remark 12.3. One may prove that if X is an affine algebraic variety and F a sheaf of
OX modules, then F is (quasi-)coherent) in the sense of Definition 12.6 if and only if it
is (quasi-)coherent in the sense of Definition 12.7.
Remark 12.4. One may prove that there is an equivalence between locally free sheaves
on X and vector bundles on X.
12.3. Quasi-coherent sheaves on projective varieties. The construction of quasi-
coherent sheaves on affine algebraic varieties generalizes to the projective case, provided
the grading is taking into account. This is analogous to the construction of the coordinate
ring of a projective algebraic varieties.
We recall few things from §7.1. Let R = ⊕Rd be a graded ring, and let M = ⊕Md be
a graded R-module. If f ∈ Rd is a homogeneous element, then
m
M(f ) = { k : m homogeneous and deg(m) = k deg(f )}.
f
This is an additive subgroup of Mf ; more importantly, it is a module over R(f ) , the
subring of Rf consisting of fractions of total degree 0.
Let X ⊂ Pn be a projective subvariety with projective coordinate ring R := Γp (X).

Definition 12.8. Let M be a graded R-module. Define the OX -module M

f as follows:
f(D+ ) := M(f ) .
if f ∈ R is homogeneous of degree > 0, define M f

e = OX .
Example 12.7. Of course, R
Remark 12.5. The sheaf M f is quasi-coherent; furthermore, if M is finitely generated,
it is coherent. Indeed, we may cover X by distinguished open sets Df+ . Then observe
that Df+ is affine and that M f| + = M
(Df )
g +
(f ) . (Note that Γ(Df ) = (Γp (X))(f ) , and that

Mg (f ) is an (Γp (X))(f ) -module.)

 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 67

Proposition 12.8. The correspondence M 7→ M f is functorial, it preserves exact se-

quence, and it commutes with direct sums and tensor products.
Sketch of the proof. Essentially the same as the proof of Theorem 12.1, except for the
proof of the statement that it commutes with tensor products. First, if M, N are graded
R-modules, the tensor product M ⊗R N is given the grading
(M ⊗R N )k = ⊕i+j=k Mi ⊗k Nj .
Then one constructs as isomorphism of OX -modules M
f⊗e N
R
e → M^
⊗R N . □
For further use we record the following lemma. (It may also be used for the exactness
statement in the previous proposition.)
Lemma 12.9. Let φ : M → N be an R-module homomorphism of degree 0. If φ :
Mn → Nn is surjective for n ≫ 0 then φ f→N
e:M e is surjective.

Proof. Choose a distinguished open set Df+ ⊂ X, where f is homogeneous of degree > 0.
It suffices to show that φf : M(f ) → N(f ) is surjective. To do this, observe that any
n
element fyr ∈ N(f ) may be written as fyf n
n+r and deg(f y) may be made arbitrarily large.

The hypothesis implies that such elements are in the image of φ; the claim follows from
this. □
Example 12.10. As in the affine case, if Y ⊂ X is closed, the sequence 0 → I →
Γp (X) → Γp (Y ) = Γp (X)/I → 0 gives the natural sequence from Equation (12.1) asso-
ciated to Y .
We now come to the main difference from the affine case: the ability to re-define the
grading on a graded module.
4
L
Definition 12.9. Let R be a graded ring and let M = n∈Z be a graded R-module.
Fix d ∈ Z. The module M (d) is the graded module M , but with the shifted grading
M (d)n := Md+n .
Definition 12.10. Let X ⊂ Pn be a projective algebraic variety and let R := Γp (X) be
] If F is any sheaf of
the projective coordinate ring. Define the sheaf OX (d) := R(d).
OX -modules, define
F(d) := F ⊗OX OX (d).
The sheaves OX (d) are coherent (in fact, locally free of rank 1), and they depend
fundamentally on the grading on Γp (X). Sometimes these are called the Serre twisting
sheaves. Observe also that if M is a graded R-module then by commutativity of tensor
products
]
M f ⊗O OX (d) = M
(d) = M f(d).
X

4We are grading over Z, but this does not make a difference.
68 LEONARDO CONSTANTIN MIHALCEA

Proposition 12.11. Let R := Γp (Pn ) = k[x0 , . . . , xn ] graded by deg xi = 1, and let Rd

be the degree d part of R. Then
(
Rd if d ≥ 0;
Γ(OPn (d)) =
0 otherwise .
In particular, Γ(OPn ) = k.
Proof. Let P be a global section of OPn (d). Its restrictions to Ui := Dx+i are of the form
Pi
di where Pi ∈ R is homogeneous and deg Pi = di + d. W.l.o.g. we may assume that xi
xi
does not divide Pi . The restrictions must agree on the overlaps Ui ∩ Uj = Dx+i xj , giving
equalities of the form
Pi Pj d
di
= dj ⇐⇒ xj j Pi = xdi i Pj ∈ k(x0 , . . . , xn ).
xi xj
Since xi does not divide Pi , nor xj , it follows that (all) di = 0. Therefore P = Pi is a
homogeneous polynomial of degree d. This proves the claim. □
A natural question is whether we can recover the graded R-module out of the quasi-
coherent sheaf F = M
f. The answer is more subtle that in the affine case. First, define
M
Γ∗ (F) := Γ(F(d)).
d∈Z

This may be given a graded R-module structure by: s ∈ Rd = Γ(OPn (d)) and t ∈
Γ∗ (F(d′ )) then one may multiply: s.t ∈ Γ(F(d + d′ )). A geometric fact which currently
is outside our reach is the following (cf. [Hartshorne, Ch.5 and Ex. II.5.9]):
Theorem 12.2. Let M f ≃ F be a quasi-coherent module. Then the following hold:
(a) There exists an isomorphism of OX -modules Γ^
∗ (F) ≃ F
(b) There exists a graded R-module homomorphism M → Γ∗ (F) such that for d ≫ 0,
Md ≃ Γ(F(d)) as k-vector spaces.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 69

13. Hilbert polynomials and Bézout’s theorem

13.1. The Hilbert function and the Hilbert polynomial. Let F be a coherent
OX -module. The global sections of F may be considered as generalized functions on X,
which carry much geometric information. If X is a projective variety, then it may be
proved that Γ(F) is a finite dimensional vector space. (This is not always the case for
affine varieties.) A fundamental problem is to determine Γ(F), or at least its dimension.
In general this is very difficult, but it turns out that the sections of a large enough twist
F(d) are easier to calculate. The reason for this (with details which are beyond the
scope of this course) is that in this case
dim Γ(F(d)) = χ(X; F(d)),
where the latter is the sheaf cohomology Euler characteristic. Various methods allow
for the determination of the Euler characteristic.5
Here we pursue a slightly different approach, based on commutative algebra tech-
niques. It leads to explicit calculations, and algorithms implemented in mathematical
software such as Maple, Mathematica, Macaulay2 etc.
We start with the following definition. As usual R := k[x0 , . . . , xn ] which we regard
as a graded ring with deg xi = 1.
Definition 13.1. Let M be a graded R-module. The Hilbert function is defined by
hM (d) := dimk Md .
Now let X ⊂ Pn be a projective algebraic variety. The Hilbert function of X is defined
by
hX := hΓp (X)
The fundamental theorem about Hilbert functions is that their asymptotics is given
by a single (Hilbert) polynomial.
Theorem 13.1. Let M be a finitely generated graded R-module. Then there exists a
(single) polynomial PM (x) ∈ Q[x] such that
PM (d) = hM (d) for d ≫ 0.
Furthermore, deg PM = dim Vp (Ann(M )) where Ann(M ) is the annihilator
Ann(M ) := {r ∈ R : r.m = 0, ∀m ∈ M }.
Proof. TODO in Spring semester. □
Corollary 13.1. Let X ⊂ Pn be a projective algebraic variety with projective coordinate
ring Γp (X). Then there exists a polynomial PX ∈ Q[d] such that for all d ≫ 0,
PX (d) = dimk Γ(OX (d)).
5Among the most popular is the Atiyah-Bott localization theorem. For an equivariant sheaf on a
variety with a group action, the Euler characteristic is a sum of ‘localized’ data coming from the fixed
loci of the group.
70 LEONARDO CONSTANTIN MIHALCEA

This polynomial is of the form:

deg(X) n
PX (d) = d + lower order terms
n!
where n = dim X, and deg(X) is an integer called the degree of X.
Proof. From Theorem 12.2 it follows that Γ(OX (d)) ≃ (Γp (X))d for d ≫ 0. Since R is a
domain, the annihilator of the R-module Γp (X) = R/Ip (X) is precisely Ip (X). Therefore
the Hilbert polynomial of X has degree dim Vp (Ip (X)) = dim X. □
13.2. Examples of Hilbert polynomials; arithmetic genus. Throughout this sec-
tion, R := k[x0 , . . . , xn ]. We start with the following short exact sequence associated to
a hypersurface in Pn .
Lemma 13.2. Let F ∈ k[x0 , . . . , xn ] be a homogeneous polynomial of degree d > 0.
Then there exists a short exact serquence graded R-modules:

(13.1) 0 / R(−d)
·F / R / R/⟨F ⟩ / 0
If in addition F has no multiple factors, let X = Vp (F ). Then we have a short exact
sequence of coherent OPn -modules:

(13.2) 0 / OPn (−d)

·F / OPn / OX / 0
Proof. The polynomial F gives a degree 0 map of graded modules:
·F /
R(−d) R
This map is clearly injective, and its cokernel is R/⟨F ⟩. By the Nullstellensatz, ⟨F ⟩ =
Ip (X), thus R/⟨F ⟩ = Γp (X). This gives the first short exact sequence of graded R-
modules. Taking the associated short exact sequence of coherent OPn -modules proves
the statement. □
For an indeterminate z define
 
z 1
:= z(z − 1) · . . . · (z − r + 1).
r r!
Example 13.3. The Hilbert polynomial of Pn is
 
z+n 1
PPn (z) = = z n + l.o.t.
n n!
In particular, deg Pn = 1.
Example 13.4 (Two points in P1 ). Consider I = ⟨x0 x1 ⟩ ⊂ k[x0 , x1 ] and let S :=
k[x0 , x1 ]/I. Then (
1 d=0
hS (d) =
2 d>0
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 71

Therefore PS (x) = 2. We may also consider the ideal I = ⟨x20 ⟩ (a double point). The
Hilbert polynomial is the same. In particular the degree of 2 distinct points equals the
degree of one double point.
Example 13.5. While proving Corollary 13.1 we will also show that if dim X = 0 (i.e.
if X is consists of finitely many points) then hX (z) ≡ dimk Γ(X). Here we utilize that
any such X is affine, and one may show that if dim X = 0 then its coordinate ring Γ(X)
is a finite dimensional vector space (i.e. dimk Γ(X) < ∞).
Next we calculate the Hilbert polynomial of a hypersurface.
Proposition 13.6. Let F ∈ k[x0 , . . . , xn ] be a homogeneous polynomial and let M :=
k[x0 , . . . , xn ]/⟨F ⟩.    
z+n z − deg F + n
PM (z) = − .
n n
In particular, if F is reduced (i.e., multiplicity free), then deg Vp (F ) = deg F .
Proof. Taking degree d components in Equation (13.1) we obtain the short exact se-
quence of vector spaces

0 /
(R)d−deg F
·F / Rd / (R/⟨F ⟩)d /0

d+n
− d−deg(F )+n



Then hM (d) = n n
. The result follows after performing the algebraic
calculations. □
Example 13.7 (A conic). Consider the conic C := Vp (y 2 − xz) ⊂ P2 . By Proposi-
tion 13.6,
(z + 2)(z + 1) z(z − 1)
pC (z) = − = 2z + 1.
2 2
Alternatively, one may calculate directly from the definition that

1 d = 0

hC (d) = 3 d = 1

5 d = 3

Again this shows that hC (z) = 2z + 1.

Example 13.8 (A cubic curve in the plane). Consider a cubic curve C := Vp (F ) ⊂ P2
where F is a multiplicity free homogeneous polynomial of degree 3. Again by Proposi-
tion 13.6,
(z + 2)(z + 1) (z − 1)(z − 2)
pC (z) = − = 3z.
2 2
Definition 13.2. Let Y ⊂ Pn be an irreducible projective variety. The arithmetic
genus of Y is defined by
pa (Y ) := (−1)dim Y (PY (0) − 1).
72 LEONARDO CONSTANTIN MIHALCEA

It may be proved that the arithmetic genus is independent of the embedding.

Example 13.9. The arithmetic genus of a plane conic pa (C2 ) = 0; the arithmetic genus
of a plane cubic pa (C) = 1.
13.3. Bézout’s theorem. Recall that R = k[x0 , . . . , xn ]. We start with a generalization
of Lemma 13.2.
Lemma 13.10. Let X ⊂ Pn be a projective variety and let F ∈ k[x0 , . . . , xn ] such
that dim(X ∩ Vp (F )) = dim X − 1. Then there exists a short exact sequence of graded
R-modules:
0 / (R/Ip (X))(− deg F )
·F / R/Ip (X) / R/(Ip (X) + ⟨F ⟩) / 0

Tr X = X1 ∪ . . . ∪ Xr the decomposition into irreducible components. Then

Proof. Write
I(X) = i=1 I(Xj ) and each I(Xj ) is a prime ideal. By Theorem 10.4, the hypothesis
that dim(X ∩ Vp (F )) = dim X − 1 implies that F does not vanish identically on each of
the Xi , i.e. that F ∈
/ I(Xj ), for 1 ≤ j ≤ r. This implies that the map Γp (X) → Γp (X)
obtained by multiplication by F is injective. The rest follows from definitions. □
Theorem 13.2 (Bézout). Let X ⊂ Pn be a projective variety and let F ∈ k[x0 , . . . , xn ]
such that dim(X ∩Vp (F )) = dim X −1 and the ideal Ip (X)+⟨F ⟩ (i.e. the ideal generated
by Ip (X) and F ) is radical. Then
deg(X ∩ Vp (F )) = deg(X) · deg(F ).
Proof. From Lemma 13.10 we obtain the short exact sequence
0 / R/Ip (X)
·F / R/Ip (X) / R/(Ip (X) + ⟨F ⟩) / 0
Since J is radical, it follows that J = Ip (X ∩ Vp (F )) (exercise!) Taking degree d homo-
geneous components implies that
PX∩Vp (F ) (d) = PX (d) − PX (d − deg(F )).
deg X r
Let r := dim X. Write PX (d) = r!
d + cr−1 dr−1 + l.o.t. Then
PX∩Vp (F ) (d) = PX (d) − PX (d − deg(F ))
deg X r
= d + cr−1 dr−1 + l.o.t. − (d − deg(F ))r − cr−1 (d − deg(F ))r−1 − l.o.t.
r!
deg(X)
= (cr−1 + r deg(F ) − cr−1 )dr−1 + l.o.t.
r!
deg(X) deg(F ) r−1
= d + l.o.t.
(r − 1)!
□
Remark 13.1. The hypothesis on J being radical may be removed, but then we need to
consider the scheme X ∩ Vp (F ), associated to the ideal J.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 73

Corollary 13.11. Let F, G ∈ k[x, y, z] be multiplicity free homogeneous polynomials

without common factors. Assume that ⟨F, G⟩ is a radical ideal. Then
deg(Vp (F ) ∩ Vp (G)) = deg(F ) deg(G).
Proof. The hypotheses imply that Ip (Vp (F )) = ⟨F ⟩, Ip (Vp (G)) = ⟨G⟩ and Ip (Vp (F ) ∩
Vp (G)) = ⟨F, G⟩. Then apply Bézout’s theorem. □
Theorem 13.3. Let X ⊂ Pn be an irreducible projective subvariety of codimension r.
Then deg(X) equals the number of points in the intersection X ∩ H1 ∩ . . . ∩ Hr where
Hi ’s are general hyperplanes in Pn .
Here ‘general’ means that there exists an open set U ∈ (Pn )∗ × . . . × (Pn )∗ (r compo-
nents) such that (H1 , . . . , Hr ) ∈ U . To prove this we need two preparatory results.
Lemma 13.12. Let X ⊂ Pn be a closed subvariety. Then there exists a hyperplane
H ⊂ Pn such that no component of X is included in H.
Proof. It suffices to show that there exists a nonempty open set U ⊂ (Pn )∗ such that
for all H ∈ U , X is not included in H. We will abuse notation and identify H with the
linear form which determines it. Let Ip (X) = ⟨F1 , . . . , Fr ⟩ where Fi are homogeneous.
The condition that X ⊂ H amounts to (Fi )|H ≡ 0 for all i, i.e. H|Fi for all i. Since each
Fi can be divided by only finitely many linear factors, the claim follows. □

We also need the following generalization of Lemma 13.12, which is currently outside
our technology.
Theorem 13.4 (Bertini’s theorem). Let X ⊂ Pn be a projective variety. Then there
exists an open set U ⊂ (Pn )∗ such that for all H ∈ U , H ∩ X has codimension 1 and
Ip (X) + H is radical.
Proof of Theorem 13.3. Choose H as in Bertini’s theorem. Then the claim follows by
induction from Bézout’s theorem, since deg(X ∩ H) = deg(X). □

13.4. Degree of the union of two projective varieties. We now turn to the degree
of a union of projective subvarieties.
Lemma 13.13. Let I1 , I2 be two homogeneous ideals in R := k[x0 , . . . , xn ]. Then there
exists a short exact sequence of graded R-modules
(f,f ) (f,−g)
0 / R/(I1 ∩ I2 ) / R/I1 ⊕ R/I2 / R/(I1 + I2 ) / 0

where the first morphism is the diagonal, and the second sends (f, g) ∈ R/I1 ⊕ R/I2 to
f − g ∈ R/(I1 + I2 ).
Proof. Clear from definitions. □
74 LEONARDO CONSTANTIN MIHALCEA

Proposition 13.14. Let X1 , X2 ⊂ Pn be two closed subvarieties of the same dimension

r, and such that dim X1 ∩ X2 < r. Let I1 = Ip (X1 ) and I2 = Ip (X2 ). Assume that I1 ∩ I2
and I1 + I2 are radical (equivalently, I1 ∩ I2 = Ip (X1 ∪ X2 ) and I1 + I2 = Ip (X1 ∩ X2 ).
Then the following hold:
deg(X1 ∪ X2 ) = deg X1 + deg X2 .
Proof. Recall Ip (X1 ∪ X2 ) = I(X1 ) ∩ I(X2 ). From Lemma 13.13 we get a short exact
sequence
0 / Γp (X1 ∪ X2 ) / Γp (X1 ) ⊕ Γp (X2 ) / Γp (X1 ∩ X2 ) / 0
Then PX1 + PX2 = PX1 ∩X2 + PX1 ∪X2 , and since deg PX1 ∩X2 < r, the leading term of
PX1 ∪X2 must be the sum of the leading terms of PX1 and PX2 . □

END OF SEMESTER 1
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 75

14. Start of semester 2: goals

The goal in semester 2 is to introduce basic notions in intersection theory, and work
out some applications from enumerative geometry. For this, we will need to work with
schemes, define flat and proper morphisms, Weil and Cartier divisors, line bundles
and vector bundles, and Chern classes of vector bundles. We will use the projective
spaces, and, more generally, Grassmannians, as the basic models of schemes which will
be utilized to set up and solve problems in enumerative geometry. Examples of such
problems are:
• Find the numbers of lines meeting 4 given lines in P3 . (Answer: 2.)
• Find number of conics meeting 8 general lines in P3 . (Answer: 92.)
• Find number of lines on a general cubic surface in P3 . (Answer: 27.)
All these may be answered just by analyzing the intersection theory of Gr(2, 4), the
Grassmannian of subspaces of dimension 2 in C4 . (This is the same as the space
parametrizing lines in P3 .)
Unlike the first semester, where (almost) all statements have been given a complete
proof, we will need to take some statements by heart. However, I aim to give rigorous
definitions of all notions we are going to utilize.
Throughout the notes from now on, k will denote an algebraically closed field. As
usual, all rings are commutative with 1 and all rings are Noetherian.
76 LEONARDO CONSTANTIN MIHALCEA

15. Schemes
Even if one wishes to work with algebraic varieties, the appearance of schemes is
unavoidable. For example:
• the intersection of a conic and a line tangent to it is a double point.
• Consider a degree 2 morphism f : P1 → P1 , e.g. [x : y] 7→ [x2 : y 2 ]. Then a
general fibre has 2 points, but there will be fibres which are double points.
To understand better what we are generalizing, recall that an algebraic variety is (by
definition) covered by affine algebraic varieties, each of which being inside of some affine
space An . Recall the correspondence:
{Affine algebraic varieties X ⊂ An } ↔ { radical ideals I ⊂ k[x1 , . . . , xn ]}
We may generalize this correspondence in several directions:
(1) We may allow the ideals I to be non-radical, leading to non-reduced schemes.
(2) We do not have to consider ideals in a polynomial ring, leading to schemes which
are non-necessarily embedded in an affine space.
(3) We may remove the requirement that I (or the polynomial ring) is finitely gen-
erated. This leads to the theory of schemes of infinite type. We will not work in
this generality.

15.1. Spectrum of a ring. In this section we largely follow Mumford’s treatment from
[Mum99].
Definition 15.1. Let R be a ring. The spectrum of R is
Spec(R) = {p ⊂ R : p is a prime ideal }
Example 15.1. If R = K is a field then Spec(K) = (0). More generally, (0) ∈ Spec(R)
whenever R is an integral domain.
We make Spec(R) into a topological space by defining its closed sets to be of the form
V (I) := {p ∈ Spec(R) : p ⊃ I}
where I ⊂ R is an ideal. Note that V (I) may also be defined by requiring that φf (p) = 0
for all f ∈ I. This topology is again called the Zariski topology.
Proposition 15.2. The closed sets form a topology on Spec(R) such that the corre-
sponding topological space is paracompact. Furthermore, the following hold for any ideals
I, J, Is ⊂ R:
S T
• V ( s Is ) = s V (Is );
• V (I ∩ J) = V (I) ∪ V (J)
Proof. Homework! □
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 77

As usual, if f ∈ R, we denote by V (f ) = V (⟨f ⟩).

Spec(R) will be the underlying set of the points of an affine scheme. The points in the
previous notion of affine variety correspond to the maximal ideals in R, but in this new
context we have many more points than before. The interpretation of p ∈ Spec(R) is
that it corresponds to a (closed) subscheme of Spec(R), namely V (p), and if p is smaller,
then the subscheme is larger. In this context the following notion is useful:
Definition 15.2. Let Z ⊂ Spec(R) be an irreducible closed subset. A point z ∈ Z is
called a generic point of Z if {z} = Z, i.e., for any open set U ⊂ Z, z ∈ U .
Example 15.3. If R is an integral domain, its only generic point is (0). More generally,
if z ∈ Spec(R), then (homework!)
{z} = {p ∈ Spec(R) : p ⊃ z} = V (z).
In particular, the only closed points in Spec(R) are the maximal ideals, and, furthermore,
Spec(R) is not T1 .
Example 15.4. TODO. Describe Spec(C[x, y]) and the generic points of irreducible
subsets.
The following gives an idea of what to expect when two closed sets are equal. Recall
that the radical of an ideal is the intersection
√ \
I= p.
p⊃I;p is prime
√
Lemma 15.5. Let I, J be ideals in R. √Then√V (I) = V ( I). If in addition I, J are
radical then V (I) = V (J) if and only if I = J.
√
Proof. The equality V (I) = V ( I) follows from the definition of the radical. This also
implies that V (I) = V (J) for I, J radical. Conversely, if V (I) = V (J) then
√ \ \ √
I= p= p = J.
p∈V (I) p∈V (J)

□
Proposition 15.6. Let p ∈ Spec(R). Then the following hold:
(a) V (p) is irreducible and p is its unique generic point.
(b) Every irreducible closed subset of Spec(R) is of the form V (p) for some p ∈
Spec(R).
Proof. Write V (p) = Z1 ∪ Z2 some closed sets Z1 , Z2 . If p ∈ Zi then {p} ⊂ Zi = Zi , thus
equality must occur. If there is another generic point q then since {p} = {q} it follows
that p ⊂ q and q ⊂ p, thus p = q. This proves (a).
Now take Z = V (I) ⊂ Spec(R)
√ closed and irreducible, where I ⊂ R is an ideal.
By Lemma 15.5 V (I) = V ( I), therefore we may assume that I is radical. We will
78 LEONARDO CONSTANTIN MIHALCEA

prove that I is prime. Take f, g ∈ R such that f g ∈ I, and define B := I + ⟨f ⟩ and

C := I + ⟨g⟩.
We claim that I = B ∩ C. Clearly the inclusion ⊂ holds. To prove the reverse
inclusion, take h := a1 + αf = a2 + βg ∈ B ∩ C where a1 , a2 ∈ I and α, β ∈ R. Then
h2 = (a1 + αf )(a2 + βg) = a1 a2 + a1 βg + a2 αf + αβf g ∈ I.
Since I is radical, it follows that h ∈ I, which proves the claim.
From the claim, V (I) = V (B)∪V (C), and since V (I) is irreducible it follows (w.l.o.g.)
that V (I) = V (B). Then
\ \ √
B⊂ p= p = I = I,
p⊃B p⊃I

therefore B ⊂ I, hence thus f ∈ I, which proves that I is prime. □

The distinguished open set in Spec(R) associated to f is defined by
Df := {p ∈ Spec(R) : f ∈/ p} = Spec(R) \ V (f ).
S
Then Spec(R) \ V (I) = f ∈I Df , showing that the distinguished open sets form a basis
for the topology of Spec(R).
We will also need the following:
S
Lemma 15.7. (a) Spec(R) = s∈S Dfs if and only if the ideal ⟨fs : s ∈ S⟩ = R.
(b) Spec(R) is quasi-compact, i.e. from any cover we may extract a finite subcover.
Proof. Let J be the ideal generated by fs . If J = S R, then for any p ∈ Spec(R), one of
the fs ’s cannot be in p. Conversely, if Spec(R) = s∈S Dfs then each prime ideal must
be in some Dfs , i.e. not all the fs ’s are contained
S in any single prime ideal. Then J = R.
For part (b), take a cover Spec(R) = Dfs . Then fs generate R, and there must be
finitely many of them which still generate; these give the finite subcover. □
In general one may prove that Df is quasicompact. But arbitrary open sets U may
not be.
We also record the following properties of the distinguished open sets:
(a) Df ∩ Dg = Df g (homework!).p
(b) Df ⊃ Dg if and only if g ∈ ⟨f ⟩.
p T p
Proof. We have ⟨f ⟩ = f ∈p p, therefore g ∈ ⟨f ⟩ iff g ∈ p for any prime p
with f ∈ p. The statement follows by negating this. □
p p
Observe that the second statement implies that Df = Dg if and only if ⟨f ⟩ = ⟨g⟩.
Remark 15.1. The inclusion Df ⊃ Dg gives a partial order ≤1 on the elements of R
by f ≤1 g iff Df ⊃ Dg . This is a directed set, in the sense that for any f, g ∈ R,
Df g = Df ∩Dg , thus f, g ≤1 f g. By (b) above, this is isomorphic (as a directed set) to the
partial order on R given by f ≤2 g iff g = f f ′ for some f ′ ∈ R. Then (R, ≤1 ) ≃ (R, ≤2 )
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 79

as partial ordered sets, and we identify the two orderings. The limits over subsets of
elements in R will be taken with respect to these limits.
15.2. The structure sheaf of Spec(R). In this section we will define the structure
sheaf OSpec(R) . We will be associating Df 7→ Rf for each f ∈ R, and we need to check a
number of compatibilities.
Lemma 15.8. Let f, g, h ∈ R.
(a) If Dg ⊂ Df then there is a natural (restriction) map Rf → Rg . Furthermore, if
Dh ⊂ Dg ⊂ Df , then there is a commutative diagram given by restriction maps
Rf / Rg


Rh
(b) The assignment Df 7→ Rf is well defined, i.e., Rf ≃ Rg if Df = Dg .
(c) Let p ∈ Spec(R). Then the localization Rp is the direct limit
Rp = lim Rf = lim Rf .
−→ −→
f ∈R\p p∈Df
p
Proof. If Dg ⊂ Df then g ∈ ⟨f ⟩, i.e. g n = hf for some positive integer n and h ∈ R.
Then the map Rf → Rg is given by
a ah
7→ n .
f g
If Df = Dg then we get maps Rf → Rg and Rg → Rf which are inverse to each other.
(Check!) We leave part (c) as an exercise. □
We will see that the localization Rp is the stalk at p of the structure sheaf OSpecR and
part (c) is the analogue of the fact that germs of functions a point p have representatives
in a neighborhood of p.
We are ready to define the structure of Spec(R). We Q utilize the same idea as in
Lemma 5.5. The field K from this lemma is replaced by p∈Spec(R) Rp .
Definition 15.3. The structure sheaf OSpecR is Q defined as follows. Let U ⊂ Spec(R) be
open. Then OSpecR (U ) is the set of tuples (sp ) ∈ p∈U Rp such that for each q ∈ U there
exists a distinguished open set Dfi ⊂ U with the property that for all p ∈ Dfi , and all
(sp )p∈Dfi , there exists si ∈ Rfi such that sp is obtained by the restriction Rfi → Rq .
If V ⊂ U are open sets, then the restriction map ρU V : OSpecR (U ) → OSpecR (V ) is the
natural projection.
It is easy to check that each OSpecR (U ) is a ring.
Theorem 15.1. (a) OSpecR is a sheaf.
(b) OSpecR (Df ) = Rf .
(c) The stalk (OSpecR )p = Rp .
80 LEONARDO CONSTANTIN MIHALCEA

In order to prove the theorem we will need two lemmas.

S
Lemma 15.9. Assume Df = i Dfi for some fi . If g ∈ Rf has image 0 in all Rfi then
g = 0.
Proof. Write g = fbn and define the annihilator ideal A = {c : cb = 0}. The following
statements are equivalent:
• g = 0 in Rf ;
m
• There√ exists m such that f · b = 0 in R;
• f ∈ A;
• If a prime ideal p ⊃ A then f ∈ P .
Assume now that g ̸= 0. Then weS may choose a prime ideal p ⊃ A such that f ∈/ p, i.e.,
p ∈ Df . Since f ∈
/ p and Df = Dfi we may choose i such that p ∈ Dfi . Consider the
commutative diagram
Rf / Rf
i


Rp
The image of g in Rp factors through Rfi , thus it is equal to 0 ∈ Rp . Then there exists
c ∈ R \ p such that cb = 0, i.e. c ∈ A. But then c ∈ A ⊂ p by assumption, which is a
contradiction. □
S
Lemma 15.10. Assume Df = i Dfi for some fi , and assume we have elements gi ∈ Rfi
such that the restriction to Rfi fj coincide. Then there exists g ∈ Rf such that the
restriction to each Rfi is gi .
Proof. Similar to proof of the gluing property in Equation (5.3) above; see [Mum99, p.
70, Lemma 2]. □
We are now ready to prove Theorem 15.1.
Proof of Theorem 15.1. To prove (a), by Lemma 5.5, it suffices to show the restriction
and gluing property for the basis of open sets. It is clear that the restriction maps ρU V
satisfy the required compatibilites. The gluing property follows from Lemma 15.10.
To prove (b), there is a natural map
Y Y
Rf → OSpecR (Df ) ⊂ Rp = Rp
p∈Df f ∈p
/

obtained by projections Rf → Rp . This map is injective by Lemma 15.9 and surjective

by Lemma 15.10.
To calculate the stalk at a point p, we may consider the elements of the basis of open
sets which contain p. Then
(OSpecR )p = lim OSpecR (Df ) = lim Rf = Rp .
−→ −→
p∈Df f ∈p
/
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 81

□
Remark 15.2. Note that the stalks of OSpecR are Rp , and these are local rings. Then
(SpecR, OSpecR ) is called a locally ringed space.
Remark 15.3. The fact that we have stalks at non-closed points gives an additional
structure. If p1 ⊂ p2 are two prime ideals in R, then p2 ∈ {p1 }, therefore any neighbor-
hood U of p2 contains p1 . Then
Rp2 = (OSpecR )p2 = lim OSpecR (U ) → lim OSpecR (V ) = (OSpecR )p1 = Rp1 .
−→ −→
p2 ∈U p1 ∈V

In particular, germs of functions at closed points restrict to germs of functions at generic

points of larger closed subsets. Algebraically, this corresponds to the natural map Rp2 →
Rp1 .
Compare this to the fact that in the previous world of varieties, we had natural maps
between the stalk at a closed point x0 ∈ X and the function field on some irreducible
variety X. The map was simply fg 7→ fg where g(x0 ) ̸= 0 for f, g regular functions in
some neighborhood of x0 .
We may still regard elements in R as functions, except that not in the same field. If
p ∈ Spec(R) then define recall that Rp is a local ring, with maximal ideal pRp . Define
the residue field k(p) = Rp /pRp . Since quotient and localization commute [AM69,
Cor. 3.4], k(p) = (R/p)⟨0⟩ (the fraction field of R/p).
Then any r ∈ R may be regarded as a function
[
φr : Spec(R) → k(p); p 7→ image of f under R → Rp /pRp = k(p).
p∈Spec(R)

Observe that if R = k[x1 , . . . , xn ]/I and p is maximal then (e.g., by Nullstellensatz)

k(p) = k. This recovers our previous point of view that elements in the coordinate ring
of an affine variety may be regarded as functions to the ground field. But if R is an
integral domain, then ⟨0⟩ is prime in R, and k(⟨0⟩) = F rac(R), the fraction field of R.
However, assume that an elementTr ∈ R satisfies φr (p) = 0 for all p ∈ Spec(R). Then
r ∈ p for all p ⊂ R prime, thus r ∈ p∈Spec(R) p = N il(R) (the nilradical of R). In other
words, r must be nilpotent.
Example 15.11. Let R = k[x]. Then
(
k if p = ⟨x − a⟩, a ∈ C;
k(p) =
k(x) if p = 0.
Example 15.12. Consider R = k[x]/⟨x2 ⟩. Then Spec(R) = {⟨x⟩} consists of a single
point, but R is not a field. Observe that the residue field is k(⟨x⟩) = k. The functions 0
and x in R have the same value on Spec(R), but they are different as functions.
The global functions on SpecR are just R = k[x]/⟨x2 ⟩, i.e. functions f ∈ k[x] modulo
functions vanishing on the second order at x = 0. We will see later that the natural
82 LEONARDO CONSTANTIN MIHALCEA

projection map k[x] → R gives an embedding SpecR ,→ A1 . At the level of functions,

the restriction of f ∈ k[x] to R remembers the information about its value of f at 0 and
also its derivative. From this point, one may interpret embeddings of Spec(R) as a point
together with a tangent direction at it.
Proposition 15.13. Let R be a ring and f ∈ R. Then there is an isomorphism of
ringed spaces
((SpecR)f , O(SpecR)f ) ≃ (Spec(Rf ), OSpecRf ),
where (Spec(R))f is the distinguished open set corresponding to f .
For the proof, we need the following lemma (cf. [AM69, Prop. 3.11(iv)]:
Lemma 15.14. Let R be a ring an let S ⊂ R be a multiplicative set. Then there is a
one to one correspondence between the prime ideals in RS and the prime ideals q ⊂ R
such that S ∩ q = ∅. The correspondence is given by extension and contraction of ideals:
q ⊂ R 7→ S −1 q ⊂ S −1 R; q ⊂ S −1 R 7→ i−1 (e
e q),
where i : R → S −1 R is the natural morphism.
Proof of Proposition 15.13. By Lemma 15.14, there is a bijective map between (SpecR)f →
Spec(Rf ) sending p 7→ pRf . Furthermore, for any g ∈ R, the same lemma implies that
there is a bijection
(SpecR)f g ≃ (SpecRf )g .
(Check!) This bijection induces a ring isomorphism at the level of sections over these
distinguished open sets, because both rings equal to Rf g , by Theorem 15.1(b). This
implies that the bijection (SpecR)f → Spec(Rf ) is a homeomorphism. □
15.3. Preschemes and morphisms of preschemes.
Definition 15.4. An affine scheme is a ringed space isomoprhic to (SpecR, OSpecR )
for some ring R. S
A prescheme is a ringed space (X, OX ) such that there exists an open cover X = Xi
such that (Xi , (OX )|Xi ) is isomorphic to an affine scheme.
Note that if X is a prescheme and U ⊂ X is openSthen (U, (OX )|U ) is also a prescheme.
Indeed, cover X by affine schemes Xi . Then U = (U ∩ Xi ) is an open cover, and each
U ∩ Xi is open in the affine scheme Xi . By Proposition 15.13 it follows that U ∩ Xi is
covered by some distinguished open sets in Xi , therefore U ∩ Xi is a prescheme, and so
is U .
Morphisms of preschemes are the same as morphisms of locally ringed spaces. To
define these we recall few facts from commutative algebra.
A ring (R, m) is local if it contains a unique maximal ideal m. The most important
example for us is Rp , where p ⊂ R is a prime ideal. A homomorphism of local rings
φ : (R, m) → (S, n) is called a local ring map if
(a) φ(m) ⊂ n, or, equivalently,
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 83

(b) φ−1 (n) = m.

(The nontrivial implication (a) ⇒ (b) uses that φ−1 (n) is a prime ideal containing the
maximal ideal m, thus they must be equal.)
Example 15.15. Let φ : R → S be a ring homomorphism. Let q ⊂ S be a prime ideal
and p := φ−1 (q). Then the induced map φp : Rp → Sq is a local ring map.
Furthermore, in this case φ also induces a map k(p) → k(q) between the residue fields.
Definition 15.5. Let (X, OX ) and (Y, OY ) be two preschemes. Then a morphism f :
(X, OX ) → (Y, OY ) of preschemes is a morphism of locally ringed spaces, i.e.:
• A continuous map f : X → Y ;
• A collection of ring homomorphisms for each U ⊂ Y open
fU∗ : OY (U ) → OX (f −1 (U ));
This data satisfies the following conditions:
• they are compatible with restrictions, i.e., for any V ⊂ U , the following diagram
commutes:
f∗ U
OY (U ) −−−→ OX (f −1 (U ))
 
resf −1 (U )f −1 (V )  resU V y

y
f∗
V
OY (V ) −−−→ OX (f −1 (V ))
• Let f (p) = q. By taking limits, the homomorphims fU∗ induce a homomorphism
fp∗ : lim OY (U ) = OY,q → lim OX (f −1 (U )) → OX,p .
−→ −−1
→
q∈U p∈f (U )

We require that this is a local ring map, i.e.,

(fp∗ )−1 (pOX,p ) = qOX,q ⇐⇒ (fp )∗ (qOX,q ) ⊂ pOX,p .
The definition of morphisms is slightly different from the one for algebraic varieties
from the previous semester. A key distinction is that if X is an algebraic variety, the
structure sheaf consisted of rational functions a : U → k, where k is a fixed field. For a
continuous map f : X → Y and U ⊂ Y open, there is a diagram
f
f −1 (U ) / U
a
f ∗ (a)=a◦f
# 
k
Therefore one can easily pull back rational functions. However, in the case of preschemes,
the pull backs are now part of the data.
84 LEONARDO CONSTANTIN MIHALCEA

Proposition 15.16. For any rings R, S there is a bijection

Scheme morphisms f : (SpecR, OSpecR ) → (SpecS, OSpecS ) ←→
ring homomorphisms f ∗ : S → R
This gives a contravariant equivalence of categories
Sch → Rings
between the category of affine schemes and the category of (commutative) rings.
Proof. If f : (SpecR, OSpecR ) → (SpecS, OSpecS ) is a morphism of affine schemes then by
definition and Theorem 15.1 there is a ring homomorphism f ∗ : OSpecS (SpecS) ≃ S →
OSpecR (SpecR) ≃ R.
Conversely, assume that we are given a ring homomorphism φ : S → R. Then for any
prime ideal p ∈ R, q := φ−1 (p) is a prime ideal in S. (Check!) The map φ induces a
map f : SpecR → SpecS defined by p → φ−1 (p). This map is continuous because
f −1 (V (J)) = {p : φ−1 (p) ⊃ J} = {p : p ⊃ φ(J)} = V (⟨φ(J)⟩).
For q = φ−1 (p), φ also induces a local ring map φp : Sq → Rp . The sections over open
sets U ⊂ SpecS and V ⊂ SpecR are tuples in the local rings Rp and Sq , and one may
check that the homomorphisms φp glue to give the required ring homomorphisms fU∗ .
Finally, one needs to check that the two constructions are inverse to each other - this
is left as an exercise. □
Example 15.17. All morphisms of varieties from the previous semester are morphisms
of locally ringed spaces. Homework!
Example 15.18. Let R = k[x]⟨x−a⟩ , i.e the local ring at a closed point in A1k . (This
is an example of a discrete valuation ring; cf. [AM69, Ch. 9]. More generally, R could
the local ring at a closed point of a nonsingular curve - in this case the maximal ideal is
principal).
Then Spec(R) = {0, m}, m is closed, and 0 is open and dense in Spec(R). Let K =
R(0) be the fraction field of R. We wish to define a morphism f : Spec(K) → Spec(R):
• At the level of points, send f (0) = 0. This is continuous, as (0) ∈ SpecR is open.
• We need to define the ring homomorphisms fU∗ : OSpec(R) (U ) → OSpec(K) (f −1 (U ))
at the level of rings of sections. Observe that there are two open sets:
– U = Spec(R). Then OSpec(R) (U ) = R and define
∗
fSpecR = R → K; r 7→ r.
– U = (0). Then U = Dr is a distinguished open set for any 0 ̸= r ∈ m. In
this case
fD∗ r = Rr → K; r 7→ r.
In other words, this is induced by the natural inclusion R ,→ K.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 85

At the level of stalks, we need to check that the map

OSpec(R),0 = R⟨0⟩ → OSpec(K),0 = K
is a local ring map. Observe that since m is a principal ideal R(0) = Rr for any 0 ̸= r ⊂ m
(since any element outside m is already invertible). Then the map R(0) → K is the
identity map, it sends 0 7→ 0 and the claim follows: this is a morphism of locally ringed
spaces.
We may also define the map f : Spec(K) → Spec(R) by requiring that f (0) = m and
given by the inclusion R ,→ K at the level of sheaves. This is a morphism of ringed
spaces, but not of locally ringed spaces. (Since the inclusion Rm → K is not a local
ring map.)
One may also prove a more general version of Proposition 15.16 (cf. [Mum99, §2,
Thm.1]:
Theorem 15.2. Let X be any prescheme, and R any ring. Then there is a bijection
HomSch (X, Spec(R)) ←→ HomRings (R, OX (X)),
which associates to any f : X → Spec(R) the ring homomorphism f ∗ : R ≃ OSpecR (Spec(R)) →
OX (X).
Applying this theorem for R = OX (X) and the identity map R → R, we obtain the
following corollary:
Corollary 15.19. (a) There exists a canonical morphism X → Spec(OX (X)).
(b) Let S be any ring and fix a morphism f : X → Spec(S). This morphism factors
through the previous morphism:
f
X / Spec(S)
7

%
Spec(OX (X))
Remark 15.4. In some situations OX (X) is very small. E.g., if X is a projective
k-variety, then OX (X) = k (This will be shown later.) Therefore it follows that any
morphism from a projective scheme to an affine scheme must be constant!
Corollary 15.20. Spec(Z) is a final object in the category of preschemes, i.e. for any
prescheme X there is a unique morphism X → Spec(Z).
Corollary 15.21. Let X be a prescheme and let x ∈ X be an arbitrary point with
residue field k(x). Then there exists a canonical morphism ix : Spec(k(x)) → X sending
0 7→ x.
Proof. Take U = Spec(R) to be an open affine neighborhood of x. There is a natural
inclusion U ,→ X, and using this we may assume X = Spec(R) is affine. Let px ⊂ R be
86 LEONARDO CONSTANTIN MIHALCEA

the prime ideal corresponding to x. Then the morphism ix is given by the natural map
R → Rpx → k(x). □
To relate preschemes to the algebraic varieties from the previous semester we give the
more general definition:
Definition 15.6. Let R be a ring. Then a prescheme X over R is a morphism πX :
X → Spec(R). A morphism f : X → Y is a morphism of preschemes over R if there is
a commutative diagram
f
X / Y

πX πY
$ {
Spec(R)
There is a category of preschemes and morphisms over R. There is an analogue of
Proposition 15.16 which establishes an equivalence between the category of affine R-
preschemes and morphisms to ring homomorphisms of R-algebras. There is a further
generalization of Theorem 15.2 to the context of R-preschemes. The most important
situation is when R = k and the sheaves of sections satisfy a finite generation condition.
In this case, the schemes over k correspond to the algebraic varieties we studied in the
prior semester.
We also record the following useful lemma:
Lemma 15.22. Let X be a prescheme and Z ⊂ X a non-empty closed irreducible subset.
Then there exists a unique point z ∈ Z such that {z} = Z. (This is called the generic
point of Z.)
Proof. Let U be any affine open scheme which intersects Z. Then Z ∩ U must be dense
in Z, and any dense point in Z must be in Z ∩ U . Observe that the latter is closed
and irreducible in the affine scheme U . By Proposition 15.6, there exists a unique point
dense in Z ∩ U (the generic point), and this finishes the proof. □
Definition 15.7. Let R be a ring. The n-dimensional affine space over R is the affine
scheme
AnR = Spec(R[x1 , . . . , xn ]).
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 87

16. Products of preschemes

Our next goal is to define schemes; these will be separated preschemes, i.e. preschemes
X such that the image of the diagonal morphism X → X × X is closed. For this, we
first need to discuss products of preschemes.
We adapt the notation and defintions from §9.1 above to the category of preschemes.
If f : X → Z and g : Y → Z are morphisms of preschemes, a product is a prescheme
X ×Z Y equipped with two morphisms pX : X ×Z Y → X and pY : X ×Z Y → Y such
the following universality property is satisfied:
• For any prescheme W and morphisms pX : W → X and pY : W → Y so that
f ◦ pX = g ◦ pY there exists a unique morphism p : W → X ×Z Y such that
πX ◦ p = pX and πY ◦ p = pY .
W
pX
p
$
X ×Z Y /& X
pY πX
πY f
 g 
Y /Z
The product X ×S Y , if it exists, is unique up to an isomorphism commuting with the
projections. The following is the main result of this section:
Theorem 16.1. (a) Let R, S be T -algebras. Then SpecR ×SpecT Spec(S), together with
natural projections, exists in the category of T -preschemes and it is isomorphic to Spec(R⊗T
S).
(b) Given any preschemes pX : X → Z and pY : Y → Z, a fibre product X ×Z Y
exists.
Proof. In the category of affine schemes, part (a) follows from the universality property
of tensor products (see Example 9.2), together with the correspondence between rings
and affine varieties from Proposition 15.16. The proof of (a) in the larger category of
preschemes follows by reducing to the affine case, utilizing that any morphism X →
Spec(R ⊗T S) factors through Spec(OX (X)) → Spec(R ⊗T S); see Corollary 15.19.
Part (b) follows by covering X and Y by affine schemes, taking fibre products of these,
then proving that they ‘glue’ to a prescheme X ×Z Y . For details please consult e.g.,
[Har77, Thm. 2.3.3] or [Mum99, Thm. II.2.3(b)]. □
16.1. Examples of fibre products. Some important situations of fibre products are
the following.
16.1.1. Fibres of morphisms.
Definition 16.1. Let f : X → Y be a morphism of preschemes, and a (possibly not
closed) point y ∈ Y with residue field k(y) and canonical morphism iy : Spec(k(y)) → Y .
88 LEONARDO CONSTANTIN MIHALCEA

The fibre f −1 (y) is the prescheme k(y) ×Y X defined by the fibre diagram below:

f −1 (y) ≃ k(y) ×Y X / X
f
 iy 
k(y) / Y
This is naturally a prescheme over the residue field of y.
Example 16.1. Let f : A1C → A1C defined by f (x) = x2 . Let a ∈ A1C be a closed point,
corresponding to the maximal ideal ma = ⟨t − a⟩ ⊂ C[t]. The residue field is k(a) =
C[x]/ma ≃ C. Note that f is induced by the C-algebra homomorphism f ∗ : C[t] → C[x]
defined by t 7→ x2 .
Then
f −1 (a) ≃ Spec(C[x] ⊗C[t] C[t]/ma ) = Spec(C[x]/⟨x2 − a⟩)
The prescheme f −1 (a) consists of two reduced points if a ̸= 0; if a = 0 then C[x]/⟨x2 ⟩ is
non-reduced, and the fibre f −1 (0) is a single non-reduced point.
Consider now the generic point 0 ∈ Spec(C[t]). The residue field is C[t](0) = C(t) (the
fraction field of C[t]). Then
f −1 (⟨0⟩) = Spec(C[x] ⊗C[t] C(t)) = Spec(C[x]C[x2 ]\0 ).
(This uses that for a multiplicative set S in a ring R and an R-module M , we have
M ⊗R (S −1 R) ≃ S −1 M as S −1 R-modules; see [AM69, Prop. 3.5].) Note that C[x]C[x2 ]\0
is a C(t)-vector space of dimension 2, with basis 1, x.
16.1.2. Base change. More generally, consider the fibre diagram
Y ′ ×Y X / X
f′ f
 
Y′ / Y
This is called a base change, or base extension (from Y to Y ′ ). It is the natural
algebraic analogue of the pull-back of a vector bundle from topology. It also leads to
the notion of extending the field of a definition. For example, for any ring R,
AnR = AnZ ×Spec(Z) Spec(R).
This follows because R ⊗Z Z[x1 , . . . , xn ] = R[x1 , . . . , xn ]. In other words we’re base
changing from coefficients in Z to coefficients in an arbitrary ring.
One important property of the base extension is that f, f ′ have the same fibres, and
have the same geometric properties (preserves flatness, properness, etc.)
Remark 16.1. The fibre product X ×Z Y depends on the base prescheme Z, and in
general there is no injective map to the set theoretic product. For instance, Spec(C ×C
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 89

C) = Spec(C) is a point. However,

G
Spec(C × C) = Spec(C ×R C) = Spec(C) Spec(C)

is a disjoint union of two points. (The two points are given by the prime ideals generated
by (1, 0) and (0, 1); note also the idempotent 1⊗1+i⊗i
2
in the tensor product, showing
that Spec(C ⊗R C) must be disconnected.) Thus the base extension may not preserve
connectedness.
However, the set theoretic and fibre product will coincide for the reduced, irreducible
preschemes of finite type over algebraically closed fields. (See §16.2 below or [Mum99,
Ch. II, §3, Prop. 5].)

16.1.3. The diagonal map and intersections. Assume we are given a morphism f : X →
Y . The diagonal ∆ : X → X ×Y X is the unique morphism given by the universality
property of fibre products:

(16.1) X
id
∆
$ &/
X ×Y X X
id
f
 f 
X / Y

Now take Y = Spec(k) (k a field) and consider X1 , X2 ⊂ X closed sub(pre)schemes with

inclusions i1 : X1 ,→ X and i2 : X2 ,→ X. (Technically, we did not talk about how these
inclusions are defined, but we have some expectations from the 1st semester; this will
be done later.) The intersection X1 ∩ X2 is the prescheme given by the fibre product:

X1 ∩ X2 / X1 ×Spec(k) X2
i1 ×i2
 
X
∆ / X ×X
Equivalently, one may also define it from the fibre diagram

X1 ∩ X2 / X1
i1
 i2

X2 / X

Of course, the intersection X1 ∩ X2 may not have good properties (e.g. expected
dimension, singularities etc.)
90 LEONARDO CONSTANTIN MIHALCEA

16.2. Preschemes and algebraic varieties.

Definition 16.2. A prescheme X over R is of finite type over R if X is quasi-compact
and for any U ⊂ X open, affine, OX (U ) is a finitely generated R-algebra.
Warning: In general, this does not imply that OX (X) is a finitely generated R-
algebra.
Definition 16.3. A prescheme X if reduced if for any U ⊂ X open, OX (U ) has no
Equivalently (check!) the stalks OX,p have no nilpotents, or there
nilpotent sections. S
exists a cover X = Ui with Ui ≃ Spec(RI ) open affine and Ri are reduced rings.
The notion of a variety in the following theorem is the same as the one from Def-
inition 6.2. In particular, at this time we do not require that an algebraic variety is
separated.
Theorem 16.2. Let k be an algebraically closed field. Then there is an equivalence of
categories between:
(a) Category of reduced, irreducible, preschemes of finite type over k, and k-morphisms;
(b) Category of irreducible algebraic varieties over k and morphisms of these.
Furthermore, if X, Y are preschemes over k as above, then the fibre product X ×Spec(k)
Y in the category of preschemes coincides with the product X × Y in the category of
algebraic varieties.
Rough idea of proof - see [Har77, Prop. II.2.6] . One needs to define functors in both
directions.
(a) ⇒ (b): If X is a prescheme, take X to be the set of closed points of X . Then
one may prove that if f : X → Y is a morphism of k-varieties, it takes closed points to
closed points. (Homework!) Therefore it will also define a map X → Y , which one
may check it will be a morphism of varieties.
(b) ⇒ (a) Take X be a variety. We need to define the points of a scheme X . We
take X to be the points in X union the symbols [Z] for any Z ⊂ X which is closed and
irreducible. If f : X → Y is a morphism, then we send x ∈ X to f (x) ∈ Y , and [Z] to
[f (Z)]. Again one checks that this satisfies the required properties. □
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 91

17. Proj(R) and projective schemes

17.1. Proj(R) and its structure sheaf.
Definition 17.1. Let R = d≥0 Rd be a graded ring, and denote by R+ = d>0 Rd the
L L
ideal of elements of positive degree. We define by
Proj(R) = {p ⊂ Spec(R) : p is homogeneous and R+ ⊈ p}.
For I ⊂ R homogeneous ideal, define
Vp (I) := {p ∈ Proj(R) : p ⊃ I}.
The next lemma is the projective analogue of Proposition 15.2 with almost the same
proof.
Lemma 17.1. The following hold for any homogeneous ideals I, J, Is ⊂ Proj(R):
S T
• Vp ( s Is ) = s Vp (Is );
• Vp (I ∩ J) = Vp (I) ∪ Vp (J).
In particular, the sets Vp (I) form the closed sets on a topology of Proj(R), called
again the Zariski topology. Next we define a sheaf of functions on Proj(R), by adapting
the definition of OSpec(R) to the projective situation, in analogy to the definition of the
structure sheaves on projective spaces from the first semester. First, for the graded ring
R = ⊕Rd and p ∈ Proj(R), define
f
R(p) = { : f, g ∈ Rd some d, g ∈ / p} ⊂ Rp .
g
This is a subring of Rp . Next, let U ⊂ Proj(R) be an open set. Define OX (U ) by
Y
OX (U ) = {s = (sp )p∈U ∈ R(p) :
p∈U
f
s. t. s is locally given by , f, g ∈ Rd }
g
In other words, for any p ∈ U there exists V ⊂ U open and f, g ∈ Rd such that
sq = fg ∈ R(q) for all q ∈ V .
Proposition 17.2. (a) OProj(R) is a sheaf.
(b) For any p ∈ Proj(R), the stalk OProj(R),p = R(p) .
(c) Let f ∈ R+ = ⊕d>0 Rd be a homogeneous element and consider the open sets
Df+ = Proj(R) \ Vp (f ) = {p ∈ Proj(R) : f ∈
/ p}.
Then the following hold:
• The sets Df+ cover Proj(R);
• Each Df+ is an affine scheme. More precisely, Df+ ≃ SpecR(f ) , where
a
R(f ) = { n : a is homogeneous and deg a = n deg f } ⊂ Rp .
f
92 LEONARDO CONSTANTIN MIHALCEA

In particular, Proj(R) is a prescheme.

Proof. Homework! □
For any graded ring R, we denote by PnR = Proj(R[x0 , . . . , xn ]).
17.2. Projective subschemes. From now on we will take R = k[x0 , . . . , xn ], for k alge-
braically closed. As in the case of projective algebraic varieties (from the first semester),
we would like to have a correspondence between the closed subschemes of Pn and ideals
of R.
Definition 17.2. Let I be a homogeneous ideal in k[x0 , . . . , xn ]. The saturation of I
is
I = {P ∈ k[x0 , . . . , xn ] : ∃m s.t. xm
i P ∈ I ∀0 ≤ i ≤ n}.
We say that I is saturated if I = I.
Example 17.3. ⟨f x0 , f x1 , . . . , f xn ⟩ = ⟨f ⟩. In other words, the saturation removes the
part coming from the ‘irrelevant ideal’ ⟨x0 , . . . , xn ⟩.
Proposition 17.4. Let I, J ⊂ k[x0 , . . . , xn ] be homogeneous ideals. Then the following
hold:
(a) The saturation I is a homogeneous ideal;
(b) Proj(k[x0 , . . . , xn ]/I) = Proj(k[x0 , . . . , xn ]/I);
(c) Proj(k[x0 , . . . , xn ]/I) = Proj(k[x0 , . . . , xn ]/J) if and only if I = J.
d
(d) I d = I for d ≫ 0, i.e., an ideal and its saturation are the same for degree d large
enough.
Proof. To prove (a), let P ∈ I. Then xm i P ∈ I for some m and each 0 ≤ i ≤ n. Then
the graded pieces
(xm i P)
(d)
= xmi P
(d−m)
∈ I,
(d−m)
therefore by definition P ∈ I.
For (b), the inclusion Proj(k[x0 , . . . , xn ]/I) ⊃ Proj(k[x0 , . . . , xn ]/I) is obvious from
definition. Let p ∈ Proj(k[x0 , . . . , xn ]/I). Its preimage is a prime ideal q ⊂ k[x0 , . . . , xn ]
which contains I. If q does not contain I then there exists P ∈ I \ q. By definition,
there exists a positive integer m such that xm i P ∈ I ⊂ q. Since P ∈
/ q it follows that
⟨x0 , . . . , xn ⟩ ⊂ q, which is a contradiction.
To prove (c), we need to recover the ideal I from Proj(k[x0 , . . . , xn ]/I). For this,
consider the cover by the open affine sets Dx+j . By Proposition 17.2,
x0 x1 xn
Dx+j ∩ Proj(k[x0 , . . . , xn ]/I) = Spec(k[ , , . . . , ]/Ij ),
xj xj xj
where Ii is an ideal. One may recover Ij as the kernel of the morphism k[ xx0j , xx1j , . . . , xxnj ] →
k[ xx0j , xx1j , . . . , xxnj ]/Ij . Then (homework!),
P
I = ⟨{P ∈ k[x0 , . . . , xn ] homogeneous : ∈ Ij , 0 ≤ j ≤ n}⟩.
xdeg
j
P
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 93

d
For (d), it suffices to show that I ⊂ I d for d ≫ 0. Since I is finitely generated, take
P1 , . . . , Ps be a set of homogeneous generators. Take D1 large enough such that xdi Pj ∈ I
for d ≥ D1 and all 0 ≤ i ≤ n, 1 ≤ j ≤ s. Take D2 the maximum degrees of Pj . Now let
d
d ≥ D2 + (n + 1)D1 and let P be any element in I , and write
X
P = aj P j
for some homogeneous aj . Then deg aj ≥ d − D2 = (n + 1)D1 . This shows that each
term of aj contains a factor which is a multiple of xD
i
1
for some i. Then aj Pj ∈ I and
we are done. □
Obviously Proj(k[x0 , . . . , xn ]/I) = Vp (I) is closed subset of Pnk ; this is called a pro-
jective subscheme. (We will see in the next section the precise meaning of the word
‘subscheme’; roughly, this means that we have a locally ringed space with an induced
topology, and surjections at the level of stalks, induced from stalks of the structure sheaf
of the ambient space.)
Corollary 17.5. There is a one-to-one correspondence between projective subschemes
of Pn and saturated ideals in k[x0 , . . . , xn ]. The correspondence associates the saturated
homogeneous ideal I to Proj(k[x, . . . , xn ]/I).
94 LEONARDO CONSTANTIN MIHALCEA

18. More properties of schemes

In this section we gather several results we’ll need later. Most of these are not proved
in these notes; a good reference is [Har77, Ch. 3 , Ch. 4].
18.1. Integral schemes; open and closed embeddings. We start with two defini-
tions:
Definition 18.1. Let X be a prescheme.
• X is connected if X is connected as a topological space.
• X is integral if for any open set U ⊂ X, OX (U ) is an integral domain.
Example 18.1. Let X = Spec(R) be an affine scheme. Then (homework!):
• X is reduced iff N il(R) = 0;
• X is irreducible iff N il(R) is prime.
• X is integral iff R is an integral domain.
More generally, one may prove the following:
Proposition 18.2. A prescheme is integral if and only if it is irreducible and reduced.
Definition 18.2. A prescheme X is locally Noetherian if it can be covered by affine
open sets Spec(Ri ) where Ri is a Noetherian ring; it is Noetherian if it is locally
Noetherian and quasicompact. (Equivalently, X is Noetherian if it may be covered by
finitely many open affines Spec(Ri ) with Ri Noetherian.
Remark 18.1. One can show that if Spec(R) is a Noetherian prescheme then Spec(R)
is Noetherian as a topological space (i.e., every descending chain of closed subsets stabi-
lizes). However, the converse is not true. For example, take
R = k[x1 , x2 , . . .]/⟨x21 , x22 . . . , ⟩.
The only prime ideal is the (maximal) ideal m = ⟨x1 , x2 , . . . , ⟩. Therefore Spec(R) is a
Noetherian topological space. But R is not Noetherian, since m is not finitely generated.
The following shows that the property of being (locally) Noetherian is independent of
the choice of the cover.
Proposition 18.3. A prescheme X is locally Noetherian if and only if for every open
affine Spec(Ri ) ⊂ X, the ring Ri is Noetherian. In particular Spec(R) is Noetherian if
and only if R is Noetherian.
Proof. See [Har77, Prop. II.3.2]. □
Definition 18.3. An open sub(pre)scheme of a prescheme X is an open prescheme
U ⊂ X equipped with the restriction structure sheaf (OX )|U . An open embedding if a
morphism i : Y → X such that i(Y ) ⊂ X is open and the restriction Y → i(Y ) is an
isomorphism.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 95

Definition 18.4. A closed embedding is a morphism of preschemes f : Y → X such

that f (Y ) is a closed subspace of X, f induces a homeomorphism Y → f (Y ), and the
induced map f ∗ : OX → f∗ (OY ) of sheaves is surjective. (Equivalently, if f (p) = q, then
the induced map of stalks fp∗ : OX,q → OY,p is surjective.)
A closed subprescheme of a prescheme X is an equivalence class of closed embed-
dings, where f : Y → X and f ′ : Y ′ → X are equivalent if there is a diagram
Y′
i / Y

f′ ~ f
X
′
where i : Y → Y is an isomorphism.
Example 18.4. (Closed subschemes of affine schemes.) Let X = Spec(R) and I ⊂ R be
an ideal. The projection i∗ : R → R/I induces an injective morphism i : Spec(R/I) →
Spec(R). This is a closed sub(pre)scheme, because the map i∗ induces surjective maps
after localization.
In particular, observe that for any closed subset V (J) ⊂ Spec(R) there are many
closed subscheme structures corresponding to ideals I such that V (I) = V (J). √ The
smallest such √ subscheme corresponds to the largest such ideal, which is necessarily J.
This gives V ( J) the reduced (induced) scheme structure.
Definition 18.5 (Dimension). If X is a prescheme, dim X is the dimension when re-
garded as a (locally) ringed space. In other words, if X is irreducible, then dim X = n
if there is a maximal chain of irreducible closed sets
∅ ⊊ X0 ⊊ X1 ⊊ . . . ⊊ Xn = X.
If X is not irreducible, then dim X is the supremum of dimensions of the irreducible
components of X.
Example 18.5. If Spec(R) is irreducible, then dim Spec(R) = dim R (the Krull dimen-
sion of R).
18.2. Separated morphisms; schemes.
Definition 18.6. Let f : X → Y be a morphism of preschemes. The diagonal mor-
phism is the unique morphism ∆f : X → X ×Y X induced by universality property of
the fibre product X ×Y X; see Eq. (16.1).
The morphism f is separated if ∆f is a closed embedding. In this case X is said to
be separated over Y . A prescheme is called a scheme if it is separated over Spec(Z).
Remark 18.2. Most often we will work with schemes separated over a field k. It follows
from Theorem 18.2 below that if X is separated, then X is separated over k.
Example 18.6. A prescheme which is not separated is the affine line with two origins;
see §9.3 above.
96 LEONARDO CONSTANTIN MIHALCEA

Proposition 18.7. Any morphism of affine schemes f : Spec(S) → Spec(R) is sepa-

rated. In particular, any affine scheme is separated.
Proof. By Proposition 15.16, the diagonal ∆f is induced by the ring homomorphism
∆∗f : S ⊗R S → S obtained by multiplication: a ⊗ a′ 7→ aa′ . This is obviously surjective,
therefore the induced map ∆f is a closed embedding. □
Corollary 18.8. A morphism f : X → Y is separated if and only if the image of the
diagonal map ∆f : X → X ×Y X is closed.
Proof. If f is separated the the image of ∆f is closed by definition. To prove the
converse, we need to show that ∆f is a closed embedding, i.e. that ∆f : X → ∆f (X) is
a homeomorphism and that OX×Y X → ∆∗ (OX ) is surjective. Note that by definition
pr1 ◦ ∆f = pr2 ◦ ∆f = idX ,
therefore the restrictions of the projections pr1 , pr2 to ∆f (X) agree. We denote by
pr : ∆f (X) → X the common restriction. To show that ∆f is a homeomorphism, we
show that pr : ∆f (X) → X is a continuous inverse. Consider the diagram
∆f (X)
pr
∆f ◦pr
(
∆f (X) ⊂ X ×Y X /* X
pr pr
pr f
%  f 
X / Y
Since pr ◦ ∆f = idX , it follows that all diagrams commute. Then by the universality
property of the fibre product it follows that ∆f ◦ pr = id∆f (X) , thus ∆f is a homeomor-
phism.
The surjectivity of sheaves is a local question. We need to show that there is a
surjective map of sheaves OX×Y X → (∆f )∗ (OX ). This amounts to showing surjectivity
at the level of stalks. Further, observe that the stalk (∆f )∗ (OX )∆f (x) = OX,x and that
the induced map is the same as (∆f )∗x : OX×Y X,∆f (x) → OX,x . Take x ∈ X and let U be
an open affine neighborhood. We may take U to be small enough so that f (U ) ⊂ V is
included in some open affine subscheme. Then U ×V U is an open affine neighborhood
of ∆f (x), and by Proposition 18.7 ∆f : U → U ×V U is a closed embedding. Then the
pull-back map at the level of sheaves is surjective. □
Here is a list of properties which guarantee separatedness; cf. [Har77, Cor. 4.6].
Theorem 18.1. Assume all preschemes are Noetherian. Then the following hold:
(a) Open and closed embeddings are separated.
(b) Composition of separated morphisms is separated.
(c) Separated morphisms are stable under base extension.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 97

(d) A morphism f : X → Y is separated if and only if Y may be covered by open

subsets Vi such that f −1 (Vi ) → Vi is separated for each i.
18.3. Proper morphisms.
Definition 18.7. A morphism f : X → Y of preschemes is locally of finite type if
there exists a covering of Y by open affines Vi = Spec(Si ) such that for each i, f −1 (Vi )
may be covered by open affines Uij = Spec(Rij ) where each Rij is a finitely generated
Si -algebra.
The morphism f is of finite type if the covering of each f −1 (Vi ) may be taken to be
finite.
The morphism f is finite if in addition each f −1 (Vi ) = Spec(Ri ) is affine and Ri is
a finitely generated Si -module.
Definition 18.8. A morphism f : X → Y is proper if it is separated, of finite type,
and universally closed.
Here a morphism is closed if for any Z ⊂ X closed, f (Z) ⊂ Y is again closed; it is
universally closed if it is closed and for any morphism Y ′ → Y , the base extension
f ′ : X ×Y Y ′ → Y is a closed morphism.
Example 18.9. The line A1k over a field k is separated, but not proper. Indeed, consider
the base extension π : A1k ×Spec(k) A1k → A1k . It is induced by the k-algebra homomorphism
k[x] ,→ k[x, y], and at the level of closed points, this is simply the first projection. This
is not a closed map: for instance the image of the hyperbola xy − 1 = 0 is given by x ̸= 0,
which is not closed.
Theorem 18.2. Assume all preschemes are Noetherian. Then the following hold:
(a) Closed embeddings are proper.
(b) Composition of proper morphisms is proper.
(c) Proper morphisms are stable under base extension.
(d) If f : X → Y and g : Y → Z are two morphisms such that g ◦ f is proper and g
separated, then f is proper.
Proof. See [Har77, Cor.4.8]. □
Theorem 18.3. Let f : X → Y be a morphism of Noetherian schemes. Then the
following are equivalent:
(a) f is finite;
(b) f is proper with finite fibres.
Proof. See Lemma 30.21.1 in The Stacks Project. □
Definition 18.9. If Y is a scheme, define the projective n-space over Y as PnY :=
Pn ×SpecZ Y . A morphism of schemes f : X → Y is projective if it factors as
X ,→ PnY → Y
into a closed embedding followed by the projection.
98 LEONARDO CONSTANTIN MIHALCEA

The morphism f is quasi-projective if it factors into an open embedding X → X ′

followed by a projective morphism g : X ′ → Y .
Example 18.10. Let A be a ring, R be a graded ring with R0 = A and which as an A-
algebra is finitely generated by R1 . Then Proj(R) → Spec(A) is a projective morphism.
Indeed, since R is a finitely generated over R1 , it means there is a surjective A-algebra
homomorphism A[x0 , . . . , xn ] → R. This induces a closed embedding Proj(R) ,→ PnA and
the morphism Proj(R) → Spec(A) factors through PnA .
Example 18.11. Let i : X ⊂ Pnk be a projective subscheme of some projective spaces
and f : X → Y any morphism. Then any morphism f : X → Y is projective. Indeed,
the morphism f factors as:
f
X /9 Y

pr2
(i,f ) &
Pnk ×Spec(Z) Y
Definition 18.10. A scheme X over a field k is called complete if the morphism
X → Spec(k) is proper. The scheme is called projective if X → Spec(k) is a projective
morphism; equivalently, there is a closed embedding X ,→ Pnk .
A key property of projective morphisms is that they are proper:
Theorem 18.4. A projective morphism if Noetherian schemes is proper. A quasi-
projective morphism of Noetherian schemes is of finite type and separated.
Remark 18.3. A consequence of the theorem is that any projective morphism is closed.
This is an instance of the elimination theory: assume we have a system of equations



 f1 (x0 , . . . , xn ; y1 , . . . , yp ) = 0
f2 (x0 , . . . , xn ; y1 , . . . , yp ) = 0

..


 .

f (x , . . . , x ; y , . . . , y ) = 0
s 0 n 1 p

where fi ’s are polynomials in x’s and y ′ over some field k, and such that they are ho-
mogeneous in variables x. Then there exist polynomials g1 (y1 , . . . , yp ), . . . gℓ (y1 , . . . , yp )
such that
g1 (y1 , . . . , yp ) = . . . = gℓ (y1 , . . . , yp ) = 0.
In other words, one may ‘eliminate’ the variables x from the system above. This follows
from the fact that pr2 : Pn ×Spec(k) Ap → Ap is projective, thus it is proper, and thus the
image of any closed subset under pr2 is closed.
If k is algebraically closed, it turns out that (quasi-)projective algebraic varieties from
the first semester correspond to (quasi-)projective schemes over k, in the sense above.
This is proved in [Har77, Prop. II.4.10]. From now on, a variety will mean a
separated, integral scheme of finite type over an algebraically closed field k.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 99

19. Relative differentials

19.1. (Quasi-)coherent sheaves. Let X = Spec(R) be an affine scheme, and let M
be an R-module. Then one may associate a sheaf of OX -modules M f in a similar way to
Q OX : if U ⊂ X is open then M (U ) is the module over OX (U ) ngiven
the definition of by
f
tuples (mp ) ∈ p∈ U Mp which are ‘locally given’ by elements of the form mi /fi i ∈ Mfi .
The proof of the following lemma follows closely the proof for properties of OX :
Lemma 19.1. M f is a sheaf of OX -modules with the following properties:
(a) The sections Mf(Xf ) = Mf ;
(b) The stalk Mfp = Mp .
in particular, one may recover M from M f by taking global sections: M = Mf(X).

Furthermore, the assignment M → M

f gives an equivalence of categories:

Proposition 19.2. The assignment M → M f gives an equivalence between the category

of R-modules and the category of OX -modules of the form M f.
This equivalence preserves exactness, in the sense that the sequence of R-modules
0 → M → N → P → 0 is exact if and only if the corresponding sequence of OX -modules
0→M f→N e → Pe → 0 is exact.
Furthermore, the equivalence is compatible with taking direct sums, tensor products,
duals.
Proof. Homework! □
Definition 19.1. Let X be a (pre)scheme and F a sheaf of OX -modules. The sheaf F
is called quasi-coherent if X may be covered by open affine schemes Ui = Spec(Ri )
such that F|Ui ≃ M fi for some Ri -module Mi .
If in addition each Mi can be taken to be a finitely generated Ri -module then F is
called coherent.
As usual, the check of (quasi-)coherence may be done for one/any affine cover of a
(pre)scheme X. More precisely, the following holds:
Theorem 19.1. Let X be a scheme and F an OX -module. The following are equivalent:
(1) For any U ⊂ X open, affine, F|U ≃ M f for some OX (U )-module M .
(2) There exists an open affine cover Ui of X such that F|Ui ≃ M
fi for OX (Ui )-modules
Mi .
(3) For all x ∈ X, there exists a neighborhood U of x and an exact sequence of
(OX )|U -modules
(OX )⊕I |U → (OX )⊕J |U → F|U → 0.
(4) For all V ⊂ U affine, the canonical map
F(U ) ⊗OX (U ) OX (V ) → F(V )
is an isomorphism.
100 LEONARDO CONSTANTIN MIHALCEA

Proof. See [Mum99, §III.1]. □

All this admits a ‘projective’ version, generalizing in the obvious way the content from
§12.3 above.
Definition 19.2. A sheaf F is called locally free of rank r if there is an open cover
{Ui } of X such that F|Ui ≃ (OX )⊕r |Ui . A locally free sheaf of rank 1 is called invertible.
19.2. The module of relative differentials. In this section we define a quasi-coherent
sheaf to any morphism f : X → Y .
Fix φ : A → B be a homomorphism making B an A-algebra. Define the B-module
of relative differentials ΩB/A as the quotient of the free B-module with symbols
{dβ : β ∈ B} modulo the following relations:
(1) d(β1 + β2 ) = dβ1 + dβ2 ;
(2) (Leibniz rule) For α, β ∈ B, d(αβ) = αdβ + βdα;
(3) d(φ(a)) = 0 for any a ∈ A.
For any B-module M , an A-derivation is a B−module map D : B → M which satisfies
the 3 conditions above.
Lemma 19.3. There is a one-to-one correspondence HomB (ΩB/A , M ) to the module of
A-derivations D : B → M .
Proof. If τ ∈ HomB (ΩB/A , M ) then an A-derivation is defined by Dτ (β) = τ (dβ). Con-
versely, if D : B → M is an A-derivation, define τD by
X X
τD ( αi dβi ) = ai D(βi ).
One checks that the two operations are inverse to each other. □
Remark 19.1. An important situation is when A = M = k with k algebraically closed,
and B is a finitely generated k-algebra. Consider X = Spec(B) and a closed point
x = mx ∈ X with k = k(x) a structure of a B-module given via the natural map
B → B/mx ≃ k. Then
ΩB/k(x) ∗ = HomB (ΩB/k(x) , k)
is the same as the vector space of k-derivations D : B → k. The localization at x will
be the stalk of the tangent sheaf of X at x.
The module of relative differentials behaves well with respect to localization: if S ⊂ B
is a multiplicative system, then
S −1 ΩB/A = ΩS −1 B/A .
We give next two explicit realizations for ΩB/A , first in the ‘global’ case, then in the
‘local’ case. Consider the ring homomorphism
(19.1) δ : B ⊗A B → B; β1 ⊗ β2 7→ β1 β2 .
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 101

Define I := ker(δ); this isP a B ⊗A B-module. Then I/I 2 is a (B ⊗A B)/I-module, and

observe that for any κ := ai ⊗bj ∈ I, and any β ∈ B, the element (1⊗β)κ−(β ⊗1)κ ∈
I 2 . This gives I/I 2 a structure of a B-module by
β.κ = (1 ⊗ β).κ .
Proposition 19.4. I/I 2 is isomorphic to ΩB/A .

Proof. Define a module map Φ : ΩB/A → I/I 2 by sending dβ 7→ β ⊗ 1 − 1 ⊗ β. One

needs to check that this is well defined. Clearly condition (1) (additivity) holds, and
since 1 ⊗A a = a ⊗A 1 for any a ∈ A, condition (3) holds. For condition (2), we have
αβ ⊗ 1 − 1 ⊗ αβ = αβ ⊗ 1 − α ⊗ β + α ⊗ β − 1 ⊗ αβ
= α ⊗ 1(β ⊗ 1 − 1 ⊗ β) + 1 ⊗ β(α ⊗ 1 − 1 ⊗ α)
This shows that Φ(d(αβ)) = αΦ(dβ)+βΦ(dα). One can prove that Φ is an isomorphism
- see [Mum99, §III.1]. □
Example 19.5. Take A = k and B = k[x1 , . . . , xn ]. Then ΩB/A is a free B-module with
basis dx1 , . . . , dxn and
n
X ∂P
dP = dxi .
i=1
∂xi
More generally, consider B = k[x1 , . . . , xn ]/⟨f1 , . . . fs ⟩. Then ΩB/A is generated (as a
B-module by dx1 , . . . , dxn subject to relations
n
X ∂fi
dfj = dxi = 0.
i=1
∂x i

Theorem 19.2. Let K be a finitely generated field over some field k. Then dimK ΩK/k ≥
trdegk K and equality holds if and only if K is separably generated over k.
Theorem 19.3. Let (B, m) be a local ring which contains a field k isomorphic to the
residue field B/m. Consider the map
δ ′ : m/m2 → ΩB/k ⊗B k; m 7→ dm ⊗ 1.
Then δ ′ is a vector space isomorphism.
Let k be algebraically closed and perfect, i.e. either char(k) = 0 or if char(k) = p
then x 7→ xp is a field automorphism. A Noetherian local ring (B, m) is a regular local
ring if the minimum number of generators of m equals the Krull dimension of B.
Theorem 19.4. Let (B, m) be a local ring which contains a perfect field k isomorphic to
the residue field B/m. Assume further that B is the localization of a finitely generated
k-algebra. Then ΩB/k is a free B-module of rank dim B if and only if B is a regular local
ring.
102 LEONARDO CONSTANTIN MIHALCEA

19.3. The sheaf of relative differentials. Consider a separated morphism f : X → Y

and the associated closed embedding ∆ : X ,→ X ×Y X. Recall that if X = Spec(B)
and Y = Spec(A), then ∆ is determined by the morphism δ from Equation (19.1) above.
Let I be the ideal sheaf of ∆(X) inside X ×Y X.
Definition 19.3. ΩX/Y is the quasi-coherent OX -module obtained by carrying the OX×Y X /I-
module I/I 2 from X ×Y X to X via ∆.
If M is an R module and I ⊂ R is an ideal then M ⊗R R/I is an R/I-module. The
projection R → R/I induces the closed embedding Y := Spec(R/I) → X := Spec(R).
If F = Mf is the quasi-coherent OX -module determined by M , then the restriction of M
f
to Y is simply
(19.2) (M ^
f)|Y = M ⊗ R (R/I).

Recall that varieties are separated integral schemes of finite type over an integrally
closed field k.
Definition 19.4. A variety X is called non-singular at x if the local ring OX,x is a
regular local ring. It is non-singular if X is non-singular at every point.
One may show that being non-singular is an open condition, i.e. if X is non-singular
at x then there exists an open set U containing x such that U is non-singular.
The main theorem of this section is the following:
Theorem 19.5. Let X be a variety. Then X is non-singular if and only if the sheaf of
regular differential ΩX/k is a locally free sheaf of rank n = dim X.
Proof. See [Har77, Thm. II.8.15]. □
Definition 19.5. Let X be a non-singular variety of dimension n. The tangent sheaf
is defined as the dual TX := Ω∗X/k . The canonical sheaf ωX is defined as ωX := ∧n ΩX .
Each is a locally free sheaf, of rank n, respectively, rank 1.
The geometric genus of X is defined as pg := dimk ωX (X).
Key tools to understand the sheaves of differentials are the following two theorems.
(We prove the second one.)
Theorem 19.6. Let X be a non-singular variety, and let Y ⊂ X a closed subvariety
defined by the sheaf of ideals I. (I.e., Y → X is a closed embedding.) Then Y is
non-singular if and only if:
(a) ΩX/Y is locally free, and,
(b) There is a short exact sequence of OY -modules:
(19.3) 0 → I/I 2 → (ΩX/k )|Y → ΩY /k → 0.
Here the first map is induced by the natural map I/I 2 ≃ ΩX/Y → ΩX/k , and the second
is the restriction. (Both maps may be defined locally.)
If this is the case then I is locally generated by c := codim(Y, X) elements, and I/I 2
is locally free sheaf of rank c.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 103

The locally free sheaf I/I 2 is called the conormal sheaf of Y in X.

Remark 19.2. We will prove later on that there is a one-to-one correspondence between
locally free sheaves and vector bundles. Then the sequence above is the dual of the usual
sequence defining the normal bundle of a submanifold Z ⊂ M :
0 → TZ → (TM )|Z → NM Z → 0.
Corollary 19.6 (Adjunction formula). Let i : Y ⊂ X be a closed embedding of codimen-
sion c of non-singular varieties. Then there is an isomorphism of locally free sheaves of
rank 1 on Y :
(ωX )|Y ≃ ωY ⊗ ∧c (I/I 2 ).
Proof. If 0 → V1 → V2 → V3 → 0 is a short exact sequence of free R-modules of ranks
a, b, c respectively, then there is a canonical isomorphism
∧a V1 ⊗ ∧c V3 → ∧b V2 ; (e1 ∧ . . . ea ) ⊗ (f1 ∧ . . . ∧ fc ) 7→ e1 ∧ . . . ea ∧ f1′ ∧ . . . ∧ fc′ ,
where fi′ ∈ V2 are any elements mapping to fi . This induces a global map of locally free
OY -modules ωY ⊗ I/I 2 → ωX which is a local isomorphism, therefore it is also a global
isomorphism. □
The next result shows how the sheaf of differentials on Pn = Pnk is related to standard
sheaves on the projective space. In what follows we will use the notation from §12.3
about quasi-coherent sheaves on projective varieties associated to graded modules.
Proposition 19.7 (Euler sequence). There is an exact sequence of locally free sheaves
on Pn :
0 → ΩPn /k → OPn (−1)⊕(n+1) → OPn → 0.
Taking duals, one obtains the Euler sequence:
0 → OPn → OPn (1)⊕(n+1) → TPn → 0.
Proof. Let S := k[x0 , . . . , xn ] regarded as a graded module with deg xi = 1. Let E :=
S(−1)n+1 be the free S module of rank n + 1 with a basis e0 , . . . , en in degree 1. (I.e.,
this is simply the S-module S n+1 , but with basis shifted to degree 1.) Define a degree 0
module homomorphism
φ : E → S; ei 7→ xi (0 ≤ i ≤ n).
Let M := ker(φ); this is a graded module, and it fits into an exact sequence of graded
modules
0→M →E→S
Note that the last map is not surjective, as S0 = k is not in the image. However,
En → Sn is surjective for n ≥ 1. Therefore the induced map on sheaves is surjective,
giving a short exact sequence
f → OPn (−1)⊕(n+1) → OPn → 0.
0→M
104 LEONARDO CONSTANTIN MIHALCEA

We need to show that M f ≃ ΩPn /k . For that, we define local isomorphisms, and we need to
check that these glue to a global isomorphism. Cover Pn by the standard affine open sets
Ui = Spec[ xx0i , . . . , xxni ] ≃ An for 0 ≤ i ≤ n. Observe that the induced homomorphism of
free Sxi -modules Exi → Sxi is surjective, with kernel the free module of rank n generated
x
by ej − xji ei . Then the restriction of M f|U is a free module of rank n generated by sections
i

ej xj
− 2 ei
xi x i

(we need the extra factor of xi in the denominator to get elements of degree 0; see §12.3
above). Recall from Example 19.5 that ΩAn /k is locally free of rank n with basis given
by sections dxj for 1 ≤ j ≤ n. Then define

xj ej xj
ψi : ΩUi /k → M
f|U
i
d( )= − 2 ei .
xi xi xi

Since both are free modules of the same rank, this is an isomorphism. We need to prove
these isomorphisms glue over the intersections Ui ∩ Uj , i.e.,

xk xk
ψj (d( )) = ψi (d( )).
xi xi

To this aim, observe that in OPn (Ui ∩ Uj ),

xk xk xj
= · .
xi xj xi

Then
xk x k xj
ψj (d( )) = ψj (d( · ))
xi xj xi
xk xj xj x k
= ψj ( · d( ) + d( ))
xj xi xi xj
xk xj xj xk
= · ψj (d( )) + ψj (d( )).
xj xi xi xj

xi xj
Observe also that xj
· xi
= 1, therefore on Ui ∩ Uj ,

xi xj xj xi xj x2j xi
d( ) + d( ) = 0 =⇒ d( ) = − 2 d( ).
xj xi xi xj xi xi xj
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 105

Then
xk xk xj xj xk
ψj (d( )) = · ψj (d( )) + ψj (d( ))
xi xj xi xi xj
2
xj xk xi xj xk
= − 2 · ψj (d( )) + ψj (d( ))
xi xj xj xi xj
xj xk ei xi xj ek xk
= − 2 ( − 2 ej ) + ( − 2 ej )
xi xj xj xi xj xj
xk ek
= − 2 ei +
xi xi
xk
= ψi (d( )).
xi
One deduces that the morphisms ψi , ψj coincide on a set of module generators of ΩPn /k
over Ui ∩ Uj . This finishes the proof. □
Lemma 19.8. Let I ⊂ R be an ideal in a ring R. Then there is a canonical isomorphism
of R/I-modules
(19.4) I/I 2 → I ⊗R R/I; a + I 2 7→ a ⊗ 1.
Proof. Consider the short exact sequence of R-modules 0 → I → R → R/I → 0. By
[AM69, Ex. 2.2], tensoring with the R-module I preserves exactness, giving
0 → I 2 ≃ I ⊗R I → I ≃ I ⊗R R → I ⊗R R/I → 0.
The statement follows from the first isomorphism theorem applied to the second map.
□
This isomorphism is key to study the important special case when X = Pn and
Y = Vp (F ) ⊂ Pn is a hypersurface given by a homogeneous polynomial F of degree d.
Denote by i : Y ,→ Pn the inclusion. The ideal sheaf Ie = ⟨F
g⟩ induces the standard exact
sequence
(19.5) 0 → I → OPn → i∗ OY → 0.
On the other side, let S := k[x0 , . . . xn ] and consider the short exact sequence of graded
S-modules
0 / S(−d)
·F / S / S/⟨F ⟩ / 0
giving the sequence of OPn -modules
(19.6) 0 → OPn (−d) → OPn → i∗ OY → 0.
Proposition 19.9. (a) There is an isomorphism of locally free OPn -modules
I ≃ OPn (−d).
106 LEONARDO CONSTANTIN MIHALCEA

(b) Assume in addition that Y = Vp (F ) is non-singular in Pn . Then there is an

isomorphism of locally free OY -modules of rank 1,
I/I 2 ≃ (OPn (−d))|Y .
Proof. The first isomorphism follows from comparing the sequences (19.5) and (19.6).
The second part follows because the restriction corresponds to tensoring by ⊗OPn OY ,
then using the isomorphism (19.4). □
The following are the first explicit calculations of canonical sheaves.
Corollary 19.10. The following holds:
ωPn ≃ OPn (−n − 1).
More generally, if Y = Vp (F ) is a non-singular hypersurface in Pn of degree d, then
ωY = OY (d − n − 1) = (OPn (d − n − 1))|Y .
Proof. The first sequence follows by taking top exterior powers in the short exact se-
quence from Proposition 19.7. The second follows from the adjunction formula (Corol-
lary 19.6) combined with part (b) of Proposition 19.9. □
Recall the definition of the arithmetic genus pa from Definition 13.2. One may show
that if Y is non-singular, then pa (Y ) = pg (Y ) (the geometric genus).
Example 19.11. Let C ⊂ P2 be a non-singular plane curve given by a polynomial of
degree d. Then
ωC ≃ OC (d − 3).
We have the following special cases:
• If C is a line (d = 1) then ωC = OC (−2) (the restriction of OP2 (−2) to C).
• If C is a conic (d = 2) then ωC = OC (−1). Recall that C ≃ P1 , but the
isomorphism is given by the (restriction of ) Veronese embedding ν2 : P1 → P2 ,
of degree 2. By taking the corresponding map of graded modules, one may prove
that
ν2∗ (OC (−1)) = OP1 (−2),
in agreement to the line case.
• If C is an elliptic curve (d = 3), then ωC ≃ OC is trivial.
These calculations eventually lead to a classification of non-singular curves. For exam-
ple, the knowledge of ωC leads to knowing what is the topological Euler characteristic:
χtop (C) = 2 − 2g,
where g is the genus of C; see Definition 13.2. For example, if C is a non-singular curve
of degree d, one may use the Hilbert polynomial techniques to show that g = d−1


2
,
giving
χtop (C) = 2 − (d − 1)(d − 2).
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 107

Therefore a line and a conic have Euler characteristic 2, a cubic has Euler characteristic
0 etc. This is (of course!) consistent with the ‘real picture’ of these curves: complex lines
and conics are homeomorphic to a (real!) sphere S 2 ; an elliptic curve is homeomorphic
to a torus S 1 × S 1 etc.
108 LEONARDO CONSTANTIN MIHALCEA

20. Locally free sheaves and vector bundles

20.1. Definition and equivalence to locally free sheaves and vector bundles.
We follow the definition from [Mum99, III.2].

Definition 20.1. Let X be a scheme. A vector bundle π : E → X of rank r is a

scheme E together with the ‘projection’ morphism π which satisfy the following proper-
ties:
There exists an open covering (‘trivialization’) Ui and isomorphisms of schemes ψi−1 (Ui ) ≃
Ui × Ar over Ui :
ψi
π −1 (Ui ) / Ui × Ar

π pr1
# {
Ui
such that for all i, j, the restriction

ψi ◦ ψj−1 : (Ui ∩ Uj ) × Ar → (Ui ∩ Uj ) × Ar

of morphisms over Ui ∩ Uj

ψi ψj
(Ui ∩ Uj ) × Ar o π −1 (Ui ) ∩ π −1 (Uj ) / (Ui ∩ Uj ) × Ar
π
pr1 pr1
*  t
Ui ∩ Uj

is of the form

ψi ◦ ψj−1 = id × ψi,j ; ψi,j : Ar × (Ui ∩ Uj ) → Ar × (Ui ∩ Uj )

and ψi,j is determined by a linear isomorphism (with coefficients in OUi ∩Uj (Ui ∩ Uj )):
∗
ψi,j : OUi ∩Uj (Ui ∩ Uj ) ⊗k k[x1 , . . . , xn ] → OUi ∩Uj (Ui ∩ Uj ) ⊗k k[x1 , . . . , xn ];
Xr
∗ (i,j)
ψi,j (xk ) = ak,ℓ (u)xℓ .
k=1

∗
In other words, ψi,j is an element of GLr (OUi ∩Uj (Ui ∩ Uj )).

Proposition 20.1. There is a one-to-one correspondence between vector bundles of rank

r and locally free sheaves of rank r.

We explain this correspondence next.

 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 109

20.1.1. From vector bundles to locally free sheaves. Let π : E → X be a vector bundle
of rank r. We define a sheaf E on X by taking local sections of E: for any open set
U ⊂ X,
E(U ) = {s : U → E : π ◦ s = idU }
We need to make E into an OX -module. The sections over the trivializations E|Ui ≃
Ui × Ar are given by morphisms s : Ui → Ar , i.e. by k-algebra homomorphisms
s∗ : k[x1 , . . . , xr ] → OUi (Ui )); xk 7→ fk ∈ OUi (Ui ).
(See also Theorem 15.2.) The fibres Ex = π −1 (x) over points x ∈ X are k-vector
spaces of dimension r, and because of linearity of the transition maps, the vector space
structures arising from different trivializations are the same. Therefore the space of
sections over U has a OX (U )-algebra structure. This may be enlarged to an OX -module
structure by taking trivializations over Ui ’s (i.e., the trivializations ensure that sections
may be multiplied by regular elements in OX (Ui )). Furthermore, there are OX (Ui )-
module isomorphisms
E(Ui ) ≃ OX (Ui )⊕r ; s 7→ (s∗ (x1 ), . . . , s∗ (xr )),
which are again compatible under restriction to intersections Ui ∩ Uj . This makes E a
locally free sheaf of rank r.
20.1.2. From locally free sheaves to vector bundles. Conversely, if one is given a locally
free sheaf E, then it must have trivializations
E(Ui ) ≃ OX (Ui )⊕r ; f 7→ (f1 , . . . , fr ),
where fi ∈ OX (Ui ). These are maps of OX (Ui )-modules, and determine sections si :
Ui → Ar obtained from s∗i (xj ) 7→ fj . We obtain ‘trivial vector bundles’ Ui × Ar → Ui ,
and one needs to prove that these glue over the intersections Ui ∩ Uj . The key fact is
that over Ui ∩ Uj the two structures of OX -modules are given by OX (Ui ∩ Uj )-module
isomorphisms, i.e. precisely by matrices
(gi,j ) ∈ GLr (OX (Ui ∩ Uj )).
This is the same data which determines trivializations of vector bundles.
Finally, we note that if L is a locally free sheaf of rank 1, then the evaluation map
gives an isomorphism of locally free modules L∗ ⊗OX L ≃ OX . (At the level of transition
functions, (gi,j (L∗ ) = gi,j (L)−1 .) Because of this, line bundles are sometimes called
invertible sheaves.
Generally speaking, if φi : E(Ui ) → Ui × Ar and
gi,j = φj ◦ φ−1
i

then local sections fi ∈ O(Ui × Ar ) have to transform under the law

fj
= gi,j =⇒ fj = gi,j fi .
fi
110 LEONARDO CONSTANTIN MIHALCEA

In what follows we will need the following notion. Let X be a scheme and U =
Spec(R) ⊂ X open affine. Define K(U ) to be the localization RS , where S is the
multiplicative set of non-zero divisors. These form a presheaf of rings on X, and we
denote by KX its sheafification. It is called the sheaf of total quotient rings.

20.2. Example: Line bundles on Pn . As usual, let S = k[x0 , . . . , xn ] be the polyno-

mial ring, graded as usual, and let Ui be the standard affine cover. The graded modules
S(d) determine locally free sheaves of rank 1. Locally,
g i ) = { P : deg P = s + d homogeneous }.
S(d)(U
xsi
Let S(xi ) = k[ xx0i , . . . , xxni ] be the submodule of Sxi consisting of degree 0 elements. Then
g i ) is a free, rank 1, S(x ) -module with basis xd (regarded as a fraction).
S(d)(U i i
For example, if d < 0, then if
X
P = ap0 ,...,pn xp00 · . . . · xpi i · . . . · xpnn ∈ k[x0 , . . . , xn ]
is a homogeneous polynomial of degree s, then
P X xp00 1 xpnn
= a p ,...,p n p0 · . . . · · . . . · ,
xd+s
i
0
xi xdi xpnn
1
and the right hand side is an element in S .
xdi (xi )
To find transition maps, write isomor-
phisms
ϕ∗i ϕ∗j
(S(xi ) )(xj ) = S(xi xj ) o S(d)(xi xj ) / (S(xj ) )(xi ) = S(xi xj )

To describe the transition maps, we distinguish between the cases d ≥ 0 and d < 0:
• If d ≥ 0 then
xdi xdi
(ϕ∗j ◦ (ϕ∗i )−1 )(1) = ϕ∗j (xdi ) = ϕ∗i (xdj ) = .
xdj xdj
• If d < 0 then
−d
1 1 xj x−d
j
(ϕ∗j ◦ (ϕ∗i )−1 )(1) = ϕ∗j ( ) = ϕ∗
j ( ) = .
x−d
i x −d −d
j x i x −d
i

In other words, we proved that for any d the transition matrix is

xdi
(20.1) gi,j (OPn (d)) = ϕ∗j ◦ (ϕ∗i )−1 = .
xdj
In particular, since transition functions are inverses to each other,
OPn (d) ⊗OPn OPn (−d) ≃ OPn .
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 111

20.3. Example: Line bundles from hypersurfaces in Pn (Cartier divisors I).

Assume now that d > 0 and consider F ∈ Sd a homogeneous polynomial of degree d.
The multiplication by F gives an inclusion of sheaves
·F /
OPn (−d) OPn .
We analyze this inclusion when restricted to standard open sets. The sections
P
OPn (−d)|Ui = { s+d : deg P = s}
xi
The inclusion is given by
P FP
OPn (−d)|Ui ,→ OPn ; s+d
7→ s+d .
xi xi
Observe that the image is the ideal in S(xi ) generated by Fi := xFd , the dehomogenization
i
of F . Utilizing this, we can identify OPn (−d)|Ui with the ideal sheaf IF , the ideal sheaf
of Vp (F ). In other words, we proved
IVp (F ) ≃ OPn (−d).
The collection of dehomogenizations of F gives a collection D := {(Ui , Fi )} called a
Cartier divisor. Such a collection D determines an invertible sheaf LD together with
a global section s ∈ LD (X), as follows: let LD (Ui ) given by the
1
(20.2) LD (Ui ) = OPn ⊂ K(Ui ).
Fi
In other words, LD is the invertible subsheaf of the sheaf of total quotient rings given
by the isomorphism
ϕ−1 1
OPn (Ui )
i
/ LD (Ui ) , 1 7→ .
Fi
One calculates the transition maps by
F
xd
   
−1 1 Fj 1 Fj xdj
gi,j = ϕj ◦ ϕi (1) = ϕj = ϕj = = F = id .
Fi Fi Fj Fi xd
xj
i

Comparing to the transition functions from Equation (20.1) this shows that we have an
isomorphism of invertible sheaves
OPn (d) ≃ LD
As a ‘reality check’, note that the ‘local sections’ Fi ∈ OPn (Ui ) (corresponding to 1 ∈
LD (Ui )) satisfy
Fj = gi,j Fi
therefore they glue to a global section s = F .
112 LEONARDO CONSTANTIN MIHALCEA

21. Cartier divisors

Let X be a scheme. Recall that we defined the sheaf of total quotients KX by KX (U ) =
RP , where U = Spec(R) ⊂ X is open affine and P ⊂ R is the subset of non-zero divisors
in R.
Example 21.1. The following are homework exercises:
• If X is integral then KX is a constant sheaf: for any U ⊂ X open, KX (U ) = K,
where K is the fraction field of any open affine subset of X.
• KAn = k(x1 , . . . , xn );
P (x0 ,...,xn )
• KPn = { Q(x 0 ,...,xn )
} where P, Q are homogeneous polynomials and deg P = deg Q.
In what follows we will need the following notion. Let X be a scheme and U =
Spec(R) ⊂ X open affine. Define K(U ) to be the localization RS , where S is the
multiplicative set of non-zero divisors. These form a presheaf of rings on X, and we
denote by KX its sheafification. It is called the sheaf of total quotient rings.
∗ ∗
denotes the sheaf of total quotient rings. We denote by OX , respectively KX , the
sheaves of multiplicative groups given by the invertible elements in OX and KX .
∗ ∗
Definition 21.1. • A Cartier divisor on X is a global section of KX /OX . In
other words, a Cartier divisor is given by a collection (Ui , fi ) where Ui cover X,
fi ∈ K(Ui )∗ , and fi /fj ∈ OX (Ui ∩ Uj )∗ .
• We say that a Cartier divisor is principal if there exists a global section f ∈
KX (X)∗ such that each fi is the restriction of f . (Since the restrictions have to
agree in K∗ /O∗ , this means that f |Ui = cfi where c ∈ O∗ (Ui ).)
• Two Cartier divisors (Ui , fi ) and (Ui , gi ) are linearly equivalent if (Ui , fi /gi )
is principal.
By convention, we use additive notation for operations on Cartier divisors. For ex-
ample, if D1 , D2 are given by (Ui , fi ) and (Ui , gi ), then D1 + D2 is given by (Ui , fi · gi ),
while D1 − D2 is given by (Ui , fi /gi ).
Example 21.2. The following are homework exercises:
• Any Cartier divisor in An is principal.
• If F, G ∈ k[x0 , . . . , xn ] are homogeneous polynomials such that deg P = deg Q
then the dehomogenization of F/G over each of the standard affines induces a
principal divisor.
• If F ∈ k[x0 , . . . , xn ] is a homogeneous polynomial, then its dehomogenization over
each standard affine is a Cartier divisor, but it is not principal unless F ≡ c (a
constant).
Consider now a Cartier divisor D = (Ui , fi ). To it we associate an invertible subsheaf
LD ,→ KX by
1
LF (Ui ) = OX (Ui ) ⊂ KX (Ui ).
fi
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 113

(I.e., this is the rank 1 module over OX generated by 1/fi .) As before the transition
function is given by
   
∗ −1 1 fj 1 fj
gi,j = ϕj ◦ (ϕi ) (1) = ϕj = ϕj · = .
fi fi fj fi
We need the following lemma.
Lemma 21.3. Let R be a ring and T ⊂ R the subset of non-zero divisors. Let M be
a free rank 1 R-module generated by an element m0 . Assume that there is an injective
R-module homomorphism φ : M → RT and let φ(m0 ) = at . Then φ(m0 ) is invertible.
Proof. It suffices to show that a ∈ T . If not, take r ∈ R \ 0 such that ra = 0. Then
φ(rm0 ) = 0, contradicting injectivity. □
Proposition 21.4. Let X be a scheme.
(a) There is a one-to-one correspondence D ↔ LD between Cartier divisors and in-
vertible OX -modules which are subsheaves of KX .
(b) LD1 −D2 ≃ LD1 ⊗ L−1 D2 .
(c) D1 −D2 is a principal Cartier divisor if and only if LD1 ≃ LD2 as abstract invertible
sheaves, i.e., not as subsheaves of KX .
Proof. To prove (a), we need to show that we can recover the Cartier divisor D out of
the invertible sheaf LD together with its embedding in KX . Take fi to be the inverse
of a local generator of LD (Ui ) ⊂ KX (Ui ) as a OX (Ui )-module. Two such choices must
generate the same module, therefore they must differ by an element in OX (Ui )∗ . As such,
they give the same element of KX (Ui )∗ /OX (Ui )∗ , and thus the same Cartier divisor.
Part (b) follows from the formula for the transition functions. Finally, to prove (c),
we utilize part (b). Then it suffices to show that D is principal if and only if LD ≃ OX .
If D is principal, then it is given by a global section f ∈ KX (X)∗ . This global section
generates LD , in the sense that its restriction to each Ui generated LD (Ui ). Then we
obtain a global isomorphism OX → LD by sending 1 7→ f −1 . Conversely, assume that
we are given an isomorphism OX → LD and consider the image of 1 ∈ OX (X). This
must be a global section f in LD , Since LD is a subsheaf of KX , f must be an invertible
element in KX (X)∗ , and this element determines a principal divisor on X. □
There are many common situations when every invertible sheaf on a scheme X is a
subsheaf of KX . Some examples are:
• X is integral (see [Har77, Prop. II.6.15]);
• X is projective over a field.
In all these situations, the (additive) group of Cartier divisors on X modulo equivalence
is isomorphic to the (multiplicative) group of invertible sheaves. This is called the
Picard group of X, denoted by Pic(X).
One can prove that Pic(Pn ) = Z (generated by OPn (1)), and that Pic(An ) = 0.
114 LEONARDO CONSTANTIN MIHALCEA

Definition 21.2. Let D = (Ui , fi ) be a Cartier divisor. We say that D is effective if the
‘local equations’ are regular functions, i.e., fi ∈ OX (Ui ). For an effective Cartier divisor
D one may associate a codimension 1 subscheme Y ⊂ X given locally by the equations
fi = 0. Sometimes we denote by OX (D), respectively by OX (−D), the invertible sheaf
corresponding to an effective divisor D, respectively −D (the inverse of D).
If D is an effective divisor, then OX (D) has a canonical global section sD , which under
the isomorphism OX (D)|Ui = f1i OX (Ui ) is defined locally by the equations (sD )Ui = fi ∈
OX (Ui ). Indeed,
fj
fj = fi = gi,j fi
fi
showing that this satisfies the required glueing properties.
Example 21.5. If X = Pn and F is a homogeneous polynomial of degree d, it determines
a Cartier divisor D = (Ui , xFd ). The corresponding line bundle LD has a section sD = F ,
i
with vanishing locus Vp (F ) ⊂ Pn .
The following follows from definitions:
Corollary 21.6. Let D be an effective Cartier divisor on a scheme X which determines
the codimension 1 subscheme Y . Then OX (−D) ≃ IY .
Definition 21.3. Let D = (Ui , fi ) be a Cartier divisor on a scheme X. The support of
D, denoted by supp(D), consists of those points x ∈ X with the property that the image
of fi in the local ring (OX,x , mx ) is in mx .
One can show that the support of a closed subset of X. If U = X \ supp(D), then
OX (D) has a section sD ∈ OX (D)(U ) which does not vanish anywhere. This section
extends to a section of OX (D) which is ‘meromorphic’, with poles on the locus where D
is not effective.
Often one can write fi ∈ KX (Ui )∗ as fi = abii where ai , bi ∈ OX (Ui )∗ and such that
A = (Ui , ai ) and B = (Ui , bi ) are Cartier divisors. Then the Cartier divisor D = (Ui , fi )
may be written as a difference of two effective divisors: D = A − B. This will allow us
to define the first Chern class
c1 (LD ) = c1 (LA ) − c1 (LB ) = [V (sA )] − [V (sB )]
where sA , sB are the canonical sections.
Definition 21.4. Let D be a Cartier divisor on a scheme X. A meromorphic section
∗
of OX (D) is any global section of OX (D)⊗OX KX . In other words, a meromorphic section
is not required to have local equations as regular functions in OX (Ui ), but in a localization
of it.
From definition, it follows that sD = (fi ) is a meromorphic section of OX (D). The
divisor D is effective precisely when sD is a (‘holomorphic’) section of OX (D).
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 115

22. Rational equivalence and the Chow group

Recall that we use variety to mean a reduced and irreducible scheme.
22.1. Rational equivalence. Let X be a variety and V ⊂ X a codimension 1 subva-
riety. Consider A := OX,V , the localization at the generic point of V . There is a well
defined group homomorphism
ordV : KX (X)∗ → Z
called the order of vanishing. It satisfies ord(rs) = ord(r) + ord(s). If we write
r ∈ KX (X)∗ as r = ab with a, b ∈ A, then
a
ord = ord(a) − ord(b).
b
An important example is when X is smooth at V , i.e. if (OX,V , mV ) is the local ring,
then dim mV /m2V = 1. In this case OX,V is a discrete valuation ring (DVR) and mV
is principal. Then for any r ∈ KX (X)∗ , we can write (its image) as r = atm where
∗
a ∈ OX,V . In this case
ordV (r) = m.
More generally the order function may be defined as the length ℓA (A/⟨r⟩) of the A-
module A/⟨r⟩; see [Ful84, App. A].
Lemma 22.1. For a fixed r ∈ KX (X)∗ , there are only finitely many V ’s such that
ordV (r) ̸= 0.
Proof. See [Ful84, App. B]. □
Example 22.2. Assume X is a smooth curve. Then OX,V = OX,x are local rings at the
points in X, and the image of r in OX,x corresponds to the evaluation at x. For all but
finitely many points, this evaluation is non-zero.
Example 22.3. Let X = P2 and r = xx01 . Let Hi = Vp (xi ) ⊂ P2 be the standard
P (x0 ,x1 ,x2 )
hyperplanes. Then OX,Hi = { Q(x 0 ,x1 ,x2 )
: deg P = deg Q, Q ∈
/ ⟨xi ⟩} and the maximal
ideal is generated by xi . It follows that
ordH0 (r) = −1; ordH1 (r) = 1; ordH2 (r) = 0.
We now define the analogues of the fundamental classes from topology.
P
Definition 22.1. Let X be a scheme. A k-cycle is a finite formal combination ni [Vi ]
where Vi ⊂ X are irreducible subvarieties of dimension k. A k-cycle is effective if each
coefficient ni ≥ 0. P
We denote this by ni [Vi ] ≥ 0. The (additive) group of k-cycles is denoted by Zk (X).
Let r ∈ KX (X)∗ and W ⊂ X a k + 1 dimensional subvariety. There is an associated
cycle in Zk (X) defined by X
div(r) = ordV (r)[V ],
V ⊂W
116 LEONARDO CONSTANTIN MIHALCEA

where the sum is over subvarieties V of dimension k.

Example 22.4. An important example of this construction is when X is smooth in
codimension 1, i.e. every codimension 1 subscheme is locally given by a single equation.
(In technical terms, ‘every Weyl divisor is Cartier’.) Then every r ∈ KX (X)∗ determines
a Cartier divisor X
div(r) = ordY (r)[Y ] ∈ Pic(X),
Y ⊂X
where the sum is over codimension 1 subschemes Y . For a Cartier divisor D one defines:
H 0 (X, OX (D)) = {f ∈ KX (X)∗ : div(f ) + D ≥ 0},
i.e. the vector space of meromorphic sections with ‘poles of order at most the order of
D’. If D is an effective divisor, then H 0 (X, OX (D)) = OX (D)(X) (the global sections of
OX (D)). Studying this vector space is closely related to the Grothendieck-Riemann-Roch
theorem.
Definition 22.2. A k-cycle α is rationally equivalent to 0 if there exist finitely many
subvarieties Wi with dim Wi = k + 1 and elements ri ∈ KWi (Wi )∗ such that
X
α= [div(ri )].
The (k-th) Chow group is defined as
Ak (X) := Zk (X)/Bk (X)
where Bk (X) is the subgroup of cycles rationally equivalent to 0. The Chow group is
defined to be
A∗ (X) = ⊕k≥0 Ak (X).
Example 22.5. A scheme X and its induced reduced structure Xred have the same
subvarieties, therefore A∗ (X) = A∗ (Xred ).
S
Example 22.6. Assume X = Xi is a union of irreducible schemes. Then Zk (X) =
⊕Zk (Xi ) and Ak (X) = ⊕Ak (Xi ). (homework!)
Example 22.7. Assume dim X = n. Then An (X) = Zn (X) is the free abelian group
on the n-dimensional irreducible components of X.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 117

23. Proper push-forward and flat pull-back

23.1. Proper push-forward. Recall from Definition 18.8 the definition of a proper
morphism. Examples of proper morphisms are closed embeddings, morphisms of pro-
jective varieties, and base changes of proper morphisms.
If f : X → Y is a proper morphism of schemes, then for any (closed) subvariety
V ⊂ X, the image f (V ) ⊂ Y is closed. Since the restriction f : V → W := f (V ) is a
dominant map of irreducible schemes, it induces an injective map f ∗ : KW ,→ KV .
Lemma 23.1. The field extension KW ,→ KV is finite if and only if dim V = dim W .
If this is the case, then f : V → W is generically finite.
If, furthermore, k is algebraically closed, then the general fibre is a zero dimensional
scheme with deg(KV : KW ) reduced points.
Proof. Refer to [Ful84, App. B2]. □
Define (
deg(KV : KW ) if dim V = dim W ;
deg(V /W ) =
0 if dim W < dim V .
Define f∗ : Zk (X) → Zk (Y ) by
(23.1) f∗ [V ] = deg(V /f (V ))[f (V )],
extended by linearity. A fundamental result is that rational equivalence is preserved by
taking proper push-forward.
Theorem 23.1. Let f : X → Y be a proper morphism of schemes, and let α ∈ Ak (X)
such that α ≃ 0. Then f∗ (α) ≃ 0.
In particular, Equation (23.1) gives a well-defined push-forward (group homomor-
phism) f : Ak (X) → Ak (Y ).
A key particular case is when X is a complete scheme over a field k. Then by definition
the morphism p : X → Spec(k) is proper, and there is a degree map
Z
deg = : A0 (X) → Z; α 7→ p∗ (α).
X
P
If k is algebraically
R closed and α = nP [P ] ∈ A0 (X) then p∗ [P ] = [p(P )]; in general
p∗ [P ] = X [P ] = deg(K(P ) : k)[p(P )]. R
One may extend the degree morphism to X : A∗ (X) → Z by sending α 7→ 0 for
α ∈ Ai (X), i > 0. Observe also that if f : X → Y is a proper morphism and X, Y are
proper, then Z Z
α= f∗ (α).
X Y
(This is a special case of functoriality, and an analogue of the chain rule from calculus.)
118 LEONARDO CONSTANTIN MIHALCEA

Example 23.2. Let f : P1 → P1 define by [x0 : x1 ] 7→ [x20 : x21 ]. Then

f∗ [P1 ] = 2[P1 ]; f∗ [pt] = [pt].
Example 23.3 (Gauss-Bonnet; assumes knowledge of Chern classes). Let X be a com-
plex, projective manifold, with tangent bundle TX . Then
Z
c(TX ) = χ(X),
X
the Euler characteristic of X. Here c(TX ) is interpreted as a non-homogeneous class in
A∗ (X).
Example 23.4 (Enumerative invariants; assumes knowledge of the intersection prod-
uct). Let X be a complex projective manifold, and let Md (X) be the ‘moduli space of
stable maps’ which compactifies the (scheme!) of morphisms f : P1 → X of some fixed
class f∗ [P1 ] = d ∈ A1 (X). Let evi : Md (X) → X be the evaluation maps, obtained by
evaluating at 0, 1, ∞. For subvarieties Ω1 , Ω2 , Ω3 ⊂ X define
Z
⟨Ω1 , Ω2 , Ω3 ⟩d = [ev−1 −1 −1
1 Ω1 ] · [ev2 Ω2 ] · [ev3 Ω3 ].
X
These count the (virtual) number of curves f : P1 → X of degree d that pass through
subvarieties Ω1 , Ω2 , Ω3 ; Gromov-Witten invariants are examples.
23.2. Fundamental class of a subscheme. We defined the group Zk (X) of cycles
on a scheme X as formal combinations of classes of irreducible subvarities. However,
we we need to defined analogues of fundamental classes for possibly reducible, possibly
non-reduced schemes. The construction below provides this construction.
Let X be a pure dimensional scheme with irreducible components X1 , . . . , Xr . The
local rings OX,Xi have dimension 0. Define the geometric multiplicity of Xi is defined
by mi = ℓOX,Xi (OX,Xi ). If k is algebraically closed, then
mi = dimk OX,Xi .
The cycle associated to X, or the fundamental class of X is
X
[X] = mi [Xi ] ∈ Adim X (X).
More generally, if X ⊂ Y is closed in some larger scheme Y , then we regard [X] as a
class in A∗ (Y ), via the natural closed embedding.
23.3. Flat pull-back. It is known from commutative algebra (cf. e.g., [AM69, Prop.2.18])
that if M1 → M2 → M3 → 0 is a short exact sequence of R-modules, then for any R-
module N , M1 ⊗ N → M2 ⊗ N → M3 ⊗ N → 0 remains exact.
Definition 23.1. An R-module N is flat if for any short exact sequence 0 → M1 →
M2 → M3 → 0, the sequence
0 → M1 ⊗ N → M2 ⊗ N → M3 ⊗ N → 0
remains exact.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 119

Definition 23.2. Let f : X → Y be a morphism of pure-dimensional schemes. The

morphism f is flat if for any open affine sets U ⊂ X, V ⊂ Y such that f (U ) ⊂ V ,
OX (U ) is a flat OY (V )-module.
We say that f has relative dimension n if for any subvariety Z ⊂ Y , f −1 (Z) is
pure dimensional of dimension dim Z + n.
Theorem 23.2. Assume f : X → Y is flat and that Y is irreducible and X is pure
dimensional. Then f : X → Y has relative dimension n := dim X − dim Y , and all base
extensions X ×Y Y ′ → Y ′ are flat of relative dimension n.
Proof. See [Har77, III.9], especially Cor. 9.6. □
Example 23.5. Here are some examples of flat morphisms:
• An open embedding i : U ,→ X where U ⊂ X is open.
• The projection from a vector bundle π : E → X, or from the projectivization of
a vector bundle P(E) → X.
• The projection X × Y → X.
• Any dominant morphism X → C from an irreducible subvariety of dimension
dim X ≥ 1 to a non-singular curve.
Let f : X → Y be a flat morphism of relative dimension n. Define a pull back
f ∗ : Zk (Y ) → Zk+n (X), [W ] 7→ [f −1 (W )].
We extend this definition by linearity.
For any closed subscheme, let f −1 (Z) denote its inverse image, defined as the fibre
product
X ×Y Z /X

f
 
Z / Y
Theorem 23.3. Let Z ⊂ Y be any closed subscheme. Then there is a well defined
pull-back homomorphism f ∗ : Ak (Y ) → Ak+n (X) defined by:
f ∗ [Z] = [f −1 (Z)] ∈ A∗ (X).
Proof. See [Ful84, Theorem 1.7]. □
Proposition 23.6. Consider a fibre diagram
g′
X′ / X
f′ f
 g 
Y′ / Y
where g is flat and f is proper. Then g ′ is flat, f ′ is proper and for any α ∈ Z∗ (X),
f∗′ (g ′ )∗ (α) = g ∗ f∗ (α) ∈ Z∗ (Y ′ ).
120 LEONARDO CONSTANTIN MIHALCEA

23.4. An exact sequence.

Proposition 23.7. Let Y be a closed subscheme of a scheme X and let U := X \ Y .
Denote by i : Y → X and j : U → X the inclusions. Then we have a short exact
sequence
i∗ j∗
Ak (Y ) / Ak (X) / Ak (U ) / 0.
Proof. The analogous sequence of cycles is exact:
i∗ j∗
Zk (Y ) / Zk (X) / Zk (U ) / 0,
showing surjectivity. It is also clear from definitions that j ∗ i∗P
(β) = 0. It remains to show
∗ ∗
that any α ∈ ker(j ) is in the image of i∗ . Write j (α) = div(ri ) for some Wi ⊂ U
subvarieties. By taking closures of Wi , and observing that K(Wi ) = K(Wi ), we obtain
X
j ∗ (α − div(ri )) = 0 ∈ Zk (U ),
P
where ri are extensions of ri to Wi . Then α − div(ri ) = i∗ (β) for some β ∈ Zk (X),
and we are done. □
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 121

24. Chern classes of line bundles

In this section we will define the first Chern class of a line bundle L over a scheme
X, as an operator c1 (L) : Ak (X) → Ak−1 (X). The idea is the following. Take V ⊂ X
be a k-dimensional subvariety. The restriction of L to V determines a Cartier divisor
DV = {(Ui , fi )}, up to linear equivalence. Then
c1 (L) ∩ [V ] = [DV ].
Of course, one needs to prove that this construction is independent of choices of repre-
sentatives in a linear equivalence class of Cartier divisors.
One can actually give a stronger statement in the case L is given by a Cartier divisor
D on X. In this case, the class is supported on Ak−1 (V ∩ |D|) where |D| is the support
of D. A key technical point is that occasionally V ⊂ |D|. To deal with the situation
when L restricted to V one needs to

24.1. Cartier divisors, Weil divisors, pseudodivisors. Assume for now that X is a
variety of dimension n. (Later it will be a scheme.) In particular, there is a well defined
function field K(X). A Weil divisor is an element of Zn−1 (X).
If D = {(Ui .fi )} is a Cartier divisor, and V ⊂ X is a subvariety then one defines
ordV (D) := ordV (fi ).
This is well defined since fi /fj is invertible on the overlaps. Using this, the Cartier
divisor defines a Weyl divisor
X
[D] = ordV (D)[V ],
V

where the sum is over all codimension 1 subvarieties V ⊂ X. Observe that if two
Cartier divisors D ∼ D′ are linearly equivalent then D − D′ is a principal divisor, i.e.,
D′ − D = div(f ) for some f ∈ K(X). But then the associated Weyl divisors [D′ ]
and [D] are rationally equivalent (by definition). This implies that we have a group
homomorphism
CaCl(X) → An−1 (X); D 7→ [D].
Recall that if D = {(Ui , fi )} is a Cartier divisor, its support is defined to be
|D| = {x ∈ X : some fi is not a unit}
(Observe that fi is in general a rational function, i.e. an element of K(X)∗ . The invertible
∗
elements OX,x ⊂ K(X)∗ are a subset of this. Therefore it makes sense to ask whether fi
is in the smaller subset, although fi may not have a well defined image in OX,x .) The
support is a closed subset of X.
Also recall that D is effective if each fi ∈ OX (Ui ) and it is not a zero divisor.
Definition 24.1 (Pseudovisors). Let X be a scheme. A pseudodivisor on X is a triple
(L, Z, s) where L is a line bundle, Z ⊂ X is closed and s is a section of L over X \ Z
122 LEONARDO CONSTANTIN MIHALCEA

which does not vanish. Equivalently, s provides an isomorphism (i.e., a trivialization)

L|X\Z ≃ (OX )|X\Z .
The data (L′ , Z ′ , s′ ) defines the same pseudodivisor if Z = Z ′ and there is an isomor-
phism σ : L ≃ L′ such that the restriction σ|X\Z sends s to s′ .
We call L, Z, s the line bundle, support, respectively the section associated to (L, Z, s).
According to this definition, a pseudodivisor (L, X, s) with support X is simply an
equivalence class of line bundles.
The pseudodivisors form an abelian group with the following operations:
(L, Z, s) + (L′ , Z ′ , s′ ) = (L ⊗ L′ , Z ∪ Z ′ , s ⊗ s′ ); −(L, Z, s) = (L−1 , Z, 1/s).
If f : X ′ → X is a morphism, then one can pull-back a pseudodivisor by
f ∗ (L, Z, s) = (f ∗ (L), f −1 (Z), f ∗ (s)).
These operations agree with those involving Cartier divisors, whenever objects are de-
fined.
An important source of pseudodivisors comes from Cartier divisors. Recall (cf. Propo-
sition 21.4 above) that a Cartier divisor D determines a line bundle LD . Then D de-
termines a pseudodivisor (LD , |D|, sD ) where |D| is the support of D, and sD = (fi ) is
the canonical meromorphic section of D; cf. Definition 21.4 and after. The section sD is
nowhere vanishing outside the support |D|.
Definition 24.2. A Cartier divisor D represents a pseudodivisor (L, Z, s) if |D| ⊂ Z
and there is an isomorphism L ≃ LD such that, outside Z, sends s 7→ sD .
For example, a Cartier divisor D represents any pseudodivisor (L, X, ∅) where L ≃ LD .
The main result about pseudodivisors is the following:
Theorem 24.1. Let X be a variety. Then any pseudodivisor (L, Z, s) is represented by
a Cartier divisor. Furthermore,
• If Z ̸= X, then D is uniquely determined.
• If Z = X, then D is determined only up to linear equivalence.
Proof. Later, if there is time; see §2.2, [Ful84]. □
The following is the main definition underlying the construction of Chern classes via
intersection products.
Definition 24.3. Let D̃ = (L, |D|, s) be a pseudodivisor on an n-dimensional variety
X. The associated Weyl divisor class
[D] ∈ An−1 (X)
is defined as follows. Take any Cartier divisor D which represents D̃.
• If |D| =
̸ X then by Theorem 24.1 D is uniquely defined, and [D] is the associated
Weyl divisor, regarded in An−1 (|D|).
• If |D| = X then D is only defined up to linear equivalence, but the associated
Weyl divisor [D] ∈ An−1 (X) is well defined.
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 123

24.2. Intersections by divisors and the first Chern class. We are now ready to
define intersections by divisors and the first Chern class of a line bundle. Assume X is
a scheme, D̃ = (L, |D|, s) a pseudivisor on X, and V ⊂ X a k-dimensional subvariety
with inclusion j :,→ X. Then define an intersection product
D̃.[V ] := [j ∗ D̃] ∈ Ak−1 (V )
In other words:
• If V ̸⊂ |D| then D restrict to a unique Cartier divisor DV on V , and D.[V ] =
[DV ];
• If V ⊂ |D|, then restrict LD to V , and take any Cartier divisor DV corresponding
to this restriction. Then D.[V ] = [DV ].
One extends this product by linearity to define an intersection product
D · α ∈ Ak−1 (X)
for any α ∈ Ak (X).
Definition 24.4 (First Chern class). Let L → X be a line bundle over a scheme X.
Define the group homomorphism c1 (L) ∩ : Ak (X) → Ak−1 (X) by
c1 (L) ∩ [V ] = [C],
where C is the Weil divisor associated to the line bundle L|C .
If L = LD is the line bundle associated to a pseudo-divisor D, then c( L)∩[V ] = D·[V ].
We will list formal properties of the Chern classes later ADDREF, once we define Chern
classes of any order; those would be essentially the same as those defined in topology.
Example 24.1. A homogeneous polynomial F ∈ k[x0 , . . . , xn ] of degree d determines a
Cartier divisor with corresponding line bundle OPn (d). Since the Cartier divisor associ-
ated to F is linearly equivalent to that for any xdi , it follows that
c1 (OPn (d)) = dc1 (OPn (1)).
n
Furthermore, if V ⊂ P is a subvariety such that F |V ̸≡ 0,
(24.1) c1 (OPn (d)) ∩ [V ] = [V ∩ Vp (F )].
Example 24.2 (Chow group of Pn ). One may use Proposition 23.7 to show that Ak (Pn )
is generated by [Lk ], the fundamental class of a k-dimensional plane. We prove that
[Lk ] ̸= 0 in Ak (Pn ). From (24.1) it follows that
c1 (OPn (1))k ∩ [Lk ] = 1.[pt] ∈ A0 (Pn ).
One can use again the sequence from Proposition 23.7 to get that [pt] ̸= 0. (Remove
hyperplanes to reduce to A0 (pt) = Z.) This implies that [Lk ] ̸= 0.
Example 24.3 (Degrees). Let α ∈ Ak (Pn ) ≃ Z. The degree of α is
Z
deg(α) = c1 (OPn (1))k ∩ α.
Pn
124 LEONARDO CONSTANTIN MIHALCEA

Let D be an effective Cartier divisor. We abuse notation and denote by D ⊂ X

the associated codimension 1 subscheme (given by the local equations of the Cartier
divisor). Let i : D ,→ X be the (closed) embedding.
Definition 24.5 (Gysin maps for divisors). Define
i∗ : Ak (X) → Ak−1 (D); [V ] 7→ D · α;
(intersection product with the divisor D).
Note that i∗ is arrow reversing, although i is not flat. These types of (arrow reversing)
morphisms in a ‘homology theory’ are called Gysin morphisms in topology.
We list some useful properties of this pull-back, including a version of the self-
intersection formula.
Proposition 24.4. The morphism i∗ satisfies the following properties:
(a) For any α ∈ Ak (X),
i∗ i∗ (α) = c1 (LD ) ∩ α.
(b) (The self-intersection formula for divisors) If α ∈ Ak (D), then
i∗ i∗ (α) = c1 (i∗ LD ) ∩ α.
(Note that, at least if D, X are smooth, then i∗ LD is the normal bundle of D in X; cf.
Proposition 19.9.)
(c) If X is pure-dimensional, then i∗ [X] = [D].
(d) For any line bundle L → X,
i∗ (c1 (L) ∩ α) = c1 (i∗ L) ∩ i∗ (α).
Example 24.5 (Bèzout). Let Z1 = Vp (F1 ), Z2 = Vp (F2 ) ⊂ P2 be plane curves given
by homogeneous polynomials of degree d1 respectively d2 . Both Z1 and Z2 are Cartier
divisors. We can utilize the intersection with divisors product to calculate [Z1 ] · [Z2 ].
Since [Z1 ] = d1 [H] and [Z2 ] = d2 [H] (H any hyperplane), it follows that
Z
deg([Z1 ] · [Z2 ]) = d1 d2 [H]2 = d1 d2 .
P2
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 125

25. Chern classes of vector bundles

We are now ready to define higher Chern classes. Let E → X be a vector bundle
of rank e + 1 over a scheme X. To it, we associate the projective bundle (or the
projectivization) P(E) → X. This is defined as follows. If E ≃ X → Ae+1 is trivial,
then P(E) = X × Pe . In general, one glues these local schemes.
The projectivization is endowed with a projection π : P(E) → X, with the property
that π −1 (X) = P(Ex ), where Ex is the fibre of E → X over x. Of course, π is flat (this
is a local property, so it suffices to check it locally). It is also proper, since Pn is proper,
and locally it is a base change from Pn → pt.
Recall that over Pn there is a tautological sequence of vector bundles:
0 → OPn (−1) → OP⊕n+1
n → Q → 0.
The fibre of OPn (−1) over a line ℓ ⊂ k n+1 is simply the line ℓ. By pulling back, then
gluing, this gives a tautological sequence of vector bundles on P(E):
0 → OE (−1) → π ∗ E → Q → 0.
(We abuse notation and denote the quotient bundle again by Q.) The fibre of OE (−1)
over a point [ℓx ] ∈ P(E) corresponding to a line ℓx ⊂ Ex is ℓx .
Definition 25.1 (Segre classes of vector bundles). Let i ≥ 0. The i-th Segre class is
the group homomorphism si (E) ∩ : Ak (X) → Ak−i (X) defined by
si (E) ∩ α = π∗ (c1 (OE (1))e+i ∩ π ∗ (α)).
Consider the formal power series
X
st (E) = si (E)ti .
i≥0
i
P
The Chern polynomial ct (E) = i≥0 ci (E)t is the (formal) inverse of this power
series:
ct (E) = st (E)−1 .
Explicitly,
c0 (E) = 1; c1 (E) = −s1 (E); c2 (E) = s1 (E)2 − s2 (E), . . .
Here are the formal properties of Chern classes (cf. Thm. 3.2 in [Fulton, IT] ADDREF.)
Theorem 25.1. Let E → X be a vector bundle of rank e over a scheme X. The Chern
classes satisfy the following properties.
(1) (Vanishing) ci (E) = 0 for i > e;
(2) (Commutativity) ci (E) ∩ (cj (E) ∩ α) = cj (E) ∩ (ci (E) ∩ α);
(3) (Projection formula) For any proper morphism f : Y → X,
f∗ (f ∗ (ci (E)) ∩ α) = ci (E) ∩ f∗ (α).
126 LEONARDO CONSTANTIN MIHALCEA

(4) (Pull-back) For any flat morphism f : Y → X,

ci (f ∗ (E) ∩ f ∗ (α) = f ∗ (ci (E) ∩ α).
(5) (Whitney formula) Consider a short exact sequence 0 → F → E → G → 0 be a
short exact sequence of vector bundles on X. Then
ct (E) = ct (F )ct (G).
(6) (Normalization) If E is a line bundle such that E ≃ OX (D) for a Cartier divisor
D on X, then
c1 (E) ∩ [X] = [D].
More generally, if E has sections and s ∈ H 0 (X; E) is a general section with
zero locus Z(s) ⊂ X, then
(25.1) ce (E) ∩ [X] = [Z(s)].
The class
c(E) = 1 + c1 (E) + . . . + ce (E)
is called the total Chern class of E. It can also be regarded as the Chern polynomial
ct (E) evaluated at t = 1.
25.1. Chern roots and the splitting principle. One can actually show that the
properties above determine the Chern classes. The idea is as follows. The splitting
principle states that if E → X is a vector bundle, then there exists a flat map f : X ′ → X
such that:
• The pull back f ∗ : H ∗ (X) → H ∗ (X ′ ) is injective;
• The pull back bundle f ∗ E has a filtration
0 ⊂ E1 ⊂ E2 ⊂ . . . ⊂ Ee = f ∗ E
such that Ei → X ′ is a vector bundle on X ′ of rank i.
By the Whitney formula, it follows that
c(Ee ) = c(E1 ) · c(E2 /E1 ) · . . . · c(Ee /Ee−1 ).
Since each quotient Ei /Ei−1 is a line bundle, the normalization formula (applied only
for line bundles) calculates each class c(Ei /Ei−1 ). Therefore the linear operator c(f ∗ E)
is calculated. Now by injectivity the class c(E) ∩ α may be identified to c(f ∗ E) ∩ f ∗ (α).
(A subtlety: in the topological category one needs to ensure that f ∗ (α) is well defined.
For instance this happens if the Poincaré dual of α exists, e.g. if X is non-singular.
This is not needed if one works in the algebraic category, and uses Chow groups; in this
case flatness ensures that pull-backs exist.) To construct f : X ′ → X with the required
properties, one may take X ′ = Fl(E). This is the variety parametrizing complete flags
of vector bundles E1 ⊂ E2 ⊂ . . . ⊂ Ee = E, equipped with its natural projection to
X. For instance, if X = pt, then E is just a vector space, and this is the complete flag
variety Fl(E). For details, we refer to [Ful84, §3.2].
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 127

The splitting principle allows one to formally define the Chern roots of a vector bundle.
If rank(E) = e, these are formal indeterminates x1 , . . . , xe such that
X
c(E) = (1 + x1 )(1 + x2 ) · . . . · (1 + xe ) = ei (x1 , . . . , xe ).
i≥0

Then the Chern class ci (E) = ei (x1 , . . . , xe ). The idea is that xi ’s are only formal, but
their symmetrizations are actual classes. By the splitting principle, one may actually
identify xi = c1 (Ei /Ei−1 ).
The Chern roots are very useful tools relating Chern classes of vector bundles. I will
list several properties below - all follow from the judicious use of the splitting principle.
Lemma 25.1. In all statements x1 , . . . , xe are the Chern roots of E. Then the following
hold:
• (Dual bundles) The Chern roots of the dual bundle E ∨ are −x1 , . . . , −xe . In
particular,
ci (E ∨ ) = (−1)i ci (E).
• (Tensor products) Let F → X be a vector bundle with Chern roots y1 , . . . , yf .
Then the Chern roots of E ⊗ F are xi + yj , for 1 ≤ i ≤ e, 1 ≤ j ≤ f .
• (Symmetric powers) The Chern roots of Symp E are xi1 + xi2 + . . . + xip where
1 ≤ i1 ≤ i2 ≤ . . . ≤ ip ≤ e.
• (Exterior powers) The Chern roots of p E are xi1 + xi2 + . . . + xip where 1 ≤
V
i1 < i2 < . . . < ip ≤ e.
We illustrate this by some examples.
Example 25.2 (The Chern classes of the tangent bundle of Pn ). We use the Euler
sequence
0 / OPn / OPn (1)⊕(n+1) / TPn / 0.
to calculate the Chern classes of the tangent bundle on Pn . Let H = c1 (OPn (1). By
the normalization property this is the class of a hyperplane in Pn . From the Whitney
formula we obtain that the total Chern class
c(O)c(TPn ) = c(OPn (1)⊕(n+1) = c(OPn (1))n+1 = (1 + H)n+1 .
Since c(O) = 1, it follows that
n  
n+1
X n+1
c(TPn ) = (1 + H) = Hk.
k=0
k

For instance, cn (TPn ) = (n + 1)[pt], reflecting the fact that the topological Euler charac-
teristic of Pn is n + 1. (This is an instance of the Gauss-Bonnet theorem.)
128 LEONARDO CONSTANTIN MIHALCEA

Example 25.3 (The degree of Gr(2, 4).). Consider the Grassmannian Gr(2, 4) parametriz-
ing linear subspaces of dimension 2 in C4 . We have seen in a previous problem (Fall
final exam!) that this is a quadric hypersurface in P5 . More precisely, if
 
a1 b1 c1 d1
A=
a2 b2 c2 d2
and if we denote by pi,j the determinant of the 2 × 2 minor in columns i, j, then
p12 p34 − p13 p24 + p14 p23 = 0.
Let D := {V ∈ Gr(2, 4) : dim V ∩ F2 ≥ 1} where F2 = ⟨e1 , e2 ⟩. You also proved that this
variety is given by the equation p34 = 0 (even scheme theoretically). This says that D is
in fact the intersection of the hyperplane divisor p34 = 0 in P5 by the quadric Gr(2, 4).
Then
Z Z
4
deg(Gr(2, 4) = c1 (OP5 (1)) ∩ [Gr(2, 4)] = c1 (OP5 (1))4 ∩ 2c1 (OP5 (1)) = 2.
P5 P5

This has the following enumerative interpretation. The product c1 (OP5 (1))4 represents
the intersection [D]4 . Since D is the variety of projective lines in P3 meeting the line
P(F2 ). To calculate [D]4 , we are allowed to take divisors which are linearly equivalent
to D, then intersect. In particular, we may move the ‘reference line’ P(F2 ) and define
4 linear equivalent Cartier divisors Di by taking 4 lines ‘sufficiently general’. Then we
are saying that there are 2 lines in P3 meeting meeting 4 given lines in general position.
Example 25.4 (Lines on cubic surfaces). Let Σ ⊂ P3 be a general surface, i.e. Σ = Z(F )
where F is a general homogeneous polynomial of degree 3 in variables x0 , x1 , x2 , x3 . We
are interested in how many lines are included in P3 . The space of lines in P3 is the same
as the Grassmannian Gr(2, 4). Let S be the tautological subbundle. The polynomial F
is a section of Sym3 (C4 )∗ , and the condition that a line P(V ) ∈ Gr(2, 4) is included in
Z(F ) means that F |V ≡ 0. In other words, F gives a (general) section
s ∈ H 0 (Gr(2, 4), Sym3 (S ∗ )), V 7→ F |V .
We are interested in the zero locus of this section. By the normalization property (25.1)
[Σ] = [Z(s)] = cr (Sym3 (S ∗ )) ∩ [Gr(2, 4)],
where r = rank(Sym3 S ∗ ) = 4. Let x1 , x2 be the Chern roots of S ∗ . Then by Lemma
Lemma 25.1 the Chern roots of Sym3 S ∗ are 3x1 , 2x1 + x2 , x1 + 2x2 , 3x2 . It follows that
c4 (Sym3 S ∗ ) = 9x1 x2 (x1 + 2x2 )(2x1 + x2 )
= 9x1 x2 (2x21 + 5x1 x2 + 2x22 )
= 9x1 x2 (2(x1 + x2 )2 + x1 x2 )
= 18x1 x2 (x1 + x2 )2 + 9(x1 x2 )2
= 18c2 (S ∗ )c1 (S ∗ )2 + 9c2 (S ∗ )2 .
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 129

Using the intersection theory of Gr(2, 4), one may prove that
c2 (S ∗ )c1 (S ∗ )2 ∩ [Gr(2, 4)] = c2 (S ∗ )2 ∩ [Gr(2, 4)] = [pt].
Therefore
c4 (Sym3 S ∗ ) ∩ [Gr(2, 4)] = 18[pt] + 9[pt] = 27[pt].
The enumerative geometry interpretation of this is that a general cubic surface will con-
tain 27 lines.
25.2. Some answers to enumerative geometry questions. You might enjoy think-
ing about these numbers:
• 2: number of lines passing through 4 given lines in P3 ;
• 92: number of conics in P3 meeting 8 lines;
• 1: non-singular conics tangent to 5 given lines;
• 3264: number of conics tangent to 5 given conics;
• 4407296: number of conics tagent to 8 general quadrics surfaces;
• 2875: number of lines on a quintic threefold in P4 ;
• 5819539783680: number of twisted cubics tangent to 12 quadrics in P3 . (This
was found by H. Schubert in 1870’s.)
THAT’S ALL FOLKS!
130 LEONARDO CONSTANTIN MIHALCEA

26. Appendix: Results from commutative algebra

Theorem 26.1 (Hilbert basis theorem). Let k be a field. Then any ideal I in the poly-
nomial ring k[x1 , . . . , xn ] is finitely generated, i.e. there exist polynomials P1 , . . . , Ps ∈
k[x1 , . . . , xn ] such that
I = ⟨P1 , . . . , Ps ⟩.
In particular, from any subset S ⊂ I such that ⟨S⟩ = I, one may extract a finite
subset S ′ ⊂ S such that ⟨S ′ ⟩ = I.
Proof. See e.g. [CLO15, §2.5]. □
Theorem 26.2 (Noether normalization lemma). Let R be an integral domain, finitely
generated over a field k. Assume that the transcendence degree of R over k equals to n.
Then there exists elements y1 , . . . , yn ∈ R such that:
• y1 , . . . , yn are algebraically independent over k;
• R is integrally dependent on the subring k[y1 , . . . , yn ].
Proof. See a proof in [Mum99, §.1] (attributed to Nagata). □
Lemma 26.1 (A consequence of Cohen-Seidenberg ‘Going Up’). Let R be a field and
S ⊂ R a subring such that R is integrally dependent on S. Then S is a field.
Proof. Let 0 ̸= a ∈ S. Then 1/a is a root of a monic polynomial with coefficients in S,
meaning that
(1/a)n + cn−1 (1/a)n−1 + . . . + c0 = 0
in R, with ci ∈ S. After multiplying by an−1 we obtain that
1
= −cn−1 − cn−2 a − . . . − c0 an−1 ∈ S.
a
□
26.1. Rings and modules of fractions; localization. Let R be a commutative ring
with 1, and M an R-module. A set S ⊂ R is called —bf multiplicative if 1 ∈ S and for
any a, b ∈ S, the product ab ∈ S.
There are two important examples of multiplicative sets.
Example 26.2. Let f ∈ R. Take S = {f n : n ∈ Z≥0 }.
Example 26.3. Let p ⊂ R be a prime ideal. Then S := R \ p is multiplicative.
For a multiplicative set S ⊂ R and an R-module M one may define an equivalence
relation on M × S by
(m1 , s1 ) ≃ (m2 , s2 ) ⇐⇒ ∃s ∈ S s(s1 m2 − s2 m1 ) = 0.
m
The equivalence class of (m, s) is denoted by .
Denote by
s
m
S −1 M := { : m ∈ M, s ∈ S}.
s
 LECTURE NOTES FOR ALGEBRAIC GEOMETRY 131

Then S −1 R is a ring and S −1 M is a S −1 R-module. There is a natural ring homomor-

phism
r
R → S −1 R; r 7→ .
1
This may not be injective. (E.g., if f is nilpotent, then Rf = 0.)
There is a similar module homomorphism M → S −1 M .
Lemma 26.4. Let R be a ring and 0 ̸= f ∈ R. If R has no nilpotents then Rf has no
nilpotents.
r
Proof. Let fa
∈ Rf such that
r
n rn
= = 0.
fa f an
Then f p rn = 0 in R for some p, implying that (f p r)n = 0. Since R has no nilpotents,
this implies that f p r = 0, i.e. fra = 0 in Rf . This proves the claim. □

26.2. Primary and irreducible ideals. The main reference for this section is [AM69,
Ch.4 and Ch. 7].
Definition 26.1. Let R be an arbitrary ring and I ̸= R an ideal. The ideal I is primary
if
xy ∈ I =⇒ x ∈ I or y n ∈ I.
The ideal I is called irreducible if
I = I1 ∩ I2 =⇒ I = I1 or I = I2 .
√
Proposition 26.5. Let I be a primary ideal. Then the radical I is prime and it is the
smallest prime containing I.
Proof. See [AM69, Prop. 4.1]. □
Proposition 26.6. Let R be a Noetherian ring.
(a) Any ideal I is a finite intersection of irreducible ideals.
(b) Any irreducible ideal is primary.
Proof. See Lemmas 7.11 and 7.12 in [AM69]. □

26.3. Topology. Let X be a topological space. A basis for the topology of X is a

collection U of open sets such that every open set in X is a union of basis elements in
U.

26.4. Limits, stalks, and localization. TODO.

132 LEONARDO CONSTANTIN MIHALCEA

References
[AM69] M. F. Atiyah and I. G. Macdonald. Introduction to commutative algebra. Addison-Wesley
Publishing Co., Reading, Mass.-London-Don Mills, Ont., 1969.
[CLO15] David A. Cox, John Little, and Donal O’Shea. Ideals, varieties, and algorithms. Undergradu-
ate Texts in Mathematics. Springer, Cham, fourth edition, 2015. An introduction to compu-
tational algebraic geometry and commutative algebra.
[Ful84] William Fulton. Intersection theory. Springer-Verlag, Berlin, 1984.
[Har77] Robin Hartshorne. Algebraic geometry. Graduate Texts in Mathematics, No. 52. Springer-
Verlag, New York-Heidelberg, 1977.
[Mum99] David Mumford. The red book of varieties and schemes, volume 1358 of Lecture Notes in
Mathematics. Springer-Verlag, Berlin, expanded edition, 1999. Includes the Michigan lectures
(1974) on curves and their Jacobians, With contributions by Enrico Arbarello.