INTRODUCTION TO ALGEBRAIC GEOMETRY

Note del corso di

Istituzioni di Geometria Superiore 2
A.A. 2005/06

Emilia Mezzetti

Dipartimento di Matematica e Informatica

Università degli Studi di Trieste
Via Valerio 12/1
34127 Trieste, ITALY
e-mail: [email protected]

Introduction

Algebraic Geometry is the field of mathematics which studies the sets of solu-
tions of systems of algebraic equations, i.e. of equations given by polynomials. The
origins of Algebraic Geometry go back to the Ancient Babilonians and Greeks and,
since them, this fascinating topic has attracted mathematicians of all times and
countries. During the 19th and the beginning of last century, important progresses
were made, mainly by the so called Italian School of Algebraic Geometry. Then,
starting from 1950, the subject was completely refounded, taking into account
the advent of Modern Algebra. This work was initiated by the Russian-Italian-
American mathematician Oscar Zariski, and pushed on mainly by the ingenious
French mathematician Alexander Grothendieck. Recently, important results and
answers to classical problems have been given. Unfortunately, the techniques de-
veloped have become more and more intricated, so that the approach to this topic
is not easy.

These notes are basically a revision of those originally written for a course
of algebraic geometry of the Diploma Course in Mathematics at ICTP. They are
devoted to the students of the Laurea Specialistica (Magistrale) in Matematica
of the University of Trieste. They are not exhaustive of all topics treated in the
lectures. For more reference, I warmly recommend the books of Šafarevič [7],
Hartshorne [3] and Harris [2].

An asterisk * near an exercise denotes that it is quoted in the text.

2 Mezzetti

1. Affine and projective space.

Let K be a field. By definition, the affine space of dimension n over K is simply

the set K n : on it, the additive group of K n acts naturally by translation. The
affine space will be denoted AnK or simply An . So the points of AnK are n−tuples
(a1 , . . . , an ), where ai ∈ K for i = 1, . . . , n.
The natural action of K n on AnK , is the map t defined by

t : K n × AnK −→ AnK

((x1 , . . . , xn ), (a1 , . . . , an )) −→ (x1 + a1 , . . . , xn + an ).

Note that: t(0, P ) = P, where 0 is the zero vector of K n and P ∈ AnK , and
t(w, t(v, P )) = t(v + w, P ), for v, w ∈ K n and P ∈ AnK .
The action of a vector v on a point P is “ by translation”. The point t(v, P )
will be denoted P + v. The action t is faithful and transitive: this means that,
for any choice of P, Q ∈ AnK , there exists one and only one v ∈ V such that
Q = t(v, P ): for this vector, the notation Q − P will be sometimes used.
Let Q ∈ AnK be a point, and W ⊂ K n be a vector subspace. We define the
affine subspace of AnK passing through Q with orienting space W (or of direction
W ) as follows:
S = {P ∈ AnK | P = Q + w, w ∈ W }.

S can be seen as “ W translated in Q”. Note that affine subspaces of AnK do not
necessarily pass through the origin. Two affine subspaces of An with a common
orienting space are called parallel. If dim W = m, we also define dim S = m. The
subspaces of dimension 1 are called lines, those of dimension 2 planes, those of
dimension n − 1 (or of codimension 1) hyperplanes.
The points of an affine subspace of An can be characterized as solutions of a
system of equations. These are of two types:
a) Parametric equations of a subspace.
Let S be the subspace passing through Q(y1 , . . . , yn ) with orienting space W , and
let w1 , . . . , ws be a basis of W , with wi = (wi1 , . . . , win ). Then P (x1 , . . . , xn ) ∈ S
if and only if there exist t1 , . . . , ts ∈ K such that

(x1 , . . . , xn ) = (y1 , . . . , yn ) + t1 w1 + . . . + ts ws ,

or equivalently
x = y1 + t1 w11 + . . . + ts ws1

 1

x2 = y2 + t1 w12 + . . . + ts ws2
...

...
As (t1 , . . . , ts ) varies in K s we get in this way all points of S.
 Introduction to algebraic geometry 3

For example, if S is the line through Q of direction W = hwi, with w =

(b1 , . . . , bn), then
x = y1 + tb1

 1

x2 = y2 + tb2
...

xn = yn + tbn
are parametric equations of S.
b) Cartesian equations of a subspace.
Let s = dim W , W ⊂ K n . Then W is the set of vectors whose coordinates are solu-
tions of a homogeneous linear system of rank n − s in n indeterminates z1 , . . . , zn :
(a z + ... + a z = 0
11 1 1n n

...
an−s,1 z1 + . . . + an−s,n zn = 0.

Hence P (x1 , . . . , xn ) belongs to S if and only if P = Q + w, where w is a solution

of the previous system, i.e. if and only if the following equations are satisfied:

 a11 (x1 − y1 ) + . . . + a1n (xn − yn ) = 0
...
an−s,1 (x1 − y1 ) + . . . + an−s,n (xn − yn ) = 0


i.e. (x1 , . . . , xn ) is a solution of the system:


 a11 x1 + . . . + a1n xn + b1 = 0
... =0
an−s,1 x1 + . . . + an−s,n xn + bn


where we have put bi = −(ai1 y1 + . . . + ain yn ), for i = 1, . . . , n − s. For example a

hyperplane is represented by a unique linear equation of the form:

a1 x1 + . . . + an xn + b = 0.

Let V be a K−vector space, of dimension n + 1. Let V ∗ = V \ {0} be the

subset of non–zero vectors. The following relation in V ∗ is an equivalence relation
(relation of proportionality):
v ∼ v 0 if and only if ∃λ 6= 0, λ ∈ K such that v 0 = λv.
∗
The quotient set V∼ is called the projective space associated to V and denoted
P(V ). The points of P(V ) are the lines of V (through the origin) without the origin.
In particular, P(K n+1 ) is denoted PnK (or simply Pn ) and called the numerical
projective n-space. By definition, the dimension of P(V ) is equal to dim V − 1.
There is a canonical surjection p : V ∗ → P(V ) which takes a vector v to its
equivalence class [v]. If (x0 , . . . , xn ) ∈ (K n+1 )∗ , then the corresponding point of
4 Mezzetti

Pn is denoted [x0 , . . . , xn ]. So [x0 , . . . , xn ]=[x00 , . . . , x0n ] if and only if ∃λ ∈ K ∗ such

that x00 = λx0 , . . . , x0n = λxn .
If a basis e0 , . . . , en of V is fixed, then a system of homogeneous coordinates
is introduced in V , in the following way: if v = x0 e0 + . . . + xn en , then x0 , . . . , xn
are called homogeneous coordinates of the corresponding point P =[v]= p(v) in
P(V ). We also write P [x0 , . . . , xn ]. Note that homogeneous coordinates of a point
P are not uniquely determined by P , but are defined only up to multiplication
by a non–zero constant. If dim V = n + 1, a system of homogeneous coordinates
allows to define a bijection
P(V ) −→ Pn
P = [v] −→ [x0 , . . . , xn ]
where v = x0 e0 + . . . + xn en .
The points E0 [1, 0, . . . , 0], . . . , En [0, 0, . . . , 1] are called the fundamental points
and U [1, . . . , 1] the unit point for the given system of coordinates.
A projective (or linear) subspace of P(V ) is a subset of the form P(W ), where
W ⊂ V is a subspace.
Assume that dim W = s + 1 and that W is represented by a linear homoge-
neous system (a x + ... + a x = 0
10 0 1n n

(∗) . . .
an−s,0 x0 + . . . + an−s,n xn = 0.
Note that a (n + 1)-tuple (x̄0 , . . . , x̄n ) is a solution of the system if and only
if (λx̄0 , . . . , λx̄n ) is, with λ 6= 0. So these solutions can also be interpreted as
representing the points of P(W ) and the equations (*) as a system of Cartesian
equations of P(W ). To write down parametric equations of P(W ) it is enough to
fix a basis of W , formed by vectors w0 , . . . , ws . Then a general point of P(W ) is
parametrically represented by [λ0 w0 + . . . + λs ws ], as λ0 , . . . , λs vary in Ps .
If W, U are vector subspaces of V , the following Grassmann relation holds:

dim U + dim W = dim(U ∩ W ) + dim(U + W ).

From this relation, observing that P(U ∩ W ) = P(U ) ∩ P(W ), we get in P(V ):

dim P(U ) + dim P(W ) = dim(P(U ) ∩ P(W )) + dim P(U + W ).

Note that P(U + W ) is the minimal linear subspace of P(V ) containing both P(U )
and P(W ): it is denoted P(U ) + P(W ).

1.1. Example. Let V = K 3 , P(V ) = P2 , U, W ⊂ K 3 subspaces of dimension

2. Then P(U ), P(V ) are lines in the projective plane. There are two cases:
(i) U = W = U + W = U ∩ W ;
(ii) U 6= W , dim U ∩ W = 1, U + W = K 3 .
 Introduction to algebraic geometry 5

In case (i) the two lines in P3 coincide; in case (ii) P(U ) ∩ P(W ) = P(U ∩ W ) = [v],
if v 6= 0 is a vector generating U ∩ W. Observe that never P(U ) ∩ P(W ) = ∅.
Let T ⊂ P(V ) be a non–empty set. The linear span hT i of T is the intersection
of the projective subspaces of P(V ) containing T , i.e. the minimum subspace
containing T . For example, if T = {P1 , . . . , Pt }, a finite set, then hP1 , . . . , Pt i =
P(W ), where W is the vector subspace of V generated by vectors v1 , . . . , vt such
that P1 = [v1 ], . . . , Pt = [vt ]. So dimhP1 , . . . , Pt i ≤ t − 1 and equality holds if and
only if v1 , . . . , vt are linearly independent; in this case, also the points P1 , . . . , Pt
are called linearly independent. In particular, for t = 2, two points are linearly
independent if they generate a line, for t = 3, three points are linearly independent
if they generate a plane, etc. It is clear that, if P1 , . . . , Pt are linearly independent,
then t ≤ n + 1, and any subset of {P1 , . . . , Pt } is formed by linearly independent
points.
P1 , . . . , Pt are said to be in general position if either t ≤ n + 1 and they are
linearly independent or t > n + 1 and any n + 1 points among them are linearly
independent.

1.2. Proposition. The fundamental points E0 , . . . , En and the unit point

U of a system of homogeneous coordinates on Pn are n + 2 points in general
position. Conversely, if P0 , . . . , Pn , Pn+1 are n + 2 points in general position,
then there exists a system of homogeneous coordinates in which P0 , . . . , Pn are the
fundamental points and Pn+1 is the unit point.
Proof. If e0 , . . . , en is a basis, then clearly the n + 1 vectors e0 , . . . , eˆi , . . . , en , e0 +
. . . + en are linearly independent: this proves the first claim. To prove the second
claim, we fix vectors v0 , . . . , vn+1 such that Pi = [vi ] for all i. So v0 , . . . , vn is a
basis and there exist λ0 , . . . , λn in K such that vn+1 = λ0 v0 + . . . + λn vn . The
assumption of general position easily implies that λ0 , . . . , λn are all different from 0,
hence λ0 v0 , . . . , λn vn is a new basis such [λi vi ] = Pi and Pn+1 is the corresponding
unit point. 

Let H0 = hE1 , . . . , En i, H1 = hE0 , E2 , . . . , En i, . . . , Hn = hE0 , . . . , En−1 i be

n+1 hyperplanes in Pn . Note that the equation of Hi is simply xi = 0. These hyper-
planes are called the fundamental hyperplanes. Let Ui = Pn \Hi = {P [x0 , . . . , xn ] |
xi 6= 0}. Note that Pn = U0 ∪ U1 ∪ . . . ∪ Un , because no point in Pn has all
coordinates equal to zero. There is a map φ0 : U0 −→ An (= K n ) defined by
φ0 ([x0 , . . . , xn ]) = ( xx01 , . . . , xxn0 ). φ0 is bijective and the inverse map is j0 : An −→ U0
such that j0 (y1 , . . . , yn ) = [1, y1 , . . . , yn ].
So φ0 and j0 establish a bijection between the affine space An and the subset U0
of the projective space Pn . There are other similar maps φi and ji for i = 1, . . . , n.
So Pn is covered by n + 1 subsets, each one in natural bijection with An .
There is a natural way of thinking of Pn as a completion of An ; this is done
by identifying An with Ui via φi , and by interpreting the points of Hi (= Pn \ Ui )
6 Mezzetti

as points at infinity of An , or directions in An . We do this explicity for i = 0. First

of all we identify An with U0 via φ0 and j0 . So if P [a0 , . . . , an ] ∈ Pn , either a0 6= 0
and P ∈ An , or a0 = 0 and P [0, a1 , . . . , an ] ∈
/ An . Then we consider in An the line
L, passing through O(0, . . . , 0) and of direction given by the vector (a1 , . . . , an ).
Parametric equations for L are the following:
x1 = a1 t


x2 = a2 t


 ..
 .
xn = an t
with t ∈ K. The points of L are identified with points of U0 (via j0 ) with homoge-
neous coordinates x0 , . . . , xn given by:
x =1

 0

x1 = a1 t
 x2 = a2 t

...
or equivalently, if t 6= 0, by:  1
 x0 = t

x1 = a1 .
 x2 = a2

...
Now, roughly speaking, if t tends to infinity, this point goes to P [0, a1 , . . . , an ].
Clearly this is not a rigorous argument, but just a hint to the intuition.
In this way Pn can be interpreted as An with the points at infinity added,
each point at infinity corresponding to one direction in An .

Exercise to §1.
1*. Let V be a vector space of finite dimension over a field K. Let V̌ denote the
dual of V . Prove that P(V̌ ) can be put in bijection with the set of the hyperplanes
of P(V ) (hint: the kernel of a non-zero linear form on V is a subvector space of V
of codimension one).

2. Algebraic sets.

Roughly speaking, algebraic subsets of the affine or of the projective space are
sets of solutions of systems of algebraic equations, i.e. common roots of sets of
polynomials.
Examples of algebraic sets are: linear subspaces of both the affine and the pro-
jective space, plane algebraic curves, quadrics, graphics of polynomials functions,
...
Algebraic geometry is the branch of mathematics which studies algebraic sets
(and their generalizations). Our first aim is to give a formal definition of algebraic
sets.
 Introduction to algebraic geometry 7

Let K[x1 , . . . , xn ] be the polynomial ring in n variables over the field K. If

P (a1 , . . . , an ) ∈ An , and F = F (x1 , . . . , xn ) ∈ K[x1 , . . . , xn ], we can consider the
value of F at P , i.e. F (P ) = F (a1 , . . . , an ) ∈ K. We say that P is a zero of F if
F (P ) = 0.
For example the points P1 (1, 0), P2 (−1, 0), P3 (0, 1) are zeroes of F = x21 +
x22 − 1 over any field. If G = x21 + x22 + 1 then G has no zeroes in A2R , but it does
have zeroes in A2C .
2.1. Definition. A subset X of AnK is an affine algebraic set if X is the set of
common zeroes of a family of polynomials of K[x1 , . . . , xn ].
This means that there exists a subset S ⊂ K[x1 , . . . , xn ] such that

X = {P ∈ An | F (P ) = 0 ∀ F ∈ S}.

In this case X is called the zeroes set of S and is denoted V (S) (or in some books
Z(S), e.g. this is the notation of Hartshorne’s book). In particular, if S = {F },
then V (S) will be simply denoted by V (F ).
2.2. Examples and remarks.
1. S = K[x1 , . . . , xn ]: then V (S) = ∅, because S contains non–zero constants.
2. S = {0}: then V (S) = An .
3. S = {xy − 1} : then V (xy − 1) is the hyperbola.
4. If S ⊂ T , then V (S) ⊃ V (T ).
Let S ⊂ K[x1 , . . . , xn ] be a set of polynomials, let α := hSi be the ideal generated
by S. Recall that α = {finite sums of products of the form HF where F ∈ S, H ∈
K[x1 , . . . , xn ]}.

2.3. Proposition. V (S) = V (α).

Proof. If P ∈ V (α), then F (P ) = 0 for any F ∈ α; in particular for any F ∈ S
because S ⊂ α. P
Conversely,Pif P ∈ V (S), let
P G = i Hi Fi be a polynomial of α (Fi ∈ S ∀ i).
Then G(P ) = ( Hi Fi )(P ) = Hi (P )Fi (P ) = 0. 

The above Proposition is important in view of the following:

Hilbert’ Basis Theorem. If R is a Noetherian ring, then the polynomial ring
R[x] is Noetherian.
Proof. See for instance [4]. 

2.4. Corollary. Any affine algebraic set X ⊂ An is the zeroes set of a finite
number of polynomials, i.e. there exist F1 , . . . , Fr ∈ K[x1 , . . . , xn ] such that X =
V (F1 , . . . , Fr ).

8 Mezzetti

Note that V (F1 , . . . , Fr ) = V (F1 )∩. . .∩V (Fr ), so every algebraic set is a finite
intersection of algebraic sets of the form V (F ), i.e. zeroes of a unique polynomial
F . If F = 0, then V (0) = An ; if F = c ∈ K \ {0}, then V (c) = ∅; if deg F > 0,
then V (F ) is called a hypersurface.

2.5. Proposition. The affine algebraic sets of An satisfy the axioms of the
closed sets of a topology, called the Zariski topology.
Proof. It is enough to check that finite unions and arbitrary intersections of alge-
braic sets are again algebraic sets.
Let V (α), V (β) be two algebraic sets, with α, β ideals of K[x1 , . . . , xn ]. Then
V (α) ∪ V (β) = V (α ∩ β) = V (αβ), where αβ is the product ideal, defined by:
X
αβ = { ai bi | ai ∈ α, bi ∈ β}.
fin
In fact: αβ ⊂ α ∩ β so V (α ∩ β) ⊂ V (αβ), and both α ∩ β ⊂ α and α ∩ β ⊂ β
so V (α) ∪ V (β) ⊂ V (α ∩ β). Assume now that P ∈ V (αβ) and P ∈ / V (α): hence
∃F ∈ α such that F (P ) 6= 0; on the other hand, if G ∈ β then F G ∈ αβ so
(F G)(P ) = 0 = F (P )G(P ), which implies G(P ) = 0.
∈ I, be a familyPof algebraic sets, αi ⊂ K[x1 , . . . , xn ]. Then
Let V (αi ), i P
∩
Pi∈I V (αi ) = V ( i∈IP αi ), where i∈I αi is thePsum ideal of αi0 s. In fact αi ⊂
i∈I αi ∀i, hence V ( P i αi ) ⊂ V (αi ) ∀i Pand V ( i αi ) ⊂ ∩i V (αi ).PConversely, if
P ∈ V (αi ) ∀i, and F ∈ i αi , then F = i Fi ; therefore F (P ) = Fi (P ) = 0.


2.6. Examples.
1. The Zariski topology of the affine line A1 .
Let us recall that the polynomial ring K[x] in one variable is a PID (principal
ideal domain), so every ideal I ⊂ K[x] is of the form I = hF i. Hence every closed
subset of A1 is of the form X = V (F ), the set of zeroes of a unique polynomial
F (x). If F = 0, then V (F ) = A1 , if F = c ∈ K ∗ , then V (F ) = ∅, if deg F = d > 0,
then F can be decomposed in linear factors in polynomial ring over the algebraic
closure of K; it follows that V (F ) has at most d points.
We conclude that the closed sets in the Zariski topology of A1 are: A1 , ∅ and
the finite sets.
2. If K = R or C, then the Zariski topology and the usual topology on An
can be compared. Every closed set in the Zariski topology is closed also in the
usual topology: this is better seen working with open sets. Let X = V (F1 , . . . , Fr )
be a closed set in the Zariski topology, and U := An \ X; if P ∈ U , then ∃ Fi such
that Fi (P ) 6= 0, so there exists an open neighbourhood of P in the usual topology
in which Fi does not vanish.
Conversely, there exist closed sets in the usual topology which are not Zariski
closed, for example the balls. The first case, of an interval in the real affine line,
follows from part 1.
 Introduction to algebraic geometry 9

We want to define now the projective algebraic sets in Pn . Let K[x0 , x1 , . . . , xn ]

be the polynomial ring in n + 1 variables. Fix a polynomial G(x0 , x1 , . . . , xn ) ∈
K[x0 , x1 , . . . , xn ] and a point P [a0 , a1 , . . . , an ] ∈ Pn : then, in general,

G(a0 , . . . , an ) 6= G(λa0 , . . . , λan ),

so the value of G at P is not defined.

2.7. Example. Let G = x1 + x0 x1 + x22 , P [0, 1, 2] = [0, 2, 4] ∈ P2R . So G(0, 1, 2) =

1 + 4 6= G(0, 2, 4) = 2 + 16. But if Q = [1, 0, 0] = [λ, 0, 0], then G(1, 0, 0) =
G(λ, 0, 0) = 0 for all λ.

2.8. Definition. Let G ∈ K[x0 , x1 , . . . , xn ]: G is homogeneous of degree d if G

is a linear combination of monomials of degree d.

2.9. Lemma. If G is homogeneous of degree d, G ∈ K[x0 , x1 , . . . , xn ], and t is a

new variable, then G(tx0 , . . . , txn ) = td G(x0 , . . . , xn ).
Proof. It is enough to prove the equality for monomials, i.e. for

G = axi00 xi11 . . . xinn with i0 + i1 + . . . + in = d :

G(tx0 , . . . , txn ) = a(tx0 )i0 (tx1 )i1 . . . (txn )in = ati0 +i1 +...+in xi00 xi11 . . . xinn =
= td G(x0 , . . . , xn ).


2.10. Definition. Let G be a homogeneous polynomial of K[x0 , x1 , . . . , xn ]. A

point P [a0 , . . . , an ] ∈ Pn is a zero of G if G(a0 , . . . , an ) = 0. In this case we write
G(P ) = 0.
Note that by Lemma 2.9 if G(a0 , . . . , an ) = 0, then

G(λa0 , . . . , λan ) = λdeg G G(a0 , . . . , an ) = 0

for every choice of λ ∈ K ∗ .

2.11. Definition. A subset Z of Pn is a projective algebraic set if Z is the set

of common zeroes of a set of homogeneous polynomials of K[x0 , x1 , . . . , xn ].
If T is such a subset of K[x0 , x1 , . . . , xn ], then the corresponding algebraic set
will be denoted by VP (T ).
Let α = hT i be the Pideal generated by the (homogeneous) polynomials of T .
If F ∈ α, then P F = i Hi Fi , Fi ∈ T : if P ∈ VP (T ), and P [a0 , . . . , an ], then
F (a0 , . . . , an ) = Hi (a0 , . . . , an )Fi (a0 , . . . , an ) = 0, for any choice of coordinates
of P , regardless if F is homogeneous or not. We say that P is a projective zero of
F.
10 Mezzetti

If F is a polynomial, then F can be written in a unique way as a sum of

homogeneous polynomials, called the homogeneous components of F : F = F0 +
F1 + . . . + Fd . More in general, we give the following:

2.12. Definition. Let A be a ring. A is called a graded ring over Z if there exists
a family of additive subgroups {Ai }i∈Z such that A = ⊕i∈Z Ai and Ai Aj ⊂ Ai+j
for all pair of indices.
The elements of Ai are called homogeneous of degree i and Ai is the homoge-
neous component of degree i. The standard example of graded ring is the polyno-
mial ring with coefficients in a ring R. In this case the homogeneous components
of negative degrees are all zero.
2.13 Proposition - Definition. Let I ⊂ A be an ideal of a graded ring. I is
called homogeneous if the following equivalent conditions are fulfilled:
(i) I is generated by homogeneous elements;
(ii) I = ⊕k∈Z (I ∩ Ak ), i.e. if F = Σk∈Z Fk ∈ I, then all homogeneous components
Fk of F belong to I.
Proof of the equivalence.
“ (ii)⇒(i)”: given a system of generators of I, write each of them as sum of its
homogeneous components: Fi = Σk∈Z Fik . Then a set of homogeneous generators
of I is formed by all the elements Fik .
“ (i)⇒(ii)”: let I be generated by a family of homogeneous elements {Gα },
with deg Gα = dα . If F ∈ I, then F is a combination of the elements Gα with
suitable coefficients Hα ; write each Hα as sum of its homogeneous components:
Hα = ΣHαk . Note that the product Hαk Gα is homogeneous of degree k + dα . By
the unicity of the expression of F as sum of homogeneous elements, it follows that
all of them are combinations of the generators {Gα } and therefore they belong to
I. 

Let I ⊂ K[x0 , x1 , . . . , xn ] be a homogeneous ideal. Note that, by the noethe-

rianity, I admits a finite set of homogeneous generators.
Let P [a0 , . . . , an ] ∈ Pn . If F ∈ I, F = F0 + . . . + Fd , then F0 ∈ I, . . . , Fd ∈ I.
We say that P is a zero of I if P is a projective zero of any polynomial of I or,
equivalently, of any homogeneous polynomial of I. This also means that P is a
zero of any homogeneous polynomial of a set generating I. The set of zeroes of I
will be denoted VP (I): all projective algebraic subsets of Pn are of this form.
As in the affine case, the projective algebraic subsets of Pn satisfy the axioms
of the closed sets of a topology called the Zariski topology of Pn (see also Exercise
3).
Note that also all subsets of An and Pn have a structure of topological space,
with the induced topology, which is still called the Zariski topology.

Exercises to §2.
 Introduction to algebraic geometry 11

1. Let F ∈ K[x1 , . . . , xn ] be a non–constant polynomial. The set An \ V (F )

will be denoted AnF . Prove that {AnF |F ∈ K[x1 , . . . , xn ] \ K} is a topology basis
for the Zariski topology.
2. Let B ⊂ Rn be a ball. Prove that B is not Zariski closed.
3*. Let I, J be homogeneous ideals of K[x0 , x1 , . . . , xn ]. Prove that I + J, IJ
and I ∩ J are homogeneous ideals.
4*. Prove that the map φ : A1 → A3 defined by t → (t, t2 , t3 ) is a homeomor-
phism between A1 and its image, for the Zariski topology.
5. Let X ⊂ A2R be the graph of the map R → R such that x → sin x. Is X
closed in the Zariski topology? (hint: intersect X with a line....)

3. Examples of algebraic sets.

a) In the Zariski topology both of An and of Pn all points are closed.

If P (a1 , . . . , an ) ∈ An : P = V (x1 − a1 , . . . , xn − an ). If P [a0 , . . . , an ] ∈ Pn :
P = VP (hai xj − aj xi ii,j=0,...,n ).
Note that in the projective case the polynomials defining P as closed set are
homogeneous. They can be seen as minors of order 2 of the matrix
 
a0 a1 . . . an
x0 x1 . . . xn

with entries in K[x0 , x1 , . . . , xn ].

b) Hypersurfaces.
Let us recall that the polynomial ring K[x1 , . . . , xn ] is a UFD (unique fac-
torization domain), i. e. every non-constant polynomial F can be expressed in
a unique way (up to the order and up to units) as F = F1r1 F2r2 . . . Fsrs , where
F1 , . . . , Fs are irreducible polynomials, two by two distinct, and ri ≥ 1∀ i =
1, . . . , s. Hence the hypersurface of An defined by F is

X := V (F ) = V (F1r1 F2r2 . . . Fsrs ) = V (F1 F2 . . . Fs ) = V (F1 ) ∪ V (F2 ) ∪ . . . ∪ V (Fs ).

The equation F1 F2 . . . Fs = 0 is √ called the reduced equation of X. Note

that F1 F2 . . . Fs generates the radical F . If s = 1, X is called an irreducible
hypersurface; by definition its degree is the degree of its reduced equation. Any
hypersurface is a finite union of irreducible hypersurfaces.
In a similar way one defines hypersurfaces of Pn , i. e. projective algebraic
sets of the form Z = VP (G), with G ∈ K[x0 , x1 , . . . , xn ], G homogeneous. Since
the irreducible factors of G are homogeneous (see Exercise 3.6), any projective
hypersurface Z has a reduced equation (whose degree is, by definition, the degree
of Z) and Z is a finite union of irreducible hypersurfaces. The degree of a projective
hypersurface has the following important geometrical meaning.
12 Mezzetti

3.1. Proposition. Let K be an algebraically closed field. Let Z ⊂ Pn be a

projective hypersurface of degree d. Then a line of Pn , not contained in Z, meets
Z at exactly d points, counting multiplicities.
Proof. Let G be the reduced equation of Z and L ⊂ Pn be any line.
We fix two points on L: A = [a0 , . . . , an ], B = [b0 , . . . , bn ]. So L admits
parametric equations of the form

x = λa0 + µb0

 0

x1 = λa1 + µb1
...

xn = λan + µbn

The points of Z ∩ L are obtained from the homogeneous pairs [λ, µ] which are
solutions of the equation G(λa0 + µb0 , . . . , λan + µbn ) = 0. If L ⊂ Z, then this
equation is identical. Otherwise, G(λa0 + µb0 , . . . , λan + µbn ) is a non-zero homo-
geneous polynomial of degree d in two variables. Being K algebraically closed, it
can be factorized in linear factors:

G(λa0 + µb0 , . . . , λan + µbn ) = (µ1 λ − λ1 µ)d1 (µ2 λ − λ2 µ)d2 . . . (µr λ − λr µ)dr

with d1 + d2 + . . . + dr = d. Every factor corresponds to a point in Z ∩ L, to be

counted with the same multiplicity as the factor. 
If K is not algebraically closed, considering the algebraic closure of K and
using Proposition 3.1, we get that d is un upper bound on the number of points
of Z ∩ L.
c) Affine and projective subspaces.
The subspaces introduced in §1, both in the affine and in the projective case,
are examples of algebraic sets.
d) Product of affine spaces.
Let An , Am be two affine spaces over the field K. The cartesian product
An × Am is the set of pairs (P, Q), P ∈ An , Q ∈ Am : it is in natural bijection with
An+m via the map
φ : An × Am −→ An+m
such that φ((a1 , . . . , an ), (b1 , . . . , bm )) = (a1 , . . . , an , b1 , . . . , bm ).
From now on we will always identify An × Am with An+m . We get two
topologies on An × Am : the Zariski topology and the product topology.

3.1. Proposition. The Zariski topology is strictly finer than the product topol-
ogy.
Proof. If X = V (α) ⊂ An , α ⊂ K[x1 , . . . , xn ] and Y = V (β) ⊂ Am , β ⊂
K[y1 , . . . , ym ], then X × Y ⊂ An × Am is Zariski closed, precisely X × Y =
 Introduction to algebraic geometry 13

V (α ∪ β) where the union is made in the polynomial ring in n + m variables

K[x1 , . . . , xn , y1 , . . . , ym ]. Hence, if U = An \ X, V = Am \ Y are open subsets of
An and Am in the Zariski topology, then U ×V = An ×Am \((An ×Y )∪(X ×Am ))
is open in An × Am in the Zariski topology.
Conversely, we prove that A1 × A1 = A2 contains some subsets which are
Zariski open but are not open in the product topology. The proper open subsets
in the product topology are of the form A1 × A1 \ { finite unions of “ vertical” and
“ horizontal” lines}.

– Fig. 1 –
Let X = A2 \ V (x − y): it is Zariski open but does not contain any non-empty
subset of the above form, so it is not open in the product topology. There are
similar examples in An × Am for any n, m. 
Note that there is no similar construction for Pn × Pm .

e) Embedding of An in Pn .
Let Hi be the hyperplane of Pn of equation xi = 0, i = 0, . . . , n; it is closed
in the Zariski topology, and the complementar set Ui is open. So we have an open
covering of Pn : Pn = U0 ∪ U1 ∪ . . . ∪ Un . Let us recall that for all i there is a
bijection φi : Ui → An such that φi ([x0 , . . . , xi , . . . , xn ]) = ( xx0i , . . . , 1̂, . . . , xxni ). The
inverse map is ji : An → Ui such that ji (y1 , . . . , yn ) = [y1 , . . . , 1, . . . , yn ].

3.2. Proposition. The map φi is a homeomorphism, for i = 0, . . . , n.

Proof. Assume i = 0 (the other cases are similar).
We introduce two maps:
(i) dehomogeneization of polynomials with respect to x0 .
14 Mezzetti

a
It is a map : K[x0 , x1 , . . . , xn ] → K[y1 , . . . , yn ] such that
a
(F (x0 , . . . , xn )) = a F (y1 , . . . , yn ) := F (1, y1 , . . . , yn ).

Note that a is a ring homomorphism.

(ii) homogeneization of polynomials with respect to x0 .
It is a map h : K[y1 , . . . , yn ] → K[x0 , x1 , . . . , xn ] defined by
h x1 xn
(G(y1 , . . . , yn )) = h G(x0 , . . . , xn ) := xdeg
0
G
G( ,..., )
x0 x0
.
h
G is always a homogeneous polynomial of the same degree as G. The map h
is clearly not a ring homomorphism. Note that always a (h G) = G but in general
h a
( F ) 6= F ; what we can say is that, if F (x0 , . . . , xn ) is homogeneous, then ∃r ≥ 0
such that F = xr0 (h (a F )).
Let X ⊂ U0 be closed in the topology induced by the Zariski topology of
the projective space, i.e. X = U0 ∩ VP (I) where I is a homogeneous ideal of
K[x0 , x1 , . . . , xn ]. Define a I = {a F | F ∈ I}: it is an ideal of K[y1 , . . . , yn ]
(because a is a ring homomorphism). We prove that φ0 (X) = V (a I). For:
let P [x0 , . . . , xn ] be a point of U0 ; then φ0 (P ) = ( xx10 , . . . , xxn0 ) ∈ φ0 (X) ⇐⇒
P [x0 , . . . , xn ] = [1, xx10 , . . . , xxn0 ] ∈ X = VP (I) ⇐⇒ F (1, xx01 , . . . , xxn0 ) = 0 ∀ a F ∈
a
I ⇐⇒ φ0 (P ) ∈ V (a I).
Conversely: let Y = V (α), α ideal of K[y1 , . . . , yn ], be a Zariski closed set
of A . Let h α be the homogeneous ideal of K[x0 , x1 , . . . , xn ] generated by the set
n

{h G | G ∈ α}. We prove that φ−1 h

0 (Y ) = VP ( α) ∩ U0 . In fact: [1, x0 , . . . , xn ] ∈
φ−1
0 (Y ) ⇐⇒ (x1 , . . . , xn ) ∈ Y ⇐⇒ G(x1 , . . . , xn ) =
h
G(1, x1 , . . . , xn ) = 0 ∀ G ∈
h
α ⇐⇒ [1, x1 , . . . , xn ] ∈ VP ( α). 

From now on we will often identify An with U0 via φ0 (and similarly with Ui
via φi ). So if P [x0 , . . . , xn ] ∈ U0 , we will refer to x0 , . . . , xn as the homogeneous
coordinates of P and to xx10 , . . . , xxn0 as the non–homogeneous or affine coordinates
of P .

Exercises to §3.
1*. Let n ≥ 2. Prove that, if K is an algebraically closed field, then in AnK
both any hypersurface and any complementar set of a hypersurface have infinitely
many points.
2. Prove that the Zariski topology on An is T1 .
3*. Let F ∈ K[x0 , x1 , . . . , xn ] be a homogeneous polynomial. Check that its
irreducible factors are homogeneous. (hint: consider a product of two polynomials
not both homogeneous...)

4. The ideal of an algebraic set and the Hilbert Nullstellensatz.

 Introduction to algebraic geometry 15

Let X ⊂ An be an algebraic set, X = V (α), α ⊂ K[x1 , . . . , xn ]. The ideal α

defining X is not unique: for example, let X = {0} ⊂ A2 ; then 0 = V (x1 , x2 ) =
V (x21 , x2 ) = V (x21 , x32 ) = V (x21 , x1 , x2 , x22 ) = . . . Nevertheless, there is an ideal we
can canonically associate to X, i.e. the biggest one. Precisely:
4.1. Definition. Let Y ⊂ An be any set.
The ideal of Y is I(Y ) = {F ∈ K[x1 , . . . , xn ] | F (P ) = 0 for any P ∈ Y } =
{F ∈ K[x1 , . . . , xn ] | Y ⊂ V (F )}: it is formed by all polynomials vanishing on Y .
Note that I(Y ) is in fact an ideal.
For instance, if P (a1 , . . . , an ) is a point, then I(P ) = hx1 − a1 , . . . , xn − an i.
Indeed all its polynomials vanish on P , and, on the other side, it is maximal.
The following relations follow immediately by the definition:
(i) if Y ⊂ Y 0 , then I(Y ) ⊃ I(Y 0 );
(ii) I(Y ∪ Y 0 ) = I(Y ) ∩ I(Y 0 );
(iii) I(Y ∩ Y 0 ) ⊃ I(Y ) + I(Y 0 ).
Similarly, if Z ⊂ Pn is any set, the homogeneous ideal of Z is, by def-
inition, the homogeneous ideal of K[x0 , x1 , . . . , xn ] generated by the set {G ∈
K[x0 , x1 , . . . , xn ] | G is homogeneous and VP (G) ⊃ Z}. It is denoted Ih (Z).
Relations similar to (i),(ii),(iii) are satisfied. Ih (Z) is also the set of polyno-
mials F (x0 , . . . , xn ) such that every point of Z is a projective zero of F .
√
Let α ⊂ K[x1 , . . . , xn ] be an ideal. Let α denote the radical of α, i.e. √ the
ideal {F ∈ K[x1 , . . . , xn ] | ∃r ≥ 1 s.t. F r ∈ α}. Note that always α ⊂ α; if
equality holds, then α is called a radical ideal.

4.2. Proposition.
1) For any X ⊂ An , I(X) is a radical ideal.
2) For any Z ⊂ Pn , Ih (Z) is a homogeneous radical ideal.
p
Proof. 1) If F ∈ I(X), let r ≥ 1 such that F r ∈ I(X): hence if P ∈ X, then
(F r )(P ) = 0 = (F (P ))r in the base field K. Therefore F (P ) = 0.
2) is similar, taking into account that Ih (Z) is a homogeneous ideal (see
Exercise 4.7.). 

We can interpret I as a map from P(An ), the set of subsets of the affine space,
to P(K[x1 , . . . , xn ]). On the other hand, V can be seen as a map in the opposite
sense. We have:

4.3. Proposition. Let α ⊂ K[x1 , . . . , xn ] be an ideal, Y ⊂ An be any subset.

Then:
(i) α ⊂ I(V (α));
(ii) Y ⊂ V (I(Y ));
(iii) V (I(Y )) = Y : the closure of Y in the Zariski topology of An .
16 Mezzetti

Proof. (i) If F ∈ α and P ∈ V (α), then F (P ) = 0, so F ∈ I(V (α)).

(ii) If P ∈ Y and F ∈ I(Y ), then F (P ) = 0, so P ∈ V (I(Y )).
(iii) Taking closures in (ii), we get: Y ⊂ V (I(Y )) = V (I(Y )). Conversely,
let X = V (β) be any closed set containing Y : X = V (β) ⊃ Y . Then I(Y ) ⊃
I(V (β)) ⊃ β by (i); we apply V again: V (β) = X ⊃ V (I(Y )) so any closed set
containing Y contains V (I(Y )) so Y ⊃ V (I(Y )). 

Similar properties relate homogeneous ideals of K[x0 , x1 , . . . , xn ] and subsets

of Pn ; in particular, if Z ⊂ Pn , then VP (Ih (Z)) = Z, the closure of Z in the Zariski
topology of Pn .
There does not exist any characterization of I(V (α)) in general.
√ We can only
say that it is a radical ideal containing α, so it contains also α. To characterize
I(V (α)) we need some extra assumption on the base field.

4.4. Hilbert Nullstellensatz (Theorem of zeroes). Let K be√an algebraically

closed field. Let α ⊂ K[x1 , . . . , xn ] be an ideal. Then I(V (α)) = α.

The assumption on K is necessary. Let me recall that K is algebraically

closed if any non–constant polynomial of K[x] has at least one root in K, or,
equivalently, if any irreducible polynomial of K[x] has degree 1. So if K is not
algebraically closed, there exists F ∈ K[x], irreducible of degree d > 1. Therefore
F has no zero in K, hence V (F ) ⊂ A1K is empty. So I(V (F )) = I(∅) =p {G ∈
K[x] | ∅ ⊂pV (G)} = K[x]. But hF ip is a maximal ideal of K[x], and hF i ⊂ hF i.
If hF i =
6 hF i, by thepmaximality hF i = h1i, so ∃r ≥ 1 such that 1r = 1 ∈ hF i,
which is false. Hence hF i = hF i = 6 K[x] = I(V (F )).

Proof. of the Nullstellensatz. It will follow from two facts:

1st fact. Let K be an algebraically closed field, let M ⊂ K[x1 , . . . , xn ] be a
maximal ideal. Then, there exist a1 , . . . , an ∈ K such that M = hx1 − a1 , . . . , xn −
an i.
2nd fact (Weak Nullstellensatz).
Let K be an algebraically closed field, let α ⊂ K[x1 , . . . , xn ] be a proper ideal.
Then V (α) 6= ∅ i.e. the polynomials of α have at least one common zero in AnK .
Indeed, since α is proper, there exists a maximal ideal M containing α. Then
V (α) ⊃ V (M). By 1st Step, M = hx1 − a1 , . . . , xn − an i, so V (M) = {P } with
P (a1 , . . . , an ), hence P ∈ V (α).
3rd Step (Rabinowitch method).
√
Let K be an algebraically closed field: we will prove that I(V (α)) ⊂ α.
Since the reverse inclusion always holds, this will conclude the proof.
Let F ∈ I(V (α)), F 6= 0 and let α = hG1 , . . . , Gr i. The assumption on
F means: if G1 (P ) = . . . = Gr (P ) = 0, then F (P ) = 0. Let us consider the
polynomial ring in n + 1 variables K[x1 , . . . , xn+1 ] and let β be the ideal β =
 Introduction to algebraic geometry 17

hG1 , . . . , Gr , xn+1 F − 1i: β has no zeroes in An+1 , hence, by Step 1, 1 ∈ β, i.e.

there exist H1 , . . . , Hr+1 ∈ K[x1 , . . . , xn+1 ] such that

1 = H1 G1 + . . . + Hr Gr + Hr+1 (xn+1 F − 1).

We introduce the homomorphism ψ : K[x1 , . . . , xn+1 ] → K(x1 , . . . , xn ) defined

by H(x1 , . . . , xn+1 ) → H(x1 , . . . , xn , F1 ).
The polynomials G1 , . . . , Gr do not contain xn+1 so ψ(Gi ) = Gi ∀ i = 1, . . . , r.
Moreover ψ(xn+1 F − 1) = 0, ψ(1) = 1. Therefore

1 = ψ(H1 G1 + . . . + Hr Gr + Hr+1 (xn+1 F − 1)) = ψ(H1 )G1 + . . . + ψ(Hr )Gr

where ψ(Hi ) is a rational function with denominator a power of F . By multiplying

this relation by a common denominator, we get an expression of the form:

F m = H10 G1 + . . . + Hr0 Gr ,
√
so F ∈ α. 

4.10. Corollaries. Let K be an algebraically closed field.

1. There is a bijection between algebraic subsets of An and radical ideals of
K[x1 , . . . , xn ]. The bijection is given by α → V (α) and X → I(X). In fact, if X
is closed in the Zariski topology, then V (I(X)) = X; if α is a radical ideal, then
I(V (α)) = α.
2. Let X, Y ⊂ An bepclosed sets. Then
(i) I(X ∩ Y ) = I(X) + I(Y ); p
(ii) I(X ∪ Y ) = I(X) ∩ I(Y ) = I(X)I(Y ).
Proof. 2. follows from next lemma, using the Nullstellensatz.

4.11.
p√Lemma. Let α, β be ideals of K[x1 , . . . , xn ]. Then
√
a) α = α;
√ p√ √
b) √ α + β = √ α +√ β; √
c) α ∩ β = αβ = α ∩ β.
Proof. p√ √
a) if F ∈ α, there exists r ≥ 1 such that F r ∈ α, hence there exists s ≥ 1
rs
√ F ∈ α.
such that √ √ √ √ p√ √
b) α ⊂ α, β ⊂ β imply α + β ⊂ α + β√hence√ α + β √ ⊂ √α + β.
Conversely, α ⊂ α +pβ, β ⊂ α + β pimply α ⊂ α + β, β ⊂ α + β, hence
√ √ √ √ √ √ √
α + β ⊂ α + β so α+ β ⊂ α +√β = α+ √β. √ √
c) αβ√ ⊂ α√∩ β ⊂ α (resp. ⊂ β) therefore αβ ⊂ α ∩ β ⊂ α ∩ β. If
F ∈ √ α ∩ β, then F r ∈ α, F s ∈ β for suitable r, s ≥ 1, hence F r+s ∈ αβ, so
F ∈ αβ. 
18 Mezzetti

Part 2.(i) of 4.10. implies that, if I(X ∩ Y ) 6= I(X) + I(Y ), then I(X) + I(Y )
is not radical.

We move now to projective space. There exist proper homogeneous ideals of

K[x0 , x1 , . . . , xn ] without zeroes in Pn , also assuming K algebraically closed: for
example the maximal ideal hx0 , x1 , . . . , xn i. The following characterization holds:

4.12. Proposition. Let K be an algebraically closed field and let I be a

homogeneous ideal of K[x0 , x1 , . . . , xn ].
The following are equivalent:
(i) VP (I) = ∅; √
(ii) either I = K[x0 , x1 , . . . , xn ] or I = hx0 , x1 , . . . , xn i;
(iii) ∃d ≥ 1 such that I ⊃ K[x0 , x1 , . . . , xn ]d , the subgroup of K[x0 , x1 , . . . , xn ]
formed by the homogeneous polynomials of degree d.
Proof.
(i)⇒(ii) Let p : An+1 − {0} → Pn be the canonical surjection. We have:
VP (I) = p(V (I)−{0}), where V (I) ⊂ An+1 . So if VP (I) = ∅, then either V (I) = ∅
or V (I) = {0}. If V (I) = ∅ then I(V √ (I)) = I(∅) = K[x0 , x1 , . . . , xn ]; if V (I) = {0},
then I(V (I)) = hx0 ,√ x1 , . . . , xn i = I by the Nullstellensatz. √
√ (ii)⇒(iii) Let I = K[x 0 , x1 , . . . , xn ], then 1 ∈ I so 1r = 1 ∈ I(r ≥ 1). If
I = hx0 , x1 , . . . , xn i, then for all variable xk there exists an index ik ≥ 1 such
that xikk ∈ I. If d ≥ i0 + i1 + . . . + in , then any monomial of degree d is in I, so
K[x0 , x1 , . . . , xn ]d ⊂ I.
(iii)⇒(i) because no point in Pn has all coordinates equal to 0. 

4.13. Theorem. Let K be an algebraically closed field and I be a homogeneous

ideal of K[x0 , x1 , . . . , xn ]. If F is a homogeneous non–constant
√ polynomial such
that VP (F ) ⊃ VP (I) (i.e. F vanishes on VP (I), then F ∈ I.
Proof. We have p(V (I) − {0}) = VP (I) ⊂ VP (F ). Since F is non–constant, we
have also V√(F ) = p−1 (VP (F
p)) ∪ {0}, so V (F ) ⊃ V (I); by the Nullstellensatz
I(V (I)) = I ⊃ I(V (F )) = (F ) 3 F . 

4.14. Corollary (homogeneous Nullstellensatz). Let I be a homogeneous √

ideal of K[x0 , x1 , . . . , xn ] such that VP (I) 6= ∅, K algebraically closed. Then I =
Ih (VP (I)). 

√
4.15. Definition. A homogeneous ideal of K[x0 , x1 , . . . , xn ] such that I =
hx0 , x1 , . . . , xn i is called irrelevant.

4.16. Corollary. Let K be an algebraically closed field. There is a bijec-

tion between the set of projective algebraic subsets of Pn and the set of radical
 Introduction to algebraic geometry 19

homogeneous non–irrelevant ideals of K[x0 , x1 , . . . , xn ]. 

Remark. Let X ⊂ Pn be an algebraic set, X 6= ∅. The affine cone of

X, denoted C(X), is the following subset of An+1 : C(X) = p−1 (X) ∪ {0}. If
X = VP (F1 , . . . , Fr ), with F1 , . . . , Fr homogeneous, then C(X) = V (F1 , . . . , Fr ).
By the Nullstellensatz, if K is algebraically closed, I(C(X)) = Ih (X).

Exercises to §4.
√
1. Give a non-trivial example of an ideal α of K[x1 , . . . , xn ] such that α 6= α.
2. Show that the following closed subsets of the affine plane Y = V (x2 +y 2 −1)
and Y 0 = V (y − 1) are such that equality does not hold in the following relation:
I(Y ∩ Y 0 ) ⊃ I(Y ) + I(Y 0 ).
√
3. Let α ⊂ K[x1 , . . . , xn ] be an ideal. Prove that α = α if and only if the
quotient ring K[x1 , . . . , xn ]/α does not contain non–zero nilpotents.
4*. Let I be a homogeneous ideal of K[x1 , . . . , xn ] satisfying the following
condition: if F is a homogeneous polynomial such that F r ∈ I for some positive
integer r, then F ∈ I. Prove that I is a radical ideal.

5. The projective closure of an affine algebraic set.

Let X ⊂ An be Zariski closed. Fix an index i ∈ {0, . . . , n} and embed An into Pn

φi
as the open subset Ui . So X ⊂ An ,→ Pn .

5.1. Definition. The projective closure of X, X, is the closure of X in the

Zariski topology of Pn .
Since the map φi is a homeomorphism (see Proposition 3.2.), we have: X ∩
An = X because X is closed in An . The points of X ∩ Hi , where Hi = VP (xi ),
are called the “ points at infinity” of X in the fixed embedding.
Note that, if K is an infinite field, then the projective closure of An is Pn :
indeed, let F be a homogeneous polynomial vanishing along An = U0 . We can
write F = F0 xd0 + F1 xd−1 0 + . . . + Fd . By assumption, for every P (a1 , . . . , an ) ∈ An ,
P ∈ VP (F ), i.e. F (1, a1 , . . . , an ) = 0 = a F (a1 , . . . , an ). So a F ∈ I(An ). We
claim that I(An ) = (0): if n = 1, this follows from the principle of identity of
polynomials, because K is infinite. If n ≥ 2, assume that F (a1 , ..., an ) = 0 for
all (a1 , ..., an ) ∈ K n and consider F (a1 , ..., an−1 , x): either it has positive degree
in x for some choice of (a1 , ..., an ), but then it has finitely many zeroes against
the assumption; or it is always constant in x, so F belongs to K[x1 , ..., xn−1 ] and
we can conclude by induction. So the claim is proved. We get therefore that
F0 = F1 = . . . = Fd = 0 and F = 0.
20 Mezzetti

5.2. Proposition. Let X ⊂ An be an affine algebraic set, X be the projective

closure of X. Then

Ih (X) = h I(X) := hh F |F ∈ I(X)i.

Proof. Assume An = U0 ⊂ Pn .
Let F ∈ Ih (X) be a homogeneous polynomial. If P (a1 , . . . , an ) ∈ X, then
[1, a1 , . . . , an ] ∈ X, so F (1, a1 , . . . , an ) = 0 = a F (a1 , . . . , an ). Hence a F ∈ X.
There exists k ≥ 0 such that F = (xk0 )h (a F ) (see Proposition 3.2), so F ∈ h I(X).
Hence Ih (X) ⊂ h I(X).
Conversely, if G ∈ I(X) and P (a1 , . . . , an ) ∈ X, then G(a1 , . . . , an ) = 0 =
h
G(1, a1 , . . . , an ), so h G ∈ Ih (X) (here X is seen as a subset of Pn ). So h I(X) ⊂
Ih (X). Since Ih (X) = Ih (X) (see Exercise 5.1), we have the claim.


In particular, if X is a hypersurface and I(X) = hF i, then Ih (X) = hh F i.

Next example will show that, in general, it is not true that, if I(X) =
hF1 , . . . , Fr i, then h I(X) = hh F1 , . . . , h Fr i. Only in the last twenty years, thanks
to the development of symbolic algebra and in particular of the theory of Groeb-
ner bases, the problem of characterizing the systems of generators of I(X), whose
homogeneization generates h I(X), has been solved.

5.3. Example. The skew cubic.

Let K be an algebraically closed field. The affine skew cubic is the following
closed subset of A3 : X = V (y − x2 , z − x3 ) (we use variables x, y, z). X is the
image of the map φ : A1 → A3 such that φ(t) = (t, t2 , t3 ). Note that φ : A1 → X
is a homeomorphism (see Exercise 2.4). The ideal α = hy − x2 , y − x3 i defines X
and is prime: indeed the quotient ring K[x, y, z]/α is isomorphic to K[x], hence
an integral domain. Therefore α is radical so α = I(X).
Let X be the projective closure of X in P3 . We are going to prove that X
is the image of the map ψ : P1 → P3 such that ψ([λ, µ]) = [λ3 , λ2 µ, λµ2 , µ3 ]. We
identify A1 with the open subset of P1 defined by λ 6= 0 i.e. U0 , and A3 with
the open subset of P3 defined by x0 6= 0 (U0 too). Note that ψ|A1 = φ, because
ψ([1, t]) = [1, t, t2 , t3 ] = via the identification of A3 with U0 = (t, t2 , t3 ) = φ(t).
Moreover ψ([0, 1]) = [0, 0, 0, 1]. So ψ(P1 ) = X ∪ {[0, 0, 0, 1]}.
If G is a homogeneous polynomial of K[x0 , x1 , . . . , x3 ] such that X ⊂ VP (G),
then G(1, t, t2 , t3 ) = 0 ∀t ∈ K, so G(λ3 , λ2 µ, λµ2 , µ3 ) = 0 ∀µ ∈ K, ∀λ ∈ K ∗ .
Since K is infinite, then G(λ3 , λ2 µ, λµ2 , µ3 ) is the zero polynomial in λ and µ, so
G(0, 0, 0, 1) = 0 and VP (G) ⊃ ψ(P1 ), therefore X ⊃ ψ(P1 ).
Conversely, it is easy to prove that ψ(P1 ) is Zariski closed, in fact that ψ(P1 ) =
VP (x21 − x0 x2 , x1 x2 − x0 x3 , x22 − x1 x3 ). So ψ(P1 ) = X.
 Introduction to algebraic geometry 21

The three polynomials F0 := x1 x3 − x22 , F1 := x1 x2 − x0 x3 , F2 := x0 x2 − x21

are the 2 × 2 minors of the matrix
 
x0 x1 x2
M=
x1 x2 x3

with entries in K[x0 , x1 , . . . , x3 ]. Let F = y − x2 , G = z − x3 be the two generators

of I(X); h F = x0 x2 −x21 , h G = x20 x3 −x31 , hence VP (h F, h G) = VP (x0 x2 −x21 , x20 x3 −
x31 ) 6= X, because VP (h F, h G) contains the whole line VP (x0 , x1 ).

We shall prove now the non-trivial fact:

5.4. Proposition. Ih (X) = hF0 , F1 , F2 i.

Proof. For all integer number d ≥ 0, let I
h (X)d := Ih (X) ∩ K[x0 , x1 , . . . , x3 ]d : it
is a K-vector space of dimension ≤ d+3 3 . We define a K-linear map ρd having
Ih (X)d as kernel:
ρd : K[x0 , x1 , . . . , x3 ]d → K[λ, µ]3d
such that ρd (F ) = F (λ3 , λ2 µ, λ2 µ2 , µ3 ). Since ρd is clearly surjective, we compute
 
d+3
dim Ih (X)d = − (3d + 1) = (d3 + 6d2 − 7d)/6.
3

For d ≥ 2, we define now a second K-linear map

φd : K[x0 , x1 , . . . , x3 ]d−2 ⊕ K[x0 , x1 , . . . , x3 ]d−2 ⊕ K[x0 , x1 , . . . , x3 ]d−2 → Ih (X)d

such that φd (G0 , G1 , G2 ) = G0 F0 + G1 F1 + G2 F2 . Our aim is to prove that φd is

surjective. The elements of its kernel are called the syzygies of degree d among
the polynomials F0 , F1 , F2 . Two obvious syzygies of degree 3 are constructed by
developing, according to the Laplace rule, the determinant of the matrix obtained
repeating one of the rows of M , for example
 
x0 x1 x2
 x0 x1 x2 
x1 x2 x3

We put H1 = (x0 , x1 , x2 ) and H2 = (x1 , x2 , x3 ), they both belong to ker φ3 . Note

that H1 and H2 give raise to syzygies of all degrees ≥ 3, in fact we can construct
a third linear map

ψd : K[x0 , x1 , . . . , x3 ]d−3 ⊕ K[x0 , x1 , . . . , x3 ]d−3 → ker φd

putting ψd (A, B) = H1 A + H2 B = (x0 , x1 , x2 )A + (x1 , x2 , x3 )B.

22 Mezzetti

Claim. ψd is an isomorphism.
Assuming the claim, we are able to compute dim ker φd = 2 d3 , therefore

   
d+1 d
dim Im φd = 3 −2
3 3

which coincides with the dimension of Ih (X)d previously computed. This proves
that φd is surjective for all d and concludes the proof of the Proposition.
Proof of the Claim. Let (G0 , G1 , G2 ) belong to ker φd . This means that the
following matrix N with entries in K[x0 , x1 , . . . , x3 ] is degenerate:
 
G0 G1 G2
N :=  x0 x1 x2 
x1 x2 x3

Therefore, the rows of N are linearly dependent over the quotient field of the
polynomial ring K(x0 , . . . , x3 ). Since the last two rows are independent, there
exist reduced rational functions aa01 , bb10 ∈ K(x0 , x1 , x2 , x3 ), such that

a1 b1 a1 b0 x0 + a0 b1 x1
G0 = x0 + x1 =
a0 b0 a0 b0

and similarly
a1 b0 x1 + a0 b1 x2 a1 b0 x2 + a0 b1 x3
G1 = , G2 =
a0 b0 a0 b0
The Gi ’s are polynomials, therefore the denominator a0 b0 divides the numerator
in each of the three expressions on the right hand side. Moreover, if p is a prime
factor of a0 , then p divides the three products b0 x0 , b0 x1 , b0 x2 , hence p divides b0 .
We can repeat the reasoning for a prime divisor of b0 , so obtaining that a0 = b0
(up to invertible constants). We get:

a1 x0 + b1 x1 a1 x1 + b1 x2 a1 x2 + b1 x3
G0 = , G1 = , G2 = ,
b0 b0 b0

therefore b0 divides the numerators

c0 := a1 x0 + b1 x1 , c1 := a1 x1 + b1 x2 , c2 := a1 x2 + b1 x3 .

Hence b0 divides also x1 c0 − x0 c1 = b1 (x21 − x0 x1 ) = −b1 F2 , and similarly x2 c0 −

x0 c2 = b1 F1 , x2 c1 − x1 c2 = −b1 F0 . But F0 , F1 , F2 are irreducible and coprime, so
we conclude that b0 | b1 . But b0 and b1 are coprime, so finally we get b0 = a0 = 1.

 Introduction to algebraic geometry 23

As a by-product of the proof of Proposition 5.4 we have the minimal free

resolution of the R-module Ih (X), where R = K[x0 , x1 , . . . , x3 ]:
ψ φ
0 → R⊕2 −→ R⊕3 −→ Ih (X) → 0
where ψ is represented by the transposed of the matrix M and φ by the triple of
polynomials (F0 , F1 , F2 ).

Exercises to §5.
1*. Let X ⊂ An be a closed subset, X be its projective closure in Pn . Prove
that Ih (X) = Ih (X).
2. Find a system of generators of the ideal of the affine skew cubic X, such
that, if you homogeneize them, you get a system of generators for Ih (X).

6. Irreducible components.

6.1. Definition. Let X 6= ∅ be a topological space. X is irreducible if the

following condition holds: if X1 , X2 are closed subsets of X such that X = X1 ∪X2 ,
then either X = X1 or X = X2 . Equivalently, X is irreducible if for all pair of non–
empty open subsets U , V we have U ∩ V 6= ∅. By definition, ∅ is not irreducible.

6.2. Proposition. X is irreducible if and only if any non–empty open subset U

of X is dense.
Proof. Let X be irreducible, let P be a point of X and IP be an open neighbourhood
of P in X. IP and U are non–empty and open, so IP ∩ U 6= ∅, therefore P ∈ U .
This proves that U = X.
Conversely, assume that open subsets are dense. Let U , V 6= ∅ be open
subsets. Let P ∈ U be a point. By assumption P ∈ V = X, so V ∩ U 6= ∅ (U is
an open neighbourhood of P ). 

Examples.
1. If X = {P } a unique point, then X is irreducible.
2. Let K be an infinite field. Then A1 is irreducible, because proper closed
subsets are finite sets. The same holds for P1 .
3. Let f : X → Y be a continuous map of topological spaces. If X is
irreducible and f is surjective, then Y is irreducible.
4. Let Y ⊂ X be a subset, give it the induced topology. Then Y is irreducible
if and only if the following holds: if Y ⊂ Z1 ∪ Z2 , with Z1 and Z2 closed in X,
then either Y ⊂ Z1 or Y ⊂ Z2 ; equivalently: if Y ∩ U 6= ∅, Y ∩ V 6= ∅, with U , V
open subsets of X, then Y ∩ U ∩ V 6= ∅.

6.3. Proposition. Let X be a topological space, Y a subset of X. Y is

irreducible if and only if Y is irreducible.
24 Mezzetti

Proof. Note first that if U ⊂ X is open and U ∩ Y = ∅ then U ∩ Y = ∅. Otherwise,

if P ∈ U ∩ Y , let A be an open neighbourhood of P : then A ∩ Y 6= ∅. In particular,
U is an open neighbourhood of P so U ∩ Y 6= ∅.
Let Y be irreducible. If U and V are open subsets of X such that U ∩ Y 6= ∅,
V ∩ Y 6= ∅, then U ∩ Y 6= ∅ and V ∩ Y 6= ∅ so Y ∩ U ∩ V 6= ∅ by irreducibility of
Y . Hence Y ∩ (U ∩ V ) 6= ∅. So Y is irreducible. If Y is irreducible, we get the
irreducibility of Y in a completely analogous way. 

6.4. Corollary. Let X be an irreducible topological space and U be a non–empty

open subset of X. Then U is irreducible.
Proof. By Proposition 6.2 U = X which is irreducible. By Proposition 6.3 U is
irreducible. 

For algebraic sets (both affine and projective) irreducibility can be expressed
in a purely algebraic way.

6.5. Proposition. Let X ⊂ An ( resp. Pn ) be an algebraic set. X is irreducible

if and only if I(X) (resp. Ih (X)) is prime.
Proof. Assume first that X is irreducible, X ⊂ An . Let F, G polynomials of
K[x1 , . . . , xn ] such that F G ∈ I(X): then

V (F ) ∪ V (G) = V (F G) ⊃ V (I(X)) = X

hence either X ⊂ V (F ) or X ⊂ V (G). In the former case, if P ∈ X then F (P ) = 0,

so F ∈ I(X), in the second case G ∈ I(X); hence I(X) is prime.
Assume now that I(X) is prime. Let X = X1 ∪ X2 be the union of two closed
subsets. Then I(X) = I(X1 ) ∩ I(X2 ) (see §4). Assume that X1 6= X, then I(X1 )
strictly contains I(X) (otherwise V (I(X1 )) = V (I(X)). So there exists F ∈ I(X1 )
such that F 6∈ I(X). But for every G ∈ I(X2 ), F G ∈ I(X1 )∩I(X2 ) = I(X) prime:
since F 6∈ I(X), then G ∈ I(X). So I(X2 ) ⊂ I(X) hence I(X2 ) = I(X).
If X ⊂ Pn , the proof is similar, taking into account the following:
6.6. Lemma Let P ⊂ K[x0 , x1 , . . . , xn ] be a homogeneous ideal. Then P is prime
if and only if, for every pair of homogeneous polynomials F, G such that F G ∈ P,
either F ∈ P or G ∈ P.
Proof of the Lemma. Let H, K be any polynomials such that HK ∈ P. Let
H = H0 + H1 + . . . + Hd , K = K0 + K1 + . . . + Ke (with Hd 6= 0 6= Ke ) be their
expressions as sums of homogeneous polynomials. Then HK = H0 K0 + (H0 K1 +
H1 K0 ) + . . . + Hd Ke : the last product is the homogeneous component of degree
d + e of HK. P being homogeneous, Hd Ke ∈ P; by assumption either Hd ∈ P or
Ke ∈ P. In the former case, HK − Hd K = (H − Hd )K belongs to P while in the
second one H(K − Ke ) ∈ P. So in both cases we can proceed by induction. 
 Introduction to algebraic geometry 25

We list now some consequences of the previous Proposition.

1. Let K be an infinite field. Then An and Pn are irreducible, because
I(A ) = Ih (Pn ) = (0).
n

2. Let Y ⊂ Pn be closed. Y is irreducible if and only if its affine cone C(Y )

is irreducible.
3. Let Y = V (F ) ⊂ An , be a hypersurface over an algebraically closed field
K. If F is irreducible, then Y is irreducible.
4. Let K be algebraically closed. There is a bijection between prime ideals of
K[x1 , . . . , xn ] and irreducible algebraic subsets of An . In particular, the maximal
ideals correspond to the points. Similarly, there is a bijection between homo-
geneous non–irrelevant prime ideals of K[x0 , x1 , . . . , xn ] and irreducible algebraic
subsets of Pn .

6.7. Definition. A topological space X is called noetherian if it satisfies the

following equivalent conditions:
(i) the ascending chain condition for open subsets;
(ii) the descending chain condition for closed subsets;
(iii) any non–empty set of open subsets of X has maximal elements;
(iv) any non–empty set of closed subsets of X has minimal elements.
The proof of the equivalence is standard.
Example. An is noetherian: if the following is a descending chain of closed
subsets
Y1 ⊃ Y2 ⊃ . . . ⊃ Yk ⊃ . . . ,
then
I(Y1 ) ⊂ I(Y2 ) ⊂ . . . ⊂ I(Yk ) ⊂ . . .
is an ascending chain of ideals of K[x1 , . . . , xn ] hence stationary from a suitable m
on; therefore V (I(Ym )) = Ym = V (I(Ym )) = Ym+1 = . . ..

6.8. Proposition. Let X be a noetherian topological space and Y be a non–

empty closed subset of X. Then Y can be written as a finite union Y = Y1 ∪
. . . ∪ Yr of irreducible closed subsets. The maximal Yi ’s in the union are uniquely
determined by Y and called the “ irreducible components” of Y . They are the
maximal irreducible subsets of Y .
Proof. By contradiction. Let S be the set of the non–empty closed subsets of X
which are not a finite union of irreducible closed subsets: assume S 6= ∅. By
noetherianity S has minimal elements, fix one of them Z. Z is not irreducible, so
Z = Z1 ∪ Z2 , Zi 6= Z for i = 1, 2. So Z1 , Z2 6∈ S, hence Z1 , Z2 are both finite
unions of irreducible closed subsets, so such is Z: a contradiction.
Now assume that Y = Y1 ∪ . . . ∪ Yr , with Yi 6⊆ Yj if i 6= j and Yi irreducible
closed for all i. If there is another similar expression Y = Y10 ∪ . . . ∪ Ys0 , Yi0 6⊆ Yj0
26 Mezzetti

Sr
for i 6= j, then Y10 ⊂ Y1 ∪ . . . Yr , so Y10 = i=1 (Y10 ∪ Yi ), hence Y10 ⊂ Yi for some
i, and we can assume i = 1. Similarly, Y1 ⊂ Yj0 , for some j, so Y10 ⊂ Y1 ⊂ Yj0 ,
so j = 1 and Y1 = Y10 . Now let Z = Y − Y1 = Y2 ∪ . . . ∪ Yr = Y20 ∪ . . . ∪ Ys0 and
proceed by induction.


6.9. Corollary. Any algebraic subset of An (resp. of Pn ) is in a unique way the

finite union of its irreducible components. 
Note that the irreducible components of X are its maximal algebraic subsets.
They correspond to the minimal prime ideals over I(X). Since I(X) is radical,
these minimal prime ideals coincide with the primary ideals appearing in the pri-
mary decomposition of I(X).

6.10. Definition. An irreducible closed subset of An is called an affine variety.

Similarly, an irreducible closed subset of Pn is a projective variety. A locally closed
subset in Pn is the intersection of an open and a closed subset. An irreducible
locally closed subset of Pn is a quasi–projective variety.

6.11. Proposition. Let X ⊂ An and Y ⊂ Am be affine varieties. Then X × Y

is irreducible, i.e. a subvariety of An+m .
Proof. Let X × Y = W1 ∪ W2 , with W1 , W2 closed. For all P ∈ X the map
{P } × Y → Y which takes (P, Q) to Q is a homeomorphism, so {P } × Y is
irreducible. {P } × Y = (W1 ∩ ({P } × Y )) ∪ (W2 ∩ ({P } × Y )), so ∃i ∈ {1, 2} such
that {P } × Y ⊂ Wi . Let Xi = {P ∈ X | {P } × Y ⊂ Wi }, i = 1, 2. Note that
X = X1 ∪ X2 .
Claim. Xi is closed in X.
Let X i (Q) = {P ∈ X | (P, Q) ∈ Wi }, Q ∈ Y . We have: (X × {Q}) ∩ Wi =
X i (Q) × {Q} ' X i (Q); X × {Q} and Wi are closed in X × Y , so X i (Q) × {Q}
i
T in Xi × Y and also in X × {Q}, so X (Q) is closed in X. Note that
is closed
Xi = Q∈Y X (Q), hence Xi is closed, which proves the Claim.
Since X is irreducible, X = X1 ∪ X2 implies that either X = X1 or X = X2 ,
so either X × Y = W1 or X × Y = W2 . 

Exercises to §6.
1. Let X 6= ∅ be a topological space. Prove that X is irreducible if and only
if all non–empty open subsets of X are connected.
2*. Prove that the cuspidal cubic Y ⊂ A2C of equation x3 −y 2 = 0 is irreducible.
(Hint: express Y as image of A1 in a continuous map...)
 Introduction to algebraic geometry 27

3. Give an example of two irreducible subvarieties of P3 whose intersection is

reducible.
4. Find the irreducible components of the following algebraic sets over the
complex field:
a) V (y 4 − x2 , y 4 − x2 y 2 + xy 2 − x3 ) ⊂ A2 ;
b) V (y 2 − xz, z 2 − y 3 ) ⊂ A3 .
5*. Let Z be a topological space and {Uα }α∈I be an open covering of Z such
that Uα ∩ Uβ 6= ∅ for α 6= β and that all Uα ’s are irreducible. Prove that Z is
irreducible.

7. Dimension.

Let X be a topological space.

7.1. Definition. The topological dimension of X is the supremum of the lengths
of the chains of distinct irreducible closed subsets of X, where by definiton the
following chain has length n:

X0 ⊂ X1 ⊂ X2 ⊂ . . . ⊂ Xn .

Example.
1. dim A1 = 1: the maximal length chains have the form {P } ⊂ A1 .
2. dim An = n: a chain of length n is

{0} = V (x1 , . . . , xn ) ⊂ V (x1 , . . . , xn−1 ) ⊂ . . . ⊂ V (x1 ) ⊂ An ;

note that V (x1 , . . . , xi ) is irreducible ∀i, because the ideal hx1 , . . . , xi i is prime.
Indeed the quotient K[x1 , . . . , xn ]/hx1 , . . . , xi i is isomorphic to K[xi+1 , . . . , xn ].
Therefore dim An ≥ n. On the other hand, from every chain of irreducible closed
subsets of An , passing to the ideals, we get a chain of the same length of prime
ideals in K[x1 , . . . , xn ], therefore dim An ≤ dim K[x1 , . . . , xn ] = n.
3. Let X be irreducible. Then dim X = 0 if and only if X is the closure of
every point of it.

We prove now some useful relations between the dimension of X and the
dimensions of its subspaces.

7.2. Proposition.
1. If Y ⊂ X, then dim Y ≤ dim X. In particular, if dim X is finite, then also
dim Y is (in this case, the number dim X − dim Y is called the codimension of Y
in X). S
2. If X = i∈I Ui is an open covering, then dim X = sup{dim Ui }.
3. If X is noetherian and X1 , . . . , Xs are its irreducible components, then
dim X = supi dim Xi .
28 Mezzetti

4. If Y ⊂ X is closed, X is irreducible, dim X is finite and dim X = dim Y ,

then Y = X.
Proof.
1. Let Y0 ⊂ Y1 ⊂ . . . ⊂ Yn be a chain of irreducible closed subsets of Y . Then
their closures are irreducible and form the following chain: Y0 ⊆ Y1 ⊆ . . . ⊆ Yn .
Note that for all i Yi ∩ Y = Yi , because Yi is closed into Y , so if Yi = Yi+1 , then
Yi = Yi+1 . Therefore the two chains have the same length and we can conclude
that dim Y ≤ dim X.
2. Let X0 ⊂ X1 ⊂ . . . ⊂ Xn be a chain of irreducible closed subsets of X. Let
P ∈ X0 be a point: there exists an index i ∈ I such that P ∈ Ui . So ∀k = 0, . . . , n
Xk ∩ Ui 6= ∅: it is an irreducible closed subset of Ui , irreducible because open in
Xk which is irreducible. Consider X0 ∩ Ui ⊂ X1 ∩ Ui ⊂ . . . ⊂ Xn ∩ Ui ; it is a chain
of length n, because Xk ∩ Ui = Xk : in fact Xk ∩ Ui is open in Xk hence dense.
Therefore, for all chain of irreducible closed subsets of X, there exists a chain of
the same length of irreducible closed subsets of some Ui . So dim X ≤ sup dim Ui .
By 1., equality holds.
3. Any chain of irreducible closed subsets of X is completely contained in an
irreducible component of X. The conclusion follows as in 2.
4. If Y0 ⊂ Y1 ⊂ . . . ⊂ Yn is a chain of maximal length in Y , then it is a
maximal chain in X, because dim X = dim Y . Hence X = Yn ⊂ Y . 

7.3. Corollary. dim Pn = dim An .

Proof. Because Pn = U0 ∪ . . . ∪ Un , and Ui is homeomorphic to An for all i. 

If X is noetherian and all its irreducible components have the same dimension
r, then X is said to have pure dimension r.
Note that the topological dimension is invariant by homeomorphism. By
definition, a curve is an algebraic set of pure dimension 1; a surface is an algebraic
set of pure dimension 2.
We want to study the dimensions of affine algebraic sets. The following defi-
nition results to be very important.
7.4. Definition. Let X ⊂ An be an algebraic set. The coordinate ring of X is
K[X] := K[x1 , . . . , xn ]/I(X).
It is a finitely generated K–algebra without non–zero nilpotents, because I(X)
is radical. There is the canonical epimorphism K[x1 , . . . , xn ] → K[X] such that
F → [F ]. The elements of K[X] can be interpreted as polynomial functions on
X: to a polynomial F , we can associate the function f : X → K such that
P (a1 , . . . , an ) → F (a1 , . . . , an ).
Two polynomials F , G define the same function on X if, and only if, F (P ) =
G(P ) for every point P ∈ X, i.e. if F − G ∈ I(X), which means exactly that F
and G have the same image in K[X].
 Introduction to algebraic geometry 29

K[X] is generated as K–algebra by [x1 ], . . . , [xn ]: these can be interpreted as

the functions on X called coordinate functions, and generally denoted t1 , . . . , tn .
In fact ti : X → K is the function which associates to P (a1 , . . . , an ) the constant
ai . Note that the function f can be interpreted as F (t1 , . . . , tn ): the polynomial
F evalued at the n– tuple of the coordinate functions.
In the projective space we can do an analogous construction. If Y ⊂ Pn is
closed, then the homogeneous coordinate ring of Y is

S(Y ) := K[x0 , x1 , . . . , xn ]/Ih (Y ).

It is also a finitely generated K–algebra, but its elements have no interpretation

as functions on Y . They are functions on the cone C(Y ).

7.5. Definition. Let R be a ring. The Krull dimension of R is the supremum

of the lengths of the chains of prime ideals of R

P0 ⊂ P1 ⊂ . . . ⊂ Pr .

Similarly, the heigth of a prime ideal P is the sup of the lengths of the chains of
prime ideals contained in P: it is denoted htP.

7.6. Proposition. Let K be an algebraically closed field. Let X be an affine

algebraic set contained in An . Then dim X = dim K[X].
Proof.
By the Nullstellensatz and by 6.5 the chains of irreducible closed subsets of
X correspond bijectively to the chains of prime ideals of K[x1 , . . . , xn ] containing
I(X), hence to the chains of prime ideals of the quotient ring K[X]. 
The dimension theory for commutative rings contains some important theo-
rems about dimension of K–algebras. The following two results are very useful.

7.7. Theorem. Let K be any field.

1. Let B be a finitely generated K–algebra and an integral domain. Then
dim B = tr.d.Q(B)/K, where Q(B) is the quotient field of B. In particular dim B
is finite.
2. Let B be as above and P ⊂ B be any prime ideal. Then dim B =
htP + dim B/P.
Proof. For 1. see Portelli’s notes. For a proof of 2., see for instance [4], Ch. II,
Proposition 3.4. It relies on the normalization lemma and the lying over theorem.


7.8. Corollary. Let K be an algebraically closed field.

1. dim An = dim Pn = n.
30 Mezzetti

2. If X is an affine variety, then dim X = tr.d.K(X)/K, where K(X) denotes

the quotient field of K[X].
2. If X ⊂ An is closed and irreducible, then dim X = n − htI(X). 

The following is an important characterization of the algebraic subsets of An

of codimension 1.

7.9. Proposition. Let X ⊂ An be closed. Then X is a hypersurface if and only

if X is of pure dimension n − 1.
Proof. We give here an elementary direct proof. It can be proved more quickly
using the Krull principal ideal theorem.
Let X ⊂ An be a hypersurface, with I(X) = (F ) = (F1r1 . . . Fsrs ), where
F1 , . . . , Fs are the irreducible factors of F with multiplicities r1 , . . . , rs respectively.
Then V (F1 ),. . ., V (Fs ) are the irreducible components of X, whose ideals are (F1 ),
. . ., (Fs ). So it is enough to prove that ht(Fi ) = 1, for i = 1, . . . , s.
If P ⊂ (Fi ) is a prime ideal, then either P = (0) or there exists G ∈ P, G 6= 0.
In the second case, let A be an irreducible factor of G belonging to P: A ∈ (Fi )
so A = HFi . Since A is irreducible, either H or Fi is invertible; Fi is irreducible,
so H is invertible, hence (A) = (Fi ) ⊂ P. Therefore either P = (0) or P = (Fi ),
and ht(Fi ) = 1.
Conversely, assume that X is irreducible of dimension n − 1. Since X 6= An ,
there exists F ∈ I(X), F 6= 0. By the irreducibility, some irreducible factor of
F , call it H, also vanishes along X. Therefore X ⊂ V (H), which is irreducible of
dimension n − 1, by the first part. So X = V (H) (by Proposition 7.2, 3). 
This proposition does not generalize to higher codimension. There exist codi-
mension 2 algebraic subsets of An whose ideal is not generated by two polynomials.
An example in A3 is the curve X parametrized by (t3 , t4 , t5 ). A system of gener-
ators of I(X) is hx3 − yz, y 2 − xz, z 2 − x2 yi. One can show that I(X) cannot be
generated by two polynomials. For a discussion of this and other similar examples,
see [4], Chapter V.

7.10. Proposition. Let X ⊂ An , Y ⊂ Am be irreducible closed subsets. Then

dim X × Y = dim X + dim Y .
Proof. Let r = dim X, s = dim Y ; let t1 , . . . , tn (resp. u1 , . . . , um ) be coordinate
functions on An (resp. Am ). We can assume that t1 , . . . , tr be a transcendence ba-
sis of Q(K[X]) and u1 , . . . , us be a transcendence basis of Q(K[Y ]). By definition,
K[X × Y ] is generated as K–algebra by t1 , . . . , tn , u1 , . . . , um : we want to show
that t1 , . . . , tr , u1 , . . . , us is a transcendence basis of Q(K[X × Y ]) over K. Assume
that F (x1 , . . . , xr , y1 , . . . , ys ) is a polynomial which vanishes on t1 , . . . , tr , u1 , . . . , us ,
i.e. F defines the zero function on X × Y . Then, ∀ P ∈ X, F (P ; y1 , . . . , ys )
is zero on Y , i.e. F (P ; u1 , . . . , us ) = 0. Since u1 , . . . , us are algebraically in-
dependent, every coefficient ai (P ) of F (P ; y1 , . . . , ys ) is zero, ∀ P ∈ X. Since
 Introduction to algebraic geometry 31

t1 , . . . , tr are algebraically independent, the polynomials ai (x1 , . . . , xr ) are zero, so

F (x1 , . . . , xr , y1 , . . . , ys ) = 0. So t1 , . . . , tr , u1 , . . . , us are algebraically independent.
Since this is certainly a maximal algebraically free set, it is a transcendence basis.


Exercises to §7.
1*. Prove that a proper closed subset of an irreducible curve is a finite set.
Deduce that any bijection between irreducible curves is a homeomorphism.
2*. Let X ⊂ A2 be the cuspidal cubic of equation: x3 − y 2 = 0, let K[X] be
its coordinate ring. Prove that all elements of K[X] can be written in a unique
way in the form f (x) + yg(x), where f, g are polynomial in the variable x. Deduce
that K[X] is not isomorphic to a polynomial ring.

8. Regular and rational functions.

Let X ⊂ Pn be a locally closed subset and P be a point of X. Let φ : X → K be

a function.

8.1. Definition. φ is regular at P if there exists a suitable neighbourhood of

P in which it can be expressed as a quotient of homogeneous polynomials of the
same degree; more precisely, if there exist an open neighbourhood U of P in X and
homogeneous polynomials F , G ∈ K[x0 , x1 , . . . , xn ] with deg F = deg G, such that
U ∩ VP (G) = ∅ and φ(Q) = F (Q)/G(Q), for all Q ∈ U . Note that the quotient
F (Q)/G(Q) is well defined.
φ is regular on X if φ is regular at every point P of X.
The set of regular functions on X is denoted O(X): it contains K (identified with
the set of constant functions), and can be given the structure of a K–algebra, by
the definitions:
(φ + ψ)(P ) = φ(P ) + ψ(P )

(φψ)(P ) = φ(P )ψ(P ),

for P ∈ X. (Check that φ + ψ and φψ are indeed regular on X.)

8.2. Proposition. Let φ : X → K be a regular function. Let K be identified

with A1 with Zariski topology. Then φ is continuous.
Proof. It is enough to prove that φ−1 (c) is closed in X, ∀ c ∈ K. For all P ∈ X,
choose an open neighbourhood UP and homogeneous polynomials FP , GP such
that φ|P = FP /GP . Then

φ−1 (c) ∩ UP = {Q ∈ UP |FP (Q) − cGP (Q) = 0} = UP ∩ VP (FP − cGP )

32 Mezzetti

is closed in UP . The proposition then follows from:

8.3. Lemma. Let T be a topological space, T = ∪i∈I Ui be an open covering of

T , Z ⊂ T be a subset. Then Z is closed if and only if Z ∩ Ui is closed in Ui for all
i.
Proof. Assume that Ui = X \ Ci and Z ∩ Ui = Zi ∩ Ui , with Ci and Zi closed in
X.
T
Claim: Z = i∈I (Zi ∪ Ci ), hence it is closed.
In fact: if P ∈ Z, then P ∈ Z ∩ Ui for a suitable i. Therefore P ∈ Zi ∩ Ui , so
P ∈ Zi ∪ Ci . If P ∈ / Zj ∩ Uj for some j, then P ∈ / Uj so P ∈ Cj and therefore
P ∈ Zj ∪ Cj .
T
Conversely, if P ∈ i∈I (Zi ∪ Ci ), then ∀ i, either P ∈ Zi or P ∈ Ci . Since ∃j
such that P ∈ Uj , hence P ∈/ Cj , so P ∈ Zj , so P ∈ Zj ∩ Uj = Z ∩ Uj .


8.4. Corollary.
1. Let φ ∈ O(X): then φ−1 (0) is closed. It is denoted V (φ) and called the
set of zeroes of φ.
2. Let X be a quasi–projective variety and φ, ψ ∈ O(X). Assume that there
exists U , open non –empty subset such that φ|U = ψ|U . Then φ = ψ.
Proof. φ − ψ ∈ O(X) so V (φ − ψ) is closed. By assumption V (φ − ψ) ⊃ U , which
is dense, because X is irreducible. So V (φ − ψ) = X.


If X ⊂ An is locally closed, we can use on X both homogeneous and non–

homogeneous coordinates. In the second case, a regular function is locally rep-
resented as a quotient F/G, with F and G ∈ K[x1 , . . . , xn ]. In particular all
polynomial functions are regular, so, if X is closed, K[X] ⊂ O(X).
T
If α ⊂ K[X] is an ideal, we can consider V (α) := φ∈α V (φ): it is closed
into X. Note that α is of the form α = α/I(X), where α is the image of α in
the canonical epimorphism, it is an ideal of K[x1 , . . . , xn ] containing I(X), hence
V (α) = V (α) ∩ X = V (α).
If K is algebraically closed, from the Nullstellensatz it follows that, if α is
proper, then V (α) 6= ∅. Moreover the following relative form of the Nullstellensatz
holds: if f ∈ K[X] and f vanishes at all points P ∈ X such that g1 (P ) = . . . =
gm (P ) = 0 (g1 , . . . , gm ∈ K[X]), then f r ∈ hg1 , . . . , gm i ⊂ K[X], for some r ≥ 1.

8.5. Theorem. Let K be an algebraically closed field. Let X ⊂ AnK be closed

in the Zariski topology. Then O(X) ' K[X]. It is an integral domain if and only
if X is irreducible.
Proof. Let f ∈ O(X).
 Introduction to algebraic geometry 33

(i) Assume first that X is irreducible. For all P ∈ X fix an open neigh-
bourhood UP of P and polynomials FP , GP such that VP (GP ) ∩ UP = ∅ and
f |UP = FP /GP . Let fP , gP be the functions of K[X] defined by FP and GP . Then
gP f = fP holds on UP , so it holds on X (by Corollary 8.3). Let α ⊂ K[X] be
the ideal α = hgP iP ∈X ; α has no zeroes on X, becauseP gP (P ) 6= 0, so α = K[X].
Therefore there exists hP ∈ K[X] such that 1 = P ∈X P hP gP (sum P with finite
support).
P Hence in O(X) we have the relation: f = f h g
P P = hP (gP f ) =
hP fP ∈ K[X].
(ii) Let X be reducible: for all P ∈ X, there exists R ∈ K[x1 , . . . , xn ] such that
R(P ) 6= 0 and R ∈ V (X \ UP ), so r ∈ O(X) is zero outside UP . So rgP f = fP r
on X and we conclude as above by replacing gP with gP r and fP with fP r.


The characterization of regular functions on projective varieties is completely dif-

ferent: we will see in §12 that, if X is a projective variety, then O(X) ' K, i.e.
the unique regular functions are constant.
This gives the motivation for introducing the following weaker concept.

8.6. Definition. Let X be a quasi–projective variety. A rational function on X

is a germ of regular functions on some open non–empty subset of X.
Precisely, let K be the set {(U, f )|U 6= ∅, open subset of X, f ∈ O(U )}. The
following relation on K is an equivalence relation:

(U, f ) ∼ (U 0 , f 0 ) if and only if f |U ∩U 0 = f 0 |U ∩U 0 .

Reflexive and symmetric properties are quite obvious. Transitive property: let
(U, f ) ∼ (U 0 , f 0 ) and (U 0 , f 0 ) ∼ (U 00 , f 00 ). Then f |U ∩U 0 = f 0 |U ∩U 0 and f 0 |U 0 ∩U 00 =
f 00 |U 0 ∩U 00 , hence f |U ∩U 0 ∩U 00 = f 00 |U ∩U 0 ∩U 00 . U ∩ U 0 ∩ U 00 is a non–empty open
subset of U ∩ U 00 (which is irreducible and quasi–projective), so by Corollary 8.4
f |U 0 ∩U 00 = f 00 |U 0 ∩U 00 .
Let K(X) := K/ ∼: its elements are by definition rational functions on X.
K(X) can be given the structure of a field in the following natural way.
Let hU, f i denote the class of (U, f ) in K(X). We define:

hU, f i + hU 0 , f 0 i = hU ∩ U 0 , f + f 0 i,

hU, f ihU 0 , f 0 i = hU ∩ U 0 , f f 0 i
(check that the definitions are well posed!).
There is a natural inclusion: K → K(X) such that c → hX, ci. Moreover, if
hU, f i 6= 0, then there exists hU, f i−1 = hU \ V (f ), f −1 i: the axioms of a field are
all satisfied.
There is also an injective map: O(X) → K(X) such that φ → hX, φi.
34 Mezzetti

8.7. Proposition. If X ⊂ An is affine, then K(X) ' Q(O(X)) = K(t1 , . . . , tn ),

where t1 , . . . , tn are the coordinate functions on X.
Proof. The isomorphism is as follows:
(i) ψ : K(X) → Q(O(X))
If hU, φi ∈ K(X), then there exists V ⊂ U , open and non–empty, such that φ |V =
F/G, where F, G ∈ K[x1 , . . . , xn ] and V (G) ∩ V = ∅. We set ψ(hU, φi) = f /g.
(ii) ψ 0 : Q(O(X)) → K(X)
If f /g ∈ Q(O(X)), we set ψ 0 (f /g) = hX \ V (g), f /gi.
It is easy to check that ψ and ψ 0 are well defined and inverse each other. 

8.8. Corollary. If X is an affine variety, then dim X is equal to the transcendence

degree over K of its field of rational functions..

8.9. Proposition. If X is quasi–projective and U 6= ∅ is an open subset, then

K(X) ' K(U ).
Proof. We have the maps: K(U ) → K(X) such that hV, φi → hV, φi, and K(X) →
K(U ) such that hA, ψi → hA ∩ U, ψ |A∩U i: they are K–homomorphisms inverse
each other. 

8.10. Corollary. If X is a projective variety contained in Pn , if i is an index

such that X ∩ Ui 6= ∅, then dim X = dim X ∩ Ui = tr.d.K(X)/K.
Proof. By Proposition 7.2 dim X = sup dim(X ∩ Ui ). By 8.8 and 8.9, if X ∩ Ui is
non–empty, dim(X ∩ Ui ) = tr.d.K(X ∩ Ui )/K = tr.d.K(X)/K is independent of
i. 

If hU, φi ∈ K(X), we can consider all possible representatives

S of it, i.e. all
pairs hUi , φi i such that hU, φi = hUi , φi i. Then U = i Ui is the maximum open
subset of X on which φ can be seen as a function: it is called the domain of
definition (or of regularity) of hU, φi, or simply of φ. It is sometimes denoted
domφ. If P ∈ U , we say that φ is regular at P.
We can consider the set of rational functions on X which are regular at P : it
is denoted OP,X . It is a subring of K(X) containing O(X), called the local ring
of X at P . In fact, OP,X is a local ring, whose maximal ideal, denoted MP,X , is
formed by rational functions φ such that φ(P ) is defined and φ(P ) = 0. To see
this, observe that an element of OP,X can be represented as hU, F/Gi: its inverse
in K(X) is hU \ VP (G), G/F i, which belongs to OP,X if and only if F (P ) 6= 0.
We’ll see in 8.12 that OP,X is the localization K[X]IX (P ) .
As in Proposition 8.9 for the fields of rational functions, also for the local rings
of points it can easily be proved that, if U 6= ∅ is an open subset of X containing
P , then OP,X ' OP,U . So the ring OP,X only depends on the local behaviour of
 Introduction to algebraic geometry 35

X in the neighbourhood of P .
The residue field of OP,X is the quotient OP,X /MP,X : it is a field which results
to be naturally isomorphic to the base field K. In fact consider the evaluation
map OP,X → K such that φ goes to φ(P ): it is surjective with kernel MP,X , so
OP,X /MP,X ' K.

8.11. Examples.
1. Let Y ⊂ A2 be the curve V (x31 − x22 ). Then F = x2 , G = x1 define the
function φ = x2 /x1 which is regular at the points P (a1 , a2 ) such that a1 6= 0.
Another representation of the same function is: φ = x21 /x2 , which shows that
φ is regular at P if a2 6= 0. If φ admits another representation F 0 /G0 , then
G0 x2 − F 0 x1 vanishes on an open subset of X, which is irreducible (see Exercise
6.2), hence G0 x2 − F 0 x1 vanishes on X, and therefore G0 x2 − F 0 x1 ∈ hx31 − x22 i.
This shows that there are essentially only the above two representations of φ. So
φ ∈ K(X) and its domain of regularity is Y \ {0, 0}.

2. The stereographic projection.

Let X ⊂ P2 be the curve VP (x21 +x22 −x20 ). Let f := x1 /(x0 −x2 ) denote the germ of
the regular function defined by x1 /(x0 − x2 ) on X \ VP (x0 − x2 ) = X \ {[1, 0, 1]} =
X \ {P }. On X we have x21 = (x0 − x2 )(x0 + x2 ) so f is represented also as
(x0 + x2 )/x1 on X \ VP (x1 ) = X \ {P, Q}, where Q = [1, 0, −1]. If we identify K
with the affine line VP (x2 ) \ VP (x0 ) (the points of the x1 –axis lying in the affine
plane U0 ), then f can be interpreted as the stereographic projection of X centered
at P , which takes A[a0 , a1 , a2 ] to the intersection of the line AP with the line
VP (x2 ). To see this, observe that AP has equation a1 x0 + (a2 − a0 )x1 − a1 x2 = 0;
and AP ∩ VP (x2 ) is the point [a0 − a2 , a1 , 0].

8.12. The algebraic characterization of the local ring OP,X .

Let us recall the construction of the ring of fractions of a ring A with respect
to a multiplicative subset S.
Let A be a ring and S ⊂ A be a multiplicative subset. The following relation
in A × S is an equivalence relation:
(a, s) ' (b, t) if and only if ∃u ∈ S such that u(at − bs) = 0.
Then the quotient A × S/' is denoted S −1 A or AS and [(a, s)] is denoted as . AS
becomes a commutative ring with unit with operations as + bt = at+bs st and as bt = ab
st
(check that they are well–defined). With these operations, AS is called the ring of
fractions of A with respect to S, or the localization of A in S.
There is a natural homomorphism j : A → S −1 A such that j(a) = a1 , which
makes S −1 A an A–algebra. Note that j is the zero map if and only if 0 ∈ S. More
precisely if 0 ∈ S then S −1 A is the zero ring: this case will always be excluded in
what follows. Moreover j is injective if and only if S is formed by regular elements.
In this case j(A) will be identified with A.
36 Mezzetti

Examples.
1. Let A be an integral domain and set S = A \ {0}. Then AS = Q(A): the
quotient field of A.
2. If P ⊂ A is a prime ideal, then S = A \ P is a multiplicative set and AS is
denoted AP and called the localization of A at P.
3. If f ∈ A, then the multiplicative set generated by f is

S = {1, f, f 2 , . . . , f n , . . .} :

AS is denoted Af .
4. If S = {x ∈ A | x is regular}, then AS is called the total ring of fractions
of A: it is the maximum ring in which A can be canonically embedded.

It is easy to verify that the ring AS enjoys the following universal property:
(i) if s ∈ S, then j(s) is invertible;
(ii) if B is a ring with a given homomorphism f : A → B such that if s ∈ S,
then f (s) is invertible, then f factorizes through AS , i.e. there exists a unique
homomorphism f such that f ◦ j = f .

We will see now the relations between ideals of AS and ideals of A.

If α ⊂ A is an ideal, then αAS = { as | a ∈ α} is called the extension of α in
AS and denoted also αe . It is an ideal, precisely the ideal generated by the set
{ a1 | a ∈ α}.
If β ⊂ AS is an ideal, then j −1 (β) =: β c is called the contraction of β and is
clearly an ideal.
We have:

8.13. Proposition.
1. ∀α ⊂ A : αec ⊃ α;
2. ∀β ⊂ AS : β = β ce ;
3. αe is proper if and only if α ∩ S = ∅;
4. αec = {x ∈ A | ∃s ∈ S such that sx ∈ α}.
Proof.
1. and 2. are straightforward.
3. if 1 = as ∈ αe , then there exists u ∈ S such that u(s − a) = 0, i.e.
us = ua ∈ S ∩ α. Conversely, if s ∈ S ∩ α then 1 = ss ∈ αe .
4.
x
αec = {x ∈ A | j(x) = ∈ αe } =
1
x a
= {x ∈ A | ∃a ∈ α, t ∈ S such that = } =
1 t
= {x ∈ A | ∃a ∈ α, t, u ∈ S such that u(xt − a) = 0}.
 Introduction to algebraic geometry 37

Hence, if x ∈ αec , then: (ut)x = ua ∈ α. Conversely: if there exists s ∈ S such

that sx = a ∈ α, then x1 = as , i.e. j(x) ∈ αe . 

If α is an ideal of A such that α = αec , α is called saturated with S. For

example, if P is a prime ideal and S ∩ P = ∅, then P is saturated and P e is prime.
Conversely, if Q ⊂ AS is a prime ideal, then Qc is prime in A.
Therefore: there is a bijection between the set of prime ideals of AS and the
set of prime ideals of A not intersecting S. In particular, if S = A \ P, P prime,
the prime ideals of AP correspond bijectively to the prime ideals of A contained
in P, hence AP is a local ring with maximal ideal P e , denoted PAP , and residue
field AP /PAP . Moreover dim AP = htP.
In particular we get the characterization of OP,X . Let X ⊂ An be an affine
variety, let P be a point of X and I(P ) ⊂ K[x1 , . . . , xn ] be the ideal of P . Let
IX (P ) := I(P )/I(X) be the ideal of K[X] formed by regular functions on X
vanishing at P . Then we can construct the localization

f
O(X)IX (P ) = { |f, g ∈ O(X), g(P ) 6= 0} ⊂ K(X) :
g

it is canonically identified with OP,X . In particular: dim OP,X = ht IX (P ) =

dim O(X) = dim X.
There is a bijection between prime ideals of OP,X and prime ideals of O(X)
contained in IX (P ); they also correspond to prime ideals of K[x1 , . . . , xn ] contained
in I(P ) and containing I(X).

If X is affine, it is possible to define the local ring OP,X also if X is reducible,

simply as localization of K[X] at the maximal ideal IX (P ). The natural map j
from K[X] to OP,X is injective if and only if K[X] \ IX (P ) does not contain any
zero divisor. A non-zero function f is a zero divisor if there exists a non-zero g
such that f g = 0, i.e. X = V (f ) ∪ V (g) is an expression of X as union of proper
closed subsets. For j to be injective it is required that every zero divisor f belongs
to IX (P ), which means that all the irreducible components of X pass through P .

Exercises to §8.
1. Prove that the affine varieties and the open subsets of affine varieties are
quasi–projective.
2. Let X = {P, Q} be the union of two points in an affine space over K.
Prove that O(X) is isomorphic to K × K.

9. Regular and rational maps.

Let X, Y be quasi–projective varieties (or more generally locally closed sets). Let
φ : X → Y be a map.
38 Mezzetti

9.1. Definition. φ is a regular map or a morphism if

(i) φ is continuous;
(ii) φ preserves regular functions, i.e. for all U ⊂ Y (U open and non–empty) and
for all f ∈ O(U ), then f ◦ φ ∈ O(φ−1 (U )):

φ
X −→ Y
↑ ↑
φ| f
φ−1 (U ) −→ U → K

Note that:
a) for all X the identity map 1X : X → X is regular;
φ ψ
b) for all X, Y , Z and regular maps X → Y , Y → Z, the composite map ψ ◦ φ is
regular.
An isomorphism of varieties is a regular map which possesses regular inverse,
i.e. a regular φ : X → Y such that there exists a regular ψ : Y → X verifying
the conditions ψ ◦ φ = 1X and φ ◦ ψ = 1Y . In this case X and Y are said to be
isomorphic, and we write: X ' Y .
If φ : X → Y is regular, there is a natural K–homomorphism φ∗ : O(Y ) →
O(X), called the comorphism associated to φ, defined by: f → φ∗ (f ) := f ◦ φ.
The construction of the comorphism is functorial, which means that:
∗
a) 1X = 1O(X) ;
b) (ψ ◦ φ)∗ = φ∗ ◦ ψ ∗ .
This implies that, if X ' Y , then O(X) ' O(Y ). In fact, if φ : X → Y is an
isomorphism and ψ is its inverse, then φ ◦ ψ = 1Y , so (φ ◦ ψ)∗ = ψ ∗ ◦ φ∗ = (1Y )∗ =
1O(Y ) and similarly ψ ◦ φ = 1X implies φ∗ ◦ ψ ∗ = 1O(X) .

9.2. Examples.
1) The homeomorphism φi : Ui → An of Proposition 3.2 is an isomorphism.

2) There exist homeomorphisms which are not isomorphisms. Let Y = V (x3 −

y ) ⊂ A2 . We have seen (see Exercise 7.2) that K[X] 6' K[A1 ], hence Y is not
2

isomorphic to the affine line. Nevertheless, the following map is regular, bijective
and also a homeomorphism (see Exercise 7.1):
φ : A1 → Y such that t → (t2 , t3 ); 
y
if x 6= 0
φ−1 : Y → A1 is defined by (x, y) → x
0 if (x, y) = (0, 0).

9.3. Proposition. Let φ : X → Y ⊂ An be a map. Then φ is regular if and only

if φi := ti ◦ φ is a regular function on X, for all i = 1, . . . , n (where t1 , . . . , tn are
the coordinate functions on Y ).
Proof. If φ is regular, then φi = φ∗ (ti ) is regular by definition.
 Introduction to algebraic geometry 39

Conversely: if F ∈ K[x1 , . . . , xn ], then

φ−1 (V (F ) ∩ Y ) = {P ∈ X|F (φ(P )) = F (φ1 , . . . , φn )(P ) = 0} = V (F (φ1 , . . . , φn ))

(note that F (φ1 , . . . , φn ) ∈ O(X): it is the composition of F with the regular func-
tions φ1 , . . . , φn ). In particular φ−1 (V (F )∩Y ) is closed, so we can conclude that φ
is continuous. If U ⊂ Y and f ∈ O(U ), for all points P of U choose open neighbour-
hoods UP such that f = FP /GP on UP . So f ◦φ = FP (φ1 , . . . , φn )/GP (φ1 , . . . , φn )
on φ−1 (UP ), hence it is regular on each φ−1 (UP ) and by consequence on φ−1 (U ).

If φ : X → Y is a regular map and Y ⊂ An , by Proposition 9.2. we can rep-
resent φ in the form φ = (φ1 , . . . , φn ), where φ1 , . . . , φn ∈ O(X) and φi = φ∗ (ti ).
If Y is closed in An , let us recall that t1 , . . . , tn generate O(Y ), hence φ1 , . . . , φn
generate φ∗ (O(Y )). This observation is the key for the following important result.

9.4. Theorem. Let X be a locally closed set and Y be an affine algebraic set. Let
Hom(X, Y ) denote the set of regular maps from X to Y and Hom(O(Y ), O(X))
denote the set of K– homomorphisms from O(Y ) to O(X).
Then the map Hom(X, Y ) → Hom(O(Y ), O(X)), such that φ : X → Y goes
to φ∗ : O(Y ) → O(X), is bijective.
Proof. Let Y ⊂ An and let t1 , . . . , tn be the coordinate functions on Y , so O(Y ) =
K[t1 , . . . , tn ]. Let u : O(Y ) → O(Y ) be a K–homomorphism: we want to define a
morphism u] : X → Y whose associated comorphism is u. By the remark above,
if u] exists, its components have to be u(t1 ), . . . , u(tn ). So we define

u] : X → An
P → (u(t1 )(P )), . . . , u(tn )(P )).

This is a morphism by Proposition 9.3 and we claim that u] (X) ⊂ Y . Let F ∈ I(Y )
and P ∈ X: then

(F (u] (P )) = F (u(t1 )(P ), . . . , u(tn )(P )) =

= F (u(t1 ), . . . , u(tn ))(P ) =
= u(F ((t1 , . . . , tn ))(P ) because u is K-homomorphism =
= u(0)(P ) =
= 0(P ) = 0.

So u] is a regular map from X to Y .

We consider now (u] )∗ : O(Y ) → O(X): it takes a function f to f ◦ u] =
f (u(t1 ), . . . , u(tn )) = u(f ), so (u] )∗ = u. Conversely, if φ : X → Y is regular, then
(φ∗ )] takes P to (φ∗ (t1 )(P ), . . . , φ∗ (tn )(P )) = (φ1 (P ), . . . , φn (P )), so (φ∗ )] = φ.

40 Mezzetti

Note that, by definition, 1]O(X) = 1X , for all affine X; moreover (v◦u)] = u] ◦v ]

for all u : O(Z) → O(Y ), v : O(Y ) → O(X), K–homomorphisms of affine algebraic
sets: this means that also this construction is functorial.

9.5. Corollary. Let X, Y be affine algebraic sets. Then X ' Y if and only if
O(X) ' O(Y ). 

If X and Y are quasi–projective varieties and φ : X → Y is regular, it is not

always possible to define a comorphism K(Y ) → K(X). If f is a rational function
on Y with domf = U , it can happen that φ(X) ∩ domf = ∅, in which case f ◦ φ
does not exist. Nevertheless, if we assume that φ is dominant, i.e. φ(X) = Y ,
then certainly φ(X) ∩ U 6= ∅, hence hφ−1 (U ), f ◦ φi ∈ K(X). We obtain a K–
homomorphism, which is necessarily injective, K(Y ) → K(X), also denoted φ∗ .
Note that in this case, we have: dim X ≥ dim Y . As above, it is possible to check
that, if X ' Y , then K(X) ' K(Y ), hence dim X = dim Y . Moreover, if P ∈ X
and Q = φ(P ), then φ∗ induces a map OQ,Y → OP,X , such that φ∗ MQ,Y ⊂ MP,X .
Also in this case, if φ is an isomorphism, then OQ,Y ' OP,X .

Let now X ⊂ Pn be a quasi–projective variety and φ : X → Pm be a map.

9.6. Proposition. φ is a morphism if and only if, for all P ∈ X, there exist
an open neighbourhood UP of P and n + 1 homogeneous polynomials F0 , . . . , Fm
of the same degree, in K[x0 , x1 , . . . , xn ], such that, if Q ∈ UP , then φ(Q) =
[F0 (Q), . . . , Fm (Q)]. In particular, for all Q ∈ UP , there exists an index i such
that Fi (Q) 6= 0.

Proof. “ ⇒” Let P ∈ X, Q = φ(P ) and assume that Q ∈ U0 . Then U := φ−1 (U0 )

is an open neighbourhood of P and we can consider the restriction φ|U : U → U0 ,
which is regular. Possibly after restricting U , using non–homogeneous coordi-
nates on Y0 , we can assume that φ|U = (F1 /G1 , . . . , Fm /Gm ), where (F1 , G1 ),
. . ., (Fm , Gm ) are pairs of homogeneous polynomials of the same degree such
that VP (Gi ) ∩ U = ∅ for all index i. We can reduce the fractions Fi /Gi to a
common denominator F0 , so that deg F0 = deg F1 = . . . = deg Fm and φ|U =
(F1 /F0 , . . . , Fm /F0 ) = [F0 , F1 , . . . , Fm ], with F0 (Q) 6= 0 for Q ∈ U .
“ ⇐” Possibly after restricting UP , we can assume Fi (Q) 6= 0 for all Q ∈ UP
and suitable i. Let i = 0: then φ|UP : UP → U0 operates as follows: φ|UP (Q) =
(F1 (Q)/F0 (Q), . . . , Fm (Q)/F0 (Q)), so it is a morphism by Proposition 9.3. From
this remark, one deduces that also φ is a morphism. 

9.7. Examples.
1. Let X ⊂ P2 , X = VP (x21 + x22 − x20 ), the projective closure of the unitary
 Introduction to algebraic geometry 41

circle. We define φ : X → P1 by

[x0 − x2 , x1 ] if (x0 − x2 , x1 ) 6= (0, 0);
[x0 , x1 , x2 ] →
[x1 , x0 + x2 ] if (x1 , x0 + x2 ) 6= (0, 0).

φ is well–defined because on X x21 = (x0 − x2 )(x0 + x2 ). Moreover

(x1 , x0 − x2 ) 6= (0, 0) ⇔ [x0 , x1 , x2 ] ∈ X \ {[1, 0, 1]},

(x0 + x2 , x1 ) 6= (0, 0) ⇔ [x0 , x1 , x2 ] ∈ X \ {[1, 0, −1]}.

The map φ is the natural extension of the rational function f : X \{[1, 0, 1]} →
K such that [x0 , x1 , x2 ] → x1 /(x0 − x2 ) (Example 8.9, 2). Now the point P [1, 0, 1],
the centre of the stereographic projection, goes to the point at infinity of the line
VP (x2 ).
By geometric reasons φ is invertible and φ−1 : P1 → X takes [λ, µ] to [λ2 +
µ , 2λµ, λ2 − µ2 ] (note the connection with the Pitagorean triples!).
2

Indeed: the line through P and [λ, µ, 0] has equation: µx0 − λx1 − µx2 = 0.
Its intersections with X are represented by the system:

µx0 − λx1 − µx2 = 0
x21 + x22 − x20 = 0

Assuming µ 6= 0 this system is equivalent to the following:


µx0 − λx1 − µx2 = 0
µ2 x20 = µ2 (x21 + x22 ) = (λx1 + µx2 )2 .

(µ2 − λ2 )x1 − 2λµx2 = 0
Therefore, either x1 = 0 and x0 = x2 , or , which gives
µx0 = λx1 + µx2
the required expression.

2. Affine transformations.
Let A = (aij ) be a n × n–matrix with entries in K, let B = (b1 , . . . , bn ) ∈ An
be a point. The map τA : An → An defined by (x1 , . . . , xn ) → (y1 , . . . , yn ), such
that X
{yi = aij xj + bi , i = 1, . . . , n,
j

is a regular map called an affine transformation of An . In matrix notation τA

is Y = AX + B. If A is of rank n, then τA is said non–degenerate and is an
isomorphism: the inverse map τA−1 is represented by X = A−1 Y − A−1 B. More in
general, an affine transformation from An to Am is a map represented in matrix
form by Y = AX + B, where A is a n × m matrix and B ∈ Am . It is injective if
and only if rkA = n and surjective if and only if rkA = m.
42 Mezzetti

The isomorphisms of an algebraic set X in itself are called automorphisms

of X: they form a group for the usual composition of maps, denoted Aut X. If
X = An , the non–degenerate affine transformations form a subgroup of Aut An .
If n = 1 and the characteristic of K is 0, then Aut A1 coincides with this sub-
group. In fact, let φ : A1 → A1 be an automorphism: it is represented by a polyno-
mial F (x) such that there exists G(x) satisfying the condition G(F (t)) = t for all
t ∈ A1 , i.e. G(F (x)) = x in the polynomial ring K[x]. Then G0 (F (x))F 0 (x) = 1,
which implies F 0 (t) 6= 0 for all t ∈ K, so F 0 (x) is a non–zero constant. Hence, F
is linear and G is linear too.
If n ≥ 2, then Aut An is not completely described. There exist non–linear
automorphisms of degree d, for all d. For example, for n = 2: let φ : A2 → A2
be given by (x, y) → (x, y + P (x), where P is any polynomial of K[x]. Then
φ−1 : (x0 , y 0 ) → (x0 , y 0 − P (x0 )). A very important open problem is the Jacobian
conjecture, stating that a regular map φ : An → An is an automorphism if and
only if the Jacobian determinant | J(φ) | is a non-zero constant.

3. Projective transformations.
Let A be a (n + 1) × (n + 1)–matrix with entries in K. Let P [x0 , . . . , xn ] ∈ Pn :
then [a00 x0 + . . . + a0n xn , . . . , an0 x0 + . . . + ann xn ] is a point of Pn if and only if it
is different from [0, . . . , 0]. So A defines a regular map τ : Pn → Pn if and only if
rkA = n+1. If rkA = r < n+1, then A defines a regular map whose domain is the
quasi–projective variety Pn \ P(kerA). If rkA = n + 1, then τ is an isomorphism,
called a projective transformation. Note that the matrices λA, λ ∈ K ∗ , all define
the same projective transformation. So P GL(n + 1, K) := GL(n + 1, K)/K ∗ acts
on Pn as the group of projective transformations.
If X, Y ⊂ Pn , they are called projectively equivalent if there exists a projective
transformation τ : Pn → Pn such that τ (X) = Y .

9.8. Theorem. Fundamental theorem on projective transformations.

Let two (n+2)–tuples of points of Pn in general position be fixed: P0 , . . . , Pn+1
and Q0 , . . . , Qn+1 . Then there exists one isomorphic projective transformation τ
of Pn in itself, such that τ (Pi ) = Qi for all index i.
Proof. Put Pi = [vi ], Qi = [wi ], i = 0, . . . , n + 1. So {v0 , . . . , vn } and {w0 , . . . , wn }
are two bases of K n+1 , hence there exist scalars λ0 , . . . , λn , µ0 , . . . , µn such that
vn+1 = λ0 v0 + . . . + λn vn , wn+1 = µ0 w0 + . . . + µn wn ,
where the coefficients are all different from 0, because of the general position
assumption. We replace vi with λi vi and wi with µi wi and get two new bases, so
there exists a unique automorphism of K n+1 transforming the first basis in the
second one and, by consequence, also vn+1 in wn+1 . This automorphism induces
the required projective transformation on Pn . 
An immediate consequence of the above theorem is that projective subspaces
of the same dimension are projectively equivalent. Also two subsets of Pn formed
 Introduction to algebraic geometry 43

both by k points in general position are projectively equivalent if k ≤ n + 2. If

k > n + 2, this is no longer true, already in the case of four points on a projective
line. The problem of describing the classes of projective equivalence of k–tuples
of points of Pn , for k > n + 2, is one the first problems of the classical invariant
theory. The solution in the case k = 4, n = 1 is given by the notion of cross–ratio.

4. Let X ⊂ An be an affine variety, then XF = X \ V (F ) is isomorphic to

a closed subset of An+1 , i.e. to Y = V (xn+1 F − 1, G1 , . . . , Gr ), where I(X) =
hG1 , . . . , Gr i. Indeed, the following regular maps are inverse each other:
φ : XF → Y such that (x1 , . . . , xn ) → (x1 , . . . , xn , 1/F (x1 , . . . , xn )),
ψ : Y → XF such that (x1 , . . . , xn , xn+1 ) → (x1 , . . . , xn ).
Hence, XF is a quasi–projective variety contained in An , not closed in An , but
isomorphic to a closed subset of another affine space.
From now on, the term affine variety will denote a quasi–projective variety
isomorphic to some affine closed set.
If X is an affine variety and precisely X ' Y , with Y ⊂ An closed, then
O(X) ' O(Y ) = K[t1 , . . . , tn ] is a finitely generated K–algebra. In particular,
if K is algebraically closed and α is an ideal strictly contained in O(X), then
V (α) ⊂ X is non–empty, by the relative form of the Nullstellensatz. From this
observation, we can deduce that the quasi–projective variety of next example is
not affine.

5. A2 \ {(0, 0)} is not affine.

Set X = A2 \ {(0, 0)}: first of all we will prove that O(X) ' K[x, y] = O(A2 ),
i.e. any regular function on X can be extended to a regular function on the whole
plane.
Indeed: let f ∈ O(X): if P 6= Q are points of X, then there exist polynomials
F, G, F 0 , G0 such that f = F/G on a neighbourhood UP of P and f = F 0 /G0 on
a neighbourhood UQ of Q. So F 0 G = F G0 on UP ∩ UQ 6= ∅, which is open also
in A2 , hence dense. Therefore F 0 G = F G0 in K[x, y]. We can clearly assume that
F and G are coprime and similarly for F 0 and G0 . So by the unique factorization
property, it follows that F 0 = F and G0 = G. In particular f admits a unique
representation as F/G on X and G(P ) 6= 0 for all P ∈ X. Hence G has no zeroes
on A2 , so G = c ∈ K ∗ and f ∈ O(X).
Now, the ideal hx, yi has no zeroes in X and is proper: this proves that X is
not affine.

On the other hand, the following property holds: 9.9. Proposition. Let
X ⊂ Pn be quasi–projective. Then X admits an open covering by affine varieties.
Proof. Let X = X0 ∪ . . . ∪ Xn be the open covering of X where Xi = Ui ∩ X
= {P ∈ X|P [a0 , . . . , an ], ai 6= 0}. So, fixed P , there exists an index i such that
P ∈ Xi . We can assume that P ∈ X0 : X0 is open in some affine variety Y of An
(identified with U0 ); set X0 = Y \ Y 0 , where Y , Y 0 are both closed. Since P 6∈ Y 0 ,
44 Mezzetti

there exists F such that F (P ) 6= 0 and V (F ) ⊃ Y 0 . So P ∈ Y \ V (F ) ⊂ Y \ Y 0

and Y \ V (F ) is an affine open neighbourhood of P in Y \ Y 0 = X0 ⊂ X.

6. The Veronese maps.
Let n, d be positive integers; put N (n, d) = n+d − 1. Note that n+d



d d is
equal to the number of (monic) monomials of degree d in the variables x0 , . . . , xn ,
that is equal to the number of n + 1–tuples (i0 , . . . , in ) such that i0 + . . . + in = d,
ij ≥ 0. Then in PN (n,d) we can use coordinates {vi0 ...in }, where i0
, . . . , in ≥ 0 and
i0 + . . . + in = d. For example: if n = 2, d = 2, then N (2, 2) = 42 − 1 = 5. In P5
we can use coordinates v200 , v110 , v101 , v020 , v011 , v002 .
For all n, d we define the map vn,d : Pn → PN (n,d) such that [x0 , . . . , xn ] →
[vd00...0 , vd−1,10...0 , . . . , v0...00d ] where vi0 ...in = xi00 xi11 . . . xinn : vn,d is clearly a mor-
phism, its image is denoted Vn,d and called the Veronese variety of type (n, d). It
is in fact the projective variety of equations:

(∗){vi0 ...in vj0 ...jn − vh0 ...hn vk0 ...kn , ∀i0 + j0 = h0 + k0 , i1 + j1 = h1 + k1 , . . .

We prove this statement in the particular case n = d = 2; the general case is

similar.
First of all, it is clear that the points of vn,d (Pn ) satisfy the system (∗).
Conversely, assume that P [v200 , v110 , . . .] ∈ P5 satisfies the equations (∗), which
become:
2
v200 v020 = v110


2
v v = v101


 200 002


2
v002 v020 = v011

 v200 v011 = v110 v101
 v020 v101 = v110 v011



v110 v002 = v011 v101
Then, at least one of the coordinates v200 , v020 , v002 is different from 0.
Therefore, if v200 6= 0, then P = v2,2 ([v200 , v110 , v101 ]); if v020 6= 0, then
P = v2,2 ([v110 , v020 , v011 ]); if v002 6= 0, then P = v2,2 ([v101 , v011 , v002 ]). Note that,
if two of these three coordinates are different from 0, then the points of P2 found
in this way have proportional coordinates, so they coincide.
We have also proved in this way that v2,2 is an isomorphism between P2 and
V2,2 , called the Veronese surface of P5 . The same happens in the general case.
If n = 1, v1,d : P1 → Pd takes [x0 , x1 ] to [xd0 , x0d−1 x1 , . . . , xd1 ]: the image is
called the rational normal curve of degree d, it is isomorphic to P1 . If d = 3, we
find the skew cubic.
Let now X ⊂ Pn be a hypersurface of degree d: X = VP (F ), with
X
F = ai0 ...in xi00 . . . xinn .
i0 +...+in =d
 Introduction to algebraic geometry 45

Then vn,d (X) ' X: it is the set of points

X
{vi0 ...in ∈ PN (n,d) | ai0 ...in vi0 ...in = 0 and [vi0 ...in ] ∈ Vn,d }.
i0 +...+in =d

It coincides with Vn,d ∩ H, where H is a hyperplane of PN (n,d) : a hyperplane

section of the Veronese variety.

The Veronese surface V of P5 enjoys a lot of interesting properties. Most of

them follow from its property of being covered by a 2-dimensional family of conics,
which are precisely the images via v2,2 of the lines of the plane.
To see this, we’ll use as coordinates in P5 w00 , w01 , w02 , w11 , w12 , w22 , so that
v2,2 sends [x0 , x1 , x2 ] to the point of coordinates wij = xi xj . With this choice of
coordinates, the equations of V are obtained by annihilating the 2 × 2 minors of
the symmetric matrix:
 
w00 w01 w02
M =  w01 w11 w12 
w02 w12 w22

Let ` be a line of P2 of equation b0 x0 + b1 x1 + b2 x2 = 0. Its image is the set of

points of P5 with coordinates wij = xi xj , such that there exists a non-zero triple
[x0 , x1 , x2 ] with b0 x0 + b1 x1 + b2 x2 = 0. But this last equation is equivalent to the
system:

 b0 x20 + b1 x0 x1 + b2 x0 x2 = 0
b x x + b1 x21 + b2 x1 x2 = 0
 0 0 1
b0 x0 x2 + b1 x1 x2 + b2 x22 = 0

It represents the intersection of V with the plane

b0 w00 + b1 w01 + b2 w02 = 0

(
(∗) b0 w01 + b1 w11 + b2 w12 = 0 ,
b0 w02 + b1 w12 + b2 w22 = 0

so v2,2 (`) is a plane curve. Its degree is the number of points in its intersection
with a general hyperplane in P5 : this corresponds to the intersection in P2 of `
with a conic (a hypersurface of degree 2). Therefore v2,2 (`) is a conic.
So the isomorphism v2,2 transforms the geometry of the lines in the plane
in the geometry of the conics on the Veronese surface. In particular, given two
distinct points on V , there is exactly one conic contained in V and passing through
them.
From this observation it is easy to deduce that the secant lines of V , i.e. the
lines meeting V at two points, are precisely the lines of the planes generated by
the conics contained in V , so that the (closure of the) union of these secant lines
46 Mezzetti

coincides with the union of the planes of the conics of V . This union results to be
the cubic hypersurface defined by the equation
 
w00 w01 w02
det M = det  w01 w11 w12  = 0.
w02 w12 w22
Indeed a point of P5 , of coordinates [wij ] belongs to the plane of a conic contained
in V if and only if there exists a non-zero triple [b0 , b1 , b2 ] which is solution of the
homogeneous system (*).

Let X, Y be quasi–projective varieties.

9.10. Definition. The rational maps from X to Y are the germs of regular maps
from open subsets of X to Y , i.e. equivalence classes of pairs (U, φ), where U 6= ∅
is open in X and φ : U → Y is regular, with respect to the relation: (U, φ) ∼ (V, ψ)
if and only if φ|U ∩V = ψ|U ∩V . The following Lemma guarantees that the above
defined relation satisfies the transitive property.

9.11. Lemma. Let φ, ψ : X → Y be regular maps between quasi-projective vari-

eties. If φ|U = ψ|U for U ⊂ X open and non–empty, then φ = ψ.
Proof. Let P ∈ X and consider φ(P ), ψ(P ) ∈ Y . There exists a hyperplane H
such that φ(P ) 6∈ H and ψ(P ) 6∈ H. Up to a projective transformation, we can
assume that H = VP (x0 ), so φ(P ), ψ(P ) ∈ U0 . Set V = φ−1 (U0 ) ∩ ψ −1 (U0 ): an
open neighbourhood of P . Consider the restrictions of φ and ψ from V to Y ∩ U0 :
they are regular maps which coincide on V ∩ U , hence their coordinates φi , ψi ,
i = 1, . . . , n, coincide on V ∩ U . So φi |V = ψi |V . In particular φ(P ) = ψ(P ). 

A rational map from X to Y will be denoted φ : X 99K Y . As for rational

functions, the domain of definition of φ, dom φ, is the maximum open subset of
X such that φ is regular at the points of dom φ.
The following proposition follows from the characterization of rational func-
tions on affine varieties.

9.12. Proposition. Let X, Y be affine algebraic sets, with Y closed in An . Then

φ : X 99K Y is a rational map if and only if φ = (φ1 , . . . , φn ), where φ1 , . . . , φn ∈
K(X). 

If X ⊂ Pn , Y ⊂ Pm , then a rational map X 99K Y is assigned by giving m + 1

homogeneous polynomials of K[x0 , x1 , . . . , xn ] of the same degree, F0 , . . . , Fm , such
that at least one of them is not identically zero on X.
A rational map φ : X 99K Y is called dominant if the image of X via φ is
dense in X, i.e. if φ(U ) = X, where U = dom φ. If φ : X 99K Y is dominant
 Introduction to algebraic geometry 47

and ψ : Y 99K Z is any rational map, then dom ψ ∩ Imφ 6= ∅, so we can define
ψ ◦ φ : X 99K Z: it is the germ of the map ψ ◦ φ, regular on φ−1 (dom ψ ∩ Imφ).

9.13. Definition. A birational map from X to Y is a rational map φ : X 99K Y

such that φ is dominant and there exists ψ : Y 99K X, a dominant rational map,
such that ψ ◦ φ = 1X and φ ◦ ψ = 1Y as rational maps. In this case, X and Y are
called birationally equivalent or simply birational.
If φ : X 99K Y is a dominant rational map, then we can define the comorphism
φ∗ : K(Y ) → K(X) in the usual way: it is an injective K–homomorphism.

9.14. Proposition. Let X, Y be quasi–projective varieties, u : K(Y ) → K(X)

be a K–homomorphism. Then there exists a rational map φ : X 99K Y such that
φ∗ = u.
Proof. Y is covered by open affine varieties Yα , α ∈ I (by Proposition 9.9): for all
index α, K(Y ) ' K(Yα ) (Prop. 8.8) and K(Yα ) ' K(t1 , . . . , tn ), where t1 , . . . , tn
can be interpreted as coordinate functions on Yα . Then u(t1 ), . . . , u(tn ) ∈ K(X)
and there exists U ⊂ X, non–empty open subset such that u(t1 ), . . . , u(tn ) are all
regular on U . So u(K[t1 , . . . , tn ]) ⊂ O(U ) and we can consider the regular map
u] : U → Yα ,→ Y . The germ of u] gives a rational map X 99K Y . It is possible
to check that this rational map does not depend on the choice of Yα and U . 

9.15. Theorem. Let X, Y be quasi–projective varieties. The following are equiv-

alent:
(i) X is birational to Y ;
(ii) K(X) ' K(Y );
(iii) there exist non–empty open subsets U ⊂ X and V ⊂ Y such that U ' V .
Proof.
(i) ⇔ (ii) via the construction of the comorphism φ∗ associated to φ and of
u] , associated to u : K(Y ) → K(X). One checks that both constructions are
functorial.
(i) ⇒ (iii) Let φ : X 99K Y , ψ : Y 99K X be inverse each other. Put
U = dom φ and V 0 = dom ψ. By assumption, ψ ◦ φ is defined on φ−1 (V 0 )
0

and coincides with 1X there. Similarly, ψ ◦ φ is defined on ψ −1 (U 0 ) and equal

to 1Y . Then φ and ψ establish an isomorphism between the corresponding sets
U := φ−1 (ψ −1 (U 0 )) and V := ψ −1 (φ−1 (V 0 )).
(iii) ⇒ (ii) U ' V implies K(U ) ' K(V ); but K(U ) ' K(X) and K(V ) '
K(Y ) (Prop.8.8), so K(X) ' K(Y ) by transitivity. 

9.16. Corollary. If X is birational to Y , then dim X = dim Y . 

9.17. Examples.
48 Mezzetti

a) The cuspidal cubic Y = V (x3 − y 2 ) ⊂ A2 .

We have seen that Y is not isomorphic to A1 , but in fact Y and A1 are
birational. Indeed, the regular map φ : A1 → Y , t → (t2 , t3 ), admits a rational
inverse ψ : Y 99K A1 , (x, y) → xy . ψ is regular on Y \ {(0, 0)}, ψ is dominant and
ψ◦φ = 1A1 , φ◦ψ = 1Y as rational maps. In particular, φ∗ : K(Y ) → K(X) is a field
isomorphism. Recall that K[Y ] = K[t1 , t2 ], with t21 = t32 , so K(Y ) = K(t1 , t2 ) =
K(t2 /t1 ), because t1 = (t2 /t1 )2 = t22 /t21 = t31 /t21 and t2 = (t2 /t1 )3 = t32 /t31 = t32 /t22 ,
so K(Y ) is generated by a unique transcendental element. Notice that φ and ψ
establish isomorphisms between A1 \ {0} and Y \ {(0, 0)}.

b)Rational maps from P1 to Pn .

Let φ : P1 99K Pn be rational: on some open U ⊂ P1 ,

φ([x0 , x1 ]) = [F0 (x0 , x1 ), . . . , Fn (x0 , x1 )],

with F0 , . . . , Fn homogeneous of the same degree, without non–trivial common

factors. Assume that Fi (P ) = 0 for a certain index i, with P = [a0 , a1 ]. Then
Fi ∈ Ih (P ) = ha1 x0 − a0 x1 i, i.e. a1 x0 − a0 x1 is a factor of Fi . This remark implies
that ∀ Q ∈ P1 there exists i ∈ {0, . . . , n} such that Fi (Q) 6= 0, because otherwise
F0 , . . . , Fn would have a common factor of degree 1. Hence we conclude that φ is
regular.

c) Projections.
Let φ : Pn 99K Pm be given in matrix form by Y = AX, where A is a
(m + 1) × (n + 1)-matrix, with entries in K. Then φ is a rational map, regular on
Pn \ P(KerA). Put Λ := P(KerA). If A = (aij ), this means that Λ has cartesian
equations
a00 x0 + . . . + a0n xn = 0


a10 x0 + . . . + a1n xn = 0


...

am0 x0 + . . . + amn xn = 0
The map φ has a geometric interpretation: it can be seen as the projection
of centre Λ to a complementar linear space. First of all, we can assume that rk
A = m + 1, otherwise we replace Pm with P(Im A); hence dim Λ = n − (m + 1).
Consider first the case Λ : x0 = . . . = xm = 0; we identify Pm with the
subspace of Pn of equations xm+1 = . . . = xn = 0, so Λ and Pm are complementar
subspaces, i.e. Λ ∩ Pm = ∅ and the linear span of Λ and Pm is Pn . Then, for
Q ∈ Pn \ Λ, φ(Q) = [x0 , . . . , xm , 0, . . . , 0]: it is the intersection of Pm with the
linear span of Λ and Q. In fact, if Q[a0 , . . . , an ] then ΛQ has equations

{ai xj − aj xi = 0, i, j = 0, . . . , m (check!)

so ΛQ ∩ Pm has coordinates [a0 , . . . , am , 0, . . . , 0].

 Introduction to algebraic geometry 49

In the general case, if Λ = VP (L0 , . . . , Lm ), with L0 , . . . , Lm linearly inde-

pendent forms, we can identify Pm with VP (Lm+1 , . . . , Ln ), where L0 , . . . , Lm ,
Lm+1 , . . . , Ln is a basis of (K n+1 )∗ . Then L0 , . . . , Lm can be interpreted as coor-
dinate functions on Pm .
If m = n − 1, then Λ is a point P and φ, often denoted πP , is the projection
from P to a hyperplane not containing P .

d)Rational and unirational varieties.

A quasi–projective variety X is called rational if it is birational to a projective
space Pn (or equivalently to An ). By Theorem 9.15, X is rational if and only if
K(X) ' K(Pn ) = K(x1 , . . . , xn ) for some n, i.e. if K(X) is an extension of K
generated by a transcendence basis. In an equivalent way, X is rational if there
exists a rational map φ : Pn 99K X which is dominant and is an isomorphism
if restricted to a suitable open subset U ⊂ Pn . Hence X admits a birational
parametrization by n parameters.
A weaker notion is that of unirational variety: X is unirational if there exists
a dominant rational map Pn 99K X i.e. if K(X) is contained in the quotient field of
a polynomial ring. Hence X can be parametrized, but not necessarily generically
one–to–one.
It is clear that, if X is rational, then it is unirational. The converse implication
has been an important open problem, up to 1971, when it has been solved in
the negative, for varieties of dimension ≥ 3 (Clemens–Griffiths and Iskovskih–
Manin). The result is true for curves (Theorem of Lüroth, 1880) and for surfaces
if charK = 0 (Theorem of Castelnuovo, 1894).
As an example of rational variety, let us consider the following quadric of
maximal rank in P3 : X = VP (x0 x3 − x1 x2 ): an irreducible hypersurface of de-
gree 2. Let πP : P3 99K P2 be the projection of centre P [1, 0, 0, 0], such that
πP ([y0 , y1 , y2 , y3 ]) = [y1 , y2 , y3 ]. The restriction of πP to X is a rational map
π˜P : X 99K P2 , regular on X \ {P }. π˜P has a rational inverse: indeed consider
the rational map ψ : P2 99K X, [y1 , y2 , y3 ] → [y1 y2 , y1 y3 , y2 y3 , y32 ]. The equation
of X is satisfied by the points of ψ(P2 ): (y1 y2 )y32 = (y1 y3 )(y2 y3 ). ψ is regular on
P2 \ VP (y1 y2 , y3 ). Let us compose ψ and π̃P :

π ψ
P
[y0 , . . . , y3 ] ∈ X → [y1 , y2 , y3 ] → [y1 y2 , y1 y3 , y2 y3 , y32 ];

y1 y2 = y0 y3 implies ψ ◦ πP = 1X . In the opposite order:

ψ π
[y1 , y2 , y3 ] → [y1 y2 , y1 y3 , y2 y3 , y32 ] →
P
[y1 y3 , y2 y3 , y32 ] = [y1 , y2 , y3 ].

So X is birational to P2 hence it is a rational surface.

e) A birational non–regular map from P2 to P2 .

50 Mezzetti

The following rational map is called the standard quadratic map:

Q : P2 99K P2 , [x0 , x1 , x2 ] → [x1 x2 , x0 x2 , x0 x1 ].

Q is regular on U := P2 \ {A, B, C}, where A[1, 0, 0], B[0, 1, 0], C[0, 0, 1] are the
fundamental points (see Fig. 2)
Let a be the line through B and C: a = VP (x0 ), and similarly b = VP (x1 ),
c = VP (x2 ). Then Q(a) = A, Q(b) = B, Q(c) = C. Outside these three lines Q is
an isomorphism. Precisely, put U 0 = P2 \ {a ∪ b ∪ c}; then Q : U 0 → P2 is regular,
the image is U 0 and Q−1 : U 0 → U 0 coincides with Q. Indeed,
Q Q
[x0 , x1 , x2 ] → [x1 x2 , x0 x2 , x0 x1 ] → [x20 x1 x2 , x0 , x21 x2 , x0 x1 x22 ].

So Q ◦ Q = 1P2 as rational map, hence Q is birational and Q = Q−1 .

– Fig. 2 –
The set of the birational maps P2 99K P2 is a group, called the Cremona
group. At the end of XIX century, Max Noether proved that the Cremona group
is generated by P GL(3, K) and by the single standard quadratic map above. The
analogous groups for Pn , n ≥ 3, are much more complicated and a complete
description is still unknown.

Exercises to §9.
1. Let φ : A1 → An be the map defined by t → (t, t2 , . . . , tn ).
a) Prove that φ is regular and describe φ(A1 );
b) prove that φ : A1 → φ(A1 ) is an isomorphism;
∗
c) give a description of φ∗ and φ−1 .
2. Let f : A2 → A2 be defined by: (x, y) → (x, xy).
 Introduction to algebraic geometry 51

a) Describe f (A2 ) and prove that it is not locally closed in A2 .

b) Prove that f (A2 ) is a constructible set in the Zariski topology of A2 (i.e.
a finite union of locally closed sets).
3. Prove that the Veronese variety Vn,d is not contained in any hyperplane of
N (n,d)
P .
4. Let GLn (K) be the set of invertible n × n matrices with entries in K.
Prove that GLn (K) can be given the structure of an affine variety.
5. Show the unicity of the projective transformation τ of Theorem 9.8.
6. Let φ : X → Y be a regular map and φ∗ its comorphism. Prove that the
kernel of φ∗ is the ideal of φ(X) in O(Y ). Deduce that φ is dominant if and only
if φ∗ is injective.
7. Prove that O(XF ) is isomorphic to O(X)f , where X is an affine algebraic
set, F a polynomial and f the function on X defined by F .

10. Products of quasi–projective varieties.

Let Pn , Pm be projective spaces over the same field K. The cartesian product
Pn × Pm is simply a set: we want to define an injective map from Pn × Pm to a
suitable projective space, so that the image be a projective variety, which will be
identified with our product.
Let N = (n(+1)(m + 1) − 1 and define σ : Pn × Pm → PN in the following
way: ([x0 , . . . , xn ], [y0 , . . . , ym ]) → [x0 y0 , x0 y1 , . . . , xi yj , . . . , xn ym ]. Using coordi-
nates wij , i = 0, . . . , n, j = 0, . . . , m, in PN , σ is given by

{wij = xi yj , i = 0, . . . , n, j = 0, . . . , m.

It is easy to observe that σ is a well–defined map.

If Pn = P(V ) and Pm = P(W ), note that PN ' P(V ⊗ W ) and σ is induced
by the natural map V × W → V ⊗ W .
Let Σn,m (or simply Σ) denote the image σ(Pn × Pm ).

10.1. Proposition. σ is injective and Σn,m is a closed subset of PN .

Proof. If σ([x], [y]) = σ([x0 ], [y 0 ]), then there exists λ 6= 0 such that x0i yj0 = λxi yj
for all i, j. In particular, if xh 6= 0, yk 6= 0, then also x0h 6= 0, yk0 6= 0, and for all i
x0i = λ yyk0 xi , so [x0 , . . . , xn ] = [x00 , . . . , x0n ]. Similarly for the second point.
k
To prove the second assertion, I claim: Σn,m is the closed set of equations:

(∗){wij whk = wik whj , i, h = 0, . . . , n; j, k = 0 . . . , m.

It is clear that if [wij ] ∈ Σ, then it satisfies (*). Conversely, assume that [wij ]
52 Mezzetti

satisfies (*) and that wαβ 6= 0. Then

[w00 , . . . , wij , . . . , wnm ] = [w00 wαβ , . . . , wij wαβ , . . . , wnm wαβ ] =

= [w0β wα0 , . . . , wiβ wαj , . . . , wnβ wαm ] =
= σ([w0β , . . . , wnβ ], [wα0 , . . . , wαm ]).


σ is called the Segre map and Σn,m the Segre variety or biprojective space. Note
that Σ is covered by the affine open subsets Σij = Σ ∩ Wij , where Wij = PN \
VP (wij ). Moreover Σij = σ(Ui × Vj ), with Ui × Vj ' An+m .

10.2. Proposition. σ|Ui ×Vj : Ui × Vj ' An+m → Σij is an isomorphism of

varieties.
Proof. Assume by simplicity i = j = 0. Choose non–homogeneous coordinates on
U0 : ui = xi /x0 and on V0 : vj = yj /y0 . So u1 , . . . un , v1 , . . . , vm are coordinates on
U0 × V0 . Take non–homogeneous coordinates also on W00 : zij = wij /w00 . Using
these coordinates we have:

σ|Ui ×Vj :(u1 , . . . un , v1 , . . . , vm ) → (v1 , . . . , vm , u1 , u1 v1 , . . . , u1 vm , . . . , un vm )

||
([1, u1 , . . . , un ], [1, v1 , . . . , vm ])

i.e. σ(u1 , . . . , vm ) = (z01 , . . . , znm ), where


 zi0 = ui , if i = 1, . . . , n;
z0j = vj , if j = 1, . . . , m;
zij = ui vj = zi0 z0j otherwise.


Hence σ|U0 ×V0 is regular.

The inverse map takes (z01 , . . . , znm ) to (z10 , . . . , zn0 , z01 , . . . , z0m ), so it is also
regular. 

10.3. Corollary. Pn × Pm is irreducible and birational to Pn+m .

Proof. The first assertion follows from Ex.5, Ch.6, considering the covering of Σ
by the open subsets Σij . For the second assertion, by Theorem 9.15, it is enough
to note that Σn,m and Pn+m contain isomorphic open subsets, i.e. Σij and An+m .

From now on, we shall identify Pn × Pm with Σn,m . If X ⊂ Pn , Y ⊂ Pm are
any quasi–projective varieties, then X × Y will be automatically identified with
σ(X × Y ) ⊂ Σ.
 Introduction to algebraic geometry 53

10.4. Proposition. If X and Y are projective varieties (resp. quasi–projective

varieties), then X × Y is projective (resp. quasi–projective).
Proof. [
σ(X × Y ) = (σ(X × Y ) ∩ Σij ) =
i,j
[
= (σ(X × Y ) ∩ (Ui × Vj )) =
i,j
[
= (σ((X ∩ Ui ) × (Y ∩ Vj ))).
i,j

If X and Y are projective varieties, then X ∩ Ui is closed in Ui and Y ∩ Vj is closed

in Vj , so their product is closed in Ui × Vj ; since σ|Ui ×Vj is an isomorphism, also
σ(X × Y ) ∩ Σij is closed in Σij , so σ(X × Y ) is closed in Σ, by Lemma 8.3.
If X, Y are quasi–projective, the proof is similar. As for the irreducibility, see
Exercise 10.1. 

10.5. Example. P1 × P1
σ : P1 × P1 → P3 is given by {wij = xi yj , i = 0, 1, j = 0, 1. Σ has only one
non–trivial equation: w00 w11 − w01 w10 , hence Σ is a quadric. The equation of Σ
can be written as
w00 w01
(∗) = 0.
w10 w11

Σ contains two families of special closed subsets parametrized by P1 , i.e.

{σ(P × P1 )}P ∈P1 and {σ(P1 × Q)}Q∈P1 .

If P [a0 , a1 ], then σ(P × P1 ) is given by the equations:

w = a y
 00 0 0
w = a y
01 0 1

 w10 = a1 y0

w11 = a1 y1

hence it is a line. Cartesian equations of σ(P × P1 ) are:


a1 w00 − a0 w10 = 0
a1 w01 − a0 w11 = 0;

they express the proportionality of the rows of the matrix (*) with coefficients
[a1 , −a0 ]. Similarly, σ(P1 × Q) is the line of equations

a1 w00 − a0 w01 = 0
n
a1 w10 − a0 w11 = 0.
54 Mezzetti

Hence Σ contains two families of lines: two lines of the same family are clearly
disjoint while two lines of different families intersect at one point (σ(P, Q)). Con-
versely, through any point of Σ there pass two lines, one for each family. Note
that Σ is exactly the quadric surface of Example 9.17, d) and that the projection
of centre [1, 0, 0, 0] realizes an explicit birational map between P1 × P1 and P2 .

Exercises to §10.
1. Using Ex. 5 of §6, prove that, if X ⊂ Pn , Y ⊂ Pm are irreducible projective
varieties, then X × Y is irreducible.
2. (*) Let X ⊂ An , Y ⊂ An . Show that X ∩ Y ' (X × Y ) ∩ ∆An , where ∆An
is the diagonal subvariety.
3. Let L, M, N be the following lines in P3 :

L : x0 = x1 = 0, M : x2 = x3 = 0, N : x0 − x2 = x1 − x3 = 0.

Let X be the union of lines meeting L, M and N : write equations for X and
describe it: is it a projective variety? If yes, of what dimension and degree?
4. Let X, Y be quasi–projective varieties, identify X × Y with its image via
p1 p2
the Segre map. Check that the two projection maps X × Y → X, X × Y → Y are
regular. (Hint: use the open covering of the Segre variety by the Σij ’s.)

11. The dimension of an intersection.

Our aim in this section is to prove the following theorem:

11.1. Theorem. Let K be an algebraically closed field. Let X, Y ⊂ Pn be

quasi–projective varieties. Assume that X ∩ Y 6= ∅. Then if Z is any irreducible
component of X ∩ Y , then dim Z ≥ dim X + dim Y − n.
The proof uses in an essential way the Krull’s principal ideal theorem (see for
instance Atiyah–MacDonald [1]).
The proof of Theorem 11.1 will be divided in three steps. Note first that we
can assume that X and Y both intersect U0 ' An , so, possibly after restricting
X and Y , we may work with closed subsets of the affine space. Put r = dim X,
s = dim Y .
Step 1. Assume that X = V (F ) is an irreducible hypersurface, with F ir-
reducible polynomial of K[x1 , . . . , xn ]. The irreducible components of X ∩ Y cor-
respond, by the Nullstellensatz, to the minimal p prime ideals containing
p I(X ∩ Y )
in K[x1 , . . . , xn ]. Let me recall that I(X ∩ Y )= I(X) + I(Y )= hI(Y ), F i. So
those prime ideals are the minimal ones over hI(Y ), F i. They correspond bijec-
tively to minimal prime ideals containing hf i in O(Y ), where f is the regular
function on Y defined by F . We distinguish two cases:
- if Y ⊂ X = V (F ), then f = 0 and Y ∩ X = Y ; s = dim Y > r + s − n =
(n − 1) + s − n. So the theorem is true.
 Introduction to algebraic geometry 55

- if Y 6⊂ X, then f 6= 0, moreover f is not invertible, otherwise X ∩ Y = ∅:

hence the minimal prime ideals over hf i in O(Y ) have all height one, so for all Z,
irreducible component of X ∩ Y , dim Z = dim Y − 1 = r + s − n (Theorem 7.7).
Step 2. Assume that I(X) is generated by n − r polynomials (where n − r is
the codimension of X): I(X) = hF1 , . . . , Fn−r i. Then we can argue by induction
on n − r: we first intersect Y with V (F1 ), whose irreducible components are all
hypersurfaces, and apply Step 1: all irreducible components of Y ∩ V (F1 ) have
dimension either s or s − 1. Then we intersect each of these components with
V (F2 ), and so on.
Step 3. We use the isomorphism ψ : X ∩ Y ' (X × Y ) ∩ ∆An (see Ex.2,
§10). ψ preserves the irreducible components and their dimensions, so we consider
instead of X and Y , the algebraic sets X × Y and ∆An , contained in A2n . We have
dim X ×Y = r+s (Proposition 7.10). ∆An is a linear subspace of A2n , so it satisfies
the assumption of Step 2. Hence, for all Z we have: dim Z ≥ (r + s) + n − 2n =
r + s − n. 

The above theorem can be seen as a generalization of the Grassmann relation

for linear subspaces. It is not an existence theorem, because it says nothing about
X ∩ Y being non–empty. But for projective varieties, the following more precise
version of the theorem holds:

11.2. Theorem. Let X, Y ⊂ Pn be projective varieties of dimensions r, s. If

r + s − n ≥ 0, then X ∩ Y 6= ∅.
Proof. Let C(X), C(Y ) be the affine cones associated to X and Y . Then C(X) ∩
C(Y ) is certainly non–empty, because it contains the origin O(0, 0, . . . , 0). Assume
we know that C(X) has dimension r + 1 and C(Y ) has dimension s + 1: then
by Theorem 11.1 all irreducible components Z of C(X) ∩ C(Y ) have dimension
≥ (r + 1) + (s + 1) − (n + 1) = r + s − n + 1 ≥ 1, hence Z contains points different
from O. These points give raise to points of Pn belonging to X ∩ Y . It remains to
show:

11.3. Proposition. Let Y ⊂ Pn be a projective variety.

Then dim Y = dim C(Y ) − 1. If S(Y ) denotes the homogeneous coordinate
ring, hence also dim Y = dim S(Y ) − 1.
Proof. Let p : An+1 \ {O} → Pn be the canonical map. Let us recall that C(Y ) =
p−1 (Y ) ∪ {O}. Assume that Y0 := Y ∩ U0 6= ∅ and consider also C(Y0 ) = p−1 (Y0 ) ∪
{O}. Then we have:

C(Y0 ) = {(λ, λa1 , . . . , λan ) | λ ∈ K, (a1 , . . . , an ) ∈ Y0 }.

So we can define a birational map between C(Y0 ) and Y0 × A1 as follows:

(y0 , y1 , . . . , yn ) ∈ C(Y0 ) → ((y1 /y0 , . . . , yn /y0 ), y0 ) ∈ Y0 × A1 ,

56 Mezzetti

((a1 , . . . , an ), λ) ∈ Y0 × A1 → (λ, λa1 , . . . , λan ) ∈ C(Y0 ).

Therefore dim C(Y0 ) = dim(Y0 × A1 ) = dim Y0 + 1. To conclude, it is enough to
remark that dim Y = dim Y0 and dim C(Y ) = dim C(Y0 ) = dim S(Y ). 

11.4. Corollaries.
1. If X, Y ⊂ P2 are projective curves over an algebraically closed field, then
X ∩ Y 6= ∅.
2. P1 × P1 is not isomorphic to P2 .
Proof. 1. is a straightforward application of Theorem 11.2. To prove 2., assume
by contradiction that φ : P1 × P1 → P2 is an isomorphism. If L, L0 are skew lines
on P1 × P1 , then φ(L), φ(L0 ) are rational disjoint curves of P2 , which contradicts
1.
If X, Y ⊂ Pn are varieties of dimensions r, s, then r + s − n is called the
expected dimension of X ∩ Y . If all irreducible components Z of X ∩ Y have the
expected dimension, then we say that the intersection X ∩ Y is proper or that X
and Y intersect properly.
For example, two plane projective curves X, Y intersect properly if they don’t
have any common irreducible component. In this case, it is possible to preview
the number of points of intersections. Precisely, it is possible to associate to every
point P ∈ X ∩ Y a number P i(P ), called the multiplicity of intersection of X and
Y at P , in such a way that P ∈X∩Y i(P ) = dd0 , where d is the degree of X and
d0 is the degree of Y . This result is known as Theorem of Bézout, and is the first
result of the branch of algebraic geometry called Intersection Theory. For a proof
of the Theorem of Bézout, see for instance the classical book of Walker [8], or the
book of Fulton on Algebraic Curves [5].
Let X be a closed subvariety of Pn (resp. of An ) of codimension r. X is called
a complete intersection if Ih (X) (resp. I(X)) is generated by r polynomials.
Hence, if X is a complete intersection of codimension r, then X is certainly the
intersection of r hypersurfaces. Conversely, if X is intersection of r hypersurfaces,
then, by Theorem 11.1, using induction, we deduce that dim X ≥ n − r; also
assuming equality, we cannot conclude that X is a complete intersection, but
simply that I(X) is the radical of an ideal generated by r polynomials.

11.5. Example. Let X ⊂ P3 be the skew cubic. The homogeneous ideal of X

contains the three polynomials F1 , F2 , F3 , the 2 × 2–minors of the matrix
 
x0 x1 x2
M= ,
x1 x2 x3

which are linearly independent polynomials of degree 2. Note that Ih (X) does not
contain any linear polynomial, because X is not contained in any hyperplane, and
that the homogeneous component of minimal degree 2 of Ih (X) is a vector space
 Introduction to algebraic geometry 57

of dimension 3. Hence Ih (X) cannot be generated by two polynomials, i.e. X is

not a complete intersection.
Nevertheless, X is the intersection of the surfaces VP (F ), VP (G), where

x0 x1 x2
x0 x1
F = and G = x1 x2 x3 .
x1 x2
x2 x3 x0

In fact: G = x0 F − x3 (x0 x3 − x1 x2 ) + x2 (x1 x3 − x22 ). If P [x0 , . . . , x3 ] ∈ VP (F ) ∩

VP (G), then P is a zero of x0 x23 − 2x1 x2 x3 + x32 , and also of

x2 (x0 x23 − 2x1 x2 x3 + x32 ) = x21 x23 − 2x1 x22 x3 + x42 = (x1 x3 − x22 )2 .

Hence P is a zero also of x1 x3 − x22 . So P annihilates x3 (x0 x3 − x1 x2 ). If x3 =

0, then also x2 = 0 and x1 = 0: P = [1, 0, 0, 0] ∈ X. If x3 6= 0, then P ∈
VP (F1 , F2 , F3 ) = X.
The situation is that the skew cubic X is the set-theoretic intersection of a
quadric and a cubic, which are tangent along X, so their intersection is X counted
with multiplicity 2.
This example motivates the following definition: X is a set–theoretic complete
intersection if codimX = r and the ideal of X is the radical of an ideal generated by
r polynomials. It is an open problem if all irreducible curves of P3 are set–theoretic
complete intersections. For more details, see [4].

Exercises to §11.
1. Let X ⊂ P2 be the union of three points not lying on a line. Prove that
the homogeneous ideal of X cannot be generated by two polynomials.

12. Complete varieties.

We work over an algebraically closed field K.

12.1. Definition. Let X be a quasi–projective variety. X is complete if, for
any quasi–projective variety Y , the natural projection on the second factor p2 :
X × Y → Y is a closed map. (Note that both projections p1 , p2 are morphisms:
see Exercise 4 to §10.)

Example. The affine line A1 is not complete: let X = Y = A1 , p2 :

A1 × A1 = A2 → A1 is the map such that (x1 , x2 ) → x2 . Then Z := V (x1 x2 − 1)
is closed in A2 but p2 (Z) = A1 \ {O} is not closed.

12.2. Proposition. (i) If f : X → Y is a regular map and X is complete, then

f (X) is a closed complete subvariety of Y .
(ii) If X is complete, then all closed subvarieties of X are complete.
58 Mezzetti

Proof. (i) Let Γf ⊂ X × Y be the graph of f : Γf = {(x, f (x)) | x ∈ X}. It is clear

that f (X) = p2 (Γf ), so to prove that f (X) is closed it is enough to check that Γf is
closed in X ×Y . Let us consider the diagonal of Y : ∆Y = {(y, y) | y ∈ Y } ⊂ Y ×Y .
If Y ⊂ Pn , then ∆Y = ∆Pn ∩(Y ×Y ), so it is closed because ∆Pn is the closed subset
defined in Σn,n by the equations wij − wji = 0, i, j = 0, . . . , n. There is a natural
map f × 1Y : X × Y → Y × Y , (x, y) → (f (x), y), such that (f × 1Y )−1 (∆Y ) = Γf .
It is easy to see that f × 1Y is regular, so Γf is closed.
Let now Z be any variety and consider p2 : f (X) × Z → Z and the regular
map f × 1Z : X × Z → f (X) × Z. There is a commutative diagram:
p0
2
X ×Z −→ Z
↓ f ×1Z % p2

f (X) × Z

If T ⊂ f (X) × Z, then (f × 1Z )−1 (T ) is closed and p2 (T ) = p02 ((f × 1Z )−1 (T )) is

closed because X is complete. We conclude that f (X) is complete.
(ii) Let T ⊂ X be a closed subvariety and Y be any variety. We have to prove
that p2 : T × Y → Y is closed. If Z ⊂ T × Y is closed, then Z is closed also in
X × Y , hence p2 (Z) is closed because X is complete. 

12.3. Corollaries.
1. If X is a complete variety, then O(X) ' K.
2. If X is an affine complete variety, then X is a point.
Proof. 1. If f ∈ O(X), f can be interpreted as a regular map f : X → A1 . By
Proposition 12.2, (i), f (X) is a closed complete subvariety of A1 , which is not
complete. Hence f (X) is a point, so f ∈ K.
2. By 1., O(X) ' K. But O(X) ' K[x1 , . . . , xn ]/I(X), hence I(X) is
maximal. By the Nullstellensatz, X is a point. 

12.4. Theorem. Let X be a projective variety. Then X is complete.

Proof. (sketch, see Šafarevič [7].)
1. It is enough to prove that p2 : Pn × Am → Am is closed, for all n, m. This
can be observed by using the local character of closedness and the affine open
coverings of quasi–projective varieties.
2. If x0 , . . . , xn are homogeneous coordinates on Pn and y1 , . . . , ym are coor-
dinates on Am , then any closed subvariety of Pn × Am can be characterized as the
set of common zeroes of a set of polynomials in the variables x0 , . . . , xn , y1 , . . . , ym ,
homogeneous in the first group of variables x0 , . . . , xn .
3. Let Z ⊂ Pn × Am be closed. Then Z is the set of solutions of a system of
equations
{Gi (x0 , . . . , xn ; y1 , . . . , ym ) = 0, i = 1, . . . , t
 Introduction to algebraic geometry 59

where Gi is homogeneous in the x’s. A point P (y 1 , . . . , y m ) is in p2 (Z) if and only

if the system
{Gi (x0 , . . . , xn ; y 0 , . . . , y m ) = 0, i = 1, . . . , t
has a solution in Pn , i.e. if the ideal of K[x0 , . . . , xn ] generated by G1 (x; y),. . . ,
Gt (x; y) has at least one zero in Pn . Hence

p2 (Z) = {(y 1 , . . . , y m )| ∀ d ≥ 1 hG1 (x; y), . . . , Gt (x; y)i 6⊃ K[x0 , . . . , xn ]d } =

\
= {(y 1 , . . . , y m )| hG1 (x; y), . . . , Gt (x; y)i 6⊃ K[x0 , . . . , xn ]d }.
d≥1

Let {Mα }α∈I be the set of the monomials of degree d in K[x0 , . . . , xn ]; let di =
deg Gi (x; y); let {Niβ } be the set of the monomials of degree d − di ; let finally
Td = {(y 1 , . . . , y m )| hG1 (x; y), . . . , Gt (x; y)i 6⊃ K[x0P
, . . . , xn ]d }.
Then P (y 1 , . . . , y m ) 6∈ Td if and only if Mα = i Gi (x; y)Fi,α (x0 , . . . , xn ), for
all α and for suitable polynomials Fi,α homogeneous of degree d − di . So P 6∈ Td if
and only if for all α Mα is a linear combination of the polynomials {Gi (x; y)Niβ },
i.e. the matrix A of the coefficients of the polynomials Gi (x; y)Niβ with respect to
the basis {Mα } has maximal rank n+d


d . So Td is the set of zeroes of the minors
of a fixed order of the matrix A, hence it is closed. 

12.5. Corollary. Let X be a projective variety. Then O(X) ' K.

12.6. Corollary. Let X be a projective variety, φ : X → Y be any regular map.

Then φ(X) is a projective variety. In particular, if X ' Y , then Y is projective.

13. The tangent space.

We define the tangent space TX,p at a point P of an affine variety X as the union
of the lines passing through P and touching X at P . Then we will find a “local”
characterization of TX,p , only depending on the local ring OX,p : this will allow to
define the tangent space at a point of any quasi–projective variety.
Assume first that X ⊂ An is closed and P = (0, . . . , 0). Let L be a line
through P : if A(a1 , . . . , an ) is another point of L, then a general point of L has
coordinates (ta1 , . . . , tan ), t ∈ K. If I(X) = (F1 , . . . , Fm ), then the intersection
X ∩ L is determined by the following system of equations in the undeterminate t:

F1 (ta1 , . . . , tan ) = . . . = Fm (ta1 , . . . , tan ) = 0.

The solutions of these equations are the roots of the greatest common divisor
G(t) of the polynomials F1 (ta1 , . . . , tan ), . . . , Fm (ta1 , . . . , tan ) in K[t]. We may
factorize G(t) as G(t) = cte (t − α1 )e1 . . . (t − αs )es , where e > 0 if and only if
P ∈ X ∩ L, and α1 , . . . , αs 6= 0. The number e is by definition the intersection
60 Mezzetti

multiplicity at P of X and L. If G(t) is identically zero, i.e. if L ⊂ X, the

intersection multiplicity is, by definition, +∞.
Note that the polynomial G(t) doesn’t depend on the choice of the generators
F1 , . . . , Fm of I(X), but only on the ideal I(X) and on L.

13.1. Definition. The line L is tangent to the variety X at P if the inter-

section multiplicity of L and X at P is at least 2 (in particular if L ⊂ X). The
tangent space to X at P is the union of the lines that are tangent to X at P ;
it is denoted TP,X .

Assume now that P ∈ X: then the polynomials Fi may be written in the

form Fi = Li + Gi , where Li is a linear polynomial (possibly zero) and Gi contains
only terms of degree ≥ 2. Then

Fi (ta1 , . . . , tan ) = tLi (a1 , . . . , an ) + Gi (ta1 , . . . , tan ),

where the last term is divisible by t2 . So L is tangent to X at P if and only if

Li (a1 , . . . , an ) = 0 for all i = 1, . . . , m.
Therefore a point A(a1 , . . . , an ) belongs to TP,X if and only if

L1 (a1 , . . . , an ) = . . . = Lm (a1 , . . . , an ) = 0.

This shows that TP,X is a linear subspace of An , whose equations are the linear
parts of the equations defining X.

13.2. Examples.
(i) TO,An = An , because I(An ) = (0).
(ii) If X is a hypersurface, I(X) = (F ), F = L + G, then TP,X = V (L):
so TP,X is either a hypersurface if L 6= 0, or the whole space An if L = 0. For
instance, if X is the affine plane cuspidal cubic V (x3 − y 2 ) ⊂ A2 , TP,X = A2 .

Assume now that P ∈ X has coordinates (y1 , . . . , yn ). With a linear trans-

formation we may translate P to the origin (0, . . . , 0), taking as new coordinates
functions on An x1 − y1 , . . . , xn − yn . This corresponds to the K-isomorphism
K[x1 , . . . , xn ] −→ K[x1 − y1 , . . . , xn − yn ], which takes a polynomial F (x1 , . . . , xn )
to its Taylor expansion
(2)
G(x1 − y1 , . . . , xn − yn ) = F (y1 , . . . , yn ) + dP F + dP F + . . . ,

(i)
where dP F denotes the ith differential of F at P : it is a homogeneous polynomial
of degree i in the variables x1 − y1 , . . . , xn − yn . In particular the linear term is

∂F ∂F
dP F = (P )(x1 − y1 ) + . . . + (P )(xn − yn ).
∂x1 ∂xn
 Introduction to algebraic geometry 61

We get that, if I(X) = (F1 , . . . , Fm ), then TP,X is the linear subspace of An defined
by the equations
dP F1 = . . . = dP Fm = 0.

The affine space An , which may identified with K n , has a natural structure
of K-vector space with origin P , so in a natural way TP,X is a vector subspace
(with origin P ). The functions x1 − y1 , . . . , xn − yn form a basis of the dual space
(K n )∗ and their restrictions generate TP,X ∗ ∗
. Note moreover that dim TP,X = k
if and only if n − k is the maximal number of polynomials linearly independent
among dP F1 , . . . , dP Fm . If dP F1 , . . . , dP Fn−k are these polynomials, then they
⊥
form a base of the orthogonal TP,X of the vector space TP,X in (K n )∗ , because
they vanish on TP,X .

Let us define now the differential of a regular function. Let f ∈ O(X) be

a regular function on X. We want to define the differential of f at P . Since
X is closed in An , f is induced by a polynomial F ∈ K[x1 , . . . , xn ] and by all
polynomials of the form F + G with G ∈ I(X). Fix P ∈ X: then dP (F + G) =
dP F + dP G so the differentials of two polynomials inducing the same function f
on X differ by the term dP G with G ∈ I(X). By definition, dP G is zero along
TP,X , so we may define dp f as a regular function on TP,X , the differential of f at
P : it is the function on TP,X induced by dP F . Since dP F is a linear combination
∗
of x1 − y1 , . . . , xn − yn , dp f can also be seen as an element of TP,X .

∗
There is a natural map dp : O(X) → TP,X , which sends f to dp f . It is clear
that dP (f + g) = dP f + dP g and dP (f g) = f (P )dP g + g(P )dP f . In particular,
if c ∈ K, dp (cf ) = cdP f . So dp is a linear map of K-vector spaces. We denote
again by dP the restriction of dP to IX (P ), the maximal ideal of the regular
functions on X which are zero at P . Since clearly f = f (P ) + (f − f (P )) then
dP f = dP (f − f (P )), so this restriction doesn’t modify the image of the map.

∗
13.3. Proposition. The map dP : IX (P ) → TP,X is surjective and its kernel is
2 ∗ 2
IX (P ) . Therefore TP,X ' IX (P )/IX (P ) as K-vector spaces.
∗
Proof. Let φ ∈ TP,X be a linear form on TP,X . φ is the restriction of a linear
n
form on K : λ1 (x1 − y1 ) + . . . + λn (xn − yn ), with λ1 , . . . , λn ∈ K. Let G be the
polynomial of degree 1 λ1 (x1 − y1 ) + . . . + λn (xn − yn ): the function g induced by
G on X is zero at P and coincides with its own differential, so dp is surjective.
Let now g ∈ IX (P ) such that dp g = 0, g induced by a polynomial G. Note that
dP G may be interpreted as a linear form on K n which vanishes on TP,X , hence as
⊥
an element of TP,X . So dP G = c1 dp F1 +. . .+cm dp Fm (c1 , . . . , cm suitable elements
of K). Let us consider the polynomial G − c1 F1 − . . . − cm Fm : since its differential
at P is zero, it doesn’t have any term of degree 0 or 1 in x1 − y1 , . . . , xn − yn , so
it belongs to I(P )2 . Since G − c1 F1 − . . . − cm Fm defines the function g on X, we
conclude that g ∈ IX (P )2 . 
62 Mezzetti

13.4. Corollary. The tangent space TP,X is isomorphic to (IX (P )/IX (P )2 )∗ as

an abstract K-vector space. 

13.5. Corollary. If φ : X → Y is an isomorphism of affine varieties and P ∈ X,

then the tangent spaces TP,X and Tφ(P ),Y are isomorphic.
Proof. φ induces the comorphism φ∗ : O(Y ) → O(X), which is an isomorphism
such that φ∗ IY (φ(P )) = IX (P ) and φ∗ IY (φ(P ))2 = IX (P )2 . So there is an induced
homomorphism
IY (φ(P ))/IY (φ(P ))2 → IX (P )/IX (P )2

which is an isomorphism of K-vector spaces. By dualizing we get the claim. 

The above map from TP,X to Tφ(P ),Y is called the differential of φ at P and
is denoted by dP φ.
Now we would like to find a “more local” characterization of TP,X . To this
end we consider the local ring of P in X OP,X . We recall that O(X) is contained
in OP,X , which is the localization O(X)IX (P ) . It is natural to extend the map
∗
dP : O(X) → TP,X to OP,X setting

f g(P )dP f − f (P )dP g

dP ( ) = .
g g(P )2

∗
As in the proof of Proposition 13.3 one proves that the map dP : OP,X → TP,X
∗
induces an isomorphism MP,X /M2P,X → TP,X , where MP,X is the maximal ideal
of OP,X . So by duality we have: TP,X ' (MP,X /M2P,X )∗ . This proves that the
tangent space TP,X is a local invariant of P in X.

13.6. Definition. Let X be any quasi-projective variety, P ∈ X. The Zariski

tangent space of X at P is the vector space (MP,X /M2P,X )∗ .

It is an abstract vector space, but if X ⊂ An is closed, taking the dual of

the comorphism associated to the embedding of X in An , we have an embedding
of TP,X into TP,An = An . If X ⊂ Pn and P ∈ Ui = An , then TP,X ⊂ Ui : its
projective closure is called the embedded tangent space to X at P .

As we have seen the tangent space TP,X is invariant by isomorphism. In

particular its dimension is invariant. If X ⊂ An is closed, I(X) = (F1 , . . . , Fm ),
then dim TP,X = n − r, where r is the maximal number of linearly independent
elements in the set {dP F1 , . . . , dp Fm }.

∂Fi ∂Fi
Since dP Fi = ∂x1 (P )(x1 − y1 ) + . . . + ∂xn (P )(xn − yn ), r is the rank of the
 Introduction to algebraic geometry 63

following m × n matrix, the Jacobian matrix of X at P :

∂F1 ∂F1
 
∂x1 (P )
... ∂xn (P )
J(P ) =  ... ... ... .
∂Fm ∂Fm
∂x1 (P ) . . . ∂xn (P )

The generic Jacobian matrix of X is instead the following matrix with entries
in O(X):
 ∂F ∂F1

∂x1
1
. . . ∂xn
J =  ... ... ... .
∂Fm
∂x1 . . . ∂F
∂xn
m

The rank of J is ρ when all minors of order ρ + 1 are functions identically zero
on X, while some order ρ minor is different from zero at some point. Hence, for
all P ∈ X rk J(P ) ≤ ρ, and rk J(P ) < ρ if and only if all minors of order ρ of
J vanish at P . It is then clear that there is a non-empty open subset of X where
dim TP,X is minimal, equal to n − ρ, and a proper (possibly empty) closed subset
formed by the points P such that dim TP,X > n − ρ.

13.7. Definition. The points of an irreducible variety X for which dim TP,X =
n − ρ (the minimal) are called smooth or non-singular or simple points of X. The
remaining points are called singular (or multiple). X is called smooth if all its
points are smooth.
If X is quasi-projective, the same argument may be repeated for any affine
open subset.

13.8. Example. Let X ⊂ An be the irreducible hypersurface V (F ). Then

∂F ∂F ∂F
J = ( ∂x 1
. . . ∂xn
) is a row matrix. So rk J = 0 or 1. If rk J = 0, then ∂x i
= 0 in
∂F ∂F
O(X) for all i. So ∂xi ∈ I(Y ) = (F ). Since the degree of ∂xi is ≤ deg F − 1, it
∂F
follows that ∂x i
= 0 in the polynomial ring. If the characteristic of k is zero this
means that F a constant: a contradiction. If char K = p, then F ∈ K[xp1 , . . . , xpn ];
since K is algebraically closed, then F = Gp for a suitable G, but this is impossible
because F is irreducible. So always rk J = 1 = ρ. Hence for P general in X,
i.e. for P varying in a suitable non-empty open subset, dim TP,X = n − 1. For
some particular points, the singular points of X, we can find dim TP,X = n, i.e.
TP,X = An .

So in the case of a hypersurface dim TP,X ≥ dim X for every point P in X,

and equality holds in the smooth locus of X. In general, one reduces to the case
of hypersurfaces using the following theorem:

13.9. Theorem. Every quasi-projective variety X is birational to a hypersurface

in some affine space.
64 Mezzetti

We will not give here the proof (see for instance Šafarevič’s book). Note
that one can reduce the proof to the case in which X is affine and that one
has to prove that the field of rational functions K(X) is isomorphic to a field
of the form K(t1 , . . . , tn ), where t1 , . . . , tn satisfy only one non-trivial relation
F (t1 , . . . , tn ) = 0, where F is an irreducible polynomial with coefficients in K. This
is generally proved using the “Abel’s primitive element” concerning extensions of
fields. One shows that the elements t1 , . . . , tn can be chosen as linear combinations
of the original coordinate functions on X: this means that the birational map is
in fact a suitable projection.

13.10. Theorem. The dimension of the tangent space at a non-singular point of

a variety X is equal to dim X.
Proof. It is enough to prove the claim under the assumption that X is affine. Let
Y be an affine hypersurface birational to X (existing by the previous theorem)
and φ : X 99K Y be a birational map. There exist open non-empty subsets
U ⊂ X and V ⊂ Y such that φ : U → Y is an isomorphism. The set of smooth
points of Y is an open subset W of Y such that W ∩ V is non-empty and that
dim TP,Y = dim Y = dim X for all P ∈ W ∩ V . But φ−1 (W ∩ V ) ⊂ U is open non-
empty and dim TQ,X = dim X for all Q ∈ φ−1 (W ∩ V ). This proves the theorem.


We would like now to study a variety X in a neighborhood of a smooth

point. We have seen that P is smooth for X if and only if dim TP,X = dim X.
Assume X affine: in this case the local ring of P in X is OP,X ' O(X)IX (P ) .
But by Theorem 7.7 we have: dim OP,X = htIX (P ) = dim O(X) = dim X and
dim TP,X = dimK MP,X /M2P,X . Therefore P is smooth if and only if

dimK MP,X /M2P,X = dim OP,X

(the first one is a dimension as K-vector space, the second one is a Krull dimen-
sion). By the Nakayama’s Lemma a base of MP,X /M2P,X corresponds bijectively
to a minimal system of generators of the ideal MP,X (observe that the residue
field of OP,X is K). Therefore P is smooth for X if and only if MP,X is minimally
generated by r elements, where r = dim X, in other words if and only if OP,X is
a regular local ring.

For example, if X is a curve, P is smooth if and only if TP,X has dimension

1, i.e. MP,X is principal: MP,X = (t). This means that the equation t = 0 only
defines the point P , i.e. P has one local equation in a suitable neighborhood of P .

Let P be a smooth point of X and dim X = n. Functions u1 , . . . , un ∈ OP,X

are called local parameters at P if u1 , . . . , un ∈ MP,X and their residues ū1 , . . . , ūn
 Introduction to algebraic geometry 65

∗
in MP,X /M2P,X (= TP,X ) form a base, or equivalently if u1 , . . . , un is a minimal
set of generators of MP,X . Recalling the isomorphism
∗
dP : MP,X /M2P,X → TP,X

we deduce that u1 , . . . , un are local parameters if and only if dP ū1 , . . . , dP ūn are
linearly independent forms on TP,X (which is a vector space of dimension n), if
and only if the system of equations on TP,X

dP ū1 = . . . = dP ūn = 0

has only the trivial solution P (which is the origin of the vector space TP,X .

Let u1 , . . . , un be local parameters at P . There exists an open affine neighbor-

hood of P on which u1 , . . . , un are all regular. We replace X by this neighborhood,
so we assume that X is affine and that u1 , . . . , un are polynomial functions on X.
Let Xi be the closed subset V (ui ) of X: it has codimension 1 in X, because ui is
not identically zero on X (u1 , . . . , un is a minimal set of generators of MP,X ).

13.11. Proposition.
T In this notation, P is a smooth point of Xi , for all i =
1, . . . , n and TP,Xi = {P }.
Proof. Assume that Ui is a polynomial inducing ui , then Xi = V (Ui ) ∩ X =
V (I(X) + (Ui )). So I(Xi ) ⊃ I(X) + (Ui ). By considering the linear parts of
the polynomials of the previous ideal, we get: TP,Xi ⊂ TP,X ∩ V (dP Ui ). By the
assumption on the ui , it follows that TP,X ∩ V (dP U1 ) ∩ . . . ∩ V (dP Un ) = {P }.
Since dim TP,X = n, we can deduce that TP,X ∩ V (dP Ui ) is strictly contained in
TP,X , and dim TP,X ∩ V (dP Ui ) = n − 1. So dim TP,Xi ≤ n − 1 = dim Xi , hence P
T a smooth point on Xi , equality holds and TP,Xi = TP,X ∩ V (dP Ui ). Moreover
is
TP,Xi = {P }. 
T
Note that Xi has no positive-dimensional component passing through P :
otherwise the T
tangent space to T at P would be contained in TP,Xi for all i, against
the fact that TP,Xi = {P }.

13.12. Definition. Let X beTa smooth variety. Subvarieties Y1 , . . . , Yr of X are

called transversal at P , P ∈ Yi , if the intersectionTof the tangent
P spaces TP,Yi
has dimension as small as possible, i.e. if codimTP,X ( TP,Yi ) = codimX Yi .
Taking TP,X as ambient variety, one gets the relation:
\ X
dim TP,Yi ≥ dim TP,Yi − (r − 1) dim TP,X ;

hence
\ \ X
codimTP,X ( TP,Yi ) = dim TP,X − dim TP,Yi ≤ (dim TP,X − dim TP,Yi ) =
66 Mezzetti
X X
= codimTP,X (TP,Yi ) ≤ codimX Yi .

If equality holds, P is a T
smooth point for Yi for all i, moreover we get that P is a
simple point for the set Yi .
For example if X is a surface and P ∈ X is smooth, there is a nbhd U of
P such that P is the transversal intersection of two curves in U , corresponding
to local parameters u1 , u2 . If P is singular we need three functions u1 , u2 , u3 to
generate the maximal ideal MP,X .

References.

[1] M. Atiyah - I. MacDonald, Introduction to Commutative Algebra, Addison

- Wesley (1969)
[2] J. Harris, Algebraic Geometry, Springer (1992)
[3] R. Hartshorne, Algebraic Geometry, Springer (1977)
[4] E. Kunz, Introduction to Commutative Algebra and Algebraic Geometry,
Birkhäuser (1980)
[5] W. Fulton, Algebraic Curves, Benjamin (1969)
[6] S. Lang, Algebra, 2nd ed., Addison-Wesley (1984)
[7] I. Šafarevič, Basic Algebraic Geometry, Springer (1974)
[8] R. Walker, Algebraic Curves, Springer (1978)
[9] J. Semple - L. Roth, Algebraic Geometry, 1957
[10] P. Samuel, O. Zariski, Commutative algebra (2 vol.), Van Nostrand, 1958
[11] J. Dieudonné, Cours de géométrie algébrique, 2 / Précis de géométrie
algébrique élémentaire, Presses Universitaires de France, 1974