Gathmann Algebraic Geometry

Algebraic Geometry
Andreas Gathmann
Class Notes TU Kaiserslautern 2014
Contents
0.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
Introduction . . . . . . . . . .
Affine Varieties . . . . . . . . .
The Zariski Topology . . . . . . .
The Sheaf of Regular Functions . . . .
Morphisms. . . . . . . . . . .
Varieties . . . . . . . . . . .
Projective Varieties I: Topology . . . .
Projective Varieties II: Ringed Spaces . .
Grassmannians . . . . . . . . .
Birational Maps and Blowing Up . . .
Smooth Varieties . . . . . . . . .
The 27 Lines on a Smooth Cubic Surface .
Hilbert Polynomials and Bzouts Theorem
Applications of Bzouts Theorem . . .
Divisors on Curves . . . . . . . .
Elliptic Curves . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
9
16
24
32
38
46
57
66
72
82
90
95
105
113
120
References . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
128
129
0.
0.
Introduction
Introduction
In this introductory chapter we will explain in a very rough sketch what algebraic geometry is about
and what it can be used for. We will stress the many correlations with other fields of research, such
as complex analysis, topology, differential geometry, singularity theory, computer algebra, commutative algebra, number theory, enumerative geometry, and even theoretical physics. The goal of this
chapter is just motivational; you will not find definitions or proofs here (and maybe not even a mathematically precise statement). In the same way, the exercises in this chapter are not designed to be
solved in a mathematically precise way. Rather, they are just given as some food for thought if
you want to think a little further about the examples presented here.
To start from something that you probably know, we can say that algebraic geometry is the combination of linear algebra and algebra:
In linear algebra (as e. g. in the Foundations of Mathematics class [G2]), we study systems
of linear equations in several variables.
In algebra (as e. g. in the Introduction to Algebra class [G3]), one of the main topics is the
study of polynomial equations in one variable.
Algebraic geometry combines these two fields of mathematics by studying systems of polynomial
equations in several variables.
Given such a system of polynomial equations, what sort of questions can we ask? Note that we
cannot expect in general to write down explicitly all the solutions: we know from algebra that even
a single complex polynomial equation of degree d > 4 in one variable can in general not be solved
exactly [G3, Problem 0.2]. So we are more interested in statements about the geometric structure of
the set of solutions. For example, in the case of a complex polynomial equation of degree d, even if
we cannot compute the solutions we know that there are exactly d of them (if we count them with the
correct multiplicities). Let us now see what sort of geometric structure we can find in polynomial
equations in several variables.
Example 0.1. Probably the easiest example that is covered neither in linear algebra nor in algebra
is that of a single polynomial equation in two variables. Let us consider the example
Cn = {(x1 , x2 ) C2 : x22 = (x1 1)(x1 2) (x1 2n)}
C2 ,
where n N>0 . Note that in this case it is actually possible to write down all the solutions, because
the equation is (almost) solved for x2 already: we can pick x1 to be any complex number, and then
get two values for x2 unless x1 {1, . . . , 2n}, in which case there is only one value for x2 (namely
0).
So it seems that Cn looks like two copies of the complex plane, with the two copies of each point
1, . . . , 2n identified: the complex plane parametrizes the values for x1 , and the two copies of it correspond to the two possible values for x2 , i. e. the two roots of the number (x1 1) (x1 2n).
This is not the correct topological picture however, because a complex non-zero number does not
have a distinguished first and second root that could correspond to the first and second copy of the
complex plane. Rather, the two roots of a complex number get exchanged if you run around the
origin once: if we consider a closed path
z = r ei
for 0 2 and fixed r > 0
around the complex origin, the square root of this number would have to be defined by
i
z = re 2
Andreas Gathmann
which gives opposite values at = 0 and = 2. In other words, if in Cn we run around one of the
points 1, . . . , 2n (i. e. around a point at which x2 is the square root of 0), we go from one copy of the
plane to the other. The way to draw this topologically is to cut the two planes along the real intervals
[1, 2], . . . , [2n 1, 2n], and to glue the two planes along these lines as in this picture for n = 3 (lines
marked with the same letter are to be identified):
C
1 A 3 C
5 E
B 2
D 4
F 6
glue
C
1 B 3 D
5 F
A 2
C 4
E 6
To make the picture a little nicer, we can compactify our set by adding two points at infinity one
for each copy of the plane in the same way as we can compactify the complex plane C by adding
a point . The precise construction of this compactification will be given in Example 5.6. If we do
this here, we end up with a compact surface with n 1 handles:
add points
at infinity
Such an object is called a surface of genus n 1; the example above shows a surface of genus 2.
Example 0.2. What happens in the previous Example 0.1 if we move the points 1, . . . , 2n for x1 at
which we have only one value for x2 , i. e. if we consider
Cn = {(x1 , x2 ) C2 : x22 = f (x1 )}
C2
with f some polynomial in x1 of degree 2n? Obviously, as long as the 2n roots of f are still distinct,
the topological picture above does not change. But if two of the roots approach each other and finally
coincide, this has the effect of shrinking one of the tubes connecting the two planes until it reduces
to a singular point (also called a node), as in the following picture on the left:
glue
Obviously, we can view this as a surface with one handle less, where in addition we identify two of
the points (as illustrated in the picture on the right). Note that we can still see the handles when
we draw the surface like this, just that one of the handles results from the gluing of the two points.
0.
Introduction
Note that our examples so far were a little cheated because we said before that we want to figure
out the geometric structure of equations that we cannot solve explicitly. In the examples above
however, the polynomial equation was chosen so that we could solve it, and in fact we used this
solution to construct the geometric picture. Let us now see what we can still do if we make the
polynomial more complicated.
Example 0.3. Let d N>0 , and consider
Cd = {(x1 , x2 ) C2 : f (x1 , x2 ) = 0}
C2 ,
where f is an arbitrary polynomial of degree d. This is an equation that we certainly cannot solve
directly if f is sufficiently general. Can we still deduce the geometric structure of Cd ?
In fact, we can do this with the idea of Example 0.2. We saw there that the genus of the surface
does not change if we perturb the polynomial equation, even if the surface acquires singular points
(provided that we know how to compute the genus of such a singular surface). So why not deform
the polynomial f to something singular that is easier to analyze? Probably the easiest thing that
comes into mind is to degenerate the polynomial f of degree d into a product of d linear equations
l1 , . . . , ld : consider
Cd0 = {(x1 , x2 ) C2 : l1 (x1 , x2 ) ld (x1 , x2 ) = 0}
C2 ,
which should have the same genus as the original Cd .

It is easy to see what Cd0 looks like: of course it is just a union of d complex planes. Any two of them
intersect in a point, and we can certainly choose them so that no three of them intersect in a point.
The picture below shows Cd0 for d = 3 (note that every complex plane is after compactifying it
with a point at infinity just a sphere).
What is the genus of this surface? In the picture above it is obvious that we have one loop; so if
d = 3 we get a surface of genus 1. In the general
case we have d spheres, and every two of them
d
connect in a pair of points, so in total we have 2 connections. But d 1 of them are needed to glue
the d spheres to a connected chain without loops; only the remaining ones then add a handle each.
So the genus of Cd0 (and hence of Cd ) is

d
d 1
(d 1) =
.
2
2
This is commonly called the degree-genus formula for complex plane curves. We will show it in
Proposition 13.17.
Remark 0.4 (Real vs. complex dimension). One of the trivial but common sources for misunderstandings is whether we count dimensions over C or over R. The examples considered above are real
surfaces (the dimension over R is 2), but complex curves (the dimension over C is 1). We have used
the word surface so far as this fitted best to the pictures that we have drawn. When looking at the
theory however, it is usually more natural to call these objects (complex) curves. In what follows,
we always mean the dimension over C unless stated otherwise.
Exercise 0.5. What do we get in Example 0.1 if we consider the equation
Cn0 = {(x1 , x2 ) C2 : x22 = (x1 1)(x1 2) (x1 (2n 1))} C2
for n N>0 instead?
Andreas Gathmann
Exercise 0.6. In Example 0.3, we argued

that a polynomial of degree d in two complex variables

gives rise to a surface of genus d1
.
In
Example
0.1 however, a polynomial of degree 2n gave us a
2
surface of genus n 1. Can you see why these two results do not contradict each other?
Remark 0.7. Here is what we should learn from the examples considered so far:
Algebraic geometry can make statements about the topological structure of objects defined
by polynomial equations. It is therefore related to topology and differential geometry (where
similar statements are deduced using analytic methods).
The geometric objects considered in algebraic geometry need not be smooth (i. e. they
need not be manifolds). Even if our primary interest is in smooth objects, degenerations
to singular objects can greatly simplify a problem (as in Example 0.3). This is a main
point that distinguishes algebraic geometry from other geometric theories (e. g. differential
or symplectic geometry). Of course, this comes at a price: our theory must be strong enough
to include such singular objects and make statements how things vary when we pass from
smooth to singular objects. In this regard, algebraic geometry is related to singularity theory
which studies precisely these questions.
Remark 0.8. Maybe it looks a bit restrictive to allow only algebraic (polynomial) equations to describe our geometric objects. But in fact it is a deep theorem that for compact objects, we would not
get anything different if we allowed holomorphic equations too. In this respect, algebraic geometry
is very much related (and in certain cases identical) to complex (analytic) geometry. The easiest
example of this correspondence is that a holomorphic map from the compactified complex plane
C {} to itself must in fact be a rational map, i. e. a quotient of two polynomials.
Example 0.9. Let us now turn our attention to the next more complicated objects, namely complex
surfaces in 3-dimensional space. We just want to give one example here: let X be the cubic surface
X = {(x1 , x2 , x3 ) : 1 + x13 + x23 + x33 (1 + x1 + x2 + x3 )3 = 0}
C3 .
As X has real dimension 4, it is impossible to draw pictures of it that reflect its topological properties
correctly. Usually, we overcome this problem by just drawing the real part, i. e. we look for solutions
of the equation over the real numbers. This then gives a real surface in R3 that we can draw. We
should just be careful about which statements we can claim to see from this incomplete geometric
picture.
The following picture shows the real part of the surface X.
In contrast to our previous examples, we have now used a linear projection to map the real 3dimensional space onto the drawing plane (and not just a topologically correct picture).
We see that there are some lines contained in X. In fact, one can show that (after a suitable compactification) every smooth cubic surface has exactly 27 lines on it, see Chapter 11. This is another sort of
question that one can ask about the solutions of polynomial equations, and that is not of topological
nature: do they contain curves with special properties (in this case lines), and if so, how many? This
branch of algebraic geometry is usually called enumerative geometry.
0.
Introduction
Remark 0.10. It is probably surprising that algebraic geometry, in particular enumerative geometry,
is very much related to theoretical physics. In fact, many results in enumerative geometry have been
found by physicists first.
Why are physicists interested e. g. in the number of lines on the cubic surface? We try to give a short
answer to this (that is necessarily vague and incomplete): There is a branch of theoretical physics
called string theory whose underlying idea is that the elementary particles (electrons, quarks, . . . )
might not be point-like, but rather 1-dimensional objects (the so-called strings), that are just so small
that their 1-dimensional structure cannot be observed directly by any sort of physical measurement.
When these particles move in time, they sweep out a surface in space-time. For some reason this
surface has a natural complex structure coming from the underlying physical theory.
Now the same idea applies to space-time in general: string theorists believe that space-time is not
4-dimensional as we observe it, but rather has some extra dimensions that are again so small in size
that we cannot observe them directly. (Think e. g. of a long tube with a very small diameter of
course this is a 2-dimensional object, but if you look at this tube from very far away you cannot see
the small diameter any more, and the object looks like a 1-dimensional line.) These extra dimensions
are parametrized by a space that sometimes has a complex structure too; it might for example be the
complex cubic surface that we looked at above.
So in this case we are in fact looking at complex curves in a complex surface. A priori, these curves
can sit in the surface in any way. But there are equations of motion that tell you how these curves
will sit in the ambient space, just as in classical mechanics it follows from the equations of motion
that a particle will move on a straight line if no forces apply to it. In our case, the equations of
motion say that the curve must map holomorphically to the ambient space. As we said in Remark
0.8 above, this is equivalent to saying that we must have algebraic equations that describe the curve.
So we are looking at exactly the same type of questions as we did in Example 0.9 above.
Example 0.11. Let us now have a brief look at curves in 3-dimensional space. Consider the example
C = {(x1 , x2 , x3 ) = (t 3 ,t 4 ,t 5 ) : t C}
C3 .
We have given this curve parametrically, but it is in fact easy to see that we can describe it equally
well in terms of polynomial equations:
C = {(x1 , x2 , x3 ) : x13 = x2 x3 , x22 = x1 x3 , x32 = x12 x2 }.
What is striking here is that we have three equations, although we would expect that a 1-dimensional
object in 3-dimensional space should be given by two equations. But in fact, if you leave out any of
the above three equations, you are changing the set that it describes: if you leave out e. g. the last
equation x32 = x12 x2 , you would get the whole x3 -axis {(x1 , x2 , x3 ) : x1 = x2 = 0} as additional points
that do satisfy the first two equations, but not the last one.
So we see another important difference to linear algebra: it is not true that every object of codimension d can be given by d equations. Even worse, if you are given d equations, it is in general a
difficult task to figure out what dimension their solution has. There do exist algorithms to find this
out for any given set of polynomials, but they are so complicated that you will in general want to
use a computer program to do that for you. This is a simple example of an application of computer
algebra to algebraic geometry.
Exercise 0.12. Show that if you replace the three equations defining the curve C in Example 0.11
by
x13 = x2 x3 , x22 = x1 x3 , x32 = x12 x2 +
for a (small) non-zero number C, the resulting set of solutions is in fact 0-dimensional, as you
would expect from three equations in 3-dimensional space. So we see that very small changes in the
equations can make a very big difference in the resulting solution set. Hence we usually cannot apply
numerical methods to our problems: very small rounding errors can change the result completely.
Remark 0.13. Especially the previous Example 0.11 is already very algebraic in nature: the question
that we asked there does not depend at all on the ground field being the complex numbers. In fact,
Andreas Gathmann
this is a general philosophy: even if algebraic geometry describes geometric objects (when viewed
over the complex numbers), most methods do not rely on this, and therefore should be established
in purely algebraic terms. For example, the genus of a curve (that we introduced topologically in
Example 0.1) can be defined in purely algebraic terms in such a way that all the statements from
complex geometry (e. g. the degree-genus formula of Example 0.3) extend to this more general
setting. Many geometric questions then reduce to pure commutative algebra, which is in some sense
the foundation of algebraic geometry.
Example 0.14. The most famous application of algebraic geometry to ground fields other than the
complex numbers is certainly Fermats Last Theorem: this is just the statement that the algebraic
curve over the rational numbers
C = {(x1 , x2 ) Q2 : x1n + x2n = 1}
Q2
contains only the trivial points where x1 = 0 or x2 = 0. Note that this is very different from the case
of the ground field C, where we have seen in Example 0.3 that C is a curve of genus n1
2 . But a
large part of the theory of algebraic geometry applies to the rational numbers (and related fields) as
well, so if you look at the proof of Fermats Theorem you will notice that it uses e. g. the concepts
of algebraic curves and their genus at many places, although the corresponding point set C contains
only some trivial points. So, in some sense, we can view (algebraic) number theory as a part of
algebraic geometry.
With this many relations to other fields of mathematics (and physics), it is obvious that we have
to restrict our attention in this class to quite a small subset of the possible applications. Although
we will develop the general theory of algebraic geometry, our focus will mainly be on geometric
questions, neglecting number-theoretic aspects most of the time. So, for example, if we say let K
be an algebraically closed field, feel free to read this as let K be the complex numbers and think
about geometry rather than algebra.
Every now and then we will quote results from or give applications to other fields of mathematics.
This applies in particular to commutative algebra, which provides some of the basic foundations of
algebraic geometry. So to fully understand algebraic geometry, you will need to get some background in commutative algebra as well, to the extent as covered e. g. in [AM] or [G5]. However,
we will not assume this here although this is probably not the standard approach it is perfectly
possible to follow these notes without any knowledge of commutative algebra. To make this easier,
all commutative algebra results that we will need will be stated clearly (and easy to understand), and
you can learn the algebraic techniques to prove them afterwards. The only algebraic prerequisite
needed for this class is some basic knowledge on groups, rings, fields, and vector spaces as e. g.
taught in the Algebraic Structures and Foundations of Mathematics courses [G1, G2].
1.
1.
Affine Varieties
Affine Varieties
As explained in the introduction, the goal of algebraic geometry is to study solutions of polynomial
equations in several variables over a fixed ground field. So let us now make the corresponding
definitions.
Convention 1.1. Throughout these notes, K will always denote a fixed base field (which we will
require to be algebraically closed after our discussion of Hilberts Nullstellensatz in Proposition
1.17). Rings are always assumed to be commutative with a multiplicative unit 1. By K[x1 , . . . , xn ] we
will denote the polynomial ring in n variables x1 , . . . , xn over K, i. e. the ring of finite formal sums
ai1 ,...,in x1i1 xnin
i1 ,...,in N
with all ai1 ,...,in K (see e. g. [G1, Chapter 9] how this concept of formal sums can be defined in a
mathematically rigorous way).
Definition 1.2 (Affine varieties).
(a) We call
An := AnK := {(c1 , . . . , cn ) : ci K for i = 1, . . . , n}
the affine n-space over K.
Note that AnK is just K n as a set. It is customary to use two different notations here since
K n is also a K-vector space and a ring. We will usually use the notation AnK if we want to
ignore these additional structures: for example, addition and scalar multiplication are defined
on K n , but not on AnK . The affine space AnK will be the ambient space for our zero loci of
polynomials below.
(b) For a polynomial
f=
K[x1 , . . . , xn ]
i1 ,...,in N
and a point c = (c1 , . . . , cn ) An we define the value of f at c to be

f (c) =
ai1 ,...,in ci11 cinn
K.
i1 ,...,in N
If there is no risk of confusion we will often denote a point in An by the same letter x as we
used for the formal variables, writing f K[x1 , . . . , xn ] for the polynomial and f (x) for its
value at a point x AnK .
(c) For a subset S K[x1 , . . . , xn ] of polynomials we call
V (S) := {x An : f (x) = 0 for all f S}
An
the (affine) zero locus of S. Subsets of An of this form are called (affine) varieties. If
S = ( f1 , . . . , fk ) is a finite set, we will write V (S) = V ({ f1 , . . . , fk }) also as V ( f1 , . . . , fk ).
Remark 1.3. Some authors refer to zero loci of polynomials in An as in Definition 1.2 (c) as (affine)
algebraic sets, reserving the name affine variety for such zero loci that are in addition irreducible
(a concept that we will introduce in Definition 2.6 (a)).
Example 1.4. Here are some simple examples of affine varieties:
(a) Affine n-space itself is an affine variety, since An = V (0). Similarly, the empty set 0/ = V (1)
is an affine variety.
(b) Any single point in An is an affine variety: we have (c1 , . . . , cn ) = V (x1 c1 , . . . , xn cn ).
10
Andreas Gathmann
(c) Linear subspaces of An = K n are affine varieties.

(d) If X An and Y Am are affine varieties then so is the product X Y An Am = An+m .
(e) All examples from the introduction in Chapter 0 are affine varieties: e. g. the curves of
Examples 0.1 and 0.3, and the cubic surface of Example 0.9.
Remark 1.5 (Affine varieties are zero loci of ideals). Let f and g be polynomials that vanish on a
certain subset X An . Then f + g and h f for any polynomial h clearly vanish on X as well. This
means that the set S K[x1 , . . . , xn ] defining an affine variety X = V (S) is certainly not unique: for
any f , g S and any polynomial h we can add f + g and h f to S without changing its zero locus. In
other words, if
I = (S) = {h1 f1 + + hm fm : m N, f1 , . . . , fm S, h1 , . . . , hm K[x1 , . . . , xn ]}
is the ideal generated by S, then V (I) = V (S). Hence any affine variety in An can be written as the
zero locus of an ideal in K[x1 , . . . , xn ].
Example 1.6 (Affine varieties in A1 ). Let X be an affine variety in A1 . By Remark 1.5 we can then
write X = V (I) for an ideal I E K[x]. But K[x] is a principal ideal domain [G1, Example 10.39].
Hence we have I = ( f ) for some polynomial f K[x], and thus X = V ( f ).
As zero loci of non-zero polynomials in one variable are always finite, this means that any affine
variety in A1 not equal to A1 itself must be a finite set. Conversely, any finite subset {a1 , . . . , an } =
V ((x a1 ) (x an )) of A1 is an affine variety, and thus we conclude that the affine varieties in
A1 are exactly the finite sets and A1 itself.
Unfortunately, for more than one variable we cannot use a similar argument to classify the affine
varieties in An as the multivariate polynomial rings K[x1 , . . . , xn ] are not principal ideal domains.
However, we still have the following result that we will borrow from commutative algebra.
Proposition 1.7 (Hilberts Basis Theorem [G5, Proposition 7.13 and Remark 7.15]). Every ideal
in the polynomial ring K[x1 , . . . , xn ] can be generated by finitely many elements.
Remark 1.8 (Affine varieties are zero loci of finitely many polynomials). Let X = V (S) be an affine
variety. Then the ideal generated by S can be written as (S) = ( f1 , . . . , fm ) for some f1 , . . . , fm S
by Proposition 1.7, and hence X = V (S) = V ( f1 , . . . , fm ) by Remark 1.5. So every affine variety is
the zero locus of finitely many polynomials.
Exercise 1.9. Prove that every affine variety X An consisting of only finitely many points can be
written as the zero locus of n polynomials.
(Hint: interpolation.)
There is another reason why Remark 1.5 is important: it is in some sense the basis of algebraic
geometry since it relates geometric objects (affine varieties) to algebraic objects (ideals). In fact,
it will be the main goal of this first chapter to make this correspondence precise. We have already
assigned affine varieties to ideals in Definition 1.2 (c) and Remark 1.5, so let us now introduce an
operation that does the opposite job.
Definition 1.10 (Ideal of a subset of An ). Let X An be any subset. Then
I(X) := { f K[x1 , . . . , xn ] : f (x) = 0 for all x X}
is called the ideal of X (note that this is indeed an ideal by Remark 1.5).
Example 1.11 (Ideal of a point). Let a = (a1 , . . . , an ) AnK be a point. Then the ideal of the one-point
set {a} is I(a) := I({a}) = (x1 a1 , . . . , xn an ):
If f I(a) then f (a) = 0. This means that replacing each xi by ai in f gives zero, i. e. that f
is zero modulo (x1 a1 , . . . , xn an ). Hence f (x1 a1 , . . . , xn an ).
If f (x1 a1 , . . . , xn an ) then f = ni=1 (xi ai ) fi for some f1 , . . . , fn K[x1 , . . . , xn ], and
so certainly f (a) = 0, i. e. f I(a).
1.
Affine Varieties
11
We have now constructed operations

{affine varieties in An }
X
V (J)
{ideals in K[x1 , . . . , xn ]}
7 I(X)
7 J
and should check whether they actually give a bijective correspondence between ideals and affine
varieties. The following lemma tells us the positive results in this direction.
Lemma 1.12. Let S and
S0
be subsets of K[x1 , . . . , xn ], and let X and
X0
be subsets of
An .
(a) If X X 0 then I(X 0 ) I(X).

If S S0 then V (S0 ) V (S).
We say that the operations V ( ) and I( ) reverse inclusions.
(b) X V (I(X)) and S I(V (S)).
(c) If X is an affine variety then V (I(X)) = X.
Proof.
(a) Let X X 0 . If f I(X 0 ), i. e. f (x) = 0 for all x X 0 , then certainly also f (x) = 0 for all
x X, and hence f I(X). The second statement follows analogously.
(b) Let x X. Then f (x) = 0 for every f I(X), and thus by definition we have x V (I(X)).
Again, the second inclusion follows in the same way.
(c) By (b) it suffices to prove . As X is an affine variety we can write X = V (S) for some
S K[x1 , . . . , xn ]. Then S I(V (S)) by (b), and thus V (S) V (I(V (S))) by (a). Replacing
V (S) by X again now gives the required inclusion.

By this lemma, the only thing left that would be needed for a bijective correspondence between
ideals and affine varieties would be I(V (J)) J for any ideal J (so that then I(V (J)) = J by part
(b)). Unfortunately, the following example shows that there are two reasons why this is not true in
general.
Example 1.13 (The inclusion J I(V (J)) is strict in general).
(a) Let J E C[x] be a non-zero ideal. As C[x] is a principal ideal domain [G1, Example 10.39]
and C is algebraically closed, we must have

J = (x a1 )k1 (x an )kn
for some n N, distinct a1 , . . . , an C, and k1 , . . . , kn N>0 . Obviously, the zero locus of
this ideal in A1 is V (J) = {a1 , . . . , an }. The polynomials vanishing on this set are precisely
those that contain each factor x ai for i = 1, . . . , n at least once, i. e. we have

I(V (J)) = (x a1 ) (x an ) .
If at least one of the numbers k1 , . . . , kn is greater than 1, this is a bigger ideal than J. In
other words, the zero locus of an ideal does not see powers of polynomials: as a power f k of
a polynomial f has the same zero locus as f itself, the information about this power is lost
when applying the operation I(V ( )).
(b) The situation is even worse for ground fields that are not algebraically closed: the ideal
J = (x2 + 1) E R[x] has an empty zero locus in A1 , and so we get I(V (J)) = I(0)
/ = R[x].
So in this case the complete information on the ideal J is lost when applying the operation
I(V ( )).
To overcome the first of these problems, we just have to restrict our attention to ideals with the
property that they contain a polynomial f whenever they contain a power f k of it. The following
definition accomplishes this.
Definition 1.14 (Radicals and radical ideals). Let I be an ideal in a ring R.
01
12
Andreas Gathmann
(a) We call
I := { f R : f n I for some n N}
the radical of I.
(b) The ideal I is said to be a radical ideal if
I = I.
Remark 1.15. Again let I be an ideal in a ring R.
(a) The radical I of I is always an ideal:
We have 0 I, since 0 I.
If f , g I, i. e. f n I and gm I for some n, m N, then

n+m
n + m k n+mk
n+m
( f + g)
=
f g
k
k=0
is again an element of I, since in each summand we must have that the power of f is at
least n (in whichcase f k I) or the power of g is at least m (in which case gn+mk I).
Hence f + g I.
If h R and f I, i. e. f n I for some n N, then (h f )n = hn f n I, and hence

h f I.
Moreover, it is obvious that

always take n = 1 in Definition 1.14 (a)). Hence
I I (we can
I is radical if and only if I I, i. e. if f n I for some n N implies f I.
(b) As expected from the terminology, the radical of I is a radical ideal:

if f n I for some
f R and n N then ( f n )m = f nm I for some m N, and hence f I.
(c) If I is the ideal of an affine variety X then I is radical: if f I then f k vanishes on X, hence
f vanishes on X and we also have f I.
Example 1.16. Continuing Example 1.13 (a), the radical of the ideal

J = (x a1 )k1 (x an )kn
E C[x]
consists of all polynomials f C[x] such that (x a1 )k1 (x an )kn divides f k for large enough
k. This is obviously the set of all polynomials containing each factor x ai for i = 1, . . . , n at least
once, i. e. we have

J = (x a1 ) (x an ) .
One should note however that the explicit computation of radicals is in general hard and requires
algorithms of computer algebra.
In our example at hand we therefore see that I(V (J)) = J, resp. that I(V (J)) = J if J is radical.
In fact, this holds in general for ideals in polynomial rings over algebraically closed fields. This
statement is usually referred to as Hilberts Nullstellensatz (theorem of the zeroes); it is another
fact that we will quote here from commutative algebra.
Proposition 1.17 (Hilberts Nullstellensatz [G5, Corollary 10.14]). Let
K be an algebraically
closed field. Then for every ideal J E K[x1 , . . . , xn ] we have I(V (J)) = J. In particular, there is
an inclusion-reversing one-to-one correspondence
{affine varieties in An }
X
V (J)
{radical ideals in K[x1 , . . . , xn ]}
7 I(X)
7 J.
Proof. The main statement I(V (J)) = J is proven in [G5, Corollary 10.14]. The correspondence
then follows from what we have already seen:
I( ) maps affine varieties to radical ideals by Remark 1.15 (c);
we have V (I(X)) = X for any affine variety X by Lemma 1.12 (c) and I(V (J)) = J for any
radical ideal J by our main statement;
1.
Affine Varieties
13
the correspondence reverses inclusions by Lemma 1.12 (a).
As we have already mentioned, this result is absolutely central for algebraic geometry since it allows
us to translate geometric objects into algebraic ones. Note however that the introduction of radical
ideals allowed us to solve Problem (a) in Example 1.13, but not Problem (b): for ground fields that
are not algebraically closed the statement of Proposition 1.17 is clearly false since e. g. the ideal
J = (x2 + 1) E R[x] is radical but has an empty zero locus, so that I(V (J)) = R[x] 6= (x2 + 1) = J.
Let us therefore agree:
From now on, our ground field K will always be assumed to be algebraically closed.
Remark 1.18.
(a) Let J E K[x1 , . . . , xn ] be an ideal in the polynomial ring (over an algebraically closed field).
If
we would have
J 6= K[x1 , . . . , xn ] then J has a zero, i. e. V (J) is non-empty: otherwise
J = I(V (J)) = I(0)
/ = K[x1 , . . . , xn ] by Proposition 1.17, which means 1 J and gives us
the contradiction 1 J. This statement can be thought of as a generalization of the algebraic
closure property that a non-constant univariate polynomial has a zero. It is the origin of the
name Nullstellensatz for Proposition 1.17.
(b) Another easy consequence of Proposition 1.17 is that polynomials and polynomial functions
on An agree: if f , g K[x1 , . . . , xn ] are two polynomials defining the same function on An ,
i. e. such that f (x) = g(x) for all x An , then
p
f g I(An ) = I(V (0)) = (0) = (0)
and hence f = g in K[x1 , . . . , xn ]. So K[x1 , . . . , xn ] can be thought of as the ring of polynomial
functions on An . Note that this is false for general fields, since e. g. the polynomial x2 + x
Z2 [x] defines the zero function on A1Z2 , although it is not the zero polynomial.
More generally, if X is an affine variety then two polynomials f , g K[x1 , . . . , xn ] define the
same polynomial function on X, i. e. f (x) = g(x) for all x X, if and only if f g I(X).
So the quotient ring K[x1 , . . . , xn ]/I(X) can be thought of as the ring of polynomial functions
on X. Let us make this into a precise definition.
Definition 1.19 (Polynomial functions and coordinate rings). Let X An be an affine variety. A
polynomial function on X is a map X K that is of the form x 7 f (x) for some f K[x1 , . . . , xn ].
By Remark 1.18 (b) the ring of all polynomial functions on X is just the quotient ring
A(X) := K[x1 , . . . , xn ]/I(X).
It is usually called the coordinate ring of the affine variety X.
According to this definition, we can think of the elements of A(X) in the following both as functions
on X and as elements of the quotient ring K[x1 , . . . , xn ]/I(X). We can use this ring to define the
concepts introduced so far in a relative setting, i. e. consider zero loci of ideals in A(Y ) and varieties
contained in Y for a fixed ambient affine variety Y that is not necessarily An :
Construction 1.20 (Relative version of the correspondence between varieties and radical ideals).
Let Y An be an affine variety. The following two constructions are then completely analogous to
those in Definitions 1.2 (c) and 1.10:
(a) For a subset S A(Y ) of polynomial functions on Y we define its zero locus as
V (S) := VY (S) := {x Y : f (x) = 0 for all f S}
Y.
The subsets that are of this form are obviously precisely the affine varieties contained in X.
They are called affine subvarieties of Y .
(b) For a subset X Y the ideal of X in Y is defined to be
I(X) := IY (X) := { f A(Y ) : f (x) = 0 for all x X}
E A(Y ).
14
Andreas Gathmann
With the same arguments as above, all results considered so far then hold in this relative setting as
well. Let us summarize them here again:
Proposition 1.21. Let Y be an affine variety in An .
(a) (Hilberts Basis Theorem) Every ideal in A(Y ) can be generated by finitely many elements.
(b) (Hilberts Nullstellensatz) For any ideal J E A(Y ) we have IY (VY (J)) = J. In particular,
there is an inclusion-reversing one-to-one correspondence
{affine subvarieties of Y }
X
VY (J)
{radical ideals in A(Y )}
7 IY (X)
7 J.
(c) For a subvariety X of Y we have A(X)

= A(Y )/IY (X).
Proof. As in our earlier version, the proof of (a) is covered by [G5, Proposition 7.13 and Remark
7.15], the proof of (b) by [G5, Corollary 10.14] and Proposition 1.17. The statement (c) follows in
the same way as in Remark 1.18 (b).

Exercise 1.22. Determine the radical of the ideal (x13 x26 , x1 x2 x23 ) E C[x1 , x2 ].
Exercise 1.23. Let X be an affine variety. Show that the coordinate ring A(X) is a field if and only
if X is a single point.
In the rest of this chapter we want to study the basic properties of the operations V ( ) and I( ).
Lemma 1.24 (Properties of V ( )). Let X be an affine variety.
(a) If J is any index set and {Si : i J} a family of subsets of A(X) then
in X.
iJ V (Si ) = V (
iJ Si )
(b) For S1 , S2 A(X) we have V (S1 ) V (S2 ) = V (S1 S2 ) in X, where as usual we set S1 S2 :=
{ f g : f S1 , g S2 }.
In particular, arbitrary intersections and finite unions of affine subvarieties of X are again affine
subvarieties of X.
Proof.
(a) We have x iJ VS(Si ) if and only if f (x) = 0 for all f Si for all i J, which is the case if
and only if x V ( iJ Si ).
T
(b) If x V (S1 ) V (S2 ) then f (x) = 0 for all f S1 or g(x) = 0 for all g S2 . In any case
this means that ( f g)(x) = 0 for all f S1 and g S2 , i. e. that x V (S1 S2 ).
If x
/ V (S1 ) V (S2 ), i. e. x
/ V (S1 ) and x
/ V (S2 ), then there are f S1 and g S2
with f (x) 6= 0 and g(x) 6= 0. Then ( f g)(x) 6= 0, and hence x
/ V (S1 S2 ).

Remark 1.25 (Ideal-theoretic version of the properties of V ( )). If we want to consider zero loci
of ideals rather than of general subsets of A(X), then the properties of Lemma 1.24 take a slightly
different form. To see this, let J1 and J2 be any ideals in A(X).
(a) The ideal generated by J1 J2 is just the sum of ideals J1 + J2 = { f + g : f J1 , g J2 }. So
with Remark 1.5 the result of Lemma 1.24 (a) translates into
V (J1 ) V (J2 ) = V (J1 + J2 ).
(b) In the same way as in (a), Lemma 1.24 (b) implies that V (J1 ) V (J2 ) is equal to the zero
locus of the ideal generated by J1 J2 . Unfortunately, the usual convention is that for two
ideals J1 and J2 (instead of arbitrary sets) the notation J1 J2 denotes the ideal generated by
all products f g with f J1 and g J2 , which is called the product of the ideals J1 and J2
rather than the set of all such products f g itself. So we get
V (J1 ) V (J2 ) = V (J1 J2 )
1.
Affine Varieties
15
with this modified definition of the product J1 J2 .

(c) Another common operation on ideals is the intersection J1 J2 . In general, this ideal is
different from the ones considered above, but we can show that there is always the relation
J1 J2 = J1 J2 :
If f J1 J2
then f n J1 J2 for some n. This means that f 2n = f n f n J1 J2 , and
hence that f J1 J2 .
For f J1 J2 we have f n J1 J2 for some n. Then f n J1 J2 , and thus f J1 J2 .

By Proposition 1.21 (b) this means that I(V (J1 J2 )) = I(V (J1 J2 )), and hence by applying
V ( ) that
V (J1 J2 ) = V (J1 J2 ) = V (J1 ) V (J2 )
by (b).
Finally, for completeness let us also formulate the properties of Lemma 1.24 and Remark 1.25 in
terms of the operation I( ) rather than V ( ).
Lemma 1.26 (Properties of I( )). Let X be an affine variety, and let Y1 and Y2 be affine subvarieties
of X. Then:
p
(a) I(Y1 Y2 ) = I(Y1 ) + I(Y2 );
(b) I(Y1 Y2 ) = I(Y1 ) I(Y2 ).
Proof.
(a) We have
I(Y1 Y2 ) = I(V (I(Y1 )) V (I(Y2 ))) (Proposition 1.21 (b))
= I(V (I(Y1 ) + I(Y2 )))
p
= I(Y1 ) + I(Y2 ).
(Remark 1.25 (a))

(Proposition 1.21 (b))
(b) A polynomial function f A(X) is contained in I(Y1 Y2 ) if and only if f (x) = 0 for all
x Y1 and all x Y2 , which is the case if and only if f I(Y1 ) I(Y2 ).

Remark 1.27. Recall from Remark 1.15 (c) that ideals of affine varieties are always radical. So in
particular, Lemma 1.26 (b) shows that intersections of radical ideals in A(X) are again radical
which could of course also be checked directly. In contrast, sums of radical ideals are in general not
radical, and hence taking the radical in Lemma 1.26 (a) is really necessary.
In fact, there is also a geometric interpretation behind this fact. Consider
for example the affine varieties Y1 ,Y2 A1C with ideals I(Y1 ) = (x2 x12 )
and I(Y2 ) = (x2 ) whose real points are shown in the picture on the right.
Their intersection Y1 Y2 is obviously the origin with ideal I(Y1 Y2 ) =
(x1 , x2 ). But
I(Y1 ) + I(Y2 ) = (x2 x12 , x2 ) = (x12 , x2 )
is not a radical ideal; only its radical is equal to I(Y1 Y2 ) = (x1 , x2 ).
Y1
Y2
Y1 Y2
The geometric meaning of the non-radical ideal I(Y1 ) + I(Y2 ) = (x12 , x2 ) is that Y1 and Y2 are tangent
at the intersection point: if we consider the function x2 x12 defining Y1 on the x1 -axis Y2 (where it is
equal to x12 ) we see that it vanishes to order 2 at the origin. This means that Y1 and Y2 share the x1 axis as common tangent direction, so that the intersection Y1 Y2 can be thought of as extending to
an infinitesimally small amount in the x1 -direction, and we can consider Y1 and Y2 as intersecting
with multiplicity 2 at the origin. We will see later in Definition 12.23 (b) how such intersection
multiplicities can be defined rigorously.
02
16
2.
Andreas Gathmann
The Zariski Topology
In this chapter we will define a topology on an affine variety X, i. e. a notion of open and closed
subsets of X. We will see that many properties of X can be expressed purely in terms of this topology,
e. g. its dimension or the question whether it consists of several components. The advantage of this
is that these concepts can then easily be reused later in Chapter 5 when we consider more general
varieties which are still topological spaces, but arise in a slightly different way.
Compared to e. g. the standard topology on subsets of real vector spaces, the properties of our topology on affine varieties will be very unusual. Consequently, most concepts and results covered in a
standard introductory course on topology will be trivial or useless in our case, so that we will only
need the very first definitions of general topology. Let us quickly review them here.
Remark 2.1 (Topologies). A topology on a set X is given by declaring some subsets of X to be
closed, such that the following properties hold:
(a) the empty set 0/ and the whole space X are closed;
(b) arbitrary intersections of closed sets are closed;
(c) finite unions of closed sets are closed.
Given such a topology on X, a subset U of X is then called open if its complement X\U is closed.
The closure A of a subset A X is defined to be the smallest closed subset containing A, or more
precisely the intersection of all closed subsets containing A (which is closed again by (b)).
A topology on X induces a subspace topology on any subset A X by declaring the subsets of A to
be closed that are of the form A Y for a closed subset Y of X (or equivalently the subsets of A to be
open that are of the form A U for an open subset U of X). Subsets of topological spaces will always
be equipped with this subspace topology unless stated otherwise. Note that if A is closed itself then
the closed subsets of A in the subspace topology are exactly the closed subsets of X contained in A;
if A is open then the open subsets of A in the subspace topology are exactly the open subsets of X
contained in A.
A map : X Y between topological spaces is called continuous if inverse images of closed subsets
of Y under are closed in X, or equivalently if inverse images of open subsets are open.
Note that the standard definition of closed subsets in Rn (or more generally in metric spaces) that you
know from real analysis satisfies the conditions (a), (b), and (c), and leads with the above definitions
to the well-known notions of open subsets, closures, and continuous functions.
With these preparations we can now define the standard topology used in algebraic geometry.
Definition 2.2 (Zariski topology). Let X be an affine variety. We define the Zariski topology on X
to be the topology whose closed sets are exactly the affine subvarieties of X, i. e. the subsets of the
form V (S) for some S A(X). Note that this in fact a topology by Example 1.4 (a) and Lemma 1.24.
Unless stated otherwise, topological notions for affine varieties (and their subsets, using the subspace
topology of Remark 2.1) will always be understood with respect to this topology.
Remark 2.3. Let X An be an affine variety. Then we have just defined two topologies on X:
(a) the Zariski topology on X, whose closed subsets are the affine subvarieties of X; and
(b) the subspace topology of X in An , whose closed subsets are the sets of the form X Y , with
Y a variety in An .
These two topologies agree, since the subvarieties of X are precisely the affine varieties contained
in X and the intersection of two affine varieties is again an affine variety. Hence it will not lead to
confusion if we consider both these topologies to be the standard topology on X.
2.
17
Exercise 2.4. Let X An be an arbitrary subset. Prove that V (I(X)) = X.

Example 2.5 (Topologies on complex varieties). Compared to the classical metric topology in the
case of the ground field C, the Zariski topology is certainly unusual:
(a) The metric unit ball A = {x A1C : |x| 1} in A1C is clearly closed in the classical topology,
but not in the Zariski topology. In fact, by Example 1.6 the Zariski-closed subsets of A1 are
only the finite sets and A1 itself. In particular, the closure of A in the Zariski topology is all
of A1 .
Intuitively, we can say that the closed subsets in the Zariski topology are very small, and
hence that the open subsets are very big (see also Remark 2.18). Any Zariski-closed subset
is also closed in the classical topology (since it is given by equations among polynomial
functions, which are continuous in the classical topology), but as the above example shows
only very few closed subsets in the classical topology are also Zariski-closed.
(b) Let : A1 A1 be any bijective map. Then is continuous in the Zariski topology, since
inverse images of finite subsets of A1 under are finite.
This statement is essentially useless however, as we will not define morphisms of affine
varieties as just being continuous maps with respect to the Zariski topology. In fact, this
example gives us a strong hint that we should not do so.
(c) In general topology there is a notion of a product topology: if X and Y are topological spaces
then X Y has a natural structure of a topological space by saying that a subset of X Y is
open if and only if it is a union of products Ui Vi for open subsets Ui X and Vi Y with
i in an arbitrary index set.
With this construction, note however that the Zariski topology of an affine product variety
X Y is not the product topology: e. g. the subset V (x1 x2 ) = {(a, a) : a K} A2 is
closed in the Zariski topology, but not in the product topology of A1 A1 . In fact, we will
see in Proposition 4.10 that the Zariski topology is the correct one, whereas the product
topology is useless in algebraic geometry.
But let us now start with the discussion of the topological concepts that are
actually useful in the Zariski topology. The first ones concern components of an
affine variety: the affine variety X = V (x1 x2 ) A2 as in the picture on the right
can be written as the union of the two coordinate axes X1 = V (x2 ) and X2 =
V (x1 ), which are themselves affine varieties. However, X1 and X2 cannot be
decomposed further into finite unions of smaller affine varieties. The following
notion generalizes this idea.
X2
X1
X = X1 X2
Definition 2.6 (Irreducible and connected spaces). Let X be a topological space.

(a) We say that X is reducible if it can be written as X = X1 X2 for closed subsets X1 , X2 ( X.
Otherwise X is called irreducible.
(b) The space X is called disconnected if it can be written as X = X1 X2 for closed subsets
X1 , X2 ( X with X1 X2 = 0.
/ Otherwise X is called connected.
Remark 2.7. Although we have given this definition for arbitrary topological spaces, you will usually want to apply the notion of irreducibility only in the Zariski topology. For example, in the
classical topology, the complex plane A1C is reducible because it can be written e. g. as the union of
closed subsets as
A1C = {z C : |z| 1} {z C : |z| 1}.
In the Zariski topology however, A1 is irreducible by Example 1.6 (as it should be).
In contrast, the notion of connectedness can be used in the usual topology too and does mean there
what you think it should mean.
In the Zariski topology, the algebraic characterization of the irreducibility and connectedness of
affine varieties is the following.
18
Andreas Gathmann
Proposition 2.8. Let X be a disconnected affine variety, with X = X1 X2 for two disjoint closed
subsets X1 , X2 ( X. Then A(X)
= A(X1 ) A(X2 ).
Proof. By Proposition 1.21 (c) we have A(X1 )
= A(X)/I(X1 ) and A(X2 )
= A(X)/I(X2 ). Hence there
is a ring homomorphism
: A(X) A(X1 ) A(X2 ), f 7 ( f , f ).
We have to show that it is bijective.
is injective: If ( f , f ) = (0, 0) then f I(X1 ) I(X2 ) = I(X1 X2 ) = I(X) = (0) by Lemma
1.26 (b).
is surjective: Let ( f1 , f2 ) A(X1 ) A(X2 ). Note that
p
A(X) = I(0)
/ = I(X1 X2 ) = I(X1 ) + I(X2 )
p
by Lemma 1.26 (a). Thus 1 I(X1 ) + I(X2 ), and hence 1 I(X1 ) + I(X2 ), which means
I(X1 )+I(X2 ) = A(X). We can therefore find g1 I(X1 ) and g2 I(X2 ) with f1 f2 = g1 g2 ,
so that f1 g1 = f2 g2 . This latter element of A(X) then maps to ( f1 , f2 ) under .

Proposition 2.9. An affine variety X is irreducible if and only if A(X) is an integral domain.
Proof. : Assume that A(X) is not an integral domain, i. e. there are non-zero f1 , f2 A(X) with
f1 f2 = 0. Then X1 = V ( f1 ) and X2 = V ( f2 ) are closed, not equal to X (since f1 and f2 are non-zero),
and X1 X2 = V ( f1 ) V ( f2 ) = V ( f1 f2 ) = V (0) = X. Hence X is reducible.
: Assume that X is reducible, with X = X1 X2 for closed subsets X1 , X2 ( X. By Proposition
1.21 (b) this means that I(Xi ) 6= (0) for i = 1, 2, and so we can choose non-zero fi I(Xi ). Then f1 f2
vanishes on X1 X2 = X. Hence f1 f2 = 0 A(X), i. e. A(X) is not an integral domain.

Remark 2.10. If X is an affine subvariety of an affine variety Y we know by Proposition 1.21 (c)
that A(X) = A(Y )/I(X). So A(X) is an integral domain, i. e. X is irreducible, if and only if for all
f , g A(Y ) the relation f g I(X) implies f I(X) or g I(X). In commutative algebra, ideals
with this property are called prime ideals. So in other words, in the one-to-one correspondence
of Proposition 1.21 (b) between affine subvarieties of Y and radical ideals in A(Y ) the irreducible
subvarieties correspond exactly to prime ideals.
Example 2.11.
(a) A finite affine variety is irreducible if and only if it is connected: namely if and only if it
contains at most one point.
(b) Any irreducible space is connected.
(c) The affine space An is irreducible (and thus connected) by Proposition 2.9 since its coordinate ring A(An ) = K[x1 , . . . , xn ] is an integral domain. More generally, this holds for any
affine variety given by linear equations, since again its coordinate ring is isomorphic to a
polynomial ring.
(d) The union X = V (x1 x2 ) A2 of the two coordinate axes X1 = V (x2 ) and X2 = V (x1 ) is not
irreducible, since X = X1 X2 . But X1 and X2 themselves are irreducible by (c). Hence we
have decomposed X into a union of two irreducible spaces.
As already announced, we now want to see that such a decomposition into finitely many irreducible
spaces is possible for any affine variety. In fact, this works for a much larger class of topological
spaces, the so-called Noetherian spaces:
Definition 2.12 (Noetherian topological spaces). A topological space X is called Noetherian if there
is no infinite strictly decreasing chain
X0 ) X1 ) X2 )
of closed subsets of X.
2.
19
Lemma 2.13. Any affine variety is a Noetherian topological space.

Proof. Let X be an affine variety. Assume that there is an infinite chain X0 ) X1 ) X2 ) of
subvarieties of X. By Proposition 1.21 (b) there is then a corresponding infinite chain
I(X0 ) ( I(X1 ) ( I(X2 ) (
of ideals in A(X). It is checked immediately that the union I :=
n=0 I(Xi ) is then an ideal as well
[G1, Exercise 10.46 (a)]. By Proposition 1.21 (a) it is finitely generated, i. e. we have I = ( f1 , . . . , fn )
for some f1 , . . . , fn I. All these polynomials have to lie in one of the ideals I(Xm ) and in fact
in the same one since these ideals form a chain. But then we have I = ( f1 , . . . , fn ) I(Xm ) ( I, a
contradiction.

S
Remark 2.14 (Subspaces of Noetherian spaces are Noetherian). Let A be a subset of a Noetherian
topological space X. Then A is also Noetherian: otherwise we would have an infinite strictly descending chain of closed subsets of A, which by definition of the subspace topology we can write
as
A Y0 ) A Y1 ) A Y2 )
for closed subsets Y0 ,Y1 ,Y2 , . . . of X. Then
Y0 Y0 Y1 Y0 Y1 Y2
is an infinite decreasing chain of closed subsets of X. In fact, in contradiction to our assumption it
is also strictly decreasing, since Y0 Yk = Y0 Yk+1 for some k N would imply A Yk =
A Yk+1 by intersecting with A.
Combining Lemma 2.13 with Remark 2.14 we therefore see that any subset of an affine variety
is a Noetherian topological space. In fact, all topological spaces occurring in this class will be
Noetherian, and thus we can safely restrict our attention to this class of spaces.
Proposition 2.15 (Irreducible decomposition of Noetherian spaces). Every Noetherian topological
space X can be written as a finite union X = X1 Xr of irreducible closed subsets. If one
assumes that Xi 6 X j for all i 6= j, then X1 , . . . , Xr are unique (up to permutation). They are called
the irreducible components of X.
Proof. To prove existence, assume that there is a topological space X for which the statement is
false. In particular, X is reducible, hence X = X1 X10 as in Definition 2.6 (a). Moreover, the
statement of the proposition must be false for at least one of these two subsets, say X1 . Continuing
this construction, one arrives at an infinite chain X ) X1 ) X2 ) of closed subsets, which is a
contradiction as X is Noetherian.
To show uniqueness, assume that we have two decompositions
X = X1 Xr = X10 Xs0 .
()
Then for any fixed i {1, . . . , r} we have Xi j X j0 , so Xi = j (Xi X j0 ). But Xi is irreducible, and
so we must have Xi = Xi X j0 , i. e. Xi X j0 for some j. In the same way we conclude that X j0 Xk
for some k, so that Xi X j0 Xk . By assumption this is only possible for i = k, and consequently
Xi = X j0 . Hence every set appearing on the left side of () also appears on the right side (and vice
versa), which means that the two decompositions agree.

S
Remark 2.16 (Computation of irreducible decompositions). In general, the actual computation of

the irreducible decomposition of an affine variety is quite difficult and requires advanced algorithmic
methods of computer algebra. In fact, the corresponding question in commutative algebra is to find
the isolated primes of a so-called primary decomposition of an ideal [G5, Chapter 8]. But in simple
cases the irreducible decomposition might be easy to spot geometrically, as e. g. in Example 2.11
(d).
Exercise 2.17. Find the irreducible components of the affine variety V (x1 x2 x3 , x1 x3 x22 ) A3C .
03
20
Andreas Gathmann
Remark 2.18 (Open subsets of irreducible spaces are dense). We have already seen in Example 2.5
(a) that open subsets tend to be very big in the Zariski topology. Here are two precise statements
along these lines. Let X be an irreducible topological space, and let U and U 0 be non-empty open
subsets of X. Then:
(a) The intersection U U 0 is never empty. In fact, by taking complements this is just equivalent
to saying that the union of the two proper closed subsets X\U and X\U 0 is not equal to X,
i. e. to the definition of irreducibility.
(b) The closure U of U is all of X one says that U is dense in X. This is easily seen: if
Y X is any closed subset containing U then X = Y (X\U), and since X is irreducible and
X\U 6= X we must have Y = X.
Exercise 2.19. Let A be a subset of a topological space X. Prove:
(a) If Y A is closed in the subspace topology of A then Y A = Y .
(b) A is irreducible if and only if A is irreducible.
Exercise 2.20. Let {Ui : i I} be an open cover of a topological space X, and assume that Ui U j 6= 0/
for all i, j I. Show:
(a) If Ui is connected for all i I then X is connected.
(b) If Ui is irreducible for all i I then X is irreducible.
Exercise 2.21. Let f : X Y be a continuous map of topological spaces. Prove:
(a) If X is irreducible then so is f (X).
(b) If X is connected then so is f (X).
Exercise 2.22. Let X An and Y Am be irreducible affine varieties. Prove that the coordinate
ring A(X Y ) of their product is an integral domain, and hence that X Y is irreducible as well.
An important application of the notion of irreducibility is the definition of the dimension of an affine
variety (or more generally of a topological space but as with our other concepts above you will
only want to apply it to the Zariski topology). Of course, at least in the case of complex varieties we
have a geometric idea what the dimension of an affine variety should be: the number of coordinates
that you need to describe X locally around any point. Although there are algebraic definitions of
dimension that mimic this intuitive one [G5, Proposition 11.31], the standard definition of dimension
that we will give here uses only the language of topological spaces. Finally, all these definitions are
of course equivalent and describe the intuitive notion of dimension, but it is actually quite hard to
prove this rigorously.
The idea to construct the dimension in algebraic geometry using the Zariski topology is rather simple: if X is an irreducible topological space, then any closed subset of X not equal to X should have
smaller dimension. The resulting definition is the following.
Definition 2.23 (Dimension and codimension). Let X be a non-empty topological space.
(a) The dimension dim X N {} is the supremum over all n N such that there is a chain
0/ 6= Y0 ( Y1 ( ( Yn X
of length n of irreducible closed subsets Y1 , . . . ,Yn of X.
(b) If Y X is a non-empty irreducible closed subset of X the codimension codimX Y of Y in X
is again the supremum over all n such that there is a chain
Y Y0 ( Y1 ( ( Yn X
of irreducible closed subsets Y1 , . . . ,Yn of X containing Y .
To avoid confusion, we will always denote the dimension of a K-vector space V by dimK V , leaving
the notation dim X (without an index) for the dimension of a topological space X as above.
2.
21
According to the above idea, one should imagine each Yi as having dimension i in a maximal chain
as in Definition 2.23 (a), so that finally dim X = n. In the same way, each Yi in Definition 2.23 (b)
should have dimension i + dimY in a maximal chain, so that n = dim X dimY can be thought of as
the difference of the dimensions of X and Y .
Example 2.24.
(a) If X is a (non-empty) finite affine variety then dim X = 0. In fact, since points are closed in
X all subsets of X will be closed, and thus the only irreducible closed subsets of X are single
points. There are therefore only chains of length 0 of irreducible closed subsets of X.
(b) In contrast to (a), general finite topological spaces need not have dimension 0. For example,
the two-pointed topological space X = {a, b} whose closed subsets are exactly 0,
/ {a}, and
X has dimension 1 since {a} ( X is a chain of length 1 of irreducible closed subsets of X
(and there are certainly no longer ones).
However, this will not be of further importance for us since all topological spaces occurring
in this class will have the property that points are closed.
(c) By Example 1.6 the affine space A1 has dimension 1: maximal chains of irreducible closed
subsets of A1 are {a} ( A1 for any a A1 .
(d) It is easy to see that the affine space An for n N>0 has dimension at least n, since there is
certainly a chain
V (x1 , . . . , xn ) ( V (x2 , . . . , xn ) ( ( V (xn ) ( V (0) = An
of irreducible (linear) closed subsets of An of length n.
Of course, we would expect geometrically that the dimension of An is equal to n. Although this turns
out to be true, the proof of this result is unfortunately rather difficult and technical. It is given in the
Commutative Algebra class, where dimension is one of the major topics. In fact, our Definition
2.23 is easy to translate into commutative algebra: since irreducible closed subvarieties of an affine
variety X correspond exactly to prime ideals in A(X) by Remark 2.10, the dimension of X is the
supremum over all n such that there is a chain I0 ) I1 ) ) In of prime ideals in A(X) and this
can be studied algebraically.
Let us now quote the results on the dimension of affine varieties that we will use from commutative
algebra. They are all very intuitive: besides the statement that dim An = n they say that for irreducible
affine varieties the codimension of Y in X is in fact the difference of the dimensions of X and Y , and
that cutting down an irreducible affine variety by one non-trivial equation reduces the dimension by
exactly 1.
Proposition 2.25 (Properties of dimension). Let X and Y be non-empty irreducible affine varieties.
(a) We have dim(X Y ) = dim X + dimY . In particular, dim An = n.
(b) If Y X we have dim X = dimY + codimX Y . In particular, codimX {a} = dim X for every
point a X.
(c) If f A(X) is non-zero every irreducible component of V ( f ) has codimension 1 in X (and
hence dimension dim X 1 by (b)).
Proof. Statement (a) is [G5, Proposition 11.9 (a) and Exercise 11.33 (a)], (b) is [G5, Example 11.13
(a)], and (c) is [G5, Corollary 11.19].

Example 2.26. Let X = V (x2 x12 ) A2C be the affine variety whose real
points are shown in the picture on the right. Then we have as expected:
x2
(a) X is irreducible by Proposition 2.9 since its coordinate ring

A(X) = C[x1 , x2 ]/(x2 x12 )
= C[x1 ] is an integral domain.
(b) X has dimension 1 by Proposition 2.25 (c), since it is the zero locus
of one non-zero polynomial in the affine space A2 , and dim A2 = 2
by Proposition 2.25 (a).
x1
22
Andreas Gathmann
Remark 2.27 (Infinite-dimensional spaces). One might be tempted to think that the finiteness
condition of a Noetherian topological space X ensures that dim X is always finite. This is not
true however: if we equip the natural numbers X = N with the topology in which (except 0/ and
X) exactly the subsets Yn := {0, . . . , n} for n N are closed, then X is Noetherian, but has chains
Y0 ( Y1 ( ( Yn of non-empty irreducible closed subsets of arbitrary length.
However, Proposition 2.25 (a) together with the following exercise shows that this cannot happen
for arbitrary subsets of affine varieties. In fact, all topological spaces considered in this class will
have finite dimension.
Exercise 2.28. Let A be an arbitrary subset of a topological space X. Prove that dim A dim X.
Remark 2.29. Depending on where our chains of irreducible closed subvarieties end resp. start, we
can break up the supremum in Definition 2.23 into irreducible components or local contributions:
(a) If X = X1 Xr is the irreducible decomposition of a Noetherian topological space as in
Proposition 2.15, we have
dim X = max{dim X1 , . . . , dim Xr } :
Assume that dim X n, so that there is a chain Y0 ( ( Yn of non-empty irreducible
closed subvarieties of X. Then Yn = (Yn X1 ) (Yn Xr ) is a union of closed
subsets. So as Yn is irreducible we must have Yn = Yn Xi , and hence Yn Xi , for some
i. But then Y0 ( ( Yn is a chain of non-empty irreducible closed subsets in Xi , and
hence dim Xi n.
Let max{dim X1 , . . . , dim Xr } n. Then there is a chain of non-empty irreducible
closed subsets Y0 ( ( Yn in some Xi . This is also such a chain in X, and hence
dim X n.
So for many purposes it suffices to consider the dimension of irreducible spaces.
(b) We always have dim X = sup{codimX {a} : a X}:
If dim X n there is a chain Y0 ( ( Yn of non-empty irreducible closed subsets of
X. For any a Y0 this chain then shows that codimX {a} n.
If codimX {a} n for some a X there is a chain {a} Y0 ( ( Yn of non-empty
irreducible closed subsets of X, which also shows that dim X n.
The picture below illustrates these two equations: the affine variety X = V (x1 x3 , x2 x3 ) A3 is a
union of two irreducible components, a line V (x1 , x2 ) of dimension 1 and a plane V (x3 ) of dimension
2 (see Proposition 2.25 (a)). So by (a) we have dim X = 2 (with a maximal chain of length 2 as in
Definition 2.23 (a) given by Y0 ( Y1 ( Y2 ).
X1
X0
(
Y0
Y1
Y2
As for (b), the codimension of the point Y0 is 2, whereas the codimension of the point X0 is 1, as
illustrated by the chains in the picture. Note that this codimension of a point can be interpreted
2.
23
geometrically as the local dimension of X at this point. Hence Proposition 2.25 (b) can also be
interpreted as saying that the local dimension of an irreducible variety is the same at every point.
In practice, we will usually be concerned with affine varieties all of whose components have the same
dimension. These spaces have special names that we want to introduce now. Note however that (as
with the definition of a variety, see Remark 1.3) these terms are not used consistently throughout the
literature sometimes e. g. a curve is required to be irreducible, and sometimes it might be allowed
to have additional components of dimension less than 1.
Definition 2.30 (Pure-dimensional spaces).
(a) A Noetherian topological space X is said to be of pure dimension n if every irreducible
component of X has dimension n.
(b) An affine variety is called . . .
an affine curve if it is of pure dimension 1;
an affine surface if it is of pure dimension 2;
an affine hypersurface of an irreducible affine variety Y X if it is of pure dimension
dimY 1.
Exercise 2.31. Let X be the set of all 2 3 matrices over a field K that have rank at most 1, considered as a subset of A6 = Mat(2 3, K).
Show that X is an affine variety. Is it irreducible? What is its dimension?
Exercise 2.32. Show that the ideal I = (x1 x2 , x1 x3 , x2 x3 ) E C[x1 , x2 , x3 ] cannot be generated by fewer
than three elements. What is the zero locus of I?
Exercise 2.33. Let X be a topological space. Prove:
(a) If {Ui : i I} is an open cover of X then dim X = sup{dimUi : i I}.
(b) If X is an irreducible affine variety and U X a non-empty open subset then dim X = dimU.
Does this statement hold more generally for any irreducible topological space?
Exercise 2.34. Prove the following (maybe at first surprising) statements:
(a) Every affine variety in the real affine space AnR is the zero locus of one polynomial.
(b) Every Noetherian topological space is compact. In particular, every open subset of an affine
variety is compact in the Zariski topology. (Recall that by definition a topological space X
is compact if every open cover of X has a finite subcover.)
(c) The zero locus of a non-constant polynomial in C[x1 , x2 ] is never compact in the classical
topology of A2C = C2 .
(For those of you who know commutative algebra: can you prove that an affine variety over
C containing infinitely many points is never compact in the classical topology?)
24
3.
Andreas Gathmann
The Sheaf of Regular Functions
After having defined affine varieties, our next goal must be to say what kind of maps between them
we want to consider as morphisms, i. e. as nice maps preserving the structure of the variety. In
this chapter we will look at the easiest case of this: the so-called regular functions, i. e. maps to the
ground field K = A1 . They should be thought of as the analogue of continuous functions in topology,
differentiable functions in real analysis, or holomorphic functions in complex analysis.
So what kind of nice algebraic functions should we consider on an affine variety X? First of all,
as in the case of continuous or differentiable functions, we should not only aim for a definition of
functions on all of X, but also on an arbitrary open subset U of X. In contrast to the coordinate
ring A(X) of polynomial functions on the whole space X, this allows us to consider quotients gf of
polynomial functions f , g A(X) with f 6= 0 as well, since we can exclude the zero set V ( f ) of the
denominator from the domain of definition of the function.
But taking our functions to be quotients of polynomials turns out to be a little bit too restrictive. The
problem with this definition would be that it is not local : recall that the condition on a function to
be continuous or differentiable is local in the sense that it can be checked at every point, with the
whole function then being continuous or differentiable if it has this property at every point. Being a
quotient of polynomials however is not a condition of this type we would have to find one global
representation as a quotient of polynomials that is then valid at every point. Imposing such non-local
conditions is usually not a good thing to do, since it would be hard in practice to find the required
global representations of the functions.
The way out of this problem is to consider functions that are only locally quotients of polynomials,
i. e. functions : U K such that each point a U has a neighborhood in U in which = gf holds
for two polynomials f and g (that may depend on a). In fact, we will see in Example 3.5 that passing
from global to local quotients of polynomials really makes a difference. So let us now give the
corresponding formal definition of regular functions.
Definition 3.1 (Regular functions). Let X be an affine variety, and let U be an open subset of X. A
regular function on U is a map : U K with the following property: for every a U there are
polynomial functions f , g A(X) with f (x) 6= 0 and
(x) =
g(x)
f (x)
for all x in an open subset Ua with a Ua U. The set of all such regular functions on U will be
denoted OX (U).
Notation 3.2. We will usually write the condition (x) = g(x)
f (x) for all x Ua of Definition 3.1
g
simply as = f on Ua . This is certainly an intuitive notation that should not lead to any confusion. However, a word of warning in particular for those of you who know commutative algebra
already: this also means that (unless stated otherwise) the fraction gf of two elements of A(X) will
always denote the pointwise quotient of the two corresponding polynomial functions and not the
algebraic concept of an element in a localized ring as introduced later in Construction 3.12.
Remark 3.3 (OX (U) as a ring and K-algebra). It is obvious that the set OX (U) of regular functions
on an open subset U of an affine variety X is a ring with pointwise addition and multiplication.
However, it has an additional structure: it is also a K-vector space since we can multiply a regular
function pointwise with a fixed scalar in K. In algebraic terms, this means that OX (U) is a K-algebra,
which is defined as follows.
3.
25
Definition 3.4 (K-algebras [G5, Definition 1.23 and Remark 1.24]).

(a) A K-algebra is a ring R that is at the same time a K-vector space such that the ring multiplication is K-bilinear.
(b) For two K-algebras R and R0 a morphism (or K-algebra homomorphism) from R to R0 is a
map f : R R0 that is a ring homomorphism as well as a K-linear map.
Example 3.5 (Local 6= global quotients of polynomials). Consider the 3-dimensional affine variety
X = V (x1 x4 x2 x3 ) A4 and the open subset
U = X\V (x2 , x4 ) = {(x1 , x2 , x3 , x4 ) X : x2 6= 0 or x4 6= 0}
Then
(x
: U K, (x1 , x2 , x3 , x4 ) 7
x2
x3
x4
X.
if x2 6= 0,
()
if x4 6= 0
is a regular function on U: it is well-defined since the defining equation for X implies xx12 = xx43
whenever x2 6= 0 and x4 6= 0, and it is obviously locally a quotient of polynomials. But none of the
two representations in () as quotients of polynomials can be used on all of U, since the first one does
not work e. g. at the point (0, 0, 0, 1) U, whereas the second one does not work at (0, 1, 0, 0) U.
Algebraically, this is just the statement that A(X) is not a unique factorization domain [G5, Definition
8.1] because of the relation x1 x4 = x2 x3 .
In fact, one can show that there is also no other global representation of as a quotient of two
polynomials. We will not need this statement here, and so we do not prove it we should just keep
in mind that representations of regular functions as quotients of polynomials will in general not be
valid on the complete domain of definition of the function.
As a first result, let us prove the expected statement that zero loci of regular functions are always
closed in their domain of definition.
Lemma 3.6 (Zero loci of regular functions are closed). Let U be an open subset of an affine variety
X, and let OX (U) be a regular function on U. Then
V () := {x U : (x) = 0}
is closed in U.
Proof. By Definition 3.1 any point a U has an open neighborhood Ua in U on which =
some fa , ga A(X) (with fa nowhere zero on Ua ). So the set
ga
fa
for
{x Ua : (x) 6= 0} = Ua \V (ga )
is open in X, and hence so is their union over all a U, which is just U\V (). This means that V ()
is closed in U.

Remark 3.7 (Identity Theorem for regular functions). A simple but remarkable consequence of
Lemma 3.6 is the following: let U V be non-empty open subsets of an irreducible affine variety
X. If 1 , 2 OX (V ) are two regular functions on V that agree on U, then they must agree on all
of V : the locus V (1 2 ) where the two functions agree contains U and is closed in V by Lemma
3.6, hence it also contains the closure U of U in V . But V = X by Remark 2.18 (b), hence V is
irreducible by Exercise 2.19 (b), which again by Remark 2.18 (b) means that the closure of U in V
is V . Consequently, we have 1 = 2 on V .
Note that this statement is not really surprising since the open subsets in the Zariski topology are
so big: over the ground field C, for example, it is also true in the classical topology that the closure
of U in V is V , and hence the equation 1 = 2 on V already follows from 1 |U = 2 |U by the
(classical) continuity of 1 and 2 . The interesting fact here is that the very same statement holds in
complex analysis for holomorphic functions as well (or more generally, in real analysis for analytic
functions): two holomorphic functions on a (connected) open subset U Cn must be the same if
they agree on any smaller open subset V U. This is called the Identity Theorem for holomorphic
04
26
Andreas Gathmann
functions. In complex analysis this is a real theorem because the open subset V can be very small,
so the statement that the extension to U is unique is a lot more surprising than it is here in algebraic
geometry. Still this is an example of a theorem that is true in literally the same way in both algebraic
and complex geometry, although these two theories are quite different a priori. We will see another
case of this in Example 3.14.
Let us now go ahead and compute the K-algebras OX (U) in some cases. A particularly important
result in this direction can be obtained if U is the complement of the zero locus of a single polynomial
function f A(X). In this case it turns out that (in contrast to Example 3.5) the regular functions on
U can always be written with a single representation as a fraction, whose denominator is in addition
a power of f .
Definition 3.8 (Distinguished open subsets). For an affine variety X and a polynomial function
f A(X) on X we call
D( f ) := X\V ( f ) = {x X : f (x) 6= 0}
the distinguished open subset of f in X.
Remark 3.9. The distinguished open subsets of an affine variety X behave nicely with respect to
intersections and unions:
(a) For any f , g A(X) we have D( f )D(g) = D( f g), since f (x) 6= 0 and g(x) 6= 0 is equivalent
to ( f g)(x) 6= 0 for all x X. In particular, finite intersections of distinguished open subsets
are again distinguished open subsets.
(b) Any open subset U X is a finite union of distinguished open subsets: by definition of the
Zariski topology it is the complement of an affine variety, which in turn is the zero locus
of finitely many polynomial functions f1 , . . . , fk A(X) by Proposition 1.21 (a). Hence we
have
U = X\V ( f1 , . . . , fk ) = D( f1 ) D( fk ).
We can therefore think of the distinguished open subsets as the smallest open subsets of X in
topology, the correct notion for this would be to say that they form a basis of the Zariski topology
on X.
Proposition 3.10 (Regular functions on distinguished open subsets). Let X be an affine variety, and
let f A(X). Then

g
:
g
A(X),
n
N
.
OX (D( f )) =
fn
In particular, setting f = 1 we see that OX (X) = A(X), i. e. the regular functions on all of X are
exactly the polynomial functions.
Proof. The inclusion is obvious: every function of the form fgn for g A(X) and n N is clearly
regular on D( f ).
For the opposite inclusion , let : D( f ) K be a regular function. By Definition 3.1 we
obtain for every a D( f ) a local representation = gfaa for some fa , ga A(X) which is valid on an
open neighborhood of a in D( f ). After possibly shrinking these neighborhoods we may assume by
Remark 3.9 (b) that they are distinguished open subsets D(ha ) for some ha A(X). Moreover, we
can change the representations of by replacing ga and fa by ga ha and fa ha (which does not change
their quotient on D(ha )) to assume that both ga and fa vanish on the complement V (ha ) of D(ha ).
Finally, this means that fa vanishes on V (ha ) and does not vanish on D(ha ) so ha and fa have the
same zero locus, and we can therefore assume that ha = fa .
As a consequence, note that
ga fb = gb fa for all a, b D( f ) :
()
these two functions agree on D( fa )D( fb ) since = gfaa = gf b there, and they are both zero otherwise
b
since by our construction we have ga (x) = fa (x) = 0 for all x V ( fa ) and gb (x) = fb (x) = 0 for all
x V ( fb ).
3.
27
Now all our open neighborhoods cover D( f ), i. e. we have D( f ) = aD( f ) D( fa ). Passing to the
complement we obtain
\

V(f) =
V ( fa ) = V { fa : a D( f )} ,
aD( f )
and thus by Proposition 1.21 (b)

p
f I(V ( f )) = I V { fa : a D( f )} = ( fa : a D( f )).
This means that f n = a ka fa for some n N and ka A(X) for finitely many elements a D( f ).
Setting g := a ka ga , we then claim that = fgn on all of D( f ): for all b D( f ) we have = gf b and
b
()
g fb = ka ga fb =
a
ka gb fa = gb f
on D( fb ), and these open subsets cover D( f ).
Remark 3.11. In the proof of Proposition 3.10 we had to use Hilberts Nullstellensatz again. In fact,
the statement is false if the ground field is not algebraically closed, as you can see from the example
of the function x21+1 that is regular on all of A1R , but not a polynomial function.
Proposition 3.10 is deeply linked to commutative algebra. Although we considered the quotients fgn
in this statement to be fractions of polynomial functions, there is also a purely algebraic construction
of fractions in a (polynomial) ring in the same way as we could regard the elements of A(X) either
geometrically as functions on X or algebraically as elements in the quotient ring K[x1 , . . . , xn ]/I(X).
In these notes we will mainly use the geometric interpretation as functions, but it is still instructive
to see the corresponding algebraic construction. For reasons that will become apparent in Lemma
3.21 it is called localization; it is also one of the central topics in the Commutative Algebra class.
Construction 3.12 (Localizations [G5, Chapter 6]). Let R be a ring. A multiplicatively closed
subset of R is a subset S R with 1 S and f g S for all f , g S.
For such a multiplicatively closed subset S we then consider pairs (g, f ) with g R and f S, and
call two such pairs (g, f ) and (g0 , f 0 ) equivalent if there is an element h S with h (g f 0 g0 f ) = 0.
The equivalence class of a pair (g, f ) will formally be written as a fraction gf , the set of all such
equivalence classes is denoted S1 R. Together with the usual rules for the addition and multiplication
of fractions, S1 R is again a ring. It is called the localization of R at S.
By construction we can think of the elements of S1 R as formal fractions, with the numerators in
R and the denominators in S. In Proposition 3.10 our set of denominators is S = { f n : n N} for a
fixed element f R; in this case the localization S1 R is usually written as R f . We will meet other
sets of denominators later in Lemma 3.21 and Exercise 9.8 (a).
So let us now prove rigorously that the K-algebra OX (D( f )) of Proposition 3.10 can also be interpreted algebraically as a localization.
Lemma 3.13 (Regular functions as localizations). Let X be an affine variety, and let f A(X). Then
OX (D( f )) is isomorphic (as a K-algebra) to the localized ring A(X) f .
Proof. There is an obvious K-algebra homomorphism
g
g
7 n
fn
f
that interprets a formal fraction in the localization A(X) f as an actual quotient of polynomial func0
tions on D( f ). It is in fact well-defined: if fgn = fgm as formal fractions in the localization A(X) f
A(X) f OX (D( f )),
then f k (g f m g0 f n ) = 0 in A(X) for some k N, which means that g f m = g0 f n and thus

as functions on D( f ).
g
fn
g0
fm
The homomorphism is surjective by Proposition 3.10. It is also injective: if fgn = 0 as a function on

D( f ) then g = 0 on D( f ) and hence f g = 0 on all of X, which means f (g 1 0 f n ) = 0 in A(X)
and thus fgn = 01 as formal fractions in the localization A(X) f .
28
Andreas Gathmann
Example 3.14 (Regular functions on A2 \{0}). Probably the easiest case of an open subset of an
affine variety that is not a distinguished open subset is the complement U = A2 \{0} of the origin in
the affine plane X = A2 . We are going to see that
OA2 (A2 \{0}) = K[x1 , x2 ]
and thus that OX (U) = OX (X), i. e. every regular function on U can be extended to X. Note that this
is another result that is true in the same way in complex analysis: there is a Removable Singularity
Theorem that implies that every holomorphic function on C2 \{0} can be extended holomorphically
to C2 .
To prove our claim let OX (U). Then is regular on the distinguished open subsets D(x1 ) =
(A1 \{0}) A1 and D(x2 ) = A1 (A1 \{0}), and so by Proposition 3.10 we can write = xfm on
1
D(x1 ) and = xgn on D(x2 ) for some f , g K[x1 , x2 ] and m, n N. Of course we can do this so that
2
x1 6 | f and x2 6 | g.
On the intersection D(x1 ) D(x2 ) both representations of are valid, and so we have f x2n = g x1m on
D(x1 ) D(x2 ). But the locus V ( f x2n g x1m ) where this equation holds is closed, and hence we see
that f x2n = g x1m also on D(x1 ) D(x2 ) = A2 . In other words, we have f x2n = g x1m in the polynomial
ring A(A2 ) = K[x1 , x2 ].
Now if we had m > 0 then x1 must divide f x2n , which is clearly only possible if x1 | f . This is a
contradiction, and so it follows that m = 0. But then = f is a polynomial, as we have claimed.
Exercise 3.15. For those of you who know some commutative algebra already: generalize the proof
of Example 3.14 to show that OX (U) = OX (X) = A(X) for any open subset U of an affine variety X
such that A(X) is a unique factorization domain [G5, Definition 8.1] and U is the complement of an
irreducible subvariety of codimension at least 2 in X.
Recall that we have defined regular functions on an open subset U of an affine variety as set-theoretic
functions from U to the ground field K that satisfy some local property. Local constructions of
function-like objects occur in many places in algebraic geometry (and also in other topological
fields of mathematics), and so we will spend the rest of this chapter to formalize the idea of such
objects. This will have the advantage that it gives us an automatic definition of morphisms between
affine varieties in Chapter 4, and in fact also between more general varieties in Chapter 5.
Definition 3.16 (Sheaves). A presheaf F (of rings) on a topological space X consists of the data:
for every open set U X a ring F (U) (think of this as the ring of functions on U),
for every inclusion U V of open sets in X a ring homomorphism V,U : F (V ) F (U)
called the restriction map (think of this as the usual restriction of functions to a subset),
such that
F (0)
/ = 0,
U,U is the identity map on F (U) for all U,
for any inclusion U V W of open sets in X we have V,U W,V = W,U .
The elements of F (U) are usually called the sections of F over U, and the restriction maps V,U
are written as 7 |U .
A presheaf F is called a sheaf of rings if it satisfies the following gluing property: if U X is
an open set, {Ui : i I} an arbitrary open cover of U and i F (Ui ) sections for all i such that
i |Ui U j = j |Ui U j for all i, j I, then there is a unique F (U) such that |Ui = i for all i.
05
Example 3.17. Intuitively speaking, any function-like object forms a presheaf; it is a sheaf if the
conditions imposed on the functions are local (i. e. if they can be checked on an open cover). The
following examples illustrate this.
3.
29
(a) Let X be an affine variety. Then the rings OX (U) of regular functions on open subsets
U X, together with the usual restriction maps of functions, form a sheaf OX on X. In
fact, the presheaf axioms are obvious, and the gluing property just means that a function
: U K is regular if it is regular on each element of an open cover of U (which follows
from the definition that is regular if it is locally a quotient of polynomial functions). We
call OX the sheaf of regular functions on X.
(b) Similarly, on X = Rn the rings
F (U) = { : U R continuous}
for open subsets U X form a sheaf F on X with the usual restriction maps. In the same way
we can consider on X the sheaves of differentiable functions, analytic functions, arbitrary
functions, and so on.
(c) On X = Rn let
F (U) = { : U R constant function}
with the usual restriction maps. Then F is a presheaf, but not a sheaf, since being a constant
function is not a local condition. More precisely, let U1 and U2 be any non-empty disjoint
open subsets of X, and let 1 F (U1 ) and 2 F (U2 ) be constant functions with different
values. Then 1 and 2 trivially agree on U1 U2 = 0,
/ but there is still no constant function
on U = U1 U2 that restricts to both 1 on U1 and 2 on U2 . Hence F does not satisfy
the gluing property. Note however that we would obtain a sheaf if we considered locally
constant functions instead of constant ones.
In order to get used to the language of sheaves let us now consider two common constructions with
them.
Definition 3.18 (Restrictions of (pre-)sheaves). Let F be a presheaf on a topological space X, and
let U X be an open subset. Then the restriction of F to U is defined to be the presheaf F |U on
U with
F |U (V ) := F (V )
for every open subset V U, and with the restriction maps taken from F . Note that if F is a sheaf
then so is F |U .
Construction 3.19 (Stalks of (pre-)sheaves). Again let F be a presheaf on a topological space X.
Fix a point a X and consider pairs (U, ) where U is an open neighborhood of a and F (U).
We call two such pairs (U, ) and (U 0 , 0 ) equivalent if there is an open subset V with a V U U 0
and |V = 0 |V (it is easy to check that this is indeed an equivalence relation). The set of all such
pairs modulo this equivalence relation is called the stalk Fa of F at a; it inherits a ring structure
from the rings F (U). The elements of Fa are called germs of F at a.
Remark 3.20. The geometric interpretation of the germs of a sheaf is that they are functions (resp.
sections of the sheaf) that are defined in an arbitrarily small neighborhood of the given point we
will also refer to these objects as local functions at this point. Hence e. g. on the real line the germ
of a differentiable function at a point a allows you to compute the derivative of this function at a, but
none of the actual values of the function at any point except a. On the other hand, we have seen in
Remark 3.7 that holomorphic functions on a (connected) open subset of Cn are already determined
by their values on any smaller open set. So in this sense germs of holomorphic functions carry
much more information than germs of differentiable functions.
In algebraic geometry, the situation is similar: let 1 and 2 be two regular functions on an open
subset U of an irreducible affine variety X. If there is a point of U at which the germs of 1 and 2
are the same then 1 and 2 have to agree on a non-empty open subset, which means by Remark 3.7
that 1 = 2 on U. In other words, the germ of a regular function determines the function uniquely
already. Note that the corresponding statement is clearly false for differentiable functions as we have
seen above.
30
Andreas Gathmann
In fact, germs of regular functions on an affine variety X can also be described algebraically in terms
of localizations as introduced in Construction 3.12 which is the reason why this algebraic concept
is called localization. As one might expect, such a germ at a point a X, i. e. a regular function in
a small neighborhood of a, is given by an element in the localization of A(X) for which we allow as
denominators all polynomials that do not vanish at a.
Lemma 3.21 (Germs of regular functions as localizations). Let a be a point on an affine variety X,
and let S = { f A(X) : f (a) 6= 0}. Then the stalk OX,a of OX at a is isomorphic (as a K-algebra) to
the localized ring

g
: f , g A(X), f (a) 6= 0 .
S1 A(X) =
f
It is called the local ring of X at a.
Proof. Note that S is clearly multiplicatively closed, so that the localization S1 A(X) exists. Consider the K-algebra homomorphism

g
g
S1 A(X) OX,a ,
7 D( f ),
f
f
that maps a formal fraction
g
f
to the corresponding quotient of polynomial functions on the open
neighborhood of a where the denominator does not vanish. It is well-defined: if

g
f
g0
f0
g
f
g0
f0
in the
localization then h(g f 0 g0 f ) = 0 for some h S. Hence the functions and agree on the open
neighborhood D(h) of a, and thus they determine the same element in the stalk OX,a .
The K-algebra homomorphism is surjective since by definition any regular function in a sufficiently
small neighborhood of a must be representable by a fraction gf with g A(X) and f S. It is also
injective: assume that a function gf represents the zero element in the stalk OX,a , i. e. that it is zero in
an open neighborhood of a. By possibly shrinking this neighborhood we may assume by Remark 3.9
(b) that it is a distinguished open subset D(h) containing a, i. e. with h S. But then h(g 1 0 f )

is zero on all of X, hence zero in A(X), and thus gf = 01 in the localization S1 A(X).
Local rings will become important later on when we construct tangent spaces (see Lemma 10.5) and
vanishing multiplicities (see Definition 12.23). We will then mostly use their algebraic description
of Lemma 3.21 and write the elements of OX,a as quotients gf with f , g A(X) such that f (a) 6= 0.
Algebraically, the most important property of the local ring OX,a is that it has only one maximal
ideal in the sense of the following lemma. In fact, in commutative algebra a local ring is defined to
be a ring with only one maximal ideal.
Lemma and Definition 3.22 (Maximal ideals). Let a be a point on an affine variety X. Then every
proper ideal of the local ring OX,a is contained in the ideal

g
: f , g A(X), g(a) = 0, f (a) 6= 0
Ia := I(a) OX,a :=
f
of all local functions vanishing at the point a. The ideal Ia is therefore called the maximal ideal of
OX,a .
Proof. It is easily checked that Ia is in fact an ideal. Now let I E OX,a be any ideal not contained in
Ia . By definition, this means that there is an element gf I with f (a) 6= 0 and g(a) 6= 0. But then gf
exists in OX,a as well. Hence 1 =
f
g
gf I, and we conclude that I = OX,a .
Exercise 3.23. Let X An be an affine variety, and let a X be a point. Show that OX,a
=
OAn ,a /I(X) OAn ,a , where I(X) OAn ,a denotes the ideal in OAn ,a generated by all quotients 1f for
f I(X).
3.
31
Exercise 3.24. Let F be a sheaf on a topological space X, and let a X. Show that the stalk Fa is
a local object in the following sense: if U X is an open neighborhood of a then Fa is isomorphic
to the stalk of F |U at a on the topological space U.
Remark 3.25 (Sheaves for other categories). In Definition 3.16 we have constructed (pre-)sheaves
of rings. In the same way one can define (pre-)sheaves of K-algebras, Abelian groups, or other
suitable categories, by requiring that all F (U) are objects and all restriction maps are morphisms
in the corresponding category. Note that the stalks of such a (pre-)sheaf then inherit this structure.
For example, all our (pre-)sheaves considered so far have also been (pre-)sheaves of K-algebras for
some field K, and thus their stalks are all K-algebras. In fact, starting in the next chapter we will
restrict ourselves to this situation.
32
4.
Andreas Gathmann
Morphisms
So far we have defined and studied regular functions on an affine variety X. They can be thought
of as the morphisms (i. e. the nice maps) from open subsets of X to the ground field K = A1 . We
now want to extend this notion of morphisms to maps to other affine varieties than just A1 (and
in fact also to maps between more general varieties in Chapter 5). It turns out that there is a very
natural way to define these morphisms once you know what the regular functions are on the source
and target variety. So let us start by attaching the data of the regular functions to the structure of an
affine variety, or rather more generally of a topological space.
Definition 4.1 (Ringed spaces).
(a) A ringed space is a topological space X together with a sheaf of rings on X. In this situation
the given sheaf will always be denoted OX and called the structure sheaf of the ringed
space. Usually we will write this ringed space simply as X, with the structure sheaf OX
being understood.
(b) An affine variety will always be considered as a ringed space together with its sheaf of
regular functions as the structure sheaf.
(c) An open subset U of a ringed space X (e. g. of an affine variety) will always be considered
as a ringed space with the structure sheaf being the restriction OX |U as in Definition 3.18.
With this idea that the regular functions make up the structure of an affine variety the obvious idea
to define a morphism f : X Y between affine varieties (or more generally ringed spaces) is now
that they should preserve this structure in the sense that for any regular function : U K on an
open subset U of Y the composition f : f 1 (U) K is again a regular function.
However, there is a slight technical problem with this approach. Whereas there is no doubt about
what the composition f above should mean for a regular function on an affine variety, this
notion is a priori undefined for general ringed spaces: recall that in this case by Definition 3.16 the
structure sheaves OX and OY are given by the data of arbitrary rings OX (U) and OY (V ) for open
subsets U X and V Y . So although we usually think of the elements of OX (U) and OY (V ) as
functions on U resp. V there is nothing in the definition that guarantees us such an interpretation, and
consequently there is no well-defined notion of composing these sections of the structure sheaves
with the map f : X Y . So in order to be able to proceed without too much technicalities let us
assume from now on that all our sheaves are in fact sheaves of functions with some properties:
Convention 4.2 (Sheaves = sheaves of K-valued functions). For every sheaf F on a topological
space X we will assume from now on that the rings F (U) for open subsets U X are subrings of the
rings of all functions from U to K (with the usual pointwise addition and multiplication) containing
all constant functions, and that the restriction maps are the ordinary restrictions of such functions.
In particular, this makes every sheaf also into a sheaf of K-algebras, with the scalar multiplication
by elements of K again given by pointwise multiplication. So in short we can say:
From now on, every sheaf is assumed to be a sheaf of K-valued functions.
With this convention we can now go ahead and define morphisms between ringed spaces as motivated above.
Definition 4.3 (Morphisms of ringed spaces). Let f : X Y be a map of ringed spaces.
4.
Morphisms
33
(a) For any map : U K from an open subset U of Y to K we denote the composition f :
f 1 (U) K (which is well-defined by Convention 4.2) by f . It is called the pull-back
of by f .
(b) The map f is called a morphism (of ringed spaces) if it is continuous, and if for all open
subsets U Y and OY (U) we have f OX ( f 1 (U)). So in this case pulling back by
f yields K-algebra homomorphisms
f : OY (U) OX ( f 1 (U)), 7 f .
(c) We say that f is an isomorphism (of ringed spaces) if it has a two-sided inverse, i. e. if it is
bijective, and both f : X Y and f 1 : Y X are morphisms.
Morphisms and isomorphisms of (open subsets of) affine varieties are morphisms (resp. isomorphisms) as ringed spaces.
Remark 4.4.
(a) The requirement of f being continuous is necessary in Definition 4.3 (b) to formulate the
second condition: it ensures that f 1 (U) is open in X if U is open in Y , i. e. that OX ( f 1 (U))
is well-defined.
(b) Without our Convention 4.2, i. e. for ringed spaces without a natural notion of a pull-back
of elements of OY (U), one would actually have to include suitable ring homomorphisms
OY (U) OX ( f 1 (U)) in the data needed to specify a morphism. In other words, in this
case a morphism is no longer just a set-theoretic map satisfying certain properties. Although
this would be the correct notion of morphisms of arbitrary ringed spaces, we will not
do this here as it would clearly make our discussion of morphisms more complicated than
necessary for our purposes.
Remark 4.5 (Properties of morphisms). The following two properties of morphisms are obvious
from the definition:
(a) Compositions of morphisms are morphisms: if f : X Y and g : Y Z are morphisms of
ringed spaces then so is g f : X Z.
(b) Restrictions of morphisms are morphisms: if f : X Y is a morphism of ringed spaces and
U X and V Y are open subsets such that f (U) V then the restricted map f |U : U V
is again a morphism of ringed spaces.
Conversely, morphisms satisfy a gluing property similar to that of a sheaf in Definition
3.16:
Lemma 4.6 (Gluing property for morphisms). Let f : X Y be a map of ringed spaces. Assume
that there is an open cover {Ui : i I} of X such that all restrictions f |Ui : Ui Y are morphisms.
Then f is a morphism.
Proof. By Definition 4.3 (b) we have to check two things:
(a) The map f is continuous: Let V Y be an open subset. Then
f 1 (V ) =
[
iI
(Ui f 1 (V )) =
( f |Ui )1 (V ).
iI
But as all restrictions f |Ui are continuous the sets ( f |Ui )1 (V ) are open in Ui , and hence open
in X. So f 1 (V ) is open in X, which means that f is continuous.
Of course, this is just the well-known topological statement that continuity is a local property.
(b) The map f pulls back sections of OY to sections of OX : Let V Y be an open subset and
OY (V ). Then ( f )|Ui f 1 (V ) = ( f |Ui f 1 (V ) ) OX (Ui f 1 (V )) since f |Ui (and
thus also f |Ui f 1 (V ) by Remark 4.5 (b)) is a morphism. By the gluing property for sheaves
in Definition 3.16 this means that f OX ( f 1 (V )).
34
Andreas Gathmann
Let us now apply our definition of morphisms to (open subsets of) affine varieties. The following
proposition can be viewed as a confirmation that our constructions above were reasonable: as one
would certainly expect, a morphism to an affine variety Y An is simply given by an n-tuple of
regular functions whose image lies in Y .
Proposition 4.7 (Morphisms between affine varieties). Let U be an open subset of an affine variety
X, and let Y An be another affine variety. Then the morphisms f : U Y are exactly the maps of
the form
f = (1 , . . . , n ) : U Y, x 7 (1 (x), . . . , n (x))
with i OX (U) for all i = 1, . . . , n.
In particular, the morphisms from U to A1 are exactly the regular functions in OX (U).
Proof. First assume that f : U Y is a morphism. For i = 1, . . . , n the i-th coordinate function yi on
Y An is clearly regular on Y , and so i := f yi OX ( f 1 (Y )) = OX (U) by Definition 4.3 (b). But
this is just the i-th component function of f , and so we have f = (1 , . . . , n ).
Conversely, let now f = (1 , . . . , n ) with 1 , . . . , n OX (U) and f (U) Y . First of all f is
continuous: let Z be any closed subset of Y . Then Z is of the form V (g1 , . . . , gm ) for some g1 , . . . , gm
A(Y ), and
f 1 (Z) = {x U : gi (1 (x), . . . , n (x)) = 0 for all i = 1, . . . , m}.
But the functions x 7 gi (1 (x), . . . , n (x)) are regular on U since plugging in quotients of polynomial functions for the variables of a polynomial gives again a quotient of polynomial functions.
Hence f 1 (Z) is closed in U by Lemma 3.6, and thus f is continuous. Similarly, if OY (W ) is a
regular function on some open subset W Y then
f = f : f 1 (W ) K, x 7 (1 (x), . . . , n (x))
is regular again, since if we replace the variables in a quotient of polynomial functions by other
quotients of polynomial functions we obtain again a quotient of polynomial functions. Hence f is a
morphism.

For affine varieties themselves (rather than open subsets) we obtain as a consequence the following
useful corollary that translates our geometric notion of morphisms entirely into the language of
commutative algebra.
Corollary 4.8. For any two affine varieties X and Y there is a one-to-one correspondence
{morphisms X Y }
{K-algebra homomorphisms A(Y ) A(X)}
f .
In particular, isomorphisms of affine varieties correspond exactly to K-algebra isomorphisms in this

way.
Proof. By Definition 4.3 any morphism f : X Y determines a K-algebra homomorphism f :
OY (Y ) OX (X), i. e. f : A(Y ) A(X) by Proposition 3.10.
Conversely, let g : A(Y ) A(X) be a K-algebra homomorphism. Assume that Y An and denote by
y1 , . . . , yn the coordinate functions of An . Then i := g(yi ) A(X) = OX (X) for all i = 1, . . . , n. If we
set f = (1 , . . . , n ) : X An then f (X) Y = V (I(Y )) since all polynomials h I(Y ) represent the
zero element in A(Y ), and hence h f = g(h) = 0. Hence f : X Y is a morphism by Proposition
4.7. It has been constructed so that f = g, and so we get the one-to-one correspondence as stated in
the corollary.
The additional statement about isomorphisms now follows immediately since ( f g) = g f and
(g f ) = f g for all f : X Y and g : Y X.

06
4.
Morphisms
35
Example 4.9 (Isomorphisms 6= bijective morphisms). Let X = V (x12 x23 ) A2 be the curve as in
the picture below on the right. It has a singular point at the origin (a notion that we will introduce
in Definition 10.7 (a)) where it does not look like the graph of a differentiable function.
Now consider the map
f : A1 X, t 7 (t 3 ,t 2 )
A1
which is a morphism by Proposition 4.7. Its corresponding K-algebra

homomorphism f : A(X) A(A1 ) as in Corollary 4.8 is given by
K[x1 , x2 ]/(x12 x23 ) K[t]

x1 7 t 3
x2 7 t 2
X = V (x12 x23 )
which can be seen by composing f with the two coordinate functions of

A2 .
Note that f is bijective with inverse map
(
f
: X A , (x1 , x2 ) 7
x1
x2
if x2 6= 0
if x2 = 0.
But f is not an isomorphism (i. e. f 1 is not a morphism), since otherwise by Corollary 4.8 the map
f above would have to be an isomorphism as well which is false since the linear polynomial
t is clearly not in its image. So we have to be careful not to confuse isomorphisms with bijective
morphisms.
Another consequence of Proposition 4.7 concerns our definition of the product X Y of two affine
varieties X and Y in Example 1.4 (d). Recall from Example 2.5 (c) that X Y does not carry the
product topology which might seem strange at first. The following proposition however justifies
this choice, since it shows that our definition of the product satisfies the so-called universal property
that giving a morphism to X Y is the same as giving a morphism each to X and Y . In fact, when we
introduce more general varieties in the next chapter we will define their products using this certainly
desirable universal property.
Proposition 4.10 (Universal property of products). Let X and Y be
affine varieties, and let X : X Y X and Y : X Y Y be the
projection morphisms from the product onto the two factors. Then for
every affine variety Z and two morphisms fX : Z X and fY : Z Y
there is a unique morphism f : Z X Y such that fX = X f and
fY = Y f .
In other words, giving a morphism to the product X Y is the same as
giving a morphism to each of the factors X and Y .
fX
f
X Y
X
X
fY
Y
Y
Proof. Obviously, the only way to obtain the relations fX = X f and fY = Y f is to take the map
f : Z X Y, z 7 ( fX (z), fY (z)). But this is clearly a morphism by Proposition 4.7: as fX and fY
must be given by regular functions in each coordinate, the same is then true for f .

Remark 4.11. If you know commutative algebra you will have noticed that the universal property of
the product in Proposition 4.10 corresponds exactly to the universal property of tensor products using
the translation between morphisms of affine varieties and K-algebra homomorphisms of Corollary
4.8. Hence the the coordinate ring A(X Y ) of the product is just the tensor product A(X) K A(Y ).
Exercise 4.12. An affine conic is the zero locus in A2 of a single irreducible polynomial in K[x1 , x2 ]
of degree 2. Show that every affine conic over a field of characteristic not equal to 2 is isomorphic
to exactly one of the varieties X1 = V (x2 x12 ) and X2 = V (x1 x2 1), with an isomorphism given by
a linear coordinate transformation followed by a translation.
36
Andreas Gathmann
Exercise 4.13. Let X A2 be the zero locus of a single polynomial i+ jd ai, j x1i x2j of degree at
most d. Show that:
(a) Any line in A2 (i.e. any zero locus of a single polynomial of degree 1) not contained in X
intersects X in at most d points.
(b) Any affine conic (as in Exercise 4.12 over a field with char K 6= 2) not contained in X intersects X in at most 2d points.
(This is a (very) special case of Bzouts theorem that we will prove in Chapter 12.)
Exercise 4.14. Let f : X Y be a morphism of affine varieties and f : A(Y ) A(X) the corresponding homomorphism of the coordinate rings. Are the following statements true or false?
(a) f is surjective if and only if f is injective.
(b) f is injective if and only if f is surjective.
(c) If f : A1 A1 is an isomorphism then f is affine linear, i. e. of the form f (x) = ax + b for
some a, b K.
(d) If f : A2 A2 is an isomorphism then f is affine linear, i. e. it is of the form f (x) = Ax + b
for some A Mat(2 2, K) and b K 2 .
Construction 4.15 (Affine varieties from finitely generated K-algebras). Corollary 4.8 allows us
to construct affine varieties up to isomorphisms from finitely generated K-algebras: if R is such an
algebra we can pick generators a1 , . . . , an for R and obtain a surjective K-algebra homomorphism
g : K[x1 , . . . , xn ] R, f 7 f (a1 , . . . , an ).
By the homomorphism theorem we therefore see that R
= K[x1 , . . . , xn ]/I, where I is the kernel of
g. If we assume that I is radical (which is the same as saying that R does not have any nilpotent
elements except 0) then X = V (I) is an affine variety in An with coordinate ring A(X)
= R.
Note that this construction of X from R depends on the choice of generators of R, and so we can get
different affine varieties that way. However, Corollary 4.8 implies that all these affine varieties will
be isomorphic since they have isomorphic coordinate rings they just differ in their embeddings in
affine spaces.
This motivates us to make a (very minor) redefinition of the term affine variety to allow for objects
that are isomorphic to an affine variety in the old sense, but that do not come with an intrinsic
description as the zero locus of some polynomials in a fixed affine space.
Definition 4.16 (Slight redefinition of affine varieties). From now on, an affine variety will be a
ringed space that is isomorphic to an affine variety in the old sense of Definition 1.2 (c).
Note that all our concepts and results immediately carry over to an affine variety X in this new sense:
for example, all topological concepts are defined as X is still a topological space, regular functions
are just sections of the structure sheaf OX , the coordinate ring A(X) can be considered to be OX (X)
by Proposition 3.10, and products involving X can be defined using any embedding of X in affine
space (yielding a product that is unique up to isomorphisms).
Probably the most important examples of affine varieties in this new sense that do not look like affine
varieties a priori are our distinguished open subsets of Definition 3.8:
Proposition 4.17 (Distinguished open subsets are affine varieties). Let X be an affine variety, and let
f A(X). Then the distinguished open subset D( f ) is an affine variety; its coordinate ring A(D( f ))
is the localization A(X) f .
Proof. Clearly,
Y := {(x,t) X A1 : t f (x) = 1}
X A1
4.
Morphisms
37
is an affine variety as it is the zero locus of the polynomial t f (x) 1 in the affine variety X A1 . It
is isomorphic to D( f ) by the map

1
f : Y D( f ), (x,t) 7 x
with inverse
f 1 : D( f ) Y, x 7 x,
.
f (x)
So D( f ) is an affine variety, and by Proposition 3.10 and Lemma 3.13 we see that its coordinate ring
is A(D( f )) = OX (D( f )) = A(X) f .

Example 4.18 (A2 \{0} is not an affine variety). As in Example 3.14 let X = A2 and consider the
open subset U = A2 \{0} of X. Then even in the new sense of Definition 4.16 the ringed space U is
not an affine variety: otherwise its coordinate ring would be OX (U) by Proposition 3.10, and thus
just the polynomial ring K[x, y] by Example 3.14. But this is the same as the coordinate ring of
X = A2 , and hence Corollary 4.8 would imply that U and X are isomorphic, with the isomorphism
given by the identity map. This is obviously not true, and hence we conclude that U is not an affine
variety.
However, we can cover U by the two (distinguished) open subsets
D(x1 ) = {(x1 , x2 ) : x1 6= 0}
and
D(x2 ) = {(x1 , x2 ) : x2 6= 0}
which are affine by Proposition 4.17. This leads us to the idea that we should also consider ringed
spaces that can be patched together from affine varieties. We will do this in the next chapter.
Exercise 4.19. Which of the following ringed spaces are isomorphic over C?
(a) A1 \{1}
(b) V (x12 + x22 ) A2
(c) V (x2 x12 , x3 x13 )\{0} A3
(d) V (x1 x2 ) A2
(e) V (x22 x13 x12 ) A2
(f) V (x12 x22 1) A2
38
5.
Andreas Gathmann
Varieties
In this chapter we will now finally introduce the main objects of study in this class, the so-called
varieties. As already announced in Example 4.18 they will be spaces that are not necessarily affine
varieties themselves, but that can be covered by affine varieties. This idea is completely analogous
to the definition of manifolds: recall that to construct them one first considers open subsets of Rn
that are supposed to form the patches of your space, and then defines a manifold to be a topological
space that looks locally like these patches. In our algebraic case we can say that the affine varieties
form the basic patches of the spaces that we want to consider, and that general varieties are then
spaces that look locally like affine varieties.
One of the main motivations for this generalization is that in the classical topology affine varieties
over C are never bounded, and hence never compact, unless they are a finite set (see Exercise 2.34
(c)). As compact spaces are usually better-behaved than non-compact ones, it is therefore desirable
to have a method to compactify an affine variety by adding some points at infinity. Technically, this
can be achieved by gluing it to other affine varieties that contain the points at infinity. The complete
space can then obviously be covered by affine varieties. We will see this explicitly in Examples 5.5
(a) and 5.6, and much more generally when we construct projective varieties in Chapters 6 and 7.
So let us start by defining spaces that can be covered by affine varieties. They are called prevarieties
since we will want to impose another condition on them later in Definition 5.19, which will then
make them into varieties.
Definition 5.1 (Prevarieties). A prevariety is a ringed space X that has a finite open cover by affine
varieties. Morphisms of prevarieties are simply morphisms as ringed spaces. In accordance with
Definition 3.1, the elements of OX (U) for an open subset U X will be called regular functions
on U.
Remark 5.2. Note that the open cover in Definition 5.1 is not part of the data needed to specify a
prevariety it is just required that such a cover exists. Any open subset of a prevariety that is an
affine variety is called an affine open set.
Example 5.3. Of course, any affine variety is a prevariety. More generally, every open subset of an
affine variety is a prevariety: it has a finite open cover by distinguished open subsets by Remark 3.9
(b), and these are affine open sets by Proposition 4.17.
The basic way to construct new prevarieties is to glue them together from previously known patches.
For simplicity, let us start with the case when we only have two spaces to glue.
Construction 5.4 (Gluing two prevarieties). Let X1 , X2 be two prevarieties (e. g. affine varieties),
and let U1,2 X1 and U2,1 X2 be non-empty open subsets. Moreover, let f : U1,2 U2,1 be an
isomorphism. Then we can define a prevariety X obtained by gluing X1 and X2 along f , as shown in
the picture below:
f
U1,2
X1
glue
U2,1
X2
As a set, the space X is just the disjoint union X1 X2 modulo the equivalence relation given
by a f (a) and f (a) a for all a U1,2 (in addition to a a for all a X1 X2 ). Note
that there are then natural embeddings i1 : X1 X and i2 : X2 X that map a point to its
equivalence class in X1 X2 .
5.
Varieties
39
1
As a topological space, we call a subset U X open if i1
1 (U) X1 and i2 (U) X2 are
open. In topology, this is usually called the quotient topology of the two maps i1 and i2 .
As a ringed space, we define the structure sheaf OX by
1
OX (U) = { : U K : i1 OX1 (i1
1 (U)) and i2 OX2 (i2 (U))}
for all open subsets U X, where i1 and i2 denote the pull-backs of Definition 4.3 (a).
Intuitively, this means that a function on the glued space is regular if it is regular when
restricted to both patches. It is obvious that this defines a sheaf on X.
With this construction it is checked immediately that the images of i1 and i2 are open subsets of X
that are isomorphic to X1 and X2 . We will often drop the inclusion maps from the notation and say
that X1 and X2 are open subsets of X. Since X1 and X2 can be covered by affine open subsets, the
same is true for X, and thus X is a prevariety.
Example 5.5. As the simplest example of the above gluing construction, let X1 = X2 = A1 and
U1,2 = U2,1 = A1 \{0} in the notation of Construction 5.4. We consider two different choices of
gluing isomorphism f : U1,2 U2,1 :
(a) Let f : U1,2 U2,1 , x 7 1x . By construction, the affine line X1 = A1 is an open subset of X.
Its complement X\X1 = X2 \U2,1 is a single point that corresponds to 0 in X2 and therefore
to = 01 in the coordinate of X1 . Hence we can think of the glued space X as A1 {},
and thus as a compactification of the affine line. We denote it by P1 ; it is a special case of
the projective spaces that we will introduce in Chapter 6 (see Exercise 7.3 (a)).
In the case K = C the resulting space X is just the Riemann sphere C {}. Its subset
of real points is shown in the picture below (with the dotted arrows indicating the gluing
isomorphism), it is homeomorphic to a circle.
X1
2
0
21
1
2
X2
glue
12
1
2
As an example of gluing morphisms as in Lemma 4.6, the morphisms

X1 X2 P1 , x 7 x
and
X2 X1 P1 , x 7 x
(that correspond to a reflection across the horizontal axis in the picture above) glue together
to a single morphism P1 P1 that can be thought of as x 7 1x in the interpretation of P1 as
A1 {}.
(b) Let f : U1,2 U2,1 be the identity map. Then the space X obtained by gluing X1 and X2
along f is shown in the picture below, it is an affine line with two zero points.
3 2 1
3 2 1
X1
glue
X2
0
0
Obviously this is a somewhat weird space. Speaking in analytic terms in the case K = C, a
sequence of points tending to zero would have two possible limits in X, namely the two zero
points. Also, as in (a) the two morphisms
X1 X2 X, x 7 x
and
X2 X1 X, x 7 x
40
Andreas Gathmann
glue again to a morphism g : X X; this time it exchanges the two zero points and thus
leads to the set {x X : g(x) = x} = A1 \{0} not being closed in X, although it is given by
an equality of continuous maps.
Usually we want to exclude such spaces from the objects we consider. In Definition 5.19 we
will therefore impose an additional condition on our prevarieties that ensures that the above
phenomena do not occur (see e. g. Proposition 5.21 (b)).
Example 5.6. Consider again the complex affine curve
X = {(x1 , x2 ) A2C : x22 = (x1 1)(x1 2) (x1 2n)}
of Example 0.1. We have already seen in the introduction that X can (and should) be compactified
by adding two points at infinity, corresponding to the limit x1 and the two possible values for
x2 . Let us now construct this compactified space rigorously as a prevariety.
To be able to add a limit point x1 = to our space, let us make a coordinate change x1 =
x1 6= 0), so that the equation of the curve becomes
1
x1
(where
x22 x12n = (1 x1 )(1 2x1 ) (1 2nx1 ).

If we make an additional coordinate change x2 = x2 x1n , this becomes
x22 = (1 x1 )(1 2x1 ) (1 2nx1 ).
In these coordinates we can now add our two points at infinity, as they correspond to x1 = 0 (and
therefore x2 = 1).
Hence the compactified curve of Example 0.1 can be constructed as the prevariety obtained by
gluing the two affine varieties
X1 = {(x1 , x2 ) A2C : x22 = (x1 1)(x1 2) (x1 2n)} = X
and
X2 = {(x1 , x2 ) A2C : x22 = (1 x1 )(1 2x1 ) (1 2nx1 )}
along the isomorphism

f : U1,2 U2,1 , (x1 , x2 ) 7 (x1 , x2 ) =
1 x2
,
x1 x1n
with inverse
f

: U2,1 U1,2 , (x1 , x2 ) 7 (x1 , x2 ) =
1 x2
,
x1 x1n

,
where U1,2 = {(x1 , x2 ) : x1 6= 0} X1 and U2,1 = {(x1 , x2 ) : x1 6= 0} X2 .

Let us now turn to the general construction to glue more than two spaces together. In principle this
works in the same way as in Construction 5.4; we just have an additional technical compatibility
condition on the gluing isomorphisms.
Construction 5.7 (General gluing construction). For a finite index set I let Xi be a prevariety for all
i I. Moreover, as in the picture below suppose that for all i, j I with i 6= j we are given open
subsets Ui, j Xi and isomorphisms fi, j : Ui, j U j,i such that for all distinct i, j, k I we have
(a) f j,i = fi,1
j ;
1
(b) Ui, j fi,1
j (U j,k ) Ui,k , and f j,k f i, j = f i,k on Ui, j f i, j (U j,k ).
5.
Xi
Varieties
41
Xj
Ui, j
a
fi, j
U j,i
Ui,k
U j,k
fi,k
Uk,i
glue
f j,k
Uk, j
Xk
In analogy to Construction 5.4 we can then define a set X by taking the disjoint union of all Xi for
i I, modulo the equivalence relation a fi, j (a) for all a Ui, j Xi (in addition to a a for all a).
In fact, the conditions (a) and (b) above ensure precisely that this relation is symmetric and transitive,
respectively. It is obvious that we can now make X into a prevariety by defining its topology and
structure sheaf in the same way as in Construction 5.4. We call it the prevariety obtained by gluing
the Xi along the isomorphisms fi, j .
Exercise 5.8. Show:
(a) Every morphism f : A1 \{0} P1 can be extended to a morphism A1 P1 .
(b) Not every morphism f : A2 \{0} P1 can be extended to a morphism A2 P1 .
(c) Every morphism f : P1 A1 is constant.
Exercise 5.9.
(a) Show that every isomorphism f : P1 P1 is of the form f (x) =
where x is an affine coordinate on A1 P1 .
ax+b
cx+d
for some a, b, c, d K,
(b) Given three distinct points a1 , a2 , a3 P1 and three distinct points b1 , b2 , b3 P1 , show that
there is a unique isomorphism f : P1 P1 such that f (ai ) = bi for i = 1, 2, 3.
Exercise 5.10. If X and Y are affine varieties we have seen in Proposition 3.10 and Corollary 4.8
that there is a one-to-one correspondence
{morphisms X Y } {K-algebra homomorphisms OY (Y ) OX (X)}
f 7 f .
Does this statement still hold
(a) if X is an arbitrary prevariety (but Y is still affine);
(b) if Y is an arbitrary prevariety (but X is still affine)?
We have just seen how we can construct prevarieties by gluing affine varieties. For the rest of the
chapter let us now study some of their basic properties. Of course, all topological concepts (like
connectedness, irreducibility, and dimension) carry over immediately to the case of prevarieties.
The irreducible decomposition of Proposition 2.15 is also applicable since a prevariety is always
Noetherian:
Exercise 5.11. Prove:
(a) Any prevariety is a Noetherian topological space.
(b) If X = X1 Xm is the irreducible decomposition of a prevariety X, then the local dimension codimX {a} of X at any point a X is
codimX {a} = max{dim Xi : a Xi }.
07
42
Andreas Gathmann
(c) The statement of (a) would be false if we had defined a prevariety to be a ringed space that
has an arbitrary (not necessarily finite) open cover by affine varieties.
As for properties of prevarieties involving the structure as ringed spaces, we should first of all figure
out to what extent their subsets, images and inverse images under morphisms, and products are again
prevarieties.
Construction 5.12 (Open and closed subprevarieties). Let X be a prevariety.
(a) Let U X be an open subset. Then U is again a prevariety (as usual with the structure
sheaf OU = OX |U as in Definition 4.1 (c)): as X can be covered by affine varieties, U can
be covered by open subsets of affine varieties, which themselves can be covered by affine
varieties by Example 5.3.
We call U (with this structure as a prevariety) an open subprevariety of X.
(b) The situation is more complicated for a closed subset Y X: as an open subset U of Y is in
general not open in X we cannot define a structure sheaf on Y by simply setting OY (U) to be
OX (U). Instead, we define OY (U) to be the K-algebra of functions U K that are locally
restrictions of functions on X, or formally
OY (U) := { : U K : for all a U there are an open neighborhood V of a in X
and OX (V ) with = on U V }.
By the local nature of this definition it is obvious that OY is a sheaf, thus making Y into a
ringed space. In fact, we will check in Exercise 5.13 that Y is a prevariety in this way. We
call it a closed subprevariety of X.
(c) If U is an open and Y a closed subset of X, then U Y is open in Y and closed in U, and
thus we can give it the structure of a prevariety by combining (a) with (b) in fact, one
can check that it does not matter whether we consider it to be an open subprevariety of
Y or a closed subprevariety of U. Intersections of open and closed subprevarieties (with
this structure of a ringed space) are called locally closed subprevarieties. For example,
{(x1 , x2 ) A2 : x1 = 0, x2 6= 0} is a locally closed subprevariety of A2 .
The reason why we consider all these seemingly special cases is that for a
general subset of X there is no way to make it into a prevariety in a natural
way. Even worse, the notions of open and closed subprevarieties do not mix
very well: whereas a finite union of open (resp. closed) subprevarieties is
of course again an open (resp. closed) subprevariety, the same statement is
not true if we try to combine open with closed spaces: in X = A2 the union
of the open subprevariety U = A1 (A1 \{0}) and the closed subprevariety Y = {0} as in the picture on the right does not have a natural structure
as a subprevariety of A2 (since it does not look like an affine variety in a
neighborhood of the origin).
U Y
Exercise 5.13. Let Y be a closed subset of a prevariety X, considered as a ringed space with the
structure sheaf of Construction 5.12 (b). Prove for every affine open subset U X that the ringed
space U Y (considered as an open subset of the ringed space Y as in Definition 4.1 (c)) is isomorphic
to the affine variety U Y (considered as an affine subvariety of the affine variety U).
In particular, this shows that Construction 5.12 (b) makes Y into a prevariety, and that this prevariety
is isomorphic to the affine variety Y if X is itself affine (and thus Y an affine subvariety of X).
Remark 5.14 (Properties of closed subprevarieties). By Construction 5.12 (b), a regular function on
(an open subset of) a closed subprevariety Y of a prevariety X is locally the restriction of a regular
function on X. Hence:
(a) The inclusion map Y X is a morphism (it is clearly continuous, and regular functions on
X are by construction still regular when restricted to Y ).
5.
Varieties
43
(b) If f : Z X is a morphism from a prevariety Z such that f (Z) Y then we can also regard
f as a morphism from Z to Y (the pull-back of a regular function on Y by f is locally also a
pull-back of a regular function on X, and hence regular as f : Z X is a morphism).
Remark 5.15 (Images and inverse images of subprevarieties). Let f : X Y be a morphism of
prevarieties.
(a) The image of an open or closed subprevariety of X under f is not necessarily an open or
closed subprevariety of Y . For example, for the affine variety X = V (x2 x3 1) {0} A3
and the projection morphism f : X A2 onto the first two coordinates the image f (X) is
exactly the space A1 (A1 \{0}) {0} of Construction 5.12 which is neither an open nor a
closed subprevariety of A2 .
As a substitute, one can often consider the graph of f instead of its image, see Proposition
5.21 (a).
(b) By continuity, the inverse image of an open (resp. closed) subprevariety of Y under f is
clearly again an open (resp. closed) subprevariety of X.
As for the product X Y of two prevarieties X and Y , the natural idea to construct this space as
a prevariety would be to choose finite affine open covers {Ui : i I} and {V j : j J} of X and Y ,
respectively, and then glue the affine product varieties Ui V j using Construction 5.7. If we did this
directly however, we would still have to prove that the resulting space does not depend on the chosen
affine covers. The best way out of this trouble is to define the product prevariety by a universal
property analogous to Proposition 4.10. This will then ensure the uniqueness of the product, so that
it suffices to prove its existence by gluing affine patches.
Definition 5.16 (Products of prevarieties). Let X and Y be prevarieties. A
product of X and Y is a prevariety P together with morphisms X : P X
and Y : P Y satisfying the following universal property: for any two
morphisms fX : Z X and fY : Z Y from another prevariety Z there is a
unique morphism f : Z P such that X f = fX and Y f = fY .
As in the affine case in Proposition 4.10, this means that giving a morphism
to the product P is the same as giving a morphism to each of the factors X
and Y .
fX
f
P
X
X
fY
Y
Y
Proposition 5.17 (Existence and uniqueness of products). Any two prevarieties X and Y have a
product. Moreover, this product P with morphisms X : P X and Y : P Y is unique up to
unique isomorphism: if P0 with X0 : P0 X and Y0 : P0 Y is another product there is a unique
isomorphism f : P0 P such that X f = X0 and Y f = Y0 .
We will denote this product simply by X Y .
Proof. To show existence we glue the affine products of Proposition 4.10 using Construction 5.7.
More precisely, let X and Y be covered by affine varieties Ui and V j for i I and j J, respectively.
Use Construction 5.7 to glue the affine products Ui V j , where we glue any two such products Ui V j
and Ui0 V j0 along the identity isomorphism of the common open subset (Ui Ui0 ) (V j V j0 ). Note
that these isomorphisms obviously satisfy the conditions (a) and (b) of the construction, and that the
resulting glued space P is just the usual product X Y as a set. Moreover, using Lemma 4.6 we
can glue the affine projection morphisms Ui V j Ui X and Ui V j V j Y to morphisms
X : P X and Y : P Y .
Let us now check the universal property of Definition 5.16 for our construction. If fX : Z X and
fY : Z Y are any two morphisms from a prevariety Z, the only way to achieve X f = fX and
Y f = fY is to define f : Z P as f (z) = ( fX (z), fY (z)), where we identify P set-theoretically
with X Y . By Lemma 4.6, we can check that this is a morphism by restricting it to an affine open
cover. If we first cover Z by the open subsets fX1 (Ui ) fY1 (V j ) for all i I and j J, and these
subsets then by affine open subsets by Construction 5.12 (a), we may assume that every affine open
44
Andreas Gathmann
subset in our open cover of Z is mapped to a single (and hence affine) patch Ui V j . But after this
restriction to the affine case we know by Proposition 4.10 that f is a morphism.
To show uniqueness, assume that P0 with X0 : P0 X and Y0 : P0 Y is another product. By the
universal property of P applied to the morphisms X0 : P0 X and Y0 : P0 Y , we see that there is
a unique morphism f : P0 P with X f = X0 and Y f = Y0 . Reversing the roles of the two
products, we get in the same way a unique morphism g : P P0 with X0 g = X and Y0 g = Y .
Finally, apply the universal property of P to the two morphisms X : P X and Y : P Y . Since
both
X ( f g) = X0 g = X
X idP = X
and
Y ( f g) = Y0 g = Y
Y idP = Y
the uniqueness part of the universal property shows that f g = idP . In the same way we see that
g f = idP0 , so that f is an isomorphism.

Remark 5.18. Again, a check might be in order that our constructions were consistent: let X and
Y be prevarieties, and let X 0 X and Y 0 Y be closed subprevarieties. Then we have defined two
structures of prevarieties on the set-theoretic product X 0 Y 0 : the closed subprevariety structure of
X 0 Y 0 in X Y as in Construction 5.12 (b), and the product prevariety structure of Definition 5.16.
As expected, these two structures agree: in fact, by Definition 5.16 together with Remark 5.14 the
set-theoretic identity map is a morphism between these two structures in both ways.
Finally, as already announced let us now impose a condition on prevarieties that excludes such unwanted spaces as the affine line with two zero points of Example 5.5 (b). In the theory of manifolds,
this is usually done by requiring that the topological space satisfies the so-called Hausdorff property,
i. e. that every two distinct points have disjoint open neighborhoods. This is obviously not satisfied
in our case, since the two zero points do not have such disjoint open neighborhoods.
However, in the Zariski topology the Hausdorff property does not make too much sense, as nonempty open subsets of an irreducible space can never be disjoint by Remark 2.18 (a). So we need
a suitable replacement for this condition that captures our geometric idea of the absence of doubled
points also in the Zariski topology.
The solution to this problem is inspired by a proposition in general topology stating that the Hausdorff property of a topological space X is equivalent to the condition that the so-called diagonal
X = {(x, x) : x X} is closed in X X (with the product topology). The picture below illustrates
this in the case when X is the affine line with two zero points a and b: the product X X then
contains four zero points (a, a), (a, b), (b, a), and (b, b), but by definition only two of them, namely
(a, a) and (b, b), are in X . Hence the diagonal is not closed: the other two zero points are also
contained in its closure.
X
a
b
b
a
a
b
X X
Of course, this equivalence does not really help us directly in algebraic geometry since we do not
use the product topology on X X. But the geometric idea to detect doubled points shown in the
picture above on the right is still valid in the Zariski topology and so we will just use the diagonal
condition to define the property that we want:
Definition 5.19 (Varieties).
(a) A prevariety X is called a variety (or separated) if the diagonal
X := {(x, x) : x X}
is closed in X X.
5.
Varieties
45
(b) Analogously to Definition 2.30 (b), a variety of pure dimension 1 or 2 is called a curve resp.
surface. If X is a pure-dimensional variety and Y a pure-dimensional subvariety of X with
dimY = dim X 1 we say that Y is a hypersurface in X.
So by the argument given above, the affine line with two zero points of Example 5.5 (b) is not a
variety. In contrast, the following lemma shows that most prevarieties that we will meet are also
varieties. From now on we will almost always assume that our spaces are separated, and thus talk
about varieties instead of prevarieties.
Lemma 5.20.
(a) Affine varieties are varieties.
(b) Open, closed, and locally closed subprevarieties of varieties are varieties. We will therefore
simply call them open, closed, and locally closed subvarieties, respectively.
Proof.
(a) If X An then X = V (x1 y1 , . . . , xn yn ) X X, where x1 , . . . , xn and y1 , . . . , yn are the
coordinates on the two factors, respectively. Hence X is closed.
(b) If Y X is an open, closed, or locally closed subvariety, consider the inclusion morphism i :
Y Y X X (which exists by the universal property of Definition 5.16). As Y = i1 (X )
and X is closed by assumption, Y is closed as well by the continuity of i.

For varieties, we have the following additional desirable properties in addition to the ones for prevarieties:
Proposition 5.21 (Properties of varieties). Let f , g : X Y be morphisms of prevarieties, and assume that Y is a variety.
(a) The graph f := {(x, f (x)) : x X} is closed in X Y .
(b) The set {x X : f (x) = g(x)} is closed in X.
Proof.
(a) By the universal property of products of prevarieties as in Definition 5.16 there is a morphism
( f , id) : X Y Y Y, (x, y) 7 ( f (x), y). As Y is a variety we know that Y is closed, and
hence so is f = ( f , id)1 (Y ) by continuity.
(b) Similarly to (a), the given set is the inverse image of the diagonal Y under the morphism
X Y Y, x 7 ( f (x), g(x)). Hence it is closed again since Y is closed.

Exercise 5.22. Show that the space P1 of Example 5.5 (a) is a variety.
Exercise 5.23. Let X and Y be prevarieties. Show:
(a) If X and Y are varieties then so is X Y .
(b) If X and Y are irreducible then so is X Y .
Exercise 5.24. Use diagonals to prove the following statements:
(a) The intersection of any two affine open subsets of a variety is again an affine open subset.
(b) If X,Y An are two pure-dimensional affine varieties then every irreducible component of
X Y has dimension at least dim X + dimY n.
Exercise 5.25. In Exercise 2.33 (b) we have seen that the dimension of a dense open subset U of a
topological space X need not be the same as that of X.
However, show now that dimU = dim X holds in this situation if X is a variety.
08
46
6.
Andreas Gathmann
Projective Varieties I: Topology
In the last chapter we have studied (pre-)varieties, i. e. topological spaces that are locally isomorphic
to affine varieties. In particular, the ability to glue affine varieties together allowed us to construct
compact spaces (in the classical topology over the ground field C) as e. g. P1 , whereas affine varieties
themselves are never compact unless they consist of only finitely many points (see Exercise 2.34 (c)).
Unfortunately, the description of a variety in terms of its affine patches and gluing isomorphisms is
quite inconvenient in practice, as we have seen already in the calculations in the last chapter. It
would therefore be desirable to have a global description of these spaces that does not refer to gluing
methods.
Projective varieties form a very large class of compact varieties that do admit such a global description. In fact, the class of projective varieties is so large that it is not easy to construct a variety
that is not (an open subset of) a projective variety in this class we will certainly not see one.
In this chapter we will construct projective varieties as topological spaces, leaving their structure as
ringed spaces to Chapter 7. To do this we first of all need projective spaces, which can be thought
of as compactifications of affine spaces. We have already seen P1 as A1 together with a point at
infinity in Example 5.5 (a); other projective spaces are just generalizations of this construction to
higher dimensions. As we aim for a global description of these spaces however, their definition looks
quite different from the one in Example 5.5 (a) at first.
Definition 6.1 (Projective spaces). Let n N. We define projective n-space over K, denoted PnK or
simply Pn , to be the set of all 1-dimensional linear subspaces of the vector space K n+1 .
Notation 6.2 (Homogeneous coordinates). Obviously, a 1-dimensional linear subspace of K n+1 is
uniquely determined by a non-zero vector in K n+1 , with two such vectors spanning the same linear
subspace if and only if they are scalar multiples of each other. In other words, we have
Pn = (K n+1 \{0}) /
with the equivalence relation
(x0 , . . . , xn ) (y0 , . . . , yn )
xi = yi for some K and all i,
where K = K\{0} is the multiplicative group of units of K. This is usually written as Pn =

(K n+1 \{0})/K , and the equivalence class of (x0 , . . . , xn ) will be denoted by (x0 : : xn ) Pn
(the notations [x0 : : xn ] and [x0 , . . . , xn ] are also common in the literature). So in the notation
(x0 : : xn ) for a point in Pn the numbers x0 , . . . , xn are not all zero, and they are defined only up to
a common scalar multiple. They are called the homogeneous coordinates of the point (the reason
for this name will become obvious in the course of this chapter). Note also that we will usually label
the homogeneous coordinates of Pn by x0 , . . . , xn instead of by x1 , . . . , xn+1 . This choice is motivated
by the following relation between An and Pn .
Remark 6.3 (Geometric interpretation of Pn ). Consider the map
f : An Pn , (x1 , . . . , xn ) 7 (1 : x1 : : xn ).
As in the picture below on the left we can embed the affine space An in K n+1 at the height x0 = 1,
and then think of f as mapping a point to the 1-dimensional linear subspace spanned by it.
6.
47
(1 : a1 : a2 )
x0
A2
x0
1
A2
(a1 , a2 )
x(t)
x2
b
x1
x1
The map f is obviously injective, with image U0 := {(x0 : : xn ) : x0 6= 0}. On this image the inverse
of f is given by
x
xn
1
f 1 : U0 An , (x0 : : xn ) 7
,...,
,
x0
x0
n
n+1
it sends a line through the origin to its intersection point with
A embedded in K . We can nthus
x1
xn
n
n
think of A as a subset U0 of P . The coordinates x0 , . . . , x0 of a point (x0 : : xn ) U0 P are
called its affine coordinates.
The remaining points of Pn are of the form (0 : x1 : : xn ); in the picture above they correspond to
lines in the horizontal plane through the origin, such as e. g. b. By forgetting their first coordinate
(which is zero anyway) they form a set that is naturally bijective to Pn1 . Thinking of the ground
field C we can regard them as points at infinity: consider e. g. in AnC a parametrized line
x(t) = (a1 + b1 t, . . . , an + bn t) for t C
for some starting point (a1 , . . . , an ) Cn and direction vector (b1 , . . . , bn ) Cn \{0}. Of course, there
is no limit point of x(t) in AnC as t . But if we embed AnC in PnC as above we have
1 a

a
: 1 + b1 : : n + bn
x(t) = (1 : a1 + b1 t : : an + bn t) =
t t
t
in homogeneous coordinates, and thus (in a suitable topology) we get a limit point (0 : b1 : : bn )
PnC \AnC at infinity for x(t) as t . This limit point obviously remembers the direction, but not the
position of the original line. Hence we can say that in Pn we have added a point at infinity to An in
each direction. In other words, after extension to Pn two distinct lines in An will meet at infinity if
and only if they are parallel, i. e. point in the same direction.
Usually, it is more helpful to think of the projective space Pn as the affine space An compactified
by adding some points (parametrized by Pn1 ) at infinity, rather than as the set of 1-dimensional
linear subspaces in K n+1 . In fact, after having given Pn the structure of a variety we will see in
Proposition 7.2 and Exercise 7.3 (b) that with the above constructions An and Pn1 are open and
closed subvarieties of Pn , respectively.
Remark 6.4 (PnC is compact in the classical topology). In the case K = C one can give PnC a standard
(quotient) topology by declaring a subset U Pn to be open if its inverse image under the quotient
map : Cn+1 \{0} Pn is open in the standard topology. Then PnC is compact: let
S = {(x0 , . . . , xn ) Cn+1 : |x0 |2 + + |xn |2 = 1}
be the unit sphere in Cn+1 . This is a compact space as it is closed and bounded. Moreover, as every
point in Pn can be represented by a unit vector in S, the restricted map |S : S Pn is surjective.
Hence Pn is compact as a continuous image of a compact set.
Remark 6.5 (Homogeneous polynomials). In complete analogy to affine varieties, we now want to
define projective varieties to be subsets of Pn that can be given as the zero locus of some polynomials
in the homogeneous coordinates. Note however that if f K[x0 , . . . , xn ] is an arbitrary polynomial,
it does not make sense to write down a definition like
V ( f ) = {(x0 : : xn ) : f (x0 , . . . , xn ) = 0}
Pn ,
because the homogeneous coordinates are only defined up to a common scalar. For example, if
f = x12 x0 K[x0 , x1 ] then f (1, 1) = 0 and f (1, 1) 6= 0, although (1 : 1) = (1 : 1) in P1 . To
48
Andreas Gathmann
get rid of this problem we have to require that f is homogeneous, i. e. that all of its monomials have
the same (total) degree d: in this case
f ( x0 , . . . , xn ) = d f (x0 , . . . , xn ) for all K ,
and so in particular we see that
f ( x0 , . . . , xn ) = 0
f (x0 , . . . , xn ) = 0,
so that the zero locus of f is well-defined in Pn . So before we can start with our discussion of
projective varieties we have to set up some algebraic language to be able to talk about homogeneous
elements in a ring (or K-algebra).
Definition 6.6 (Graded rings and K-algebras).
(a) A graded ring is a ring R together with Abelian subgroups Rd R for all d N, such that:
Every element f R has a unique decomposition f = dN fd such that fd Rd for
all d N and only finitely many fd are non-zero. In accordance with the direct sum
L
notation in linear algebra, we usually write this condition as R = dN Rd .
For all d, e N and f Rd , g Re we have f g Rd+e .
For f R\{0} the biggest number d N with fd 6= 0 in the decomposition f = dN fd
as above is called the degree deg f of f . The elements of Rd \{0} are said to be homogeL
neous (of degree d). We call f = dN fd and R = dN Rd as above the homogeneous
decomposition of f and R, respectively.
(b) If R is also a K-algebra in addition to (a), we say that it is a graded K-algebra if f Rd
for all d N and f Rd .
Example 6.7. The polynomial ring R = K[x0 , . . . , xn ] is obviously a graded ring with
(
)
Rd =
i
ai0 ,...,in x00 xnin : ai0 ,...,in K for all i0 , . . . , in
i0 ,...,in N
i0 ++in =d
for all d N. In the following we will always consider it with this grading.
Exercise 6.8. Let R 6= 0 be a graded ring. Show that the multiplicative unit 1 R is homogeneous
of degree 0.
Of course, we will also need ideals in graded rings. Naively, one might expect that we should
consider ideals consisting of homogeneous elements in this case. However, as an ideal has to be
closed under multiplication with arbitrary ring elements, it is virtually impossible that all of its
elements are homogeneous. Instead, the correct notion of homogeneous ideal is the following.
Definition 6.9 (Homogeneous ideals). An ideal in a graded ring is called homogeneous if it can be
generated by homogeneous elements.
Lemma 6.10 (Properties of homogeneous ideals). Let I and J be ideals in a graded ring R.
(a) The ideal I is homogeneous if and only if for all f I with homogeneous decomposition
f = dN fd we also have fd I for all d.
(b) If I and J are homogeneous then so are I + J, IJ, I J, and I.

(c) If I is homogeneous
then the quotient R/I is a graded ring with homogeneous decomposition
L
R/I = dN Rd /(Rd I).
Proof.
(a) : Let I = (h j : j J) for homogeneous elements h j R for all j, and let f I. Then
f = jJ g j h j for some (not necessarily homogeneous) g j R, of which only finitely many
6.
49
are non-zero. If we denote by g j = eN g j,e the homogeneous decompositions of these

elements, the degree-d part of f for d N is
fd =
g j,e h j I.
jJ,eN
e+deg h j =d
: Now let I = (h j : j J) for arbitrary elements h j R for all j. If h j = dN h j,d is

their homogeneous decomposition, we have h j,d I for all j and d by assumption, and thus
I = (h j,d : j J, d N) can be generated by homogeneous elements.
(b) If I and J are generated by homogeneous elements, then clearly so are I + J (which is generated by I J) and IJ. Moreover, I and J then satisfy the equivalent condition of (a), and thus
so does I J.
It remains
to be shown that I is homogeneous. We will check the condition of (a) for any
f I by induction over the degree d of f . Writing f = f0 + + fd in its homogeneous

decomposition, we get
f n = ( f0 + + fd )n = fdn + (terms of lower degree) I
for some n N, hence fdn I by (a), and thus fd I. But then f fd = f0 + + fd1 I
as well, and so by the induction hypothesis we also see that f0 , . . . , fd1 I.

(c) It is clear that Rd /(Rd I) R/I, f 7 f is an injective group homomorphism, so that we
can consider Rd /(Rd I) as a subgroup of R/I for all d.
Now let f R be arbitrary, with homogeneous decomposition f = dN fd . Then f =
dN fd with fd Rd /(Rd I), so f also has a homogeneous decomposition. Moreover, this
decomposition is unique: if dN fd = dN gd are two such decompositions of the same
element in R/I then dN ( fd gd ) lies in I, hence by (a) is of the form dN hd with all
hd Rd I. But then
( fd gd hd ) = 0,
dN
which implies that fd gd hd = 0, and thus fd = gd Rd /(Rd I) for all d.
With this preparation we can now define projective varieties in the same way as affine ones. For
simplicity, for a homogeneous polynomial f K[x0 , . . . , xn ] and a point x = (x0 : : xn ) Pn we
will write the condition f (x0 , . . . , xn ) = 0 (which is well-defined by Remark 6.5) also as f (x) = 0.
Definition 6.11 (Projective varieties and their ideals). Let n N.
(a) Let S K[x0 , . . . , xn ] be a set of homogeneous polynomials. Then the (projective) zero locus
of S is defined as
V (S) := {x Pn : f (x) = 0 for all f S}
Pn .
Subsets of Pn that are of this form are called projective varieties. For S = ( f1 , . . . , fk ) we
will write V (S) also as V ( f1 , . . . , fk ).
(b) For a homogeneous ideal I E K[x0 , . . . , xn ] we set
V (I) := {x Pn : f (x) = 0 for all homogeneous f I}
Pn .
Obviously, if I is the ideal generated by a set S of homogeneous polynomials then V (I) =

V (S).
(c) If X Pn is any subset we define its ideal to be
I(X) := f K[x0 , . . . , xn ] homogeneous : f (x) = 0 for all x X
E K[x0 , . . . , xn ].
(Note that the homogeneous polynomials vanishing on X do not form an ideal yet, so that
we have to take the ideal generated by them.)
If we want to distinguish these projective constructions from the affine ones in Definitions 1.2 (c) and
1.10 we will denote them by Vp (S) and I p (X), and the affine ones by Va (S) and Ia (X), respectively.
50
Andreas Gathmann
Example 6.12.
(a) As in the affine case, the empty set 0/ = Vp (1) and the whole space Pn = Vp (0) are projective
varieties.
(b) If f1 , . . . , fr K[x0 , . . . , xn ] are homogeneous linear polynomials then Vp ( f1 , . . . , fr ) Pn is
a projective variety. Projective varieties that are of this form are called linear subspaces of
Pn .
Exercise 6.13. Let a Pn be a point. Show that the one-point set {a} is a projective variety, and
compute explicit generators for the ideal I p ({a}) E K[x0 , . . . , xn ].
Example 6.14. Let f = x12 x22 x02 C[x0 , x1 , x2 ]. The real part of the affine zero locus Va ( f ) A3
of this homogeneous polynomial is the 2-dimensional cone shown in the picture below on the left.
According to Definition 6.11, its projective zero locus Vp ( f ) P2 is the set of all 1-dimensional
linear subspaces contained in this cone but we have seen in Remark 6.3 already that we should
rather think of P2 as the affine plane A2 (embedded in A3 at x0 = 1) together with some points at
infinity. With this interpretation the real part of Vp ( f ) consists of the hyperbola shown below on the
right (whose equation x12 x22 1 = 0 can be obtained by setting x0 = 1 in f ), together with two
points a and b at infinity. In the 3-dimensional picture on the left, these two points correspond to
the two 1-dimensional linear subspaces parallel to the plane at x0 = 1, in the 2-dimensional picture
of the affine part in A2 on the right they can be thought of as points at infinity in the corresponding
directions. Note that, in the latter interpretation, opposite points at infinity are actually the same,
since they correspond to the same 1-dimensional linear subspace in C3 .
b
x0
Va ( f )
x2
Vp ( f )
1
a
a
x1
b
Vp ( f )
a
09
We see in this example that the affine and projective zero locus of f carry essentially the same
geometric information the difference is just whether we consider the cone as a set of individual
points, or as a union of 1-dimensional linear subspaces in A3 . Let us now formalize and generalize
this correspondence.
Definition 6.15 (Cones). Let : An+1 \{0} Pn , (x0 , . . . , xn ) 7 (x0 : : xn ).
(a) An affine variety X An+1 is called a cone if 0 X, and x X for all K and x X. In
other words, it consists of the origin together with a union of lines through 0.
(b) For a cone X An+1 we call
P(X) := (X\{0}) = {(x0 : : xn ) Pn : (x0 , . . . , xn ) X}
Pn
the projectivization of X.
(c) For a projective variety X Pn we call
C(X) := {0} 1 (X) = {0} {(x0 , . . . , xn ) : (x0 : : xn ) X}
An+1
the cone over X (we will see in Remark 6.17 that this is in fact a cone in the sense of (a)).
Remark 6.16 (Cones and homogeneous ideals). If S K[x0 , . . . , xn ] is a set of homogeneous polynomials with non-empty affine zero locus in An+1 then Va (S) is a cone: clearly, we have 0 Va (S)
as every non-constant homogeneous polynomial vanishes at the origin. Moreover, let K and
x Va (S). Then f (x) = 0 for all f S, hence f ( x) = deg f f (x) = 0, and so x Va (S) as well.
6.
51
Conversely, the ideal I(X) of a cone X An+1 is homogeneous: let f I(X) with homogeneous
decomposition f = dN fd . Then for all x X we have f (x) = 0, and therefore also
0 = f ( x) =
d fd (x)
dN
for all K since X is a cone. This means that we have the zero polynomial in , i. e. that fd (x) = 0
for all d, and thus fd I(X). Hence I(X) is homogeneous by Lemma 6.10 (a).
Remark 6.17 (Cones projective varieties). Let S K[x0 , . . . , xn ] be a set of homogeneous polynomials with non-empty affine zero locus in An+1 . Then Va (S) is a cone by Remark 6.16, and by
construction we have
P(Va (S)) = Vp (S)
and
C(Vp (S)) = Va (S).
In particular, the projectivization P(Va (S)) is a projective variety, and C(Vp (S)) is a cone. Moreover,
as Remark 6.16 also shows that every cone is of the form Va (S) for a suitable set S of homogeneous
polynomials (namely generators for its homogeneous ideal), we obtain a one-to-one correspondence
{cones in An+1 }
X
C(X)
{projective varieties in Pn }
7 P(X)
7 X.
In other words, the correspondence works by passing from the affine to the projective zero locus
(and vice versa) of the same set of homogeneous polynomials, as in Example 6.14. Note that in this
way linear subspaces of An+1 correspond exactly to linear subspaces of Pn in the sense of Example
6.12 (b).
Having defined projective varieties, we can now proceed with their study as in the affine case. First
of all, we should associate a coordinate ring to a projective variety, and consider zero loci and ideals
with respect to these coordinate rings.
Construction 6.18 (Relative version of zero loci and ideals). Let Y Pn be a projective variety. In
analogy to Definition 1.19 we call
S(Y ) := K[x0 , . . . , xn ]/I(Y )
the homogeneous coordinate ring of Y . By Lemma 6.10 (c) it is a graded ring, so that it makes sense
to talk about homogeneous elements of S(Y ). Moreover, the condition f (x) = 0 is still well-defined
for a homogeneous element f S(Y ) and a point x Y , and thus we can define as in Definition 6.11
V (I) := {x Y : f (x) = 0 for all homogeneous f I}
for a homogeneous ideal I E S(Y )
(and similarly for a set of homogeneous polynomials in S(Y )), and

I(X) := f S(Y ) homogeneous : f (x) = 0 for all x X
for a subset X Y .
As before, in case of possible confusion we will decorate V and I with the subscript Y and/or p to
denote the relative and projective situation, respectively. Subsets of Y that are of the form VY (I) for a
homogeneous ideal I E S(Y ) will be called projective subvarieties of Y ; these are obviously exactly
the projective varieties contained in Y .
Remark 6.19. Let Y be a projective variety. The following results are completely analogous to the
affine case:
(a) (Hilberts Basis Theorem) Every homogeneous ideal in S(Y ) can be generated by finitely
many elements. In fact, it is finitely generated by [G5, Proposition 7.13 and Remark 7.15],
and hence also by homogeneous elements as we have seen in the proof of part of
Lemma 6.10 (a).
(b) The operations VY ( ) and IY ( ) reverse inclusions, we have X = VY (IY (X)) for every projective subvariety X of Y , and J IY (VY (J)) for any homogeneous ideal J E S(Y ) these
statements follow literally in the same way as in Lemma 1.12.
52
Andreas Gathmann
(c) The
p by Lemma 6.10 (b) the radical
p ideal IY (X) of a projective subvariety X Y is radical:
IY (X) is homogeneous, so it suffices to prove that f IY (X) implies f IY (X) for any
homogeneous f . But this is obvious since f k = 0 on Y for some k implies f = 0 on Y .
(d) By (c) the ideal I p (Y ) E K[x0 , . . . , xn ] is radical. Hence Proposition 1.17 implies that it is also
the ideal of its affine zero locus Va (I p (Y )) An+1 . But Vp (I p (Y )) = Y by (b), and so we see
by Remark 6.17 that Va (I p (Y )) = C(Y ). Therefore we conclude that I p (Y ) = Ia (C(Y )), and
thus that S(Y ) = A(C(Y )). Hence every homogeneous coordinate ring of a projective variety
can also be interpreted as a usual coordinate ring of an affine variety.
Remark 6.20. A remark that is sometimes useful is that every projective subvariety X of a projective
variety Y Pn can be written as the zero locus of finitely many homogeneous polynomials in S(Y ) of
the same degree. This follows easily from the fact that Vp ( f ) = Vp (x0d f , . . . , xnd f ) for all homogeneous
f S(Y ) and every d N. However, it is not true that every homogeneous ideal in S(Y ) can be
generated by homogeneous elements of the same degree.
Of course, we would alsoexpect a projective version of the Nullstellensatz as in Proposition 1.21
(b), i. e. that IY (VY (J)) = J for any homogeneous ideal J in the homogeneous coordinate ring of a
projective variety Y . This is almost true and can in fact be proved by reduction to the affine case
there is one exception however, since the origin in An+1 does not correspond to a point in projective
space Pn :
Example 6.21 (Irrelevant ideal). Let Y Pn be a non-empty projective variety, and let
I0 := (x0 , . . . , xn ) E S(Y ) = K[x0 , . . . , xn ]/I(Y ).
Then I0 is a homogeneous radical ideal, and its projective zero locus is empty since there is no point
in Y all of whose homogeneous coordinates are zero. Hence
IY (VY (I0 )) = IY (0)
/ = S(Y ),
which is not equal to I0 = I0 . In fact, I0 can never appear as the ideal of a projective variety,
since I0 = IY (X) for some X Y would imply X = VY (IY (X)) = VY (I0 ) = 0/ by Remark 6.19 (b), in
contradiction to IY (0)
/ = S(Y ).
We will see now however that this is the only counterexample to the projective version of the Nullstellensatz. The ideal I0 above is therefore often called the irrelevant ideal. Note that by Proposition
1.21 (b) this is the unique radical ideal whose affine zero locus is {0}.
Proposition 6.22 (Projective Nullstellensatz).
Let Y be a non-empty projective variety, and
let
J S(Y ) be a homogeneous ideal such that J is not the irrelevant ideal. Then I p (Vp (J)) = J. In
particular, we have an inclusion-reversing one-to-one correspondence

homogeneous radical ideals in S(Y )
{projective subvarieties of Y }
not equal to the irrelevant ideal
X
Vp (J)
I p (X)
J.
Proof. In this proof we will regard J as an ideal in both S(Y ) and A(C(Y )) (see Remark 6.19 (d)), so
that we can take both its projective and its affine zero locus. Note then that

I p (Vp (J)) = f S(Y ) homogeneous : f (x) = 0 for all x Vp (J)

= f S(Y ) homogeneous : f (x) = 0 for all x Va (J)\{0} .
As the affine zero locus of polynomials is closed, we can rewrite this as

I p (Vp (J)) = f S(Y ) homogeneous : f (x) = 0 for all x Va (J)\{0} .
6.
53
By Example 6.21 we have Va (J) = Va ( J) 6= {0} since J is not the irrelevant ideal. But then
Va (J)\{0} = Va (J): if Va (J) = 0/ this is trivial, and otherwise the cone Va (J)\{0} contains at least
one line without the origin, so that the origin lies in its closure. Hence we get

I p (Vp (J)) = f S(Y ) homogeneous : f (x) = 0 for all x Va (J) .
As the ideal of the cone Va (J)
is homogeneous by Remark 6.16 this can be rewritten as I p (Vp (J)) =
Ia (Va (J)), which is equal to J by the affine Nullstellensatz of Proposition 1.21 (b).
The one-to-one correspondence then follows together with the statement Vp (I p (X)) = X from Remark 6.19 (b) (note that I p (X) is always homogeneous by definition, radical by Remark 6.19 (c), and
not equal to the irrelevant ideal by Example 6.21).

Remark 6.23 (Properties of Vp ( ) and I p ( )). The operations Vp ( ) and I p ( ) satisfy the same properties as their affine counterparts in Lemma 1.24, Remark 1.25, and Lemma 1.26. More precisely,
for any projective variety X we have:
(a) For any family {Si } of subsets of S(X) we have i Vp (Si ) = Vp (
S1 , S2 S(X) we have Vp (S1 ) Vp (S2 ) = Vp (S1 S2 ).
T
i Si );
for any two subsets
(b) If J1 , J2 E S(X) are homogeneous ideals then

Vp (J1 ) Vp (J2 ) = Vp (J1 + J2 ) and Vp (J1 ) Vp (J2 ) = Vp (J1 J2 ) = Vp (J1 J2 ).
(c) For subsets Y1p
,Y2 of a projective variety X we have I p (Y1 Y2 ) = I p (Y1 ) I p (Y2 ). Moreover,
I p (Y1 Y2 ) = I p (Y1 ) + I p (Y2 ) unless the latter is the irrelevant ideal (which is only possible
if Y1 and Y2 are disjoint).
The proof of these statements is completely analogous to the affine case.
In particular, by (a) it follows that arbitrary intersections and finite unions of projective subvarieties
of X are again projective subvarieties, and hence we can define the Zariski topology on X in the same
way as in the affine case:
Definition 6.24 (Zariski topology). The Zariski topology on a projective variety X is the topology
whose closed sets are exactly the projective subvarieties of X, i. e. the subsets of the form Vp (S) for
some set S S(X) of homogeneous elements.
Of course, from now on we will always use this topology for projective varieties and their subsets.
Note that, in the same way as in Remark 2.3, this is well-defined in the sense that the Zariski topology
on a projective variety X Pn agrees with the subspace topology of X in Pn . Moreover, since we
want to consider An as a subset of Pn as in Remark 6.3 we should also check that the Zariski topology
on An is the same as the subspace topology of An in Pn . To do this, we need the following definition.
Definition 6.25 (Homogenization).
(a) Let
f=
K[x1 , . . . , xn ]
i1 ,...,in N
be a (non-zero) polynomial of degree d. We define its homogenization to be

x
xn
1
f h := x0d f
,...,
x0
x0
=
ai1 ,...,in x0di1 in x1i1 xnin
K[x0 , . . . , xn ];
i1 ,...,in N
obviously this is a homogeneous polynomial of degree d.

(b) The homogenization of an ideal I E K[x1 , . . . , xn ] is defined to be the ideal I h in K[x0 , . . . , xn ]
generated by all f h for f I.
Example 6.26. For f = x12 x22 1 K[x1 , x2 ] we have f h = x12 x22 x02 K[x0 , x1 , x2 ].
54
Andreas Gathmann
Remark 6.27. If f , g K[x1 , . . . , xn ] are polynomials of degree d and e, respectively, then f g has
degree d + e, and so we get
x
xn x1
xn
1
( f g)h = x0d+e f
,...,
g
,...,
= f h gh .
x0
x0
x0
x0
However, ( f + g)h is clearly not equal to f h + gh in general in fact, f h + gh is usually not even
homogeneous. This is the reason why in Definition 6.25 (b) we have to take the ideal generated by
all homogenizations of polynomials in I, instead of just all these homogenizations themselves.
Remark 6.28 (An as an open subset of Pn ). Recall from Remark 6.3 that we want to identify the
subset U0 = {(x0 : : xn ) Pn : x0 6= 0} of Pn with An by the bijective map
F : An U0 , (x1 , . . . , xn ) 7 (1 : x1 : : xn ).
Obviously, U0 is an open subset of Pn . Moreover, with the above identification the subspace topology
of U0 = An Pn is the affine Zariski topology:
(a) If X = Vp (S) An is closed in the subspace topology (for a subset S K[x0 , . . . , xn ] of homogeneous polynomials) then X = V ( f (1, ) : f S) is also Zariski closed.
(b) If X = V (S) An is Zariski closed (with S K[x1 , . . . , xn ]) then X = Vp ( f h : f S) An is
closed in the subspace topology as well.
In other words we can say that the map F : An U0 above is a homeomorphism. In fact, after having
given Pn the structure of a variety we will see in Proposition 7.2 that it is even an isomorphism of
varieties.
Having defined the Zariski topology on projective varieties (or more generally on subsets of Pn ) we
can now immediately apply all topological concepts of Chapter 2 to this new situation. In particular,
the notions of connectedness, irreducibility, and dimension are well-defined for projective varieties
(and have the same geometric interpretation as in the affine case). Let us study some examples using
these concepts.
Remark 6.29 (Pn is irreducible of dimension n). Of course, by symmetry of the coordinates, it
follows from Remark 6.28 that all subsets Ui = {(x0 : : xn ) : xi 6= 0} of Pn for i = 0, . . . , n are
homeomorphic to An as well. As these subsets cover Pn and have non-empty intersections, we
conclude by Exercise 2.20 (b) that Pn is irreducible, and by Exercise 2.33 (a) that dim Pn = n.
Exercise 6.30. Let L1 , L2 P3 be two disjoint lines (i. e. 1-dimensional linear subspaces in the sense
of Example 6.12 (b)), and let a P3 \(L1 L2 ). Show that there is a unique line L P3 through a
that intersects both L1 and L2 .
Is the corresponding statement for lines and points in A3 true as well?
Exercise 6.31.
(a) Prove that a graded ring R is an integral domain if and only if for all homogeneous elements
f , g R with f g = 0 we have f = 0 or g = 0.
(b) Show that a projective variety X is irreducible if and only if its homogeneous coordinate ring
S(X) is an integral domain.
Exercise 6.32. In this exercise we want to show that an intersection of projective varieties is never
empty unless one would expect it to be empty for dimensional reasons so e. g. the phenomenon
of parallel non-intersecting lines in the plane does not occur in projective space (which we have seen
already in Remark 6.3).
So let X,Y Pn be non-empty projective varieties. Show:
(a) The dimension of the cone C(X) An+1 is dim X + 1.
(b) If dim X + dimY n then X Y 6= 0.
/
6.
55
We have just seen in Remark 6.28 (b) that we can use homogenizations of polynomials to describe
an affine variety X An in terms of their homogeneous coordinates on An Pn . Let us now finish
this chapter by showing that this construction can also be used to compute the closure of X in Pn . As
this will be a compact space in the sense of Remarks 6.3 and 6.4 we can think of this closure X as
being obtained by compactifying X by some points at infinity. For example, if we start with the
affine hyperbola X = Va (x12 x22 1) A2 in the picture below on the left, its closure X P2 adds
the two points a and b at infinity as in Example 6.14.
b
x2
x1
x2
add points
x1
at infinity
X
X
a
We know already in this example that X = Vp (x12 x22 x02 ), i. e. that the closure is the projective
zero locus of the homogenization of the original polynomial x12 x22 1. Let us now prove the
corresponding general statement.
Proposition 6.33 (Computation of the projective closure). Let I EK[x1 , . . . , xn ] be an ideal. Consider
its affine zero locus X = Va (I) An , and its closure X in Pn .
(a) We have X = Vp (I h ).
(b) If I = ( f ) is a principal ideal then X = Vp ( f h ).
Proof.
(a) Clearly, the set Vp (I h ) is closed and contains X: if x = (x1 , . . . , xn ) X then f (x1 , . . . , xn ) = 0
and thus f h (1, x1 , . . . , xn ) = 0 for all f I, which implies that (1 : x1 : : xn ) Vp (I h ).
In order to show that Vp (I h ) is the smallest closed set containing X let Y X be any closed
set; we have to prove that Y Vp (I h ). As Y is closed we have Y = Vp (J) for some homogeneous ideal J. Now any homogeneous element of J can be written as x0d f h for some d N
and f K[x1 , . . . , xn ], and for this element we have
x0d f h is zero on X
f is zero on X
f Ia (X) = Ia (Va (I)) = I
(X is a subset of Y )
(x0 6= 0 on X An )
(Proposition 1.17)
f I for some m N
( f h )m = ( f m )h I h for some m N (Remark 6.27)
f h Ih
x0d f h I h .
We therefore conclude that J I h , and so Y = Vp (J) Vp ( I h ) = Vp (I h ) as desired.

(b) As ( f ) = { f g : g K[x1 , . . . , xn ]}, we have
X = V (( f g)h : g K[x1 , . . . , xn ]) = V ( f h gh : g K[x1 , . . . , xn ]) = V ( f h )
by (a) and Remark 6.27.
Example 6.34. In contrast to Proposition 6.33 (b), for general ideals it usually does not suffice to
only homogenize a set of generators. As an example, consider the ideal I = (x1 , x2 x12 ) E K[x1 , x2 ]
with affine zero locus X = Va (I) = {0} A2 . This one-point set is also closed in P2 , and thus
10
56
Andreas Gathmann
X = {(1 : 0 : 0)} is just the corresponding point in homogeneous coordinates. But if we homogenize
the two given generators of I we obtain the homogeneous ideal (x1 , x0 x2 x12 ) with projective zero
locus {(1 : 0 : 0), (0 : 0 : 1)} ) X.
For those of you who know some computer algebra: one can show however that it suffices to homogenize a Grbner basis of I. This makes the problem of finding X computationally feasible since
in contrast to Proposition 6.33 (a) we only have to homogenize finitely many polynomials.
Exercise 6.35. Sketch the set of real points of the complex affine curve X = V (x13 x1 x22 + 1) A2C
and compute the points at infinity of its projective closure X P2C .
7.
7.
Projective Varieties II: Ringed Spaces
57
After having defined projective varieties as topological spaces, we will now give them the structure
of ringed spaces to make them into varieties in the sense of Chapter 5. In other words, we have to
define a suitable notion of regular functions on (open subsets of) projective varieties.
Of course, as in the affine case in Definition 3.1 the general idea is that a regular function should
be a K-valued function that is locally a quotient of two polynomials. However, note that in contrast
to the affine situation the elements of the homogeneous coordinate ring S(X) of a projective variety
X are not well-defined functions on X: even if f S(X) is homogeneous of degree d we only have
f ( x) = d f (x) for all x X and K. So the only way to obtain well-defined functions is to
consider quotients of homogeneous polynomials of the same degree, so that the factor d cancels
out:
Definition 7.1 (Regular functions on projective varieties). Let U be an open subset of a projective
variety X. A regular function on U is a map : U K with the following property: for every
a U there are d N and f , g S(X)d with f (x) 6= 0 and
(x) =
g(x)
f (x)
for all x in an open subset Ua with a Ua U.

It is obvious that the sets OX (U) of regular functions on U are subrings of the K-algebras of all
functions from U to K, and by the local nature of the definition that they form a sheaf OX on
X.
With this definition, let us check first of all that the open subsets of a projective variety where one
of the coordinates is non-zero are affine varieties, so that projective varieties are prevarieties in the
sense of Definition 5.1.
Proposition 7.2 (Projective varieties are prevarieties). Let X Pn be a projective variety. Then
Ui = {(x0 : : xn ) X : xi 6= 0}
is an affine variety for all i = 0, . . . , n. In particular, X is a prevariety.

Proof. By symmetry it suffices to prove the statement for i = 0. Let X = Vp (h1 , . . . , hr ) for some
homogeneous polynomials h1 , . . . , hr K[x0 , . . . , xn ], and set g j (x1 , . . . , xn ) = h j (1, x1 , . . . , xn ) for all
j = 1, . . . , r. If Y = Va (g1 , . . . , gr ) we claim that
F : Y U0 , (x1 , . . . , xn ) 7 (1 : x1 : : xn )
is an isomorphism with inverse

x1
xn
F : U0 Y, (x0 : : xn ) 7
,...,
.
x0
x0
In fact, it is clear by construction that these two maps are well-defined and inverse to each
other. Moreover, similarly to Remark 6.28 they are continuous: the inverse image of a closed set
Vp ( f1 , . . . , fs ) U0 under F is the closed set Va ( f1 (1, ), . . . , fs (1, )), and the image of a closed set
Va ( f1 , . . . , fs ) Y under F is the closed set Vp ( f1h , . . . , fsh ) U0 .
1
Finally, we have to check that F and F 1 pull back regular functions to regular functions: a regular
p(x0 ,...,xn )
(with nowhere
function on (an open subset of) U0 is by Definition 7.1 locally of the form q(x
0 ,...,xn )
vanishing denominator) for two homogeneous polynomials p and q of the same degree. Then
p(x0 , . . . , xn )
p(1, x1 , . . . , xn )
F
=
q(x0 , . . . , xn )
q(1, x1 , . . . , xn )
58
Andreas Gathmann
is a quotient of polynomials and thus a regular function on Y . Conversely, F 1 pulls back a quotient
p(x1 ,...,xn )
q(x1 ,...,xn ) of two polynomials to

p xx10 , . . . , xx0n
1 p(x1 , . . . , xn )
,
(F )
= x1
q(x1 , . . . , xn )
q x0 , . . . , xxn0
which is a regular function on U0 since it can be rewritten as a quotient of two homogeneous polynomials of the same degree (by multiplying both the numerator and the denominator by x0m for
m = max(deg p, deg q)). Hence F is an isomorphism by Definition 4.3 (b), and so U0 is an affine
open subset of X.
In particular, as the open subsets Ui for i = 0, . . . , n cover X we conclude that X is a prevariety.
Exercise 7.3. Check that Definition 7.1 (together with Proposition 7.2) is compatible with our earlier
constructions in the following cases:
(a) The prevariety P1 is the same as the one introduced in Example 5.5 (a).
(b) If X Pn is a projective variety then its structure sheaf as defined above is the same as the
closed subprevariety structure of X in Pn as in Construction 5.12 (b).
Exercise 7.4. Let m, n N>0 . Use Exercise 6.32 to prove that Pm Pn is not isomorphic to Pm+n .
We have already mentioned that the major advantage of (subprevarieties of) projective varieties is
that they have a global description with homogeneous coordinates that does not refer to gluing techniques. In fact, the following proposition shows that many morphisms between projective varieties
can also be constructed without gluing.
Lemma 7.5 (Morphisms of projective varieties). Let X Pn be a projective variety, and let
f0 , . . . , fm S(X) be homogeneous elements of the same degree. Then on the open subset U :=
X\V ( f0 , . . . , fm ) these elements define a morphism
f : U Pm , x 7 ( f0 (x) : : fm (x)).
Proof. First of all note that f is well-defined set-theoretically: by definition of U the image point
can never be (0 : : 0); and if we rescale the homogeneous coordinates x0 , . . . , xn of x U we get
( f0 ( x0 : : xn ) : : fm ( x0 : : xn ))
= ( d f0 (x0 : : xn ) : : d fm (x0 : : xn ))
= ( f0 (x0 : : xn ) : : fm (x0 : : xn )),
where d is the common degree of the f0 , . . . , fm . To check that f is a morphism we want to use
the gluing property of Lemma 4.6. So let {Vi : i = 0, . . . , m} be the affine open cover of Pm with
Vi = {(y0 : : ym ) : yi 6= 0} for all i. Then the open subsets Ui := f 1 (Vi ) = {x X : fi (x) 6= 0}
cover U, and in the affine coordinates on Vi the map f |Ui is given by the quotients of polynomials
fj
fi for j = 0, . . . , m with j 6= i, which are regular functions on Ui by Definition 7.1. Hence f |Ui is a
morphism by Proposition 4.7, and so f is a morphism by Lemma 4.6.

Example 7.6.
(a) Let A GL(n + 1, K) be an invertible matrix. Then f : Pn Pn , x 7 Ax is a morphism with
inverse f 1 : Pn Pn , x 7 A1 x, and hence an isomorphism. We will refer to these maps
as projective automorphisms of Pn . In fact, we will see in Proposition 13.4 that these are
the only isomorphisms of Pn .
(b) Let a = (1 : 0 : : 0) Pn and L = V (x0 )
= Pn1 . Then the map
f : Pn \{a} Pn1 , (x0 : : xn ) 7 (x1 : : xn )
7.
59
given by forgetting one of the homogeneous coordinates is a morphism by Lemma 7.5. It can
be interpreted geometrically as in the picture below on the left: for x = (x0 : : xn ) Pn \{a}
the unique line through a and x is clearly given parametrically by
{(s :t x1 : :t xn ) : (s :t) P1 },
and its intersection point with L is just (0 : x1 : : xn ), i. e. f (x) with the identification L
=
Pn1 . We call f the projection from a to the linear subspace L. Note however that the picture
below is only schematic and does not show a standard affine open subset Ui = {(x0 : : xn ) :
xi 6= 0}, since none of these subsets contains both a and (a non-empty open subset of) L.
Of course, the same construction works for any point a Pn and any linear subspace L
of dimension n 1 not containing a the corresponding morphism then differs from the
special one considered above by a projective automorphism as in (a).
a
f (x)
f (x)
(b)
L
(c)
(c) The projection morphism f : Pn \{a} Pn1 as in (b) cannot be extended to the point a.
The intuitive reason for this is that the line through a and x (and thus also the point f (x))
does not have a well-defined limit as x approaches a. This changes however if we restrict the
projection to a suitable projective variety: for X = V (x0 x2 x12 ) as in the schematic picture
above on the right consider the map
(
(x1 : x2 ) if (x0 : x1 : x2 ) 6= (1 : 0 : 0),
1
f : X P , (x0 : x1 : x2 ) 7
(x0 : x1 ) if (x0 : x1 : x2 ) 6= (0 : 0 : 1).
It is clearly well-defined since the equation x0 x2 x12 = 0 implies (x1 : x2 ) = (x0 : x1 ) whenever
both these points in P1 are defined. Moreover, it extends the projection as in (b) to all of X
(which includes the point a), and it is a morphism since it is patched together from two
projections as above. Geometrically, the image f (a) is the intersection of the tangent to X at
a with the line L.
This geometric picture also tells us that f is bijective: for every point y L the restriction of
the polynomial x0 x2 x12 defining X to the line through a and y has degree 2, and thus this
line intersects X in two points (counted with multiplicities), of which one is a. The other
point is then the unique inverse image f 1 (y). In fact, it is easy to check that f is even an
isomorphism since its inverse is
f 1 : P1 X, (y0 : y1 ) 7 (y20 : y0 y1 : y21 ),
which is a morphism by Lemma 7.5.
Note that the example of the morphism f above also shows that we cannot expect every
morphism between projective varieties to have a global description by homogeneous polynomials as in Lemma 7.5.
(d) Now let X P2 be any projective conic, i. e. the zero locus of a single irreducible homogeneous polynomial f K[x0 , x1 , x2 ] of degree 2. Assuming that char K 6= 2, we know by
Exercise 4.12 that the affine part X A2 is isomorphic to Va (x2 x12 ) or Va (x1 x2 1) by a
linear transformation followed by a translation. Extending this map to a projective automorphism of P2 as in (a), the projective conic X thus becomes isomorphic to Vp (x0 x2 x12 )
or Vp (x1 x2 x02 ) by Proposition 6.33 (b). So by (c) we see that every projective conic is
isomorphic to P1 .
60
Andreas Gathmann
Exercise 7.7. Let us say that n + 2 points in Pn are in general position if for any n + 1 of them their
representatives in K n+1 are linearly independent.
Now let a1 , . . . , an+2 and b1 , . . . , bn+2 be two sets of points in Pn in general position. Show that there
is an isomorphism f : Pn Pn with f (ai ) = bi for all i = 1, . . . , n + 2.
Exercise 7.8. Show by example that the homogeneous coordinate ring of a projective variety is not
invariant under isomorphisms, i.e. that there are isomorphic projective varieties X,Y such that the
rings S(X) and S(Y ) are not isomorphic.
Exercise 7.9. Let f : Pn Pm be a morphism. Prove:
(a) If X Pm is the zero locus of a single homogeneous polynomial in K[x0 , . . . , xm ] then every
irreducible component of f 1 (X) has dimension at least n 1.
(b) If n > m then f must be constant.
Let us now verify that projective varieties are separated, i. e. that they are varieties and not just
prevarieties. In other words, we have to check that the diagonal X of a projective variety X is
closed in the product X X. By Lemma 5.20 (b) it suffices to show this for X = Pn .
For the proof of this statement it is useful to first find a good description of the product of projective
spaces note that by Exercise 7.4 such products are not just again projective spaces. Of course, we
could just parametrize these products by two sets of homogeneous coordinates. It turns out however
that we can also use a single set of homogeneous coordinates and thus embed products of projective
spaces as a projective variety into a bigger projective space.
Construction 7.10 (Segre embedding). Consider Pm with homogeneous coordinates x0 , . . . , xm and
Pn with homogeneous coordinates y0 , . . . , yn . Set N = (m + 1)(n + 1) 1 and let PN be the projective
space with homogeneous coordinates labeled zi, j for 0 i m and 0 j n. Then there is an
obviously well-defined set-theoretic map
f : Pm Pn PN
11
given by zi, j = xi y j for all i, j. It satisfies the following properties:

Proposition 7.11. Let f : Pm Pn PN be the map of Construction 7.10. Then:
(a) The image X = f (Pm Pn ) is a projective variety given by
X = Vp (zi, j zk,l zi,l zk, j : 0 i, k m, 0 j, l n).
(b) The map f : Pm Pn X is an isomorphism.
X is a projective variety. The isomorphism f : Pm Pn X PN is called
In particular, Pm Pn =
the Segre embedding; the coordinates z0,0 , . . . , zm,n above will be referred to as Segre coordinates
on Pm Pn .
Proof.
(a) It is obvious that the points of f (Pm Pn ) satisfy the given equations. Conversely, consider
a point z PN with homogeneous coordinates z0,0 , . . . , zm,n that satisfy the given equations.
At least one of these coordinates must be non-zero; we can assume without loss of generality
that it is z0,0 . Let us pass to affine coordinates by setting z0,0 = 1. Then we have zi, j = zi,0 z0, j
for all i = 0, . . . , m and j = 0, . . . , n. Hence by setting xi = zi,0 and y j = z0, j (in particular
x0 = y0 = 1) we obtain a point of Pm Pn that is mapped to z by f .
(b) Continuing the above notation, let z X be a point with z0,0 = 1. If f (x, y) = z for some
(x, y) Pm Pn , it follows that x0 6= 0 and y0 6= 0, so we can pass to affine coordinates here
as well and assume that x0 = 1 and y0 = 1. But then it follows that xi = zi,0 and y j = z0, j for
all i and j, i. e. f is injective and thus as a map onto its image also bijective.
The same calculation shows that f and f 1 are given (locally in affine coordinates) by polynomial maps. Hence f is an isomorphism.
7.
61
Example 7.12. According to Proposition 7.11, the product P1 P1 is (isomorphic to) the surface
X = {(z0,0 : z0,1 : z1,0 : z1,1 ) : z0,0 z1,1 = z1,0 z0,1 }
P3
by the isomorphism
f : P1 P1 X, ((x0 : x1 ), (y0 : y1 )) 7 (x0 y0 : x0 y1 : x1 y0 : x1 y1 ).
In particular, the lines {a}P1 and P1 {a} in P1 P1 where the first or second factor is constant,
respectively, are mapped to lines in X P3 . The following schematic picture shows these two
families of lines on the surface X (whose set of real points is a hyperboloid).
P1 P1
X P3
Corollary 7.13. Every projective variety is a variety.

Proof. We have already seen in proposition 7.2 that every projective variety is a prevariety. So by
Lemma 5.20 (b) it only remains to be shown that Pn is separated, i. e. that the diagonal Pn is closed
in Pn Pn . We can describe this diagonal as
Pn = {((x0 : : xn ), (y0 : : yn )) : xi y j x j yi = 0 for all i, j},
because these equations mean exactly that the matrix

x0 x1
y0 y1
xn
yn
has rank (at most) 1, i. e. that (x0 : : xn ) = (y0 : : yn ). In particular, it follows that Pn is closed
as the zero locus of the homogeneous linear polynomials zi, j z j,i in the Segre coordinates zi, j = xi y j
of Pn Pn .

Remark 7.14. If X Pm and Y Pn are projective varieties then X Y is a closed subset of Pm Pn .
As the latter is a projective variety by the Segre embedding we see that X Y is a projective variety
as well (namely a projective subvariety of Pm Pn ).
Exercise 7.15. Let X P2 be a curve given as the zero locus of a homogeneous polynomial of degree 3. Moreover, let U X X be the set of all
(a, b) X X such that a 6= b and the unique line through a and b meets
X in exactly three distinct points. Of course, two of these points are then a
and b; we will denote the third one by (a, b) X.
Show that U X X is open, and that : U X is a morphism.
b
a
(a, b)
Exercise 7.16.
(a) Prove that for every projective variety Y Pn of pure dimension n1 there is a homogeneous
polynomial f such that I(Y ) = ( f ). You may use the commutative algebra fact that every
polynomial in K[x0 , . . . , xn ] admits a unique decomposition into prime elements [G5, Remark
8.6].
(b) If X is a projective variety of dimension n, show by example that in general not every projective variety Y X of dimension n 1 is of the form V ( f ) for a homogeneous polynomial
f S(X). (One possibility is to consider the Segre embedding X of P1 P1 in P3 , and
Y = P1 {0} P1 P1 .)
62
Andreas Gathmann
The most important property of projective varieties is that they are compact in the classical topology
if the ground field is C. We have seen this already for projective spaces in Remark 6.4, and it then
follows for projective varieties as well since they are closed subsets of them. However, Exercises
2.34 (c) and 5.11 (a) show unfortunately that every prevariety is compact in the Zariski topology, and
so in particular that compactness in the Zariski topology does not capture the same geometric idea
as in the classical case. We therefore need an alternative description of the intuitive compactness
property that works in our algebraic setting of the Zariski topology.
The key idea to achieve this is that compact sets should be mapped to compact sets again under
continuous maps. In our language, this means that images of morphisms between projective varieties
should be closed. This property (that we have already seen to be false for general varieties in Remark
5.15 (a)) is what we want to prove now. We start with a special case which contains all the hard work,
and from which the general case will then follow easily.
Definition 7.17 (Closed maps). A map f : X Y between topological spaces is called closed if
f (A) Y is closed for every closed subset A X.
Proposition 7.18. The projection map : Pn Pm Pm is closed.
Proof. Let Z Pn Pm be a closed set. By Remark 6.20 we can write Z = V ( f1 , . . . , fr ) for homogeneous polynomials f1 , . . . , fr of the same degree d in the Segre coordinates of Pn Pm , i. e. for
bihomogeneous polynomials of degree d in both the coordinates x0 , . . . , xn of Pn and y0 , . . . , ym of
Pm . Now consider a fixed point a Pm ; we will determine if it is contained in the image (Z). To
do this, let gi = fi ( , a) K[x0 , . . . , xn ] for i = 1, . . . , r. Then
a
/ (Z) there is no x Pn with (x, a) Z
Vp (g1 , . . . , gr ) = 0/
p
p
(g1 , . . . , gr ) = (1) or (g1 , . . . , gr ) = (x0 , . . . , xn ) (Proposition 6.22)
there are ki N with xiki (g1 , . . . , gr ) for all i
K[x0 , . . . , xn ]k (g1 , . . . , gr ) for some k N,
where as usual K[x0 , . . . , xn ]k denotes the homogeneous degree-k part of the polynomial ring as in
Definition 6.6, and the direction of the last equivalence follows by setting k = k0 + + kn .
Of course, the last condition can only be satisfied if k d and is equivalent to K[x0 , . . . , xn ]k =
(g1 , . . . , gr )k . As (g1 , . . . , gr ) = {h1 g1 + + hr gr : h1 , . . . , hr K[x0 , . . . , xn ]} this is the same as
saying that the K-linear map
Fk : (K[x0 , . . . , xn ]kd )r K[x0 , . . . , xn ]k , (h1 , . . . , hr ) 7 h1 g1 + hr gr

is surjective, i. e. has rank dimK K[x0 , . . . , xn ]k = n+k
for some k d. This in turn is the case if
k
n+k
and only if at least one of the minors of size k of a matrix for some Fk is non-zero. But these
minors are polynomials in the coefficients of g and thus in the coordinates of a, and consequently
the non-vanishing of one of these minors is an open condition in the Zariski topology of Pm .
Hence the set of all a Pm with a
/ (Z) is open, which means that (Z) is closed.
Remark 7.19. Let us look at Proposition 7.18 from an algebraic point of view. We start with some
equations f1 (x, y) = = fr (x, y) = 0 in two sets of variables x = (x0 , . . . , xn ) and y = (y0 , . . . , ym )
and ask for the image of their common zero locus under the projection map (x, y) 7 x. The equations
satisfied on this image are precisely the equations in x alone that can be derived from the given ones
f1 (x, y) = = fr (x, y) = 0 in x and y. In other words, we want to eliminate the variables y from
the given system of equations. The statement of Proposition 7.18 is therefore sometimes called the
main theorem of elimination theory.
Corollary 7.20. The projection map : Pn Y Y is closed for any variety Y .
7.
63
Proof. Let us first show the statement for an affine variety Y Am . Then we can regard Y as a
locally closed subvariety of Pm via the embedding Am Pm . Now let Z Pn Y be closed, and let
Z be its closure in Pn Pm . If : Pn Pm Pm is the projection map then (Z) is closed in Pm by
Proposition 7.18, and thus
(Z) = (Z (Pn Y )) = (Z) Y
is closed in Y .
If Y is any variety we can cover it by affine open subsets. As the condition that a subset is closed
can be checked by restricting it to the elements of an open cover, the statement follows from the
corresponding one for the affine open patches that we have just shown.

It is in fact this property of Corollary 7.20 that captures the classical idea of compactness. Let us
therefore give it a name:
Definition 7.21 (Complete varieties). A variety X is called complete if the projection map :
X Y Y is closed for any variety Y .
Example 7.22.
(a) Pn is complete by Corollary 7.20.
(b) Any closed subvariety X 0 of a complete variety X is complete: if Z X 0 Y is closed then
Z is also closed in X Y , and hence its image under the second projection to Y is closed as
well. In particular, by (a) this means that every projective variety is complete.
(c) A1 is not complete: as in the picture below on the left, the image (Z) of the closed subset
Z = V (x1 x2 1) A1 A1 under the second projection is A1 \{0}, which is not closed.
Z
A 1 A1
A1
P1 A1
A1
The geometric reason for this is that A1 is missing a point at infinity: if we replace A1 by P1
as in the picture on the right there is an additional point in the closure Z of Z A1 A1 in
P1 A1 ; the image of this point under fills the gap and makes (Z) a closed set. Intuitively,
one can think of the name complete as coming from the geometric idea that it contains all
the points at infinity that are missing in affine varieties.
Remark 7.23. There are complete varieties that are not projective, but this is actually quite hard to
show we will certainly not meet such an example in this course. So for practical purposes you
can usually assume that the terms projective variety and complete variety are synonymous.
In any case, complete varieties now have the property that we were aiming for:
Corollary 7.24. Let f : X Y be a morphism of varieties. If X is complete then its image f (X) is
closed in Y .
Proof. By Proposition 5.21 (a) the graph f X Y is closed. But then f (X) = ( f ) for the
projection map : X Y Y , which is closed again since X is complete.

Let us conclude this chapter with two applications of this property.
Corollary 7.25. Let X be a connected complete variety. Then OX (X) = K, i. e. every global regular
function on X is constant.
64
Andreas Gathmann
Proof. A global regular function OX (X) determines a morphism : X A1 . By extension of

the target we can consider this as a morphism : X P1 = A1 {} whose image (X) P1 does
not contain the point . But (X) is also closed by Corollary 7.24 since X is complete, and hence
it must be a finite set since these are the only closed proper subsets of P1 . Moreover, Exercise 2.21
(b) implies that (X) is connected since X is. Altogether this means that (X) is a single point, i. e.
that is constant.
12
Remark 7.26. Corollary 7.25 is another instance of a result that has a counterpart in complex analysis: it can be shown that every holomorphic function on a connected compact complex manifold is
constant.
Construction 7.27 (Veronese embedding). Choose n, d N>0 , and let f0 , . . . , fN K[x0 , . . . , xn ]
for N = n+d
n 1 be the set of all monomials of degree d in the variables x0 , . . . , xn , in any order.
Consider the map
F : Pn PN , x 7 ( f0 (x) : : fN (x)).
By Lemma 7.5 this is a morphism (note that the monomials x0d , . . . , xnd , which cannot be simultaneously zero, are among the f0 , . . . , fN ). So by Corollary 7.24 the image X = F(Pn ) is a projective
variety.
We claim that F : Pn X is an isomorphism. All we have to do to prove this is to find an inverse
morphism. This is not hard: we can do this on an affine open cover, so let us e. g. consider the open
subset where x0 6= 0, i. e. x0d 6= 0. On this set we can pass to affine coordinates and set x0 = 1. The
inverse morphism is then given by xi =
monomials.
xi x0d1
x0d
for i = 1, . . . , n, which is a quotient of two degree-d
The morphism F is therefore an isomorphism and thus realizes Pn as a subvariety X of PN . It

is usually called the degree-d Veronese embedding; the coordinates on PN are called Veronese
coordinates of Pn
= X. Of course, this embedding can also be restricted to any projective variety
Y Pn and then gives an isomorphism by degree-d polynomials between Y and a projective variety
in PN .
The importance of the Veronese embedding lies in the fact that degree-d polynomials in the coordinates of Pn are translated into linear polynomials in the Veronese coordinates. An example where
this is useful will be given in Corollary 7.30.
Example 7.28.
(a) For d = 1 the Veronese embedding of Pn is just the identity Pn Pn .
(b) For n = 1 the degree-d Veronese embedding of P1 in Pd is
F : P1 Pd , (x0 : x1 ) 7 (x0d : x0d1 x1 : : x0 x1d1 : x1d ).
In the d = 2 case we have already seen in Example 7.6 (c) that this is an isomorphism.
Exercise 7.29.
Let F : Pn PN be the degree-d Veronese embedding as in Construction 7.27, with
n+d
N = n 1. By applying Corollary 7.24 we have seen already that the image X = F(Pn ) is a
projective variety. Find explicit equations describing X, i. e. generators for a homogeneous ideal I
such that X = V (I).
Corollary 7.30. Let X Pn be a projective variety, and let f S(X) be a non-zero homogeneous
element. Then X\V ( f ) is an affine variety.
Proof. If f = x0 this is just Proposition 7.2. For a general linear polynomial f the statement follows
from this after a projective automorphism as in Example 7.6 (a) that takes f to x0 , and if f is of
degree d > 1 we can reduce the claim to the linear case by first applying the degree-d Veronese
embedding of Construction 7.27.

Exercise 7.31. Recall from Example 7.6 (d) that a conic in P2 over a field of characteristic not equal
to 2 is the zero locus of an irreducible homogeneous polynomial of degree 2 in K[x0 , x1 , x2 ].
7.
65
(a) Considering the coefficients of such polynomials, show that the set of all conics in P2 can be
identified with an open subset U of the projective space P5 .
(b) Let a P2 . Show that the subset of U consisting of all conics passing through a is the zero
locus of a linear equation in the homogeneous coordinates of U P5 .
(c) Given 5 points in P2 , no three of which lie on a line, show that there is a unique conic in P5
passing through all these points.
Exercise 7.32. Let X P3 be the degree-3 Veronese embedding of P1 , i. e. the image of the morphism
P1 P3 , (x0 : x1 ) 7 (y0 : y1 : y2 : y3 ) = (x03 : x02 x1 : x0 x12 : x13 ).
Moreover, let a = (0 : 0 : 1 : 0) P3 and L = V (y2 ) P3 , and consider the projection f from a to L as
in Example 7.6 (b).
(a) Show that f is a morphism.
(b) Determine an equation of the curve f (X) in L
= P2 .
(c) Is f : X f (X) an isomorphism onto its image?
66
8.
Andreas Gathmann
Grassmannians
After having introduced (projective) varieties the main objects of study in algebraic geometry
let us now take a break in our discussion of the general theory to construct an interesting and
useful class of examples of projective varieties. The idea behind this construction is simple: since
the definition of projective spaces as the sets of 1-dimensional linear subspaces of K n turned out to
be a very useful concept, let us now generalize this and consider instead the sets of k-dimensional
linear subspaces of K n for an arbitrary k = 0, . . . , n.
Definition 8.1 (Grassmannians). Let n N>0 , and let k N with 0 k n. We denote by G(k, n)
the set of all k-dimensional linear subspaces of K n . It is called the Grassmannian of k-planes in K n .
Remark 8.2. By Example 6.12 (b) and Exercise 6.32 (a), the correspondence of Remark 6.17 shows
that k-dimensional linear subspaces of K n are in natural one-to-one correspondence with (k 1)dimensional linear subspaces of Pn1 . We can therefore consider G(k, n) alternatively as the set
of such projective linear subspaces. As the dimensions k and n are reduced by 1 in this way, our
Grassmannian G(k, n) of Definition 8.1 is sometimes written in the literature as G(k 1, n 1)
instead.
Of course, as in the case of projective spaces our goal must again be to make the Grassmannian
G(k, n) into a variety in fact, we will see that it is even a projective variety in a natural way. For
this we need the algebraic concept of alternating tensor products, a kind of multilinear product on
K n generalizing the well-known cross product

a1
b1
a2 b3 a3 b2
a2 b2 = a3 b1 a1 b3
a3
b3
a1 b2 a2 b1
on K 3 whose coordinates are all the 2 2 minors of the matrix

a1 a2 a3
.
b1 b2 b3
If you have seen ordinary tensor products in commutative algebra already [G5, Chapter 5], you
probably know that the best way to introduce these products is by a universal property similar to
the one for products of varieties in Definition 5.16. Although the same is true for our alternating
tensor products, we will follow a faster and more basic approach here, whose main disadvantage is
that it is not coordinate-free. Of course, if you happen to know the better definition of alternating
tensor products using their universal property already, you can use this definition as well and skip
the following construction.
Construction 8.3 (Alternating tensor products). Let (e1 , . . . , en) denote the standard basis of K n . For
k N we define k K n to be a K-vector space of dimension nk with basis vectors formally written
as
ei1 ei2 eik
()
for all multi-indices (i1 , . . . , ik ) of natural numbers with 1 i1 < i2 < < ik n. Note that the set
of these strictly increasing multi-indices is in natural bijection
with the set of all k-element subsets
{i1 , . . . , ik } of {1, . . . , n}, so that there are in fact exactly nk of these basis vectors. In particular,
k K n is the zero vector space if k > n.
We extend the notation () to arbitrary (i. e. not strictly increasing) multi-indices (i1 , . . . , ik ) with
1 i1 , . . . , ik n by setting ei1 eik := 0 if any two of the i1 , . . . , ik coincide, and
ei1 eik := sign ei (1) ei (k)
8.
Grassmannians
67
if all i1 , . . . , ik are distinct, and is the unique permutation of {1, . . . , k} such that i (1) < < i (k) .
We can then extend this notation multilinearly to a product (K n )k k K n : for v1 , . . . , vk K n with
basis expansions v j = ni=1 a j,i ei for some a j,i K we define
v1 vk :=
a1,i1 ak,ik ei1 eik
k K n .
i1 ,...,ik
More generally, we obtain bilinear and associative products k K n l K n k+l K n by a bilinear

extension of
(ei1 eik ) (e j1 e jl ) := ei1 eik e j1 e jl .
The vector space k K n is usually called the k-fold alternating or antisymmetric tensor product of
K n , the elements of k K n are referred to as alternating or antisymmetric tensors.
Example 8.4.
(a) By definition we have 0 = K and 1 K n = K n ; a basis of 1 K n is again (e1 , . . . , en ). We also
have n K n
= K, with single basis vector e1 en .
(b) As in (a), 2 K 2 is isomorphic to K with basis vector e1 e2 . For two arbitrary vectors
v = a1 e1 + a2 e2 and w = b1 e1 + b2 e2 of K 2 their alternating tensor product is
v w = a1 b1 e1 e1 + a1 b2 e1 e2 + a2 b1 e2 e1 + a2 b2 e2 e2
= (a1 b2 a2 b1 ) e1 e2 ,
so under the isomorphism 2 K 2
= K it is just the determinant of the coefficient matrix of v
and w.
(c) Similarly, for v = a1 e1 + a2 e2 + a3 e3 and w = b1 e1 + b2 e2 + b3 e3 in K 3 we have
v w = (a1 b2 b2 a1 ) e1 e2 + (a1 b3 b3 a1 ) e1 e3 + (a2 b3 a3 b2 ) e2 e3
2 K 3
= K3,
so (up to a simple change of basis) v w is just the cross product v w considered in the
introduction to this chapter.
As we will see now, it is in fact a general phenomenon that the coordinates of alternating tensor
products can be interpreted as determinants.
Remark 8.5 (Alternating tensor products and determinants). Let 0 k n, and let v1 , . . . , vk K n
with basis expansions v j = i a j,i ei for j = 1, . . . , k. For a strictly increasing multi-index ( j1 , . . . , jk )
let us determine the coefficient of the basis vector e j1 e jk in the tensor product v1 vk .
As in Construction 8.3 we have
v1 vk =
a1,i1 ak,ik ei1 eik .
i1 ,...,ik
Note that the indices i1 , . . . , ik in the products ei1 eik in the terms of this sum are not necessarily
in strictly ascending order. So to figure out the coefficient of e j1 e jk in v1 vk we have to
sort the indices in each sum first; the resulting coefficient is then
sign a1, j (1) ak, j (k) ,

where the sum is taken over all permutations . By definition this is exactly the determinant of the
maximal quadratic submatrix of the coefficient matrix (ai, j )i, j obtained by taking only the columns
j1 , . . . , jk . In other words, the coordinates of v1 vk are just all the maximal minors of the matrix
whose rows are v1 , . . . , vk . So the alternating tensor product can be viewed as a convenient way to
encode all these minors in a single object.
As a consequence, alternating tensor products can be used to encode the linear dependence and linear
spans of vectors in a very elegant way.
Lemma 8.6. Let v1 , . . . , vk K n for some k n. Then v1 vk = 0 if and only if v1 , . . . , vk are
linearly dependent.
68
Andreas Gathmann
Proof. By Remark 8.5, we have v1 vk = 0 if and only if all maximal minors of the matrix with
rows v1 , . . . , vk are zero. But this is the case if and only if this matrix does not have full rank [G2,
Exercise 17.25], i. e. if and only if v1 , . . . , vk are linearly dependent.

Remark 8.7.
(a) By construction, the alternating tensor product is antisymmetric in the sense that for all
v1 , . . . , vk K n and all permutations we have
v1 vk = sign v (1) v (k) .
Moreover, Lemma 8.6 tells us that v1 vk = 0 if two of the vectors v1 , . . . , vk coincide.
(b) We have constructed the alternating tensor product using a fixed basis e1 , . . . , en of K n . However, if v1 , . . . , vn is an arbitrary basis of K n it is easy to see that the alternating tensors
vi1 vik for strictly increasing multi-indices (i1 , . . . , ik ) form a basis of k K n as well:
there are nk of these vectors, and they generate k K n since every standard unit vector ei is
a linear combination of v1 , . . . , vn , and hence every k-fold alternating product ei1 eik is
a linear combination of k-fold alternating products of v1 , . . . , vn which can be expressed
by (a) in terms of such products with strictly increasing indices.
Lemma 8.8. Let v1 , . . . , vk K n and w1 , . . . , wk K n both be linearly independent. Then v1 vk
and w1 wk are linearly dependent in k K n if and only if Lin(v1 , . . . , vk ) = Lin(w1 , . . . , wk ).
Proof. As we have assumed both v1 , . . . , vk and w1 , . . . , wk to be linearly independent, we know by
Lemma 8.6 that v1 vk and w1 wk are both non-zero.
Assume that Lin(v1 , . . . , vk ) 6= Lin(w1 , . . . , wk ), so without loss of generality that w1
/
Lin(v1 , . . . , vk ). Then w1 , v1 , . . . , vk are linearly independent, and thus w1 v1 vk 6= 0
by Lemma 8.6. But by assumption we know that v1 vk = w1 wk for some
K, and hence
0 6= w1 v1 vk = w1 w1 wk
in contradiction to Remark 8.7 (a).
If v1 , . . . , vk and w1 , . . . , wk span the same subspace of K n then the basis w1 , . . . , wk of this
subspace can be obtained from v1 , . . . , vk by a finite sequence of basis exchange operations
vi vi + v j and vi vi for K and i 6= j. But both these operations change the
alternating product of the vectors at most by a multiplicative scalar, since
v1 vi1 (vi + v j ) vi+1 vn = v1 vi vn
and
v1 ( vi ) vn = v1 vn
by multilinearity and Remark 8.7 (a).

We can now use our results to realize the Grassmannian G(k, n) as a subset of a projective space.
n
Construction 8.9 (Plcker embedding). Let 0 k n, and consider the map f : G(k, n) P(k)1
n
given by sending a linear subspace Lin(v1 , . . . , vk ) G(k, n) to the class of v1 vk k K n
= K (k )
n
in P(k)1 . Note that this is well-defined: v1 vk is non-zero by Lemma 8.6, and representing
n
the same subspace by a different basis does not change the resulting point in P(k)1 by the part
of Lemma 8.8. Moreover, the map f is injective by the part of Lemma 8.8. We call it the
Plcker embedding of G(k, n); for a k-dimensional linear subspace L G(k, n) the (homogeneous)
n
coordinates of f (L) in P(k)1 are the Plcker coordinates of L. By Remark 8.5, they are just all the
maximal minors of the matrix whose rows are v1 , . . . , vk .
n
In the following, we will always consider G(k, n) as a subset of P(k)1 using this Plcker embedding.
8.
Grassmannians
69
Example 8.10.
(a) The Plcker embedding of G(1, n) simply maps a linear subspace Lin(a1 e1 + + an en ) to
n
the point (a1 : : an ) P(1)1 = Pn1 . Hence G(1, n) = Pn1 as expected.
(b) Consider the 2-dimensional subspace L = Lin(e1 + e2 , e1 + e3 ) G(2, 3) of K 3 . As
(e1 + e2 ) (e1 + e3 ) = e1 e2 + e1 e3 + e2 e3 ,
3
the coefficients (1 : 1 : 1) of this vector are the Plcker coordinates of L in P(2)1 = P2 .
Alternatively, these are the three maximal minors of the matrix

1 1 0
1 0 1
whose rows are the given spanning vectors e1 + e2 and e1 + e3 of L. Note that a change of
these spanning vectors will just perform row operations on this matrix, which changes the
maximal minors at most by a common constant factor. This shows again in this example that
the homogeneous Plcker coordinates of L are well-defined.
So far we have embedded the Grassmannian G(k, n) into a projective space, but we still have to see
that it is a closed subset, i. e. a projective variety. So by Construction 8.9 we have to find suitable
equations describing the alternating tensors in k K n that can be written as a so-called pure tensor, i. e.
as v1 vk for some v1 , . . . , vk K n and not just as a linear combination of such expressions.
The key lemma to achieve this is the following.
Lemma 8.11. For a fixed non-zero k K n with k < n consider the K-linear map
f : K n k+1 K n , v 7 v .
Then rk f n k, with equality holding if and only if = v1 vk for some v1 , . . . , vk K n .
Example 8.12. Let k = 2 and n = 4.
(a) For = e1 e2 the map f of Lemma 8.11 is given by
f (a1 e1 + a2 e2 + a3 e3 + a4 e4 ) = (a1 e1 + a2 e2 + a3 e3 + a4 e4 ) e1 e2
= a3 e1 e2 e3 + a4 e1 e2 e4 ,
for a1 , a2 , a3 , a4 K, and thus has rank rk f = 2 = n k in accordance with the statement of
the lemma.
(b) For = e1 e2 + e3 e4 we get
f (a1 e1 + a2 e2 + a3 e3 + a4 e4 )
= (a1 e1 + a2 e2 + a3 e3 + a4 e4 ) (e1 e2 + e3 e4 )
= a1 e1 e3 e4 + a2 e2 e3 e4 + a3 e1 e2 e3 + a4 e1 e2 e4
instead, so that rk f = 4. Hence Lemma 8.11 tells us that there is no way to write as a pure
tensor v1 v2 for some vectors v1 , v2 K 4 .
Proof of Lemma 8.11. Let v1 , . . . , vr be a basis of ker f (with r = n rk f ), and extend it to a basis
v1 , . . . , vn of K n . By Remark 8.7 (b) the alternating tensors vi1 vik with 1 i1 < < ik n
then form a basis of k K n , and so we can write
=
ai1 ,...,ik vi1 vik
i1 <<ik
for suitable coefficients ai1 ,...,ik K. Now for i = 1, . . . , r we know that vi ker f , and thus
0 = vi =
ai1 ,...,ik vi vi1 vik .
()
i1 <<ik
Note that vi vi1 vik = 0 if i {i1 , . . . , ik }, and in the other cases these products are (up to
sign) different basis vectors of k+1 K n . So the equation () tells us that we must have ai1 ,...,ik = 0
13
70
Andreas Gathmann
whenever i
/ {i1 , . . . , ik }. As this holds for all i = 1, . . . , r we conclude that the coefficient ai1 ,...,ik = 0
can only be non-zero if {1, . . . , r} {i1 , . . . , ik }.
But at least one of these coefficients has to be non-zero since 6= 0 by assumption. This obviously
requires that r k, i. e. that rk f = n r n k. Moreover, if we have equality then only the
coefficient a1,...,k can be non-zero, which means that is a scalar multiple of v1 vk .
Conversely, if = w1 wk for some (necessarily linearly independent) w1 , . . . , wk K n then
w1 , . . . , wk ker f . Hence in this case dim ker f k, i. e. rk f n k, and together with the above
result rk f n k we have equality.

Corollary 8.13 (G(k, n) as a projective variety). With the Plcker embedding of Construction 8.9,
n
the Grassmannian G(k, n) is a closed subset of P(k)1 . In particular, it is a projective variety.
Proof. As G(n, n) is just a single point (and hence clearly a variety) we may assume that k < n.
n
Then by construction a point P(k)1 lies in G(k, n) if and only if it is the class of a pure tensor
v1 vk . Lemma 8.11 shows that this is the case if and only if the rank of the linear map
f : K n k+1 K n , v 7 v is n k. As we also know that the rank of this map is always at least
n k, this condition can be checked by the vanishing of all (n k + 1) (n k + 1) minors of the
matrix corresponding to f [G2, Exercise 17.25]. But these minors are polynomials in the entries of
this matrix, and thus in the coordinates of . Hence we see that the condition for to be in G(k, n)
is closed.

Example 8.14. By the proof of Corollary 8.13, the Grassmannian G(2, 4) is given by the vanishing
of all sixteen 3 3 minors of a 4 4 matrix corresponding to a linear map K 4 3 K 4 , i. e. it is a
4
subset of P(2)1 = P5 given by 16 cubic equations.
As you might expect, this is by no means the simplest set of equations describing G(2, 4) in fact,
we will see in Exercise 8.19 (a) that a single quadratic equation suffices to cut out G(2, 4) from P5 .
Our proof of Corollary 8.13 is just the easiest way to show that G(k, n) is a variety; it is not suitable
in practice to find a nice description of G(k, n) as a zero locus of simple equations.
However, there is another useful description of the Grassmannian in terms of affine patches, as we
will see now. This will then also allow us to easily read off the dimension of G(k, n) which would
be very hard to compute from its equations as in Corollary 8.13.
n
Construction 8.15 (Affine cover of the Grassmannian). Let U0 G(k, n) P(k)1 be the affine
open subset where the e1 ek -coordinate is non-zero. Then by Remark 8.5 a linear subspace
L = Lin(v1 , . . . , vk ) G(k, n) is in U0 if and only if the k n matrix A with rows v1 , . . . , vk is of the
form A = (B |C) for an invertible k k matrix B and an arbitrary k (n k) matrix C. This in turn
is the case if and only if A is equivalent by row transformations, i. e. by a change of basis for L, to
a matrix of the form (Ek | D), where Ek denotes the k k unit matrix and D Mat(k (n k), K):
namely by multiplying A with B1 from the left to obtain (Ek | D) with D = B1C. Note that this is
in fact the only choice for D, so that we get a bijection
f : Ak(nk) = Mat(k (n k), K) U0 ,

D 7 the linear subspace spanned by the rows of (Ek | D).
As the Plcker coordinates of this subspace, i. e. the maximal minors of (Ek | D), are clearly polynomial functions in the entries of D, we see that f is a morphism. Conversely, the (i, j)-entry of D can
be reconstructed (up to sign) from f (D) as the maximal minor of (Ek | D) where we take all columns
of Ek except the i-th, together with the j-th column of D. Hence f 1 is a morphism as well, showing
that f is an isomorphism and thus U0
= Ak(nk) is an affine space (and not just an affine variety,
which is already clear from Proposition 7.2).
Of course, this argument holds in the same way for all other affine patches where one of the Plcker
coordinates is non-zero. Hence we conclude:
Corollary 8.16. G(k, n) is an irreducible variety of dimension k(n k).
8.
Grassmannians
71
Proof. We have just seen in Construction 8.15 that G(k, n) has an open cover by affine spaces
Ak(nk) . As any two of these patches have a non-empty intersection (it is in fact easy to write
down a k n matrix such that any two given maximal minors are non-zero), the result follows from
Exercises 2.20 (b) and 2.33 (a).

Remark 8.17. The argument of Construction 8.15 also shows an alternative description of the Grassmannian: it is the space of all full-rank k n matrices modulo row transformations. As we know
that every such matrix is equivalent modulo row transformations to a unique matrix in reduced row
echelon form, we can also think of G(k, n) as the set of full-rank k n matrices in such a form. For
example, in the case k = 1 and n = 2 (when G(1, 2) = P1 by Example 8.10 (a)) the full-rank 1 2
matrices in reduced row echelon form are
( 1 ) corresponding to A1 P1
and
( 0 1 ) corresponding to P1
as in the homogeneous coordinates of P1 .

The affine cover of Construction 8.15 can also be used to show the following symmetry property of
the Grassmannians.
Proposition 8.18. For all 0 k n we have G(k, n)
= G(n k, n).
Proof. There is an obvious well-defined set-theoretic bijection f : G(k, n) G(n k, n) that sends
a k-dimensional linear subspace L of K n to its orthogonal complement
L = {x K n : hx, yi = 0 for all y L},
where hx, yi = ni=1 xi yi denotes the standard bilinear form. It remains to be shown that f (and analogously f 1 ) is a morphism. By Lemma 4.6, we can do this on the affine coordinates of Construction
8.15. So let L G(k, n) be described as the subspace spanned by the rows of a matrix (Ek | D), where
the entries of D Mat(k (n k), K) are the affine coordinates of L. As

D
(Ek | D)
= 0,
Enk
we see that L is the subspace spanned by the rows of (DT | Enk ). But the maximal minors of this
matrix, i. e. the Plcker coordinates of L , are clearly polynomials in the entries of D, and thus we
conclude that f is a morphism.

Exercise 8.19. Let G(2, 4) P5 be the Grassmannian of lines in P3 (or of 2-dimensional linear
subspaces of K 4 ). We denote the homogeneous Plcker coordinates of G(2, 4) in P5 by xi, j for
1 i < j 4. Show:
(a) G(2, 4) = V (x1,2 x3,4 x1,3 x2,4 + x1,4 x2,3 ).
(b) Let L P3 be an arbitrary line. Show that the set of lines in P3 that intersect L, considered
as a subset of G(2, 4) P5 , is the zero locus of a homogeneous linear polynomial.
How many lines in P3 would you expect to intersect four general given lines?
Exercise 8.20. Show that the following sets are projective varieties:
(a) the incidence correspondence
{(L, a) G(k, n) Pn1 : L Pn1 a (k 1)-dimensional linear subspace and a L};
(b) the join of two disjoint varieties X,Y Pn , i. e. the union in Pn of all lines intersecting both
X and Y .
72
9.
Andreas Gathmann
Birational Maps and Blowing Up
In the course of this class we have already seen many examples of varieties that are almost the
same in the sense that they contain isomorphic dense open subsets (although the varieties are not
isomorphic themselves). Let us quickly recall some of them.
Example 9.1 (Irreducible varieties with isomorphic non-empty open subsets).
(a) The affine space An and the projective space Pn have the common open subset An by Proposition 7.2. Consequently, Pm Pn and Pm+n have the common open subset Am An = Am+n
but they are not isomorphic by Exercise 7.4.
(b) Similarly, the affine space Ak(nk) and the Grassmannian G(k, n) have the common open
subset Ak(nk) by Construction 8.15.
(c) The affine line A1 and the curve X = V (x12 x23 ) A2 of Example 4.9 have the isomorphic
open subsets A1 \{0} resp. X\{0} in fact, the morphism f given there is an isomorphism
after removing the origin from both the source and the target curve.
We now want to study this situation in more detail and present a very general construction the
so-called blow-ups that gives rise to many examples of this type. But first of all we have to set
up some notation to deal with morphisms that are defined on dense open subsets. For simplicity,
we will do this only for the case of irreducible varieties, in which every non-empty open subset is
automatically dense by Remark 2.18.
Definition 9.2 (Rational maps). Let X and Y be irreducible varieties. A rational map f from X to
Y , written f : X 99K Y , is a morphism f : U Y (denoted by the same letter) from a non-empty open
subset U X to Y . We say that two such rational maps f1 : U1 Y and f2 : U2 Y with U1 ,U2 X
are the same if f1 = f2 on a non-empty open subset of U1 U2 .
Remark 9.3. Strictly speaking, Definition 9.2 means that a rational map f : X 99K Y is an equivalence class of morphisms from non-empty open subsets of X to Y . Note that the given relation is in
fact an equivalence relation: reflexivity and symmetry are obvious, and if f1 : U1 Y agrees with
f2 : U2 Y on a non-empty open subset U1,2 and f2 with f3 : U3 Y on a non-empty open subset
U2,3 then f1 and f3 agree on U1,2 U1,3 , which is again non-empty by Remark 2.18 (a) since X is
irreducible. For the sake of readability it is customary however not to indicate these equivalence
classes in the notation and to denote the rational map f : X 99K Y and the morphism f : U Y by
the same letter.
If we now want to consider rational maps with an inverse, i. e. rational maps f : X 99K Y such that
there is another rational map g : Y 99K X with g f = idX and f g = idY , we run into problems: if
e. g. f is a constant map and g is not defined at the point f (X) then there is no meaningful way to
compose it with f . So we need to impose a technical condition first to ensure that compositions are
well-defined:
Definition 9.4 (Birational maps). Again let X and Y be irreducible varieties.
(a) A rational map f : X 99K Y is called dominant if its image contains a non-empty open subset
U of Y . In this case, if g : Y 99K Z is another rational map, defined on a non-empty open
subset V of Y , we can construct the composition g f : X 99K Z as a rational map since we
have such a composition of ordinary morphisms on the non-empty open subset f 1 (U V ).
(b) A rational map f : X 99K Y is called birational if it is dominant, and if there is another
dominant rational map g : Y 99K X with g f = idX and f g = idY .
(c) We say that X and Y are birational if there is a birational map f : X 99K Y between them.
9.
73
Remark 9.5. By definition, two irreducible varieties are birational if and only if they contain isomorphic non-empty open subsets. In particular, Exercise 5.25 then implies that birational irreducible
varieties have the same dimension.
An important case of rational maps is when the target space is just the ground field, i. e. if we consider
regular functions on open subsets.
Construction 9.6 (Rational functions and function fields). Let X be an irreducible variety.
A rational map : X 99K A1 = K is called a rational function on X. In other words, a rational
function on X is given by a regular function OX (U) on some non-empty open subset U X,
with two such regular functions defining the same rational function if and only if they agree on a
non-empty open subset. The set of all rational functions on X will be denoted K(X).
Note that K(X) is a field: for 1 OX (U1 ) and 2 OX (U2 ) we can define 1 + 2 and 1 2
on U1 U2 6= 0,
/ the additive inverse 1 on U1 , and for 1 6= 0 the multiplicative inverse 11 on
U1 \V (1 ). We call K(X) the function field of X.
Remark 9.7. If U X is a non-empty open subset of an irreducible variety X then K(U)
= K(X):
an isomorphism is given by
K(U) K(X)
OU (V ) 7 OX (V )
with inverse
K(X) K(U)
OX (V ) 7 |V U OU (V U).
In particular, birational irreducible varieties have isomorphic function fields.

Exercise 9.8. Let X be an irreducible affine variety. Show:
(a) The function field K(X) is isomorphic to the so-called quotient field of the coordinate ring
A(X), i. e. to the localization of the integral domain A(X) at the multiplicatively closed subset
A(X)\{0}.
(b) Every local ring OX,a for a X is naturally a subring of K(X).
Exercise 9.9. Let X Pn be a quadric, i. e. an irreducible variety which is the zero locus of an
irreducible homogeneous polynomial of degree 2. Show that X is birational, but in general not
isomorphic, to the projective space Pn1 .
The main goal of this chapter is now to describe and study a general procedure to modify an irreducible variety to a birational one. In its original form, this construction depends on given polynomial functions f1 , . . . , fr on an affine variety X but we will see in Construction 9.17 that it can
also be performed with a given ideal in A(X) or subvariety of X instead, and that it can be glued in
order to work on arbitrary varieties.
Construction 9.10 (Blowing up). Let X An be an affine variety. For some r N>0 let f1 , . . . , fr
A(X) be polynomial functions on X, and set U = X\V ( f1 , . . . , fr ). As f1 , . . . , fr then do not vanish
simultaneously at any point of U, we obtain a well-defined morphism
f : U Pr1 , x 7 ( f1 (x) : . . . : fr (x)).
We consider its graph
f = {(x, f (x)) : x U}
U Pr1
which is isomorphic to U (with inverse morphism the projection to the first factor). Note that f is
closed in U Pr1 by Proposition 5.21 (a), but in general not closed in X Pr1 . The closure of f
in X Pr1 then contains f as a dense open subset. It is called the blow-up of X at f1 , . . . , fr ; we
Note that there is a natural projection morphism : X X to the first
will usually denote it by X.
factor. Sometimes we will also say that this morphism is the blow-up of X at f1 , . . . , fr .
Before we give examples of blow-ups let us introduce some more notation and easy general results
that will help us to deal with them.
14
74
Andreas Gathmann
Remark 9.11 (Exceptional sets). In construction 9.10, the graph f is isomorphic to U, with | f :
f U being an isomorphism. By abuse of notation, one often uses this isomorphism to identify
Its complement X\U
f with U, so that U becomes an open subset of X.

= 1 (V ( f1 , . . . , fr )), on
which is usually not an isomorphism, is called the exceptional set of the blow-up.
If X is irreducible and f1 , . . . , fr do not vanish simultaneously on all of X, then U = X\V ( f1 , . . . , fr )
is a non-empty and hence dense open subset of X. So its closure in the blow-up, which is all of X by
definition, is also irreducible. We therefore conclude that X and X are birational in this case, with
common dense open subset U.
Remark 9.12 (Strict transforms and blow-ups of subvarieties). In the notation of Construction 9.10,
let Y be a closed subvariety of X. Then we can blow up Y at f1 , . . . , fr as well. By construction, the
in fact it is the closure
resulting space Y Y Pr1 X Pr1 is then also a closed subvariety of X,
of Y U in X (using the isomorphism f
= U of Remark 9.11 to identify Y U with a subset of
If we consider Y as a subset of X in this way it is often called the strict transform of Y in the
X).
blow-up of X.
In particular, if X = X1 Xm is the irreducible decomposition of X then Xi X for i = 1, . . . , m.
Moreover, since taking closures commutes with finite unions it is immediate from Construction 9.10
that
X = X1 Xm ,
i. e. that for blowing up X we just blow up its irreducible components individually. For many purposes it therefore suffices to consider blow-ups of irreducible varieties.
Example 9.13 (Trivial cases of blow-ups). Let r = 1 in the notation of Construction 9.10, i. e.
consider the case when we blow up X at only one function f1 . Then X X P0
= X, and f
= U.
So X is just the closure of U in X under this isomorphism. If we assume for simplicity that X is
irreducible we therefore obtain the following two cases:
(a) If f1 6= 0 then U = X\V ( f1 ) is a non-empty open subset of X, and hence X = X by Remark
2.18 (b).
(b) If f1 = 0 then U = 0,
/ and hence also X = 0.
/
So in order to obtain interesting examples of blow-ups we will have to consider cases with r 2.
In order to understand blow-ups better, one of our main tasks has to be to find an explicit description
of them that does not refer to taking closures. The following inclusion is a first step in this direction.
Lemma 9.14. The blow-up X of an affine variety X at f1 , . . . , fr A(X) satisfies
X {(x, y) X Pr1 : yi f j (x) = y j fi (x) for all i, j = 1, . . . , r}.
Proof. Let U = X\V ( f1 , . . . , fr ). Then any point (x, y) U Pr1 on the graph f of the function
f : U Pr1 , x 7 ( f1 (x) : : fr (x)) satisfies (y1 : : yr ) = ( f1 (x) : : fr (x)), and hence yi f j (x) =
y j fi (x) for all i, j = 1, . . . , r. As these equations then also have to hold on the closure X of f , the
lemma follows.

Example 9.15 (Blow-up of An at the coordinate functions). Our first non-trivial (and in fact the most
important) case of a blow-up is that of the affine space An at the coordinate functions x1 , . . . , xn . This
fn is then isomorphic to An on the open subset U = An \V (x1 , . . . , xn ) = An \{0}, and by
blow-up A
Lemma 9.14 we have
fn {(x, y) An Pn1 : yi x j = y j xi for all i, j = 1, . . . , n} =: Y.
A
(1)
We claim that this inclusion is in fact an equality. To see this, let us consider the open subset
U1 = {(x, y) Y : y1 6= 0} with affine coordinates x1 , . . . , xn , y2 , . . . , yn in which we set y1 = 1. Note
that for given x1 , y2 , . . . , yn the equations (1) for Y then say exactly that x j = x1 y j for j = 2, . . . , n.
Hence there is an isomorphism
An U1 An Pn1 , (x1 , y2 , . . . , yn ) 7 ((x1 , x1 y2 , . . . , x1 yn ), (1 : y2 : . . . : yn )).
(2)
9.
75
Of course, the same holds for the open subsets Ui of Y where yi 6= 0 for i = 2, . . . , n. Hence Y can
be covered by n-dimensional affine spaces. By Exercises 2.20 (b) and 2.33 (a) this means that Y is
fn which is also irreducible of
irreducible of dimension n. But as Y contains the closed subvariety A
fn .
dimension n by Remarks 9.5 and 9.11, we conclude that we must already have Y = A
fn (with equality, as we have just seen) and the affine coordinates
In fact, both the description (1) of A
of (2) are very useful in practice for explicit computations on this blow-up.
fn An of Construction 9.10.
Let us now also study the blow-up (i. e. projection) morphism : A
n
We know already that this map is an isomorphism on U = A \{0}. In contrast, the exceptional set
1 (0) is given by setting x1 , . . . , xn to 0 in the description (1) above. As all defining equations
xi y j = x j yi become trivial in this case, we simply get
1 (0) = {(0, y) An Pn1 }
= Pn1 .
fn leaves all points except 0 unchanged, whereas the origin is
In other words, passing from An to A
n1
replaced by a projective space P . This is the geometric reason why this construction is called
blowing up in fact, we will slightly extend our terminology in Construction 9.17 (a) so that we
can then call the example above the blow-up of An at the origin, instead of at the functions x1 , . . . , xn .
fn as An with
Because of this behavior of the inverse images of one might be tempted to think of A
n1
a projective space P
attached at the origin, as in the picture below on the left. This is not correct
fn is.
however, as one can see already from the fact that this space would not be irreducible, whereas A
n
n
To get the true geometric picture for A let us consider the strict transform of a line L A through
fn . We will give a general recipe to
the origin, i. e. the blow-up L of L at x1 , . . . , xn contained in A
compute such strict transforms in Exercise 9.22, but in the case at hand this can also be done without
much theory: by construction, over the complement of the origin every point (x, y) L L Pn1
must have y being equal to the projective point corresponding to L K n . Hence the same holds on
and thus the strict transform L meets the exceptional set 1 (0)
the closure L,
= Pn1 above exactly
in the point corresponding to L. In other words, the exceptional set parametrizes the directions in
An at 0; two lines through the origin with distinct directions will become separated after the blowup. The following picture on the right illustrates this in the case of the plane: we can imagine the
f2 as a helix winding around the central line 1 (0)
blow-up A
= P1 (in fact, it winds around this
exceptional set once, so that one should think of the top of the helix as being glued to the bottom).
exceptional set
1 (0)
f2
A
P1
L
A2
A2
A2
0
Wrong picture
Correct picture
As already mentioned, the geometric interpretation of Example 9.15 suggests that we can think of
this construction as the blow-up of An at the origin instead of at the functions x1 , . . . , xn . To justify
this notation let us now show that the blow-up construction does not actually depend on the chosen
functions, but only on the ideal generated by them.
76
Andreas Gathmann
Lemma 9.16. The blow-up of an affine variety X at f1 , . . . , fr A(X) depends

only on the ideal ( f1 , . . . , fr ) E A(X).
precisely, if f10 , . . . , fs0 A(X) with ( f1 , . . . , fr )
X and 0 : X 0 X are the corresponding
( f10 , . . . , fs0 ) E A(X),
More
=
and
: X
blow-ups, there is an isomorphism F : X X 0 with 0 F = . In other words, we get a commutative
diagram as in the picture on the right.
X 0
0
Proof. By assumption we have relations

s
fi =
gi, j f j0 for all i = 1, . . . , r
and
f j0 =
j=1
h j,k fk for all j = 1, . . . , s

k=1
in A(X) for suitable gi, j , h j,k A(X). We claim that then

r
0
0
F : X X , (x, y) 7 (x, y ) := x, h1,k (x) yk : :

k=1
hs,k (x) yk

k=1
is an isomorphism between X X Pr1 and X 0 X Ps1 as required. This is easy to check:

The homogeneous coordinates of y0 are not simultaneously 0: note that by construction
we have the relation (y1 : : yr ) = ( f1 : : fr ) on U = X\V ( f1 , . . . , fr ) X X Pr1 ,
i. e. these two vectors are linearly dependent (and non-zero) at each point in this set.
Hence the linear relations fi = j,k gi, j h j,k fk in f1 , . . . , fr imply the corresponding relations
So if we had
yi = j,k gi, j h j,k yk in y1 , . . . , yr on this set, and thus also on its closure X.
y0j = k h j,k yk = 0 for all j then we would also have yi = j gi, j y0j = 0 for all i, which is a
contradiction.
The image of F lies in X 0 : by construction we have
r

r

F(x, y) = x, h1,k (x) fk (x) : : hs,k (x) fk (x)
= x, ( f10 (x) : : fs0 (x))
X 0
k=1
k=1
on the open subset U, and hence also on its closure X.

F is an isomorphism: by symmetry the same construction as above can also be done in the
other direction and gives us an inverse morphism F 1 .
It is obvious that 0 F = .
15
Construction 9.17 (Generalizations of the blow-up construction).
(a) Let X be an affine variety. For an ideal I E A(X) we define the blow-up of X at I to be the
blow-up of X at any set of generators of I which is well-defined up to isomorphisms by
Lemma 9.16. If Y X is a closed subvariety the blow-up of X at I(Y ) E A(X) will also be
called the blow-up of X at Y . So in this language we can say that Example 9.15 describes
the blow-up of An at the origin.
(b) Now let X be an arbitrary variety, and let Y X be a closed subvariety. For an affine open
cover {Ui : i I} of X, let U i be the blow-up of Ui at the closed subvariety Ui Y . It is then
We will call it
easy to check that these blow-ups U i can be glued together to a variety X.
again the blow-up of X at Y .
In the following, we will probably only need this in the case of the blow-up of a point, where
the construction is even easier as it is local around the blown-up point: let X be a variety,
and let a X be a point. Choose an affine open neighborhood U X of a, and let U be the
blow-up of U at a. Then we obtain X by gluing X\{a} to U along the common open subset
U\{a}.
9.
77
(c) With our current techniques the gluing procedure of (b) only works for blow-ups at subvarieties for the general construction of blowing up ideals we would need a way to patch
ideals. This is in fact possible and leads to the notion of a sheaf of ideals, but we will not do
this in this class.
Note however that blow-ups of a projective variety X can be defined in essentially the same
way as for affine varieties: if f1 , . . . , fr S(X) are homogeneous of the same degree the
blow-up of X at f1 , . . . , fr is defined as the closure of the graph
= {(x, ( f1 (x) : : fr (x)) : x U}
U Pr1
(for U = X\V ( f1 , . . . , fr )) in X Pr1 ; by the Segre embedding as in Remark 7.14 it is again

a projective variety.
1
f3 be the blow-up of A3 at the line V (x , x )
Exercise 9.18. Let A
1 2 = A . Show that its exceptional
1
1
set is isomorphic to A P . When do the strict transforms of two lines in A3 through V (x1 , x2 )
intersect in the blow-up? What is therefore the geometric meaning of the points in the exceptional
set (corresponding to Example 9.15 in which the points of the exceptional set correspond to the
directions through the blown-up point)?
Exercise 9.19. Let X An be an affine variety, and let Y1 ,Y2 ( X be irreducible, closed subsets,
no-one contained in the other. Moreover, let X be the blow-up of X at the ideal I(Y1 ) + I(Y2 ).
Show that the strict transforms of Y1 and Y2 in X are disjoint.
One of the main applications of blow-ups is the local study of varieties. We have seen already in
Example 9.15 that the exceptional set of the blow-up of An at the origin parametrizes the directions
of lines at this point. It should therefore not come as a surprise that the exceptional set of the blow-up
of a general variety X at a point a X parametrizes the tangent directions of X at a.
Construction 9.20 (Tangent cones). Let a be a point on a variety X. Consider the blow-up :
X X of X at a; its exceptional set 1 (a) is a projective variety (e. g. by choosing an affine
open neighborhood U An of a = (a1 , . . . , an ) in X and blowing up U at x1 a1 , . . . , xn an ; the
exceptional set is then contained in the projective space {a} Pn1 U Pn1 ).
The cone over this exceptional set 1 (a) (as in Definition 6.15 (c)) is called the tangent cone Ca X
of X at a. Note that it is well-defined up to isomorphisms by Lemma 9.16. In the special case (of an
affine patch) when X An and a X is the origin, we will also consider Ca X C(Pn1 ) = An as a
closed subvariety of the same ambient affine space as for X by blowing up at x1 , . . . , xn .
Example 9.21. Consider the three complex affine curves X1 , X2 , X3 A2C with real parts as in the
picture below.
x2
x2
x2
X3
x1
x1
x1
X2
X1
X1 = V (x2 + x12 )
X2 = V (x22 x12 x13 )
X3 = V (x22 x13 )
Note that by Remark 9.12 the blow-ups Xi of these curves at the origin (for i = 1, 2, 3) are contained
f2 of the affine plane at the origin as in Example 9.15. They can
as strict transforms in the blow-up A
thus be obtained geometrically as in the following picture by lifting the curves Xi \{0} by the map
f2 A2 and taking the closure in A
f2 . The additional points in these closures (drawn as dots
:A
in the picture below) are the exceptional sets of the blow-ups. By definition, the tangent cones C0 Xi
78
Andreas Gathmann
then consist of the lines corresponding to these points, as shown in gray below. They can be thought
of as the cones, i. e. unions of lines, that approximate Xi best around the origin.
f2
A
f2
A
f2
A
X1
X3
X2
C0 X1
C0 X2
A2
C0 X3
A2
X1
X2
A2
X3
Let us now study how these tangent cones can be computed rigorously. For example, for a point
f2 A2 P1 we have x2 x2 x3 = 0 (as the equation of the curve)
((x1 , x2 ), (y1 : y2 )) X2 A
2
1
1
and y1 x2 y2 x1 = 0 by Lemma 9.14. The latter means that the vectors (x1 , x2 ) and (y1 , y2 ) are
linearly dependent, i. e. that y1 = x1 and y2 = x2 away from the origin for some non-zero K.
Multiplying the equation of the curve with 2 thus yields
2 (x22 x12 x13 ) = 0
y22 y21 y21 x1 = 0
on X2 \ 1 (0), and thus also on its closure X2 . On 1 (0), i. e. if x1 = x2 = 0, this implies

y22 y21 = 0
(y2 y1 )(y2 + y1 ) = 0,
so that the exceptional set consists of the two points with (y1 : y2 ) P1 equal to (1 : 1) or (1 : 1).
Consequently, the tangent cone C0 X2 is the cone in A2 with the same equation
(x2 x1 )(x2 + x1 ) = 0,
i. e. the union of the two diagonals in A2 as in the picture above.
Note that the effect of this computation was exactly to pick out the terms of minimal degree of the
defining equation x22 x12 x13 = 0 in this case of degree 2 to obtain the equation x22 x12 = 0
of the tangent cone at the origin. This obviously yields a homogeneous polynomial (so that its affine
zero locus is a cone), and it fits well with the intuitive idea that for small values of x1 and x2 the
higher powers of the coordinates are much smaller, so that we get a good approximation for the
curve around the origin when we neglect them.
In fact, the following exercise (which is similar in style to proposition 6.33) shows that taking the
terms of smallest degree of the defining equations is the general way to compute tangent cones
explicitly after the coordinates have been shifted so that the point under consideration is the origin.
Exercise 9.22 (Computation of tangent cones). Let I E K[x1 , . . . , xn ] be an ideal, and assume that the
fn
corresponding affine variety X = V (I) An contains the origin. Consider the blow-up X A
An Pn1 at x1 , . . . , xn , and denote the homogeneous coordinates of Pn1 by y1 , . . . , yn .
fn can be covered by affine spaces, with one coordinate
(a) By Example 9.15 we know that A
patch being
fn An Pn1 ,
An A
(x1 , y2 , . . . , yn ) 7 ((x1 , x1 y2 , . . . , x1 yn ), (1 : y2 : : yn )).
9.
79
Prove that on this coordinate patch the blow-up X is given as the zero locus of the polynomials
f (x1 , x1 y2 , . . . , x1 yn )
min deg f
x1
for all non-zero f I, where min deg f denotes the smallest degree of a monomial in f .
(b) Prove that the exceptional hypersurface of X is
Vp ( f in : f I)
{0} Pn1 ,
where f in is the initial term of f , i. e. the sum of all monomials in f of smallest degree.
Consequently, the tangent cone of X at the origin is
C0 X = Va ( f in : f I)
An .
(c) If I = ( f ) is a principal ideal prove that C0 X = Va ( f in ). However, for a general ideal I show
that C0 X is in general not the zero locus of the initial terms of a set of generators for I.
In Example 9.15 above, blowing up the n-dimensional variety An at (x1 , . . . , xn ) has replaced the
fn . We will now see that this is in fact a general
origin by a variety Pn1 of codimension 1 in A
phenomenon.
Proposition 9.23 (Dimension of the exceptional set). Let : X X be the blow-up of an irreducible affine variety X at f1 , . . . , fr A(X). Then every irreducible component of the exceptional
It is therefore often called the exceptional hypersurset 1 (V ( f1 , . . . , fr )) has codimension 1 in X.
face of the blow-up.
Proof. It is enough to prove the statement on the affine open subsets Ui X X Pr1 for i =
But note
1, . . . , r where the i-th projective coordinate yi is non-zero, since these open subsets cover X.
that for a Ui the condition fi (a) = 0 implies f j (a) = 0 for all j by Lemma 9.14. So the exceptional
set is given by one equation fi = 0 on Ui . Moreover, if Ui is non-empty then this polynomial fi
would be contained in the
is not identically zero on Ui : otherwise Ui , and thus also its closure X,
exceptional set which is a contradiction since this implies U = 0/ and thus X = 0.
/ The statement
of the lemma thus follows from Proposition 2.25 (c).

Corollary 9.24 (Dimension of tangent cones). Let a be a point on a variety X. Then the dimension
dimCa X of the tangent cone of X at a is the local dimension codimX {a} of X at a.
Proof. Note that both dimCa X and codimX {a} are local around the point a. By passing to an open
neighborhood of a we can therefore assume that every irreducible component of X meets a, and that
X An is affine. We may also assume that X is not just the one-point set {a}, since otherwise the
statement of the corollary is trivial.
Now let X = X1 Xm be the irreducible decomposition of X. Note that X 6= {a} implies that all
of these components have dimension at least 1. By Proposition 9.23 every irreducible component of
the exceptional set of the blow-up Xi of Xi at a has dimension dim Xi 1, and so by Exercise 6.32
(a) every irreducible component of the tangent cone Ca Xi has dimension dim Xi . As the maximum of
these dimensions is just the local dimension codimX {a} (see Exercise 5.11 (b)) it therefore suffices
to show that all these exceptional sets (and hence also the tangent cones) are non-empty.
Assume the contrary, i. e. that the exceptional set of Xi is empty for some i. Extending this to the
n
projective closure Pn of An we obtain an irreducible variety
Xi P containing a whose blow-up
e
e
fn has an empty exceptional set. This means that Xi = Xi \{a}, where : P
fn Pn is the
Xi in P
blow-up map. As Xe is a projective (and hence complete) variety by Construction 9.17 (c) this is a
i
contradiction to Corollary 7.24 since Xi \{a} is not closed (recall that Xi has dimension at least 1, so
that Xi \{a} 6= 0).
/
80
Andreas Gathmann
Exercise 9.25. Let X = V (x22 x12 x13 ) A2 . Show that X is not isomorphic to A1 , but that the
blow-up of X at the origin is.
Can you interpret this result geometrically?
Exercise 9.26.
(a) Show that the blow-up of A2 at the ideal (x12 , x1 x2 , x22 ) is isomorphic to the blow-up of A2 at
the ideal (x1 , x2 ).
(b) Let X be an affine variety, and let I E A(X) be
an ideal. Is it true in general that the blow-up
of X at I is isomorphic to the blow-up of X at I?
We will now discuss another important application of blow-ups that follows more or less directly
from the definitions: they can be used to extend morphisms defined only on an open subset of a
variety.
Remark 9.27 (Blowing up to extend morphisms). Let X An be an affine variety, and let f1 , . . . , fr
be polynomial functions on X. Note that the morphism f : x 7 ( f1 (x) : : fr (x)) to Pr1 is only
well-defined on the open subset U = X\V ( f1 , . . . , fr ) of X. In general, we can not expect that this
morphism can be extended to a morphism on all of X. But we can always extend it after blowing up
the ideal ( f1 , . . . , fr ) of the indeterminacy locus: there is an extension f : X Pr1 of f (that agrees
with f on U), namely just the projection from X X Pr1 to the second factor Pr1 . So blowing
up is a way to extend morphisms to bigger sets on which they would otherwise be ill-defined. Let us
consider a concrete example of this idea in the next lemma and the following remark.
Lemma 9.28. P1 P1 blown up in one point is isomorphic to P2 blown up in two points.
Proof. We know from Example 7.12 that P1 P1 is isomorphic to the quadric surface
X = {(x0 : x1 : x2 : x3 ) : x0 x3 = x1 x2 }
P3 .
Let X be blow-up of X at a = (0 : 0 : 0 : 1) X, which can be realized as in Construction 9.17 (c) as

the blow-up X P3 P2 of X at x0 , x1 , x2 .
f2 P2 P3 be the blow-up of P2
On the other hand, let b = (0 : 1 : 0), c = (0 : 0 : 1) P2 , and let P
at y20 , y0 y1 , y0 y2 , y1 y2 . Note that these polynomials do not generate the ideal I({b, c}) = (y0 , y1 y2 ),
but this does not matter: the blow-up is a local construction, so let us check that we are locally just
blowing up b, and similarly c. There is an open affine neighborhood around b given by y1 6= 0,
where we can set y1 = 1, and on this neighborhood the given functions y20 , y0 , y0 y2 , y2 generate the
f2 is actually the blow-up of P2 at b and c.
ideal (y , y ) of b. So P
0
Now we claim that an isomorphism is given by

f2 , ((x : x : x : x ), (y : y : y )) 7 ((y : y : y ), (x : x : x : x )).
f : X 7 P
0 1 2 3
0 1 2
0 1 2
0 1 2 3
In fact, this is easy to prove: obviously, f is an isomorphism from P3 P2 to P2 P3 , so we only have
f2 , and that f 1 maps P
f2 to X.
Note that it suffices to check this on a dense
to show that f maps X to P
open subset. But this is easy: on the complement of the exceptional set in X we have x0 x3 = x1 x2
and (y0 : y1 : y2 ) = (x0 : x1 : x2 ), so on the (smaller) complement of V (x0 ) we get the correct equations
(x0 : x1 : x2 : x3 ) = (x02 : x0 x1 : x0 x2 : x0 x3 ) = (x02 : x0 x1 : x0 x2 : x1 x2 ) = (y20 : y0 y1 : y0 y2 : y1 y2 )
16
f2 . Conversely, on the complement of the exceptional set in P

f2
for the image point under f to lie in P
we have (x0 : x1 : x2 : x3 ) = (y20 : y0 y1 : y0 y2 : y1 y2 ), so we conclude that x0 x3 = x1 x2 and (y0 : y1 : y2 ) =
(x0 : x1 : x2 ) where y0 6= 0.

Remark 9.29. The proof of Lemma 9.28 is short and elegant, but not very insightful. So let us try
to understand geometrically what is going on. As in the proof above, we think of P1 P1 as the
quadric surface
X = {(x0 : x1 : x2 : x3 ) : x0 x3 = x1 x2 } P3 .
9.
81
Let us project X from a = (0 : 0 : 0 : 1) X to Vp (x3 )

= P2 . The corresponding morphism f is shown
in the picture below; as in Example 7.6 (b) it is given by f (x0 : x1 : x2 : x3 ) = (x0 : x1 : x2 ) and welldefined away from a.
X
= P1 P1
a
x
f (x)
P2
Recall that, in the corresponding case of the projection of a quadric curve in Example 7.6 (c), the
morphism f could be extended to the point a. This is now no longer the case for our quadric surface
X: to construct f (a) we would have to take the limit of the points f (x) for x approaching a, i. e.
consider lines through a and x for x a. These lines will then become tangent lines to X at a
but X, being two-dimensional, has a one-parameter family of such tangent lines. This is why f (a) is
ill-defined. But we also see from this discussion that blowing up a on X, i. e. replacing it by the set of
all tangent lines through a, will exactly resolve this indeterminacy. Hence f becomes a well-defined
morphism from X to Vp (x3 )
= P2 .
Let us now check if there is an inverse morphism. By construction, it is easy to see what it would
have to look like: the points of X\{a} mapped to a point y Vp (x3 ) are exactly those on the line ay
through a and y. In general, this line intersects X in two points, one of which is a. So there is then
exactly one point on X which maps to y, leading to an inverse morphism f 1 . This reasoning is only
false if the whole line ay lies in X. Then this whole line would be mapped to y, so that we cannot
have an inverse f 1 there. But of course we expect again that this problem can be taken care of by
blowing up y in P2 , so that it is replaced by a P1 that can then be mapped bijectively to ay.
There are obviously two such lines ab and ac, given by b = (0 : 1 : 0) and c = (0 : 0 : 1). If you think
of X as P1 P1 again, these lines are precisely the horizontal and vertical lines passing through
a where the coordinate in one of the two factors is constant. So we would expect that f can be made
into an isomorphism after blowing up b and c, which is exactly what we have shown in Lemma 9.28.
Exercise 9.30 (Cremona transformation). Let a = (1 : 0 : 0), b = (0 : 1 : 0), and c = (0 : 0 : 1) be the
three coordinate points of P2 , and let U = P2 \{a, b, c}. Consider the morphism
f : U P2 , (x0 : x1 : x2 ) 7 (x1 x2 : x0 x2 : x0 x1 ).
(a) Show that there is no morphism P2 P2 extending f .
f2 be the blow-up of P2 at {a, b, c}. Show that f can be extended to an isomorphism
(b) Let P
f2 P
f2 . This isomorphism is called the Cremona transformation.
f : P
82
Andreas Gathmann
10.
Smooth Varieties
Let a be a point on a variety X. In the last chapter we have introduced the tangent cone Ca X as a
way to study X locally around a (see Construction 9.20). It is a cone whose dimension is the local
dimension codimX {a} (Corollary 9.24), and we can think of it as the cone that best approximates X
around a. In an affine open chart where a is the origin, we can compute Ca X by choosing an ideal
with zero locus X and replacing each polynomial in this ideal by its initial term (Exercise 9.22 (b)).
However, in practice one often wants to approximate a given variety by a linear space rather than
by a cone. We will therefore study now to what extent this is possible, and how the result compares
to the tangent cones that we already know. Of course, the idea to construct this is just to take the
linear terms instead of the initial terms of the defining polynomials when considering the origin
in an affine variety. For simplicity, let us therefore assume for a moment that we have chosen an
affine neighborhood of the point a such that a = 0 we will see in Lemma 10.5 that the following
construction actually does not depend on this choice.
Definition 10.1 (Tangent spaces). Let a be a point on a variety X. By choosing an affine neighborhood of a we assume that X An and that a = 0 is the origin. Then
Ta X := V ( f1 : f I(X))
An
is called the tangent space of X at a, where f1 K[x1 , . . . , xn ] denotes the linear term of a polynomial
f K[x1 , . . . , xn ] as in Definition 6.6 (a).
As in the case of tangent cones, we can consider Ta X either as an abstract variety (leaving its dimension as the only invariant since it is a linear space) or as a subspace of An .
Remark 10.2.
(a) In contrast to the case of tangent cones in Exercise 9.22 (c), it always suffices in Definition
10.1 to take the zero locus only of the linear parts of a set S of generators for I(X): if f , g S
are polynomials such that f1 and g1 vanish at a point x An then
( f + g)1 (x) = f1 (x) + g1 (x) = 0
and
(h f )1 (x) = h(0) f1 (x) + f (0) h1 (x) = h(0) 0 + 0 h1 (x) = 0
for an arbitrary polynomial h K[x1 , . . . , xn ], and hence x Ta X.

(b) However, again in contrast to the case of tangent cones in Exercise 9.22 it is crucial in
Definition 10.1 that we take the radical ideal of X and not just any ideal with zero locus X:
the ideals (x) and (x2 ) in K[x] have the same zero locus {0} in A1 , but the zero locus of the
linear term of x is the origin again, whereas the zero locus of the linear term of x2 is all of
A1 .
(c) For polynomials vanishing at the origin, a non-vanishing linear term is clearly always initial. Hence by Exercise 9.22 (b) it follows that Ca X Ta X, i. e. that the tangent space
always contains the tangent cone. In particular, this means by Corollary 9.24 that dim Ta X
codimX {a}.
Example 10.3. Consider again the three curves X1 , X2 , X3 of Example 9.21. By taking the initial
resp. linear term of the defining polynomials we can compute the tangent cones and spaces of these
curves at the origin:
X1 = V (x2 + x12 ): C0 X1 = T0 X1 = V (x2 );
X2 = V (x22 x12 x13 ): C0 X2 = V (x22 x12 ), T0 X2 = V (0) = A2 ;
X3 = V (x22 x13 ): C0 X3 = V (x22 ) = V (x2 ), T0 X3 = V (0) = A2 .
10.
Smooth Varieties
83
The following picture shows these curves together with their tangent cones and spaces. Note that for
the curve X1 the tangent cone is already a linear space, and the notions of tangent space and tangent
cone agree. In contrast, the tangent cone of X2 at the origin is not linear. By Remark 10.2 (c), the
tangent space T0 X2 must be a linear space containing C0 X2 , and hence it is necessarily all of A2 .
However, the curve X3 shows that the tangent space is not always the linear space spanned by the
tangent cone.
T0 X2
x2
T0 X3
x2
T0 X1 = C0 X1
x2
X3
C0 X2
C0 X3
x1
x1
x1
X2
X1
X1 = V (x2 + x12 )
X2 = V (x22 x12 x13 )
X3 = V (x22 x13 )
Before we study the relation between tangent spaces and cones in more detail, let us show first of all
that the (dimension of the) tangent space is actually an intrinsic local invariant of a variety around a
point, i. e. that it does not depend on a choice of affine open subset or coordinates around the point.
We will do this by establishing an alternative description of the tangent space that does not need any
such choices. The key observation needed for this is the isomorphism of the following lemma.
Lemma 10.4. Let X An be an affine variety containing the origin. Moreover, let M :=
(x1 , . . . , xn ) = I(0) E A(X) be the ideal of the origin in X. Then there is a natural vector space
isomorphism
M/M 2
= HomK (T0 X, K).
In other words, the tangent space T0 X is naturally the vector space dual to M/M 2 .
Proof. Consider the K-linear map
: M HomK (T0 X, K), f 7 f1 |T0 X
sending the class of a polynomial modulo I(X) to its linear term, regarded as a map restricted to
the tangent space. By definition of the tangent space, this map is well-defined. Moreover, note
that is surjective since any linear map on T0 X can be extended to a linear map on An . So by the
homomorphism theorem it suffices to prove that ker = M 2 :
Consider the vector subspace W = {g1 : g I(X)} of K[x1 , . . . , xn ], and let k be its dimension.
Then its zero locus T0 X has dimension n k, and hence the space of linear forms vanishing
on T0 X has dimension k again. As it clearly contains W , we conclude that W must be equal
to the space of linear forms vanishing on T0 X.
So if f ker , i. e. the linear term of f vanishes on T0 X, we know that there is a polynomial
g I(X) with g1 = f1 . But then f g has no constant or linear term, and hence f = f g
M2.
If f , g M then ( f g)1 = f (0) g1 + g(0) f1 = 0 g1 + 0 f1 = 0, and hence ( f g) = 0.
In order to make Lemma 10.4 into an intrinsic description of the tangent space we need to transfer
it from the affine coordinate ring A(X) (which for a general variety would require the choice of an
affine coordinate chart that sends the given point a to the origin) to the local ring OX,a (which is
independent of any choices). To do this, recall by Lemma 3.21 that with the notations from above
we have OX,a
= S1 A(X), where S = A(X)\M = { f A(X) : f (a) 6= 0} is the multiplicatively closed
subset of polynomial functions that are non-zero at the point a. In this ring
ng
o
: g, f A(X) with g(a) = 0 and f (a) 6= 0
S1 M =
f
84
Andreas Gathmann
is just the maximal ideal Ia of all local functions vanishing at a as in Definition 3.22. Using these
constructions we obtain the following result.
Lemma 10.5. With notations as above we have
M/M 2
= (S1 M)/(S1 M)2 .
In particular, if a is a point on a variety X and Ia = { OX,a : (a) = 0} is the maximal ideal of
local functions in OX,a vanishing at a, then Ta X is naturally isomorphic to the vector space dual to
Ia /Ia2 , and thus independent of any choices.
Proof. This time we consider the vector space homomorphism
: M (S1 M)/(S1 M)2 , g 7
g
1
(S1 M)2 .
where the bar denotes classes modulo

In order to deduce the lemma from the homomorphism theorem we have to show the following three statements:
is surjective: Let
under since
g
f
S1 M. Then
g
f (0)
M is an inverse image of the class of this fraction
g
g f (0) f
g
=
f
f (0)
f
f (0)
(S1 M)2
(note that f (0) f lies in M as it does not contain a constant term).

ker M 2 : Let g ker , i. e.
g
1
(S1 M)2 . This means that

g
hi ki
=
1
fi
i
()
for a finite sum with elements hi , ki M and fi S. By bringing this to a common denominator we can assume that all fi are equal, say to f . The equation () in OX,a then means
f ( f g i hi ki ) = 0 in A(X) for some f S by Construction 3.12. This implies f f g M 2 .
But (( f f )(0) f f ) g M 2 as well, and hence ( f f )(0) g M 2 , which implies g M 2 since
( f f )(0) K .
M 2 ker is trivial.
Exercise 10.6. Let f : X Y be a morphism of varieties, and let a X. Show that f induces a linear
map Ta X T f (a)Y between tangent spaces.
We have now constructed two objects associated to the local structure of a variety X at a point a X:
the tangent cone Ca X, which is a cone of dimension codimX {a}, but in general not a linear
space; and
the tangent space Ta X, which is a linear space, but whose dimension might be bigger than
codimX {a}.
Of course, we should give special attention to the case when these two notions agree, i. e. when X
can be approximated around a by a linear space whose dimension is the local dimension of X at a.
Definition 10.7 (Smooth and singular varieties). Let X be a variety.
(a) A point a X is called smooth, regular, or non-singular if Ta X = Ca X. Otherwise it is
called a singular point of X.
(b) If X has a singular point we say that X is singular. Otherwise X is called smooth, regular, or
non-singular.
10.
Smooth Varieties
85
Example 10.8. Of the three curves of Example 10.3, exactly the first one is smooth at the origin.
As in our original motivation for the definition of tangent spaces, this is just the statement that X1
can be approximated around the origin by a straight line in contrast to X2 and X3 , which have a
multiple point resp. a corner there. A more precise geometric interpretation of smoothness can
be obtained by comparing our algebraic situation with the Implicit Function Theorem from analysis,
see Remark 10.14.
Lemma 10.9. Let X be a variety, and let a X be a point. The following statements are equivalent:
(a) The point a is smooth on X.
(b) dim Ta X = codimX {a}.
(c) dim Ta X codimX {a}.
Proof. The implication (a) (b) follows immediately from Corollary 9.24, and (b) (c) is trivial.
To prove (c) (a), note first that (c) together with Remark 10.2 (c) implies dim Ta X = codimX {a}.
But again by Remark 10.2 (c), the tangent space Ta X contains the tangent cone Ca X, which is of
the same dimension by Corollary 9.24. As Ta X is irreducible (since it is a linear space), this is only
possible if Ta X = Ca X, i. e. if a is a smooth point of X.

Remark 10.10 (Smoothness in commutative algebra). Let a be a point on a variety X.
(a) Let Ia E OX,a be the maximal ideal of local functions vanishing at a as in Definition 3.22.
Combining Lemma 10.5 with Lemma 10.9 we see that a is a smooth point of X if and only
if the vector space dimension of Ia /Ia2 is equal to the local dimension codimX {a} of X at a.
This is a property of the local ring OX,a alone, and one can therefore study it with methods
from commutative algebra. A ring with these properties is usually called a regular local ring
[G5, Definition 11.38], which is also the reason for the name regular point in Definition
10.7 (a).
(b) It is a result of commutative algebra that a regular local ring as in (a) is always an integral
domain [G5, Proposition 11.40]. Translating this into geometry as in Proposition 2.9, this
yields the intuitively obvious statement that a variety is locally irreducible at every smooth
point a, i. e. that X has only one irreducible component meeting a. Equivalently, any point
on a variety at which two irreducible components meet is necessarily a singular point.
The good thing about smoothness is that it is very easy to check using (formal) partial derivatives:
Proposition 10.11 (Affine Jacobi criterion). Let X An be an affine variety with ideal I(X) =
( f1 , . . . , fr ), and let a X be a point. Then X is smooth at a if and only if the rank of the r n
Jacobian matrix

fi
(a)
xj
i, j
is at least n codimX {a}.
Proof. Let x = (x1 , . . . , xn ) be the coordinates of An , and let y := x a be the shifted coordinates
in which the point a becomes the origin. By a formal Taylor expansion, the linear term of the
polynomial fi in these coordinates y is nj=1 xfij (a) y j . Hence the tangent space Ta X is by Definition
10.1 and Remark
10.2 (a) the zero locus of these linear terms, i. e. the kernel of the Jacobian matrix

fi
J = x j (a) i, j . So by Lemma 10.9 the point a is smooth if and only if dim ker J codimX {a},
which is equivalent to rk J n codimX {a}.

To check smoothness for a point on a projective variety, we can of course restrict to an affine open
subset of the point. However, the following exercise shows that there is also a projective version of
the Jacobi criterion that does not need these affine patches and works directly with the homogeneous
coordinates instead.
17
86
Andreas Gathmann
Exercise 10.12.
(a) Show that
n
xi xi = d f
i=0
for any homogeneous polynomial f K[x0 , . . . , xn ] of degree d.

(b) (Projective Jacobi criterion) Let X Pn be a projective variety with homogeneous ideal
I(X) = ( f1 , . . . , fr ), and let a X. Prove
that X is smooth at a if and only if the rank of the

r (n + 1) Jacobian matrix
fi
x j (a) i, j
In this criterion, note that the entries
is at least n codimX {a}.
fi
x j (a)
of the Jacobian matrix are not well-defined:
multiplying the coordinates of a by a scalar K will multiply xfij (a) by di 1 , where di

is the degree of fi . However, these are just row transformations of the Jacobian matrix, which
do not affect its rank. Hence the condition in the projective Jacobi criterion is well-defined.
Remark 10.13 (Variants of the Jacobi criterion). There are a few ways to extend the Jacobi criterion
even further. For simplicity, we will discuss this here in the case of an affine variety X as in Proposition 10.11, but it is easy to see that the corresponding statements hold in the projective setting of
Exercise 10.12 (b) as well.
(a) If X is irreducible then codimX {a} = dim X for all a X by Proposition 2.25 (b). So in
this case a is a smooth point of X if and only if the rank of the Jacobian matrix is at least
n dim X = codim X.
(b) Let f1 , . . . , fr K[x1 , . . . , xn ] be polynomials such that X = V ( f1 , . . . , fr ), but that do not
necessarily generate the ideal of X (as required in Proposition 10.11). Then theJacobi
criterion still holds in one direction: assume that the rank of the Jacobian matrix xfij (a)
i, j
is at least n codimX {a}. The proof of the affine Jacobi criterion then shows that the zero
locus of all linear terms of the elements of ( f1 , . . . , fr ) has dimension at most codimX {a}.
The
p same then necessarily holds for the zero locus of all linear terms of the elements of
( f1 , . . . , fr ) = I(X) (which might only be smaller). By Proposition 10.11 this means that
a is a smooth point of X.
The converse is in general false, as we have already seen in Remark 10.2 (b).
(c) Again let f1 , . . . , fr K[x1, . . . , xn ]be polynomials with X = V ( f1 , . . . , fr ). This time assume
that the Jacobian matrix
fi
x j (a) i, j
has maximal row rank, i. e. that its rank is equal to
r. As every irreducible component of X has dimension at least n r by Proposition 2.25

(c) we know moreover that codimX {a} n r. Hence the rank of the Jacobian matrix is
r n codimX {a}, so X is smooth at a by (b).
Remark 10.14 (Relation to the Implicit Function Theorem). The version of the Jacobi criterion of
Remark 10.13 (c) is closely related to the Implicit Function Theorem from analysis. Given real
polynomials f1 , . . . , fr R[x1 , . . . , xn ] (or more generally continuously differentiable functions on an
n
open subset
of R
) and a point a in their common zero locus X = V ( f1 , . . . , fr ) such that the Jacobian
matrix
fi
x j (a) i, j
has rank r, this theorem states roughly that X is locally around a the graph of a
continuously differentiable function [G2, Proposition 26.9] so that in particular it does not have
any corners. It can be shown that the same result holds over the complex numbers as well. So in
the case K = C the statement of Remark 10.13 (c) that X is smooth at a can also be interpreted in
this geometric way.
Note however that there is no algebraic analogue of the Implicit Function Theorem itself: for example, the polynomial equation f (x1 , x2 ) := x2 x12 = 0 cannot be solved for x1 by a regular function
locally around the point (1, 1), although xf (1, 1) = 2 6= 0 it can only be solved by a continu1
ously differentiable function x1 = x2 .
10.
Smooth Varieties
87
Example 10.15. Consider again the curve X3 = V (x22 x13 ) A2C of Examples 9.21 and 10.3. The
Jacobian matrix of the single polynomial f = x22 x13 is

f f
= (3x12 2x2 ),
x1 x2
so it has rank (at least) 2 dim X = 1 exactly if (x1 , x2 ) A2 \{0}. Hence the Jacobi criterion as
in Remark 10.13 (a) does not only reprove our observation from Example 10.3 that the origin is a
singular point of X3 , but also shows simultaneously that all other points of X3 are smooth.
In the picture on the right we have also drawn the blow-up X3 of X3
at its singular point again. We have seen already that its exceptional
f2
A
set consists of only one point a X3 . Let us now check that this is a
X3
smooth point of X3 as we would expect from the picture.
f2 A2 P1 , the point
In the coordinates ((x1 , x2 ), (y1 : y2 )) of X3 A
a is given as ((0, 0), (1 : 0)). So around a we can use the affine open
chart U1 = {((x1 , x2 ), (y1 : y2 )) : y1 6= 0} with affine coordinates x1 and
y2 as in Example 9.15. By Exercise 9.22 (a), the blow-up X3 is given
in these coordinates by
(x1 y2 )2 x13
= 0,
x12
As the Jacobian matrix

i. e. g(x1 , y2 ) := y22 x1 = 0.
A2
0
g g
x1 y2

= (1 2y2 )
X3
of this polynomial has rank 1 at every point, the Jacobi criterion tells us that X3 is smooth. In
fact, from the defining equation y22 x1 = 0 we see that on the open subset U1 the curve X3 is just
f2 (which is given on U by the equation
the standard parabola tangent to the exceptional set of A
1
x1 = 0 by the proof of Proposition 9.23).
It is actually a general statement that blowing up makes singular points nicer, and that successive
blow-ups will eventually make all singular points smooth. This process is called resolution of singularities. We will not discuss this here in detail, but the following exercise shows an example of this
process.
Exercise 10.16. For k N let Xk be the affine curve Xk := V (x22 x12k+1 ) A2 . Show that Xk is not
isomorphic to Xl if k 6= l.
(Hint: Consider the blow-up of Xk at the origin.)
Exercise 10.17. Let X P3 be the degree-3 Veronese embedding of P1 as in Exercise 7.32. Of
course, X must be smooth since it is isomorphic to P1 . Verify this directly using the projective
Jacobi criterion of Exercise 10.12 (b).
Corollary 10.18. The set of singular points of a variety is closed.
Proof. It suffices to prove the statement in the case of an affine variety X. We show that the subset
U X of smooth points is open. So let a U. By possibly restricting to a smaller affine subset,
we may assume by Remark 10.10 (b) that X is irreducible. Then by Remark 10.13 (a) we know that
U is exactly the set of points at which the rank of the Jacobian matrix of generators of I(X) is at
least codim X. As this is an open condition (given by the non-vanishing of at least one minor of size
codim X), the result follows.

As the set of smooth points of a variety is open in the Zariski topology by Corollary 10.18, it is very
big unless it is empty, of course. Let us quickly study whether this might happen.
88
Andreas Gathmann
Remark 10.19 (Generic smoothness). Let f K[x1 , . . . , xn ] be a non-constant irreducible polynomial, and let X = V ( f ) An . We claim that X has a smooth point, so that the set of smooth points
of X is a non-empty open subset by Corollary 10.18, and thus dense by Remark 2.18.
Assume the contrary, i. e. that all points of X are singular. Then by Remark 10.13 (a) the Jacobian
matrix of f must have rank 0 at every point, which means that xfi (a) = 0 for all a X and i = 1, . . . , n.
Hence
f
xi
I(V ( f )) = ( f ) by the Nullstellensatz. But since f is irreducible and the polynomial
has smaller degree than f this is only possible if
f
xi
f
xi
= 0 for all i.
In the case char K = 0 this is already a contradiction to f being non-constant. If char K = p is

positive, then f must be a polynomial in x1p , . . . , xnp , and so
!p
f=
ai1 ,...,in x1pi1 xnpin =
bi1 ,...,in x1i1 xnin
i1 ,...,in
i1 ,...,in
for p-th roots bi1 ,...,in of ai1 ,...,in . This is a contradiction since f was assumed to be irreducible.
In fact, one can show this generic smoothness statement for any variety X: the set of smooth points
of X is dense in X. A proof of this result can be found in [H, Theorem I.5.3].
Example 10.20 (Fermat hypersurfaces). For given n, d N>0 consider the Fermat hypersurface
X := Vp (x0d + + xnd )
Pn .
We want to show that X is smooth for all choices of n, d, and K. For this we use the Jacobian matrix
(dx0d1 dxnd1 ) of the given polynomial:
(a) If char K 6 | d the Jacobian matrix has rank 1 at every point, so X is smooth by Exercise 10.12
(b).
(b) If p = char K | d we can write d = k pr for some r N>0 and p 6 | k. Since
r
x0d + + xnd = (x0k + + xnk ) p ,

we see again that X = Vp (x0k + + xnk ) is smooth by (a).
Exercise 10.21. Let X be a projective variety of dimension n. Prove:
(a) There is an injective morphism X P2n+1 .
(b) There is in general no such morphism that is an isomorphism onto its image.
Exercise 10.22. Let n 2. Prove:
(a) Every smooth hypersurface in Pn is irreducible.
(b) A general hypersurface in PnC is smooth (and thus by (a) irreducible). More precisely, for a

given d N>0 the vector space C[x0 , . . . , xn ]d has dimension n+d
n , and so the space of all
homogeneous degree-d polynomials in x0 , . . . , xn modulo scalars can be identified with the
n+d
projective space P( n )1 . Show that the subset of this projective space of all (classes of)
polynomials f such that f is irreducible and Vp ( f ) is smooth is dense and open.
Exercise 10.23 (Dual curves). Assume that char K 6= 2, and let f K[x0 , x1 , x2 ] be a homogeneous
polynomial whose partial derivatives xfi for i = 0, 1, 2 do not vanish simultaneously at any point of
X = Vp ( f ) P2 . Then the image of the morphism

f
f
f
2
F : X P , a 7
(a) :
(a) :
(a)
x0
x1
x2
is called the dual curve to X.
(a) Find a geometric description of F. What does it mean geometrically if F(a) = F(b) for two
distinct points a, b X?
(b) If X is a conic, prove that its dual F(X) is also a conic.
10.
Smooth Varieties
89
(c) For any five lines in P2 in general position (what does this mean?) show that there is a unique
conic in P2 that is tangent to all of them.
90
Andreas Gathmann
11.
The 27 Lines on a Smooth Cubic Surface
As an application of the theory that we have developed so far, we now want to study lines on cubic
surfaces in P3 . In Example 0.9, we have already mentioned that every smooth cubic surface has
exactly 27 lines on it. Our goal is now to show this, to study the configuration of these lines, and to
prove that every smooth cubic surface is birational (but not isomorphic) to P2 . All these results are
classical, dating back to the 19th century. They can be regarded historically as being among the first
non-trivial statements in projective algebraic geometry.
The results of this chapter will not be needed later on. Most proofs will therefore not be given in
every detail here. The aim of this chapter is rather to give an idea of what can be done with our
current methods.
For simplicity, we will restrict ourselves to the case of the ground field K = C. By a smooth cubic
surface we will always mean a smooth hypersurface in P3 that can be written as the zero locus of an
irreducible homogeneous polynomial of degree 3. Let us start with the discussion of a special case
of such a cubic surface: the Fermat cubic Vp (x03 + x13 + x23 + x33 ) P3 as in Example 10.20.
Lemma 11.1. The Fermat cubic X = Vp (x03 + x13 + x23 + x33 ) P3 contains exactly 27 lines.
Proof. Up to a permutation of coordinates, every line in P3 is given by two linear equations of the
form x0 = a2 x2 + a3 x3 and x1 = b2 x2 + b3 x3 for suitable a2 , a3 , b2 , b3 C. Such a line lies in X if
and only if
(a2 x2 + a3 x3 )3 + (b2 x2 + b3 x3 )3 + x23 + x33 = 0
as a polynomial in C[x2 , x3 ], so by comparing coefficients if and only if
a32 + b32 = 1,
(1)
a33 + b33 = 1,
(2)
a22 a3
a2 a23
= b22 b3 ,
= b2 b23 .
(3)
(4)
If a2 , a3 , b2 , b3 are all non-zero, then (3)2 /(4) gives a32 = b32 , in contradiction to (1). Hence for a
line in the cubic at least one of these numbers must be zero. Again after possibly renumbering the
coordinates we may assume that a2 = 0. Then b32 = 1 by (1), b3 = 0 by (3), and a33 = 1 by (2).
Conversely, for such values of a2 , a3 , b2 , b3 the above equations all hold, so that we really obtain a
line in the cubic.
We thus obtain 9 lines in X by setting b2 = j and a3 = k for 0 j, k 2 and = exp( 2i
3 )
a primitive third root of unity. So by finally allowing permutations of the coordinates we find that
there are exactly the following 27 lines on X:
x0 + x3 k = x1 + x2 j = 0, 0 j, k 2,
x0 + x2 k = x3 + x1 j = 0, 0 j, k 2,
x0 + x1 k = x3 + x2 j = 0, 0 j, k 2.
18
Corollary 11.2. Let X P3 again be the Fermat cubic as in Lemma 11.1.
(a) Given any line L in X, there are exactly 10 other lines in X that intersect L.
(b) Given any two disjoint lines L1 , L2 in X, there are exactly 5 other lines in X meeting both L1
and L2 .
11.
91
Proof. As we know all the lines in X by the proof of Lemma 11.1, this is just simple checking. For
example, to prove (a) we may assume by permuting coordinates and multiplying them with suitable
third roots of unity that L is given by x0 + x3 = x1 + x2 = 0. The other lines meeting L are then
exactly the following:
4 lines of the form x0 + x3 k = x1 + x2 j = 0,
( j, k) = (1, 0), (2, 0), (0, 1), (0, 2),
0 j 2,
0 j 2.
3 lines of the form x0 + x2 = x3 + x1 = 0,

3 lines of the form x0 + x1 = x3 + x2 = 0,
The proof of part (b) is analogous.
Let us now transfer these results to an arbitrary smooth cubic surface. This is where it gets interesting, since the equations determining the lines lying in the cubic as in the proof of Lemma 11.1 will
in general be too complicated to solve them directly. Instead, we will only show that the number of
lines in a smooth cubic must be the same for all cubics, so that we can then conclude by Lemma 11.1
that this number must be 27. In other words, we have to consider all smooth cubic surfaces at once.
Construction 11.3 (The incidence correspondence of lines in smooth cubic surfaces). As in Exer3+3
cise 10.22 (b), let P19 = P( 3 )1 be the projective space of all homogeneous degree-3 polynomials
in x0 , x1 , x2 , x3 modulo scalars, so that the space of smooth cubic surfaces is a dense open subset U
of P19 . More precisely, a smooth cubic surface can be given as the zero locus of an irreducible polynomial fc := c x = 0 in multi-index notation, i. e. runs over all quadruples of non-negative
indices (0 , 1 , 2 , 3 ) with i i = 3. The corresponding point in U P19 is then the one with
homogeneous coordinates c = (c ) .
Moreover, we know already that the lines in P3 are parametrized by the Grassmannian G(2, 4) introduced in Chapter 8. We can therefore consider the incidence correspondence
M := {(X, L) : L is a line contained in the smooth cubic X}
U G(2, 4).
Note that it comes with a natural projection map : M U sending a pair (X, L) to X, and that the
number of lines in a cubic surface is just its number of inverse images under .
To show that this number of inverse images is constant on U, we will pass from the algebraic to the
analytic category and prove the following statement.
Lemma 11.4. With notations as in Construction 11.3, the
incidence correspondence M is. . .
M U G(2, 4)
(a) closed in the Zariski topology of U G(2, 4);

(b) locally in the classical topology the graph of a continuously differentiable function U G(2, 4), as shown
in the picture on the right.
Proof. Let (X, L) M. By a linear change of coordinates we can assume that L is given by the equations x2 = x3 = 0. Locally around this point L G(2, 4) in the Zariski topology we can use the affine
coordinates on the Grassmannian as in Construction 8.15, namely a2 , a3 , b2 , b3 C corresponding
to the line in P3 spanned by the rows of the matrix

1 0 a2 a3
,
0 1 b2 b3
with the point (a2 , a3 , b2 , b3 ) = (0, 0, 0, 0) corresponding to L. On the space U of smooth cubic
surfaces we use the coordinates (c ) as in Construction 11.3. In these coordinates (a, b, c) =
92
Andreas Gathmann
(a2 , a3 , b2 , b3 , c ) on U G(2, 4), the incidence correspondence M is then given by

(a, b, c) M fc (s (1, 0, a2 , a3 ) + t (0, 1, b2 , b3 )) = 0 for all s,t
c s0 t 1 (s a2 + t b2 )2 (s a3 + t b3 )3 = 0 for all s,t
: sit 3i Fi (a, b, c) = 0 for all s,t

i
Fi (a, b, c) = 0 for 0 i 3.
This shows (a), since F0 , . . . , F3 are polynomial functions in a, b, c. The claim of (b) is that these
four equations determine (a2 , a3 , b2 , b3 ) locally around the origin in the classical topology in terms
of c. Of course, we will prove this with (the complex version of) the Implicit Function Theorem
0 ,F1 ,F2 ,F3 )
[G2, Proposition 26.9]. All we have to show is therefore that the Jacobian matrix J := (F
(a2 ,a3 ,b2 ,b3 )
is invertible at a = b = 0.
So let us compute this Jacobian matrix. Note that

i 3i
=
f
(s,t,
s
a
+
t
b
,
s
a
+
t
b
)
s
t
F

c
i
2
2
3
3
a2
a
a=b=0
a=b=0
2
i
=s
fc
(s,t, 0, 0).
x2
The (s,t)-coefficients of this polynomial are the first column in the matrix J. Similarly, the other
columns are obviously s xfc (s,t, 0, 0), t xfc (s,t, 0, 0), and t xfc (s,t, 0, 0). Hence, if the matrix J was
3
2
3
not invertible, there would be a relation
(2 s + 2t)
fc
fc
(s,t, 0, 0) + (3 s + 3t)
(s,t, 0, 0) = 0
x2
x3
identically in s,t, with (2 , 2 , 3 , 3 ) C4 \{0}. As homogeneous polynomials in two variables

always decompose into linear factors, this means that xfc (s,t, 0, 0) and xfc (s,t, 0, 0) must have a
2
common linear factor, i. e. that there is a point p = (p0 , p1 , 0, 0) L with
fc
fc
x2 (p) = x3 (p) = 0.
But as the line L lies in the cubic Vp ( fc ), we also have fc (s,t, 0, 0) = 0 for all s,t. Differentiating this
with respect to s and t gives xfc (p) = 0 and xfc (p) = 0, respectively. Hence all partial derivatives
0
1
of fc vanish at p L X. By the Jacobi criterion of Exercise 10.12 (b) this means that p would be
a singular point of X, in contradiction to our assumption. Hence J must be invertible, and part (b) of
the lemma follows.

Corollary 11.5. Every smooth cubic surface contains exactly 27 lines.
Proof. In this proof we will work with the classical topology throughout. Let X U be a fixed
smooth cubic, and let L P3 be an arbitrary line. We distinguish two cases:
Case 1: If L lies in X, Lemma 11.4 (b) shows that there is an open neighborhood VL WL of (X, L) in
U G(2, 4) in which the incidence correspondence M is the graph of a continuously differentiable
function. In particular, every cubic in VL contains exactly one line in WL .
Case 2: If L does not lie in X there is an open neighborhood VL WL of (X, L) such that no cubic in
VL contains any line (since the incidence correspondence is closed by Lemma 11.4 (a)).
Now let L vary. As the Grassmannian G(2, 4) is projective, and hence compact, there are finitely
many WL that cover G(2, 4). Let V be the intersection of the corresponding VL , which is then again
an open neighborhood of X. By construction, in this neighborhood V all cubic surfaces have the
same number of lines (namely the number of WL coming from case 1). As this argument holds for
any cubic, we conclude that the number of lines contained in a cubic surface is a locally constant
function on U.
11.
93
To see that this number is also globally constant, it therefore suffices to show that U is connected.
But this follows from Exercise 10.22 (b): we know that U is the complement of a proper Zariskiclosed subset in P19 . But as such a closed subset has complex codimension at least 1 and hence real
codimension at least 2, taking this subset away from the smooth and connected space P19 leaves us
again with a connected space.

Remark 11.6.
(a) In topological terms, the argument of the proof of Corollary 11.5 says that the map : M U
of Construction 11.3 is a 27-sheeted covering map.
(b) Applying the methods of Lemma 11.4 and Corollary 11.5 to suitable incidence correspondences involving two resp. three lines in cubic surfaces, one can show similarly that the
statements of Corollary 11.2 hold for an arbitrary smooth cubic surface X as well: there
are exactly 10 lines in X meeting any given one, and exactly 5 lines in X meeting any two
disjoint given ones.
Note that a cubic surface X is clearly not isomorphic to P2 : by Remark 11.6 (b) there are two disjoint
lines on X, whereas in P2 any two curves intersect by Exercise 6.32 (b). However, we will now see
that X is birational to P2 , and that it is in fact isomorphic to a blow-up of P2 at six points.
Proposition 11.7. Any smooth cubic surface is birational to P2 .
L
Proof. By Remark 11.6 (b) there are two disjoint lines L1 , L2 X.

The following mutually inverse rational maps X 99K L1 L2 and
L1 L2 99K X show that X is birational to L1 L2
= P1 P1 , and
2
hence to P :
X 99K L1 L2 : By Exercise 6.30, for every point a not on L1 or L2
there is a unique line L in P3 through L1 , L2 , and a. Take the rational
map from X to L1 L2 sending a to (a1 , a2 ) := (L1 L, L2 L),
which is obviously well-defined away from L1 L2 .
L1
a1
a
L2
a2
X
L1 L2 99K X: Map any pair of points (a1 , a2 ) L1 L2 to the third intersection point of X with
the line L through a1 and a2 . This is well-defined whenever L is not contained in X.

Proposition 11.8. Any smooth cubic surface is isomorphic to P2 blown up in 6 (suitably chosen)
points.
Proof. We will only sketch the proof. Let X be a smooth cubic surface, and let f : X 99K L1 L2
=
P1 P1 be the rational map as in the proof of Proposition 11.7.
First of all we claim that f is actually a morphism. To see this, note that there is a different description
for f : if a X\L1 , let H be the unique plane in P3 that contains L1 and a, and set f2 (a) = H L2 .
If one defines f1 (a) similarly, then f (a) = ( f1 (a), f2 (a)). Now if the point a lies on L1 , let H be the
tangent plane to X at a, and again set f2 (a) = H L2 . Extending f1 similarly, one can show that this
extends f = ( f1 , f2 ) to a well-defined morphism X P1 P1 on all of X.
Now let us investigate where the inverse map P1 P1 99K X is not well-defined. As already mentioned in the proof of Proposition 11.7, this is the case if the point (a1 , a2 ) L1 L2 is chosen so that
a1 a2 X. In this case, the whole line a1 a2 will be mapped to (a1 , a2 ) by f , and it can be checked
that f is actually locally the blow-up of this point. By Remark 11.6 (b) there are exactly 5 such lines
a1 a2 on X. Hence X is the blow-up of P1 P1 in 5 points, i. e. by Lemma 9.28 the blow-up of P2 in
6 suitably chosen points.

Remark 11.9. It is interesting to see the lines on a cubic surface X in the picture of Proposition 11.8
in which we think of X as a blow-up of P2 in 6 points. It turns out that the 27 lines correspond to the
following curves that we know already (and that are all isomorphic to P1 ):
the 6 exceptional hypersurfaces,
94
Andreas Gathmann
the strict transforms of the

the strict transforms of the
7.31 (c)).
6
2 = 15 lines through two of the blown-up points,
6
5 = 6 conics through five of the blown-up points
(see Exercise
In fact, it is easy to check by the above explicit description of the isomorphism of X with the blow-up
of P2 that these curves on the blow-up actually correspond to lines on the cubic surface.
It is also interesting to see again in this picture that every such line meets 10 of the other lines,
as mentioned in Remark 11.6 (b):
Every exceptional hypersurface intersects the 5 lines and the 5 conics that pass through this
blown-up point.
Every line through two of the blown-up points meets
the 2 exceptional hypersurfaces of the blown-up points,

the 42 = 6 lines through two of the four remaining points,
the 2 conics through the four remaining points and one of the blown-up points.
Every conic through five of the blown-up points meets the 5 exceptional hypersurfaces at
these points, as well as the 5 lines through one of these five points and the remaining point.
4+5
Exercise 11.10. As in Exercise 10.22 (b) let U P( 4 )1 = P125 be the set of smooth (3dimensional) hypersurfaces of degree 5 in P4 . Prove:
(a) The incidence correspondence

{(X, L) U G(2, 5) : L is a line contained in X}
is smooth of dimension 125, i. e. of the same dimension as U.
(b) Although (a) suggests that a smooth hypersurface of degree 5 in P4 contains only finitely
many lines, the Fermat hypersurface Vp (x05 + + x45 ) P4 contains infinitely many lines.
12.
12.
95
After our study of smooth cubic surfaces in the last chapter, let us now come back to the general
theory of algebraic geometry. Our main goal of this chapter will be to determine the number of
intersection points of given varieties (in case this number is finite). For example, let X and Y be
two plane curves, with (principal) ideals generated by two polynomials f and g, respectively. If they
do not have a common irreducible component, their intersection will be zero-dimensional, and we
can ask for the number of points in X Y . We will see in Bzouts Theorem as in Corollaries 12.20
(b) and 12.26 (b) that this number of points is at most deg f deg g, and that we can even make this
an equality if we count the points with suitable multiplicities. We have seen a special case of this
already in Exercise 4.13, where one of the two curves was a line or a conic.
In particular, this statement means that the number of points in X Y (counted with multiplicities)
depends only on the degrees of the defining polynomials, and not on the polynomials themselves.
One can view this as a direct generalization of the statement that a degree-d polynomial in one
variable always has d zeroes, again counted with multiplicities.
In order to set up a suitable framework for Bzouts Theorem, we have to take note of the following
two technical observations:
As mentioned above, we have to define suitable intersection multiplicities, e. g. for two plane
curves X and Y . We have motivated in Remark 1.27 already that such multiplicities are
encoded in the (possibly non-radical) ideal I(X) + I(Y ). Most constructions in this chapter
are therefore based on ideals rather than on varieties, and consequently commutative algebra
will play a somewhat greater role than before.
The simplest example of Bzouts Theorem is that two distinct lines in the plane always meet
in one point. This would clearly be false in the affine setting, where two such lines might
be parallel. We therefore have to work with projective varieties that can have intersection
points at infinity in such cases.
Taking these two points into account, we see that our main objects of study will have to be homogeneous ideals in polynomial rings. The central concept that we will need is the Hilbert function of
such an ideal.
Definition 12.1 (Hilbert functions).
(a) Let I E K[x0 , . . . , xn ] be a homogeneous ideal. Then K[x0 , . . . , xn ]/I is a finite-dimensional
graded K-algebra by Lemma 6.10 (c). We can therefore define the function
hI : N N, d 7 dimK K[x0 , . . . , xn ]d /Id
encoding the dimensions of the graded parts of this quotient algebra. It is called the Hilbert
function of I.
(b) For a projective variety X Pn we set hX := hI(X) , so that
hX : N N, d 7 dimK S(X)d ,
where S(X) = K[x0 , . . . , xn ]/I(X) is the homogeneous coordinate ring of X as in Construction
6.18. We call hX the Hilbert function of X.
Remark 12.2. Note that the Hilbert function of an ideal is invariant under projective automorphisms as in Example 7.6 (a): an invertible matrix corresponding to an automorphism Pn Pn also
defines an isomorphism An+1 An+1 , and hence by Corollary 4.8 an isomorphism K[x0 , . . . , xn ]
K[x0 , . . . , xn ] of K-algebras. As this isomorphism respects the grading, any ideal has the same Hilbert
function as its image under this isomorphism.
19
96
Andreas Gathmann
Example 12.3.
(a) The Hilbert function of Pn is given by hPn (d) = dimK K[x0 , . . . , xn ]d =
n+d
n
for d N.
(b) Let I E K[x0 , . . . , xn ] be a homogeneous ideal with Vp (I) = 0.

/ Then I = (x0 , . . . , xn ) or
I = (1) by the projective Nullstellensatz of Proposition 6.22. In both cases we have xiki I
for suitable ki N for all i. This means that all monomials of degree at least k := k0 + + kn
are contained in I. Hence Id = K[x0 , . . . , xn ]d for all d k, or in other words
hI (d) = 0
for almost all d N,
where as usual we use the term almost all for all but finitely many.
(c) Let X = {a} Pn be a single point. To compute its Hilbert function we may assume by Remark 12.2 that a = (1 : 0 : : 0), so that I(a) = (x1 , . . . , xn ). Then S(X) =
K[x0 , . . . , xn ]/I(a)
= K[x0 ], and hence
for all d N.
hX (d) = 1
Exercise 12.4. Compute the Hilbert function of. . .
(a) the ideal (x02 x12 , x03 ) E K[x0 , x1 ];
(b) two intersecting lines in P3 ;
(c) two non-intersecting lines in P3 .
In order to work with Hilbert functions it is convenient to adopt the language of exact sequences
from commutative algebra. The only statement that we will need about them is the following.
Lemma and Definition 12.5 (Exact sequences). Let f : U V and g : V W be linear maps
of vector spaces over K. Assume that f is injective, g is surjective, and that im f = ker g. These
assumptions are usually summarized by saying that
f
0 U V W 0
is an exact sequence [G5, Chapter 4].
Then dimK V = dimK U + dimK W .
Proof. This is just standard linear algebra: we have
dimK V = dimK ker g + dimK im g = dimK im f + dimK im g = dimK U + dimK W,
with the last equation following since f is injective and g is surjective.
Proposition 12.6. For any two homogeneous ideals I, J E K[x0 , . . . , xn ] we have

hIJ + hI+J = hI + hJ .
Proof. Set R = K[x0 , . . . , xn ]. It is easily checked that
0 R/(I J) R/I R/J R/(I + J) 0

f
7
f, f

f ,g
7 f g
is an exact sequence. Taking its degree-d part and applying Lemma 12.5 gives the desired result.
Example 12.7.
(a) Let X and Y be disjoint projective varieties in Pn . Then I(X) I(Y ) = I(X Y ) by Remark 6.23. Moreover, by the same remark the ideal I(X) + I(Y ) has empty zero locus since
V (I(X) + I(Y )) = V (I(X)) V (I(Y )) = X Y = 0,
/ and hence its Hilbert function is almost
everywhere zero by Example 12.3 (b). Proposition 12.6 thus implies that
hXY (d) = hX (d) + hY (d) for almost all d N.
12.
97
In particular, this means by Example 12.3 (c) that for a finite set X = {a1 , . . . , ar } of r points
we have
hX (d) = r for almost all d N.
(b) Let I = (x12 ) E K[x0 , x1 ]. It is a non-radical ideal whose projective zero locus consists of the
single point (1 : 0) P1 . In fact, it can be viewed as an ideal describing this point with
multiplicity 2 as in Remark 1.27.
The Hilbert function remembers this multiplicity: as K[x0 , x1 ]d /Id has basis x0d and x0d1 x1
for d 1, we see that hI (d) = 2 for almost all d, in the same way as for the Hilbert function
of two distinct points as in (a).
(c) Let X P2 be three points lying on a line. Then there is a homogeneous linear
polynomial in K[x0 , x1 , x2 ] vanishing on X, so that dimK I(X)1 = 1. Hence hX (1) =
dimK K[x0 , x1 , x2 ]1 /I(X)1 = 3 1 = 2. On the other hand, if X consists of three points
not on a line, then no linear polynomial vanishes on X, and consequently hX (1) =
dimK K[x0 , x1 , x2 ]1 /I(X)1 = 3 0 = 3. So in particular, we see that in contrast to Remark
12.2 the Hilbert function is not invariant under arbitrary isomorphisms, since any set of three
points is isomorphic to any other such set.
Together with the result of (a), for a finite set X Pn we can say intuitively that hX (d)
encodes the number of points in X for large values of d, whereas it gives some information
on the relative position of these points for small values of d.
Note that the intersection X Y of two varieties X and Y corresponds to the sum of their ideals.
To obtain a formula for the number of points in X Y we therefore have to compute the Hilbert
functions of sums of ideals. The following lemma will help us to do this in the case when one of the
ideals is principal.
Lemma 12.8. Let I E K[x0 , . . . , xn ] be a homogeneous ideal, and let f K[x0 , . . . , xn ] be a homogeneous polynomial of degree e. Assume that there is a number d0 N with the following property:
for all homogeneous g K[x0 , . . . , xn ] of degree at least d0 with f g I we have g I.
Then hI+( f ) (d) = hI (d) hI (d e) for almost all d N.
Proof. Let R = K[x0 , . . . , xn ]. There is an exact sequence
f
0 Rde /Ide Rd /Id Rd /(I + ( f ))d 0

for all d with d e d0 , where the second map is just the quotient map. In fact, it is obvious that this
quotient map is surjective, and that its kernel is exactly the image of the first map. The injectivity of
the first map is precisely the assumption of the lemma.
The desired statement now follows immediately from Lemma 12.5.
Before we can apply this lemma, we have to analyze the geometric meaning of the somewhat technical assumption that f g I implies g I for all polynomials g (of sufficiently large degree).
Remark 12.9. Assume that I = I(X) is the (radical) ideal of a projective variety X. Consider the
irreducible decomposition X = X1 Xr of X, so that I = I(X1 ) I(Xr ) by Remark 6.23 (c).
We claim that the assumption of Lemma 12.8 is then satisfied if f does not vanish identically on any
Xi . In fact, in this case f is non-zero in the integral domain S(Xi ) for all i (see Exercise 6.31 (b)).
Hence g f I, i. e. g f = 0 S(Xi ), implies g = 0 S(Xi ) for all i, and thus g I.
If I is not radical, a similar statement holds but in order for this to work we need to be able to
decompose I as an intersection of ideals corresponding to irreducible varieties again. This so-called
primary decomposition of I is one of the main topics in the Commutative Algebra class [G5, Chapter
8]. We will therefore just quote the results that we are going to need.
98
Andreas Gathmann
Remark 12.10 (Primary decompositions). Let I E K[x0 , . . . , xn ] be an arbitrary ideal. Then there is
a decomposition
I = I1 Ir
into primary ideals I1 , . . . , Ir , which means by definition that g f Ii implies g Ii or f Ii for all
i and all polynomials f , g K[x0 , . . . , xn ] [G5, Definition 8.9 and Proposition 8.16]. Moreover, this
decomposition satisfies the following properties:
(a) The zero loci Va (Ii ) are irreducible:
by Proposition 2.9 and the Nullstellensatz, this is the
same as saying that K[x0, . . . , xn ]/ Ii is an integral domain. So assume that f and g are two
polynomials with g f Ii . Then gk f k Ii for some k N.
But this implies that gk Ii or
p
k
f Ii since Ii is primary, and hence that g Ii or f
Ii = Ii .
(b) Applying Remark 1.25 (c) to our decomposition, we see that
Va (I) = Va (I1 ) Va (Ir ).
In particular, by (a) all irreducible components of Va (I) must be among the varieties
Va (I1 ), . . . ,Va (Ir ). Moreover, we can assume that no two of these
p varieties coincide: if
Va (Ii ) = Va (I j ) for some i 6= j, i. e. by the Nullstellensatz Ii = I j , we can replace the
two ideals Ii and I j by the single ideal Ii I j in the decomposition, which is easily seen to be
primary again.
However, it may well happen that there are (irreducible) varieties among Va (I1 ), . . . ,Va (Ir )
that are strictly contained in an irreducible component of Va (I) [G5, Example 8.23]. These
varieties are usually called the embedded components of I. In the primary decomposition,
the ideals corresponding to the irreducible components are uniquely determined, whereas
the ones corresponding to the embedded components are usually not [G5, Example 8.23 and
Proposition 8.34].
Using these primary decompositions, we can now show for a homogeneous ideal I E K[x0 , . . . , xn ]
that, by a suitable homogeneous linear change of coordinates, we can always achieve that f = x0
satisfies the condition of Lemma 12.8. In fact, assume that g is a homogeneous polynomial such that
gx0 Ii for all i. We distinguish two cases:
If Va (Ii ) {0} then Ii (x0 , . . . , xn ) by the Nullstellensatz. Hence K[x0 , . . . , xn ]d Ii for

large d in the same way as in Example 12.3 (b), which means that g Ii if the degree of g is
big enough.
If Va (Ii ) 6 {0} a general homogeneous linear change of coordinates will assure that Va (Ii )
is not contained inthe hypersurface Va (x0 ). Then x0 is not identically zero on Va (Ii ), so that
x0
/ Ia (Va (Ii )) = Ii . Since Ii is primary, we conclude that g Ii .
Let us now come back to our study of Hilbert functions. We have already seen that the important
information in hI concerning the number of intersection points of varieties is contained in its values
hI (d) for large d. We therefore have to study the behavior of hI (d) as d . The central result in
this direction is that the Hilbert function is eventually polynomial, with particularly the degree and
the leading coefficient of this polynomial deserving special attention.
Proposition and Definition 12.11 (Hilbert polynomials). Let I E K[x0 , . . . , xn ] be a homogeneous
ideal. Then there is a unique polynomial I Q[d] such that I (d) = hI (d) for almost all d N.
Moreover,
(a) The degree of I is m := dimVp (I).
(b) If Vp (I) 6= 0,
/ the leading coefficient of I is
1
m!
times a positive integer.
The polynomial I is called the Hilbert polynomial of I. For a projective variety X Pn we set
X := I(X) .
12.
99
Proof. It is obvious that a polynomial with infinitely many fixed values is unique. So let us prove
the existence of I by induction on m = dimVp (I). The start of the induction follows from Example
12.3 (b): for Vp (I) = 0/ we obtain the zero polynomial I .
Let us now assume that Vp (I) 6= 0.
/ By a homogeneous linear change of coordinates (which does not
affect the Hilbert function by Remark 12.2) we can assume that the polynomial x0 does not vanish
identically on any irreducible component of Vp (I). Hence dimVp (I + (x0 )) m 1, and so by our
induction on m we know that d 7 hI+(x0 ) (d) is a polynomial of degree at most m 1 for large d.

d
as a basis of the vector space of
For reasons that will be apparent later, let us choose d0 , . . . , m1
polynomials in d of degree at most m 1, so that for suitable c0 , . . . , cm1 Q we can write
m1
d
for almost all d N.
hI+(x0 ) (d) = ci
i
i=0
Moreover, by Remark 12.10 we can assume that Lemma 12.8 is applicable for f = x0 , so that
m1
d
hI (d) hI (d 1) = hI+(x0 ) (d) = ci
for almost all d N.
i
i=0
We will now show by induction on d that there is a constant c Q such that

m1
d +1
hI (d) = c + ci
for almost all d N.
i+1
i=0
(1)
(2)
The start of the induction is trivial, since we can always adjust c so that this equation holds at a single
value of d (chosen so that (1) holds for all larger values of d). But then for all larger d we have

m1
m1
m1
d +1
d +1
d +1
d +2
hI (d + 1) = hI (d) + ci
= c + ci
+
= c + ci
i
i+1
i
i+1
i=0
i=0
i=0
by (1) and the induction assumption. As the right hand side of (2) is a polynomial in d (of degree at
most m), this proves the existence of I .
1
times a positive integer, thus
Finally, if Vp (I) 6= 0/ let us show that the d m -coefficient of I is m!
proving the additional statements (a) and (b). We will do this again by induction on m.
m = 0: In this case I is a constant, and it is clearly a non-negative integer, since it is by

definition the dimension of K[x0 , . . . , xn ]d /Id for large d. Moreover, it cannot be zero, since
otherwise Id = K[x0 , . . . , xn ]d for some d, which implies xid Id for all i and thus Vp (I) = 0.
/
m > 0: In this case Vp (I) has an irreducible component of dimension m (and none of bigger
dimension). In our proof above, the zero locus of x0 on this component is non-empty by
Exercise 6.32 (b), and of dimension m 1 by Proposition 2.25 (c). Hence dimVp (I + (x0 )) =
m 1, and so by induction I+(x0 ) is a polynomial of degree m 1, with (m 1)! times the
leading coefficient being a positive integer. But note that in the proof above this integer is
just cm1 , which is then also m! times the leading coefficient of I by (2).

Remark 12.12. Of course, all our statements concerning the values of the Hilbert function d 7 hI (d)
at large values of d can be transferred immediately to the Hilbert polynomial. For example, Example
12.3 implies that I = 0 (as a polynomial) if Vp (I) = 0,
/ and I = 1 if I is the ideal of a point. Similar
statements hold for Proposition 12.6, Example 12.7, and Lemma 12.8.
Definition 12.13 (Degree). Let I E K[x0 , . . . , xn ] be a homogeneous ideal with non-empty projective
zero locus, and let m = dimVp (I). Then m! times the leading coefficient of I , which is a positive
integer by Proposition 12.11, is called the degree deg I of I. The reason for this name will become
clear in Example 12.17.
For a projective variety X, its degree is defined as deg X := deg I(X).
Example 12.14.
(a) The degree of Pn is n! times the d n -coefficient of dimK K[x0 , . . . , xn ]d =
By Example 12.3 (c), the degree of a single point is 1 as well.
n+d
n
n , i. e. deg P = 1.
100
Andreas Gathmann
More generally, if X Pn is any linear space then its homogeneous coordinate ring is isomorphic to a polynomial ring, and hence as above deg X = 1 again.
(b) Let X and Y be projective varieties in Pn of the same dimension m, and assume that they
do not have a common irreducible component. Then the zero locus X Y of I(X) + I(Y )
has dimension smaller than m, so that I(X)+I(Y ) has degree less than m. Moreover, we have
I(X) I(Y ) = I(X Y ), and hence considering m! times the degree-m coefficients in the
Hilbert polynomials of I(X) I(Y ), I(X) + I(Y ), I(X), and I(Y ) yields
deg(X Y ) = deg X + degY
by Proposition 12.6 (which of course holds for the Hilbert polynomials as well as for the
Hilbert functions).
(c) Let I E K[x0 , . . . , xn ] be a homogeneous ideal
with finite zero locus, consisting of r points.
Then I = r by Example 12.7 (a). But I I also implies I I , and so we conclude

that
deg I = I I = r.
In fact, in Corollary 12.26 we will refine this statement by interpreting deg I as a sum of multiplicities for each point in Vp (I), with each of these multiplicities being a positive integer.
Exercise 12.15.
n+m
n .
of Pn is
(a) Show that the degree of the Segre embedding of Pm Pn is

20
(b) Show that the degree of the degree-d Veronese embedding
dn.
We are now ready to prove the main result of this chapter.

Proposition 12.16 (Bzouts Theorem). Let X Pn be a projective variety of dimension at least
1, and let f K[x0 , . . . , xn ] be a homogeneous polynomial that does not vanish identically on any
irreducible component of X. Then
deg(I(X) + ( f )) = deg X deg f .
Proof. Let m = dim X. By Definition 12.13, the Hilbert polynomial of X is given by
deg X m
d + a d m1 + (terms of degree less than m 1)
m!
for some a Q. So by Remark 12.9 we can apply Lemma 12.8 and obtain with e := deg f
X (d) =
I(X)+( f ) (d) = X (d) X (d e)

deg X m
=
(d (d e)m ) + a (d m1 (d e)m1 ) + (terms of degree less than m 1)
m!
e deg X m1
d
+ (terms of degree less than m 1).
=
(m 1)!
By Definition 12.13 again, this means that deg(I(X) + ( f )) = e deg X = deg X deg f .
Example 12.17. Let I = ( f ) E K[x0 , . . . , xn ] be a principal ideal. Then Bzouts Theorem together
with Example 12.14 (a) implies
deg I = deg((0) + ( f )) = deg Pn deg f = deg f .
In particular, if X Pn is a hypersurface, so that I(X) = ( f ) for some homogeneous polynomial f
by Exercise 7.16 (a), then deg X = deg f . This justifies the name degree in Definition 12.13.
Exercise 12.18. Prove that every pure-dimensional projective variety of degree 1 is a linear space.
Notation 12.19. Let X Pn be an irreducible projective variety. In certain cases there are commonly
used names to describe the degree and / or the dimension of X that we have probably used informally
already several times:
12.
101
(a) If deg X = 1 then X is called a line if dim X = 1, a plane if dim X = 2, and a hyperplane if
dim X = n 1.
(b) For any dimension, X is called a quadric if deg X = 2, a cubic if deg X = 3, a quartic if
deg X = 4, and so on.
Corollary 12.20 (Bzouts Theorem for curves).
(a) Let X Pn be a curve, and let f K[x0 , . . . , xn ] be a homogeneous polynomial that does not
vanish identically on any irreducible component of X. Then
|X V ( f )| deg X deg f .
(b) For any two curves X and Y in P2 without a common irreducible component we have
|X Y | deg X degY.
Proof.
(a) As I(X) + ( f ) is an ideal with zero locus X V ( f ), the statement follows from Bzouts
Theorem together with Example 12.14 (c).
(b) Apply (a) to a polynomial f generating I(Y ), and use Example 12.17.
For the remaining part of this chapter we will focus on the case of curves as in Corollary 12.20. Our
goal is to assign a natural multiplicity to each point in X V ( f ) (resp. X Y ) so that the inequality
becomes an equality when all points are counted with their respective multiplicities. In order to
achieve this we have to study the degree of a homogeneous ideal with zero-dimensional zero locus
from a local point of view. It is convenient to do this in an affine chart of Pn , and then finally in the
local rings.
Exercise 12.21. Let I E K[x0 , . . . , xn ] be a homogeneous ideal with dimVp (I) = 0. Assume that we
have chosen coordinates so that all points in Vp (I) have a non-vanishing x0 -coordinate. Prove that
the degree of I is then
deg I = I = dimK K[x1 , . . . , xn ]/J,
where J = { f (1, x1 , . . . , xn ) : f I} E K[x1 , . . . , xn ].
The following lemma now expresses this dimension as a sum of local dimensions. In case you
have attended the Commutative Algebra class already you will probably recognize this as precisely
the Structure Theorem for Artinian rings, stating that an Artinian ring is always the product of its
localizations [G5, Proposition 7.20].
Lemma 12.22. Let J E K[x1 , . . . , xn ] be an ideal with finite affine zero locus Va (J) = {a1 , . . . , ar }.
Then
K[x1 , . . . , xn ]/J
= OAn ,a1 /J OAn ,a1 OAn ,ar /J OAn ,ar ,
where J OAn ,ai denotes the ideal in OAn ,ai generated by all elements
f
1
for f J.
Proof. Consider the primary decomposition of J as in Remark 12.10. By part (b) of this remark it
is of the form J = J1 Jr for some ideals J1 , . . . , Jr with Va (Ji ) = {ai } for all i. Moreover, note
that Ji OAn ,a j is the unit ideal for i 6= j since a j
/ Va (Ji ) implies that there is a polynomial in Ji not
vanishing at a j , so that it is a unit in the local ring OAn ,a j . Hence it suffices to prove that
K[x1 , . . . , xn ]/J
= OAn ,a1 /J1 OAn ,a1 OAn ,ar /Jr OAn ,ar .
We will do this by showing that the K-algebra homomorphism
: K[x1 , . . . , xn ]/J OAn ,a1 /J1 OAn ,a1 OAn ,ar /Jr OAn ,ar , f 7 f , . . . , f
is bijective.
102
Andreas Gathmann

is injective: Let f be a polynomial with f = 0. Then f lies in Ji OAn ,ai for all i, i. e.
f
gi
1 = fi for some gi , f i with gi Ji and f i K[x0 , . . . , xn ] such that f (ai ) 6= 0. This means that
hi ( fi f
gi ) = 0 for some hi with hi (ai ) 6= 0, and hence hi fi f Ji . But hi fi
/ I(ai ) means
hi fi
/ Ji , and thus f Ji since Ji is primary. As this holds for all i, we conclude that f J,
i. e. f = 0 in K[x1 , . . . , xn ]/J.
is surjective: By symmetry of the factors it suffices toprove that (1, 0, . . . , 0) im . As
V (J1 + Ji ) = {a1 } {ai } = 0/ for all i > 1 we see that 1 J1 + Ji , and hence also 1 J1 + Ji .
There are thus ai J1 and bi Ji with ai + bi = 1, so that bi 0 mod Ji and bi 1 mod J1 .
Hence the product b2 br is an inverse image of (1, 0, . . . , 0) under .
It is now straightforward to translate Bzouts Theorem for curves into a local version.
Definition 12.23 (Multiplicities).
(a) Let I E K[x0 , . . . , xn ] be a homogeneous ideal with finite projective zero locus, and let a Pn .
Choose an affine patch of Pn containing a, and let J be the corresponding affine ideal as in
Exercise 12.21. Then
multa (I) := dimK OAn ,a /J OAn ,a
is called the multiplicity of I at a.
(b) Let a be a point on a projective curve X Pn . For a homogeneous polynomial f
K[x0 , . . . , xn ] that does not vanish identically on any irreducible component of X, the number
multa (X, f ) := multa (I(X) + ( f ))
is called the (vanishing) multiplicity of f at a. Note that multa (X, f ) depends only on the
class of f modulo I(X) and not on f itself, so that we can also construct the multiplicity
multa (X, f ) for f S(X). In this case, we will also often simplify its notation to multa ( f ).
If n = 2 and Y P2 is another curve that does not share a common irreducible component
with X, the intersection multiplicity of X and Y at a is defined as
multa (X,Y ) := multa (I(X) + I(Y )).
Remark 12.24 (Positivity of multiplicities). Continuing the notation of Definition 12.23 (a), note
that 1
/ J OAn ,a if and only if a Vp (I). It follows that multa (I) 1 if and only if a Vp (I).
Applying this to Definition 12.23 (b), we see that the vanishing multiplicity multa (X, f ) is at least
1 if and only if f (a) = 0, and the intersection multiplicity multa (X,Y ) is at least 1 if and only if
a X Y . In fact, we will show in Exercise 12.27 that there is also an easy geometric criterion for
when multa (X,Y ) = 1.
Remark 12.25 (Vanishing and intersection multiplicities in local rings). It is often useful to express the multiplicities of Definition 12.23 (b) in terms of local rings as in Definition 12.23 (a). As
above, we choose an affine patch {x Pn : xi 6= 0}
= An of Pn containing a. By abuse of notation, if
f K[x0 , . . . , xn ] is a homogeneous polynomial, we will also denote by f the (not necessarily homogeneous) polynomial obtained from it by setting xi equal to 1, and then also its quotient by 1 in the
local ring OAn ,a (see Exercise 3.24). Then Definition 12.23 can be formulated as follows:
(a) Let X Pn be a curve, and let f K[x0 , . . . , xn ] be a homogeneous polynomial not vanishing
identically on any irreducible component of X. Denote by U = X An the affine part of X in
the chosen patch, and let J = I(U) be its ideal. Then the vanishing multiplicity of f at a Pn
is equal to dimK OAn ,a /(J + ( f )) OAn ,a by Definition 12.23. But OAn ,a /J OAn ,a
= OX,a by
Exercise 3.23, and so we conclude that
multa (X, f ) = dimK OX,a /( f ).
Note that we could use the same formula to define the vanishing multiplicity for any local
function f OX,a that does not vanish identically on any irreducible component of X through
a. In fact, for an irreducible variety we will even define such a multiplicity for rational
12.
103
functions in Construction 14.5, which then includes the case of local functions (see Exercise
9.8 (b) and Remark 14.7).
(b) For two curves X,Y P2 without common irreducible component and ideals I(X) = ( f ) and
I(X) = (g) their intersection multiplicity is
multa (X,Y ) = dimK OA2 ,a /( f , g),
or alternatively with (a)
multa (X,Y ) = dimK OX,a /(g) = dimK OY,a /( f ).
Of course, as we have defined the multiplicities above using affine charts, we could construct them
equally well for affine instead of projective varieties. However, the projective case is needed for the
local version of Bzouts Theorem, which we can now prove.
Corollary 12.26 (Bzouts Theorem for curves, local version).
(a) Let X Pn be a curve, and let f K[x0 , . . . , xn ] be a homogeneous polynomial that does not
vanish identically on any irreducible component of X. Then
multa (X, f ) = deg X deg f .
aXV ( f )
(b) For any two curves X and Y in P2 without a common irreducible component we have
multa (X,Y ) = deg X degY.
aXY
Proof.
(a) By Exercise 12.21, Lemma 12.22, and the definition of multiplicities, all applied to the ideal
I(X) + ( f ), we have
deg(I(X) + ( f )) =
multa (X, f ).
aXV ( f )
Hence the statement follows immediately from Proposition 12.16.

(b) This follows from (a) for f a polynomial generating I(Y ), since degY = deg f by Example
12.17.

Exercise 12.27 (Geometric interpretation of intersection multiplicities). Let X,Y A2 be two affine
curves containing the origin. Moreover, let I(X) = ( f ) and I(Y ) = (g) be their ideals. Show that the
following statements are equivalent:
(a) dimK OA2 ,0 /( f , g) = 1 (i. e. the intersection multiplicity of X and Y at the origin is 1).
(b) X and Y are smooth at 0 and have different tangent spaces there (i. e. X and Y intersect
transversely at the origin).
Example 12.28. Consider the two projective curves
X
X = V (x0 x2 x12 ) and Y = V (x2 )

in P2 , whose affine parts (which we have already considered in Remark
1.27) are shown in the picture on the right. Note that deg X = 2 and
degY = 1 by Example 12.17, and that a := (1 : 0 : 0) is the only point in
the intersection X Y . As X and Y have the same tangent space at a, we
must have multa (X,Y ) 2 by Exercise 12.27.
Y
a
In fact, it is easy to compute multa (X,Y ) explicitly: by definition we have

multa (X,Y ) = multa (x0 x2 x12 , x2 ) = dimK OA2 ,0 /(x2 x12 , x2 ) = dimK OA2 ,0 /(x12 , x2 ),
and since a is the only intersection point of X and Y we can rewrite this by Lemma 12.22 as
multa (X,Y ) = dimK K[x1 , x2 ]/(x12 , x2 ) = dimK K[x1 ]/(x12 ) = 2.
104
Andreas Gathmann
Note that this is in accordance with Bzouts Theorem as in Corollary 12.26 (b), since multa (X,Y ) =
2 = deg X degY .
13.
13.
Applications of Bzouts Theorem
105
Bzouts Theorem (as in Proposition 12.16 or Corollary 12.26) is clearly one of the most powerful
results in algebraic geometry that we will discuss in this class. To illustrate this, let us now take
some time to study several of its applications, which are in fact of quite different flavors.
Our first application is actually not much more than a simple remark. It concerns the question
whether the ideal of a given variety of pure dimension n k in affine space An or projective space
Pn can be generated by k elements. We have seen in Exercise 7.16 (a) already that this is always the
case if k = 1. On the other hand, Example 0.11 and Exercise 2.32 show that for k 2 one sometimes
needs more than k generators. Bzouts Theorem can often be used to detect when this happens, e. g.
in the following setting.
Proposition 13.1. Let X P3 be a curve that is not contained in any proper linear subspace of P3 .
If deg X is a prime number, then I(X) cannot be generated by two elements.
Proof. Assume for a contradiction that I(X) = ( f , g) for two homogeneous polynomials f , g
K[x0 , x1 , x2 , x3 ]. Clearly, g does not vanish identically on any irreducible component of V ( f ), since
otherwise the zero locus of ( f , g) would have codimension 1. By Proposition 12.16 and Example
12.17 we therefore have
deg X = deg(( f ) + (g)) = deg( f ) deg g = deg f deg g.
As deg X is a prime number, this is only possible if deg f = 1 or deg g = 1, i. e. if one of these
polynomials is linear. But then X is contained in a proper linear subspace of P3 , in contradiction to
our assumption.

Example 13.2. Let X be the degree-3 Veronese embedding of P1 as in Construction 7.27, i. e.
X = {(s3 : s2t : st 2 :t 3 ) : (s :t) P1 }
P3 .
By Exercise 12.15 (b) we know that X is a cubic curve, i. e. deg X = 3. Moreover, X is not contained
in any proper linear subspace of P3 , since otherwise the monomials s3 , s2t, st 2 ,t 3 would have to
satisfy a K-linear relation. Hence Proposition 13.1 implies that the ideal I(X) cannot be generated
by two elements.
However, one can check directly that I(X) can be generated by three elements. For example, we can
write
I(X) = (x0 x2 x12 , x1 x3 x22 , x0 x3 x1 x2 ).
In the spirit of Bzouts Theorem, we can also see geometrically why none of these three generators
is superfluous: if we leave out e. g. the last generator and consider I = (x0 x2 x12 , x1 x3 x22 ) instead,
we now have deg I = 2 2 = 4. Clearly, V (I) still contains the cubic X, and hence by the additivity of
degrees as in Example 12.14 (b) there must be another 1-dimensional component in V (I) of degree
1. In fact, this component is easy to find: we have V (I) = X L for the line L = V (x1 , x2 ).
Exercise 13.3. Let X Pn be an irreducible curve of degree d. Show that X is contained in a linear
subspace of Pn of dimension at most d.
As another application of Bzouts Theorem, we are now able to prove the result already announced
in Example 7.6 (a) that any isomorphism of Pn is linear, i. e. a projective automorphism in the sense
of this example. Note that the corresponding statement would be false in the affine case, as e. g.
f : A2 A2 , (x1 , x2 ) 7 (x1 , x2 + x12 ) is an isomorphism with inverse f 1 : (x1 , x2 ) 7 (x1 , x2 x12 ).
Proposition 13.4. Every isomorphism f : Pn Pn is linear, i. e. of the form f (x) = Ax for an
invertible matrix A GL(n + 1, K).
21
106
Andreas Gathmann
Proof. Let H Pn be a hyperplane (given as the zero locus of one homogeneous linear polynomial),
and let L Pn be a line not contained in H. Clearly, the intersection L H consists of one point
with multiplicity 1 (i. e. I(L) + I(H) has multiplicity 1 in the sense of Definition 12.23 (a)). As f
is an isomorphism, f (L) and f (H) must again be a curve resp. a hypersurface that intersect in one
point with multiplicity 1. By the local version of Bzouts Theorem as in Corollary 12.26 (a), this
means that deg f (L) deg f (H) = 1. This is only possible if deg f (H) = 1. In other words, f must
map hyperplanes to hyperplanes.
By composing f with a suitable projective automorphism (i. e. a linear map as in Example 7.6 (a)) we
can therefore assume that f maps the affine part An = Pn \V (x0 ) isomorphically to itself. Passing to
this affine part with coordinates x1 , . . . , xn , the above argument shows that f 1 (V (xi )) = V ( f xi )
is an affine linear space for all i, so that f xi must be a power of a linear polynomial. But
f : K[x1 , . . . , xn ] K[x1 , . . . , xn ] is an isomorphism by Corollary 4.8 and thus maps irreducible
polynomials to irreducible polynomials again. Hence f xi is itself linear for all i, which means that
f is an affine linear map on An , i. e. a linear map on Pn .

For the rest of this chapter we will now restrict to plane curves. One consequence of Bzouts
Theorem in this case is that it gives an upper bound on the number of singular points that such a
curve can have, in terms of its degree.

Proposition 13.5. Let X P2 be an irreducible curve of degree d. Then X has at most d1
2
singular points.
Example 13.6.
(a) A plane curve of degree 1 is a line, which is isomorphic to P1 . A curve of degree 2, i. e. a
conic, is always isomorphic to P1 as well, as we have seen in Example 7.6 (d). So in both
these cases the curve does not have any singular points, in accordance with the statement of
Proposition 13.5.
(b) By Proposition 13.5, a cubic curve in P2 can have at most one singular point. In fact, we
have already seen both a cubic with no singular points (e. g. the Fermat cubic in Example
10.20) and a cubic with one singular point (e. g. Vp (x0 x22 x0 x12 x13 ) or Vp (x0 x22 x13 ), whose
affine parts we have studied in Example 10.3).
(c) Without the assumption of X being irreducible the statement of Proposition 13.5 is false: a
union of two intersecting lines in P2 has degree 2, but a singular point (namely the intersection point). However, we will see in Exercise 13.7 that there is a similar statement with a
slightly higher bound also for not necessarily irreducible curves.
Proof of Proposition 13.5. By Example 13.6 (a) it suffices to prove the
theorem for d 3. Assume for a contradiction that there are distinct singular points a1 , . . . , a(d1)+1 of X. Moreover, pick d 3 arbitrary further
2
distinct points b1 , . . . , bd3 on X, so that the total number of points is

d 1
d
+1+d 3 =
1.
2
2
a1
a2
b1
We claim that there is a curve Y of degree at most d 2 that passes through all these points. The argument is essentially the same as in Exercise 7.31 (c): the space K[x0 , x1 , x2 ]d2 ofall homogeneous
polynomials of degree d 2 in three variables is a vector space of dimension d2 , with the coefficients of the polynomials as coordinates. Moreover, the condition that such a polynomial
vanishes

at a given point is clearly a homogeneous linear equation in these coordinates. As d2 1 homoge
neous linear equations in a vector space of dimension d2 must have a non-trivial common solution,
we conclude that there is a non-zero polynomial f K[x0 , x1 , x2 ]d2 vanishing at all points ai and
b j . The corresponding curve Y = Vp ( f ) then has degree at most d 2 (strictly less if f contains
repeated factors) and passes through all these points.
13.
107
Note that X and Y cannot have a common irreducible component, since X is irreducible and of bigger
degree than Y . Hence Corollary 12.26 (b) shows that the curves X and Y can intersect in at most
deg X degY = d(d 2) points, counted with multiplicities. But the intersection multiplicity at all ai
is at least 2 by Exercise 12.27 since X is singular there. Hence the number of intersection points that
we know already, counted with their respective multiplicities, is at least

d 1
2
+ 1 + (d 3) = d(d 2) + 1 > d(d 2),
2

d
which is a contradiction.
Exercise 13.7. Show that a (not necessarily irreducible) curve of degree d in P2 has at most 2
singular points. Can you find an example for each d in which this maximal number of singular
points is actually reached?
Let us now study smooth plane curves in more detail. An interesting topic that we have neglected
entirely so far is the classical topology of such curves when we consider them over the real or
complex numbers, e. g. their number of connected components in the standard topology. We will
now see that Bzouts Theorem is able to answer such questions.
Of course, for these results we will need some techniques and statements from topology that have
not been discussed in this class. The following proofs in this chapter should therefore rather be
considered as sketch proofs, which can be made into exact arguments with the necessary topological background. However, all topological results that we will need should be intuitively clear
although their exact proofs are often quite technical. Let us start with the real case, as real curves
are topologically simpler than complex ones.
Remark 13.8 (Real curves). Note that we have developed most of our theory only for algebraically
closed ground fields, so that our constructions and results are not directly applicable to real curves.
However, as we will not go very deep into the theory of real algebraic geometry it suffices to note
that the definition of projective curves and their ideals works over R in the same way as over C. To
apply other concepts and theorems to a real curve X with ideal I(X) = ( f ) for a real homogeneous
polynomial f , we will simply pass to the corresponding complex curve XC = Vp ( f ) P2C first. For
example, we will say that X is smooth or irreducible if XC is.
Remark 13.9 (Loops of real projective curves). Let X P2R be a smooth projective curve over R. In
the classical topology, X is then a compact 1-dimensional manifold (see Remark 10.14). This means
that X is a disjoint union of finitely many connected components, each of which is homeomorphic
to a circle. We will refer to these components as loops of X.
Note that X can consist of several loops in the classical topology even if f is irreducible (so that X is
irreducible in the Zariski topology). A convenient way to construct such curves is by deformations
of singular curves. For example, consider the singular cubic curve X in P2R whose affine part in A2R
is the zero locus of f3 := x22 x12 x13 as in Example 10.3. It has a double point at the origin, as
shown in the picture below on the left. In P2R , the curve contains one additional point at infinity that
connects the two unbounded branches, so that X is homeomorphic to two circles glued together at a
point.
f3 (x1 , x2 ) = 0
f3 (x1 , x2 ) = 0
f3 (x1 , x2 ) + = 0
108
Andreas Gathmann
Let us now perturb f3 and consider the curves f3 (x1 , x2 ) = 0 for a small number R>0 instead.
This deforms X into a smooth cubic with one or two loops depending on the sign of the perturbation,
as shown in the picture above. The same technique applied to a singular quartic curve, e. g. the
union of two ellipses given by f4 = (x12 + 2x22 1)(x22 + 2x12 1), yields two or four loops as in the
following picture.
f4 (x1 , x2 ) = 0
f4 (x1 , x2 ) = 0
f4 (x1 , x2 ) + = 0
Remark 13.10 (Even and odd loops). Although all loops of smooth curves in P2R are homeomorphic
to a circle, there are two different kinds of them when we consider their embeddings in projective
space. To understand this, recall from Remark 6.3 that P2R is obtained from A2R (which we will draw
topologically as an open disc below) by adding a point at infinity for each direction in A2R . This has
the effect of adding a boundary to the disc, with the boundary points corresponding to the points at
infinity. But note that opposite points of the boundary of the disc belong to the same direction in A2R
and hence are the same point in P2R . In other words, P2R is topologically equivalent to a closed disc
with opposite boundary points identified, as in the picture below on the left.
a
b
P2R
P2R
P2R
a
b
identify
a
Even loop
Odd loop
The consequence of this is that we have two different types of loops. An even loop is a loop such that
its complement has two connected components, which we might call its interior (shown in dark
in the picture above, homeomorphic to a disc) and exterior (homeomorphic to a Mbius strip),
respectively. In contrast, an odd loop does not divide P2R into two regions; its complement is a single
component homeomorphic to a disc. Note that the distinction between even and odd is not whether
the affine part of the curve is bounded: whereas an odd loop always has to be unbounded, an even
loop may well be unbounded, too. Instead, if you know some topology you will probably recognize
that the statement being made here is just that the fundamental group 1 (P2R ) is isomorphic to Z/2Z;
the two types of loops simply correspond to the two elements of this group.
In principle, a real curve can have even as well as odd loops. There is one restriction however: as the
complement of an odd loop is simply a disc, all other loops in this complement will have an interior
and exterior, so that they are even. In other words, a smooth curve in P2R can have at most one odd
loop.
We are now ready to find a bound on the number of loops in a smooth curve in P2R of a given
degree. Interestingly, the idea in its proof is almost identical to that in Proposition 13.5, although the
resulting statement is quite different.
Proposition 13.11 (Harnacks Theorem). A smooth real projective curve of degree d in P2R has at

most d1
2 + 1 loops.
13.
109
Example 13.12. A line (d = 1) has always exactly one loop. A conic (d = 2) is a hyperbola,
parabola, or ellipse, so in every case the number of loops is again 1 (after adding the points at
infinity). For d = 3 Harnacks Theorem gives a maximum number of 2 loops, and for d = 4 we get
at most 4 loops. We have just seen examples of these numbers of loops in Remark 13.9. In fact,
one can show that the bound given in Harnacks theorem is sharp, i. e. that for every d one can find
smooth real curves of degree d with exactly d1
2 + 1 loops.
Sketch proof of Proposition 13.11. By Example 13.12 it suffices to consider the case d 3. Assume that
the statement of the proposition is false,
i. e. that there are at least d1
2 + 2 loops. We have seen in Remark 13.10
d1
that at least 2 + 1 of these loops must be even. Hence we can pick
points a1 , . . . , a(d1)+1 on distinct even loops of X, and d 3 more points
a1
a1
a2
b1 , . . . , bd3 on another loop (which might be even

or odd). So as in the
proof of Proposition 13.5, we have a total of d2 1 points.
a2
b1
Again as in the proof of Proposition 13.5, it now follows that there is a real curve Y of degree at
most d 2 passing through all these points. Note that the corresponding complex curves XC and YC
as in Remark 13.8 cannot have a common irreducible component since XC is assumed to be smooth,
hence irreducible by Exercise 10.22 (a), and has bigger degree than YC . So Bzouts Theorem as
in Corollary 12.26 (b) implies that XC and YC intersect in at most d(d 2) points, counted with
multiplicities. But recall that the even loops of X containing the points ai divide the real projective
plane into two regions, hence if Y enters the interior of such a loop it has to exit it again at another
point ai of the same loop as in the picture above (it may also happen that Y is tangent to X at ai , in
which case their intersection multiplicity is at least 2 there by Exercise 12.27). So in any case the
total number of intersection points, counted with their respective multiplicities, is at least

d 1
2
+ 1 + (d 3) = d(d 2) + 1 > d(d 2),
2
which is a contradiction.

Let us now turn to the case of complex curves. Of course, their topology is entirely different, as they
are 2-dimensional spaces in the classical topology. In fact, we have seen such an example already in
Example 0.1 of the introduction.
Remark 13.13 (Classical topology of complex curves). Let X P2C be a smooth curve. Then X is
a compact 2-dimensional manifold in the classical topology (see Remark 10.14). Moreover, one can
show:
(a) X is always an oriented manifold in the classical topology, i. e. a two-sided surface, as
opposed to e. g. a Mbius strip. To see this, note that all tangent spaces Ta X of X for a X
are isomorphic to C, and hence admit a well-defined multiplication with the imaginary unit
i. Geometrically, this means that all tangent planes have a well-defined notion of a clockwise
rotation by 90 degrees, varying continuously with a which defines an orientation of X. In
fact, this statement holds for all smooth complex curves, not just for curves in P2C .
(b) In contrast to the real case that we have just studied, X is always connected. In short, the
reason for this is that the notion of degree as well as Bzouts Theorem can be extended
to compact oriented 2-dimensional submanifolds of P2C . Hence, if X had (at least) two
connected components X1 and X2 in the classical topology, each of these components would
be a compact oriented 2-dimensional manifold itself, and there would thus be well-defined
degrees deg X1 , deg X2 N>0 . But then X1 and X2 would have to intersect in deg X1 deg X2
points (counted with multiplicities), which is obviously a contradiction.
Of course, the methods needed to prove Bzouts Theorem in the topological setting are
entirely different from ours in Chapter 12. If you know some algebraic topology, the statement here is that the 2-dimensional homology group H2 (P2C , Z) is isomorphic to Z. With
this isomorphism, the class of a compact oriented 2-dimensional submanifold in H2 (P2C , Z)
110
Andreas Gathmann
is a positive number, and the intersection product H2 (P2C , Z) H2 (P2C , Z) H0 (P2C , Z)
=Z
(using Poincar duality) is just the product of these numbers.
22
It is now a (non-trivial but intuitive) topological result that a connected compact orientable 2-dimensional manifold X is always
homeomorphic to a sphere with some finite number of handles.
This number of handles is called the genus of X. Hence every curve
in P2C can be assigned a genus that describes its topological type in
the classical topology. The picture on the right shows a complex
curve of genus 2.
Our goal for the rest of this chapter will be to compute the genus of a smooth curve in P2C in terms of
its degree, as already announced in Example 0.3. To do this, we will need the following technique
from topology.
Construction 13.14 (Cell decompositions). Let X be a compact 2-dimensional manifold. A cell
decomposition of X is given by writing X topologically as a finite disjoint union of points, (open)
lines, and (open) discs. This decomposition should be nice in a certain sense, e. g. the boundary
points of every line in the decomposition must be points of the decomposition. We do not want to
give a precise definition here (which would necessarily be technical), but only remark that every
reasonable decomposition that one could think of will be allowed. For example, the following
picture shows three valid decompositions of the complex curve P1C , which is topologically a sphere.
In the left two pictures, we have 1 point, 1 line, and 2 discs (the two halves of the sphere), whereas
in the picture on the right we have 2 points, 4 lines, and 4 discs.
Of course, there are many possibilities for cell decompositions of X. But there is an important
number that does not depend on the chosen decomposition:
Lemma and Definition 13.15 (Euler characteristic). Let X be a compact 2-dimensional manifold.
Consider a cell decomposition of X, consisting of 0 points, 1 lines, and 2 discs. Then the number
:= 0 1 + 2
depends only on X, and not on the chosen decomposition. We call it the (topological) Euler characteristic of X.
Sketch proof. Let us first consider the case when we move from one decomposition to a finer one,
i. e. if we add points or lines to the decomposition. Such a process is always obtained by performing
the following steps a finite number of times:
Adding another point on a line: in this case we raise 0 and 1 by 1 as in the picture below,
hence the alternating sum 0 1 + 2 does not change.
Adding another line in a disc: in this case we raise 1 and 2 by 1, so again 0 1 + 2
remains invariant.
add a
point
add a
line
13.
111
We conclude that the alternating sum 0 1 + 2 does not change under refinements. But any two
decompositions have a common refinement which is essentially given by taking all the points
and lines in both decompositions, and maybe adding more points where two such lines intersect. For
example, in Construction 13.14 the decomposition in the picture on the right is a common refinement
of the other two. Hence the Euler characteristic is independent of the chosen decomposition.

Example 13.16 (Euler characteristic genus). Let X be
a connected compact orientable 2-dimensional manifold of
genus g, and consider the cell decomposition of X as shown
on the right. It has 0 = 2g + 2 points, 1 = 4g + 4 lines, and
4 discs, and hence we conclude that the Euler characteristic
of X is
= 0 1 + 2 = 2 2g.
In other words, the genus is given in terms of the Euler characteristic as g = 1 2 .
We are now ready to compute the genus of a smooth curve in P2C .
Proposition 13.17 (Degree-genus formula). A smooth curve of degree d in P2C has genus
d1
2 .
Sketch proof. Let I(X) = ( f ) for a homogeneous polynomial f of degree d. By a linear change of
coordinates we can assume that a := (1 : 0 : 0)
/ X. Then the projection
: X P1C , (x0 : x1 : x2 ) 7 (x1 : x2 )
from a as in the picture on the right is a well-defined morphism on X. Let us study its inverse images of a fixed point
(x1 : x2 ) P1C . Of course, they are given by the values of x0 such
that f (x0 , x1 , x2 ) = 0, so that there are exactly d such points
unless the polynomial f ( , x1 , x2 ) has a multiple zero in x0 at a
point in the inverse image, which happens if and only if f and
f
x are simultaneously zero.
0
a
b
X
c
(b)
(c)
P1C
If we choose our original linear change of coordinates general enough, exactly two of the zeroes of
f ( , x1 , x2 ) will coincide at these points in the common zero locus of f and xf , so that 1 (x1 : x2 )
0
then consists of d 1 instead of d points. These points, as e. g. b and c in the picture above, are
usually called the ramification points of . Note that the picture might be a bit misleading since it
suggests that X is singular at b and c, which is not the case. The correct topological picture of the
map is impossible to draw however since it would require a real 4-dimensional space, namely an
affine chart of P2C .
Let us now pick a sufficiently fine cell decomposition of P1C , containing all images of the ramification points as points of the decomposition. If 0 , 1 , 2 denote the number of points, lines, and
discs in this decomposition, respectively, we have 0 1 + 2 = 2 by Example 13.16 since P1C is
topologically a sphere, i. e. of genus 0. Now lift this cell decomposition to a decomposition of X
by taking all inverse images of the cells of P1C . By our above argument, all cells will have exactly
d inverse images except for the images of the ramification points, which have one inverse image
less. As the number of ramification points is |Vp ( f , xf )| = deg f deg xf = d(d 1) by Bzouts
0
0
Theorem, the resulting decomposition of X has d0 d(d 1) points, d1 lines, and d2 discs.
Hence by Lemma 13.15 the Euler characteristic of X is
= d0 d(d 1) d1 + d2 = 2d d(d 1) = 3d d 2 ,
which means by Example 13.16 that its genus is
g = 1

1
d 1
= (d 2 3d + 2) =
.
2
2
2
112
Andreas Gathmann
Example 13.18.
(a) A smooth curve of degree 1 or 2 in P2C is isomorphic to P1C (see Example 7.6 (d)). It is
therefore topologically a sphere, i. e. of genus 0, in accordance with Proposition 13.17.
(b) By Proposition 13.17, a smooth curve of degree 3 in P2C has genus 1, i. e. it is topologically
a torus. We will study such plane cubic curves in detail in Chapter 15.
Remark 13.19. Note that every isomorphism of complex varieties is also a homeomorphism in the
classical topology. In particular, two smooth connected projective curves over C of different genus
cannot be isomorphic. Combining this with Proposition 13.17, we see that two smooth curves in P2C
of different degree are never isomorphic, unless these degrees are 1 and 2.
Exercise 13.20 (Arithmetic genus). For a projective variety X the number (1)dim X (X (0) 1) is
called its arithmetic genus, where X denotes as usual the Hilbert polynomial of X. Show that the
arithmetic genus of a smooth curve in P2C agrees with the (geometric) genus introduced above.
In fact, one can show that this is true for any smooth projective curve over C. However, the proof
of this statement is hard and goes well beyond the scope of these notes. Note that, as the definition
of the arithmetic genus is completely algebraic, one can use it to extend the notion of genus to
projective curves over arbitrary ground fields.
Exercise 13.21. Show that
{((x0 : x1 ), (y0 : y1 )) : (x02 + x12 )(y20 + y21 ) = x0 x1 y0 y1 }
is a smooth curve of genus 1.
P1C P1C
14.
14.
Divisors on Curves
113
Divisors on Curves
In its version for curves, Bzouts Theorem determines the number of zeroes of a homogeneous
polynomial on a projective curve (see Corollaries 12.20 and 12.26). For example, if X P2 is a
cubic curve then the zero locus of a homogeneous linear polynomial f on X consists of three points,
counted with multiplicities. But of course not every collection of three points on X can arise in this
way, as three points will in general not lie on a line, and thus cannot be in the zero locus of f . So
by reducing the question of the zeroes of polynomials to just their number we are losing information
about their possible positions. To avoid this, we will now present a theory that is able to keep track
of the actual configurations of points on curves.
It turns out that these configurations, called divisors below, are parametrized by a group that is
naturally associated to the curve X. This will allow us to study and classify curves with methods
from group theory, very much in the same way as in topology the fundamental group or the homology
groups can be used to study and distinguish topological spaces. For example, using divisors we will
be able to prove in Proposition 14.19 and Remark 14.20 that a smooth plane cubic curve as above is
never isomorphic to P1 . Note that for the ground field C we have already seen this topologically in
Remark 13.19 since a smooth plane cubic is a torus whereas P1C is a sphere but of course this was
using techniques from topology that would certainly require some work to make them rigorous. In
contrast, our new proof here will be entirely self-contained and algebraic, so in particular applicable
to any ground field.
The concept of divisors can be defined for arbitrary curves. For example, in the smooth affine case
this leads to the notion of Dedekind domains studied in commutative algebra [G5, Chapter 13], and
for singular curves one needs two different concepts of divisors, called Weil divisors and Cartier
divisors. However, in our applications we will only need irreducible smooth projective curves. So
for simplicity of notation we will restrict ourselves to this case from the very beginning, even if
many of our constructions and results do not need all these assumptions.
Let us start by giving the definition of divisors. It should be noted that the name divisor in this
context has historical reasons; it is completely unrelated to the notion of divisors in an integral
domain.
Definition 14.1 (Divisors). Let X be an irreducible smooth projective curve.
(a) A divisor on X is a formal finite linear combination k1 a1 + + kn an of distinct points
a1 , . . . , an X with integer coefficients k1 , . . . , kn Z for some n N. Obviously, the divisors
on X form an Abelian group under pointwise addition of the coefficients. We will denote it
by Div X.
Equivalently, in algebraic terms Div X is just the free Abelian group generated by the points
of X (i. e. the group of maps X Z being non-zero at only finitely many points; with a point
mapping to its coefficient in the sense above).
(b) A divisor D = k1 a1 + + kn an as above is called effective, written D 0, if ki 0 for all
i = 1, . . . , n. If D1 , D2 are two divisors with D2 D1 effective, we also write this as D2 D1
or D1 D2 . In other words, we have D2 D1 if and only if the coefficient of any point in
D2 is greater than or equal to the coefficient of this point in D1 .
(c) The degree of a divisor D = k1 a1 + + kn an as above is the number deg D := k1 + + kn
Z. Obviously, the degree is a group homomorphism deg : Div X Z. Its kernel is denoted
by
Div0 X = {D Div X : deg D = 0}.
114
Andreas Gathmann
Construction 14.2 (Divisors from polynomials and intersections). Again, let X Pn be an irreducible smooth curve. Our construction of multiplicities in Definition 12.23 (b) allows us to define
divisors on X as follows.
(a) For a non-zero homogeneous polynomial f S(X) the divisor of f is defined to be
div f :=
multa ( f ) a
Div X,
aVX ( f )
where VX ( f ) denotes the zero locus of f on X as in Construction 6.18. In other words,

the divisor div f contains the data of the zeroes of f together with their multiplicities. By
Bzouts Theorem as in Corollary 12.26 (a), its degree is deg(div f ) = deg X deg f .
(b) If n = 2 and Y P2 is another curve not containing X, the intersection divisor of X and Y is
X Y :=
multa (X,Y ) a
Div X.
aXY
By definition, this divisor is just the same as div f for a generator f of I(Y ). Note that it is
symmetric in X and Y , so in particular the result can be considered as an element of DivY as
well if Y is also smooth and irreducible. By Bzouts Theorem as in Corollary 12.26 (b), we
have deg(X Y ) = deg X degY .
Example 14.3. Consider again the two projective curves X = V (x0 x2 x12 ) and Y = V (x2 ) in P2 of
Example 12.28. We have seen in this example that X and Y intersect in a single point a = (1 : 0 : 0)
with multiplicity 2. Hence X Y = 2a in Div X in the notation of Construction 14.2. Equivalently,
we can write div x2 = 2a on X, and div(x0 x2 x12 ) = 2a on Y .
Note that so far all our multiplicities have been non-negative, and hence all the divisors in Construction 14.2 are effective. Let us now extend this construction to multiplicities and divisors of rational
functions, which will lead to negative multiplicities at their poles, and thus to non-effective divisors.
To do this, we need the following lemma first.
Lemma 14.4. Let X be an irreducible smooth projective curve, and let f , g S(X) be two non-zero
polynomials. Then
multa ( f g) = multa ( f ) + multa (g)
for all a X. In particular, we have div( f g) = div f + div g in Div X.
Proof. By Remark 12.25 (a) we have to show that
dimK OX,a /( f g) = dimK OX,a /( f ) + dimK OX,a /(g).
for all a X. But this follows immediately by Lemma 12.5 from the exact sequence
g
0 OX,a /( f ) OX,a /( f g) OX,a /(g) 0

(for the injectivity of the first map note that OX,a is an integral domain since X is irreducible). Taking
these results for all a X together, we conclude that div( f g) = div f + div g.

Construction 14.5 (Multiplicities and divisors of rational functions). Let X be an irreducible smooth
projective curve, and let K(X) be a non-zero rational function (see Construction 9.6). By
Definition 7.1 we can write = gf for two homogeneous polynomials f and g of the same degree.
(a) We define the multiplicity of at a point a X to be
multa () := multa (g) multa ( f )
g0
f0
Z.
g
f
Note that this is well-defined: if = for two other homogeneous polynomials f 0 and g0 of
the same degree then g0 f f 0 g = 0 on a non-empty open subset, hence on all of X since X is
irreducible, and consequently g0 f = f 0 g S(X). Lemma 14.4 thus implies that multa (g0 ) +
multa ( f ) = multa ( f 0 ) + multa (g), i. e. that multa (g0 ) multa ( f 0 ) = multa (g) multa ( f ).
Geometrically, we can think of this multiplicity as the order of the zero (if multa () > 0)
resp. pole (if multa () < 0) of at a.
14.
Divisors on Curves
115
(b) Analogously to Construction 14.2, we define the divisor of to be

div :=
multa () a = div g div f .
aVX ( f )VX (g)
23
Example 14.6. The rational function =
x0 x1
(x0 x1 )2
on P1 has divisor
div = (1 : 0) + (0 : 1) 2 (1 : 1).
Remark 14.7 (Multiplicities of local functions). By Exercise 9.8 (b) every local function OX,a
at a point a of an irreducible smooth projective curve X can be considered as a rational function on
X. Hence Construction 14.5 also defines a multiplicity multa () for any non-zero OX,a .
Moreover, note that then has a representation of the form = gf with f , g S(X) and f (a) 6= 0.
By Remark 12.24, this means that multa ( f ) = 0, and thus multa () = multa (g) N. By the same
remark, we have multa () = 0 if and only if g(a) 6= 0 as well, i. e. if and only if (a) 6= 0.
Remark 14.8. As above, let X be an irreducible smooth projective curve.
(a) Lemma 14.4 implies that multa (1 2 ) = multa 1 + multa 2 for any a X and any two nonzero rational functions 1 , 2 K(X) . We therefore also have div(1 2 ) = div 1 + div 2 ,
i. e. the map div : K(X) Div X is a homomorphism of groups.
(b) As any non-zero rational function on X has the form = gf for two homogeneous polynomials of the same degree, we see by Construction 14.2 (a) that its divisor always has degree
0:
deg div = deg(div g div f ) = deg div g deg div f = deg X deg g deg X deg f = 0.
Hence the group homomorphism of (a) can also be viewed as a homomorphism div :
K(X) Div0 X.
This observation leads to the idea that we should give special attention to the divisors of rational
functions, i. e. to the image subgroup of the above divisor homomorphism.
Definition 14.9 (Principal divisors and the Picard group). Let X be an irreducible smooth projective
curve.
(a) A divisor on X is called principal if it is the divisor of a (non-zero) rational function. We
denote the set of all principal divisors by Prin X. Note that Prin X is just the image of the
divisor homomorphism div : K(X) Div0 X of Remark 14.8 (b), and hence a subgroup of
both Div0 X and Div X.
(b) The quotient
Pic X := Div X/ Prin X
is called the Picard group or group of divisor classes on X. Restricting to degree zero, we
also define Pic0 X := Div0 X/ Prin X. By abuse of notation, a divisor and its class in Pic X
will usually be denoted by the same symbol.
Remark 14.10. The groups Pic X and Pic0 X carry essentially the same information on X, since we
always have
Pic X/ Pic0 X
= Div X/ Div0 X
= Z.
It just depends on the specific application in mind whether it is more convenient to work with Pic X
or with Pic0 X.
By construction, the group Div X of divisors on an irreducible smooth projective curve X is a free
Abelian group with an infinite number of generators, and hence not very interesting from a grouptheoretic point of view. In contrast, the Picard group is rather small and has quite a rich structure
that we want to study now in some examples.
116
Andreas Gathmann
Example 14.11 (The Picard group of P1 is trivial). On P1 , every degree-0 divisor is principal: if
D = k1 (a1,0 : a1,1 ) + + kn (an,0 : an,1 ) for some points (ai,0 : ai,1 ) P1 and integers ki for i = 1, . . . , n
with k1 + + kn = 0, the rational function given by
n
(x0 : x1 ) = (ai,1 x0 ai,0 x1 )ki

i=1
has divisor div = D. Hence the divisor map div : K(P1 ) Div0 P1 is surjective, so that we have
Prin P1 = Div0 P1 , and consequently
Pic0 P1 = {0} and Pic P1 = Div P1 / Div0 P1
= Z,
with the isomorphism deg : Pic P1 Z.
Let us now move on to more complicated curves. We know already from Example 7.6 (d) that a
smooth conic X P2 (which is irreducible by Exercise 10.22 (a)) is isomorphic to P1 . As the Picard
group is clearly invariant under isomorphisms, this means that Pic0 X will then be the trivial group
again. So the next case to consider is a smooth cubic curve X P2 . Our main goal in this chapter
is to prove that Pic0 X is not trivial in this case, so that X cannot be isomorphic to P1 . In fact, in the
next chapter in Proposition 15.2 we will even be able to compute Pic0 X for a plane cubic explicitly.
However, even for the special case of plane cubics the computation of Pic0 X is not easy, and so
we will need some preliminaries first. The following lemma will be well-known to you if you have
attended the Commutative Algebra class already, since it essentially states in algebraic terms that the
local rings of X are discrete valuation rings [G5, Lemma 12.1]. It is the first time in this chapter that
the smoothness assumption on X is essential.
Lemma 14.12 (Local coordinates on a smooth plane curve). Let X P2 be a smooth curve, and let
Ia E OX,a be the maximal ideal in the local ring of a point a X as in Definition 3.22.
(a) The ideal Ia is principal, with Ia = (a ) for a suitable a OX,a with multa (a ) = 1.
(b) Any non-zero OX,a can be written as = c am , where c OX,a \Ia and m = multa ().
Proof.
(a) Choose a linear function a vanishing at a such that the line V (a ) is not the tangent line to
X at a. Then a vanishes on X with multiplicity 1 at a by Exercise 12.27. Hence a Ia ,
and
1 = dimK OX,a /(a ) dimK OX,a /Ia > 0.
It follows that we must have equality, so in particular that Ia = (a ).
(b) Note that
/ Iam+1 , since by (a) the elements of Iam+1 are multiples of am+1 , and thus have
multiplicity at least m + 1 at a. Hence there is a maximal n N with Ian . By (a) this
means = c an for some c OX,a . But we must have c
/ Ia since n is maximal. Hence
m = multa () = multa (c an ) = n, and the result follows.

Remark 14.13. Thinking of a smooth curve X P2C as a 1-dimensional complex manifold, we can
interpret the function a of Lemma 14.12 as a local coordinate on X at a, i. e. as a function that
gives an isomorphism of a neighborhood of a with a neighborhood of the origin in C in the classical
topology. By standard complex analysis, any local holomorphic function on X at a can then be
written as a non-vanishing holomorphic function times a power of the local coordinate [G4, Lemma
10.4]. Lemma 14.12 (b) is just the corresponding algebraic statement. Note however that this is only
a statement about the local ring in contrast to the analytic setting it does not imply that X has a
Zariski-open neighborhood of a isomorphic to an open subset of A1C !
Remark 14.14 (Infinite multiplicity). Let X P2 be a smooth curve. For the following lemma, it
is convenient to set formally multa (X, f ) = for all a X if f is a homogeneous polynomial that
vanishes identically on X. Note that by Lemma 14.12 (b) we then have for an arbitrary homogeneous
polynomial f that multa (X, f ) m if and only if f is a multiple of am in OX,a , where we interpret f
as an element in OX,a as in Remark 12.25, and a is a local coordinate as in Lemma 14.12.
14.
Divisors on Curves
117
Lemma 14.15. Let X P2 be a smooth curve, and let a X be a point.

(a) Let f , g K[x0 , x1 , x2 ] be homogeneous polynomials of the same degree with multa (X, f ) m
and multa (X, g) m for some m N. Then:
multa (X, f + g) m for all , K;
there are , K, not both zero, such that multa (X, f + g) m + 1.
(b) Let Y P2 be another curve, and set m = multa (X,Y ). If f K[x0 , x1 , x2 ] is a homogeneous
polynomial with multa (X, f ) m, then we also have multa (Y, f ) m.
Proof. As in Remark 12.25, we will consider f and g as elements in the local ring OX,a .
(a) We may assume that f and g do not vanish identically on X, since otherwise the statement is
trivial. By Remark 14.14 there are then u, v OX,a such that f = u am and g = v am in OX,a .
So for any , K, we have f + g = ( u + v) am , and thus multa ( u + v) m.
Moreover, we can pick and not both zero such that u(a) + v(a) = 0 K. Then
multa ( u + v) 1, and hence multa ( f + g) m + 1.
(b) As above we can assume that f does not vanish identically on X or Y . Let g and h be
polynomials such that I(X) = (g) and I(Y ) = (h). The assumption then means that k :=
multa (X, f ) m = multa (X, h). Hence f = u ak and h = v am for suitable units u, v OX,a
by Lemma 14.12 (b). This implies that ( f ) (h) in OX,a , so that ( f , g) (g, h) in OA2 ,a . But
then ( f , h) (g, h) in OA2 ,a as well, hence multa ( f , h) multa (g, h), and thus multa (Y, f )
multa (X,Y ) = m.

Example 14.16. The following examples show that the smoothness assumption in Lemma 14.15 (b)
is crucial.
(a) Let X and Y be two smooth plane curves such that multa (X,Y ) = 2. By Exercise 12.27 this
means that X and Y are tangent at a, as in the picture below on the left. Lemma 14.15 (b)
then states that any polynomial vanishing on X to order at least 2 at a also vanishes on Y to
order at least 2 at a. In terms of the corresponding curve V ( f ), this means using Exercise
12.27 again that any other curve that is (singular or) tangent to X at a is also (singular or)
tangent to Y at a which is obvious.
(b) In contrast to (a), in the picture on the right we have X = Vp (x23 x12 x0 ), Y = Vp (x2 ), f = x1 ,
and thus at the origin m = multa (X,Y ) = 2, multa (X, f ) = 3, but multa (Y, f ) = 1. This fits
well with the geometric interpretation of Remark 14.13: the curve X is singular at the origin,
so locally not a 1-dimensional complex manifold. Hence there is no local coordinate on X
around a, and the argument of Lemma 14.12 resp. Lemma 14.15 breaks down.
V(f)
X
a
V(f)
Y
(a)
a
(b)
Lemma 14.17. Let X P2 be a smooth curve, and let g, h S(X) be two non-zero homogeneous
polynomials.
(a) If div g = div h then g and h are linearly dependent in S(X).
(b) If h is linear and div g div h then h | g in S(X).
Proof. Let f K[x0 , x1 , x2 ] be a homogeneous polynomial with I(X) = ( f ).
118
Andreas Gathmann
(a) By assumption we have ma := multa (g) = multa (h) for all a X. Moreover, Bzouts Theorem as in Corollary 12.26 (a) implies that aX ma = deg X deg g = deg X deg h, so in
particular we see already that d := deg g = deg h.
Now pick an arbitrary point b X. By Lemma 14.15 (a) there are , K, not both zero,
such that multa ( g + h) ma for all a X, and multb ( g + h) mb + 1. Summing up,
this means that aX multa ( g + h) d deg X + 1. But g + h also has degree d, hence
by Bzout again it follows that f + g must vanish identically on X, i. e. g + h = 0
S(X).
(b) Let L = V (h), and choose a representative g K[x0 , x1 , x2 ] of g S(X). We may assume that
g does not vanish identically on L, as otherwise h | g in K[x0 , x1 , x2 ], and we are done.
24
By assumption, we have multa (X, g)

multa (X, L) for all a X L. As X is smooth, Lemma
14.15 (b) shows that then multa (L, g)
multa (X, L), and thus that div g div f on L. As
L
= P1 , we can find a homogeneous polynomial f 0 K[x0 , x1 , x2 ] of degree deg g deg f
with div f 0 = div g div f on L as in Example 14.11. Then div( f f 0 ) = div g on L, which
means by (a) that f f 0 and g are linearly dependent in S(L). But then g = f f 0 + ph for some
homogeneous polynomial p and after possibly multiplying f 0 with a non-zero scalar, which
means that g = ph in K[x0 , x1 , x2 ]/( f ) = S(X).

We are now finally ready to prove that the Picard group of a smooth cubic curve in P2 is not trivial.
Notation 14.18. Let a and b be two points on a smooth cubic curve X
P2 , not necessarily distinct. By Exercise 12.27 there is then a unique line
L P2 such that a + b L X as divisors on X (in the sense of Definition
14.1 (b)), namely the line through a and b if these points are distinct, and
the tangent line to X at a = b otherwise. But L X is an effective divisor
of degree 3 on X, and hence there is a unique point c X (which need not
be distinct from a and b) with L X = a + b + c. In the following, we will
denote this point c by (a, b).
X
b
a
a0
(a, b)
(a0 , a0 )
In geometric terms, for general a, b X (i. e. such that none of the above points coincide) the point
(a, b) is just the third point of intersection of X with the line through a and b. Hence the above
definition is a generalization of our construction in Exercise 7.15. In fact, one can show that the map
: X X X, (a, b) 7 (a, b) is a morphism, but we will not need this result here.
Proposition 14.19. Let X P2 be a smooth cubic curve. Then for all distinct a, b X we have
a b 6= 0 in Pic0 X, i. e. there is no non-zero rational function on X with div = a b.
Proof. Assume for a contradiction that the statement of the proposition is false. Then there are a
positive integer d and homogeneous polynomials f , g S(X) of degree d such that the following
conditions hold:
(a) There are points a1 , . . . , a3d1 and a 6= b on X such that
div g = a1 + + a3d1 + a
(hence div = a b for =
and
div f = a1 + + a3d1 + b
g
f ).
(b) Among the a1 , . . . , a3d1 there are at least 2d distinct points. (If this is not the case in the first
place, we can replace f and g by f l and g l, respectively, for some homogeneous linear
polynomial l that vanishes on X at three distinct points that are not among the ai . This raises
the degree of the polynomials by 1 and the number of distinct points by 3, so by doing this
often enough we can get at least 2d distinct points.)
Pick d minimal with these two properties.
If d = 1 then div g = a1 + a2 + a and div f = a1 + a2 + b, so we must have a = b = (a1 , a2 ) by
Notation 14.18, in contradiction to our assumption. Hence we can assume that d > 1. Let us relabel
14.
Divisors on Curves
119
the points a1 , . . . , a3d1 such that a2 6= a3 , and such that a1 = a2 if there are any equal points among
the ai .
Now consider linear combinations f + g for , K, not both zero. As the polynomials f and
g have different divisors they are linearly independent in S(X), and hence f + g does not vanish
identically on X. Moreover, by Lemma 14.15 (a) we have a1 + + a3d1 div( f + g) for all
and , and for any given c X there are and with a1 + + a3d1 + c div( f + g). Of
course, by Bzouts Theorem we must then have div( f + g) = a1 + + a3d1 + c.
In other words, by passing to linear combinations of f and g we can assume that the last points
a and b in the divisors of f and g are any two points we like. Let us choose a = (a1 , a2 ) and
b = (a1 , a3 ). Then
div g = (a1 + a2 + (a1 , a2 )) + a3 + a4 + + a3d1
and
div f = (a1 + a3 + (a1 , a3 )) + a2 + a4 + + a3d1 .
But a1 + a2 + (a1 , a2 ) and a1 + a3 + (a1 , a3 ) are divisors of homogeneous linear polynomials

k and l in S(X), respectively, and hence by Lemma 14.17 (b) there are homogeneous polynomials
f 0 , g0 S(X) of degree d 1 with g = kg0 and f = l f 0 , and thus with
div g0 = a4 + + a3d1 + a3
and
div f 0 = a4 + + a3d1 + a2 .
Note that these new polynomials f 0 and g0 satisfy (a) for d replaced by d 1, as a3 6= a2 by assumption. Moreover, f 0 and g0 satisfy (b) because, if there are any equal points among the ai at all, then by
our relabeling of these points there are only two distinct points among a1 , a2 , a3 , and so there must
still be at least 2d 2 distinct points among a4 , . . . , a3d1 .
This contradicts the minimality of d, and therefore proves the proposition.
Remark 14.20. In particular, Proposition 14.19 implies that Pic0 X 6= {0} for any smooth cubic
surface X P2 . So by Example 14.11 we can already see that X is not isomorphic to P1 . In fact, we
will see in the next chapter that Proposition 14.19 suffices to compute Pic0 X explicitly.
120
15.
Andreas Gathmann
Elliptic Curves
At the end of the last chapter we have used Picard groups to show in Proposition 14.19 and Remark
14.20 that smooth cubic curves in P2 are not isomorphic to P1 . In fact, if our ground field is not
necessarily C (so that we cannot apply topological methods as in Remark 13.19), this is the first
class of smooth projective curves for which we could prove rigorously that they are not isomorphic
to P1 . So let us now study these curves in more detail. We will see that they have a very rich
structure, both from an algebraic and over C from an analytic point of view.
Definition 15.1 (Elliptic curves). In this chapter, by an elliptic curve we will simply mean a smooth
cubic curve in P2 .
Usually in the literature, an elliptic curve is defined to be a smooth complete curve of genus 1 (see
Remark 13.13 and Exercise 13.20 for the definition of genus). Note that a smooth cubic curve in P2
is in fact complete by Example 7.22 (b) and of genus 1 by Example 13.18 (b) and Exercise 13.20.
Conversely, one can show that every smooth complete curve of genus 1 can be embedded as a cubic
curve in P2 . Hence our somewhat non-standard definition of an elliptic curve is consistent with the
literature.
The term elliptic curve might sound confusing at first, because the shape of a plane cubic curve has
no similarities with an ellipse, not even over the real numbers (see e. g. Remark 13.9). The historical
reason for this name is that the formula for the circumference of an ellipse can be expressed in terms
of an integral over a plane cubic curve.
Probably the single most important result about elliptic curves is that they carry a natural group
structure. The easiest, or at least the most conceptual way to prove this is by computing the degree-0
Picard group of an elliptic curve X, which (after the choice of a base point) turns out to be in natural
bijection with X itself.
Proposition 15.2. Let X P2 be an elliptic curve, and let a0 X be a point. Then the map
: X Pic0 X,
a 7 a a0
is a bijection.
Proof. As deg(a a0 ) = 0, the map is clearly well-defined. It is also injective: if (a) = (b)
for a, b X this means that a a0 = b a0 , and hence a b = 0, in Pic0 X. By Proposition 14.19
this is only possible if a = b.
To show that is surjective, let D be an arbitrary element of Pic0 X, which we can write as
D = a1 + + am b1 bm
for some m N>0 and not necessarily distinct a1 , . . . , am , b1 , . . . , bm X. Assume first that m 2.
Then there are homogeneous linear polynomials l, l 0 on X such that div l = a1 + a2 + (a1 , a2 ) and
div l 0 = b1 + b2 + (b1 , b2 ), where is as in Notation 14.18. The quotient of these polynomials is
then a rational function on X, whose divisor a1 + a2 + (a1 , a2 ) b1 b2 (b1 , b2 ) is therefore
zero in Pic0 X. It follows that we can also write
D = (b1 , b2 ) + a3 + + am (a1 , a2 ) b3 bm
Pic0 X.
We have thus reduced the number m of (positive and negative) points in D by 1. Continuing this
process, we can assume that m = 1, i. e. that D = a1 b1 for some a1 , b1 X.
In the same way, we then also have
a0 + a1 + (a0 , a1 ) b1 (a0 , a1 ) (b1 , (a0 , a1 )) = 0
Pic0 X,
15.
Elliptic Curves
121
so that D = a1 b1 = (b1 , (a0 , a1 )) a0 Pic0 X. Hence D = ((b1 , (a0 , a1 ))), i. e. is

surjective.

Remark 15.3. Let X P2 be an elliptic curve. After choosing a base point a0 X, Proposition
15.2 gives a canonical bijection between the variety X and the Abelian group Pic0 X, i. e. between
two totally different mathematical objects. So we can use this bijection to give X the structure of an
Abelian group, and Pic0 X the structure of a smooth projective variety.
In fact, Pic0 X can be made into a variety (the so-called Picard variety) for every smooth projective
curve X. It is in general not isomorphic to X, however. One can only show that the map : X
Pic0 X, a 7 a a0 of Proposition 15.2 is injective if X is not P1 , so that we can then think of X as a
subvariety of the Picard variety.
In contrast, the statement that X can be made into an Abelian group is very special to elliptic curves.
In the following, we want to explore this group structure in more detail.
Construction 15.4 (The group structure on an elliptic curve). Let a0 be a fixed base point on an
elliptic curve X P2 . As in Remark 15.3, we can use Proposition 15.2 to define a group structure
on X. More precisely, if we denote this group operation by the symbol (to distinguish it from the
addition of points in Div X or Pic X), then a b for a, b X is the unique point of X satisfying
(a b) = (a) + (b).
To find an explicit description for a b, note that as in the proof of Proposition 15.2 both
a + b + (a, b) and a0 + (a, b) + (a0 , (a, b)) are divisors of homogeneous linear polynomials,
and thus
a + b + (a, b) a0 (a, b) (a0 , (a, b)) = 0
Pic0 X.
Hence
a b = 1 ((a) + (b))
= 1 (a a0 + b a0 )
= 1 ((a0 , (a, b)) a0 )
= (a0 , (a, b)).
In other words, to construct the point a b we draw a line through a and b. Then we draw another
line through the third intersection point (a, b) of this line with X and the point a0 . The third
intersection point of this second line with X is then a b, as in the picture below on the left.
Similarly, to construct the inverse a of a in the above group structure we use the relation
a0 + a0 + (a0 , a0 ) a (a0 , a0 ) (a, (a0 , a0 )) = 0
Pic0 X
to obtain
a = 1 ((a))
= 1 (a0 a)
= 1 ((a, (a0 , a0 )) a0 )
= (a, (a0 , a0 )).
So to construct the inverse a we draw the tangent to X through a0 . Then we draw another line
through the other intersection point (a0 , a0 ) of this tangent with C and the point a. The third
intersection point of this second line with C is a, as in the following picture.
122
Andreas Gathmann
X
a0
a
a0
ab
Note that, using this geometric description, the operation could also be defined in a completely
elementary way, without referring to the theory of divisors. However, it would then be very difficult
to show that we obtain a group structure in this way, in particular to prove associativity.
Exercise 15.5. Let X and Y be two distinct elliptic curves in P2 , and assume that they intersect in 9
distinct points a1 , . . . , a9 . Prove that every elliptic curve passing through a1 , . . . , a8 also has to pass
through a9 .
Can you find a stronger version of this statement that applies in the case when the intersection
multiplicities in X Y are not all equal to 1?
Example 15.6 (Elliptic Curve Cryptography). There is an interesting application of the group structure on an elliptic curve to cryptography. The key observation is that multiplication is easy, but
division is hard. More precisely, assume that we are given a specific elliptic curve X and a base
point a0 X for the group structure.
(a) Given a X and n N, the n-fold addition n a := a a can be computed very quickly,
even for very large n (think of numbers with hundreds of digits):
By repeatedly applying the operation a 7 a a, we can compute all points 2k a for
all k such that 2k n.
Now we just have to add these points 2k a for all k such that the k-th digit in the
binary representation of n is 1.
This computes the point n a in a time proportional to log n (i. e. in a very short time).
(b) On the other hand, given two sufficiently general points a, b X it is essentially impossible
to compute an integer n N such that n a = b (in case such a number exists). Note that
this is not a mathematically precise statement there is just no known algorithm that can
perform the inverse of the multiplication of (a) in shorter time than a simple trial-and-error
approach (which would be impractical for large n).
Let us now assume that Alice and Bob want to establish an encrypted communication over an insecure channel, but that they have not met in person before, so that they could not secretly agree on
a key for the encryption. Using the above idea, they can then agree (publicly) on a ground field K,
a specific elliptic curve X over K, a base point a0 X, and another point a X. Now Alice picks
a secret (very large) integer n, computes n a as in (a), and sends (the coordinates of) this point
to Bob. In the same way, Bob chooses a secret number m, computes m a, and sends this point to
Alice.
As Alice knows her secret number n and the point m a from Bob, she can then compute the point
mn a = n (m a). In the same way, Bob can compute this point as mn a = m (n a) as well.
But except for the data of the chosen curve the only information they have exchanged publicly was
a, n a, and m a, and by (b) it is not possible in practice to recover n or m, and hence mn a,
from these data. Hence Alice and Bob can use (the coordinates of) mn a as a secret key for their
encrypted communication.
Exercise 15.7. Let X be an elliptic curve of the form
X = {(x0 : x1 : x2 ) : x22 x0 = x13 + x1 x02 + x03 }
P2
for some given , K (it can be shown that every elliptic curve can be brought into this form
by a change of coordinates if the characteristic of K is not 2 or 3). Pick the point a0 = (0 : 0 : 1) as
15.
Elliptic Curves
123
the base point for the group structure on X. For given points b = (b0 : b1 : b2 ) and c = (c0 : c1 : c2 )
compute explicitly the coordinates of the inverse b and of the sum b c. Conclude that the group
structure on X is well-defined even if the ground field K is not necessarily algebraically closed. (This
is important for practical computations, where one usually wants to work over finite fields in order
to avoid rounding errors.)
25
Let us now restrict our attention to the ground field C, so that an elliptic curve is topologically a
torus by Example 13.18 (b). In the remaining part of this chapter we want to see how these tori arise
in complex analysis in a totally different way. As we have not developed any analytic techniques in
this class we will only sketch most arguments; more details can be found e. g. in [K, Section 5.1]
(and many other books on complex analysis). Let us start by giving a quick review of what we will
need from standard complex analysis.
Remark 15.8 (Holomorphic and meromorphic functions). Let U C be an open set in the classical
topology. Recall that a function f : U C is called holomorphic if it is complex differentiable at
all points z0 U, i. e. if the limit
f 0 (z0 ) := lim
zz0
f (z) f (z0 )
z z0
exists. A function f : U C {} is called meromorphic if it is holomorphic except for some isolated singularities which are all poles, i. e. if for all z0 U there is a number n Z and a holomorphic
function f in a neighborhood V of z0 in U such that
f (z) = (z z0 )n f(z)
on V . If f does not vanish identically in a neighborhood of z0 we can moreover assume f(z0 ) 6= 0
in this representation; the number n is then uniquely determined. We will call it the order of f at z0
and denote it by ordz0 f . It is the analogue of the multiplicity of a rational function in Construction
14.5. If n > 0 we say that f has a zero of order n at z0 ; if n < 0 then f has a pole of order n
there. A meromorphic function is holomorphic around a point z0 if and only if its order at this point
is non-negative.
Of course, every regular (resp. rational) function on a Zariski-open subset of A1C = C is holomorphic
(resp. meromorphic). However, there are many holomorphic (resp. meromorphic) functions that are
not regular (resp. rational), e. g. f : C C, z 7 ez .
Remark 15.9 (Properties of holomorphic and meromorphic functions). Although the definition of
holomorphic, i. e. complex differentiable functions is formally exactly the same as that of real differentiable functions, the behavior of the complex and real cases is totally different. The most notable
differences that we will need are:
(a) Every holomorphic function is automatically infinitely differentiable: all higher derivatives
f (k) for k N exist and are again holomorphic [G4, Corollary 8.1].
(b) Every holomorphic function f is analytic, i. e. it can be represented locally around every
point z0 by its Taylor series. The radius of convergence is as large as it can be, i. e. if
f is holomorphic in an open ball U around z0 , then the Taylor series of f at z0 converges
and represents f at least on U. Consequently, a meromorphic function f of order n at z0
k
can be expanded in a Laurent series as f (z) =
k=n ck (z z0 ) [G4, Proposition 9.8]. The
coefficient c1 of this series is called the residue of f at z0 and denoted by resz0 f .
Residues are related to orders of meromorphic functions as follows: if f (z) = (z z0 )n f(z)
as in Remark 15.8 above, we obtain

f 0 (z)
n
f0 (z)
resz0
= resz0
+
= n = ordz0 f .
f (z)
z z0
f(z)
124
Andreas Gathmann
(c) (Residue Theorem) If is a closed (positively oriented) contour in C and f is a meromorphic

function in a neighborhood of and its interior, without poles on itself, then
Z
f (z) dz = 2i resz0 f (z),

z0
with the sum taken over all z0 in the interior of (at which f has poles) [G4, Proposition
11.10]. In particular, if f is holomorphic in the interior of then this integral vanishes.
(d) (Liouvilles Theorem) Every function that is holomorphic and bounded on the whole complex plane C is constant [G4, Proposition 8.2].
For our applications to elliptic curves we will need a particular
meromorphic function. To describe its construction, fix two complex numbers 1 , 2 C that are linearly independent over R, i. e.
that do not lie on the same real line in C through the origin. Then
= Z1 + Z2 = {m1 + n2 : m, n Z}
is called a lattice in C, as indicated by the points in the picture on

the right. Note that is an additive subgroup of C, and that the
quotient C/ is topologically a torus. We want to see that it can
be identified with an elliptic curve in a natural way, using a map
that we are going to introduce now.
Im
2
Re
1
Proposition and Definition 15.10 (The Weierstra -function). Let = Z1 + Z2 be a lattice

in C. There is a meromorphic function on C, called the Weierstra -function (pronounced like
the letter p), defined by

1
1
1
.
(z) = 2 +
z
(z )2 2
\{0}
It has poles of order 2 exactly at the lattice points.
Sketch proof. It is a standard fact that an (infinite) sum of holomorphic functions is holomorphic at
z0 provided that the sum converges uniformly in a neighborhood of z0 . We will only sketch the proof
of this convergence: let z0 C\ be a fixed point that is not in the lattice. Then every summand is
a holomorphic function in a neighborhood of z0 . The expansions of these summands for large are

1
1
1
2z
1
1
= 2
1 = 3 + terms of order at least 4 ,
(z )2 2
(1 z )2
so the summands grow like 3 . Let us add up these values according to the absolute value of .
Note that the number of lattice points with a given absolute value approximately equal to n N is
roughly proportional to the area of the annulus with inner radius n 12 and outer radius n + 21 , which
1
1
grows linearly with n. Hence the final sum behaves like

n=1 n n3 = n=1 n2 , which is convergent.
Note that the sum would not have been convergent without the subtraction of the constant 12 in each
summand, as then the individual terms would grow like 12 , and therefore the final sum would be of
1
the type
n=1 n , which is divergent.
Finally, the poles of order 2 at the points of are clearly visible.
Remark 15.11 (Properties of the -function). It is a standard fact that in an absolutely convergent
series as above all manipulations (reordering of the summands, term-wise differentiation) can be
performed as expected. In particular, the following properties of the -function are obvious:
(a) The -function is an even function, i. e. (z) = (z) for all z C. Hence its Laurent
series at 0 contains only even exponents.
15.
Elliptic Curves
125
2
0
0
(b) Its derivative is 0 (z) = (z)
3 . It is an odd function, i. e. (z) = (z) for all z.
In other words, its Laurent series at 0 contains only odd exponents. It has poles of order 3
exactly at the lattice points.
(c) The -function is doubly periodic with respect to , i. e. (z0 ) = (z0 + ) for all z0 C
and . To show this note first that it is obvious from (b) that 0 (z0 ) = 0 (z0 + ). Now
integrate 0 (z) along the closed contour = 1 + 2 + 3 + 4 shown in the picture below on
the right.
Im
Of course, the result is 0, since is an integral of 0 .
z0
z0 +
But also the integral along 2 cancels the integral along
0
1
4 as (z) is periodic. The integral along 3 is equal to
( 2 ) ( 2 ), so it vanishes as well since is an even
4
2
function. So we conclude that
3
Re
Z
0 (z) dz = (z0 + ) (z0 ),
0=
1
i. e. is periodic with respect to .

Lemma 15.12. The -function associated to a lattice satisfies a differential equation
0 (z)2 = c3(z)3 + c2(z)2 + c1(z) + c0
for all z C
for some constants c0 , c1 , c2 , c3 C (depending on ).

Proof. By Remark 15.11 (b) we know that (0 )2 is an even function with a pole of order 6 at the
origin. Hence its Laurent series around 0 is of the form
a6 a4 a2
0 (z)2 = 6 + 4 + 2 + a0 + (terms of positive order)
z
z
z
for some constants a6 , a4 , a2 , a0 C. The functions 3 , 2 , , and 1 are also even and have
poles at the origin of order 6, 4, 2, and 0, respectively. Hence there are constants c3 , c2 , c1 , c0 C
such that the Laurent series of the linear combination
f (z) := 0 (z)2 c3(z)3 c2(z)2 c1(z) c0
has only positive powers of z. This means that f is holomorphic around the origin and vanishes at 0.
But and 0 , and hence also f , are -periodic by Remark 15.11 (c). Hence f is holomorphic
around all lattice points. But f is also holomorphic around all other points, as and 0 are. In
other words, f is holomorphic on all of C.
Moreover, the periodicity means that every value taken on by f is already assumed on the parallelogram {x1 + y2 : x, y [0, 1]}. As f is continuous, its image on this compact parallelogram, and
hence on all of C, is bounded. So we see by Liouvilles Theorem of Remark 15.9 (d) that f must be
constant. But as we have already shown that f (0) = 0, it follows that f is the zero function, which
is exactly the statement of the lemma.

Remark 15.13. By an explicit computation one can show that the coefficients c3 , c2 , c1 , c0 in Lemma
15.12 are given by
1
1
c3 = 4, c2 = 0, c1 = 60
, and c0 = 140
.
4
6
\{0}
\{0}
The proof of Lemma 15.12 shows impressively the powerful methods of complex analysis: to prove
our differential equation, i. e. the equality of the two functions (0 )2 and c33 + c22 + c1 + c0 ,
it was sufficient to just compare four coefficients of their Laurent expansions at the origin the rest
then follows entirely from general theory.
Note also that the differential equation of Lemma 15.12 is a (non-homogeneous) cubic equation in
the two functions and 0 , which are -periodic and thus well-defined on the quotient C/. We
can therefore use it to obtain a map from C/ to an elliptic curve as follows.
126
Andreas Gathmann
Proposition 15.14. Let C be a fixed lattice, and let X P2C be the cubic curve
X = {(x0 : x1 : x2 ) : x22 x0 = c3 x13 + c2 x12 x0 + c1 x1 x02 + c0 x03 }
for the constants c3 , c2 , c1 , c0 C of Lemma 15.12. Then there is a bijection
: C/ X, z 7 (1 :(z) :0 (z)).
Proof. As and 0 are periodic with respect to and satisfy the differential equation of Lemma
15.12, it is clear that is well-defined as a map to X. (Strictly speaking, for z = 0 we have to note that
and 0 have poles of order 2 and 3, respectively, so that the given expression for (0) formally
g(z)
0
looks like (1 : : ). But by Remark 15.8 we can write (z) = fz(z)
2 and (z) = z3 locally around
the origin for some holomorphic functions f , g that do not vanish at 0, and so we have to interpret
the expression for as
(0) = lim (1 :(z) :0 (z)) = lim (z3 : z f (z) : g(z)) = (0 : 0 : 1),
z0
z0
i. e. (z) is well-defined at z = 0 as well.)

Now let (x0 : x1 : x2 ) X be a given point; we will show that it has exactly one inverse image point
under . By what we have just said this is obvious for the point at infinity (0 : 0 : 1), so let us
assume that we are not at this point and hence pass to inhomogeneous coordinates where x0 = 1.
We will first look for a number z C such that (z) = x1 . To do
so, consider the integral
Z
Im
0 (z)
dz
(z) x1
over the boundary of any parallelogram of periodicity (that does
not meet the zeroes and poles of the function z 7 (z) x1 ), as
in the picture on the right. The integrals along opposite sides of
the parallelogram vanish because of the periodicity of and 0 ,
so that the total integral must be 0. Hence by Remark 15.9 (b) and
(c) we get
0=
z0 C/
resz0
0 (z)
=
(z) x1 z
Re
ordz0 ((z) x1 ).
0 C/
In other words, the function z 7 (z) x1 has as many zeroes as it has poles in C/, counted with
multiplicities. (This is a statement in complex analysis analogous to the algebraic result of Remark
14.8 (b).) As has a pole of order 2 in the lattice points, it thus follows that there are exactly two
points in 1 (x1 ), counted with multiplicities.
For such a point z with (z) = x1 we then have by Lemma 15.12
0 (z)2 = c3(z)3 + c2(z)2 + c1(z) + c0 = c3 x12 + x2 x12 + c1 x1 + c0 = x22
since (1 : x1 : x2 ) X. So there are two possibilities:
0 (z) = 0: Then x2 = 0 as well, and z is a double zero (i. e. the only zero) of the function
z 7 (z) x1 . So there is exactly one z C/ with (z) = (1 :(z) :0 (z)) = (1 : x1 : x2 ).
0 (z) 6= 0: Then z is only a simple zero of z 7 (z) x1 . As is even and 0 odd
by Remark 15.11, we see that z must be the other zero, and it satisfies 0 (z) = 0 (z).
Hence exactly one of the equations0 (z) = x2 and0 (z) = x2 holds, and the corresponding
point is the unique inverse image of (1 : x1 : x2 ) under .
Altogether we conclude that is bijective, as we have claimed.
Remark 15.15. With Proposition 15.14 we are again in a similar situation as in Proposition 15.2:
we have a bijection between a group C/ and a variety X, so that the map of the above proposition
can be used to construct a group structure on X. In fact, we will see in Exercise 15.17 that this group
structure is precisely the same as that obtained by the map of Proposition 15.2 using divisors. But
15.
Elliptic Curves
127
the algebraic properties of this group structure is a lot more obvious in this new picture: for example,
the points of order n are easily read off to be the n2 points
1
(i1 + j2 ) for 0 i, j < n.
n
It should be said however that the analytic bijection of Proposition 15.14 differs from that of Proposition 15.2 in that both C/ and X can independently be made into a 1-dimensional complex manifold,
and the map of the above proposition is then an isomorphism between these two manifolds.
Exercise 15.16. Using the identification of an elliptic curve X with a torus C/ as in Proposition
15.14, reprove the statement of Proposition 14.19 that there is no rational function on an elliptic
curve X with divisor div = a b for distinct points a, b X.
Exercise 15.17. Let X be an elliptic curve corresponding to a torus C/. Show that the group
structure of Pic0X is isomorphic to the natural group structure of C/.
Exercise 15.18. Let C be a lattice. Given two points z, w C/, it is obviously very easy to
find a natural number n such that n w = z (in the group structure of C/), in case such a number
exists. Why is this no contradiction to the idea of the cryptographic application in Example 15.6?
26
128
References
References
[AM] M. Atiyah, I. MacDonald, Introduction to Commutative Algebra, Addison Wesley (2004)
[E]
D. Eisenbud, Commutative Algebra with a View towards Algebraic Geometry, Springer (2004)
[EGA] A. Grothendieck, J. Dieudonn, lments de Gomtrie Algbrique, Publications Mathmatiques IHES (various
volumes)
[G1] A. Gathmann, Algebraische Strukturen, class notes TU Kaiserslautern (2010),
www.mathematik.uni-kl.de/~gathmann/ags.php
[G2] A. Gathmann, Grundlagen der Mathematik, class notes TU Kaiserslautern (2011/12),
www.mathematik.uni-kl.de/~gathmann/gdm.php
[G3] A. Gathmann, Einfhrung in die Algebra, class notes TU Kaiserslautern (2010/11),
www.mathematik.uni-kl.de/~gathmann/algebra.php
[G4] A. Gathmann, Einfhrung in die Funktionentheorie, class notes TU Kaiserslautern (2014/15),
www.mathematik.uni-kl.de/~gathmann/futheo.php
[G5] A. Gathmann, Commutative Algebra, class notes TU Kaiserslautern (2014),
www.mathematik.uni-kl.de/~gathmann/commalg.php
[GP] G.-M. Greuel, G. Pfister, A Singular Introduction to Commutative Algebra, Springer (2002)
[H]
R. Hartshorne, Algebraic Geometry, Springer Graduate Texts in Mathematics 52 (1977)
[Ha]
J. Harris, Algebraic Geometry, Springer Graduate Texts in Mathematics 133 (1992)
[K]
F. Kirwan, Complex Algebraic Curves, London Mathematical Society Student Texts 23, Cambridge University Press
(1992)
[M]
R. Miranda, Algebraic Curves and Riemann Surfaces, AMS Graduate Studies in Mathematics 5 (1995)
[M1] D. Mumford, The Red Book of Varieties and Schemes, Springer Lecture Notes in Mathematics 1358 (1988)
[M2] D. Mumford, Algebraic Geometry I Complex Projective Varieties, Springer Classics in Mathematics (1976)
[R]
M. Reid, Undergraduate Algebraic Geometry, London Mathematical Society Student Texts 12, Cambridge University
Press (1988)
[S]
W. Decker, G.-M. Greuel, G. Pfister, H. Schnemann, S INGULAR a computer algebra system for polynomial
computations,
www.singular.uni-kl.de
[S1]
I. Shafarevich, Basic Algebraic Geometry I Varieties in Projective Space, Springer (1994)
[S2]
I. Shafarevich, Basic Algebraic Geometry II Schemes and Complex Manifolds, Springer (1994)
Note: This is a very extensive list of literature of varying usefulness. Here is a short recommendation
which of the references you might want to use for what:
For motivational aspects, examples, and a generally fairy-tale style introduction without
much theoretical background, see [Ha], [R], or maybe [S1] and [S2].
For questions concerning complex curves, in particular in relation to complex analysis, see
[K] or [M].
For a general reference on the commutative algebra background, see [AM] or [G5].
For commutative algebra problems involving computational aspects, see [GP].
For a good book that develops the theory thoroughly, but largely lacks motivations and examples, see [H]. You should not try to read the hard-core parts of this book without some
motivational background.
For the ultimate reference (if it is not proven there, it must be wrong), see [EGA]. Warning:
this is unreadable if you do not have a decent background in algebraic geometry yet, and
close to being unreadable even if you do.
Index
129
Index
An 9
A(X) 13
affine conic 35
affine coordinates 47
affine curve 23
affine hypersurface 23
affine Jacobi criterion 85
affine open set 38
affine space 9
affine subvariety 13
affine surface 23
affine variety 9, 36
product 10, 35
affine zero locus 9
algebra 25
graded 48
algebra homomorphism 25
algebraic set 9
almost all 96
alternating tensor 67
antisymmetric tensor 67
arithmetic genus 112
automorphism
projective 58
Basis Theorem 10, 14, 51
Bzouts Theorem 100
for curves 101
local version 103
birational map 72
birational varieties 72
blow-up
at a closed subvariety 76
at an ideal 76
at polynomials 73
of a projective variety 77
of a variety 76
of an affine variety 73
Ca X 77
C(X) 50
cell decomposition 110
class
of a divisor 115
closed map 62
closed set 16
closed subprevariety 42
locally 42
closed subvariety 45
closure 16
codimension 20
complete variety 63
component
embedded 98
irreducible 19
cone 50
over a subset 50
tangent 77
conic
affine 35
projective 59
connected space 17
continuous map 16
coordinate ring
affine 13
homogeneous 51
coordinates
affine 47
homogeneous 46
local 116
Plcker 68
Segre 60
Veronese 64
Cremona transformation 81
cross product 66
cubic 101
cubic curve 118, 120
cubic surface 6, 90
curve 45
affine 23
cubic 118, 120
dual 88
elliptic 120
X 44
decomposition
cell 110
homogeneous 48
irreducible 19
primary 19, 98
deg I 99
deg X 99
degree
in a graded ring 48
of a divisor 113
of a homogeneous ideal 99
degree-genus formula 5, 111
dense 20
diagonal 44
dimK V 20
dim X 20
dimension 20
local 23
pure 23
disconnected space 17
distinguished open subset 26
div f 114
div 115
Div X 113
Div0 X 113
divisor 113
class 115
class group 115
130
effective 113
group 113
intersection 114
of a polynomial 114
of a rational function 115
principal 115
dominant map 72
dual curve 88
effective divisor 113
elimination theory
main theorem 62
elliptic curve 120
embedded component 98
embedding
Plcker 68
Segre 60
Veronese 64
Euler characteristic 110
even loop 108
exact sequence 96
exceptional hypersurface 79
exceptional set 74
Fa 29
f h 53
f in 79
Fermat hypersurface 88
free Abelian group 113
function
Hilbert 95
holomorphic 123
local 29
meromorphic 123
rational 73
regular 24, 38, 57
function field 73
G(k, n) 66
generic smoothness 88
genus
arithmetic 112
of a curve 4, 110
germ 29
gluing
of prevarieties 38, 41
graded algebra 48
graded ring 48
graph 45
Grassmannian variety 66
group
free Abelian 113
of divisor classes 115
of divisors 113
Picard 115
hI 95
hX 95
Hausdorff space 44
Hilbert
Basis Theorem 10, 14, 51
Nullstellensatz 12, 14
Hilbert function
Index
Hilbert polynomial
holomorphic function 123
homogeneous coordinate ring 51
homogeneous coordinates 46
homogeneous decomposition 48
homogeneous element 48
homogeneous ideal 48
homogenization
of a polynomial 53
of an ideal 53
hyperplane 101
hypersurface 45
affine 23
exceptional 79
Fermat 88
I(X) 10, 13, 49, 51
Ia 30
Ia (X) 49
I h 53
I p (X) 49, 51
IY (X) 13, 51
ideal
degree 99
homogeneous 48
homogenization 53
intersection 15
irrelevant 52
maximal 30
of a set 10, 13, 49
of an affine variety 10
primary 98
prime 18
product 14
radical 12
sum 14
Identity Theorem 25
Implicit Function Theorem 86
incidence correspondence 71, 91
infinity
point in projective space 47
initial term 79
intersection divisor 114
intersection multiplicity 102
intersection of ideals 15
irreducible component 19
irreducible decomposition 19
irreducible space 17
irrelevant ideal 52
isomorphism
of ringed spaces 33
Jacobi criterion
affine 85
projective 86
Jacobian matrix 85
join 71
K(X) 73
k K n 66
lattice 124
Index
Laurent series 123

line 101
linear subspace
of Pn 50
Liouvilles Theorem 124
local coordinate 116
local dimension 23
local function 29
local ring 30
regular 85
localization
of a ring 27
locally closed subprevariety 42
locally closed subvariety 45
locally irreducible variety 85
loop
even 108
odd 108
of a real curve 107
main theorem
of elimination theory 62
map
birational 72
closed 62
continuous 16
rational 72
meromorphic 123
morphism
of algebras 25
of ringed spaces 33
multa ( f ) 102
multa (I) 102
multa () 114
multa (X, f ) 102
multa (X,Y ) 102
multiplicatively closed set 27
multiplicity
intersection 102
of a polynomial 102
of an ideal 102
Noetherian space 18
non-singular point 84
non-singular variety 84
Nullstellensatz 12, 14
projective 52
OX (U) 24, 57
OX 29, 32
OX,a 30
odd loop 108
open set 16
distinguished 26
open subprevariety 42
open subvariety 45
ordz0 f 123
order
of a meromorphic function 123
124
P1 39
Pn 46
P(X) 50
131
(a, b) 61, 118

Pic X 115
Pic0 X 115
Picard group 115
Picard variety 121
plane 101
Plcker coordinates 68
Plcker embedding 68
point
at infinity 47
non-singular 84
ramification 111
regular 84
singular 84
smooth 84
pole
polynomial 9
Hilbert 98
homogenization 53
polynomial function 13
polynomial ring 9
presheaf 28
germ 29
restriction 29
stalk 29
prevariety 38
gluing 38, 41
product 43
separated 44
primary decomposition 19, 98
primary ideal 98
prime ideal 18
Prin X 115
principal divisor 115
product
of affine varieties 10, 35
of ideals 14
of prevarieties 43
topology 17
universal property 35, 43
projection 59
projective automorphism 58
projective conic 59
projective Jacobi criterion 86
projective Nullstellensatz 52
projective space 46
projective subvariety 51
projective variety 49
projective zero locus 49
projectivization
of a cone 50
pull-back
of a function 33
pure dimension 23
pure tensor 69
quadric 73, 101
quartic 101
quotient field 73
quotient topology 39
radical
132
of an ideal 12
radical ideal 12
ramification point 111
rational function 73
rational map 72
dominant 72
reducible space 17
regular function
on a prevariety 38
on a projective variety 57
on an affine variety 24
sheaf 29
regular local ring 85
regular point 84
regular variety 84
Removable Singularity Theorem 28
resz0 f 123
residue 123
Residue Theorem 123
resolution of singularities 87
restriction
of a presheaf 29
restriction map 28
ring
graded 48
local 30
regular local 85
ringed space 32
isomorphism 33
morphism 33
S(X) 51
section
of a presheaf 28
Segre coordinates 60
Segre embedding 60
separated prevariety 44
sequence
exact 96
series
Laurent 123
set
affine open 38
algebraic 9
closed 16
dense 20
distinguished open 26
exceptional 74
multiplicatively closed 27
open 16
sheaf 28
germ 29
of regular functions 29
restriction 29
stalk 29
structure 32
singular point 84
resolution 87
singular variety 84
smooth point 84
smooth variety 84
generically 88
space
affine 9
Index
projective 46
ringed 32
tangent 82
stalk 29
strict transform 74
structure sheaf 32
subprevariety
closed 42
locally closed 42
open 42
subspace topology 16
subvariety
affine 13
closed 45
locally closed 45
open 45
projective 51
sum of ideals 14
surface 45
affine 23
cubic 6, 90
tangent cone 77
tangent space 82
tensor
pure 69
tensor product
alternating 67
antisymmetric 67
topological Euler characteristic 110
topological space
connected 17
disconnected 17
Hausdorff 44
irreducible 17
Noetherian 18
reducible 17
topology 16
product 17
quotient 39
subspace 16
Zariski 16, 53
transform
strict 74
universal property
of products 35, 43
V (S) 9, 13, 49, 51
Va (S) 49
Vp (S) 49, 51
VY (S) 13, 51
value
of a polynomial 9
vanishing multiplicity 102
variety 44
affine 9, 36
birational 72
complete 63
Grassmannian 66
non-singular 84
Picard 121
projective 49
regular 84
Index
singular 84
smooth 84
Veronese coordinates 64
Veronese embedding 64
Weierstra
-function 124
I 98
X 98
Zariski topology 16, 53
zero
zero locus 13
affine 9
projective 49
133

Gathmann Algebraic Geometry

Uploaded by

Copyright:

Available Formats

Gathmann Algebraic Geometry

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Gathmann Algebraic Geometry

Uploaded by

Copyright:

Available Formats

Algebraic Geometry

Class Notes TU Kaiserslautern 2014

for 0 2 and fixed r > 0

which should have the same genus as the original Cd .

Exercise 0.6. In Example 0.3, we argued

ai1 ,...,in x1i1 xnin

ai1 ,...,in x1i1 xnin

and a point c = (c1 , . . . , cn ) An we define the value of f at c to be

ai1 ,...,in ci11 cinn

(c) Linear subspaces of An = K n are affine varieties.

We have now constructed operations

be subsets of K[x1 , . . . , xn ], and let X and

(a) If X X 0 then I(X 0 ) I(X).

Remark 1.15. Again let I be an ideal in a ring R.

(a) The radical I of I is always an ideal:

If f , g I, i. e. f n I and gm I for some n, m N, then

If h R and f I, i. e. f n I for some n N, then (h f )n = hn f n I, and hence

Moreover, it is obvious that

(b) As expected from the terminology, the radical of I is a radical ideal:

{radical ideals in K[x1 , . . . , xn ]}

the correspondence reverses inclusions by Lemma 1.12 (a).

{radical ideals in A(Y )}

(c) For a subvariety X of Y we have A(X)

with this modified definition of the product J1 J2 .

For f J1 J2 we have f n J1 J2 for some n. Then f n J1 J2 , and thus f J1 J2 .

(Remark 1.25 (a))

The Zariski Topology

The Zariski Topology

Exercise 2.4. Let X An be an arbitrary subset. Prove that V (I(X)) = X.

Definition 2.6 (Irreducible and connected spaces). Let X be a topological space.

The Zariski Topology

Lemma 2.13. Any affine variety is a Noetherian topological space.

Remark 2.16 (Computation of irreducible decompositions). In general, the actual computation of

The Zariski Topology

(a) X is irreducible by Proposition 2.9 since its coordinate ring

The Zariski Topology

The Sheaf of Regular Functions

The Sheaf of Regular Functions

Definition 3.4 (K-algebras [G5, Definition 1.23 and Remark 1.24]).

The Sheaf of Regular Functions

and thus by Proposition 1.21 (b)

on D( fb ), and these open subsets cover D( f ).

then f k (g f m g0 f n ) = 0 in A(X) for some k N, which means that g f m = g0 f n and thus

The homomorphism is surjective by Proposition 3.10. It is also injective: if fgn = 0 as a function on

The Sheaf of Regular Functions

to the corresponding quotient of polynomial functions on the open

neighborhood of a where the denominator does not vanish. It is well-defined: if

gf I, and we conclude that I = OX,a .

The Sheaf of Regular Functions

{K-algebra homomorphisms A(Y ) A(X)}

In particular, isomorphisms of affine varieties correspond exactly to K-algebra isomorphisms in this

which is a morphism by Proposition 4.7. Its corresponding K-algebra

K[x1 , x2 ]/(x12 x23 ) K[t]

which can be seen by composing f with the two coordinate functions of

As an example of gluing morphisms as in Lemma 4.6, the morphisms

x22 x12n = (1 x1 )(1 2x1 ) (1 2nx1 ).

X2 = {(x1 , x2 ) A2C : x22 = (1 x1 )(1 2x1 ) (1 2nx1 )}

along the isomorphism

where U1,2 = {(x1 , x2 ) : x1 6= 0} X1 and U2,1 = {(x1 , x2 ) : x1 6= 0} X2 .