J.S. Milne
Version 5.22
January 13, 2012
These notes are an introduction to the theory of algebraic varieties. In contrast to most
such accounts they study abstract algebraic varieties, and not just subvarieties of afne and
projective space. This approach leads more naturally into scheme theory.
BibTeX information
@misc{milneAG,
author={Milne, James S.},
title={Algebraic Geometry (v5.22)},
year={2012},
note={Available at www.jmilne.org/math/},
pages={260}
}
v2.01 (August 24, 1996). First version on the web.
v3.01 (June 13, 1998).
v4.00 (October 30, 2003). Fixed errors; many minor revisions; added exercises; added two
sections/chapters; 206 pages.
v5.00 (February 20, 2005). Heavily revised; most numbering changed; 227 pages.
v5.10 (March 19, 2008). Minor xes; T
E
Xstyle changed, so page numbers changed; 241
pages.
v5.20 (September 14, 2009). Minor corrections; revised Chapters 1, 11, 16; 245 pages.
v5.21 (March 31, 2011). Minor changes; changed T
E
Xstyle; 258 pages.
v5.22 (January 13, 2012). Minor xes; 260 pages.
Available at www.jmilne.org/math/
Please send comments and corrections to me at the address on my web page.
The photograph is of Lake Sylvan, New Zealand.
Copyright c _ 1996, 1998, 2003, 2005, 2008, 2009, 2011, 2012 J.S. Milne.
Single paper copies for noncommercial personal use may be made without explicit permis
sion from the copyright holder.
Contents
Contents 3
Notations 6; Prerequisites 6; References 6; Acknowledgements 6
1 Preliminaries 11
Rings and algebras 11; Ideals 11; Noetherian rings 13; Unique factorization 16; Polynomial
rings 18; Integrality 19; Direct limits (summary) 21; Rings of fractions 22; Tensor Prod
ucts 25; Categories and functors 28; Algorithms for polynomials 31; Exercises 37
2 Algebraic Sets 39
Denition of an algebraic set 39; The Hilbert basis theorem 40; The Zariski topology 41;
The Hilbert Nullstellensatz 42; The correspondence between algebraic sets and ideals 43;
Finding the radical of an ideal 46; The Zariski topology on an algebraic set 46; The coor
dinate ring of an algebraic set 47; Irreducible algebraic sets 48; Dimension 50; Exercises
52
3 Afne Algebraic Varieties 55
Ringed spaces 55; The ringed space structure on an algebraic set 57; Morphisms of ringed
spaces 60; Afne algebraic varieties 60; The category of afne algebraic varieties 61;
Explicit description of morphisms of afne varieties 63; Subvarieties 65; Properties of the
regular map dened by specm() 66; Afne space without coordinates 67; Exercises 68
4 Algebraic Varieties 71
Algebraic prevarieties 71; Regular maps 72; Algebraic varieties 73; Maps fromvarieties to
afne varieties 74; Subvarieties 75; Prevarieties obtained by patching 76; Products of vari
eties 76; The separation axiom revisited 81; Fibred products 84; Dimension 85; Birational
equivalence 86; Dominant maps 87; Algebraic varieties as a functors 87; Exercises 89
5 Local Study 91
Tangent spaces to plane curves 91; Tangent cones to plane curves 93; The local ring at a
point on a curve 93; Tangent spaces of subvarieties of A
n
94; The differential of a regular
map 96; Etale maps 97; Intrinsic denition of the tangent space 99; Nonsingular points
102; Nonsingularity and regularity 103; Nonsingularity and normality 104; Etale neigh
bourhoods 105; Smooth maps 107; Dual numbers and derivations 108; Tangent cones 111;
Exercises 113
6 Projective Varieties 115
Algebraic subsets of P
n
115; The Zariski topology on P
n
118; Closed subsets of A
n
and
P
n
119; The hyperplane at innity 119; P
n
is an algebraic variety 120; The homogeneous
3
coordinate ring of a subvariety of P
n
122; Regular functions on a projective variety 123;
Morphisms from projective varieties 124; Examples of regular maps of projective vari
eties 125; Projective space without coordinates 130; Grassmann varieties 130; Bezouts
theorem 134; Hilbert polynomials (sketch) 135; Exercises 136
7 Complete varieties 137
Denition and basic properties 137; Projective varieties are complete 139; Elimination
theory 140; The rigidity theorem 142; Theorems of Chow 143; Nagatas Embedding The
orem 144; Exercises 144
8 Finite Maps 145
Denition and basic properties 145; Noether Normalization Theorem 149; Zariskis main
theorem 150; The base change of a nite map 152; Proper maps 152; Exercises 154
9 Dimension Theory 155
Afne varieties 155; Projective varieties 162
10 Regular Maps and Their Fibres 165
Constructible sets 165; Orbits of group actions 168; The bres of morphisms 170; The
bres of nite maps 172; Flat maps 173; Lines on surfaces 174; Stein factorization 180;
Exercises 180
11 Algebraic spaces; geometry over an arbitrary eld 181
Preliminaries 181; Afne algebraic spaces 185; Afne algebraic varieties. 186; Algebraic
spaces; algebraic varieties. 187; Local study 192; Projective varieties. 193; Complete
varieties. 194; Normal varieties; Finite maps. 194; Dimension theory 194; Regular maps
and their bres 194; Algebraic groups 194; Exercises 194
12 Divisors and Intersection Theory 195
Divisors 195; Intersection theory. 196; Exercises 201
13 Coherent Sheaves; Invertible Sheaves 203
Coherent sheaves 203; Invertible sheaves. 205; Invertible sheaves and divisors. 206; Direct
images and inverse images of coherent sheaves. 208; Principal bundles 209
14 Differentials (Outline) 211
15 Algebraic Varieties over the Complex Numbers 215
16 Descent Theory 219
Models 219; Fixed elds 219; Descending subspaces of vector spaces 220; Descending
subvarieties and morphisms 222; Galois descent of vector spaces 223; Descent data 224;
Galois descent of varieties 228; Weil restriction 229; Generic bres and specialization 229;
Rigid descent 230; Weils descent theorems 232; Restatement in terms of group actions
234; Faithfully at descent 236
17 Lefschetz Pencils (Outline) 239
Denition 239
18 Algebraic Schemes and Algebraic Spaces 243
4
A Solutions to the exercises 245
B Annotated Bibliography 253
Index 257
QUESTION: If we try to explain to a layman what algebraic geometry is, it seems to me that
the title of the old book of Enriques is still adequate: Geometrical Theory of Equations....
GROTHENDIECK: Yes! but your layman should know what a system of algebraic equa
tions is. This would cost years of study to Plato.
5
Notations
We use the standard (Bourbaki) notations: N={0. 1. 2. . . .], Z =ring of integers, 1=eld
of real numbers, C = eld of complex numbers, F
]
= Z,Z = eld of elements, a
prime number. Given an equivalence relation, + denotes the equivalence class containing
+. A family of elements of a set indexed by a second set 1, denoted (a
i
)
iJ
, is a function
i a
i
: 1 .
A eld k is said to be separably closed if it has no nite separable extensions of degree
>1. We use k
sep
and k
al
to denote separable and algebraic closures of k respectively.
All rings will be commutative with 1, and homomorphisms of rings are required to map
1 to 1. For a ring ,
}=1
a
i}
X
}
=b
i
. i =1. . . . . m. (1)
the starting point for algebraic geometry is the study of the solutions of systems of polyno
mial equations,
i
(X
1
. . . . . X
n
) =0. i =1. . . . . m.
i
kX
1
. . . . . X
n
.
Note immediately one difference between linear equations and polynomial equations: the
orems for linear equations dont depend on which eld k you are working over,
1
but those
for polynomial equations depend on whether or not k is algebraically closed and (to a lesser
extent) whether k has characteristic zero.
A better description of algebraic geometry is that it is the study of polynomial functions
and the spaces on which they are dened (algebraic varieties), just as topology is the study
of continuous functions and the spaces on which they are dened (topological spaces),
differential topology the study of innitely differentiable functions and the spaces on which
they are dened (differentiable manifolds), and so on:
algebraic geometry regular (polynomial) functions algebraic varieties
topology continuous functions topological spaces
differential topology differentiable functions differentiable manifolds
complex analysis analytic (power series) functions complex manifolds.
The approach adopted in this course makes plain the similarities between these different
areas of mathematics. Of course, the polynomial functions form a much less rich class than
the others, but by restricting our study to polynomials we are able to do calculus over any
eld: we simply dene
J
JX
a
i
X
i
=
i a
i
X
i1
.
1
For example, suppose that the system (1) has coefcients a
i}
k and that 1 is a eld containing k. Then
(1) has a solution in k
n
if and only if it has a solution in 1
n
, and the dimension of the space of solutions is the
same for both elds. (Exercise!)
7
8 INTRODUCTION
Moreover, calculations (on a computer) with polynomials are easier than with more general
functions.
Consider a nonzero differentiable function (.. ,. :). In calculus, we learn that the
equation
(.. ,. :) =C (2)
denes a surface S in 1
3
, and that the tangent plane to S at a point 1 = (a. b. c) has
equation
2
_
d
d.
_
1
(. a)
_
d
d,
_
1
(, b)
_
d
d:
_
1
(: c) =0. (3)
The inverse function theorem says that a differentiable map : S S
t
of surfaces is a local
isomorphism at a point 1 S if it maps the tangent plane at 1 isomorphically onto the
tangent plane at 1
t
=(1).
Consider a nonzero polynomial (.. ,. :) with coefcients in a eld k. In this course,
we shall learn that the equation (2) denes a surface in k
3
, and we shall use the equation
(3) to dene the tangent space at a point 1 on the surface. However, and this is one of the
essential differences between algebraic geometry and the other elds, the inverse function
theorem doesnt hold in algebraic geometry. One other essential difference is that 1,X is
not the derivative of any rational function of X, and nor is X
n]1
in characteristic =0
these functions can not be integrated in the ring of polynomial functions.
The rst ten chapters of the notes form a basic course on algebraic geometry. In these
chapters we generally assume that the ground eld is algebraically closed in order to be
able to concentrate on the geometry. The remaining chapters treat more advanced topics,
and are largely independent of one another except that Chapter 11 should be read rst.
The approach to algebraic geometry taken in these notes
In differential geometry it is important to dene differentiable manifolds abstractly, i.e., not
as submanifolds of some Euclidean space. For example, it is difcult even to make sense
of a statement such as the Gauss curvature of a surface is intrinsic to the surface but the
principal curvatures are not without the abstract notion of a surface.
Until the mid 1940s, algebraic geometry was concerned only with algebraic subvarieties
of afne or projective space over algebraically closed elds. Then, in order to give substance
to his proof of the congruence Riemann hypothesis for curves an abelian varieties, Weil
was forced to develop a theory of algebraic geometry for abstract algebraic varieties over
arbitrary elds,
3
but his foundations are unsatisfactory in two major respects:
Lacking a topology, his method of patching together afne varieties to form abstract
varieties is clumsy.
His denition of a variety over a base eld k is not intrinsic; specically, he xes
some large universal algebraically closed eld and denes an algebraic variety
over k to be an algebraic variety over with a kstructure.
In the ensuing years, several attempts were made to resolve these difculties. In 1955,
Serre resolved the rst by borrowing ideas from complex analysis and dening an algebraic
variety over an algebraically closed eld to be a topological space with a sheaf of functions
2
Think of S as a level surface for the function , and note that the equation is that of a plane through
(a. b. c) perpendicular to the gradient vector ( )
1
of at 1.
3
Weil, Andr e. Foundations of algebraic geometry. American Mathematical Society, Providence, R.I. 1946.
9
that is locally afne.
4
Then, in the late 1950s Grothendieck resolved all such difculties by
introducing his theory of schemes.
In these notes, we follow Grothendieck except that, by working only over a base eld,
we are able to simplify his language by considering only the closed points in the underlying
topological spaces. In this way, we hope to provide a bridge between the intuition given by
differential geometry and the abstractions of scheme theory.
4
Serre, JeanPierre. Faisceaux alg ebriques coh erents. Ann. of Math. (2) 61, (1955). 197278.
CHAPTER 1
Preliminaries
In this chapter, we review some denitions and basic results in commutative algebra and
category theory, and we derive some algorithms for working in polynomial rings.
Rings and algebras
Let be a ring. A subring of is a subset that contains 1
on and sending X
i
to .
i
is surjective. We
then write T =(i
B
).
1
. . . . . .
n
.
A ring homomorphism T is said to be of nitetype, and T is a nitely generated
algebra if T is generated by a nite set of elements as an algebra.
A ring homomorphism T is nite, and T is a nite
1
algebra, if T is nitely
generated as an module.
Let k be a eld, and let be a kalgebra. When 1
= 0 in , the map k is
injective, and we can identify k with its image, i.e., we can regard k as a subring of .
When 1
.
Ideals
Let be a ring. An ideal a in is a subset such that
(a) a is a subgroup of regarded as a group under addition;
(b) a a, r =ra a.
The ideal generated by a subset S of is the intersection of all ideals a containing S
it is easy to verify that this is in fact an ideal, and that it consists of all nite sums of the
1
The term modulenite is also used.
11
12 1. PRELIMINARIES
form
r
i
s
i
with r
i
, s
i
S. The ideal generated by the empty set is the zero ideal {0].
When S ={s
1
. s
2
. . . .], we shall write (s
1
. s
2
. . . .) for the ideal it generates.
Let a and b be ideals in . The set {a b [ a a. b b] is an ideal, denoted by ab.
The ideal generated by {ab [ a a. b b] is denoted by ab. Clearly ab consists of all nite
sums
a
i
b
i
with a
i
a and b
i
b, and if a = (a
1
. . . . . a
n
) and b = (b
1
. . . . . b
n
), then
ab =(a
1
b
1
. . . . . a
i
b
}
. . . . . a
n
b
n
). Note that
ab ab. (4)
The kernel of a homomorphismT is an ideal in . Conversely, for any ideal a in ,
the set of cosets of a in forms a ring ,a, and a aa is a homomorphism: ,a
whose kernel is a. The map b
1
(b) is a onetoone correspondence between the ideals
of ,a and the ideals of containing a.
An ideal p is prime if p = and ab p =a p or b p. Thus p is prime if and only
if ,p is nonzero and has the property that
ab =0. b =0 =a =0.
i.e., ,p is an integral domain. Note that if p is prime and a
1
a
n
p, then at least one of
the a
i
p.
An ideal m in is maximal if it is maximal among the proper ideals of . Thus m is
maximal if and only if ,m is nonzero and has no proper nonzero ideals, and so is a eld.
Note that
m maximal == m prime.
The ideals of T are all of the form a b with a and b ideals in and T. To see
this, note that if c is an ideal in T and (a. b) c, then (a. 0) = (1. 0)(a. b) c and
(0. b) =(0. 1)(a. b) c. Therefore, c =ab with
a ={a [ (a. 0) c]. b ={b [ (0. b) c].
Ideals a and b in are coprime if a b =. When a and b are coprime, there exist
a a and b b such that ab =1. For .. , , let : =a, b.; then
: b. . moda
: a, , modb.
and so the canonical map
,a,b (5)
is surjective. Clearly its kernel is ab, which contains ab. Let c ab; then
c =cacb ab.
Hence, (5) is surjective with kernel ab. This statement extends to nite sets of ideals.
THEOREM 1.1 (CHINESE REMAINDER THEOREM). Let a
1
. . . . . a
n
be ideals in a ring .
If a
i
is coprime to a
}
whenever i = , then the map
,a
1
,a
n
(6)
is surjective, with kernel
a
i
=
_
a
i
.
Noetherian rings 13
PROOF. We have just proved the statement for n = 2. For each i , there exist elements
a
i
a
1
and b
i
a
i
such that
a
i
b
i
=1, all i _2.
The product
i_2
(a
i
b
i
) =1, and lies in a
1
i_2
a
i
, and so
a
1
i_2
a
i
=.
We can now apply the theorem in the case n =2 to obtain an element ,
1
of such that
,
1
1 mod a
1
. ,
1
0 mod
i_2
a
i
.
These conditions imply
,
1
1 mod a
1
. ,
1
0 mod a
}
, all >1.
Similarly, there exist elements ,
2
. .... ,
n
such that
,
i
1 mod a
i
. ,
i
0 mod a
}
for =i.
The element . =
.
i
,
i
maps to (.
1
moda
1
. . . . . .
n
moda
n
), which shows that (6) is sur
jective.
It remains to prove that
_
a
i
=
a
i
. Obviously
_
a
i
a
i
. We complete the proof
by induction. This allows us to assume that
i_2
a
i
=
_
i_2
a
i
. We showed above that a
1
and
i_2
a
i
are relatively prime, and so
a
1
(
i_2
a
i
) =a
1
(
i_2
a
i
)
by the n =2 case. Now a
1
(
i_2
a
i
) =
i_1
a
i
and
a
1
(
i_2
a
i
) =a
1
(
_
i_2
a
i
) =
_
i_1
a
i
.
which completes the proof.
2
Noetherian rings
PROPOSITION 1.2. The following three conditions on a ring are equivalent:
(a) every ideal in is nitely generated;
(b) every ascending chain of ideals a
1
a
2
eventually becomes constant, i.e., for
some m, a
n
=a
n1
= .
(c) every nonempty set of ideals in has a maximal element (i.e., an element not prop
erly contained in any other ideal in the set).
PROOF. (a) == (b): If a
1
a
2
is an ascending chain, then a =
_
a
i
is an ideal, and
hence has a nite set {a
1
. . . . . a
n
] of generators. For some m, all the a
i
belong a
n
and then
a
n
=a
n1
= =a.
14 1. PRELIMINARIES
(b) == (c): Let be a nonempty set of ideals in . If doesnt contain a maximal
element, then every ideal in is properly contained in another ideal in . Using this, we
get an innite chain of ideals a
1
_a
2
_ in ,
2
which violates (b).
(c) == (a): Let a be an ideal, and let be the set of nitely generated ideals contained
in a. Then is nonempty because it contains the zero ideal, and so it contains a maximal
element c =(a
1
. . . . . a
i
). If c =a, then there exists an element a ac, and (a
1
. . . . . a
i
. a)
will be a nitely generated ideal in a properly containing c. This contradicts the denition
of c.
2
A ring is noetherian if it satises the equivalent conditions of the proposition. On
applying (c) to the set of all proper ideals containing a xed proper ideal, we see that every
proper ideal in a noetherian ring is contained in a maximal ideal. This is, in fact, true for
every ring, but the proof for nonnoetherian rings requires Zorns lemma (CA 2.1).
A ring is said to be local if it has exactly one maximal ideal m. Because every nonunit
is contained in a maximal ideal, for a local ring
=m.
PROPOSITION 1.3 (NAKAYAMAS LEMMA). Let be a local ring with maximal ideal m,
and let M be a nitely generated module.
(a) If M =mM, then M =0.
(b) If N is a submodule of M such that M =N mM, then M =N.
PROOF. (a) Suppose that M =0. Choose a minimal set of generators {e
1
. . . . . e
n
], n _ 1,
for M, and write
e
1
=a
1
e
1
a
n
e
n
. a
i
m.
Then
(1a
1
)e
1
=a
2
e
2
a
n
e
n
and, as (1a
1
) is a unit, e
2
. . . . . e
n
generate M, contradicting the minimality of the set.
(b) The hypothesis implies that M,N =m(M,N), and so M,N =0.
2
Now let be a local noetherian ring with maximal ideal m. When we regard m as an
module, the action of on m,m
2
factors through k =,m.
COROLLARY 1.4. The elements a
1
. . . . . a
n
of m generate m as an ideal if and only if their
residues modulo m
2
generate m,m
2
as a vector space over k. In particular, the minimum
number of generators for the maximal ideal is equal to the dimension of the vector space
m,m
2
.
PROOF. If a
1
. . . . . a
n
generate m, it is obvious that their residues generate m,m
2
. Con
versely, suppose that their residues generate m,m
2
, so that m=(a
1
. . . . . a
n
)m
2
. Because
is noetherian, m is nitely generated, and Nakayamas lemma, applied with M =m and
N =(a
1
. . . . . a
n
), shows that m=(a
1
. . . . . a
n
).
2
DEFINITION 1.5. Let be a noetherian ring.
2
We are using the axiom of dependent choice, which is strictly weaker than the axiom of choice. It can
be stated as follows: let be a nonempty set, and let 1 be a subset of with the property that, for
each a , there exists a b such that (a. b) 1; then there exists a sequence (a
n
)
nN
in such that
(a
n
. a
n1
) 1 for every n N.
Noetherian rings 15
(a) The height ht(p) of a prime ideal p in is the greatest length J of a chain of distinct
prime ideals
p =p
d
p
d1
p
0
. (7)
(b) The Krull dimension of is sup{ht(p) [ p . p prime].
Thus, the Krull dimension of a ring is the supremum of the lengths of chains of
prime ideals in (the length of a chain is the number of gaps, so the length of (7) is J).
For example, a eld has Krull dimension 0, and conversely an integral domain of Krull
dimension 0 is a eld. The height of every nonzero prime ideal in a principal ideal domain
is 1, and so such a ring has Krull dimension 1 (provided it is not a eld).
The height of every prime ideal in a noetherian ring is nite, but the Krull dimension
of the ring may be innite because it may contain a sequence of prime ideals p
1
. p
2
. p
3
. . . .
such that ht(p
i
) tends to innity (see Nagata, Local Rings, 1962, Appendix A.1, p203).
DEFINITION 1.6. A local noetherian ring of Krull dimension J is said to be regular if its
maximal ideal can be generated by J elements.
It follows from Corollary 1.4 that a local noetherian ring is regular if and only if its
Krull dimension is equal to the dimension of the vector space m,m
2
.
LEMMA 1.7. In a noetherian ring, every set of generators for an ideal contains a nite
generating subset.
PROOF. Let a be an ideal in a noetherian ring , and let S be a set of generators for a.
An ideal maximal in the set of ideals generated by nite subsets of S must contain every
element of S (otherwise it wouldnt be maximal), and so equals a.
2
THEOREM 1.8 (KRULL INTERSECTION THEOREM). In every noetherian local ring with
maximal ideal m,
_
n_1
m
n
={0].
PROOF. Let a
1
. . . . . a
i
generate m. Then m
n
consists of all nite sums
i
1
i
r
=n
c
i
1
i
r
a
i
1
1
a
i
r
i
. c
i
1
i
r
.
In other words, m
n
consists of the elements of that equal g(a
1
. . . . . a
i
) for some ho
mogeneous polynomial g(X
1
. . . . . X
i
) X
1
. . . . . X
i
 of degree n. Let S
n
be the set of
homogeneous polynomials of degree m such that (a
1
. . . . . a
i
)
_
n_1
m
n
, and let a
be the ideal in X
1
. . . . . X
i
 generated by the set
_
n
S
n
. According to the lemma, there
exists a nite set {
1
. . . . .
x
] of elements of
_
n
S
n
that generates a. Let J
i
=deg
i
, and
let J = maxJ
i
. Let b
_
n_1
m
n
; then b m
d1
, and so b = (a
1
. . . . . a
i
) for some
homogeneous polynomial of degree J 1. By denition, S
d1
a, and so
=g
1
1
g
x
x
for some g
i
X
1
. . . . . X
n
. As and the
i
are homogeneous, we can omit from each g
i
all terms not of degree deg deg
i
, since these terms cancel out. Thus, we may choose
16 1. PRELIMINARIES
the g
i
to be homogeneous of degree deg deg
i
=J 1J
i
>0. Then g
i
(a
1
. . . . . a
i
)
m, and so
b =(a
1
. . . . . a
i
) =
i
g
i
(a
1
. . . . . a
i
)
i
(a
1
. . . . . a
i
) m
_
n_1
m
n
.
Thus,
_
m
n
=m
_
m
n
, and Nakayamas lemma implies that
_
m
n
=0.
2
Unique factorization
Let be an integral domain. An element a of is irreducible if it is not zero, not a unit,
and admits only trivial factorizations, i.e.,
a =bc == b or c is a unit.
The element a is said to be prime if (a) is a prime ideal, i.e.,
a[bc == a[b or a[c.
An integral domain is called a unique factorization domain if every nonzero nonunit
in can be written as a nite product of irreducible elements in exactly one way up to units
and the order of the factors. In such a ring, an irreducible element a can divide a product
bc only if it is an irreducible factor of b or c (write bc =aq and express b. c. q as products
of irreducible elements).
PROPOSITION 1.9. Let be an integral domain, and let a be an element of that is
neither zero nor a unit. If a is prime, then a is irreducible, and the converse holds when
is a unique factorization domain.
PROOF. Assume that a is prime. If a = bc then a divides bc, and so a divides b or c.
Suppose the rst, and write b =aq. Now a =bc =aqc, which implies that qc =1, and
that c is a unit. Therefore a is irreducible.
For the converse, assume that a is irreducible. If a[bc, then (as we noted above) a[b or
a[c, and so a is prime.
2
PROPOSITION 1.10 (GAUSSS LEMMA). Let be a unique factorization domain with
eld of fractions J. If (X) X factors into the product of two nonconstant poly
nomials in JX, then it factors into the product of two nonconstant polynomials in X.
PROOF. Let =gh in JX. For suitable c. J , the polynomials g
1
=cg and h
1
=Jh
have coefcients in , and so we have a factorization
cJ =g
1
h
1
in X.
If an irreducible element of divides cJ, then, looking modulo (), we see that
0 =g
1
h
1
in (,())X.
According to Proposition 1.9, () is prime, and so (,())X is an integral domain.
Therefore, divides all the coefcients of at least one of the polynomials g
1
. h
1
, say g
1
, so
that g
1
=g
2
for some g
2
X. Thus, we have a factorization
(cJ,) =g
2
h
1
in X.
Unique factorization 17
Continuing in this fashion, we can remove all the irreducible factors of cJ, and so obtain a
factorization of in X.
2
Let be a unique factorization domain. A nonzero polynomial
=a
0
a
1
X a
n
X
n
in X is said to be primitive if the coefcients a
i
have no common factor (other than
units). Every polynomial in X can be written = c( )
1
with c( ) and
1
primitive. The element c( ), welldened up to multiplication by a unit, is called the
content of .
LEMMA 1.11. The product of two primitive polynomials is primitive.
PROOF. Let
=a
0
a
1
X a
n
X
n
g =b
0
b
1
X b
n
X
n
.
be primitive polynomials, and let be an irreducible element of . Let a
i
0
be the rst
coefcient of not divisible by and b
}
0
the rst coefcient of g not divisible by . Then
all the terms in
i}=i
0
}
0
a
i
b
}
are divisible by , except a
i
0
b
}
0
, which is not divisible
by . Therefore, doesnt divide the (i
0
0
)thcoefcient of g. We have shown that
no irreducible element of divides all the coefcients of g, which must therefore be
primitive.
2
LEMMA 1.12. For polynomials . g X,
c(g) =c( ) c(g):
hence every factor in X of a primitive polynomial is primitive.
PROOF. Let =c( )
1
and g =c(g) g
1
with
1
and g
1
primitive. Then
g =c( ) c(g)
1
g
1
with
1
g
1
primitive, and so c(g) =c( )c(g).
2
PROPOSITION 1.13. If is a unique factorization domain, then so also is X.
PROOF. From the factorization =c( )
1
, we see that the irreducible elements of X
are to be found among the constant polynomials and the primitive polynomials, but a con
stant polynomial is irreducible if and only if a is an irreducible element of (obvious)
and a primitive polynomial is irreducible if and only if it has no primitive factor of lower
degree (by 1.12). From this it is clear that every nonzero nonunit in X is a product of
irreducible elements.
From the factorization =c( )
1
, we see that the irreducible elements of X are
to be found among the constant polynomials and the primitive polynomials.
18 1. PRELIMINARIES
Let
=c
1
c
n
1
n
=J
1
J
i
g
1
g
x
be two factorizations of an element of X into irreducible elements with the c
i
. J
}
constants and the
i
. g
}
primitive polynomials. Then
c( ) =c
1
c
n
=J
1
J
i
(up to units in ),
and, on using that is a unique factorization domain, we see that m=r and the c
i
s differ
from the J
i
s only by units and ordering. Hence,
1
n
=g
1
g
x
(up to units in ).
Gausss lemma shows that the
i
. g
}
are irreducible polynomials in JX and, on using
that JX is a unique factorization domain, we see that n =s and that the
i
s differ from
the g
i
s only by units in J and by their ordering. But if
i
=
o
b
g
}
with a and b nonzero
elements of , then b
i
=ag
}
. As
i
and g
}
are primitive, this implies that b =a (up to a
unit in ), and hence that
o
b
is a unit in .
2
Polynomial rings
Let k be a eld. A monomial in X
1
. . . . . X
n
is an expression of the form
X
o
1
1
X
o
n
n
. a
}
N.
The total degree of the monomial is
a
i
. We sometimes denote the monomial by X
.
=(a
1
. . . . . a
n
) N
n
.
The elements of the polynomial ring kX
1
. . . . . X
n
 are nite sums
c
o
1
o
n
X
o
1
1
X
o
n
n
. c
o
1
o
n
k. a
}
N.
with the obvious notions of equality, addition, and multiplication. In particular, the mono
mials form a basis for kX
1
. . . . . X
n
 as a kvector space.
The degree, deg( ), of a nonzero polynomial is the largest total degree of a mono
mial occurring in with nonzero coefcient. Since deg(g) =deg( )deg(g), kX
1
. . . . . X
n

is an integral domain and kX
1
. . . . . X
n

= k
. An element of kX
1
. . . . . X
n
 is irre
ducible if it is nonconstant and =gh == g or h is constant.
THEOREM 1.14. The ring kX
1
. . . . . X
n
 is a unique factorization domain.
PROOF. Note that kX
1
. . . . . X
n1
X
n
 =kX
1
. . . . . X
n
; this simply says that every poly
nomial in n variables X
1
. . . . . X
n
can be expressed uniquely as a polynomial in X
n
with
coefcients in kX
1
. . . . . X
n1
,
(X
1
. . . . . X
n
) =a
0
(X
1
. . . . . X
n1
)X
i
n
a
i
(X
1
. . . . . X
n1
).
Since k itself is a unique factorization domain (trivially), the theorem follows by induction
from Proposition 1.13.
2
COROLLARY 1.15. A nonzero proper principal ideal ( ) in kX
1
. . . . . X
n
 is prime if and
only if is irreducible.
PROOF. Special case of (1.9).
2
Integrality 19
Integrality
Let be an integral domain, and let 1 be a eld containing . An element of 1 is said
to be integral over if it is a root of a monic
3
polynomial with coefcients in , i.e., if it
satises an equation
n
a
1
n1
. . . a
n
=0. a
i
.
THEOREM 1.16. The set of elements of 1 integral over forms a ring.
PROOF. Let and integral over . Then there exists a monic polynomial
h(X) =X
n
c
1
X
n1
c
n
. c
i
.
having and among its roots (e.g., take h to be the product of the polynomials exhibiting
the integrality of and ). Write
h(X) =
n
i=1
(X ;
i
)
with the ;
i
in an algebraic closure of 1. Up to sign, the c
i
are the elementary symmetric
polynomials in the ;
i
(cf. FT p. 74). I claim that every symmetric polynomial in the ;
i
with coefcients in lies in : let
1
.
2
. . . . be the elementary symmetric polynomials in
X
1
. . . . . X
n
; if 1 X
1
. . . . . X
n
 is symmetric, then the symmetric polynomials theorem
(ibid. 5.33) shows that 1(X
1
. . . . . X
n
) =Q(
1
. . . . .
n
) for some Q X
1
. . . . . X
n
, and
so
1(;
1
. . . . . ;
n
) =Q(c
1
. c
2
. . . .) .
The coefcients of the polynomials
1_i,}_n
(X ;
i
;
}
) and
1_i,}_n
(X (;
i
;
}
))
are symmetric polynomials in the ;
i
with coefcients in , and therefore lie in . As the
polynomials are monic and have and among their roots, this shows that these
elements are integral.
2
For a less computational proof, see CA 5.3.
DEFINITION 1.17. The ring of elements of 1 integral over is called the integral closure
of in 1.
PROPOSITION 1.18. Let be an integral domain with eld of fractions J, and let 1 be a
eld containing J. If 1 is algebraic over J, then there exists a J such that J is
integral over .
PROOF. By assumption, satises an equation
n
a
1
n1
a
n
=0. a
i
J.
3
A polynomial is monic if its leading coefcient is 1, i.e., (X) =X
n
terms of degree <n.
20 1. PRELIMINARIES
Let J be a common denominator for the a
i
, so that Ja
i
, all i , and multiply through the
equation by J
n
:
J
n
n
a
1
J
n
n1
a
n
J
n
=0.
We can rewrite this as
(J)
n
a
1
J(J)
n1
a
n
J
n
=0.
As a
1
J. . . . . a
n
J
n
, this shows that J is integral over .
2
COROLLARY 1.19. Let be an integral domain and let 1 be an algebraic extension of the
eld of fractions of . Then 1 is the eld of fractions of the integral closure of in 1.
PROOF. The proposition shows that every 1 can be written =,J with integral
over and J .
2
DEFINITION 1.20. An integral domain is integrally closed if it is equal to its integral
closure in its eld of fractions J, i.e., if
J. integral over == .
PROPOSITION 1.21. Every unique factorization domain (e.g. a principal ideal domain) is
integrally closed.
PROOF. Let a,b, a. b , be integral over . If a,b , then there is an irreducible
element of dividing b but not a. As a,b is integral over , it satises an equation
(a,b)
n
a
1
(a,b)
n1
a
n
=0, a
i
.
On multiplying through by b
n
, we obtain the equation
a
n
a
1
a
n1
b a
n
b
n
=0.
The element then divides every term on the left except a
n
, and hence must divide a
n
.
Since it doesnt divide a, this is a contradiction.
2
PROPOSITION 1.22. Let be an integrally closed integral domain, and let 1 be a nite
extension of the eld of fractions J of . An element of 1 is integral over if and only
if its minimum polynomial over J has coefcients in .
PROOF. Let be integral over , so that
n
a
1
n1
a
n
=0. some a
i
.
Let
t
be a conjugate of , i.e., a root of the minimum polynomial (X) of over J. Then
there is an Jisomorphism
4
o: J J
t
. o() =
t
4
Recall (FT 1) that the homomorphism X : JX J denes an isomorphism JX,( )
J, where is the minimum polynomial of .
Direct limits (summary) 21
On applying o to the above equation we obtain the equation
tn
a
1
tn1
a
n
=0.
which shows that
t
is integral over . Hence all the conjugates of are integral over ,
and it follows from (1.16) that the coefcients of (X) are integral over . They lie in
J, and is integrally closed, and so they lie in . This proves the only if part of the
statement, and the if part is obvious.
2
COROLLARY 1.23. Let be an integrally closed integral domain with eld of fractions J,
and let (X) be a monic polynomial in X. Then every monic factor of (X) in JX
has coefcients in .
PROOF. It sufces to prove this for an irreducible monic factor g of in JX. Let be a
root of g in some extension eld of J. Then g is the minimum polynomial , which, being
also a root of , is integral. Therefore g has coefcients in .
2
Direct limits (summary)
DEFINITION 1.24. A partial ordering _on a set 1 is said to be directed, and the pair (1. _)
is called a directed set, if for all i. 1 there exists a k 1 such that i. _k.
DEFINITION 1.25. Let (1. _) be a directed set, and let 1 be a ring.
(a) An direct system of 1modules indexed by (1. _) is a family (M
i
)
iJ
of 1modules
together with a family (
i
}
: M
i
M
}
)
i_}
of 1linear maps such that
i
i
=id
i
and
}
k
i
}
=
i
k
all i _ _k.
(b) An 1module M together with a family (
i
: M
i
M)
iJ
of 1linear maps satisfy
ing
i
=
}
i
}
all i _ is said to be a direct limit of the system in (a) if it has the
following universal property: for any other 1module N and family (
i
: M
i
N)
of 1linear maps such that
i
=
}
i
}
all i _ , there exists a unique morphism
: M N such that
i
=
i
for i .
Clearly, the direct limit (if it exists), is uniquely determined by this condition up to a unique
isomorphism. We denote it lim
(M
i
.
}
i
), or just lim
M
i
.
Criterion
An 1module M together with 1linear maps
i
: M
i
M is the direct limit of a system
(M
i
.
}
i
) if and only if
(a) M =
_
iJ
i
(M
i
), and
(b) m
i
M
i
maps to zero in M if and only if it maps to zero in M
}
for some _i .
22 1. PRELIMINARIES
Construction
Let
M =
iJ
M
i
,M
t
where M
t
is the 1submodule generated by the elements
m
i
i
}
(m
i
) all i < , m
i
M
i
.
Let
i
(m
i
) =m
i
M
t
. Then certainly
i
=
}
i
}
for all i _ . For any 1module N
and 1linear maps
}
: M
}
N, there is a unique map
iJ
M
i
N.
namely,
m
i
i
(m
i
), sending m
i
to
i
(m
i
), and this map factors through M and is
the unique 1linear map with the required properties.
Direct limits of 1algebras, etc., are dened similarly.
Rings of fractions
A multiplicative subset of a ring is a subset S with the property:
1 S. a. b S == ab S.
Dene an equivalence relation on S by
(a. s) ~(b. t ) u(at bs) =0 for some u S.
Write
o
x
for the equivalence class containing (a. s), and dene addition and multiplication
of equivalence classes in the way suggested by the notation:
o
x
b
t
=
otbx
xt
.
o
x
b
t
=
ob
xt
.
It is easy to check that these do not depend on the choices of representatives for the equiv
alence classes, and that we obtain in this way a ring
S
1
=
_
a
s
[ a . s S
_
and a ring homomorphism a
o
1
: S
1
, whose kernel is
{a [ sa =0 for some s S].
For example, if is an integral domain an 0 S, then a
o
1
is injective, but if 0 S, then
S
1
is the zero ring.
Write i for the homomorphism a
o
1
: S
1
.
PROPOSITION 1.26. The pair (S
1
. i ) has the following universal property: every ele
ment s S maps to a unit in S
1
, and any other homomorphismT with this property
factors uniquely through i :
S
1
T.
i
3
Rings of fractions 23
PROOF. If exists,
s
o
x
=a == (s)(
o
x
) =(a) == (
o
x
) =(a)(s)
1
.
and so is unique. Dene
(
o
x
) =(a)(s)
1
.
Then
o
c
=
b
d
== s(aJ bc) =0 some s S == (a)(J) (b)(c) =0
because (s) is a unit in T, and so is welldened. It is obviously a homomorphism.
2
As usual, this universal property determines the pair (S
1
. i ) uniquely up to a unique
isomorphism.
When is an integral domain and S ={0], J =S
1
is the eld of fractions of
. In this case, for any other multiplicative subset T of not containing 0, the ring T
1
o
i
h
i
denes an iso
morphism
X,(1hX)
:
h
.
(b) For every multiplicative subset S of , S
1
. lim
h
, where h runs over the
elements of S (partially ordered by division).
5
First check m is an ideal. Next, if m =
p
, then 1 m; but if 1 =
o
x
for some a p and s p, then
u(s a) =0 some u p, and so ua =us p, which contradicts a p. Finally, m is maximal because every
element of
p
not in m is a unit.
24 1. PRELIMINARIES
PROOF. (a) If h =0, both rings are zero, and so we may assume h =0. In the ring . =
X,(1 hX), 1 = h., and so h is a unit. Let : T be a homomorphism of rings
such that (h) is a unit in T. The homomorphism
a
i
X
i
(a
i
)(h)
i
: X T
factors through . because 1hX 1(h)(h)
1
=0, and, because (h) is a unit in
T, this is the unique extension of to .. Therefore . has the same universal property
as
h
, and so the two are (uniquely) isomorphic by an isomorphism that xes elements of
and makes h
1
correspond to ..
(b) When h[h
t
, say, h
t
=hg, there is a canonical homomorphism
o
h
o
h
0
:
h
h
0 ,
and so the rings
h
form a direct system indexed by the set S. When h S, the homomor
phism S
1
extends uniquely to a homomorphism
o
h
o
h
:
h
S
1
(1.26), and
these homomorphisms are compatible with the maps in the direct system. Now apply the
criterion p21 to see that S
1
is the direct limit of the
h
.
2
Let S be a multiplicative subset of a ring , and let S
1
be the corresponding ring of
fractions. Any ideal a in , generates an ideal S
1
a in S
1
. If a contains an element of
S, then S
1
a contains a unit, and so is the whole ring. Thus some of the ideal structure of
is lost in the passage to S
1
, but, as the next lemma shows, some is retained.
PROPOSITION 1.30. Let S be a multiplicative subset of the ring . The map
p S
1
p =(S
1
)p
is a bijection from the set of prime ideals of disjoint from S to the set of prime ideals of
S
1
with inverse q (inverse image of q in ).
PROOF. For an ideal b of S
1
, let b
c
be the inverse image of b in , and for an ideal a of
, let a
e
=(S
1
)a be the ideal in S
1
generated by the image of a.
For an ideal b of S
1
, certainly, b b
ce
. Conversely, if
o
x
b, a , s S, then
o
1
b, and so a b
c
. Thus
o
x
b
ce
, and so b =b
ce
.
For an ideal a of , certainly a a
ec
. Conversely, if a a
ec
, then
o
1
a
e
, and so
o
1
=
o
0
x
for some a
t
a, s S. Thus, t (as a
t
) =0 for some t S, and so ast a. If a is a prime
ideal disjoint from S, this implies that a a: for such an ideal, a =a
ec
.
If b is prime, then certainly b
c
is prime. For any ideal a of , S
1
,a
e
.
S
1
(,a)
where
S is the image of S in ,a. If a is a prime ideal disjoint from S, then
S
1
(,a) is
a subring of the eld of fractions of ,a, and is therefore an integral domain. Thus, a
e
is
prime.
We have shown that p p
e
and q q
c
are inverse bijections between the prime ideals
of disjoint from S and the prime ideals of S
1
.
2
LEMMA 1.31. Let m be a maximal ideal of a noetherian integral domain , and let n =
m
m
. For all n, the map
am
n
an
n
: ,m
n
m
,n
n
is an isomorphism. Moreover, it induces isomorphisms
m
i
,m
n
n
i
,n
n
for all r <n.
Tensor Products 25
PROOF. The second statement follows from the rst, because of the exact commutative
diagram (r <n):
0 m
i
,m
n
,m
n
,m
i
0
_
:
_
:
0 n
i
,n
n
m
,n
n
m
,n
i
0.
Let S = m, so that
m
= S
1
. Because S contains no zero divisors, the map
a
o
1
:
m
is injective, and Ill identify with its image. In order to show that the
map ,m
n
n
,n
n
is injective, we have to show that n
n
=m
n
. But n
n
=m
n
m
=
S
1
m
n
, and so we have to show that m
n
=(S
1
m
n
) . An element of (S
1
m
n
)
can be written a =b,s with b m
n
, s S, and a . Then sa m
n
, and so sa =0 in
,m
n
. The only maximal ideal containing m
n
is m (because m
t
m
n
== m
t
m), and
so the only maximal ideal in ,m
n
is m,m
n
. As s is not in m,m
n
, it must be a unit in
,m
n
, and as sa =0 in ,m
n
, a must be 0 in ,m
n
, i.e., a m
n
.
We now prove that the map is surjective. Let
o
x
m
, a , s m. The only
maximal ideal of containing m
n
is m, and so no maximal ideal contains both s and m
n
;
it follows that (s)m
n
=. Therefore, there exist b and q m
n
such that sbq =1.
Because s is invertible in
m
,n
n
,
o
x
is the unique element of this ring such that s
o
x
= a;
since s(ba) =a(1q), the image of ba in
m
also has this property and therefore equals
o
x
.
2
PROPOSITION 1.32. In every noetherian ring, only 0 lies in all powers of all maximal
ideals.
PROOF. Let a be an element of a noetherian ring . If a =0, then {b [ ba =0] is a proper
ideal, and so is contained in some maximal ideal m. Then
o
1
is nonzero in
m
, and so
o
1
(m
m
)
n
for some n (by the Krull intersection theorem), which implies that a m
n
.
2
NOTES. For more on rings of fractions, see CA 6. In particular, 1.31 holds for all noetherian rings
(CA 6.7).
Tensor Products
Tensor products of modules
Let 1 be a ring. A map : MN 1 of 1modules is said to be 1bilinear if
(. .
t
. ,) =(.. ,) (.
t
. ,). .. .
t
M. , N
(.. , ,
t
) =(.. ,) (.. ,
t
). . M. ,. ,
t
N
(r.. ,) =r(.. ,). r 1. . M. , N
(.. r,) =r(.. ,). r 1. . M. , N.
i.e., if is 1linear in each variable. An 1module T together with an 1bilinear map
: M N T is called the tensor product of M and N over 1 if it has the following
26 1. PRELIMINARIES
universal property: every 1bilinear map
t
: MN T
t
factors uniquely through ,
MN T
T
t
.
0
3
As usual, the universal property determines the tensor product uniquely up to a unique
isomorphism. We write it M
T
N.
Construction Let M and N be 1modules, and let 1
(1)
be the free 1module with
basis MN. Thus each element 1
(1)
can be expressed uniquely as a nite sum
r
i
(.
i
. ,
i
). r
i
1. .
i
M. ,
i
N.
Let 1 be the submodule of 1
(1)
generated by the following elements
(. .
t
. ,) (.. ,) (.
t
. ,). .. .
t
M. , N
(.. , ,
t
) (.. ,) (.. ,
t
). . M. ,. ,
t
N
(r.. ,) r(.. ,). r 1. . M. , N
(.. r,) r(.. ,). r 1. . M. , N.
and dene
M
T
N =1
(1)
,1.
Write . , for the class of (.. ,) in M
T
N. Then
(.. ,) . ,: MN M
T
N
is 1bilinear we have imposed the fewest relations necessary to ensure this. Every
element of M
T
N can be written as a nite sum
r
i
(.
i
,
i
). r
i
1. .
i
M. ,
i
N.
and all relations among these symbols are generated by the following
(. .
t
) , =. , .
t
,
. (, ,
t
) =. , . ,
t
r(. ,) =(r.) , =. r,.
The pair (M
T
N. (.. ,) . ,) has the following universal property:
Tensor products of algebras
Let and T be kalgebras. A kalgebra C together with homomorphisms i : C and
: T C is called the tensor product of and T if it has the following universal property:
Tensor Products 27
for every pair of homomorphisms (of kalgebras) : 1 and : T 1, there is a unique
homomorphism ;: C 1 such that ; i = and ; =:
C T
1.
i
}
3
;
If it exists, the tensor product, is uniquely determined up to a unique isomorphism by this
property. We write it
k
T.
Construction Regard and T as kvector spaces, and form the tensor product
k
T.
There is a multiplication map
k
T
k
T
k
T for which
(ab)(a
t
b
t
) =aa
t
bb
t
.
This makes
k
T into a ring, and the homomorphism
c c(11) =c 1 =1c
makes it into a kalgebra. The maps
a a1: C and b 1b: T C
are homomorphisms, and they make
k
T into the tensor product of and T in the above
sense.
EXAMPLE 1.33. The algebra T, together with the given map k T and the identity map
T T, has the universal property characterizing k
k
T. In terms of the constructive
denition of tensor products, the map c b cb: k
k
T T is an isomorphism.
EXAMPLE 1.34. The ring kX
1
. . . . . X
n
. X
n1
. . . . . X
nn
, together with the obvious in
clusions
kX
1
. . . . . X
n
 kX
1
. . . . . X
nn
 kX
n1
. . . . . X
nn

is the tensor product of kX
1
. . . . . X
n
 and kX
n1
. . . . . X
nn
. To verify this we only have
to check that, for every kalgebra 1, the map
Hom
kalg
(kX
1
. . . . . X
nn
. 1) Hom
kalg
(kX
1
. . . .. 1) Hom
kalg
(kX
n1
. . . .. 1)
induced by the inclusions is a bijection. But this map can be identied with the bijection
1
nn
1
n
1
n
.
In terms of the constructive denition of tensor products, the map
g g: kX
1
. . . . . X
n

k
kX
n1
. . . . . X
nn
 kX
1
. . . . . X
nn

is an isomorphism.
28 1. PRELIMINARIES
REMARK 1.35. (a) If (b
k
kX
1
. . . . . X
n
 .1X
1
. . . . . 1X
n
 .X
1
. . . . . X
n
.
If =kX
1
. . . . . X
n
,(g
1
. . . . . g
n
), then
k
.X
1
. . . . . X
n
,(g
1
. . . . . g
n
).
(c) If and T are algebras of kvalued functions on sets S and T respectively, then
( g)(.. ,) =(.)g(,) realizes
k
T as an algebra of kvalued functions on S T .
For more details on tensor products, see CA 8.
Extension of scalars
Let 1 be a commutative ring and an 1algebra (not necessarily commutative) such that
the image of 1 lies in the centre of . Then we have a functor M
T
M from
left 1modules to left modules.
Behaviour with respect to direct limits
PROPOSITION 1.36. Direct limits commute with tensor products:
lim
iJ
M
i
T
lim
}J
N
}
. lim
(i,})JJ
(M
i
T
N
}
).
PROOF. Using the universal properties of direct limits and tensor products, one sees easily
that lim
(M
i
T
N
}
) has the universal property to be the tensor product of lim
M
i
and
lim
N
}
.
2
Flatness
For every 1module M, the functor N M
T
N is right exact, i.e.,
M
T
N
t
M
T
N M
T
N
tt
0
is exact whenever
N
t
N N
tt
0
is exact. If M
T
N M
T
N
t
is injective whenever N N
t
is injective, then M is
said to be at. Thus M is at if and only if the functor N M
T
N is exact. Similarly,
an 1algebra is at if N
T
N is at.
Categories and functors
A category C consists of
(a) a class of objects ob(C);
Categories and functors 29
(b) for each pair (a. b) of objects, a set Mor(a. b), whose elements are called morphisms
from a to b, and are written : a b;
(c) for each triple of objects (a. b. c) a map (called composition)
(. ) : Mor(a. b) Mor(b. c) Mor(a. c).
Composition is required to be associative, i.e., (; ) =; (), and for each object
a there is required to be an element id
o
Mor(a. a) such that id
o
=, id
o
=, for
all and for which these composites are dened. The sets Mor(a. b) are required to be
disjoint (so that a morphism determines its source and target).
EXAMPLE 1.37. (a) There is a category of sets, Sets, whose objects are the sets and whose
morphisms are the usual maps of sets.
(b) There is a category Aff
k
of afne kalgebras, whose objects are the afne kalgebras
and whose morphisms are the homomorphisms of kalgebras.
(c) In Chapter 4 below, we dene a category Var
k
of algebraic varieties over k, whose
objects are the algebraic varieties over k and whose morphisms are the regular maps.
The objects in a category need not be sets with structure, and the morphisms need not
be maps.
Let C and D be categories. A covariant functor J from C to D consists of
(a) a map a J(a) sending each object of C to an object of D, and,
(b) for each pair of objects a. b of C, a map
J(): Mor(a. b) Mor(J(a). J(b))
such that J(id
) =id
T()
and J() =J() J().
A contravariant functor is dened similarly, except that the map on morphisms is
J(): Mor(a. b) Mor(J(b). J(a))
A functor J: C D is full (resp. faithful, fully faithful) if, for all objects a and b of
C, the map
Mor(a. b) Mor(J(a). J(b))
is a surjective (resp. injective, bijective).
A covariant functor J: A B of categories is said to be an equivalence of categories
if it is fully faithful and every object of B is isomorphic to an object of the form J(a),
a ob(A) (J is essentially surjective). One can show that such a functor J has a quasi
inverse, i.e., that there is a functor G: B A, which is also an equivalence, and for which
there exist natural isomorphisms G(J(a)) ~a and J(G(b)) ~b.
6
6
For each object b of B, choose an object G(b) of A and an isomorphismJ(G(b)) b. For each morphism
: b b
t
in B, let G() be the unique morphism such that
J(G(b)) b
_T()
_
J(G(b
t
)) b
t
commutes. Then G is a quasiinverse to J.
30 1. PRELIMINARIES
Similarly one denes the notion of a contravariant functor being an equivalence of cat
egories.
Any fully faithful functor J: C D denes an equivalence of C with the full subcat
egory of D whose objects are isomorphic to J(a) for some object a of C. The essential
image of a fully faithful functor J: C D consists of the objects of D isomorphic to an
object of the form J(a), a ob(C).
Let J and G be two functors CD. A morphism (or natural transformation) : J
G is a collection of morphisms (a): J(a) G(a), one for each object a of C, such that,
for every morphism u: a b in C, the following diagram commutes:
a
_
u
b
J(a)
(o)
G(a)
_
T(u)
_
G(u)
J(b)
(b)
G(b).
(**)
With this notion of morphism, the functors C D form a category Fun(C. D) (provided
that we ignore the problem that Mor(J. G) may not be a set, but only a class).
For any object V of a category C, we have a contravariant functor
h
V
: C Set
that sends an object a to the set Mor(a. V ) and sends a morphism : a b to the map
: h
V
(b) h
V
(a).
i.e., h
V
(+) = Mor(+. V ) and h
V
() = + . Let : V W be a morphism in C. The
collection of maps
h
(a): h
V
(a) h
W
(a).
is a morphism of functors.
PROPOSITION 1.38 (YONEDA LEMMA). For any contravariant functor J: C Set,
Mor(h
V
. J) .J(V ).
In particular, when J =h
W
,
Mor(h
V
. h
W
) .Mor(V. W)
to give a natural transformation h
V
h
W
is the same as to give a morphism V W.
PROOF. An element . of J(V ) denes a natural transformation h
V
J, namely,
J()(.): h
V
(T ) J(T ). h
V
(T ) =Hom(T. V ).
Conversely, a natural transformation h
V
J denes an element of J(V ), namely, the
image of the universal element id
V
under h
V
(V ) J(V ). It is easy to check that these
two maps are inverse and are natural in both J and V .
2
Algorithms for polynomials 31
Algorithms for polynomials
As an introduction to algorithmic algebraic geometry, we derive some algorithms for working with
polynomial rings. This section is little more than a summary of the rst two chapters of Cox et
al.1992 to which I refer the reader for more details. Those not interested in algorithms can skip the
remainder of this chapter. Throughout, k is a eld (not necessarily algebraically closed).
The two main results will be:
(a) An algorithmic proof of the Hilbert basis theorem: every ideal in kX
1
. . . . . X
n
 has a nite
set of generators (in fact, of a special kind).
(b) There exists an algorithm for deciding whether a polynomial belongs to an ideal.
Division in kX
The division algorithm allows us to divide a nonzero polynomial into another: let and g be
polynomials in kX with g = 0; then there exist unique polynomials q. r kX such that =
qg r with either r = 0 or degr < degg. Moreover, there is an algorithm for deciding whether
(g), namely, nd r and check whether it is zero.
In Maple,
quo(f. g. X): computes q
rem(f. g. X): computes r
Moreover, the Euclidean algorithm allows you to pass from a nite set of generators for an ideal
in kX to a single generator by successively replacing each pair of generators with their greatest
common divisor.
Orderings on monomials
Before we can describe an algorithm for dividing in kX
1
. . . . . X
n
, we shall need to choose a way
of ordering monomials. Essentially this amounts to dening an ordering on N
n
. There are two main
systems, the rst of which is preferred by humans, and the second by machines.
(Pure) lexicographic ordering (lex). Here monomials are ordered by lexicographic (dictionary)
order. More precisely, let =(a
1
. . . . . a
n
) and =(b
1
. . . . . b
n
) be two elements of N
n
; then
> and X
>X
(lexicographic ordering)
if, in the vector difference (an element of Z
n
), the leftmost nonzero entry is positive. For
example,
XY
2
>Y
3
7
4
: X
3
Y
2
7
4
>X
3
Y
2
7.
Note that this isnt quite how the dictionary would order them: it would put XXXYYZZZZ after
XXXYYZ.
Graded reverse lexicographic order (grevlex). Here monomials are ordered by total degree,
with ties broken by reverse lexicographic ordering. Thus, > if
a
i
>
b
i
, or
a
i
=
b
i
and in the rightmost nonzero entry is negative. For example:
X
4
Y
4
7
T
>X
5
Y
5
7
4
(total degree greater)
XY
5
7
2
>X
4
Y7
3
. X
5
Y7 >X
4
Y7
2
.
Orderings on kX
1
. . . . . X
n

Fix an ordering on the monomials in kX
1
. . . . . X
n
. Then we can write an element of kX
1
. . . . . X
n

in a canonical fashion by reordering its elements in decreasing order. For example, we would write
=4XY
2
747
2
5X
3
7X
2
7
2
32 1. PRELIMINARIES
as
=5X
3
7X
2
7
2
4XY
2
747
2
(lex)
or
=4XY
2
77X
2
7
2
5X
3
47
2
(grevlex)
Let =
kX
1
. . . . . X
n
. Write it in decreasing order:
=a
0
X
0
a
1
X
1
.
0
>
1
> . a
0
=0.
Then we dene:
(a) the multidegree of to be multdeg( ) =
0
;
(b) the leading coefcient of to be LC( ) =a
0
;
(c) the leading monomial of to be LM( ) =X
0
;
(d) the leading term of to be LT( ) =a
0
X
0
.
For example, for the polynomial =4XY
2
7 , the multidegree is (1. 2. 1), the leading coef
cient is 4, the leading monomial is XY
2
7, and the leading term is 4XY
2
7.
The division algorithm in kX
1
. . . . . X
n

Fix a monomial ordering in N
n
. Suppose given a polynomial and an ordered set (g
1
. . . . . g
x
) of
polynomials; the division algorithm then constructs polynomials a
1
. . . . . a
x
and r such that
=a
1
g
1
a
x
g
x
r
where either r =0 or no monomial in r is divisible by any of LT(g
1
). . . . . LT(g
x
).
STEP 1: If LT(g
1
)[LT( ), divide g
1
into to get
=a
1
g
1
h. a
1
=
LT( )
LT(g
1
)
kX
1
. . . . . X
n
.
If LT(g
1
)[LT(h), repeat the process until
=a
1
g
1
1
(different a
1
) with LT(
1
) not divisible by LT(g
1
). Now divide g
2
into
1
, and so on, until
=a
1
g
1
a
x
g
x
r
1
with LT(r
1
) not divisible by any of LT(g
1
). . . . . LT(g
x
).
STEP 2: Rewrite r
1
=LT(r
1
) r
2
, and repeat Step 1 with r
2
for :
=a
1
g
1
a
x
g
x
LT(r
1
) r
3
(different a
i
s).
STEP 3: Rewrite r
3
=LT(r
3
) r
4
, and repeat Step 1 with r
4
for :
=a
1
g
1
a
x
g
x
LT(r
1
) LT(r
3
) r
3
(different a
i
s).
Continue until you achieve a remainder with the required property. In more detail,
7
after di
viding through once by g
1
. . . . . g
x
, you repeat the process until no leading term of one of the g
i
s
divides the leading term of the remainder. Then you discard the leading term of the remainder, and
repeat.
7
This differs from the algorithm in Cox et al. 1992, p63, which says to go back to g
1
after every successful
division.
Algorithms for polynomials 33
EXAMPLE 1.39. (a) Consider
=X
2
Y XY
2
Y
2
. g
1
=XY 1. g
2
=Y
2
1.
First, on dividing g
1
into , we obtain
X
2
Y XY
2
Y
2
=(X Y )(XY 1) X Y
2
Y.
This completes the rst step, because the leading term of Y
2
1 does not divide the leading term of
the remainder X Y
2
Y . We discard X, and write
Y
2
Y =1 (Y
2
1) Y 1.
Altogether
X
2
Y XY
2
Y
2
=(X Y ) (XY 1) 1 (Y
2
1) X Y 1.
(b) Consider the same polynomials, but with a different order for the divisors
=X
2
Y XY
2
Y
2
. g
1
=Y
2
1. g
2
=XY 1.
In the rst step,
X
2
Y XY
2
Y
2
=(X 1) (Y
2
1) X (XY 1) 2X 1.
Thus, in this case, the remainder is 2X 1.
REMARK 1.40. If r =0, then (g
1
. . . . . g
x
), but, because the remainder depends on the ordering
of the g
i
, the converse is false. For example, (lex ordering)
XY
2
X =Y (XY 1) 0 (Y
2
1) X Y
but
XY
2
X =X (Y
2
1) 0 (XY 1) 0.
Thus, the division algorithm (as stated) will not provide a test for lying in the ideal generated by
g
1
. . . . . g
x
.
Monomial ideals
In general, an ideal a can contain a polynomial without containing the individual monomials of the
polynomial; for example, the ideal a =(Y
2
X
3
) contains Y
2
X
3
but not Y
2
or X
3
.
DEFINITION 1.41. An ideal a is monomial if
a and c
=0 == X
a.
PROPOSITION 1.42. Let a be a monomial ideal, and let = { [ X
, . Conversely, if is a subset of
N
n
satisfying (*), then the ksubspace a of kX
1
. . . . . X
n
 generated by {X
[ ] is a monomial
ideal.
34 1. PRELIMINARIES
PROOF. It is clear from its denition that a monomial ideal a is the ksubspace of kX
1
. . . . . X
n

generated by the set of monomials it contains. If X
a and X
kX
1
. . . . . X
n
, then X
=
X
_
_
_
N
n
J
_
_
=
,
c
(nite sums).
and so if satises (*), then the subspace generated by the monomials X
, , is an ideal.
2
The proposition gives a classication of the monomial ideals in kX
1
. . . . . X
n
: they are in one
toone correspondence with the subsets of N
n
satisfying (*). For example, the monomial ideals in
kX are exactly the ideals (X
n
), n _ 0, and the zero ideal (corresponding to the empty set ). We
write
(X
[ )
for the ideal corresponding to (subspace generated by the X
, ).
LEMMA 1.43. Let S be a subset of N
n
. Then the ideal a generated by {X
[ S] is the monomial
ideal corresponding to
def
= { N
n
[ S. some S].
In other words, a monomial is in a if and only if it is divisible by one of the X
, S.
PROOF. Clearly satises (*), and a (X
[ ). Conversely, if , then N
n
for
some S, and X
=X
X

a. The last statement follows from the fact that X
[X
N
n
.
2
Let N
2
satisfy (*). From the geometry of , it is clear that there is a nite set of elements
S ={
1
. . . . .
x
] of such that
={ N
2
[
i
N
2
. some
i
S].
(The
i
s are the corners of .) Moreover, the ideal (X
i
,
i
S. This suggests the following result.
THEOREM 1.44 (DICKSONS LEMMA). Let a be the monomial ideal corresponding to the subset
N
n
. Then a is generated by a nite subset of {X
[ ].
PROOF. This is proved by induction on the number of variables Cox et al. 1992, p70.
2
Hilbert Basis Theorem
DEFINITION 1.45. For a nonzero ideal a in kX
1
. . . . . X
n
, we let (LT(a)) be the ideal generated by
{LT( ) [ a].
LEMMA 1.46. Let a be a nonzero ideal in kX
1
. . . . . X
n
; then (LT(a)) is a monomial ideal, and it
equals (LT(g
1
). . . . . LT(g
n
)) for some g
1
. . . . . g
n
a.
PROOF. Since (LT(a)) can also be described as the ideal generated by the leading monomials (rather
than the leading terms) of elements of a, it follows from Lemma 1.43 that it is monomial. Now
Dicksons Lemma shows that it equals (LT(g
1
). . . . . LT(g
x
)) for some g
i
a.
2
THEOREM 1.47 (HILBERT BASIS THEOREM). Every ideal a in kX
1
. . . . . X
n
 is nitely gener
ated; in fact, a is generated by any elements of a whose leading terms generate LT(a).
Algorithms for polynomials 35
PROOF. Let g
1
. . . . . g
n
be as in the lemma, and let a. On applying the division algorithm, we
nd
=a
1
g
1
a
x
g
x
r. a
i
. r kX
1
. . . . . X
n
.
where either r =0 or no monomial occurring in it is divisible by any LT(g
i
). But r =
a
i
g
i
a, and therefore LT(r) LT(a) =(LT(g
1
). . . . . LT(g
x
)), which, according to Lemma 1.43, implies
that every monomial occurring in r is divisible by one in LT(g
i
). Thus r =0, and g (g
1
. . . . . g
x
).
2
Standard (Gr obner) bases
Fix a monomial ordering of kX
1
. . . . . X
n
.
DEFINITION 1.48. A nite subset S ={g
1
. . . . . g
x
] of an ideal a is a standard (Grobner, Groebner,
Gr obner) basis
8
for a if
(LT(g
1
). . . . . LT(g
x
)) =LT(a).
In other words, S is a standard basis if the leading term of every element of a is divisible by at least
one of the leading terms of the g
i
.
THEOREM 1.49. Every ideal has a standard basis, and it generates the ideal; if {g
1
. . . . . g
x
] is a
standard basis for an ideal a, then a the remainder on division by the g
i
is 0.
PROOF. Our proof of the Hilbert basis theorem shows that every ideal has a standard basis, and
that it generates the ideal. Let a. The argument in the same proof, that the remainder of on
division by g
1
. . . . . g
x
is 0, used only that {g
1
. . . . . g
x
] is a standard basis for a.
2
REMARK 1.50. The proposition shows that, for a, the remainder of on division by {g
1
. . . . . g
x
]
is independent of the order of the g
i
(in fact, its always zero). This is not true if a see the
example using Maple at the end of this chapter.
Let a = (
1
. . . . .
x
). Typically, {
1
. . . . .
x
] will fail to be a standard basis because in some
expression
cX
i
JX
}
. c. J k. (**)
the leading terms will cancel, and we will get a new leading term not in the ideal generated by the
leading terms of the
i
. For example,
X
2
=X (X
2
Y X 2Y
2
) Y (X
3
2XY )
is in the ideal generated by X
2
Y X 2Y
2
and X
3
2XY but it is not in the ideal generated by
their leading terms.
There is an algorithm for transforming a set of generators for an ideal into a standard basis,
which, roughly speaking, makes adroit use of equations of the form (**) to construct enough new
elements to make a standard basis see Cox et al. 1992, pp8087.
We now have an algorithm for deciding whether (
1
. . . . .
i
). First transform {
1
. . . . .
i
]
into a standard basis {g
1
. . . . . g
x
]. and then divide by g
1
. . . . . g
x
to see whether the remainder is
0 (in which case lies in the ideal) or nonzero (and it doesnt). This algorithm is implemented in
Maple see below.
A standard basis {g
1
. . . . . g
x
] is minimal if each g
i
has leading coefcient 1 and, for all i , the
leading term of g
i
does not belong to the ideal generated by the leading terms of the remaining
gs. A standard basis {g
1
. . . . . g
x
] is reduced if each g
i
has leading coefcient 1 and if, for all i , no
monomial of g
i
lies in the ideal generated by the leading terms of the remaining gs. One can prove
(Cox et al. 1992, p91) that every nonzero ideal has a unique reduced standard basis.
8
Standard bases were rst introduced (under that name) by Hironaka in the mid1960s, and independently,
but slightly later, by Buchberger in his Ph.D. thesis. Buchberger named them after his thesis adviser Gr obner.
36 1. PRELIMINARIES
REMARK 1.51. Consider polynomials . g
1
. . . . . g
x
kX
1
. . . . . X
n
. The algorithm that replaces
g
1
. . . . . g
x
with a standard basis works entirely within kX
1
. . . . . X
n
, i.e., it doesnt require a eld
extension. Likewise, the division algorithm doesnt require a eld extension. Because these opera
tions give welldened answers, whether we carry them out in kX
1
. . . . . X
n
 or in 1X
1
. . . . . X
n
,
1 k, we get the same answer. Maple appears to work in the subeld of C generated over by all
the constants occurring in the polynomials.
We conclude this chapter with the annotated transcript of a session in Maple applying the above
algorithm to show that
q =3.
3
,:
2
.:
2
,
3
,:
doesnt lie in the ideal
(.
2
2.: 5. .,
2
,:
3
. 3,
2
8:
3
).
A Maple session
>with(grobner):
This loads the grobner package, and lists the available commands:
finduni, finite, gbasis, gsolve, leadmon, normalf, solvable, spoly
To discover the syntax of a command, a brief description of the command, and an example, type
?command;
>G:=gbasis([x^22*x*z+5,x*y^2+y*z^3,3*y^28*z^3],[x,y,z]);
G :=.
2
2.: 5. 3,
2
8:
3
. 8.,
2
3,
3
. 9,
4
48:,
3
320,
2

This asks Maple to nd the reduced Grobner basis for the ideal generated by the three polynomials
listed, with respect to the symbols listed (in that order). It will automatically use grevlex order unless
you add ,plex to the command.
> q:=3*x^3*y*z^2  x*z^2 + y^3 + y*z;
q :=3.
3
,:
2
.:
2
,
3
:,
This denes the polynomial q.
> normalf(q,G,[x,y,z]);
9:
2
,
3
15,:
2
.
41
4
,
3
60,
2
: .:
2
:,
Asks for the remainder when q is divided by the polynomials listed in G using the symbols listed.
This particular example is amusing the program gives different orderings for G, and different
answers for the remainder, depending on which computer I use. This is O.K., because, since q isnt
in the ideal, the remainder may depend on the ordering of G.
Notes:
(a) To start Maple on a Unix computer type maple; to quit type quit.
(b) Maple wont do anything until you type ; or : at the end of a line.
(c) The student version of Maple is quite cheap, but unfortunately, it doesnt have the Grobner
package.
(d) For more information on Maple:
i) There is a brief discussion of the Grobner package in Cox et al. 1992, Appendix C, 1.
ii) The Maple V Library Reference Manual pp469478 briey describes what the Grobner
package does (exactly the same information is available on line, by typing ?command).
iii) There are many books containing general introductions to Maple syntax.
(e) Gr obner bases are also implemented in Macsyma, Mathematica, and Axiom, but for serious
work it is better to use one of the programs especially designed for Gr obner basis computa
tion, namely,
CoCoA (Computations in Commutative Algebra) http://cocoa.dima.unige.it/.
Macaulay 2 (Grayson and Stillman) http://www.math.uiuc.edu/Macaulay2/.
Exercises 37
Exercises
11. Let k be an innite eld (not necessarily algebraically closed). Show that an
kX
1
. . . . . X
n
 that is identically zero on k
n
is the zero polynomial (i.e., has all its coef
cients zero).
12. Find a minimal set of generators for the ideal
(X 2Y. 3X 6Y 37. 2X 4Y 37)
in kX. Y. 7. What standard algorithm in linear algebra will allow you to answer this
question for any ideal generated by homogeneous linear polynomials? Find a minimal set
of generators for the ideal
(X 2Y 1. 3X 6Y 3X 2. 2X 4Y 373).
CHAPTER 2
Algebraic Sets
In this chapter, k is an algebraically closed eld.
Denition of an algebraic set
An algebraic subset V(S) of k
n
is the set of common zeros of some set S of polynomials
in kX
1
. . . . . X
n
:
V(S) ={(a
1
. . . . . a
n
) k
n
[ (a
1
. . . . . a
n
) =0 all (X
1
. . . . . X
n
) S].
Note that
S S
t
== V(S) V(S
t
):
more equations means fewer solutions.
Recall that the ideal a generated by a set S consists of all nite sums
i
g
i
.
i
kX
1
. . . . . X
n
. g
i
S.
Such a sum
i
g
i
is zero at every point at which the g
i
are all zero, and so V(S) V(a),
but the reverse conclusion is also true because S a. Thus V(S) = V(a) the zero set
of S is the same as that of the ideal generated by S. Hence the algebraic sets can also be
described as the sets of the form V(a), a an ideal in kX
1
. . . . . X
n
.
EXAMPLE 2.1. (a) If S is a system of homogeneous linear equations, then V(S) is a sub
space of k
n
. If S is a system of nonhomogeneous linear equations, then V(S) is either
empty or is the translate of a subspace of k
n
.
(b) If S consists of the single equation
Y
2
=X
3
aX b. 4a
3
27b
2
=0.
then V(S) is an elliptic curve. For more on elliptic curves, and their relation to Fermats last
theorem, see my notes on Elliptic Curves. The reader should sketch the curve for particular
values of a and b. We generally visualize algebraic sets as though the eld k were 1,
although this can be misleading.
(c) For the empty set 0, V(0) =k
n
.
(d) The algebraic subsets of k are the nite subsets (including 0) and k itself.
39
40 2. ALGEBRAIC SETS
(e) Some generating sets for an ideal will be more useful than others for determining
what the algebraic set is. For example, a Gr obner basis for the ideal
a =(X
2
Y
2
7
2
1. X
2
Y
2
Y. X 7)
is (according to Maple)
X 7. Y
2
2Y 1. 7
2
1Y.
The middle polynomial has (double) root 1, and it follows easily that V(a) consists of the
single point (0. 1. 0).
The Hilbert basis theorem
In our denition of an algebraic set, we didnt require the set S of polynomials to be nite,
but the Hilbert basis theorem shows that every algebraic set will also be the zero set of a 
nite set of polynomials. More precisely, the theorem shows that every ideal in kX
1
. . . . . X
n

can be generated by a nite set of elements, and we have already observed that any set of
generators of an ideal has the same zero set as the ideal.
We sketched an algorithmic proof of the Hilbert basis theorem in the last chapter. Here
we give the slick proof.
THEOREM 2.2 (HILBERT BASIS THEOREM). The ring kX
1
. . . . . X
n
 is noetherian, i.e.,
every ideal is nitely generated.
Since k itself is noetherian, and kX
1
. . . . . X
n1
X
n
 = kX
1
. . . . . X
n
, the theorem
follows by induction from the next lemma.
LEMMA 2.3. If is noetherian, then so also is X.
PROOF. Recall that for a polynomial
(X) =a
0
X
i
a
1
X
i1
a
i
. a
i
. a
0
=0.
r is called the degree of , and a
0
is its leading coefcient.
Let a be a proper ideal in X, and let a
i
be the set of elements of that occur as the
leading coefcient of a polynomial in a of degree _i . Then a
i
is an ideal in , and
a
1
a
2
a
i
.
Because is noetherian, this sequence eventually becomes constant, say a
d
=a
d1
=. . .
(and a
d
consists of the leading coefcients of all polynomials in a).
For each i _J, choose a nite set
i1
.
i2
. . . . of polynomials in a of degree i such that
the leading coefcients a
i}
of the
i}
s generate a
i
.
Let a; we shall prove by induction on the degree of that it lies in the ideal
generated by the
i}
. When has degree 0, it is zero, and so lies in (
i}
).
Suppose that has degree s _J. Then =aX
x
with a a
d
, and so
a =
}
b
}
a
d}
. some b
}
.
The Zariski topology 41
Now
}
b
}
d}
X
xd
has degree <deg( ), and so lies in (
i}
) by induction.
Suppose that has degree s _J. Then a similar argument shows that
b
}
x}
has degree <deg( ) for suitable b
}
, and so lies in (
i}
) by induction.
2
ASIDE 2.4. One may ask how many elements are needed to generate a given ideal a in
kX
1
. . . . . X
n
, or, what is not quite the same thing, how many equations are needed to
dene a given algebraic set V . When n = 1, we know that every ideal is generated by a
single element. Also, if V is a linear subspace of k
n
, then linear algebra shows that it is the
zero set of ndim(V ) polynomials. All one can say in general, is that at least ndim(V )
polynomials are needed to dene V (see 9.7), but often more are required. Determining
exactly how many is an area of active research see (9.14).
The Zariski topology
PROPOSITION 2.5. There are the following relations:
(a) a b == V(a) V(b):
(b) V(0) =k
n
; V(kX
1
. . . . . X
n
) =0:
(c) V(ab) =V(ab) =V(a) LV(b):
(d) V(
iJ
a
i
) =
_
iJ
V(a
i
) for every family of ideals (a
i
)
iJ
.
PROOF. The rst two statements are obvious. For (c), note that
ab ab a. b == V(ab) V(ab) V(a) LV(b).
For the reverse inclusions, observe that if a V(a) LV(b), then there exist a, g b
such that (a) = 0, g(a) = 0; but then (g)(a) = 0, and so a V(ab). For (d) recall
that, by denition,
a
i
consists of all nite sums of the form
i
,
i
a
i
. Thus (d) is
obvious.
2
Statements (b), (c), and (d) show that the algebraic subsets of k
n
satisfy the axioms to be
the closed subsets for a topology on k
n
: both the whole space and the empty set are closed;
a nite union of closed sets is closed; an arbitrary intersection of closed sets is closed. This
topology is called the Zariski topology on k
n
. The induced topology on a subset V of k
n
is called the Zariski topology on V .
The Zariski topology has many strange properties, but it is nevertheless of great impor
tance. For the Zariski topology on k, the closed subsets are just the nite sets and the whole
space, and so the topology is not Hausdorff. We shall see in (2.29) below that the proper
closed subsets of k
2
are nite unions of (isolated) points and curves (zero sets of irreducible
kX. Y ). Note that the Zariski topologies on C and C
2
are much coarser (have many
fewer open sets) than the complex topologies.
42 2. ALGEBRAIC SETS
The Hilbert Nullstellensatz
We wish to examine the relation between the algebraic subsets of k
n
and the ideals of
kX
1
. . . . . X
n
, but rst we consider the question of when a set of polynomials has a common
zero, i.e., when the equations
g(X
1
. . . . . X
n
) =0. g a.
are consistent. Obviously, the equations
g
i
(X
1
. . . . . X
n
) =0. i =1. . . . . m
are inconsistent if there exist
i
kX
1
. . . . . X
n
 such that
i
g
i
=1, i.e., if 1 (g
1
. . . . . g
n
)
or, equivalently, (g
1
. . . . . g
n
) = kX
1
. . . . . X
n
. The next theorem provides a converse to
this.
THEOREM 2.6 (HILBERT NULLSTELLENSATZ).
1
Every proper ideal a in kX
1
. . . . . X
n

has a zero in k
n
.
A point 1 =(a
1
. . . . . a
n
) in k
n
denes a homomorphism evaluate at 1
kX
1
. . . . . X
n
 k. (X
1
. . . . . X
n
) (a
1
. . . . . a
n
).
whose kernel contains a if 1 V(a). Conversely, froma homomorphism: kX
1
. . . . . X
n

k of kalgebras whose kernel contains a, we obtain a point 1 in V(a), namely,
1 =((X
1
). . . . . (X
n
)).
Thus, to prove the theorem, we have to show that there exists a kalgebra homomorphism
kX
1
. . . . . X
n
,a k.
Since every proper ideal is contained in a maximal ideal, it sufces to prove this for
a maximal ideal m. Then 1
def
= kX
1
. . . . . X
n
,m is a eld, and it is nitely generated as
an algebra over k (with generators X
1
m. . . . . X
n
m). To complete the proof, we must
show 1 =k. The next lemma accomplishes this.
Although we shall apply the lemma only in the case that k is algebraically closed, in
order to make the induction in its proof work, we need to allowarbitrary ks in the statement.
LEMMA 2.7 (ZARISKIS LEMMA). Let k 1 be elds (k is not necessarily algebraically
closed). If 1 is nitely generated as an algebra over k, then 1 is algebraic over k. (Hence
1 =k if k is algebraically closed.)
PROOF. We shall prove this by induction on r, the minimum number of elements required
to generate 1 as a kalgebra. The case r = 0 being trivial, we may suppose that 1 =
k.
1
. . . . . .
i
 with r _ 1. If 1 is not algebraic over k, then at least one .
i
, say .
1
, is not
algebraic over k. Then, k.
1
 is a polynomial ring in one symbol over k, and its eld of
fractions k(.
1
) is a subeld of 1. Clearly 1 is generated as a k(.
1
)algebra by .
2
. . . . . .
i
,
and so the induction hypothesis implies that .
2
. . . . . .
i
are algebraic over k(.
1
). According
to (1.18), there exists a J k.
1
 such that J.
i
is integral over k.
1
 for all i _ 2. Let
1 = k.
1
. . . . . .
i
. For a sufciently large N, J
1
k.
1
. J.
2
. . . . . J.
i
, and so
1
Nullstellensatz = zeropointstheorem.
The correspondence between algebraic sets and ideals 43
J
1
is integral over k.
1
 (1.16). When we apply this statement to an element of k(.
1
),
(1.21) shows that J
1
k.
1
. Therefore, k(.
1
) =
_
1
J
1
k.
1
, but this is absurd,
because k.
1
 (. kX) has innitely many distinct monic irreducible polynomials
2
that
can occur as denominators of elements of k(.
1
).
2
The correspondence between algebraic sets and ideals
For a subset W of k
n
, we write 1(W) for the set of polynomials that are zero on W:
1(W) ={ kX
1
. . . . . X
n
 [ (1) =0 all 1 W].
Clearly, it is an ideal in kX
1
. . . . . X
n
. There are the following relations:
(a) V W == 1(V ) 1(W):
(b) 1(0) =kX
1
. . . . . X
n
; 1(k
n
) =0:
(c) 1(
_
W
i
) =
_
1(W
i
).
Only the statement 1(k
n
) = 0 is (perhaps) not obvious. It says that, if a polynomial is
nonzero (in the ring kX
1
. . . . . X
n
), then it is nonzero at some point of k
n
. This is true
with k any innite eld (see Exercise 11). Alternatively, it follows from the strong Hilbert
Nullstellensatz (cf. 2.14a below).
EXAMPLE 2.8. Let 1 be the point (a
1
. . . . . a
n
). Clearly 1(1) (X
1
a
1
. . . . . X
n
a
n
),
but (X
1
a
1
. . . . . X
n
a
n
) is a maximal ideal, because evaluation at (a
1
. . . . . a
n
) denes
an isomorphism
kX
1
. . . . . X
n
,(X
1
a
1
. . . . . X
n
a
n
) k.
As 1(1) is a proper ideal, it must equal (X
1
a
1
. . . . . X
n
a
n
).
PROPOSITION 2.9. For every subset W k
n
, V1(W) is the smallest algebraic subset of
k
n
containing W. In particular, V1(W) =W if W is an algebraic set.
PROOF. Let V be an algebraic set containing W, and write V = V(a). Then a 1(W),
and so V(a) V1(W).
2
The radical rad(a) of an ideal a is dened to be
{ [
i
a, some r N, r >0].
PROPOSITION 2.10. Let a be an ideal in a ring .
(a) The radical of a is an ideal.
(b) rad(rad(a)) =rad(a).
PROOF. (a) If a rad(a), then clearly a rad(a) for all . Suppose a. b rad(a),
with say a
i
a and b
x
a. When we expand (a b)
ix
using the binomial theorem, we
nd that every term has a factor a
i
or b
x
, and so lies in a.
(b) If a
i
rad(a), then a
ix
=(a
i
)
x
a for some s.
2
2
If k is innite, then consider the polynomials Xa, and if k is nite, consider the minimum polynomials
of generators of the extension elds of k. Alternatively, and better, adapt Euclids proof that there are innitely
many prime numbers.
44 2. ALGEBRAIC SETS
An ideal is said to be radical if it equals its radical, i.e., if
i
a == a. Equiva
lently, a is radical if and only if ,a is a reduced ring, i.e., a ring without nonzero nilpotent
elements (elements some power of which is zero). Since integral domains are reduced,
prime ideals (a fortiori maximal ideals) are radical.
If a and b are radical, then ab is radical, but ab need not be: consider, for example,
a = (X
2
Y ) and b = (X
2
Y ); they are both prime ideals in kX. Y , but X
2
a b,
X ab.
As
i
(1) =(1)
i
,
i
is zero wherever is zero, and so 1(W) is radical. In partic
ular, 1V(a) rad(a). The next theorem states that these two ideals are equal.
THEOREM 2.11 (STRONG HILBERT NULLSTELLENSATZ). For every ideal a in kX
1
. . . . . X
n
,
1V(a) is the radical of a; in particular, 1V(a) =a if a is a radical ideal.
PROOF. We have already noted that 1V(a) rad(a). For the reverse inclusion, we have
to show that if h is identically zero on V(a), then h
1
a for some N > 0; here h
kX
1
. . . . . X
n
. We may assume h = 0. Let g
1
. . . . . g
n
generate a, and consider the sys
tem of m1 equations in n1 variables, X
1
. . . . . X
n
. Y.
_
g
i
(X
1
. . . . . X
n
) = 0. i =1. . . . . m
1Yh(X
1
. . . . . X
n
) = 0.
If (a
1
. . . . . a
n
. b) satises the rst m equations, then (a
1
. . . . . a
n
) V(a); consequently,
h(a
1
. . . . . a
n
) = 0, and (a
1
. . . . . a
n
. b) doesnt satisfy the last equation. Therefore, the
equations are inconsistent, and so, according to the original Nullstellensatz, there exist
i
kX
1
. . . . . X
n
. Y  such that
1 =
n
i=1
i
g
i
n1
(1Yh)
(in the ring kX
1
. . . . . X
n
. Y ). On applying the homomorphism
_
X
i
X
i
Y h
1
: kX
1
. . . . . X
n
. Y  k(X
1
. . . . . X
n
)
to the above equality, we obtain the identity
1 =
n
i=1
i
(X
1
. . . . . X
n
. h
1
) g
i
(X
1
. . . . . X
n
) (*)
in k(X
1
. . . . . X
n
). Clearly
i
(X
1
. . . . . X
n
. h
1
) =
polynomial in X
1
. . . . . X
n
h
1
i
for some N
i
. Let N be the largest of the N
i
. On multiplying (*) by h
1
we obtain an
equation
h
1
=
(polynomial in X
1
. . . . . X
n
) g
i
(X
1
. . . . . X
n
).
which shows that h
1
a.
2
The correspondence between algebraic sets and ideals 45
COROLLARY 2.12. The map a V(a) denes a onetoone correspondence between the
set of radical ideals in kX
1
. . . . . X
n
 and the set of algebraic subsets of k
n
; its inverse is 1.
PROOF. We know that 1V(a) =a if a is a radical ideal (2.11), and that V1(W) =W if W
is an algebraic set (2.9). Therefore, 1 and V are inverse maps.
2
COROLLARY 2.13. The radical of an ideal in kX
1
. . . . . X
n
 is equal to the intersection of
the maximal ideals containing it.
PROOF. Let a be an ideal in kX
1
. . . . . X
n
. Because maximal ideals are radical, every
maximal ideal containing a also contains rad(a):
rad(a)
_
ma
m.
For each 1 = (a
1
. . . . . a
n
) k
n
, m
1
= (X
1
a
1
. . . . . X
n
a
n
) is a maximal ideal in
kX
1
. . . . . X
n
, and
m
1
(1) =0
(see 2.8). Thus
m
1
a 1 V(a).
If m
1
for all 1 V(a), then is zero on V(a), and so 1V(a) =rad(a). We have
shown that
rad(a)
_
1V(a)
m
1
.
2
REMARK 2.14. (a) Because V(0) =k
n
,
1(k
n
) =1V(0) =rad(0) =0:
in other words, only the zero polynomial is zero on the whole of k
n
.
(b) The onetoone correspondence in the corollary is order inverting. Therefore the
maximal proper radical ideals correspond to the minimal nonempty algebraic sets. But
the maximal proper radical ideals are simply the maximal ideals in kX
1
. . . . . X
n
, and the
minimal nonempty algebraic sets are the onepoint sets. As
1((a
1
. . . . . a
n
)) =(X
1
a
1
. . . . . X
n
a
n
)
(see 2.8), this shows that the maximal ideals of kX
1
. . . . . X
n
 are exactly the ideals of the
form (X
1
a
1
. . . . . X
n
a
n
).
(c) The algebraic set V(a) is empty if and only if a =kX
1
. . . . . X
n
, because
V(a) =0 =rad(a) =kX
1
. . . . . X
n
 =1 rad(a) =1 a.
(d) Let W and W
t
be algebraic sets. Then W W
t
is the largest algebraic subset con
tained in both W and W
t
, and so 1(W W
t
) must be the smallest radical ideal containing
both 1(W) and 1(W
t
). Hence 1(W W
t
) =rad(1(W) 1(W
t
)).
For example, let W =V(X
2
Y ) and W
t
=V(X
2
Y ); then 1(W W
t
) =rad(X
2
. Y ) =
(X. Y ) (assuming characteristic =2). Note that W W
t
={(0. 0)], but when realized as
the intersection of Y =X
2
and Y =X
2
, it has multiplicity 2. [The reader should draw
a picture.]
46 2. ALGEBRAIC SETS
ASIDE 2.15. Let 1 be the set of subsets of k
n
and let Qbe the set of subsets of kX
1
. . . . . X
n
.
Then 1: 1 Q and V : Q 1 dene a simple Galois correspondence (cf. FT 7.17).
Therefore, 1 and V dene a onetoone correspondence between 11 and VQ. But the
strong Nullstellensatz shows that 11 consists exactly of the radical ideals, and (by deni
tion) VQ consists of the algebraic subsets. Thus we recover Corollary 2.12.
Finding the radical of an ideal
Typically, an algebraic set V will be dened by a nite set of polynomials {g
1
. . . . . g
x
], and
then we shall need to nd 1(V ) =rad((g
1
. . . . . g
x
)).
PROPOSITION 2.16. The polynomial h rad(a) if and only if 1 (a. 1Yh) (the ideal in
kX
1
. . . . . X
n
. Y  generated by the elements of a and 1Yh).
PROOF. We saw that 1 (a. 1 Yh) implies h rad(a) in the course of proving (2.11).
Conversely, if h
1
a, then
1 =Y
1
h
1
(1Y
1
h
1
)
=Y
1
h
1
(1Yh) (1Yh Y
11
h
11
)
a(1Yh).
2
Since we have an algorithm for deciding whether or not a polynomial belongs to an
ideal given a set of generators for the ideal see Section 1 we also have an algorithm
deciding whether or not a polynomial belongs to the radical of the ideal, but not yet an
algorithm for nding a set of generators for the radical. There do exist such algorithms
(see Cox et al. 1992, p177 for references), and one has been implemented in the computer
algebra system Macaulay 2 (see p36).
The Zariski topology on an algebraic set
We now examine more closely the Zariski topology on k
n
and on an algebraic subset of
k
n
. Proposition 2.9 says that, for each subset W of k
n
, V1(W) is the closure of W, and
(2.12) says that there is a onetoone correspondence between the closed subsets of k
n
and
the radical ideals of kX
1
. . . . . X
n
. Under this correspondence, the closed subsets of an
algebraic set V correspond to the radical ideals of kX
1
. . . . . X
n
 containing 1(V ).
PROPOSITION 2.17. Let V be an algebraic subset of k
n
.
(a) The points of V are closed for the Zariski topology (thus V is a T
1
space).
(b) Every ascending chain of open subsets U
1
U
2
of V eventually becomes
constant, i.e., for some m, U
n
= U
n1
= ; hence every descending chain of closed
subsets of V eventually becomes constant.
(c) Every open covering of V has a nite subcovering.
PROOF. (a) Clearly {(a
1
. . . . . a
n
)] is the algebraic set dened by the ideal (X
1
a
1
. . . . . X
n
a
n
).
(b) A sequence V
1
V
2
of closed subsets of V gives rise to a sequence of radical
ideals 1(V
1
) 1(V
2
) . . ., which eventually becomes constant because kX
1
. . . . . X
n
 is
noetherian.
The coordinate ring of an algebraic set 47
(c) Let V =
_
iJ
U
i
with each U
i
open. Choose an i
0
1; if U
i
0
= V , then there
exists an i
1
1 such that U
i
0
U
i
0
LU
i
1
. If U
i
0
LU
i
1
=V , then there exists an i
2
1 etc..
Because of (b), this process must eventually stop.
2
A topological space having the property (b) is said to be noetherian. The condition
is equivalent to the following: every nonempty set of closed subsets of V has a minimal
element. A space having property (c) is said to be quasicompact (by Bourbaki at least;
others call it compact, but Bourbaki requires a compact space to be Hausdorff). The proof
of (c) shows that every noetherian space is quasicompact. Since an open subspace of a
noetherian space is again noetherian, it will also be quasicompact.
The coordinate ring of an algebraic set
Let V be an algebraic subset of k
n
, and let 1(V ) =a. The coordinate ring of V is
kV  =kX
1
. . . . . X
n
,a.
This is a nitely generated reduced kalgebra (because a is radical), but it need not be an
integral domain.
A function V k of the form 1 (1) for some kX
1
. . . . . X
n
 is said to
be regular.
3
Two polynomials . g kX
1
. . . . . X
n
 dene the same regular function on
V if only if they dene the same element of kV . The coordinate function .
i
: V k,
(a
1
. . . . . a
n
) a
i
is regular, and kV  .k.
1
. . . . . .
n
.
For an ideal b in kV , set
V(b) ={1 V [ (1) =0, all b].
Let W =V(b). The maps
kX
1
. . . . . X
n
 kV  =
kX
1
. . . . . X
n

a
kW =
kV 
b
send a regular function on k
n
to its restriction to V , and then to its restriction to W.
Write for the map kX
1
. . . . . X
n
 kV . Then b
1
(b) is a bijection from the
set of ideals of kV  to the set of ideals of kX
1
. . . . . X
n
 containing a, under which radical,
prime, and maximal ideals correspond to radical, prime, and maximal ideals (each of these
conditions can be checked on the quotient ring, and kX
1
. . . . . X
n
,
1
(b) . kV ,b).
Clearly
V(
1
(b)) =V(b).
and so b V(b) is a bijection from the set of radical ideals in kV  to the set of algebraic
sets contained in V .
For h kV , set
D(h) ={a V [ h(a) =0].
It is an open subset of V , because it is the complement of V((h)), and it is empty if and
only if h is zero (2.14a).
3
In the next chapter, well give a more general denition of regular function according to which these are
exactly the regular functions on V , and so kV  will be the ring of regular functions on V .
48 2. ALGEBRAIC SETS
PROPOSITION 2.18. (a) The points of V are in onetoone correspondence with the maxi
mal ideals of kV .
(b) The closed subsets of V are in onetoone correspondence with the radical ideals of
kV .
(c) The sets D(h), h kV , are a base for the topology on V , i.e., each D(h) is open,
and every open set is a union (in fact, a nite union) of D(h)s.
PROOF. (a) and (b) are obvious from the above discussion. For (c), we have already ob
served that D(h) is open. Any other open set U V is the complement of a set of the form
V(b), with b an ideal in kV , and if
1
. . . . .
n
generate b, then U =
_
D(
i
).
2
The D(h) are called the basic (or principal) open subsets of V . We sometimes write
V
h
for D(h). Note that
D(h) D(h
t
) V(h) V(h
t
)
rad((h)) rad((h
t
))
h
i
(h
t
) some r
h
i
=h
t
g, some g.
Some of this should look familiar: if V is a topological space, then the zero set of a
family of continuous functions : V 1 is closed, and the set where such a function is
nonzero is open.
Irreducible algebraic sets
A nonempty topological space is said to be irreducible if it is not the union of two proper
closed subsets; equivalently, if any two nonempty open subsets have a nonempty intersec
tion, or if every nonempty open subset is dense. By convention, the empty space is not
irreducible.
If an irreducible space W is a nite union of closed subsets, W =W
1
L. . . LW
i
, then
W =W
1
or W
2
L. . . LW
i
; if the latter, then W =W
2
or W
3
L. . . LW
i
, etc.. Continuing in
this fashion, we nd that W =W
i
for some i .
The notion of irreducibility is not useful for Hausdorff topological spaces, because the
only irreducible Hausdorff spaces are those consisting of a single point two points would
have disjoint open neighbourhoods contradicting the second condition.
PROPOSITION 2.19. An algebraic set W is irreducible and only if 1(W) is prime.
PROOF. ==: Suppose g 1(W). At each point of W, either is zero or g is zero, and
so W V( ) LV(g). Hence
W =(W V( )) L(W V(g)).
As W is irreducible, one of these sets, say W V( ), must equal W. But then 1(W).
This shows that 1(W) is prime.
==: Suppose W =V(a) LV(b) with a and b radical ideals we have to show that
W equals V(a) or V(b). Recall (2.5) that V(a) LV(b) =V(ab) and that ab is radical;
hence 1(W) =ab. If W =V(a), then there is an a1(W). For all g b,
g ab =1(W).
Irreducible algebraic sets 49
Because 1(W) is prime, this implies that b 1(W); therefore W V(b).
2
Thus, there are onetoone correspondences
radical ideals  algebraic subsets
prime ideals  irreducible algebraic subsets
maximal ideals  onepoint sets.
These correspondences are valid whether we mean ideals in kX
1
. . . . . X
n
 and algebraic
subsets of k
n
, or ideals in kV  and algebraic subsets of V . Note that the last correspon
dence implies that the maximal ideals in kV  are those of the form (.
1
a
1
. . . . . .
n
a
n
),
(a
1
. . . . . a
n
) V .
EXAMPLE 2.20. Let kX
1
. . . . . X
n
. As we showed in (1.14), kX
1
. . . . . X
n
 is a
unique factorization domain, and so ( ) is a prime ideal if and only if is irreducible
(1.15). Thus
is irreducible == V( ) is irreducible.
On the other hand, suppose factors,
=
n
i
i
.
i
distinct irreducible polynomials.
Then
( ) =
_
(
n
i
i
). (
n
i
i
) distinct primary
4
ideals,
rad(( )) =
_
(
i
). (
i
) distinct prime ideals,
V( ) =
_
V(
i
). V(
i
) distinct irreducible algebraic sets.
PROPOSITION 2.21. Let V be a noetherian topological space. Then V is a nite union
of irreducible closed subsets, V = V
1
L. . . LV
n
. Moreover, if the decomposition is irre
dundant in the sense that there are no inclusions among the V
i
, then the V
i
are uniquely
determined up to order.
PROOF. Suppose that V can not be written as a nite union of irreducible closed subsets.
Then, because V is noetherian, there will be a closed subset W of V that is minimal among
those that cannot be written in this way. But W itself cannot be irreducible, and so W =
W
1
LW
2
, with each W
i
a proper closed subset of W. From the minimality of W, we deduce
that each W
i
is a nite union of irreducible closed subsets, and so therefore is W. We have
arrived at a contradiction.
Suppose that
V =V
1
L. . . LV
n
=W
1
L. . . LW
n
are two irredundant decompositions. Then V
i
=
_
}
(V
i
W
}
), and so, because V
i
is ir
reducible, V
i
= V
i
W
}
for some . Consequently, there is a function : {1. . . . . m]
{1. . . . . n] such that V
i
W
((i)
for each i . Similarly, there is a function g: {1. . . . . n]
{1. . . . . m] such that W
}
V
(})
for each . Since V
i
W
((i)
V
((i)
, we must have
g(i ) =i and V
i
=W
((i)
; similarly g =id. Thus and g are bijections, and the decom
positions differ only in the numbering of the sets.
2
3
In a noetherian ring , a proper ideal q is said to primary if every zerodivisor in ,q is nilpotent.
50 2. ALGEBRAIC SETS
The V
i
given uniquely by the proposition are called the irreducible components of V .
They are the maximal closed irreducible subsets of V . In Example 2.20, the V(
i
) are the
irreducible components of V( ).
COROLLARY 2.22. A radical ideal a in kX
1
. . . . . X
n
 is a nite intersection of prime ide
als, a = p
1
. . . p
n
; if there are no inclusions among the p
i
, then the p
i
are uniquely
determined up to order.
PROOF. Write V(a) as a union of its irreducible components, V(a) =
_
V
i
, and take p
i
=
1(V
i
).
2
REMARK 2.23. (a) An irreducible topological space is connected, but a connected topo
logical space need not be irreducible. For example, V(X
1
X
2
) is the union of the coor
dinate axes in k
2
, which is connected but not irreducible. An algebraic subset V of k
n
is not connected if and only if there exist ideals a and b such that a b = 1(V ) and
ab =kX
1
. . . . . X
n
.
(b) A Hausdorff space is noetherian if and only if it is nite, in which case its irreducible
components are the onepoint sets.
(c) In kX, ((X)) is radical if and only if is squarefree, in which case is a
product of distinct irreducible polynomials, =
1
. . .
i
, and ( ) = (
1
) . . . (
i
) (a
polynomial is divisible by if and only if it is divisible by each
i
).
(d) In a noetherian ring, every proper ideal a has a decomposition into primary ideals:
a =
_
q
i
(see CA 14). For radical ideals, this becomes a simpler decomposition into
prime ideals, as in the corollary. For an ideal ( ) with =
n
i
i
, it is the decomposition
( ) =
_
(
n
i
i
) noted in Example 2.20.
Dimension
We briey introduce the notion of the dimension of an algebraic set. In Chapter 9 we shall
discuss this in more detail.
Let V be an irreducible algebraic subset. Then 1(V ) is a prime ideal, and so kV  is
an integral domain. Let k(V ) be its eld of fractions k(V ) is called the eld of rational
functions on V . The dimension of V is dened to be the transcendence degree of k(V )
over k (see FT, Chapter 8).
5
EXAMPLE 2.24. (a) Let V =k
n
; then k(V ) =k(X
1
. . . . . X
n
), and so dim(V ) =n.
(b) If V is a linear subspace of k
n
(or a translate of such a subspace), then it is an easy
exercise to show that the dimension of V in the above sense is the same as its dimension in
the sense of linear algebra (in fact, kV  is canonically isomorphic to kX
i
1
. . . . . X
i
d
 where
the X
i
j
are the free variables in the system of linear equations dening V see 5.12).
In linear algebra, we justify saying V has dimension n by proving that its elements
are parametrized by ntuples. It is not true in general that the points of an algebraic set
of dimension n are parametrized by ntuples. The most one can say is that there exists a
nitetoone map to k
n
(see 8.12).
(c) An irreducible algebraic set has dimension 0 if and only if it consists of a single
point. Certainly, for any point 1 k
n
, k1 =k, and so k(1) =k. Conversely, suppose
5
According to CA, 13.8, the transcendence degree of k(V ) is equal to the Krull dimension of kV ; cf.
2.30 below.
Dimension 51
V = V(p), p prime, has dimension 0. Then k(V ) is an algebraic extension of k, and so
equals k. From the inclusions
k kV  k(V ) =k
we see that kV  = k. Hence p is maximal, and we saw in (2.14b) that this implies that
V(p) is a point.
The zero set of a single nonconstant nonzero polynomial (X
1
. . . . . X
n
) is called a
hypersurface in k
n
.
PROPOSITION 2.25. An irreducible hypersurface in k
n
has dimension n1.
PROOF. An irreducible hypersurface is the zero set of an irreducible polynomial (see
2.20). Let
k.
1
. . . . . .
n
 =kX
1
. . . . . X
n
,( ). .
i
=X
i
p.
and let k(.
1
. . . . . .
n
) be the eld of fractions of k.
1
. . . . . .
n
. Since is not zero, some X
i
,
say, X
n
, occurs in it. Then X
n
occurs in every nonzero multiple of , and so no nonzero
polynomial in X
1
. . . . . X
n1
belongs to ( ). This means that .
1
. . . . . .
n1
are algebraically
independent. On the other hand, .
n
is algebraic over k(.
1
. . . . . .
n1
), and so {.
1
. . . . . .
n1
]
is a transcendence basis for k(.
1
. . . . . .
n
) over k.
2
For a reducible algebraic set V , we dene the dimension of V to be the maximum of
the dimensions of its irreducible components. When the irreducible components all have
the same dimension J, we say that V has pure dimension J.
PROPOSITION 2.26. If V is irreducible and 7 is a proper algebraic subset of V , then
dim(7) < dim(V ).
PROOF. We may assume that 7 is irreducible. Then 7 corresponds to a nonzero prime
ideal p in kV , and k7 =kV ,p.
Write
kV  =kX
1
. . . . . X
n
,1(V ) =k.
1
. . . . . .
n
.
Let kV . The image
of in kV ,p = k7 is the restriction of to 7. With
this notation, k7 = k .
1
. . . . . .
n
. Suppose that dim7 = J and that the X
i
have been
numbered so that .
1
. . . . . .
d
are algebraically independent (see FT 8.9) for the proof that
this is possible). I will show that, for any nonzero p, the J 1 elements .
1
. . . . . .
d
.
are algebraically independent, which implies that dimV _J 1.
Suppose otherwise. Then there is a nontrivial algebraic relation among the .
i
and ,
which we can write
a
0
(.
1
. . . . . .
d
)
n
a
1
(.
1
. . . . . .
d
)
n1
a
n
(.
1
. . . . . .
d
) =0.
with a
i
(.
1
. . . . . .
d
) k.
1
. . . . . .
d
 and not all zero. Because V is irreducible, kV  is an
integral domain, and so we can cancel a power of if necessary to make a
n
(.
1
. . . . . .
d
)
nonzero. On restricting the functions in the above equality to 7, i.e., applying the homo
morphism kV  k7, we nd that
a
n
( .
1
. . . . . .
d
) =0.
which contradicts the algebraic independence of .
1
. . . . . .
d
.
2
52 2. ALGEBRAIC SETS
PROPOSITION 2.27. Let V be an irreducible variety such that kV  is a unique factor
ization domain (for example, V = A
d
). If W V is a closed subvariety of dimension
dimV 1, then 1(W) =( ) for some kV .
PROOF. We know that 1(W) =
_
1(W
i
) where the W
i
are the irreducible components of
W, and so if we can prove 1(W
i
) = (
i
) then 1(W) = (
1
i
). Thus we may suppose
that W is irreducible. Let p =1(W); it is a prime ideal, and it is nonzero because otherwise
dim(W) = dim(V ). Therefore it contains an irreducible polynomial . From (1.15) we
know ( ) is prime. If ( ) =p , then we have
W =V(p) _V(( )) _V.
and dim(W) < dim(V( )) < dimV (see 2.26), which contradicts the hypothesis.
2
EXAMPLE 2.28. Let J(X. Y ) and G(X. Y ) be nonconstant polynomials with no common
factor. Then V(J(X. Y )) has dimension 1 by (2.25), and so V(J(X. Y )) V(G(X. Y ))
must have dimension zero; it is therefore a nite set.
EXAMPLE 2.29. We classify the irreducible closed subsets V of k
2
. If V has dimension
2, then (by 2.26) it cant be a proper subset of k
2
, so it is k
2
. If V has dimension 1,
then V =k
2
, and so 1(V ) contains a nonzero polynomial, and hence a nonzero irreducible
polynomial (being a prime ideal). Then V V( ), and so equals V( ). Finally, if V
has dimension zero, it is a point. Correspondingly, we can make a list of all the prime ideals
in kX. Y : they have the form (0), ( ) (with irreducible), or (X a. Y b).
ASIDE 2.30. Later (9.4) we shall show that if, in the situation of (2.26), 7 is a maximal
proper irreducible subset of V , then dim7 =dimV 1. This implies that the dimension of
an algebraic set V is the maximum length of a chain
V
0
V
1
V
d
with each V
i
closed and irreducible and V
0
an irreducible component of V . Note that
this description of dimension is purely topological it makes sense for every noetherian
topological space.
On translating the description in terms of ideals, we see immediately that the dimension
of V is equal to the Krull dimension of kV the maximal length of a chain of prime
ideals,
p
d
p
d1
p
0
.
Exercises
21. Find 1(W), where V =(X
2
. XY
2
). Check that it is the radical of (X
2
. XY
2
).
22. Identify k
n
2
with the set of mm matrices. Show that, for all r, the set of matrices
with rank _r is an algebraic subset of k
n
2
.
23. Let V = {(t. . . . . t
n
) [ t k]. Show that V is an algebraic subset of k
n
, and that
kV  ~kX (polynomial ring in one variable). (Assume k has characteristic zero.)
Exercises 53
24. Using only that kX. Y  is a unique factorization domain and the results of 1,2,
show that the following is a complete list of prime ideals in kX. Y :
(a) (0);
(b) ((X. Y )) for an irreducible polynomial;
(c) (X a. Y b) for a. b k.
25. Let and T be (not necessarily commutative) algebras of nite dimension over ,
and let
al
be the algebraic closure of in C. Show that if Hom
Calgebras
(
C. T
al
algebras
(
al
. T
al
) =0. (Hint: The proof takes only a few
lines.)
CHAPTER 3
Afne Algebraic Varieties
In this chapter, we dene the structure of a ringed space on an algebraic set, and then we
dene the notion of afne algebraic variety roughly speaking, this is an algebraic set with
no preferred embedding into k
n
. This is in preparation for 4, where we dene an algebraic
variety to be a ringed space that is a nite union of afne algebraic varieties satisfying a
natural separation axiom.
Ringed spaces
Let V be a topological space and k a eld.
DEFINITION 3.1. Suppose that for every open subset U of V we have a set O
V
(U) of
functions U k. Then O
V
is called a sheaf of kalgebras if it satises the following
conditions:
(a) O
V
(U) is a ksubalgebra of the algebra of all kvalued functions on U, i.e., O
V
(U)
contains the constant functions and, if . g lie in O
V
(U), then so also do g and
g.
(b) If U
t
is an open subset of U and O
V
(U), then [U
t
O
V
(U
t
).
(c) A function : U k on an open subset U of V is in O
V
(U) if [U
i
O
V
(U
i
) for
all U
i
in some open covering of U.
Conditions (b) and (c) require that a function on U lies in O
V
(U) if and only if each
point 1 of U has a neighborhood U
1
such that [U
1
lies in O
V
(U
1
); in other words, the
condition for to lie in O
V
(U) is local.
EXAMPLE 3.2. (a) Let V be any topological space, and for each open subset U of V let
O
V
(U) be the set of all continuous realvalued functions on U. Then O
V
is a sheaf of
1algebras.
(b) Recall that a function : U 1, where U is an open subset of 1
n
, is said to
be smooth (or innitely differentiable) if its partial derivatives of all orders exist and are
continuous. Let V be an open subset of 1
n
, and for each open subset U of V let O
V
(U)
be the set of all smooth functions on U. Then O
V
is a sheaf of 1algebras.
(c) Recall that a function : U C, where U is an open subset of C
n
, is said to be
analytic (or holomorphic) if it is described by a convergent power series in a neighbourhood
55
56 3. AFFINE ALGEBRAIC VARIETIES
of each point of U. Let V be an open subset of C
n
, and for each open subset U of V let
O
V
(U) be the set of all analytic functions on U. Then O
V
is a sheaf of Calgebras.
(d) Nonexample: let V be a topological space, and for each open subset U of V let
O
V
(U) be the set of all realvalued constant functions on U; then O
V
is not a sheaf, unless
V is irreducible!
1
When constant is replaced with locally constant, O
V
becomes a
sheaf of 1algebras (in fact, the smallest such sheaf).
A pair (V. O
V
) consisting of a topological space V and a sheaf of kalgebras will be
called a ringed space. For historical reasons, we often write 1(U. O
V
) for O
V
(U) and call
its elements sections of O
V
over U.
Let (V. O
V
) be a ringed space. For every open subset U of V , the restriction O
V
[U of
O
V
to U, dened by
1(U
t
. O
V
[U) =1(U
t
. O
V
), all open U
t
U.
is a sheaf again.
Let (V. O
V
) be ringed space, and let 1 V . Consider pairs (. U) consisting of an open
neighbourhood U of 1 and an O
V
(U). We write (. U) ~(
t
. U
t
) if [U
tt
=
t
[U
tt
for some open neighbourhood U
tt
of 1 contained in U and U
t
. This is an equivalence
relation, and an equivalence class of pairs is called a germ of a function at 1 (relative to
O
V
). The set of equivalence classes of such pairs forms a kalgebra denoted O
V,1
or O
1
.
In all the interesting cases, it is a local ring with maximal ideal the set of germs that are
zero at 1.
In a fancier terminology,
O
1
=lim
O
V
(U). (direct limit over open neighbourhoods U of 1).
A germ of a function at 1 is dened by a function on a neigbourhood of 1 (section of
O
V
), and two such functions dene the same germ if and only if they agree in a possibly
smaller neighbourhood of 1.
EXAMPLE 3.3. Let O
V
be the sheaf of holomorphic functions on V =C, and let c C. A
power series
n_0
a
n
(: c)
n
, a
n
C, is called convergent if it converges on some open
neighbourhood of c. The set of such power series is a Calgebra, and I claim that it is
canonically isomorphic to the Calgebra of germs of functions O
c
.
Let be a holomorphic function on a neighbourhood U of c. Then has a unique
power series expansion =
a
n
(: c)
n
in some (possibly smaller) open neighbourhood
of c (Cartan 1963
2
, II 2.6). Moreover, another holomorphic function
1
on a neighbourhood
U
1
of c denes the same power series if and only if
1
and agree on some neighbourhood
of c contained in U U
t
(ibid. I 4.3). Thus we have a welldened injective map from the
ring of germs of holomorphic functions at c to the ring of convergent power series, which
is obviously surjective.
1
If V is reducible, then it contains disjoint open subsets, say U
1
and U
2
. Let be the function on the union
of U
1
and U
2
taking the constant value 1 on U
1
and the constant value 2 on U
2
. Then is not in O
V
(U
1
LU
2
),
and so condition 3.1c fails.
2
Cartan, Henri. Elementary theory of analytic functions of one or several complex variables. Hermann,
Paris; AddisonWesley; 1963.
The ringed space structure on an algebraic set 57
The ringed space structure on an algebraic set
We now take k to be an algebraically closed eld. Let V be an algebraic subset of k
n
. An
element h of kV  denes functions
1 h(1): V k, and 1 1,h(1): D(h) k.
Thus a pair of elements g. h kV  with h =0 denes a function
1
g(1)
h(1)
: D(h) k.
We say that a function : U k on an open subset U of V is regular if it is of this form
in a neighbourhood of each of its points, i.e., if for all 1 U, there exist g. h kV  with
h(1) =0 such that the functions and
h
agree in a neighbourhood of 1. Write O
V
(U)
for the set of regular functions on U.
For example, if V =k
n
, then a function : U k is regular at a point 1 U if there
exist polynomials g(X
1
. . . . . X
n
) and h(X
1
. . . . . X
n
) with h(1) =0 such that (Q) =
(1)
h(1)
for all Q in a neighbourhood of 1.
PROPOSITION 3.4. The map U O
V
(U) denes a sheaf of kalgebras on V .
PROOF. We have to check the conditions (3.1).
(a) Clearly, a constant function is regular. Suppose and
t
are regular on U, and let
1 U. By assumption, there exist g. g
t
. h. h
t
kV , with h(1) =0 =h
t
(1) such that
and
t
agree with
h
and
0
h
0
respectively near 1. Then
t
agrees with
h
0
0
h
hh
0
near 1,
and so
t
is regular on U. Similarly
t
is regular on U. Thus O
V
(U) is a kalgebra.
(b,c) It is clear from the denition that the condition for to be regular is local.
2
Let g. h kV  and mN. Then 1 g(1),h(1)
n
is a regular function on D(h), and
well show that all regular functions on D(h) are of this form, i.e., 1(D(h). O
V
) .kV 
h
.
In particular, the regular functions on V itself are exactly those dened by elements of kV .
LEMMA 3.5. The function 1 g(1),h(1)
n
on D(h) is the zero function if and only if
and only if gh =0 (in kV ) (and hence g,h
n
=0 in kV 
h
).
PROOF. If g,h
n
is zero on D(h), then gh is zero on V because h is zero on the comple
ment of D(h). Therefore gh is zero in kV . Conversely, if gh =0, then g(1)h(1) =0
for all 1 V , and so g(1) =0 for all 1 D(h).
2
The lemma shows that the canonical map kV 
h
O
V
(D(h)) is welldened and in
jective. The next proposition shows that it is also surjective.
PROPOSITION 3.6. (a) The canonical map kV 
h
1(D(h). O
V
) is an isomorphism.
(b) For every 1 V , there is a canonical isomorphism O
1
kV 
m
P
, where m
1
is
the maximal ideal 1(1).
58 3. AFFINE ALGEBRAIC VARIETIES
PROOF. (a) It remains to show that every regular function on D(h) arises from an ele
ment of kV 
h
. By denition, we know that there is an open covering D(h) =
_
V
i
and el
ements g
i
, h
i
kV  with h
i
nowhere zero on V
i
such that [V
i
=
i
h
i
. We may assume that
each set V
i
is basic, say, V
i
=D(a
i
) for some a
i
kV . By assumption D(a
i
) D(h
i
),
and so a
1
i
=h
i
g
t
i
for some N N and g
t
i
kV  (see p48). On D(a
i
),
=
g
i
h
i
=
g
i
g
t
i
h
i
g
t
i
=
g
i
g
t
i
a
1
i
.
Note that D(a
1
i
) =D(a
i
). Therefore, after replacing g
i
with g
i
g
t
i
and h
i
with a
1
i
, we can
assume that V
i
=D(h
i
).
We nowhave that D(h) =
_
D(h
i
) and that [D(h
i
) =
i
h
i
. Because D(h) is quasicom
pact, we can assume that the covering is nite. As
i
h
i
=
j
h
j
on D(h
i
) D(h
}
) =D(h
i
h
}
),
we have (by Lemma 3.5) that
h
i
h
}
(g
i
h
}
g
}
h
i
) =0, i.e., h
i
h
2
}
g
i
=h
2
i
h
}
g
}
. (*)
Because D(h) =
_
D(h
i
) =
_
D(h
2
i
), the set V((h)) =V((h
2
1
. . . . . h
2
n
)), and so h lies in
rad(h
2
1
. . . . . h
2
n
): there exist a
i
kV  such that
h
1
=
n
i=1
a
i
h
2
i
. (**)
for some N. I claim that is the function on D(h) dened by
o
i
i
h
i
h
N
.
Let 1 be a point of D(h). Then 1 will be in one of the D(h
i
), say D(h
}
). We have the
following equalities in kV :
h
2
}
n
i=1
a
i
g
i
h
i
=
n
i=1
a
i
g
}
h
2
i
h
}
by (*)
=g
}
h
}
h
1
by (**).
But [D(h
}
) =
j
h
j
, i.e., h
}
and g
}
agree as functions on D(h
}
). Therefore we have the
following equality of functions on D(h
}
):
h
2
}
n
i=1
a
i
g
i
h
i
= h
2
}
h
1
.
Since h
2
}
is never zero on D(h
}
), we can cancel it, to nd that, as claimed, the function
h
1
on D(h
}
) equals that dened by
a
i
g
i
h
i
.
(b) In the denition of the germs of a sheaf at 1, it sufces to consider pairs (. U) with
U lying in a some basis for the neighbourhoods of 1, for example, the basis provided by
the basic open subsets. Therefore,
O
1
= lim
h(1)y0
1(D(h). O
V
)
(o)
. lim
hm
P
kV 
h
1.29(b)
. kV 
m
P
.
2
The ringed space structure on an algebraic set 59
REMARK 3.7. Let V be an afne variety and 1 a point on V . Proposition 1.30 shows
that there is a onetoone correspondence between the prime ideals of kV  contained in
m
1
and the prime ideals of O
1
. In geometric terms, this says that there is a onetoone
correspondence between the prime ideals in O
1
and the irreducible closed subvarieties of
V passing through 1.
REMARK 3.8. (a) Let V be an algebraic subset of k
n
, and let =kV . The proposition
and (2.18) allow us to describe (V. O
V
) purely in terms of :
V is the set of maximal ideals in ; for each , let D( ) ={m[ m];
the topology on V is that for which the sets D( ) form a base;
O
V
is the unique sheaf of kalgebras on V for which 1(D( ). O
V
) =
(
.
(b) When V is irreducible, all the rings attached to it are subrings of the eld k(V ). In
this case,
1(D(h). O
V
) =
_
g,h
1
k(V ) [ g kV . N N
_
O
1
={g,h k(V ) [ h(1) =0]
1(U. O
V
) =
_
1U
O
1
=
_
1(D(h
i
). O
V
) if U =
_
D(h
i
).
Note that every element of k(V ) denes a function on some dense open subset of V . Fol
lowing tradition, we call the elements of k(V ) rational functions on V .
3
The equalities
show that the regular functions on an open U V are the rational functions on V that are
dened at each point of U (i.e., lie in O
1
for each 1 U).
EXAMPLE 3.9. (a) Let V = k
n
. Then the ring of regular functions on V , 1(V. O
V
), is
kX
1
. . . . . X
n
. For every nonzero polynomial h(X
1
. . . . . X
n
), the ring of regular functions
on D(h) is
_
g,h
1
k(X
1
. . . . . X
n
) [ g kX
1
. . . . . X
n
. N N
_
.
For every point 1 =(a
1
. . . . . a
n
), the ring of germs of functions at 1 is
O
1
={g,h k(X
1
. . . . . X
n
) [ h(1) =0] =kX
1
. . . . . X
n

(A
1
o
1
,...,A
n
o
n
)
.
and its maximal ideal consists of those g,h with g(1) =0.
(b) Let U ={(a. b) k
2
[ (a. b) =(0. 0)]. It is an open subset of k
2
, but it is not a basic
open subset, because its complement {(0. 0)] has dimension 0, and therefore cant be of the
form V(( )) (see 2.25). Since U =D(X) LD(Y ), the ring of regular functions on U is
O
U
(U) =kX. Y 
A
kX. Y 
Y
(intersection inside k(X. Y )). A regular function on U can be expressed
=
g(X. Y )
X
1
=
h(X. Y )
Y
.
3
The terminology is similar to that of meromorphic function, which also are not functions on the whole
space.
60 3. AFFINE ALGEBRAIC VARIETIES
where we can assume X [ g and Y [ h. On multiplying through by X
1
Y
, we nd that
g(X. Y )Y
=h(X. Y )X
1
.
Because X doesnt divide the left hand side, it cant divide the right hand side either, and
so N = 0. Similarly, M = 0, and so kX. Y : every regular function on U extends
uniquely to a regular function on k
2
.
Morphisms of ringed spaces
A morphism of ringed spaces (V. O
V
) (W. O
W
) is a continuous map : V W such
that
1(U. O
W
) == 1(
1
U. O
V
)
for all open subsets U of W. Sometimes we write
+
( ) for . If U is an open subset of
V , then the inclusion (U. O
V
[V ) (V. O
V
) is a morphism of ringed spaces. A morphism
of ringed spaces is an isomorphism if it is bijective and its inverse is also a morphism of
ringed spaces (in particular, it is a homeomorphism).
EXAMPLE 3.10. (a) Let V and V
t
be topological spaces endowed with their sheaves O
V
and O
V
0 of continuous real valued functions. Every continuous map : V V
t
is a mor
phism of ringed structures (V. O
V
) (V
t
. O
V
0 ).
(b) Let U and U
t
be open subsets of 1
n
and 1
n
respectively, and let .
i
be the coordi
nate function (a
1
. . . . . a
n
) a
i
. Recall from advanced calculus that a map
: U U
t
1
n
is said to be smooth (innitely differentiable) if each of its component functions
i
=.
i
_
,m
:
T,n.
Recall that the image of an element of in ,m .k is denoted (m). Therefore, the
commutativity of the diagram means that, for ,
((n)) =( )(n), i.e., =. (*)
Since
1
D( ) =D( ) (obviously), it follows from (*) that
1
(D( )) =D(( )).
and so is continuous.
Let be a regular function on D(h), and write = g,h
n
, g . Then, from (*)
we see that is the function on D((h)) dened by (g),(h)
n
. In particular, it is
regular, and so maps regular functions on D(h) to regular functions on D((h)).
It follows that sends regular functions on any open subset of spm() to regular
functions on the inverse image of the open subset. Thus denes a morphism of ringed
spaces Spm(T) Spm().
Conversely, by denition, a morphism of : (V. O
V
) (W. O
W
) of afne algebraic
varieties denes a homomorphism of the associated afne kalgebras kW kV . Since
these maps are inverse, we have shown:
PROPOSITION 3.14. For any afne algebras and T,
Hom
kalg
(. T)
:
Mor(Spm(T). Spm()):
for any afne varieties V and W,
Mor(V. W)
:
Hom
kalg
(kW. kV ).
In terms of categories, Proposition 3.14 can now be restated as:
PROPOSITION 3.15. The functor Spmis a (contravariant) equivalence from the cat
egory of afne kalgebras to that of afne algebraic varieties with quasiinverse (V. O
V
)
1(V. O
V
).
Explicit description of morphisms of afne varieties 63
Explicit description of morphisms of afne varieties
PROPOSITION 3.16. Let V =V(a) k
n
, W =V(b) k
n
. The following conditions on
a continuous map : V W are equivalent:
(a) is regular;
(b) the components
1
. . . . .
n
of are all regular;
(c) kW == kV .
PROOF. (a) == (b). By denition
i
=,
i
where ,
i
is the coordinate function
(b
1
. . . . . b
n
) b
i
: W k.
Hence this implication follows directly from the denition of a regular map.
(b) == (c). The map is a kalgebra homomorphism from the ring of all
functions W k to the ring of all functions V k, and (b) says that the map sends the
coordinate functions ,
i
on W into kV . Since the ,
i
s generate kW as a kalgebra, this
implies that it sends kW into kV .
(c) == (a). The map is a homomorphism : kW kV . It therefore
denes a map spmkV  spmkW, and it remains to show that this coincides with
when we identify spmkV  with V and spmkW with W. Let 1 V , let Q = (1),
and let m
1
and m
Q
be the ideals of elements of kV  and kW that are zero at 1 and Q
respectively. Then, for kW,
( ) m
1
((1)) =0 (Q) =0 m
Q
.
Therefore
1
(m
1
) =m
Q
, which is what we needed to show.
2
REMARK 3.17. For 1 V , the maximal ideal in O
V,1
consists of the germs represented
by pairs (. U) with (1) =0. Clearly therefore, the map O
W,(1)
O
V,1
dened by
(see 3.11) maps m
(1)
into m
1
, i.e., it is a local homomorphism of local rings.
Now consider equations
Y
1
=
1
(X
1
. . . . . X
n
)
. . .
Y
n
=
n
(X
1
. . . . . X
n
).
On the one hand, they dene a regular map : k
n
k
n
, namely,
(a
1
. . . . . a
n
) (
1
(a
1
. . . . . a
n
). . . . .
n
(a
1
. . . . . a
n
)).
On the other hand, they dene a homomorphism : kY
1
. . . . . Y
n
 kX
1
. . . . . X
n
 of k
algebras, namely, that sending
Y
i
i
(X
1
. . . . . X
n
).
This map coincides with g g , because
(g)(1) =g(. . . .
i
(1). . . .) =g((1)).
64 3. AFFINE ALGEBRAIC VARIETIES
Now consider closed subsets V(a) k
n
and V(b) k
n
with a and b radical ideals. I claim
that maps V(a) into V(b) if and only if (b) a. Indeed, suppose (V(a)) V(b), and
let g b; for Q V(b),
(g)(Q) =g((Q)) =0.
and so ( ) 1V(b) =b. Conversely, suppose (b) a, and let 1 V(a); for a,
((1)) =( )(1) =0.
and so (1) V(a). When these conditions hold, is the morphism of afne varieties
V(a) V(b) corresponding to the homomorphismkY
1
. . . . . Y
n
,b kX
1
. . . . . X
n
,a de
ned by .
Thus, we see that the regular maps
V(a) V(b)
are all of the form
1 (
1
(1). . . . .
n
(1)).
i
kX
1
. . . . . X
n
.
In particular, they all extend to regular maps A
n
A
n
.
EXAMPLE 3.18. (a) Consider a kalgebra 1. From a kalgebra homomorphism: kX
1, we obtain an element (X) 1, and (X) determines completely. Moreover, (X)
can be any element of 1. Thus
(X): Hom
kalg
(kX. 1)
:
1.
According to (3.14)
Mor(V. A
1
) =Hom
kalg
(kX. kV ).
Thus the regular maps V A
1
are simply the regular functions on V (as we would hope).
(b) Dene A
0
to be the ringed space (V
0
. O
V
0
) with V
0
consisting of a single point, and
1(V
0
. O
V
0
) =k. Equivalently, A
0
=Spmk. Then, for any afne variety V ,
Mor(A
0
. V ) .Hom
kalg
(kV . k) .V
where the last map sends to the point corresponding to the maximal ideal Ker().
(c) Consider t (t
2
. t
3
): A
1
A
2
. This is bijective onto its image,
V : Y
2
=X
3
.
but it is not an isomorphism onto its image the inverse map is not regular. Because of
(3.15), it sufces to show that t (t
2
. t
3
) doesnt induce an isomorphism on the rings of
regular functions. We have kA
1
 =kT  and kV  =kX. Y ,(Y
2
X
3
) =k.. ,. The
map on rings is
. T
2
. , T
3
. k.. , kT .
which is injective, but its image is kT
2
. T
3
 =kT . In fact, k.. , is not integrally closed:
(,,.)
2
. =0, and so (,,.) is integral over k.. ,, but ,,. k.. , (it maps to T under
the inclusion k(.. ,) k(T )).
Subvarieties 65
(d) Let k have characteristic =0, and consider . .
]
: A
n
A
n
. This is a bijection,
but it is not an isomorphism because the corresponding map on rings,
X
i
X
]
i
: kX
1
. . . . . X
n
 kX
1
. . . . . X
n
.
is not surjective.
This is the famous Frobenius map. Take k to be the algebraic closure of F
]
, and write
J for the map. Recall that for each m_ 1 there is a unique subeld F
]
m of k of degree m
over F
]
, and that its elements are the solutions of X
]
m
=X (FT 4.20). Therefore, the xed
points of J
n
are precisely the points of A
n
with coordinates in F
]
m. Let (X
1
. . . . . X
n
) be
a polynomial with coefcients in F
]
m, say,
=
c
i
1
i
n
X
i
1
1
X
i
n
n
. c
i
1
i
n
F
]
m.
Let (a
1
. . . . . a
n
) =0. Then
0 =
_
a
i
1
1
a
i
n
n
_
]
m
=
a
]
m
i
1
1
a
]
m
i
n
n
.
and so (a
]
m
1
. . . . . a
]
m
n
) =0. Here we have used that the binomial theorem takes the simple
form (X Y )
]
m
=X
]
m
Y
]
m
in characteristic . Thus J
n
maps V( ) into itself, and
its xed points are the solutions of
(X
1
. . . . . X
n
) =0
in F
]
m.
In one of the most beautiful pieces of mathematics of the second half of the twentieth
century, Grothendieck dened a cohomology theory ( etale cohomology) and proved a xed
point formula that allowed himto express the number of solutions of a systemof polynomial
equations with coordinates in F
]
n as an alternating sum of traces of operators on nite
dimensional vector spaces, and Deligne used this to obtain very precise estimates for the
number of solutions. See my course notes: Lectures on Etale Cohomology.
Subvarieties
Let be an afne kalgebra. For any ideal a in , we dene
V(a) ={1 spm() [ (1) =0 all a]
={m maximal ideal in [ a m].
This is a closed subset of spm(), and every closed subset is of this form.
Now assume a is radical, so that ,a is again reduced. Corresponding to the homomor
phism ,a, we get a regular map
Spm(,a) Spm()
The image is V(a), and spm(,a) V(a) is a homeomorphism. Thus every closed subset
of spm() has a natural ringed structure making it into an afne algebraic variety. We call
V(a) with this structure a closed subvariety of V.
66 3. AFFINE ALGEBRAIC VARIETIES
ASIDE 3.19. If (V. O
V
) is a ringed space, and 7 is a closed subset of V , we can dene
a ringed space structure on 7 as follows: let U be an open subset of 7, and let be
a function U k; then 1(U. O
7
) if for each 1 U there is a germ (U
t
.
t
) of a
function at 1 (regarded as a point of V ) such that
t
[7U
t
= . One can check that
when this construction is applied to 7 =V(a), the ringed space structure obtained is that
described above.
PROPOSITION 3.20. Let (V. O
V
) be an afne variety and let h be a nonzero element of
kV . Then
(D(h). O
V
[D(h)) .Spm(
h
):
in particular, it is an afne variety.
PROOF. The map
h
denes a morphism spm(
h
) spm(). The image is D(h),
and it is routine (using (1.29)) to verify the rest of the statement.
2
If V =V(a) k
n
, then
(a
1
. . . . . a
n
) (a
1
. . . . . a
n
. h(a
1
. . . . . a
n
)
1
): D(h) k
n1
.
denes an isomorphism of D(h) onto V(a. 1 hX
n1
). For example, there is an isomor
phism of afne varieties
a (a. 1,a): A
1
{0] V A
2
.
where V is the subvariety XY =1 of A
2
the reader should draw a picture.
REMARK 3.21. We have seen that all closed subsets and all basic open subsets of an afne
variety V are again afne varieties with their natural ringed structure, but this is not true
for all open subsets U. As we saw in (3.13), if U is afne, then the natural map U
spm1(U. O
U
) is a bijection. But for U = A
2
(0. 0) = D(X) LD(Y ), we know that
1(U. O
A
2) =kX. Y  (see 3.9b), but U spmkX. Y  is not a bijection, because the ideal
(X. Y ) is not in the image. However, U is clearly a union of afne algebraic varieties
we shall see in the next chapter that it is a (nonafne) algebraic variety.
Properties of the regular map dened by specm()
PROPOSITION 3.22. Let : T be a homomorphism of afne kalgebras, and let
: Spm(T) Spm()
be the corresponding morphism of afne varieties (so that ( ) = ).
(a) The image of is dense for the Zariski topology if and only if is injective.
(b) denes an isomorphism of Spm(T) onto a closed subvariety of Spm() if and only
if is surjective.
Afne space without coordinates 67
PROOF. (a) Let . If the image of is dense, then
=0 == =0.
On the other hand, if the image of is not dense, then the closure of its image will be a
proper closed subset of Spm(), and so there will be a nonzero function that is zero
on it. Then =0.
(b) If is surjective, then it denes an isomorphism ,a T where a is the kernel of
. This induces an isomorphism of Spm(T) with its image in Spm().
2
A regular map : V W of afne algebraic varieties is said to be a dominant (or
dominating) if its image is dense in W. The proposition then says that:
is dominant : 1(W. O
W
) 1(V. O
V
) is injective.
Afne space without coordinates
Let 1 be a vector space over k of dimension n. The set A(1) of points of 1 has a natural
structure of an algebraic variety: the choice of a basis for 1 denes an bijection A(1)
A
n
, and the inherited structure of an afne algebraic variety on A(1) is independent of
the choice of the basis (because the bijections dened by two different bases differ by an
automorphism of A
n
).
We now give an intrinsic denition of the afne variety A(1). Let V be a nite
dimensional vector space over a eld k (not necessarily algebraically closed). The tensor
algebra of V is
T
+
V =
i_0
V
i
with multiplication dened by
(
1
i
) (
t
1
t
}
) =
1
i
t
1
t
}
.
It is noncommutative kalgebra, and the choice of a basis e
1
. . . . . e
n
for V denes an isomor
phism to T
+
V from the kalgebra of noncommuting polynomials in the symbols e
1
. . . . . e
n
.
The symmetric algebra S
+
(V ) of V is dened to be the quotient of T
+
V by the twosided
ideal generated by the relations
nn. . n V.
This algebra is generated as a kalgebra by commuting elements (namely, the elements
of V = V
1
), and so is commutative. The choice of a basis e
1
. . . . . e
n
for V denes an
isomorphism of kalgebras
e
1
e
i
e
1
e
i
: ke
1
. . . . . e
n
 S
+
(V )
(here ke
1
. . . . . e
n
 is the commutative polynomial ring in the symbols e
1
. . . . . e
n
). In partic
ular, S
+
(V ) is an afne kalgebra. The pair (S
+
(V ). i ) consisting of S
+
(V ) and the natural
68 3. AFFINE ALGEBRAIC VARIETIES
klinear map i : V S
+
(V ) has the following universal property: any klinear map V
from V into a kalgebra extends uniquely to a kalgebra homomorphism S
+
(V ) :
V S
+
(V )
.
i
klinear 3 kalgebra
(8)
As usual, this universal property determines the pair (S
+
(V ). i ) uniquely up to a unique
isomorphism.
We now dene A(1) to be Spm(S
+
(1
). ) (3.14)
.Hom
klinear
(1
. ) (8)
.1
k
(linear algebra).
In particular,
A(1)(k) .1.
Moreover, the choice of a basis e
1
. . . . . e
n
for 1 determines a (dual) basis
1
. . . . .
n
of 1
,
and hence an isomorphism of kalgebras k
1
. . . . .
n
 S
+
(1
.
Thus the equivalence classes are the lines through the origin in k
n1
(with the origin omit
ted). Write (a
0
: . . . : a
n
) for the equivalence class containing (a
0
. . . . . a
n
). For each i , let
U
i
={(a
0
: . . . : a
i
: . . . : a
n
) P
n
[ a
i
=0].
Then P
n
=
_
U
i
, and the map
(a
0
: . . . : a
n
) (a
0
,a
i
. . . . . a
n
,a
i
): U
i
u
i
A
n
(the term a
i
,a
i
is omitted) is a bijection. In Chapter 6 we shall show that there is a unique
structure of a (separated) algebraic variety on P
n
for which each U
i
is an open afne sub
variety of P
n
and each map u
i
is an isomorphism of algebraic varieties.
Regular maps
In each of the examples (4.1a,b,c), a morphism of manifolds (continuous map, smooth
map, holomorphic map respectively) is just a morphism of ringed spaces. This motivates
the following denition.
Let (V. O
V
) and (W. O
W
) be algebraic prevarieties. A map : V W is said to be
regular if it is a morphism of ringed spaces. A composite of regular maps is again regular
(this is a general fact about morphisms of ringed spaces).
Note that we have three categories:
(afne varieties) (algebraic prevarieties) (ringed spaces).
Each subcategory is full, i.e., the morphisms Mor(V. W) are the same in the three categories.
PROPOSITION 4.4. Let (V. O
V
) and (W. O
W
) be prevarieties, and let : V W be a con
tinuous map (of topological spaces). Let W =
_
W
}
be a covering of W by open afnes,
and let
1
(W
}
) =
_
V
}i
be a covering of
1
(W
}
) by open afnes. Then is regular if
and only if its restrictions
[V
}i
: V
}i
W
}
are regular for all i. .
Algebraic varieties 73
PROOF. We assume that satises this condition, and prove that it is regular. Let be
a regular function on an open subset U of W. Then [U W
}
is regular for each W
}
(sheaf condition 3.1(b)), and so [
1
(U) V
}i
is regular for each . i (this is our
assumption). It follows that is regular on
1
(U) (sheaf condition 3.1(c)). Thus is
regular. The converse is even easier.
2
ASIDE 4.5. A differentiable manifold of dimension n is locally isomorphic to an open
subset of 1
n
. In particular, all manifolds of the same dimension are locally isomorphic.
This is not true for algebraic varieties, for two reasons:
(a) We are not assuming our varieties are nonsingular (see Chapter 5 below).
(b) The inverse function theorem fails in our context. If 1 is a nonsingular point on
variety of dimension J, we shall see (in the next chapter) that there does exist a neighbour
hood U of 1 and a regular map : U A
d
such that map (J)
1
: T
1
T
(1)
on the
tangent spaces is an isomorphism, but also that there does not always exist a U for which
itself is an isomorphism onto its image (as the inverse function theorem would assert).
Algebraic varieties
In the study of topological manifolds, the Hausdorff condition eliminates such bizarre pos
sibilities as the line with the origin doubled (see 4.10 below) where a sequence tending to
the origin has two limits.
It is not immediately obvious how to impose a separation axiom on our algebraic va
rieties, because even afne algebraic varieties are not Hausdorff. The key is to restate the
Hausdorff condition. Intuitively, the signicance of this condition is that it prevents a se
quence in the space having more than one limit. Thus a continuous map into the space
should be determined by its values on a dense subset, i.e., if
1
and
2
are continuous
maps 7 U that agree on a dense subset of 7 then they should agree on the whole of 7.
Equivalently, the set where two continuous maps
1
.
2
: 7 U agree should be closed.
Surprisingly, afne varieties have this property, provided
1
and
2
are required to be reg
ular maps.
LEMMA 4.6. Let
1
and
2
be regular maps of afne algebraic varieties 7 V . The
subset of 7 on which
1
and
2
agree is closed.
PROOF. There are regular functions .
i
on V such that 1 (.
1
(1). . . . . .
n
(1)) identies
V with a closed subset of A
n
(take the .
i
to be any set of generators for kV  as a kalgebra).
Now .
i
1
and .
i
2
are regular functions on 7, and the set where
1
and
2
agree is
_
n
i=1
V(.
i
1
.
i
2
), which is closed.
2
DEFINITION 4.7. An algebraic prevariety V is said to be separated, or to be an algebraic
variety, if it satises the following additional condition:
Separation axiom: for every pair of regular maps
1
.
2
: 7 V with 7 an
afne algebraic variety, the set {: 7 [
1
(:) =
2
(:)] is closed in 7.
The terminology is not completely standardized: some authors require a variety to be
irreducible, and some call a prevariety a variety.
2
2
Our terminology is agrees with that of JP. Serre, Faisceaux alg ebriques coh erents. Ann. of Math. 61,
(1955). 197278.
74 4. ALGEBRAIC VARIETIES
PROPOSITION 4.8. Let
1
and
2
be regular maps 7 V from an algebraic prevariety 7
to a separated prevariety V . The subset of 7 on which
1
and
2
agree is closed.
PROOF. Let W be the set on which
1
and
2
agree. For any open afne U of 7, W U
is the subset of U on which
1
[U and
2
[U agree, and so W U is closed. This implies
that W is closed because 7 is a nite union of open afnes.
2
EXAMPLE 4.9. The open subspace U =A
2
{(0. 0)] of A
2
becomes an algebraic variety
when endowed with the sheaf O
A
2[U (cf. 3.21).
EXAMPLE 4.10. (The afne line with the origin doubled.) Let V
1
and V
2
be copies of A
1
.
Let V
+
=V
1
LV
2
(disjoint union), and give it the obvious topology. Dene an equivalence
relation on V
+
by
. (in V
1
) ~, (in V
2
) . =, and . =0.
Let V be the quotient space V =V
+
,~with the quotient topology (a set is open if and only
if its inverse image in V
+
is open). Then V
1
and V
2
are open subspaces of V , V =V
1
LV
2
,
and V
1
V
2
=A
1
{0]. Dene a function on an open subset to be regular if its restriction
to each V
i
is regular. This makes V into a prevariety, but not a variety: it fails the separation
axiom because the two maps
A
1
=V
1
V
+
. A
1
=V
2
V
+
agree exactly on A
1
{0], which is not closed in A
1
.
Let Var
k
denote the category of algebraic varieties over k and regular maps. The functor
Spmis a fully faithful contravariant functor Aff
k
Var
k
, and denes an equivalence
of the rst category with the subcategory of the second whose objects are the afne algebraic
varieties.
Maps from varieties to afne varieties
Let (V. O
V
) be an algebraic variety, and let : 1(V. O
V
) be a homomorphism from
an afne kalgebra to the kalgebra of regular functions on V . For any 1 V ,
( )(1) is a kalgebra homomorphism k, and so its kernel (1) is a maximal ideal
in . In this way, we get a map
: V spm()
which is easily seen to be regular. Conversely, from a regular map : V Spm(), we get
a kalgebra homomorphism : 1(V. O
V
). Since these maps are inverse, we
have proved the following result.
PROPOSITION 4.11. For an algebraic variety V and an afne kalgebra , there is a canon
ical onetoone correspondence
Mor(V. Spm()) .Hom
kalgebra
(. 1(V. O
V
)).
Subvarieties 75
Let V be an algebraic variety such that 1(V. O
V
) is an afne kalgebra. Then propo
sition shows that the regular map : V Spm(1(V. O
V
)) dened by id
T(V,O
V
)
has the
following universal property: any regular map from V to an afne algebraic variety U
factors uniquely through :
V Spm(1(V. O
V
))
U.
3
Subvarieties
Let (V. O
V
) be a ringed space, and let W be a subspace. For U open in W, dene O
W
(U)
to be the set of functions : U k such that there exist open subsets U
i
of V and
i
O
V
(U
i
) such that U =W (
_
U
i
) and [W U
i
=
i
[W U
i
for all i . Then (W. O
W
)
is again a ringed space.
We now let (V. O
V
) be a prevariety, and examine when (W. O
W
) is also a prevariety.
Open subprevarieties. Because the open afnes form a base for the topology on V , for
any open subset U of V , (U. O
V
[U) is a prevariety. The inclusion U V is regular, and
U is called an open subprevariety of V . A regular map : W V is an open immersion
if (W) is open in V and denes an isomorphism W (W) (of prevarieties).
Closed subprevarieties. Any closed subset 7 in V has a canonical structure of an al
gebraic prevariety: endow it with the induced topology, and say that a function on an
open subset of 7 is regular if each point 1 in the open subset has an open neighbourhood
U in V such that extends to a regular function on U. To show that 7, with this ringed
space structure is a prevariety, check that for every open afne U V , the ringed space
(U 7. O
7
[U 7) is isomorphic to U 7 with its ringed space structure acquired as a
closed subset of U (see p66). Such a pair (7. O
7
) is called a closed subprevariety of V .
A regular map : W V is a closed immersion if (W) is closed in V and denes an
isomorphism W (W) (of prevarieties).
Subprevarieties. A subset W of a topological space V is said to be locally closed if
every point 1 in W has an open neighbourhood U in V such that W U is closed in U.
Equivalent conditions: W is the intersection of an open and a closed subset of V ; W is
open in its closure. A locally closed subset W of a prevariety V acquires a natural structure
as a prevariety: write it as the intersection W =U 7 of an open and a closed subset; 7
is a prevariety, and W (being open in 7) therefore acquires the structure of a prevariety.
This structure on W has the following characterization: the inclusion map W V is
regular, and a map : V
t
W with V
t
a prevariety is regular if and only if it is regular
when regarded as a map into V . With this structure, W is called a sub(pre)variety of V .
A morphism : V
t
V is called an immersion if it induces an isomorphism of V
t
onto
a subvariety of V . Every immersion is the composite of an open immersion with a closed
immersion (in both orders).
A subprevariety of a variety is automatically separated.
76 4. ALGEBRAIC VARIETIES
Application.
PROPOSITION 4.12. A prevariety V is separated if and only if two regular maps from a
prevariety to V agree on the whole prevariety whenever they agree on a dense subset of it.
PROOF. If V is separated, then the set on which a pair of regular maps
1
.
2
: 7 V agree
is closed, and so must be the whole of the 7.
Conversely, consider a pair of maps
1
.
2
: 7 V , and let S be the subset of 7 on
which they agree. We assume V has the property in the statement of the proposition, and
show that S is closed. Let
S be the closure of S in 7. According to the above discussion,
S and
2
[
S are regular.
Because they agree on a dense subset of
S they agree on the whole of
S, and so S =
S is
closed.
2
Prevarieties obtained by patching
PROPOSITION 4.13. Let V =
_
iJ
V
i
(nite union), and suppose that each V
i
has the
structure of a ringed space. Assume the following patching condition holds:
for all i. , V
i
V
}
is open in both V
i
and V
}
and O
V
i
[V
i
V
}
=O
V
j
[V
i
V
}
.
Then there is a unique structure of a ringed space on V for which
(a) each inclusion V
i
V is a homeomorphism of V
i
onto an open set, and
(b) for each i 1, O
V
[V
i
=O
V
i
.
If every V
i
is an algebraic prevariety, then so also is V , and to give a regular map from
V to a prevariety W amounts to giving a family of regular maps
i
: V
i
W such that
i
[V
i
V
}
=
}
[V
i
V
}
.
PROOF. One checks easily that the subsets U V such that U V
i
is open for all i are the
open subsets for a topology on V satisfying (a), and that this is the only topology to satisfy
(a). Dene O
V
(U) to be the set of functions : U k such that [U V
i
O
V
i
(U V
i
)
for all i . Again, one checks easily that O
V
is a sheaf of kalgebras satisfying (b), and that
it is the only such sheaf.
For the nal statement, if each (V
i
. O
V
i
) is a nite union of open afnes, so also is
(V. O
V
). Moreover, to give a map : V W amounts to giving a family of maps
i
: V
i
W such that
i
[V
i
V
}
=
}
[V
i
V
}
(obviously), and is regular if and only [V
i
is
regular for each i .
2
Clearly, the V
i
may be separated without V being separated (see, for example, 4.10).
In (4.27) below, we give a condition on an open afne covering of a prevariety sufcient to
ensure that the prevariety is separated.
Products of varieties
Let V and W be objects in a category C. A triple
(V W. : V W V. q: V W W)
Products of varieties 77
is said to be the product of V and W if it has the following universal property: for every
pair of morphisms 7 V , 7 W in C, there exists a unique morphism 7 V W
making the diagram
7
V V W W
] q
3
commute. In other words, it is a product if the map
(. q ): Hom(7. V W) Hom(7. V ) Hom(7. W)
is a bijection. The product, if it exists, is uniquely determined up to a unique isomorphism
by this universal property.
For example, the product of two sets (in the category of sets) is the usual cartesion
product of the sets, and the product of two topological spaces (in the category of topological
spaces) is the cartesian product of the spaces (as sets) endowed with the product topology.
We shall show that products exist in the category of algebraic varieties. Suppose, for
the moment, that V W exists. For any prevariety 7, Mor(A
0
. 7) is the underlying set of
7; more precisely, for any : 7, the map A
0
7 with image : is regular, and these are
all the regular maps (cf. 3.18b). Thus, from the denition of products we have
(underlying set of V W) .Mor(A
0
. V W)
.Mor(A
0
. V ) Mor(A
0
. W)
.(underlying set of V ) (underlying set of W).
Hence, our problem can be restated as follows: given two prevarieties V and W, dene on
the set V W the structure of a prevariety such that
(a) the projection maps . q: V W V. W are regular, and
(b) a map : T V W of sets (with T an algebraic prevariety) is regular if its compo
nents . q are regular.
Clearly, there can be at most one such structure on the set V W (because the identity map
will identify any two structures having these properties).
Products of afne varieties
EXAMPLE 4.14. Let a and b be ideals in kX
1
. . . . . X
n
 and kX
n1
. . . . . X
nn
 respec
tively, and let (a. b) be the ideal in kX
1
. . . . . X
nn
 generated by the elements of a and b.
Then there is an isomorphism
g g:
kX
1
. . . . . X
n

a
k
kX
n1
. . . . . X
nn

b
kX
1
. . . . . X
nn

(a. b)
.
Again this comes down to checking that the natural map from
Hom
kalg
(kX
1
. . . . . X
nn
,(a. b). 1)
to
Hom
kalg
(kX
1
. . . . . X
n
,a. 1) Hom
kalg
(kX
n1
. . . . . X
nn
,b. 1)
78 4. ALGEBRAIC VARIETIES
is a bijection. But the three sets are respectively
V(a. b) =zeroset of (a. b) in 1
nn
.
V(a) =zeroset of a in 1
n
.
V(b) =zeroset of b in 1
n
.
and so this is obvious.
The tensor product of two kalgebras and T has the universal property to be a product
in the category of kalgebras, but with the arrows reversed. Because of the category anti
equivalence (3.15), this shows that Spm(
k
T) will be the product of Spm and SpmT
in the category of afne algebraic varieties once we have shown that
k
T is an afne
kalgebra.
PROPOSITION 4.15. Let and T be kalgebras with nitely generated.
(a) If and T are reduced, then so also is
k
T.
(b) If and T are integral domains, then so also is
k
T.
PROOF. Let
k
T. Then =
n
i=1
a
i
b
i
, some a
i
, b
i
T. If one of the b
i
s
is a linear combination of the remaining bs, say, b
n
=
n1
i=1
c
i
b
i
, c
i
k, then, using the
bilinearity of , we nd that
=
n1
i=1
a
i
b
i
n1
i=1
c
i
a
n
b
i
=
n1
i=1
(a
i
c
i
a
n
) b
i
.
Thus we can suppose that in the original expression of , the b
i
s are linearly independent
over k.
Now assume and T to be reduced, and suppose that is nilpotent. Let m be a
maximal ideal of . From a a: ,m=k we obtain homomorphisms
ab ab ab:
k
T k
k
T
:
T
The image
a
i
b
i
of under this homomorphism is a nilpotent element of T, and hence
is zero (because T is reduced). As the b
i
s are linearly independent over k, this means that
the a
i
are all zero. Thus, the a
i
s lie in all maximal ideals m of , and so are zero (see
2.13). Hence =0, and we have shown that
k
T is reduced.
Now assume that and T are integral domains, and let ,
t
k
T be such
that
t
= 0. As before, we can write =
a
i
b
i
and
t
=
a
t
i
b
t
i
with the sets
{b
1
. b
2
. . . .] and {b
t
1
. b
t
2
. . . .] each linearly independent over k. For each maximal ideal m
of , we know (
a
i
b
i
)(
a
t
i
b
t
i
) =0 in T, and so either (
a
i
b
i
) =0 or (
a
t
i
b
t
i
) =0.
Thus either all the a
i
m or all the a
t
i
m. This shows that
spm() =V(a
1
. . . . . a
n
) LV(a
t
1
. . . . . a
t
n
).
As spm() is irreducible (see 2.19), it follows that spm() equals either V(a
1
. . . . . a
n
) or
V(a
t
1
. . . . . a
t
n
). In the rst case =0, and in the second
t
=0.
2
EXAMPLE 4.16. We give some examples to illustrate that k must be taken to be alge
braically closed in the proposition.
Products of varieties 79
(a) Suppose k is nonperfect of characteristic , so that there exists an element in an
algebraic closure of k such that k but
]
k. Let k
t
= k, and let
]
= a. Then
(11) =0 in k
t
k
k
t
(in fact, the elements
i
}
, 0 _i. _1, form a basis
for k
t
k
k
t
as a kvector space), but
( 11)
]
=(a11a)
=(1a1a) (because a k)
=0.
Thus k
t
k
k
t
is not reduced, even though k
t
is a eld.
(b) Let 1 be a nite separable extension of k and let be a second eld containing k.
By the primitive element theorem (FT 5.1),
1 =k =kX,((X)).
for some 1 and its minimal polynomial (X). Assume that is large enough to split
, say, (X) =
i
X
i
with
i
. Because 1,k is separable, the
i
are distinct, and
so
k
1 .X,((X)) (1.35(b))
.
X,(X
i
) (1.1)
and so it is not an integral domain. For example,
C
1
C .CX,(X i ) CX,(X i ) .CC.
The proposition allows us to make the following denition.
DEFINITION 4.17. The product of the afne varieties V and W is
(V W. O
V W
) =Spm(kV 
k
kW)
with the projection maps . q: V W V. W dened by the homomorphisms
1: kV  kV 
k
kW and g 1g: kW kV 
k
kW.
PROPOSITION 4.18. Let V and W be afne varieties.
(a) The variety (V W. O
V W
) is the product of (V. O
V
) and (W. O
W
) in the category
of afne algebraic varieties; in particular, the set V W is the product of the sets V
and W and and q are the projection maps.
(b) If V and W are irreducible, then so also is V W.
PROOF. (a) As noted at the start of the subsection, the rst statement follows from (4.15a),
and the second statement then follows by the argument on p77.
(b) This follows from (4.15b) and (2.19).
2
COROLLARY 4.19. Let V and W be afne varieties. For any prevariety T , a map : T
V W is regular if and q are regular.
80 4. ALGEBRAIC VARIETIES
PROOF. If and q are regular, then (4.18) implies that is regular when restricted
to any open afne of T , which implies that it is regular on T .
2
The corollary shows that V W is the product of V and W in the category of prevari
eties (hence also in the categories of varieties).
EXAMPLE 4.20. (a) It follows from (1.34) that A
nn
endowed with the projection maps
A
n
]
A
nn
q
A
n
.
_
(a
1
. . . . . a
nn
) =(a
1
. . . . . a
n
)
q(a
1
. . . . . a
nn
) =(a
n1
. . . . . a
nn
).
is the product of A
n
and A
n
.
(b) It follows from (1.35c) that
V(a)
]
V(a. b)
q
V(b)
is the product of V(a) and V(b).
The topology on V W is not the product topology; for example, the topology on
A
2
=A
1
A
1
is not the product topology (see 2.29).
Products in general
We now dene the product of two algebraic prevarieties V and W.
Write V as a union of open afnes V =
_
V
i
, and note that V can be regarded as the
variety obtained by patching the (V
i
. O
V
i
); in particular, this covering satises the patching
condition (4.13). Similarly, write W as a union of open afnes W =
_
W
}
. Then
V W =
_
V
i
W
}
and the (V
i
W
}
. O
V
i
W
j
) satisfy the patching condition. Therefore, we can dene (V
W. O
V W
) to be the variety obtained by patching the (V
i
W
}
. O
V
i
W
j
).
PROPOSITION 4.21. With the sheaf of kalgebras O
V W
just dened, V W becomes the
product of V and W in the category of prevarieties. In particular, the structure of prevariety
on V W dened by the coverings V =
_
V
i
and W =
_
W
}
are independent of the
coverings.
PROOF. Let T be a prevariety, and let : T V W be a map of sets such that and
q are regular. Then (4.19) implies that the restriction of to
1
(V
i
W
}
) is regular.
As these open sets cover T , this shows that is regular.
2
PROPOSITION 4.22. If V and W are separated, then so also is V W.
PROOF. Let
1
.
2
be two regular maps U V W. The set where
1
.
2
agree is the
intersection of the sets where
1
.
2
and q
1
. q
2
agree, which is closed.
2
The separation axiom revisited 81
EXAMPLE 4.23. An algebraic group is a variety G together with regular maps
mult: GG G. inverse: G G. A
0
e
G
that make G into a group in the usual sense. For example,
SL
n
=Spm(kX
11
. X
12
. . . . . X
nn
,(det(X
i}
) 1))
and
GL
n
=Spm(kX
11
. X
12
. . . . . X
nn
. Y ,(Y det(X
i}
) 1))
become algebraic groups when endowed with their usual group structure. The only afne
algebraic groups of dimension 1 are
G
n
=GL
1
=SpmkX. X
1

and
G
o
=SpmkX.
Any nite group N can be made into an algebraic group by setting
N =Spm()
with the set of all maps : N k.
Afne algebraic groups are called linear algebraic groups because they can all be re
alized as closed subgroups of GL
n
for some n. Connected algebraic groups that can be
realized as closed algebraic subvarieties of a projective space are called abelian varieties
because they are related to the integrals studied by Abel (happily, they all turn out to be
commutative; see 7.15 below).
The connected component G
is
a nite group. An important theorem of Barsotti and Chevalley says that every connected
algebraic group G contains a unique connected linear algebraic group G
1
such that G,G
1
is an abelian variety. Thus, we have the following coarse classication: every algebraic
group G contains a sequence of normal subgroups
G G
G
1
{e]
with G,G
a nite group, G
,G
1
an abelian variety, and G
1
a linear algebraic group.
The separation axiom revisited
Now that we have the notion of the product of varieties, we can restate the separation axiom
in terms of the diagonal.
By way of motivation, consider a topological space V and the diagonal zV V ,
z
def
={(.. .) [ . V ].
If z is closed (for the product topology), then every pair of points (.. ,) z has a neigh
bourhood U U
t
such that U U
t
z=. In other words, if . and , are distinct points in
V , then there are neighbourhoods U and U
t
of . and , respectively such that U U
t
=.
Thus V is Hausdorff. Conversely, if V is Hausdorff, the reverse argument shows that z is
closed.
For a variety V , we let z=z
V
(the diagonal) be the subset {(. ) [ V ] of V V .
82 4. ALGEBRAIC VARIETIES
PROPOSITION 4.24. An algebraic prevariety V is separated if and only if z
V
is closed.
3
PROOF. Assume z
V
is closed. Let
1
and
2
be regular maps 7 V . The map
(
1
.
2
): 7 V V. : (
1
(:).
2
(:))
is regular because its composites with the projections to V are
1
and
2
. In particular, it
is continuous, and so (
1
.
2
)
1
(z) is closed. But this is precisely the subset on which
1
and
2
agree.
Conversely, suppose V is separated. This means that for any afne variety 7 and regular
maps
1
.
2
: 7 V , the set on which
1
and
2
agree is closed in 7. Apply this with
1
and
2
the two projection maps V V V , and note that the set on which they agree is
z
V
.
2
COROLLARY 4.25. For any prevariety V , the diagonal is a locally closed subset of V V .
PROOF. Let 1 V , and let U be an open afne neighbourhood of 1. Then U U is an
open neighbourhood of (1. 1) in V V , and z
V
(U U) = z
U
, which is closed in
U U because U is separated (4.6).
2
Thus z
V
is always a subvariety of V V , and it is closed if and only if V is separated.
The graph 1
of is
locally closed in V W, and it is closed if W is separated. The map (. ()) is an
isomorphism of V onto 1
V W
]
V is an inverse to (. ()): V 1
.
2
THEOREM 4.27. The following three conditions on a prevariety V are equivalent:
(a) V is separated;
(b) for every pair of open afnes U and U
t
in V , U U
t
is an open afne, and the map
g [
UU
0 g[
UU
0 : kU
k
kU
t
 kU U
t

is surjective;
3
Recall that the topology on V V is not the product topology. Thus the statement does not contradict the
fact that V is not Hausdorff.
The separation axiom revisited 83
(c) the condition in (b) holds for the sets in some open afne covering of V .
PROOF. Let U and U
t
be open afnes in V . We shall prove that
(i) if z is closed then U U
t
afne,
(ii) when U U
t
is afne,
(U U
t
) z is closed kU
k
kU
t
 kU U
t
 is surjective.
Assume (a); then these statements imply (b). Assume that (b) holds for the sets in an
open afne covering (U
i
)
iJ
of V . Then (U
i
U
}
)
(i,})JJ
is an open afne covering of
V V , and z
V
(U
i
U
}
) is closed in U
i
U
}
for each pair (i. ), which implies (a).
Thus, the statements (i) and (ii) imply the theorem.
Proof of (i): The graph of the inclusion U U
t
V is the subset (U U
t
) z of
(U U
t
) V. If z
V
is closed, then (U U
t
) z
V
is a closed subvariety of an afne
variety, and hence is afne (see p66). Now (4.26) implies that U U
t
is afne.
Proof of (ii): Assume that U U
t
is afne. Then
(U U
t
) z
V
is closed in U U
t
(. ): U U
t
U U
t
is a closed immersion
kU U
t
 kU U
t
 is surjective (3.22).
Since kU U
t
 =kU
k
kU
t
, this completes the proof of (ii).
2
In more downtoearth terms, condition (b) says that U U
t
is afne and every regular
function on U U
t
is a sum of functions of the form1 (1)g(1) with and g regular
functions on U and U
t
.
EXAMPLE 4.28. (a) Let V = P
1
, and let U
0
and U
1
be the standard open subsets (see
4.3). Then U
0
U
1
= A
1
{0], and the maps on rings corresponding to the inclusions
U
i
U
0
U
1
are
(X) (X): kX kX. X
1

(X) (X
1
): kX kX. X
1
.
Thus the sets U
0
and U
1
satisfy the condition in (b).
(b) Let V be A
1
with the origin doubled (see 4.10), and let U and U
t
be the upper and
lower copies of A
1
in V . Then U U
t
is afne, but the maps on rings corresponding to the
inclusions U
i
U
0
U
1
are
X X: kX kX. X
1

X X: kX kX. X
1
.
Thus the sets U
0
and U
1
fail the condition in (b).
(c) Let V be A
2
with the origin doubled, and let U and U
t
be the upper and lower copies
of A
2
in V . Then U U
t
is not afne (see 3.21).
84 4. ALGEBRAIC VARIETIES
Fibred products
Consider a variety S and two regular maps : V S and [: W S. Then the set
V
S
W
def
={(. n) V W [ () =[(n)]
is a closed subvariety of V W (because it is the set where and [ q agree). It is
called the bred product of V and W over S. Note that if S consists of a single point, then
V
S
W =V W.
Write
t
for the map (. n) n: V
S
W W and [
t
for the map (. n) : V
S
W V . We then have a commutative diagram:
V
S
W
0
W
_)
0
_
)
V
S.
The bred product has the following universal property: consider a pair of regular maps
: T V , : T W; then
t ((t ), (t )): T V W
factors through V
S
W (as a map of sets) if and only if =[, in which case (. ) is
regular (because it is regular as a map into V W);
T
V
S
W W
V S
(, )
)
The map
t
in the above diagram is called the base change of with respect to [.
For any point 1 S, the base change of : V S with respect to 1 S is the map
1
(1) 1 induced by , which is called the bre of V over 1.
EXAMPLE 4.29. If : V S is a regular map and U is an open subvariety of S, then
V
S
U is the inverse image of U in S.
EXAMPLE 4.30. Since a tensor product of rings
T
T has the opposite universal prop
erty to that of a bred product, one might hope that
Spm()
Spm(T)
Spm(T)
=Spm(
T
T).
This is true if
T
T is an afne kalgebra, but in general it may have nilpotent
4
elements.
For example, let 1 =kX, let =k with the 1algebra structure sending X to a, and let
4
By this, of course, we mean nonzero nilpotent elements.
Dimension 85
T =kX with the 1algebra structure sending X to X
]
. When k has characteristic =0,
then
T
T .k
kA
p
j
kX .kX,(X
]
a).
The correct statement is
Spm()
Spm(T)
Spm(T) .Spm(
T
T,N) (9)
where N is the ideal of nilpotent elements in
T
T. To prove this, note that for any
variety T ,
Mor(T. Spm(
T
T,N)) .Hom(
T
T,N. 1(T. O
T
))
.Hom(
T
T. 1(T. O
T
))
.Hom(. 1(T. O
T
))
Hom(T,T(T,O
T
))
Hom(T. 1(T. O
T
))
.Mor(V. Spm())
Mor(V,Spm(T))
Mor(V. Spm(T)).
For the rst and fourth isomorphisms, we used (4.11); for the second isomorphism, we used
that 1(T. O
T
) has no nilpotents; for the third isomorphism, we used the universal property
of
T
T.
Dimension
In an irreducible algebraic variety V , every nonempty open subset is dense and irreducible.
If U and U
t
are open afnes in V , then so also is U U
t
and
kU kU U
t
 k(U)
where k(U) is the eld of fractions of kU, and so k(U) is also the eld of fractions of
kU U
t
 and of kU
t
. Thus, we can attach to V a eld k(V ), called the eld of rational
functions on V , such that for every open afne U in V , k(V ) is the eld of fractions of
kU. The dimension of V is dened to be the transcendence degree of k(V ) over k. Note
the dim(V ) =dim(U) for any open subset U of V . In particular, dim(V ) =dim(U) for U
an open afne in V . It follows that some of the results in 2 carry over for example, if
7 is a proper closed subvariety of V , then dim(7) <dim(V ).
PROPOSITION 4.31. Let V and W be irreducible varieties. Then
dim(V W) =dim(V ) dim(W).
PROOF. We may suppose V and W to be afne. Write
kV  =k.
1
. . . . . .
n

kW =k,
1
. . . . . ,
n

where the .s and ,s have been chosen so that {.
1
. . . . . .
d
] and {,
1
. . . . . ,
e
] are maxi
mal algebraically independent sets of elements of kV  and kW. Then {.
1
. . . . . .
d
] and
{,
1
. . . . . ,
e
] are transcendence bases of k(V ) and k(W) (see FT 8.12), and so dim(V ) =J
and dim(W) =e. Then
5
kV W
def
=kV 
k
kW k.
1
. . . . . .
d

k
k,
1
. . . . . ,
e
 .k.
1
. . . . . .
d
. ,
1
. . . . . ,
e
.
5
In general, it is not true that if M
t
and N
t
are 1submodules of M and N, then M
t
T
N
t
is an 1
submodule of M
T
N. However, this is true if 1 is a eld, because then M
t
and N
t
will be direct summands
of M and N, and tensor products preserve direct summands.
86 4. ALGEBRAIC VARIETIES
Therefore {.
1
1. . . . . .
d
1. 1,
1
. . . . . 1,
e
] will be algebraically independent in kV 
k
kW. Obviously kV W is generated as a kalgebra by the elements .
i
1, 1 ,
}
,
1 _i _m, 1 _ _n, and all of them are algebraic over
k.
1
. . . . . .
d

k
k,
1
. . . . . ,
e
.
Thus the transcendence degree of k(V W) is J e.
2
We extend the denition of dimension to an arbitrary variety V as follows. An algebraic
variety is a nite union of noetherian topological spaces, and so is noetherian. Consequently
(see 2.21), V is a nite union V =
_
V
i
of its irreducible components, and we dene
dim(V ) =maxdim(V
i
). When all the irreducible components of V have dimension n. V is
said to be pure of dimension n (or to be of pure dimension n).
Birational equivalence
Two irreducible varieties V and W are said to be birationally equivalent if k(V ) ~k(W).
PROPOSITION 4.32. Two irreducible varieties V and W are birationally equivalent if and
only if there are open subsets U and U
t
of V and W respectively such that U ~U
t
.
PROOF. Assume that V and W are birationally equivalent. We may suppose that V and W
are afne, corresponding to the rings and T say, and that and T have a common eld
of fractions 1. Write T = k.
1
. . . . . .
n
. Then .
i
= a
i
,b
i
, a
i
. b
i
, and T
b
1
...b
r
.
Since Spm(
b
1
...b
r
) is a basic open subvariety of V , we may replace with
b
1
...b
r
, and
suppose that T . The same argument shows that there exists a J T such that
T
d
. Now
T T
d
=T
d
d
(T
d
)
d
=T
d
.
and so
d
= T
d
. This shows that the open subvarieties D(b) V and D(b) W are
isomorphic. This proves the only if part, and the if part is obvious.
2
REMARK 4.33. Proposition 4.32 can be improved as follows: if V and W are irreducible
varieties, then every inclusion k(V ) k(W) is dened by a regular surjective map : U
U
t
from an open subset U of W onto an open subset U
t
of V .
PROPOSITION 4.34. Every irreducible algebraic variety of dimension J is birationally
equivalent to a hypersurface in A
d1
.
PROOF. Let V be an irreducible variety of dimension J. According to FT 8.21, there exist
algebraically independent elements .
1
. . . . . .
d
k(V ) such that k(V ) is nite and separable
over k(.
1
. . . . . .
d
). By the primitive element theorem(FT 5.1), k(V ) =k(.
1
. . . . . .
d
. .
d1
)
for some .
d1
. Let kX
1
. . . . . X
d1
 be an irreducible polynomial satised by the .
i
,
and let H be the hypersurface =0. Then k(V ) ~k(H).
2
REMARK 4.35. An irreducible variety V of dimension J is said to rational if it is bira
tionally equivalent to A
d
. It is said to be unirational if k(V ) can be embedded in k(A
d
)
according to (4.33), this means that there is a regular surjective map from an open subset of
Dominant maps 87
A
dimV
onto an open subset of V . L uroths theorem (cf. FT 8.19) says that every unirational
curve is rational. It was proved by Castelnuovo that when k has characteristic zero every
unirational surface is rational. Only in the seventies was it shown that this is not true for
three dimensional varieties (Artin, Mumford, Clemens, Grifths, Manin,...). When k has
characteristic = 0, Zariski showed that there exist nonrational unirational surfaces, and
P. Blass showed that there exist innitely many surfaces V , no two birationally equivalent,
such that k(X
]
. Y
]
) k(V ) k(X. Y ).
Dominant maps
As in the afne case, a regular map : V W is said to be dominant (or dominating)
if the image of is dense in W. Suppose V and W are irreducible. If V
t
and W
t
are
open afne subsets of V and W such that (V
t
) W
t
, then (3.22) implies that the map
: kW
t
 kV
t
 is injective. Therefore it extends to a map on the elds of
fractions, k(W) k(V ), and this map is independent of the choice of V
t
and W
t
.
Algebraic varieties as a functors
Let be an afne kalgebra, and let V be an algebraic variety. We dene a point of V with
coordinates in to be a regular map Spm() V . For example, if V =V(a) k
n
, then
V() ={(a
1
. . . . . a
n
)
n
[ (a
1
. . . . . a
n
) =0 all a].
which is what you should expect. In particular V(k) =V (as a set), i.e., V (as a set) can be
identied with the set of points of V with coordinates in k. Note that
(V W)() =V() W()
(property of a product).
REMARK 4.36. Let V be the union of two subvarieties, V = V
1
LV
2
. If V
1
and V
2
are
both open, then V() = V
1
() LV
2
(), but not necessarily otherwise. For example, for
any polynomial (X
1
. . . . . X
n
),
A
n
=D
(
LV( )
where D
(
.Spm(kX
1
. . . . . X
n
. T ,(1T )) and V( ) is the zero set of , but
n
={a
n
[ (a)
] L{a
n
[ (a) =0]
in general.
THEOREM 4.37. A regular map : V W of algebraic varieties denes a family of maps
of sets, (): V() W(), one for each afne kalgebra , such that for every homo
morphism : T of afne kalgebras,
V()
()
W()
_
V()
_
V()
T V(T)
(B)
W(T)
(*)
88 4. ALGEBRAIC VARIETIES
commutes. Every family of maps with this property arises from a unique morphism of
algebraic varieties.
For a variety V , let h
aff
V
be the functor sending an afne kalgebra to V(). We can
restate as Theorem 4.37 follows.
THEOREM 4.38. The functor
V h
aff
V
: Var
k
Fun(Aff
k
. Sets)
if fully faithful.
PROOF. The Yoneda lemma (1.38) shows that the functor
V h
V
: Var
k
Fun(Var
k
. Sets)
is fully faithful. Let be a morphism h
aff
V
h
aff
V
0
, and let T be a variety. Let (U
i
)
iJ
be a
nite afne covering of T . Each intersection U
i
U
}
is afne (4.27), and so gives rise to
a commutative diagram
0 h
V
(T )
i
h
V
(U
i
)
i,}
h
V
(U
i
U
}
)
0 h
V
0 (T )
i
h
V
0 (U
i
)
i,}
h
V
0 (U
i
U
}
)
in which the pairs of maps are dened by the inclusions U
i
U
}
U
i
. U
}
. As the rows
are exact (4.13), this shows that
V
extends uniquely to a functor h
V
h
V
0 , which (by the
Yoneda lemma) arises from a unique regular map V V
t
.
2
COROLLARY 4.39. To give an afne algebraic group is the same as to give a functor
G: Aff
k
Gp such that for some n and some nite set S of polynomials in kX
1
. X
2
. . . . . X
n
,
G() is the set of zeros of S in
n
.
PROOF. Certainly an afne algebraic group denes such a functor. Conversely, the con
ditions imply that G = h
V
for an afne algebraic variety V (unique up to a unique iso
morphism). The multiplication maps G() G() G() give a morphism of functors
h
V
h
V
h
V
. As h
V
h
V
.h
V V
(by denition of V V ), we see that they arise from
a regular map V V V . Similarly, the inverse map and the identityelement map are
regular.
2
It is not unusual for a variety to be most naturally dened in terms of its points functor.
REMARK 4.40. The essential image of h h
V
: Var
aff
k
Fun(Aff
k
. Sets) consists of the
functors J dened by some (nite) set of polynomials.
We now describe the essential image of h h
V
: Var
k
Fun(Aff
k
. Sets). The bre
product of two maps
1
: J
1
J
3
,
2
: J
2
J
3
of sets is the set
J
1
T
3
J
2
={(.
1
. .
2
) [
1
(.
1
) =
2
(.
2
)].
Exercises 89
When J
1
. J
2
. J
3
are functors and
1
.
2
.
3
are morphisms of functors, there is a functor
J =J
1
T
3
J
2
such that
(J
1
T
3
J
2
)() =J
1
()
T
3
()
J
2
()
for all afne kalgebras .
To simplify the statement of the next proposition, we write U for h
U
when U is an
afne variety.
PROPOSITION 4.41. A functor J: Aff
k
Sets is in the essential image of Var
k
if and only
if there exists an afne scheme U and a morphism U J such that
(a) the functor 1
def
=U
T
U is a closed afne subvariety of U U and the maps 1U
dened by the projections are open immersions;
(b) the set 1(k) is an equivalence relation on U(k), and the map U(k) J(k) realizes
J(k) as the quotient of U(k) by 1(k).
PROOF. Let J = h
V
for V an algebraic variety. Choose a nite open afne covering
V =
_
U
i
of V , and let U =
U
i
. It is again an afne variety (Exercise 42). The functor
1 is h
U
0 where U
t
is the disjoint union of the varieties U
i
U
}
. These are afne (4.27),
and so U
t
is afne. As U
t
is the inverse image of z
V
in U U, it is closed (4.24). This
proves (a), and (b) is obvious.
The converse is omitted for the present.
2
REMARK 4.42. A variety V denes a functor 1 V(1) from the category of all k
algebras to Sets. For example, if V is afne,
V(1) =Hom
kalgebra
(kV . 1).
More explicitly, if V k
n
and 1(V ) = (
1
. . . . .
n
), then V(1) is the set of solutions in
1
n
of the system equations
i
(X
1
. . . . . X
n
) =0. i =1. . . . . m.
Again, we call the elements of V(1) the points of V with coordinates in 1.
Note that, when we allow 1 to have nilpotent elements, it is important to choose the
i
to generate 1(V ) (i.e., a radical ideal) and not just an ideal a such that V(a) =V .
6
Exercises
41. Show that the only regular functions on P
1
are the constant functions. [Thus P
1
is
not afne. When k =C, P
1
is the Riemann sphere (as a set), and one knows from complex
analysis that the only holomorphic functions on the Riemann sphere are constant. Since
regular functions are holomorphic, this proves the statement in this case. The general case
is easier.]
6
Let a be an ideal in kX
1
. . . .. If has no nonzero nilpotent elements, then every kalgebra homomor
phism kX
1
. . . . that is zero on a is also zero on rad(a), and so
Hom
k
(kX
1
. . . .,a. ) .Hom
k
(kX
1
. . . .,rad(a). ).
This is not true if has nonzero nilpotents.
90 4. ALGEBRAIC VARIETIES
42. Let V be the disjoint union of algebraic varieties V
1
. . . . . V
n
. This set has an obvious
topology and ringed space structure for which it is an algebraic variety. Show that V is
afne if and only if each V
i
is afne.
43. Show that every algebraic subgroup of an algebraic group is closed.
CHAPTER 5
Local Study
In this chapter, we examine the structure of a variety near a point. We begin with the
case of a curve, since the ideas in the general case are the same but the formulas are more
complicated. Throughout, k is an algebraically closed eld.
Tangent spaces to plane curves
Consider the curve
V : J(X. Y ) =0
in the plane dened by a nonconstant polynomial J(X. Y ). We assume that J(X. Y ) has
no multiple factors, so that (J(X. Y )) is a radical ideal and 1(V ) = (J(X. Y )). We can
factor J into a product of irreducible polynomials, J(X. Y ) =
J
i
(X. Y ), and then V =
_
V(J
i
) expresses V as a union of its irreducible components. Each component V(J
i
)
has dimension 1 (see 2.25) and so V has pure dimension 1. More explicitly, suppose for
simplicity that J(X. Y ) itself is irreducible, so that
kV  =kX. Y ,(J(X. Y )) =k.. ,
is an integral domain. If J =Xc, then . is transcendental over k and , is algebraic over
k(.), and so . is a transcendence basis for k(V ) over k. Similarly, if J =Y c, then , is
a transcendence basis for k(V ) over k.
Let (a. b) be a point on V . In calculus, the equation of the tangent at 1 = (a. b) is
dened to be
dJ
dX
(a. b)(X a)
dJ
dY
(a. b)(Y b) =0. (10)
This is the equation of a line unless both
JT
JA
(a. b) and
JT
JY
(a. b) are zero, in which case it is
the equation of a plane.
DEFINITION 5.1. The tangent space T
1
V to V at 1 = (a. b) is the space dened by
equation (10).
When
JT
JA
(a. b) and
JT
JY
(a. b) are not both zero, T
1
(V ) is a line, and we say that 1 is
a nonsingular or smooth point of V . Otherwise, T
1
(V ) has dimension 2, and we say that
1 is singular or multiple. The curve V is said to be nonsingular or smooth when all its
points are nonsingular.
91
92 5. LOCAL STUDY
We regard T
1
(V ) as a subspace of the twodimensional vector space T
1
(A
2
), which is
the twodimensional space of vectors with origin 1.
EXAMPLE 5.2. For each of the following examples, the reader (or his computer) is invited
to sketch the curve.
1
The characteristic of k is assumed to be =2. 3.
(a) X
n
Y
n
= 1. All points are nonsingular unless the characteristic divides m (in
which case X
n
Y
n
1 has multiple factors).
(b) Y
2
=X
3
. Here only (0. 0) is singular.
(c) Y
2
=X
2
(X 1). Here again only (0. 0) is singular.
(d) Y
2
=X
3
aX b. In this case,
V is singular Y
2
X
3
aX b, 2Y , and 3X
2
a have a common zero
X
3
aX b and 3X
2
a have a common zero.
Since 3X
2
a is the derivative of X
3
aXb, we see that V is singular if and only
if X
3
aX b has a multiple root.
(e) (X
2
Y
2
)
2
3X
2
Y Y
3
=0. The origin is (very) singular.
(f) (X
2
Y
2
)
3
4X
2
Y
2
=0. The origin is (even more) singular.
(g) V = V(JG) where JG has no multiple factors and J and G are relatively prime.
Then V =V(J) LV(G), and a point (a. b) is singular if and only if it is a singular
point of V(J), a singular point of V(G), or a point of V(J) V(G). This follows
immediately from the equations given by the product rule:
d(JG)
dX
=J
dG
dX
dJ
dX
G.
d(JG)
dY
=J
dG
dY
dJ
dY
G.
PROPOSITION 5.3. Let V be the curve dened by a nonconstant polynomial J without
multiple factors. The set of nonsingular points
2
is an open dense subset V .
PROOF. We can assume that J is irreducible. We have to show that the set of singular
points is a proper closed subset. Since it is dened by the equations
J =0.
dJ
dX
=0.
dJ
dY
=0.
it is obviously closed. It will be proper unless dJ,dX and dJ,dY are identically zero on V ,
and are therefore both multiples of J, but, since they have lower degree, this is impossible
unless they are both zero. Clearly dJ,dX = 0 if and only if J is a polynomial in Y (k
of characteristic zero) or is a polynomial in X
]
and Y (k of characteristic ). A similar
remark applies to dJ,dY . Thus if dJ,dX and dJ,dY are both zero, then J is constant
(characteristic zero) or a polynomial in X
]
, Y
]
, and hence a th power (characteristic ).
These are contrary to our assumptions.
2
The set of singular points of a variety is called the singular locus of the variety.
1
For (b,e,f), see p57 of: Walker, Robert J., Algebraic Curves. Princeton Mathematical Series, vol. 13.
Princeton University Press, Princeton, N. J., 1950 (reprinted by Dover 1962).
2
In common usage, singular means uncommon or extraordinary as in he spoke with singular shrewd
ness. Thus the proposition says that singular points (mathematical sense) are singular (usual sense).
Tangent cones to plane curves 93
Tangent cones to plane curves
A polynomial J(X. Y ) can be written (uniquely) as a nite sum
J =J
0
J
1
J
2
J
n
(11)
where J
n
is a homogeneous polynomial of degree m. The term J
1
will be denoted J
I
and
called the linear formof J, and the rst nonzero termon the right of (11) (the homogeneous
summand of J of least degree) will be denoted J
+
and called the leading form of J.
If 1 =(0. 0) is on the curve V dened by J, then J
0
=0 and (11) becomes
J =aX bY higher degree terms;
moreover, the equation of the tangent space is
aX bY =0.
DEFINITION 5.4. Let J(X. Y ) be a polynomial without square factors, and let V be the
curve dened by J. If (0. 0) V , then the geometric tangent cone to V at (0. 0) is the zero
set of J
+
. The tangent cone is the pair (V(J
+
). J
+
). To obtain the tangent cone at any other
point, translate to the origin, and then translate back.
EXAMPLE 5.5. (a) Y
2
=X
3
: the tangent cone at (0. 0) is dened by Y
2
it is the Xaxis
(doubled).
(b) Y
2
=X
2
(X 1): the tangent cone at (0,0) is dened by Y
2
X
2
it is the pair
of lines Y =X.
(c) (X
2
Y
2
)
2
3X
2
Y Y
3
=0: the tangent cone at (0. 0) is dened by 3X
2
Y Y
3
it is the union of the lines Y =0, Y =
_
3X.
(d) (X
2
Y
2
)
3
4X
2
Y
2
=0: the tangent cone at (0. 0) is dened by 4X
2
Y
2
=0
it is the union of the X and Y axes (each doubled).
In general we can factor J
+
as
J
+
(X. Y ) =
X
i
0
(Y a
i
X)
i
i
.
Then degJ
+
=
r
i
is called the multiplicity of the singularity, mult
1
(V ). A multiple
point is ordinary if its tangents are nonmultiple, i.e., r
i
=1 all i . An ordinary double point
is called a node, and a nonordinary double point is called a cusp. (There are many names
for special types of singularities see any book, especially an old book, on curves.)
The local ring at a point on a curve
PROPOSITION 5.6. Let 1 be a point on a curve V , and let mbe the corresponding maximal
ideal in kV . If 1 is nonsingular, then dim
k
(m,m
2
) =1, and otherwise dim
k
(m,m
2
) =2.
94 5. LOCAL STUDY
PROOF. Assume rst that 1 = (0. 0). Then m = (.. ,) in kV  = kX. Y ,(J(X. Y )) =
k.. ,. Note that m
2
=(.
2
. .,. ,
2
), and
m,m
2
=(X. Y ),(m
2
J(X. Y )) =(X. Y ),(X
2
. XY. Y
2
. J(X. Y )).
In this quotient, every element is represented by a linear polynomial c. J,, and the
only relation is J
I
(.. ,) =0. Clearly dim
k
(m,m
2
) =1 if J
I
=0, and dim
k
(m,m
2
) =2
otherwise. Since J
I
=0 is the equation of the tangent space, this proves the proposition in
this case.
The same argument works for an arbitrary point (a. b) except that one uses the variables
X
t
=X a and Y
t
=Y b; in essence, one translates the point to the origin.
2
We explain what the condition dim
k
(m,m
2
) =1 means for the local ring O
1
=kV 
m
.
Let n be the maximal ideal mkV 
m
of this local ring. The map mn induces an isomor
phism m,m
2
n,n
2
(see 1.31), and so we have
1 nonsingular dim
k
m,m
2
=1 dim
k
n,n
2
=1.
Nakayamas lemma (1.3) shows that the last condition is equivalent to n being a principal
ideal. Since O
1
is of dimension 1, n being principal means O
1
is a regular local ring of
dimension 1 (1.6), and hence a discrete valuation ring, i.e., a principal ideal domain with
exactly one prime element (up to associates) (CA 17.4). Thus, for a curve,
1 nonsingular O
1
regular O
1
is a discrete valuation ring.
Tangent spaces of subvarieties of A
m
Before dening tangent spaces at points of closed subvarietes of A
n
we review some ter
minology from linear algebra.
Linear algebra
For a vector space k
n
, let X
i
be the i th coordinate function a a
i
. Thus X
1
. . . . . X
n
is
the dual basis to the standard basis for k
n
. A linear form
a
i
X
i
can be regarded as an
element of the dual vector space (k
n
)
=Hom(k
n
. k).
Let =(a
i}
) be an nm matrix. It denes a linear map : k
n
k
n
, by
_
_
_
a
1
.
.
.
a
n
_
_
_
_
_
_
a
1
.
.
.
a
n
_
_
_
=
_
_
_
n
}=1
a
1}
a
}
.
.
.
n
}=1
a
n}
a
}
_
_
_
.
Write X
1
. . . . . X
n
for the coordinate functions on k
n
and Y
1
. . . . . Y
n
for the coordinate
functions on k
n
. Then
Y
i
=
n
}=1
a
i}
X
}
.
This says that, when we apply to a, then the i th coordinate of the result is
n
}=1
a
i}
(X
}
a) =
n
}=1
a
i}
a
}
.
Tangent spaces of subvarieties of A
n
95
Tangent spaces
Consider an afne variety V k
n
, and let a = 1(V ). The tangent space T
a
(V ) to V
at a = (a
1
. . . . . a
n
) is the subspace of the vector space with origin a cut out by the linear
equations
m
i=1
dJ
dX
i
a
(X
i
a
i
) =0. J a. (12)
Thus T
a
(A
n
) is the vector space of dimension m with origin a, and T
a
(V ) is the subspace
of T
a
(A
n
) dened by the equations (12).
Write (JX
i
)
a
for (X
i
a
i
); then the (JX
i
)
a
form a basis for the dual vector space
T
a
(A
n
)
to T
a
(A
n
) in fact, they are the coordinate functions on T
a
(A
n
)
. As in ad
vanced calculus, we dene the differential of a polynomial J kX
1
. . . . . X
n
 at a by the
equation:
(JJ)
a
=
n
i=1
dJ
dX
i
a
(JX
i
)
a
.
It is again a linear form on T
a
(A
n
). In terms of differentials, T
a
(V ) is the subspace of
T
a
(A
n
) dened by the equations:
(JJ)
a
=0. J a. (13)
I claim that, in (12) and (13), it sufces to take the J in a generating subset for a. The
product rule for differentiation shows that if G =
}
H
}
J
}
, then
(JG)
a
=
}
H
}
(a) (JJ
}
)
a
J
}
(a) (JH
}
)
a
.
If J
1
. . . . . J
i
generate a and a V(a), so that J
}
(a) = 0 for all , then this equation be
comes
(JG)
a
=
}
H
}
(a) (JJ
}
)
a
.
Thus (JJ
1
)
a
. . . . . (JJ
i
)
a
generate the kspace {(JJ)
a
[ J a].
When V is irreducible, a point a on V is said to be nonsingular (or smooth) if the
dimension of the tangent space at a is equal to the dimension of V ; otherwise it is singular
(or multiple). When V is reducible, we say a is nonsingular if dimT
a
(V ) is equal to the
maximum dimension of an irreducible component of V passing through a. It turns out then
that a is singular precisely when it lies on more than one irreducible component, or when it
lies on only one component but is a singular point of that component.
Let a =(J
1
. . . . . J
i
), and let
J =Jac(J
1
. . . . . J
i
) =
_
dJ
i
dX
}
_
=
_
_
_
_
JT
1
JA
1
. . . . .
JT
1
JA
m
.
.
.
.
.
.
JT
r
JA
1
. . . . .
JT
r
JA
m
_
_
_
_
.
Then the equations dening T
a
(V ) as a subspace of T
a
(A
n
) have matrix J(a). Therefore,
linear algebra shows that
dim
k
T
a
(V ) =mrankJ(a).
96 5. LOCAL STUDY
and so a is nonsingular if and only if the rank of Jac(J
1
. . . . . J
i
)(a) is equal to mdim(V ).
For example, if V is a hypersurface, say 1(V ) =(J(X
1
. . . . . X
n
)), then
Jac(J)(a) =
_
dJ
dX
1
(a). . . . .
dJ
dX
n
(a)
_
.
and a is nonsingular if and only if not all of the partial derivatives
JT
JA
i
vanish at a.
We can regard J as a matrix of regular functions on V . For each r,
{a V [ rankJ(a) _r]
is closed in V , because it is the set where certain determinants vanish. Therefore, there
is an open subset U of V on which rankJ(a) attains its maximum value, and the rank
jumps on closed subsets. Later (5.18) we shall show that the maximum value of rankJ(a)
is mdimV , and so the nonsingular points of V form a nonempty open subset of V .
The differential of a regular map
Consider a regular map
: A
n
A
n
. a (1
1
(a
1
. . . . . a
n
). . . . . 1
n
(a
1
. . . . . a
n
)).
We think of as being given by the equations
Y
i
=1
i
(X
1
. . . . . X
n
), i =1. . . . n.
It corresponds to the map of rings
+
: kY
1
. . . . . Y
n
 kX
1
. . . . . X
n
 sending Y
i
to 1
i
(X
1
. . . . . X
n
),
i =1. . . . n.
Let a A
n
, and let b =(a). Dene (J)
a
: T
a
(A
n
) T
b
(A
n
) to be the map such that
(JY
i
)
b
(J)
a
=
d1
i
dX
}
a
(JX
}
)
a
.
i.e., relative to the standard bases, (J)
a
is the map with matrix
Jac(1
1
. . . . . 1
n
)(a) =
_
_
_
_
J1
1
JA
1
(a). . . . .
J1
1
JA
m
(a)
.
.
.
.
.
.
J1
n
JA
1
(a). . . . .
J1
n
JA
m
(a)
_
_
_
_
.
For example, suppose a =(0. . . . . 0) and b =(0. . . . . 0), so that T
a
(A
n
) =k
n
and T
b
(A
n
) =
k
n
, and
1
i
=
m
jD1
c
i}
X
}
(higher terms), i =1. . . . . n.
Then Y
i
(J)
a
=
}
c
i}
X
}
, and the map on tangent spaces is given by the matrix (c
i}
),
i.e., it is simply t (c
i}
)t.
Let J kX
1
. . . . . X
n
. We can regard J as a regular map A
n
A
1
, whose differential
will be a linear map
(JJ)
a
: T
a
(A
n
) T
b
(A
1
). b =J(a).
When we identify T
b
(A
1
) with k, we obtain an identication of the differential of J (J
regarded as a regular map) with the differential of J (J regarded as a regular function).
Etale maps 97
LEMMA 5.7. Let : A
n
A
n
be as at the start of this subsection. If maps V =V(a)
k
n
into W =V(b) k
n
, then (J)
a
maps T
a
(V ) into T
b
(W), b =(a).
PROOF. We are given that
b = a.
and have to prove that
b =(J )
b
(J)
a
is zero on T
a
(V ).
The chain rule holds in our situation:
d
dX
i
=
n
}=1
d
dY
}
dY
}
dX
i
. Y
}
=1
}
(X
1
. . . . . X
n
). =(Y
1
. . . . . Y
n
).
If is the map given by the equations
Y
}
=1
}
(X
1
. . . . . X
n
). =1. . . . . n.
then the chain rule implies
J( )
a
=(J )
b
(J)
a
. b =(a).
Let t T
a
(V ); then
(J )
b
(J)
a
(t) =J( )
a
(t).
which is zero if b because then a. Thus (J)
a
(t) T
b
(W).
2
We therefore get a map (J)
a
: T
a
(V ) T
b
(W). The usual rules from advanced calcu
lus hold. For example,
(J[)
b
(J)
a
=J([ )
a
. b =(a).
The denition we have given of T
a
(V ) appears to depend on the embedding V
A
n
. Later we shall give an intrinsic of the tangent space, which is independent of any
embedding.
EXAMPLE 5.8. Let V be the union of the coordinate axes in A
3
, and let W be the zero set
of XY(X Y ) in A
2
. Each of V and W is a union of three lines meeting at the origin. Are
they isomorphic as algebraic varieties? Obviously, the origin o is the only singular point on
V or W. An isomorphism V W would have to send the singular point to the singular
point, i.e., o o, and map T
o
(V ) isomorphically onto T
o
(W). But V =V(XY. Y7. X7),
and so T
o
(V ) has dimension 3, whereas T
o
W has dimension 2. Therefore, they are not
isomorphic.
Etale maps
DEFINITION 5.9. A regular map : V W of smooth varieties is etale at a point 1 of V
if (J)
1
: T
1
(V ) T
(1)
(W) is an isomorphism; is etale if it is etale at all points of V .
98 5. LOCAL STUDY
EXAMPLE 5.10. (a) A regular map
: A
n
A
n
, a (1
1
(a
1
. . . . . a
n
). . . . . 1
n
(a
1
. . . . . a
n
))
is etale at a if and only if rankJac(1
1
. . . . . 1
n
)(a) = n, because the map on the tangent
spaces has matrix Jac(1
1
. . . . . 1
n
)(a)). Equivalent condition: det
_
J1
i
JA
j
(a)
_
=0
(b) Let V = Spm() be an afne variety, and let =
c
i
X
i
X be such that
X,((X)) is reduced. Let W = Spm(X,((X)), and consider the map W V
corresponding to the inclusion X,( ). Thus
X,( ) X W V A
1
V
The points of W lying over a point a V are the pairs (a. b) V A
1
such that b is a root
of
c
i
(a)X
i
. I claim that the map W V is etale at (a. b) if and only if b is a simple root
of
c
i
(a)X
i
.
To see this, write =SpmkX
1
. . . . . X
n
,a, a =(
1
. . . . .
i
), so that
X,( ) =kX
1
. . . . . X
n
,(
1
. . . . .
i
. ).
The tangent spaces to W and V at (a. b) and a respectively are the null spaces of the matrices
_
_
_
_
_
_
J(
1
JA
1
(a) . . .
J(
1
JA
m
(a) 0
.
.
.
.
.
.
J(
n
JA
1
(a) . . .
J(
n
JA
m
(a) 0
J(
JA
1
(a) . . .
J(
JA
m
(a)
J(
JA
(a. b)
_
_
_
_
_
_
_
_
_
_
J(
1
JA
1
(a) . . .
J(
1
JA
m
(a)
.
.
.
.
.
.
J(
n
JA
1
(a) . . .
J(
n
JA
m
(a)
_
_
_
_
and the map T
(a,b)
(W) T
a
(V ) is induced by the projection map k
n1
k
n
omitting the
last coordinate. This map is an isomorphism if and only if
J(
JA
(a. b)=0, because then any
solution of the smaller set of equations extends uniquely to a solution of the larger set. But
d
dX
(a. b) =
J(
i
c
i
(a)X
i
)
JX
(b).
which is zero if and only if b is a multiple root of
i
c
i
(a)X
i
. The intuitive picture is
that W V is a nite covering with deg( ) sheets, which is ramied exactly at the points
where two or more sheets cross.
(c) Consider a dominant map : W V of smooth afne varieties, corresponding to
a map T of rings. Suppose T can be written T =Y
1
. . . . . Y
n
,(1
1
. . . . . 1
n
) (same
number of polynomials as variables). A similar argument to the above shows that is etale
if and only if det
_
J1
i
JA
j
(a)
_
is never zero.
(d) The example in (b) is typical; in fact every etale map is locally of this form, provided
V is normal (in the sense dened below p105). More precisely, let : W V be etale
Intrinsic denition of the tangent space 99
at 1 W, and assume V to normal; then there exist a map
t
: W
t
V
t
with kW
t
 =
kV
t
X,((X)), and a commutative diagram
W U
1
~ U
t
1
W
t
_
0
V U
2
~ U
t
2
V
t
with the Us all open subvarieties and 1 U
1
.
In advanced calculus (or differential topology, or complex analysis), the inverse
function theorem says that a map that is etale at a point a is a local isomorphism
there, i.e., there exist open neighbourhoods U and U
t
of a and (a) such that induces an
isomorphism U U
t
. This is not true in algebraic geometry, at least not for the Zariski
topology: a map can be etale at a point without being a local isomorphism. Consider for
example the map
: A
1
{0] A
1
{0]. a a
2
.
This is etale if the characteristic is =2, because the Jacobian matrix is (2X), which has rank
one for all X =0 (alternatively, it is of the form (5.10b) with (X) =X
2
T , where T is
the coordinate function on A
1
, and X
2
c has distinct roots for c =0). Nevertheless, I claim
that there do not exist nonempty open subsets U and U
t
of A
1
{0] such that denes an
isomorphism U U
t
. If there did, then would dene an isomorphism kU
t
 kU
and hence an isomorphism on the elds of fractions k(A
1
) k(A
1
). But on the elds of
fractions, denes the map k(X) k(X), X X
2
, which is not an isomorphism.
ASIDE 5.11. There is an old conjecture that any etale map : A
n
A
n
is an isomorphism.
If we write =(1
1
. . . . . 1
n
), then this becomes the statement:
if det
_
d1
i
dX
}
(a)
_
is never zero (for a k
n
), then has a inverse.
The condition, det
_
J1
i
JA
j
(a)
_
never zero, implies that det
_
J1
i
JA
j
_
is a nonzero constant (by
the Nullstellensatz 2.6 applied to the ideal generated by det
_
J1
i
JA
j
_
). This conjecture, which
is known as the Jacobian conjecture, has not been settled even for k =C and n =2, despite
the existence of several published proofs and innumerable announced proofs. It has caused
many mathematicians a good deal of grief. It is probably harder than it is interesting. See
Bass et al. 1982
3
.
Intrinsic denition of the tangent space
The denition we have given of the tangent space at a point used an embedding of the
variety in afne space. In this section, we give an intrinsic denition that depends only on
a small neighbourhood of the point.
3
Bass, Hyman; Connell, Edwin H.; Wright, David. The Jacobian conjecture: reduction of degree and
formal expansion of the inverse. Bull. Amer. Math. Soc. (N.S.) 7 (1982), no. 2, 287330.
100 5. LOCAL STUDY
LEMMA 5.12. Let c be an ideal in kX
1
. . . . . X
n
 generated by linear forms
1
. . . . .
i
,
which we may assume to be linearly independent. Let X
i
1
. . . . . X
i
nr
be such that
{
1
. . . . .
i
. X
i
1
. . . . . X
i
nr
]
is a basis for the linear forms in X
1
. . . . . X
n
. Then
kX
1
. . . . . X
n
,c .kX
i
1
. . . . . X
i
nr
.
PROOF. This is obvious if the forms are X
1
. . . . . X
i
. In the general case, because {X
1
. . . . . X
n
]
and {
1
. . . . .
i
. X
i
1
. . . . . X
i
nr
] are both bases for the linear forms, each element of one set
can be expressed as a linear combination of the elements of the other. Therefore,
kX
1
. . . . . X
n
 =k
1
. . . . .
i
. X
i
1
. . . . . X
i
nr
.
and so
kX
1
. . . . . X
n
,c =k
1
. . . . .
i
. X
i
1
. . . . . X
i
nr
,c
.kX
i
1
. . . . . X
i
nr
.
2
Let V = V(a) k
n
, and assume that the origin o lies on V . Let a
I
be the ideal gen
erated by the linear terms
I
of the a. By denition, T
o
(V ) = V(a
I
). Let
I
=
kX
1
. . . . . X
n
,a
I
, and let m be the maximal ideal in kV  consisting of the functions zero
at o; thus m=(.
1
. . . . . .
n
).
PROPOSITION 5.13. There are canonical isomorphisms
Hom
klinear
(m,m
2
. k)
:
Hom
kalg
(
I
. k)
:
T
o
(V ).
PROOF. First isomorphism: Let n = (X
1
. . . . . X
n
) be the maximal ideal at the origin in
kX
1
. . . . . X
n
. Then m,m
2
.n,(n
2
a), and as
I
n
2
for every a, it follows
that m,m
2
.n,(n
2
a
I
). Let
1,I
. . . . .
i,I
be a basis for the vector space a
I
. From linear
algebra we know that there are n r linear forms X
i
1
. . . . . X
i
nr
forming with the
i,I
a
basis for the linear forms on k
n
. Then X
i
1
m
2
. . . . . X
i
nr
m
2
form a basis for m,m
2
as a kvector space, and the lemma shows that
I
.kX
i
1
. . . . X
i
nr
. A homomorphism
:
I
k of kalgebras is determined by its values (X
i
1
). . . . . (X
i
nr
), and they can be
arbitrarily given. Since the klinear maps m,m
2
k have a similar description, the rst
isomorphism is now obvious.
Second isomorphism: To give a kalgebra homomorphism
I
k is the same as to
give an element (a
1
. . . . . a
n
) k
n
such that (a
1
. . . . . a
n
) =0 for all a
I
, which is the
same as to give an element of T
1
(V ).
2
Let n be the maximal ideal in O
o
(=
m
). According to (1.31), m,m
2
n,n
2
, and so
there is a canonical isomorphism
T
o
(V )
:
Hom
klin
(n,n
2
. k).
We adopt this as our denition.
Intrinsic denition of the tangent space 101
DEFINITION 5.14. The tangent space T
1
(V ) at a point 1 of a variety V is dened to be
Hom
klinear
(n
1
,n
2
1
. k), where n
1
the maximal ideal in O
1
.
The above discussion shows that this agrees with previous denition
4
for 1 =o V
A
n
. The advantage of the present denition is that it obviously depends only on a (small)
neighbourhood of 1. In particular, it doesnt depend on an afne embedding of V .
Note that (1.4) implies that the dimension of T
1
(V ) is the minimum number of ele
ments needed to generate n
1
O
1
.
A regular map : V W sending 1 to Q denes a local homomorphism O
Q
O
1
,
which induces maps n
Q
n
1
, n
Q
,n
2
Q
n
1
,n
2
1
, and T
1
(V ) T
Q
(W). The last map
is written (J)
1
. When some open neighbourhoods of 1 and Q are realized as closed
subvarieties of afne space, then (J)
1
becomes identied with the map dened earlier.
In particular, an n
1
is represented by a regular map U A
1
on a neighbourhood
U of 1 sending 1 to 0 and hence denes a linear map (J )
1
: T
1
(V ) k. This is just
the map sending a tangent vector (element of Hom
klinear
(n
1
,n
2
1
. k)) to its value at
mod n
2
1
. Again, in the concrete situation V A
n
this agrees with the previous denition.
In general, for O
1
, i.e., for a germ of a function at 1, we dene
(J )
1
= (1) mod n
2
.
The tangent space at 1 and the space of differentials at 1 are dual vector spaces.
Consider for example, a V(a) A
n
, with a a radical ideal. For kA
n
 =kX
1
. . . . . X
n
,
we have (trivial Taylor expansion)
=(1)
c
i
(X
i
a
i
) terms of degree _2 in the X
i
a
i
.
that is,
(1)
c
i
(X
i
a
i
) mod m
2
1
.
Therefore (J )
1
can be identied with
c
i
(X
i
a
i
) =
d
dX
i
a
(X
i
a
i
).
which is how we originally dened the differential.
5
The tangent space T
a
(V(a)) is the zero
set of the equations
(J )
1
=0. a.
and the set {(J )
1
[
T
a
(V )
[ kX
1
. . . . . X
n
] is the dual space to T
a
(V ).
REMARK 5.15. Let 1 be a nite dimensional vector space over k. Then
T
o
(A(1)) .1.
4
More precisely, dene T
1
(V ) =Hom
klinear
(n,n
2
. k). For V =A
n
, the elements (JX
i
)
o
=X
i
n
2
for
1 _ i _ m form a basis for n,n
2
, and hence form a basis for the space of linear forms on T
1
(V ). A closed
immersion i : V A
n
sending 1 to o maps T
1
(V ) isomorphically onto the linear subspace of T
o
(A
n
) dened
by the equations
1_i_n
_
d
dX
i
_
o
(JX
i
)
o
=0. 1(iV ).
5
The same discussion applies to any O
1
. Such an is of the form
h
with h(a) =0, and has a (not
quite so trivial) Taylor expansion of the same form, but with an innite number of terms, i.e., it lies in the power
series ring kX
1
a
1
. . . . . X
n
a
n
.
102 5. LOCAL STUDY
Nonsingular points
DEFINITION 5.16. (a) A point 1 on an algebraic variety V is said to be nonsingular if it
lies on a single irreducible component V
i
of V , and dim
k
T
1
(V ) =dimV
i
; otherwise the
point is said to be singular.
(b) A variety is nonsingular if all of its points are nonsingular.
(c) The set of singular points of a variety is called its singular locus.
Thus, on an irreducible variety V of dimension J,
1 is nonsingular dim
k
T
1
(V ) =J
dim
k
(n
1
,n
2
1
) =J
n
1
can be generated by J functions.
PROPOSITION 5.17. Let V be an irreducible variety of dimension J. If 1 V is nonsin
gular, then there exist J regular functions
1
. . . . .
d
dened in an open neighbourhood U
of 1 such that 1 is the only common zero of the
i
on U.
PROOF. Let
1
. . . . .
d
generate the maximal ideal n
1
in O
1
. Then
1
. . . . .
d
are all
dened on some open afne neighbourhood U of 1, and I claim that 1 is an irreducible
component of the zero set V(
1
. . . . .
d
) of
1
. . . . .
d
in U. If not, there will be some
irreducible component 7 =1 of V(
1
. . . . .
d
) passing through 1. Write 7 =V(p) with
p a prime ideal in kU. Because V(p) V(
1
. . . . .
d
) and because 7 contains 1 and is
not equal to it, we have
(
1
. . . . .
d
) p _m
1
(ideals in kU).
On passing to the local ring O
1
=kU
m
P
, we nd (using 1.30) that
(
1
. . . . .
d
) pO
1
_n
1
(ideals in O
1
).
This contradicts the assumption that the
i
generate n
1
. Hence 1 is an irreducible compo
nent of V(
1
. . . . .
d
). On removing the remaining irreducible components of V(
1
. . . . .
d
)
from U, we obtain an open neighbourhood of 1 with the required property.
2
THEOREM 5.18. The set of nonsingular points of a variety is dense and open.
PROOF. We have to show that the singular points form a proper closed subset of every
irreducible component of V .
Closed: We can assume that V is afne, say V =V(a) A
n
. Let 1
1
. . . . . 1
i
generate a.
Then the set of singular points is the zero set of the ideal generated by the (nJ) (nJ)
minors of the matrix
Jac(1
1
. . . . . 1
i
)(a) =
_
_
_
_
J1
1
JA
1
(a) . . .
J1
1
JA
m
(a)
.
.
.
.
.
.
J1
r
JA
1
(a) . . .
J1
r
JA
m
(a)
_
_
_
_
Proper: According to (4.32) and (4.34) there is a nonempty open subset of V isomor
phic to a nonempty open subset of an irreducible hypersurface in A
d1
, and so we may
Nonsingularity and regularity 103
suppose that V is an irreducible hypersurface in A
d1
, i.e., that it is the zero set of a single
nonconstant irreducible polynomial J(X
1
. . . . . X
d1
). By (2.25), dimV = J. Now the
proof is the same as that of (5.3): if
JT
JA
1
is identically zero on V(J), then
JT
JA
1
must be
divisible by J, and hence be zero. Thus J must be a polynomial in X
2
. . . . X
d1
(char
acteristic zero) or in X
]
1
. X
2
. . . . . X
d1
(characteristic ). Therefore, if all the points of V
are singular, then J is constant (characteristic 0) or a th power (characteristic ) which
contradict the hypothesis.
2
COROLLARY 5.19. An irreducible algebraic variety is nonsingular if and only if its tangent
spaces T
1
(V ), 1 V , all have the same dimension.
PROOF. According to the theorem, the constant dimension of the tangent spaces must be
the dimension of V , and so all points are nonsingular.
2
COROLLARY 5.20. Every algebraic group G is nonsingular.
PROOF. From the theorem we know that there is an open dense subset U of G of non
singular points. For any g G, a ga is an isomorphism G G, and so gU consists
of nonsingular points. Clearly G =
_
gU. (Alternatively, because G is homogeneous, all
tangent spaces have the same dimension.)
2
In fact, any variety on which a group acts transitively by regular maps will be nonsin
gular.
ASIDE 5.21. Note that, if V is irreducible, then
dimV =min
1
dimT
1
(V )
This formula can be useful in computing the dimension of a variety.
Nonsingularity and regularity
In this section we assume two results that wont be proved until 9.
5.22. For any irreducible variety V and regular functions
1
. . . . .
i
on V , the irreducible
components of V(
1
. . . . .
i
) have dimension _dimV r (see 9.7).
Note that for polynomials of degree 1 on k
n
, this is familiar from linear algebra: a
system of r linear equations in n variables either has no solutions (the equations are incon
sistent) or its solutions form an afne space of dimension at least nr.
5.23. If V is an irreducible variety of dimension J, then the local ring at each point 1 of
V has dimension J (see 9.6).
Because of (1.30), the height of a prime ideal p of a ring is the Krull dimension of
p
. Thus (5.23) can be restated as: if V is an irreducible afne variety of dimension J, then
every maximal ideal in kV  has height J.
104 5. LOCAL STUDY
Sketch of proof of (5.23): If V =A
d
, then =kX
1
. . . . . X
d
, and all maximal ideals
in this ring have height J, for example,
(X
1
a
1
. . . . . X
d
a
d
) (X
1
a
1
. . . . . X
d1
a
d1
) . . . (X
1
a
1
) 0
is a chain of prime ideals of length J that cant be rened, and there is no longer chain. In the
general case, the Noether normalization theoremsays that kV  is integral over a polynomial
ring k.
1
. . . . . .
d
, .
i
kV ; then clearly .
1
. . . . . .
d
is a transcendence basis for k(V ), and
the going up and down theorems show that the local rings of kV  and k.
1
. . . . . .
d
 have
the same dimension.
THEOREM 5.24. Let 1 be a point on an irreducible variety V . Any generating set for the
maximal ideal n
1
of O
1
has at least J elements, and there exists a generating set with J
elements if and only if 1 is nonsingular.
PROOF. If
1
. . . . .
i
generate n
1
, then the proof of (5.17) shows that 1 is an irreducible
component of V(
1
. . . . .
i
) in some open neighbourhood U of 1. Therefore (5.22) shows
that 0 _J r, and so r _J. The rest of the statement has already been noted.
2
COROLLARY 5.25. A point 1 on an irreducible variety is nonsingular if and only if O
1
is
regular.
PROOF. This is a restatement of the second part of the theorem.
2
According to CA 17.3, a regular local ring is an integral domain. If 1 lies on two
irreducible components of a V , then O
1
is not an integral domain,
6
and so O
1
is not
regular. Therefore, the corollary holds also for reducible varieties.
Nonsingularity and normality
An integral domain that is integrally closed in its eld of fractions is called a normal ring.
LEMMA 5.26. An integral domain is normal if and only if
m
is normal for all maximal
ideals m of .
PROOF. =: If is integrally closed, then so is S
1
for any multiplicative subset S (not
containing 0), because if
b
n
c
1
b
n1
c
n
=0. c
i
S
1
.
then there is an s S such that sc
i
for all i , and then
(sb)
n
(sc
1
)(sb)
n1
s
n
c
n
=0.
6
Suppose that 1 lies on the intersection 7
1
7
2
of the distinct irreducible components 7
1
and 7
2
. Since
7
1
7
2
is a proper closed subset of 7
1
, there is an open afne neighbourhood U of 1 such that U 7
1
7
2
is a proper closed subset of U 7
1
, and so there is a nonzero regular function
1
on U 7
1
that is zero on
U 7
1
7
2
. Extend
1
to a neighbourhood of 1 in 7
1
L7
2
by setting
1
(Q) = 0 for Q 7
2
. Then
1
denes a nonzero germ of regular function at 1. Similarly construct a function
2
that is zero on 7
1
. Then
1
and
2
dene nonzero germs of functions at 1, but their product is zero.
Etale neighbourhoods 105
demonstrates that sb , whence b S
1
.
=: If c is integral over , it is integral over each
m
, hence in each
m
, and =
_
m
(if c
_
m
, then the set of a such that ac is an ideal in , not contained in any
maximal ideal, and therefore equal to itself).
2
Thus the following conditions on an irreducible variety V are equivalent:
(a) for all 1 V , O
1
is integrally closed;
(b) for all irreducible open afnes U of V , kU is integrally closed;
(c) there is a covering V =
_
V
i
of V by open afnes such that kV
i
 is integrally closed
for all i .
An irreducible variety V satisfying these conditions is said to be normal. More generally,
an algebraic variety V is said to be normal if O
1
is normal for all 1 V . Since, as
we just noted, the local ring at a point lying on two irreducible components cant be an
integral domain, a normal variety is a disjoint union of irreducible varieties (each of which
is normal).
A regular local noetherian ring is always normal (cf. CA 17.5); conversely, a normal
local integral domain of dimension one is regular. Thus nonsingular varieties are normal,
and normal curves are nonsingular. However, a normal surface need not be nonsingular: the
cone
X
2
Y
2
7
2
=0
is normal, but is singular at the origin the tangent space at the origin is k
3
. However, it is
true that the set of singular points on a normal variety V must have dimension _dimV 2.
For example, a normal surface can only have isolated singularities the singular locus
cant contain a curve.
Etale neighbourhoods
Recall that a regular map : W V is said to be etale at a nonsingular point 1 of W if the
map (J)
1
: T
1
(W) T
(1)
(V ) is an isomorphism.
Let 1 be a nonsingular point on a variety V of dimension J. A local system of parame
ters at 1 is a family {
1
. . . . .
d
] of germs of regular functions at 1 generating the maximal
ideal n
1
O
1
. Equivalent conditions: the images of
1
. . . . .
d
in n
1
,n
2
1
generate it as a
kvector space (see 1.4); or (J
1
)
1
. . . . . (J
d
)
1
is a basis for dual space to T
1
(V ).
PROPOSITION 5.27. Let {
1
. . . . .
d
] be a local system of parameters at a nonsingular
point 1 of V . Then there is a nonsingular open neighbourhood U of 1 such that
1
.
2
. . . . .
d
are represented by pairs (
1
. U). . . . . (
d
. U) and the map (
1
. . . . .
d
): U A
d
is etale.
PROOF. Obviously, the
i
are represented by regular functions
i
dened on a single open
neighbourhood U
t
of 1, which, because of (5.18), we can choose to be nonsingular. The
map =(
1
. . . . .
d
): U
t
A
d
is etale at 1, because the dual map to (J)
a
is (JX
i
)
o
(J
i
)
a
. The next lemma then shows that is etale on an open neighbourhood U of 1.
2
LEMMA 5.28. Let W and V be nonsingular varieties. If : W V is etale at 1, then it is
etale at all points in an open neighbourhood of 1.
106 5. LOCAL STUDY
PROOF. The hypotheses imply that W and V have the same dimension J, and that their
tangent spaces all have dimension J. We may assume W and V to be afne, say W A
n
and V A
n
, and that is given by polynomials 1
1
(X
1
. . . . . X
n
). . . . . 1
n
(X
1
. . . . . X
n
).
Then (J)
a
: T
a
(A
n
) T
(a)
(A
n
) is a linear map with matrix
_
J1
i
JA
j
(a)
_
, and is not etale
at a if and only if the kernel of this map contains a nonzero vector in the subspace T
a
(V ) of
T
a
(A
n
). Let
1
. . . . .
i
generate 1(W). Then is not etale at a if and only if the matrix
_
J(
i
JA
j
(a)
J1
i
JA
j
(a)
_
has rank less than m. This is a polynomial condition on a, and so it fails on a closed subset
of W, which doesnt contain 1.
2
Let V be a nonsingular variety, and let 1 V . An etale neighbourhood of a point 1
of V is pair (Q. : U V ) with an etale map from a nonsingular variety U to V and Q
a point of U such that (Q) =1.
COROLLARY 5.29. Let V be a nonsingular variety of dimension J, and let 1 V . There
is an open Zariski neighbourhood U of 1 and a map : U A
d
realizing (1. U) as an
etale neighbourhood of (0. . . . . 0) A
d
.
PROOF. This is a restatement of the Proposition.
2
ASIDE 5.30. Note the analogy with the denition of a differentiable manifold: every point
1 on nonsingular variety of dimension J has an open neighbourhood that is also a neigh
bourhood of the origin in A
d
. There is a topology on algebraic varieties for which the
open neighbourhoods of a point are the etale neighbourhoods. Relative to this topol
ogy, any two nonsingular varieties are locally isomorphic (this is not true for the Zariski
topology). The topology is called the etale topology see my notes Lectures on Etale
Cohomology.
The inverse function theorem
THEOREM 5.31 (INVERSE FUNCTION THEOREM). If a regular map of nonsingular vari
eties : V W is etale at 1 V , then there exists a commutative diagram
V
open
U
1
_
~
_
0
W
etale
U
(1)
with U
1
an open neighbourhood U of 1, U
((1)
an etale neighbourhood (1), and
t
an
isomorphism.
PROOF. According to (5.28), there exists an open neighbourhood U of 1 such that the
restriction [U of to U is etale. To get the above diagram, we can take U
1
=U, U
(1)
to be the etale neighbourhood [U: U W of (1), and
t
to be the identity map.
2
Smooth maps 107
The rank theorem
For vector spaces, the rank theorem says the following: let : V W be a linear map of
kvector spaces of rank r; then there exist bases for V and W relative to which has matrix
_
1
i
0
0 0
_
. In other words, there is a commutative diagram
V
W
_
~
_
~
k
n
(x
1
,...,x
m
)(x
1
,...,x
r
,0,...)
k
n
A similar result holds locally for differentiable manifolds. In algebraic geometry, there is
the following weaker analogue.
THEOREM 5.32 (RANK THEOREM). Let : V W be a regular map of nonsingular va
rieties of dimensions m and n respectively, and let 1 V . If rank(T
1
()) =n, then there
exists a commutative diagram
U
1
[U
P
U
(1)
_ etale
_ etale
A
n
(x
1
,...,x
m
)(x
1
,...,x
n
)
A
n
in which U
1
and U
(1)
are open neighbourhoods of 1 and (1) respectively and the
vertical maps are etale.
PROOF. Choose a local systemof parameters g
1
. . . . . g
n
at (1), and let
1
=g
1
. . . . .
n
=
g
n
. Then J
1
. . . . . J
n
are linearly independent forms on T
1
(V ), and there exist
n1
. . . . .
n
such J
1
. . . . . J
n
is a basis for T
1
(V )
. Then
1
. . . . .
n
is a local system of parameters
at 1. According to (5.28), there exist open neighbourhoods U
1
of 1 and U
(1)
of (1)
such that the maps
(
1
. . . . .
n
): U
1
A
n
(g
1
. . . . . g
n
): U
(1)
A
n
are etale. They give the vertical maps in the above diagram.
2
Smooth maps
DEFINITION 5.33. A regular map : V W of nonsingular varieties is smooth at a point
1 of V if (J)
1
: T
1
(V ) T
(1)
(W) is surjective; is smooth if it is smooth at all points
of V .
THEOREM 5.34. A map : V W is smooth at 1 V if and only if there exist open
neighbourhoods U
1
and U
(1)
of 1 and (1) respectively such that [U
1
factors into
U
1
etale
A
dimV dimW
U
(1)
q
U
(1)
.
108 5. LOCAL STUDY
PROOF. Certainly, if [U
1
factors in this way, it is smooth. Conversely, if is smooth at
1, then we get a diagram as in the rank theorem. From it we get maps
U
1
A
n
A
n U
(1)
U
(1)
.
The rst is etale, and the second is the projection of A
nn
U
(1)
onto U
(1)
.
2
COROLLARY 5.35. Let V and W be nonsingular varieties. If : V W is smooth at 1,
then it is smooth on an open neighbourhood of V .
PROOF. In fact, it is smooth on the neighbourhood U
1
in the theorem.
2
Dual numbers and derivations
In general, if is a kalgebra and M is an module, then a kderivation is a map D:
M such that
(a) D(c) =0 for all c k;
(b) D( g) =D( ) D(g);
(c) D(g) = Dg Dg (Leibnizs rule).
Note that the conditions imply that D is klinear (but not linear). We write Der
k
(. M)
for the space of all kderivations M.
For example, the map (J )
1
def
= (1) mod n
2
1
is a kderivation O
1
n
1
,n
2
1
.
PROPOSITION 5.36. There are canonical isomorphisms
Der
k
(O
1
. k)
:
Hom
klinear
(n
1
,n
2
1
. k)
:
T
1
(V ).
PROOF. The composite k
cc
O
1
( ((1)
k is the identity map, and so, when regarded
as kvector space, O
1
decomposes into
O
1
=k n
1
. ((1). (1)).
A derivation D: O
1
k is zero on k and on n
2
1
(by Leibnizs rule). It therefore denes a
klinear map n
1
,n
2
1
k. Conversely, a klinear map n
1
,n
2
1
k denes a derivation by
composition
O
1
( (d( )
P
n
1
,n
2
1
k.
2
The ring of dual numbers is kc = kX,(X
2
) where c = X (X
2
). As a kvector
space it has a basis {1. c], and (abc)(a
t
b
t
c) =aa
t
(ab
t
a
t
b)c.
PROPOSITION 5.37. The tangent space to V at 1 is canonically isomorphic to the space
of local homomorphisms of local kalgebras O
1
kc:
T
1
(V ) .Hom(O
1
. kc).
Dual numbers and derivations 109
PROOF. Let : O
1
kc be a local homomorphism of kalgebras, and write (a) =
a
0
D
(ab)c. and
(a)(b) =(a
0
D
(a)c)(b
0
D
(b)c) =a
0
b
0
(a
0
D
(b) b
0
D
(a))c.
On comparing these expressions, we see that D
dJ
dX
i
a
(X
i
a
i
) 1
with 1 a nite sum of products of at least two terms (X
i
a
i
). Now let a k
n
be a point
on V , and consider the condition for acb kc
n
to be a point on V . When we substitute
a
i
cb
i
for X
i
in the above formula and take J a, we obtain:
J(a
1
cb
1
. . . . . a
n
cb
n
) =c
_
dJ
dX
i
a
b
i
_
.
Consequently, (a
1
cb
1
. . . . . a
n
cb
n
) lies on V if and only if (b
1
. . . . . b
n
) T
a
(V ) (orig
inal denition p95).
Geometrically, we can think of a point of V with coordinates in kc as being a point
of V with coordinates in k (the image of the point under V(kc) V(k)) together with a
tangent direction
110 5. LOCAL STUDY
REMARK 5.38. The description of the tangent space in terms of dual numbers is particu
larly convenient when our variety is given to us in terms of its points functor. For example,
let M
n
be the set of nn matrices, and let 1 be the identity matrix. Write e for 1 when it
is to be regarded as the identity element of GL
n
.
(a) A matrix 1 c has inverse 1 c in M
n
(kc), and so lies in GL
n
(kc). There
fore,
T
e
(GL
n
) ={1 c [ M
n
]
.M
n
(k).
(b) Since
det(1 c) =1 ctrace()
(using that c
2
=0),
T
e
(SL
n
) ={1 c [ trace() =0]
.{ M
n
(k) [ trace() =0].
(c) Assume the characteristic =2, and let O
n
be orthogonal group:
O
n
={ GL
n
[
tr
=1].
(
tr
denotes the transpose of ). This is the group of matrices preserving the quadratic form
X
2
1
X
2
n
. The determinant denes a surjective regular homomorphism det: O
n
{1], whose kernel is dened to be the special orthogonal group SO
n
. For 1 c
M
n
(kc),
(1 c)
tr
(1 c) =1 c
tr
c.
and so
T
e
(O
n
) =T
e
(SO
n
) ={1 c M
n
(kc) [ is skewsymmetric]
.{ M
n
(k) [ is skewsymmetric].
Note that, because an algebraic group is nonsingular, dimT
e
(G) =dimG this gives
a very convenient way of computing the dimension of an algebraic group.
ASIDE 5.39. On the tangent space T
e
(GL
n
) .M
n
of GL
n
, there is a bracket operation
M. N
def
=MN NM
which makes T
e
(GL
n
) into a Lie algebra. For any closed algebraic subgroup G of GL
n
,
T
e
(G) is stable under the bracket operation on T
e
(GL
n
) and is a subLiealgebra of M
n
,
which we denote Lie(G). The Lie algebra structure on Lie(G) is independent of the em
bedding of G into GL
n
(in fact, it has an intrinsic denition in terms of left invariant deriva
tions), and G Lie(G) is a functor from the category of linear algebraic groups to that of
Lie algebras.
This functor is not fully faithful, for example, any etale homomorphism G G
t
will
dene an isomorphism Lie(G) Lie(G
t
), but it is nevertheless very useful.
Assume k has characteristic zero. A connected algebraic group G is said to be semisim
ple if it has no closed connected solvable normal subgroup (except {e]). Such a group G
Tangent cones 111
may have a nite nontrivial centre 7(G), and we call two semisimple groups G and G
t
locally isomorphic if G,7(G) ~G
t
,7(G
t
). For example, SL
n
is semisimple, with centre
j
n
, the set of diagonal matrices diag(. . . . . ),
n
=1, and SL
n
,j
n
=PSL
n
. A Lie algebra
is semisimple if it has no commutative ideal (except {0]). One can prove that
G is semisimple Lie(G) is semisimple.
and the map G Lie(G) denes a onetoone correspondence between the set of local
isomorphism classes of semisimple algebraic groups and the set of isomorphism classes of
Lie algebras. The classication of semisimple algebraic groups can be deduced from that of
semisimple Lie algebras and a study of the nite coverings of semisimple algebraic groups
this is quite similar to the relation between Lie groups and Lie algebras.
Tangent cones
In this section, I assume familiarity with parts of Atiyah and MacDonald 1969, Chapters
11, 12.
Let V = V(a) k
n
, a = rad(a), and let 1 = (0. . . . . 0) V . Dene a
+
to be the
ideal generated by the polynomials J
+
for J a, where J
+
is the leading form of J (see
p93). The geometric tangent cone at 1, C
1
(V ) is V(a
+
), and the tangent cone is the pair
(V(a
+
). kX
1
. . . . . X
n
,a
+
). Obviously, C
1
(V ) T
1
(V ).
Computing the tangent cone
If a is principal, say a = (J), then a
+
= (J
+
), but if a = (J
1
. . . . . J
i
), then it need not
be true that a
+
= (J
1+
. . . . . J
i+
). Consider for example a = (XY. X7 7(Y
2
7
2
)).
One can show that this is a radical ideal either by asking Macaulay (assuming you believe
Macaulay), or by following the method suggested in Cox et al. 1992, p474, problem 3 to
show that it is an intersection of prime ideals. Since
Y7(Y
2
7
2
) =Y (X77(Y
2
7
2
)) 7 (XY ) a
and is homogeneous, it is in a
+
, but it is not in the ideal generated by XY , X7. In fact, a
+
is the ideal generated by
XY. X7. Y7(Y
2
7
2
).
This raises the following question: given a set of generators for an ideal a, how do you
nd a set of generators for a
+
? There is an algorithm for this in Cox et al. 1992, p467. Let
a be an ideal (not necessarily radical) such that V =V(a), and assume the origin is in V .
Introduce an extra variable T such that T > the remaining variables. Make each generator
of a homogeneous by multiplying its monomials by appropriate (small) powers of T , and
nd a Gr obner basis for the ideal generated by these homogeneous polynomials. Remove
T from the elements of the basis, and then the polynomials you get generate a
+
.
Intrinsic denition of the tangent cone
Let be a local ring with maximal ideal n. The associated graded ring is
gr() =
i_0
n
i
,n
i1
.
Note that if =T
m
and n =m, then gr() =
m
i
,m
i1
(because of (1.31)).
112 5. LOCAL STUDY
PROPOSITION 5.40. The map kX
1
. . . . . X
n
,a
+
gr(O
1
) sending the class of X
i
in
kX
1
. . . . . X
n
,a
+
to the class of X
i
in gr(O
1
) is an isomorphism.
PROOF. Let m be the maximal ideal in kX
1
. . . . . X
n
,a corresponding to 1. Then
gr(O
1
) =
m
i
,m
i1
=
(X
1
. . . . . X
n
)
i
,(X
1
. . . . . X
n
)
i1
a(X
1
. . . . . X
n
)
i
=
(X
1
. . . . . X
n
)
i
,(X
1
. . . . . X
n
)
i1
a
i
where a
i
is the homogeneous piece of a
+
of degree i (that is, the subspace of a
+
consisting
of homogeneous polynomials of degree i ). But
(X
1
. . . . . X
n
)
i
,(X
1
. . . . . X
n
)
i1
a
i
=i th homogeneous piece of kX
1
. . . . . X
n
,a
+
.
2
For a general variety V and 1 V , we dene the geometric tangent cone C
1
(V ) of
V at 1 to be Spm(gr(O
1
)
red
), where gr(O
1
)
red
is the quotient of gr(O
1
) by its nilradical,
and we dene the tangent cone to be (C
1
(V ). gr(O
1
)).
Recall (Atiyah and MacDonald 1969, 11.21) that dim() =dim(gr()). Therefore the
dimension of the geometric tangent cone at 1 is the same as the dimension of V (in contrast
to the dimension of the tangent space).
Recall (ibid., 11.22) that gr(O
1
) is a polynomial ring in J variables (J =dimV ) if and
only if O
1
is regular. Therefore, 1 is nonsingular if and only if gr(O
1
) is a polynomial
ring in J variables, in which case C
1
(V ) =T
1
(V ).
Using tangent cones, we can extend the notion of an etale morphism to singular vari
eties. Obviously, a regular map : V W induces a homomorphismgr(O
(1)
) gr(O
1
).
The map on the rings kX
1
. . . . . X
n
,a
+
dened by a map of algebraic varieties is
not the obvious one, i.e., it is not necessarily induced by the same map on polyno
mial rings as the original map. To see what it is, it is necessary to use Proposition 5.40, i.e.,
it is necessary to work with the rings gr(O
1
).
We say that is etale at 1 if this is an isomorphism. Note that then there is an iso
morphism of the geometric tangent cones C
1
(V ) C
(1)
(W), but this map may be an
isomorphism without being etale at 1. Roughly speaking, to be etale at 1, we need the
map on geometric tangent cones to be an isomorphism and to preserve the multiplicities
of the components.
It is a fairly elementary result that a local homomorphism of local rings : T
induces an isomorphism on the graded rings if and only if it induces an isomorphism on the
completions (ibid., 10.23). Thus : V W is etale at 1 if and only if the map
O
(1)
O
1
is an isomorphism. Hence (5.27) shows that the choice of a local system of parameters
1
. . . . .
d
at a nonsingular point 1 determines an isomorphism
O
1
kX
1
. . . . . X
d
.
We can rewrite this as follows: let t
1
. . . . . t
d
be a local system of parameters at a non
singular point 1; then there is a canonical isomorphism
O
1
kt
1
. . . . . t
d
. For
O
1
,
the image of kt
1
. . . . . t
d
 can be regarded as the Taylor series of .
For example, let V =A
1
, and let 1 be the point a. Then t =Xa is a local parameter
at a, O
1
consists of quotients (X) =g(X),h(X) with h(a) =0, and the coefcients of
the Taylor expansion
n_0
a
n
(Xa)
n
of (X) can be computed as in elementary calculus
courses: a
n
=
(n)
(a),n.
Exercises 113
Exercises
51. Find the singular points, and the tangent cones at the singular points, for each of
(a) Y
3
Y
2
X
3
X
2
3Y
2
X 3X
2
Y 2XY :
(b) X
4
Y
4
X
2
Y
2
(assume the characteristic is not 2).
52. Let V A
n
be an irreducible afne variety, and let 1 be a nonsingular point on V . Let
H be a hyperplane in A
n
(i.e., the subvariety dened by a linear equation
a
i
X
i
=J with
not all a
i
zero) passing through 1 but not containing T
1
(V ). Show that 1 is a nonsingular
point on each irreducible component of V H on which it lies. (Each irreducible compo
nent has codimension 1 in V you may assume this.) Give an example with H T
1
(V )
and 1 singular on V H. Must 1 be singular on V H if H T
1
(V )?
53. Let 1 and Q be points on varieties V and W. Show that
T
(1,Q)
(V W) =T
1
(V ) T
Q
(W).
54. For each n, show that there is a curve C and a point 1 on C such that the tangent
space to C at 1 has dimension n (hence C cant be embedded in A
n1
).
55. Let 1 be the nn identity matrix, and let J be the matrix
_
0 1
1 0
_
. The symplectic
group Sp
n
is the group of 2n2n matrices with determinant 1 such that
tr
J =J.
(It is the group of matrices xing a nondegenerate skewsymmetric form.) Find the tangent
space to Sp
n
at its identity element, and also the dimension of Sp
n
.
56. Find a regular map : V W which induces an isomorphism on the geometric tan
gent cones C
1
(V ) C
(1)
(W) but is not etale at 1.
57. Show that the cone X
2
Y
2
= 7
2
is a normal variety, even though the origin is
singular (characteristic =2). See p105.
58. Let V =V(a) A
n
. Suppose that a =1(V ), and for a V , let T
t
a
be the subspace
of T
a
(A
n
) dened by the equations (J )
a
=0, a. Clearly, T
t
a
T
a
(V ), but need they
always be different?
CHAPTER 6
Projective Varieties
Throughout this chapter, k will be an algebraically closed eld. Recall (4.3) that we dened
P
n
to be the set of equivalence classes in k
n1
{origin] for the relation
(a
0
. . . . . a
n
) ~(b
0
. . . . . b
n
) (a
0
. . . . . a
n
) =c(b
0
. . . . . b
n
) for some c k
.
Write (a
0
: . . . : a
n
) for the equivalence class of (a
0
. . . . . a
n
), and for the map
k
n1
{origin],~P
n
.
Let U
i
be the set of (a
0
: . . . : a
n
) P
n
such that a
i
=0, and let u
i
be the bijection
(a
0
: . . . : a
n
)
_
o
0
o
i
. . . . .
o
n
o
i
_
: U
i
A
n
(
o
i
o
i
omitted).
In this chapter, we shall dene on P
n
a (unique) structure of an algebraic variety for which
these maps become isomorphisms of afne algebraic varieties. A variety isomorphic to
a closed subvariety of P
n
is called a projective variety, and a variety isomorphic to a lo
cally closed subvariety of P
n
is called a quasiprojective variety.
1
Every afne variety is
quasiprojective, but there are many varieties that are not quasiprojective. We study mor
phisms between quasiprojective varieties.
Projective varieties are important for the same reason compact manifolds are important:
results are often simpler when stated for projective varieties, and the part at innity often
plays a role, even when we would like to ignore it. For example, a famous theorem of
Bezout (see 6.34 below) says that a curve of degree m in the projective plane intersects a
curve of degree n in exactly mn points (counting multiplicities). For afne curves, one has
only an inequality.
Algebraic subsets of P
n
A polynomial J(X
0
. . . . . X
n
) is said to be homogeneous of degree J if it is a sum of terms
a
i
0
,...,i
n
X
i
0
0
X
i
n
n
with i
0
i
n
=J; equivalently,
J(tX
0
. . . . . tX
n
) =t
d
J(X
0
. . . . . X
n
)
1
A subvariety of an afne variety is said to be quasiafne. For example, A
2
{(0. 0)] is quasiafne but
not afne.
115
116 6. PROJECTIVE VARIETIES
for all t k. Write kX
0
. . . . . X
n

d
for the subspace of kX
0
. . . . . X
n
 of polynomials of
degree J. Then
kX
0
. . . . . X
n
 =
d_0
kX
0
. . . . . X
n

d
:
that is, each polynomial J can be written uniquely as a sum J =
J
d
with J
d
homoge
neous of degree J.
Let 1 =(a
0
: . . . : a
n
) P
n
. Then 1 also equals (ca
0
: . . . : ca
n
) for any c k
, and
so we cant speak of the value of a polynomial J(X
0
. . . . . X
n
) at 1. However, if J is
homogeneous, then J(ca
0
. . . . . ca
n
) =c
d
J(a
0
. . . . . a
n
), and so it does make sense to say
that J is zero or not zero at 1. An algebraic set in P
n
(or projective algebraic set) is the
set of common zeros in P
n
of some set of homogeneous polynomials.
EXAMPLE 6.1. Consider the projective algebraic subset 1 of P
2
dened by the homoge
neous equation
Y
2
7 =X
3
aX7
2
b7
3
(14)
where X
3
aXb is assumed not to have multiple roots. It consists of the points (. : , : 1)
on the afne curve 1U
2
Y
2
=X
3
aX b.
together with the point at innity (0 : 1 : 0).
Curves dened by equations of the form (14) are called elliptic curves. They can also
be described as the curves of genus one, or as the abelian varieties of dimension one. Such
a curve becomes an algebraic group, with the group law such that 1 Q1 =0 if and
only if 1, Q, and 1 lie on a straight line. The zero for the group is the point at innity.
(Without the point at innity, it is not possible to make 1 into an algebraic group.)
When a. b , we can speak of the zeros of (14) with coordinates in . They also
form a group 1(), which Mordell showed to be nitely generated. It is easy to compute
the torsion subgroup of 1(), but there is at present no known algorithm for computing the
rank of 1(). More precisely, there is an algorithm which always works, but which has
not been proved to terminate after a nite amount of time, at least not in general. There is
a very beautiful theory surrounding elliptic curves over and other number elds, whose
origins can be traced back 1,800 years to Diophantus. (See my book on Elliptic Curves for
all of this.)
An ideal a kX
0
. . . . . X
n
 is said to be homogeneous if it contains with any polyno
mial J all the homogeneous components of J, i.e., if
J a == J
d
a, all J.
It is straightforward to check that
an ideal is homogeneous if and only if it is generated by (a nite set of) homogeneous
polynomials;
the radical of a homogeneous ideal is homogeneous;
an intersection, product, or sum of homogeneous ideals is homogeneous.
For a homogeneous ideal a, we write V(a) for the set of common zeros of the homoge
neous polynomials in a. If J
1
. . . . . J
i
are homogeneous generators for a, then V(a) is the
Algebraic subsets of P
n
117
set of common zeros of the J
i
. Clearly every polynomial in a is zero on every representa
tive of a point in V(a). We write V
aff
(a) for the set of common zeros of a in k
n1
. It is
cone in k
n1
, i.e., together with any point 1 it contains the line through 1 and the origin,
and
V(a) =(V
aff
(a) (0. . . . . 0)),~.
The sets V(a) have similar properties to their namesakes in A
n
.
PROPOSITION 6.2. There are the following relations:
(a) a b =V(a) V(b):
(b) V(0) =P
n
: V(a) = rad(a) (X
0
. . . . . X
n
):
(c) V(ab) =V(ab) =V(a) LV(b):
(d) V(
a
i
) =
_
V(a
i
).
PROOF. Statement (a) is obvious. For the second part of (b), note that
V(a) =0 V
aff
(a) {(0. . . . . 0)] rad(a) (X
0
. . . . . X
n
).
by the strong Nullstellensatz (2.11). The remaining statements can be proved directly, or
by using the relation between V(a) and V
aff
(a).
2
If C is a cone in k
n1
, then 1(C) is a homogeneous ideal in kX
0
. . . . . X
n
: if J(ca
0
. . . . . ca
n
) =
0 for all c k
, then
d
J
d
(a
0
. . . . . a
n
) c
d
=J(ca
0
. . . . . ca
n
) =0.
for innitely many c, and so
J
d
(a
0
. . . . . a
n
)X
d
is the zero polynomial. For a subset S
of P
n
, we dene the afne cone over S in k
n1
to be
C =
1
(S) L{origin]
and we set
1(S) =1(C).
Note that if S is nonempty and closed, then C is the closure of
1
(S) =0, and that 1(S)
is spanned by the homogeneous polynomials in kX
0
. . . . . X
n
 that are zero on S.
PROPOSITION 6.3. The maps V and 1 dene inverse bijections between the set of alge
braic subsets of P
n
and the set of proper homogeneous radical ideals of kX
0
. . . . . X
n
.
An algebraic set V in P
n
is irreducible if and only if 1(V ) is prime; in particular, P
n
is
irreducible.
PROOF. Note that we have bijections
{algebraic subsets of P
n
] {nonempty closed cones in k
n1
]
{proper homogeneous radical ideals in kX
0
. . . . . X
n
]
S C
V J
118 6. PROJECTIVE VARIETIES
Here the top map sends S to the afne cone over S, and the maps V and 1 are in the sense
of projective geometry and afne geometry respectively. The composite of any three of
these maps is the identity map, which proves the rst statement because the composite of
the top map with 1 is 1 in the sense of projective geometry. Obviously, V is irreducible if
and only if the closure of
1
(V ) is irreducible, which is true if and only if 1(V ) is a prime
ideal.
2
Note that (X
0
. . . . . X
n
) and kX
0
. . . . . X
n
 are both radical homogeneous ideals, but
V(X
0
. . . . . X
n
) =0 =V(kX
0
. . . . . X
n
)
and so the correspondence between irreducible subsets of P
n
and radical homogeneous
ideals is not quite onetoone.
The Zariski topology on P
n
Proposition 6.2 shows that the projective algebraic sets are the closed sets for a topology on
P
n
. In this section, we verify that it agrees with that dened in the rst paragraph of this
chapter. For a homogeneous polynomial J, let
D(J) ={1 P
n
[ J(1) =0].
Then, just as in the afne case, D(J) is open and the sets of this type form a base for the
topology of P
n
.
To each polynomial (X
1
. . . . . X
n
), we attach the homogeneous polynomial of the same
degree
+
(X
0
. . . . . X
n
) =X
deg(( )
0
_
A
1
A
0
. . . . .
A
n
A
0
_
.
and to each homogeneous polynomial J(X
0
. . . . . X
n
), we attach the polynomial
J
+
(X
1
. . . . . X
n
) =J(1. X
1
. . . . . X
n
).
PROPOSITION 6.4. For the topology on P
n
just dened, each U
i
is open, and when we
endow it with the induced topology, the bijection
U
i
A
n
, (a
0
: . . . : 1 : . . . : a
n
) (a
0
. . . . . a
i1
. a
i1
. . . . . a
n
)
becomes a homeomorphism.
PROOF. It sufces to prove this with i =0. The set U
0
=D(X
0
), and so it is a basic open
subset in P
n
. Clearly, for any homogeneous polynomial J kX
0
. . . . . X
n
,
D(J(X
0
. . . . . X
n
)) U
0
=D(J(1. X
1
. . . . . X
n
)) =D(J
+
)
and, for any polynomial kX
1
. . . . . X
n
,
D( ) =D(
+
) U
0
.
Thus, under U
0
A
n
, the basic open subsets of A
n
correspond to the intersections with U
i
of the basic open subsets of P
n
, which proves that the bijection is a homeomorphism.
2
Closed subsets of A
n
and P
n
119
REMARK 6.5. It is possible to use this to give a different proof that P
n
is irreducible. We
apply the criterion that a space is irreducible if and only if every nonempty open subset is
dense (see p48). Note that each U
i
is irreducible, and that U
i
U
}
is open and dense in
each of U
i
and U
}
(as a subset of U
i
, it is the set of points (a
0
: . . . : 1 : . . . : a
}
: . . . : a
n
)
with a
}
= 0). Let U be a nonempty open subset of P
n
; then U U
i
is open in U
i
. For
some i , U U
i
is nonempty, and so must meet U
i
U
}
. Therefore U meets every U
}
, and
so is dense in every U
}
. It follows that its closure is all of P
n
.
Closed subsets of A
n
and P
n
We identify A
n
with U
0
, and examine the closures in P
n
of closed subsets of A
n
. Note that
P
n
=A
n
LH
o
. H
o
=V(X
0
).
With each ideal a in kX
1
. . . . . X
n
, we associate the homogeneous ideal a
+
in kX
0
. . . . . X
n

generated by {
+
[ a]. For a closed subset V of A
n
, set V
+
=V(a
+
) with a =1(V ).
With each homogeneous ideal a in kX
0
. X
1
. . . . . X
n
], we associate the ideal a
+
in
kX
1
. . . . . X
n
 generated by {J
+
[ J a]. When V is a closed subset of P
n
, we set V
+
=
V(a
+
) with a =1(V ).
PROPOSITION 6.6. (a) Let V be a closed subset of A
n
. Then V
+
is the closure of V in P
n
,
and (V
+
)
+
=V . If V =
_
V
i
is the decomposition of V into its irreducible components,
then V
+
=
_
V
+
i
is the decomposition of V
+
into its irreducible components.
(b) Let V be a closed subset of P
n
. Then V
+
=V A
n
, and if no irreducible component
of V lies in H
o
or contains H
o
, then V
+
is a proper subset of A
n
, and (V
+
)
+
=V .
PROOF. Straightforward.
2
EXAMPLE 6.7. (a) For
V : Y
2
=X
3
aX b.
we have
V
+
: Y
2
7 =X
3
aX7
2
b7
3
.
and (V
+
)
+
=V .
(b) Let V =V(
1
. . . . .
n
); then the closure of V in P
n
is the union of the irreducible
components of V(
+
1
. . . . .
+
n
) not contained in H
o
. For example, let V = V(X
1
. X
2
1
X
2
) ={(0. 0)]; then V(X
0
X
1
. X
2
1
X
0
X
2
) consists of the two points (1: 0: 0) (the closure
of V ) and (0: 0: 1) (which is contained in H
o
).
2
(c) For V =H
o
=V(X
0
), V
+
=0 =V(1) and (V
+
)
+
=0 =V .
The hyperplane at innity
It is often convenient to think of P
n
as being A
n
=U
0
with a hyperplane added at innity.
More precisely, identify the U
0
with A
n
. The complement of U
0
in P
n
is
H
o
={(0 : a
1
: . . . : a
n
) P
n
].
2
Of course, in this case a =(X
1
. X
2
), a
+
=(X
1
. X
2
), and V
+
={(1: 0: 0)], and so this example doesnt
contradict the proposition.
120 6. PROJECTIVE VARIETIES
which can be identied with P
n1
.
For example, P
1
= A
1
LH
o
(disjoint union), with H
o
consisting of a single point,
and P
2
=A
2
LH
o
with H
o
a projective line. Consider the line
1aX
1
bX
2
=0
in A
2
. Its closure in P
2
is the line
X
0
aX
1
bX
2
=0.
This line intersects the line H
o
= V(X
0
) at the point (0 : b : a), which equals (0 : 1 :
a,b) when b = 0. Note that a,b is the slope of the line 1 aX
1
bX
2
= 0, and so
the point at which a line intersects H
o
depends only on the slope of the line: parallel lines
meet in one point at innity. We can think of the projective plane P
2
as being the afne
plane A
2
with one point added at innity for each direction in A
2
.
Similarly, we can think of P
n
as being A
n
with one point added at innity for each
direction in A
n
being parallel is an equivalence relation on the lines in A
n
, and there is
one point at innity for each equivalence class of lines.
We can also identify A
n
with U
n
, as in Example 6.1. Note that in this case the point
at innity on the elliptic curve Y
2
=X
3
aX b is the intersection of the closure of any
vertical line with H
o
.
P
n
is an algebraic variety
For each i , write O
i
for the sheaf on U
i
P
n
dened by the homeomorphismu
i
: U
i
A
n
.
LEMMA 6.8. Write U
i}
=U
i
U
}
; then O
i
[U
i}
=O
}
[U
i}
. When endowed with this sheaf
U
i}
is an afne variety; moreover, 1(U
i}
. O
i
) is generated as a kalgebra by the functions
( [U
i}
)(g[U
i}
) with 1(U
i
. O
i
), g 1(U
}
. O
}
).
PROOF. It sufces to prove this for (i. ) =(0. 1). All rings occurring in the proof will be
identied with subrings of the eld k(X
0
. X
1
. . . . . X
n
).
Recall that
U
0
={(a
0
: a
1
: . . . : a
n
) [ a
0
=0]; (a
0
: a
1
: . . . : a
n
) (
o
1
o
0
.
o
2
o
0
. . . . .
o
n
o
0
) A
n
.
Let k
A
1
A
0
.
A
2
A
0
. . . . .
A
n
A
0
 be the subring of k(X
0
. X
1
. . . . . X
n
) generated by the quotients
A
i
A
0
it is the polynomial ring in the n symbols
A
1
A
0
. . . . .
A
n
A
0
. An element (
A
1
A
0
. . . . .
A
n
A
0
)
k
A
1
A
0
. . . . .
A
n
A
0
 denes a map
(a
0
: a
1
: . . . : a
n
) (
o
1
o
0
. . . . .
o
n
o
0
): U
0
k.
and in this way k
A
1
A
0
.
A
2
A
0
. . . . .
A
n
A
0
 becomes identied with the ring of regular functions on
U
0
. and U
0
with Spm
_
k
A
1
A
0
. . . . .
A
n
A
0

_
.
Next consider the open subset of U
0
.
U
01
={(a
0
: . . . : a
n
) [ a
0
=0, a
1
=0].
P
n
is an algebraic variety 121
It is D(
A
1
A
0
), and is therefore an afne subvariety of (U
0
. O
0
). The inclusion U
01
U
0
corresponds to the inclusion of rings k
A
1
A
0
. . . . .
A
n
A
0
 k
A
1
A
0
. . . . .
A
n
A
0
.
A
0
A
1
. An element
(
A
1
A
0
. . . . .
A
n
A
0
.
A
0
A
1
) of k
A
1
A
0
. . . . .
A
n
A
0
.
A
0
A
1
 denes the function (a
0
: . . . : a
n
) (
o
1
o
0
. . . . .
o
n
o
0
.
o
0
o
1
)
on U
01
.
Similarly,
U
1
={(a
0
: a
1
: . . . : a
n
) [ a
1
=0]; (a
0
: a
1
: . . . : a
n
) (
o
0
o
1
. . . . .
o
n
o
1
) A
n
.
and we identify U
1
with Spm
_
k
A
0
A
1
.
A
2
A
0
. . . . .
A
n
A
1

_
. Apolynomial (
A
0
A
1
. . . . .
A
n
A
1
) in k
A
0
A
1
. . . . .
A
n
A
1

denes the map (a
0
: . . . : a
n
) (
o
0
o
1
. . . . .
o
n
o
1
): U
1
k.
When regarded as an open subset of U
1
. U
01
=D(
A
0
A
1
), and is therefore an afne sub
variety of (U
1
. O
1
), and the inclusion U
01
U
1
corresponds to the inclusion of rings
k
A
0
A
1
. . . . .
A
n
A
1
 k
A
0
A
1
. . . . .
A
n
A
1
.
A
1
A
0
. An element (
A
0
A
1
. . . . .
A
n
A
1
.
A
1
A
0
) of k
A
0
A
1
. . . . .
A
n
A
1
.
A
1
A
0

denes the function (a
0
: . . . : a
n
) (
o
0
o
1
. . . . .
o
n
o
1
.
o
1
o
0
) on U
01
.
The two subrings k
A
1
A
0
. . . . .
A
n
A
0
.
A
0
A
1
 and k
A
0
A
1
. . . . .
A
n
A
1
.
A
1
A
0
 of k(X
0
. X
1
. . . . . X
n
) are
equal, and an element of this ring denes the same function on U
01
regardless of which of
the two rings it is considered an element. Therefore, whether we regard U
01
as a subvariety
of U
0
or of U
1
it inherits the same structure as an afne algebraic variety (3.8a). This
proves the rst two assertions, and the third is obvious: k
A
1
A
0
. . . . .
A
n
A
0
.
A
0
A
1
 is generated by
its subrings k
A
1
A
0
. . . . .
A
n
A
0
 and k
A
0
A
1
.
A
2
A
1
. . . . .
A
n
A
1
.
2
PROPOSITION 6.9. There is a unique structure of a (separated) algebraic variety on P
n
for
which each U
i
is an open afne subvariety of P
n
and each map u
i
is an isomorphism of
algebraic varieties.
PROOF. Endow each U
i
with the structure of an afne algebraic variety for which u
i
is
an isomorphism. Then P
n
=
_
U
i
, and the lemma shows that this covering satises the
patching condition (4.13), and so P
n
has a unique structure of a ringed space for which
U
i
P
n
is a homeomorphism onto an open subset of P
n
and O
P
n[U
i
=O
U
i
. Moreover,
because each U
i
is an algebraic variety, this structure makes P
n
into an algebraic prevariety.
Finally, the lemma shows that P
n
satises the condition (4.27c) to be separated.
2
EXAMPLE 6.10. Let C be the plane projective curve
C: Y
2
7 =X
3
and assume char(k) =2. For each a k
, there is an automorphism
(. : , : :) (a. : , : a
3
:): C
a
C.
Patch two copies of C A
1
together along C (A
1
{0]) by identifying (1. u) with
(
o
(1). a
1
), 1 C, a A
1
{0]. One obtains in this way a singular 2dimensional
variety that is not quasiprojective (see Hartshorne 1977, Exercise 7.13). It is even complete
see below and so if it were quasiprojective, it would be projective. It is known that
every irreducible separated curve is quasiprojective, and every nonsingular complete sur
face is projective, and so this is an example of minimum dimension. In Shafarevich 1994,
VI 2.3, there is an example of a nonsingular complete variety of dimension 3 that is not
projective.
122 6. PROJECTIVE VARIETIES
The homogeneous coordinate ring of a subvariety of P
n
Recall (page 50) that we attached to each irreducible variety V a eld k(V ) with the prop
erty that k(V ) is the eld of fractions of kU for any open afne U V . We now describe
this eld in the case that V =P
n
. Recall that kU
0
 =k
A
1
A
0
. . . . .
A
n
A
0
. We regard this as a
subring of k(X
0
. . . . . X
n
), and wish to identify the eld of fractions of kU
0
 as a subeld
of k(X
0
. . . . . X
n
). Any nonzero J kU
0
 can be written
J(
A
1
A
0
. . . . .
A
n
A
0
) =
J
+
(X
0
. . . . . X
n
)
X
deg(T)
0
with J
+
homogeneous of degree deg(J), and it follows that the eld of fractions of kU
0

is
k(U
0
) =
_
G(X
0
. . . . . X
n
)
H(X
0
. . . . . X
n
)
g
i
, with g
i
represented by
the homogeneous polynomial G
i
of degree i . Then J G p, and the homogeneity of p
implies that J
i
G
i
=(J G)
i
p. Therefore
i
=g
i
.
2
It therefore makes sense to speak of homogeneous elements of kV . For such an ele
ment h, we dene D(h) ={1 V [ h(1) =0].
Since k
hom
V  is an integral domain, we can form its eld of fractions k
hom
(V ). Dene
k
hom
(V )
0
=
_
g
h
k
hom
(V )
1
and
Q
dene the same function on some open afne U
t
, and so
1
=
Q
as elements
of kU
t
 k(V ). This shows that is the function dened by
1
on the whole of U.
2
REMARK 6.14. (a) The elements of k(V ) =k
hom
(V )
0
should be regarded as the algebraic
analogues of meromorphic functions on a complex manifold; the regular functions on an
open subset U of V are the meromorphic functions without poles on U. [In fact, when
k =C, this is more than an analogy: a nonsingular projective algebraic variety over C de
nes a complex manifold, and the meromorphic functions on the manifold are precisely the
3
Unless k
hom
V  is a unique factorization domain, there will be no preferred representation =
h
.
124 6. PROJECTIVE VARIETIES
rational functions on the variety. For example, the meromorphic functions on the Riemann
sphere are the rational functions in :.]
(b) We shall see presently (6.21) that, for any nonzero homogeneous h k
hom
V , D(h)
is an open afne subset of V . The ring of regular functions on it is
kD(h) ={g,h
n
[ g homogeneous of degree mdeg(h)] L{0].
We shall also see that the ring of regular functions on V itself is just k, i.e., any regular
function on an irreducible (connected will do) projective variety is constant. However, if U
is an open nonafne subset of V , then the ring 1(U. O
V
) of regular functions can be almost
anything it neednt even be a nitely generated kalgebra!
Morphisms from projective varieties
We describe the morphisms from a projective variety to another variety.
PROPOSITION 6.15. The map
: A
n1
{origin] P
n
, (a
0
. . . . . a
n
) (a
0
: . . . : a
n
)
is an open morphism of algebraic varieties. A map : P
n
V with V a prevariety is regular
if and only if is regular.
PROOF. The restriction of to D(X
i
) is the projection
(a
0
. . . . . a
n
) (
o
0
o
i
: . . . :
o
n
o
i
): k
n1
V(X
i
) U
i
.
which is the regular map of afne varieties corresponding to the map of kalgebras
k
_
A
0
A
i
. . . . .
A
n
A
i
_
kX
0
. . . . . X
n
X
1
i
.
(In the rst algebra
A
j
A
i
is to be thought of as a single symbol.) It now follows from (4.4)
that is regular.
Let U be an open subset of k
n1
{origin], and let U
t
be the union of all the lines
through the origin that meet U, that is, U
t
=
1
(U). Then U
t
is again open in k
n1
{origin], because U
t
=
_
cU, c k
, and . c. is an automorphism of k
n1
{origin].
The complement 7 of U
t
in k
n1
{origin] is a closed cone, and the proof of (6.3) shows
that its image is closed in P
n
; but (U) is the complement of (7). Thus sends open
sets to open sets.
The rest of the proof is straightforward.
2
Thus, the regular maps P
n
V are just the regular maps A
n1
{origin] V factor
ing through P
n
(as maps of sets).
REMARK 6.16. Consider polynomials J
0
(X
0
. . . . . X
n
). . . . . J
n
(X
0
. . . . . X
n
) of the same
degree. The map
(a
0
: . . . : a
n
) (J
0
(a
0
. . . . . a
n
) : . . . : J
n
(a
0
. . . . . a
n
))
Examples of regular maps of projective varieties 125
obviously denes a regular map to P
n
on the open subset of P
n
where not all J
i
vanish,
that is, on the set
_
D(J
i
) =P
n
V(J
1
. . . . . J
n
). Its restriction to any subvariety V of P
n
will also be regular. It may be possible to extend the map to a larger set by representing
it by different polynomials. Conversely, every such map arises in this way, at least locally.
More precisely, there is the following result.
PROPOSITION 6.17. Let V = V(a) P
n
and W = V(b) P
n
. A map : V W is
regular if and only if, for every 1 V , there exist polynomials
J
0
(X
0
. . . . . X
n
). . . . . J
n
(X
0
. . . . . X
n
).
homogeneous of the same degree, such that
((b
0
: . . . : b
n
)) =(J
0
(b
0
. . . . . b
n
) : . . . : J
n
(b
0
. . . . . b
n
))
for all points (b
0
: . . . : b
n
) in some neighbourhood of 1 in V(a).
PROOF. Straightforward.
2
EXAMPLE 6.18. We prove that the circle X
2
Y
2
=7
2
is isomorphic to P
1
. This equa
tion can be rewritten (X iY )(X iY ) = 7
2
, and so, after a change of variables, the
equation of the circle becomes C : X7 =Y
2
. Dene
: P
1
C, (a : b) (a
2
: ab : b
2
).
For the inverse, dene
[: C P
1
by
_
(a : b : c) (a : b) if a =0
(a : b : c) (b : c) if b =0
.
Note that,
a =0 =b. ac =b
2
==
c
b
=
b
a
and so the two maps agree on the set where they are both dened. Clearly, both and [
are regular, and one checks directly that they are inverse.
Examples of regular maps of projective varieties
We list some of the classic maps.
EXAMPLE 6.19. Let 1 =
c
i
X
i
be a nonzero linear form in n1 variables. Then the
map
(a
0
: . . . : a
n
)
_
a
0
1(a)
. . . . .
a
n
1(a)
_
is a bijection of D(1) P
n
onto the hyperplane 1(X
0
. X
1
. . . . . X
n
) = 1 of A
n1
, with
inverse
(a
0
. . . . . a
n
) (a
0
: . . . : a
n
).
Both maps are regular for example, the components of the rst map are the regular
functions
A
j
c
i
A
i
. As V(11) is afne, so also is D(1), and its ring of regular functions
126 6. PROJECTIVE VARIETIES
is k
A
0
c
i
A
i
. . . . .
A
n
c
i
A
i
. In this ring, each quotient
A
j
c
i
A
i
is to be thought of as a single
symbol, and
c
}
A
j
c
i
A
i
= 1; thus it is a polynomial ring in n symbols; any one symbol
A
j
c
i
A
i
for which c
}
=0 can be omitted (see Lemma 5.12).
For a xed 1 =(a
0
: . . . : a
n
) P
n
, the set of c =(c
0
: . . . : c
n
) such that
1
c
(1)
def
=
c
i
a
i
=0
is a nonempty open subset of P
n
(n >0). Therefore, for any nite set S of points of P
n
,
{c P
n
[ S D(1
c
)]
is a nonempty open subset of P
n
(because P
n
is irreducible). In particular, S is contained
in an open afne subset D(1
c
) of P
n
. Moreover, if S V where V is a closed subvariety
of P
n
, then S V D(1
c
): any nite set of points of a projective variety is contained in
an open afne subvariety.
EXAMPLE 6.20. (The Veronese map.) Let
1 ={(i
0
. . . . . i
n
) N
n1
[
i
}
=m].
Note that 1 indexes the monomials of degree m in n1 variables. It has
_
nn
n
_
elements
4
.
Write v
n,n
=
_
nn
n
_
1, and consider the projective space P
n;m
whose coordinates are
indexed by 1; thus a point of P
n;m
can be written (. . . : b
i
0
...i
n
: . . .). The Veronese mapping
is dened to be
: P
n
P
n;m
, (a
0
: . . . : a
n
) (. . . : b
i
0
...i
n
: . . .). b
i
0
...i
n
=a
i
0
0
. . . a
i
n
n
.
In other words, the Veronese mapping sends an n1tuple (a
0
: . . . : a
n
) to the set of mono
mials in the a
i
of degre m. For example, when n =1 and m=2, the Veronese map is
P
1
P
2
, (a
0
: a
1
) (a
2
0
: a
0
a
1
: a
2
1
).
Its image is the curve v(P
1
) : X
0
X
2
=X
2
1
, and the map
(b
2,0
: b
1,1
: b
0,2
)
_
(b
2,0
: b
1,1
) if b
2,0
=1
(b
1,1
: b
0,2
) if b
0,2
=0.
is an inverse v(P
1
) P
1
. (Cf. Example 6.19.)
5
4
This can be proved by induction on mn. If m = 0 = n, then
_
0
0
_
= 1, which is correct. A general
homogeneous polynomial of degree m can be written uniquely as
J(X
0
. X
1
. . . . . X
n
) =J
1
(X
1
. . . . . X
n
) X
0
J
2
(X
0
. X
1
. . . . . X
n
)
with J
1
homogeneous of degree m and J
2
homogeneous of degree m1. But
_
nn
n
_
=
_
nn1
n
_
_
nn1
n1
_
because they are the coefcients of X
n
in
(X 1)
nn
=(X 1)(X 1)
nn1
.
and this proves the induction.
5
Note that, although P
1
and v(P
1
) are isomorphic, their homogeneous coordinate rings are not. In fact
k
hom
P
1
 =kX
0
. X
1
, which is the afne coordinate ring of the smooth variety A
2
, whereas k
hom
v(P
1
) =
kX
0
. X
1
. X
2
,(X
0
X
2
X
2
1
) which is the afne coordinate ring of the singular variety X
0
X
2
X
2
1
.
Examples of regular maps of projective varieties 127
When n =1 and m is general, the Veronese map is
P
1
P
n
, (a
0
: a
1
) (a
n
0
: a
n1
0
a
1
: . . . : a
n
1
).
I claim that, in the general case, the image of v is a closed subset of P
n;m
and that v
denes an isomorphism of projective varieties v: P
n
v(P
n
).
First note that the map has the following interpretation: if we regard the coordinates a
i
of a point 1 of P
n
as being the coefcients of a linear form 1 =
a
i
X
i
(welldened up
to multiplication by nonzero scalar), then the coordinates of v(1) are the coefcients of the
homogeneous polynomial 1
n
with the binomial coefcients omitted.
As 1 =0 =1
n
=0, the map v is dened on the whole of P
n
, that is,
(a
0
. . . . . a
n
) =(0. . . . . 0) =(. . . . b
i
0
...i
n
. . . .) =(0. . . . . 0).
Moreover, 1
1
= c1
2
=1
n
1
= c1
n
2
, because kX
0
. . . . . X
n
 is a unique factorization do
main, and so v is injective. It is clear from its denition that v is regular.
We shall see later in this chapter that the image of any projective variety under a regular
map is closed, but in this case we can prove directly that v(P
n
) is dened by the system of
equations:
b
i
0
...i
n
b
}
0
...}
n
=b
k
0
...k
n
b
I
0
...I
n
. i
h
h
=k
h
h
, all h (*).
Obviously P
n
maps into the algebraic set dened by these equations. Conversely, let
V
i
={(. . . . : b
i
0
...i
n
: . . .) [ b
0...0n0...0
=0].
Then v(U
i
) V
i
and v
1
(V
i
) = U
i
. It is possible to write down a regular map V
i
U
i
inverse to v[U
i
: for example, dene V
0
P
n
to be
(. . . : b
i
0
...i
n
: . . .) (b
n,0,...,0
: b
n1,1,0,...,0
: b
n1,0,1,0,...,0
: . . . : b
n1,0,...,0,1
).
Finally, one checks that v(P
n
)
_
V
i
.
For any closed variety W P
n
, v[W is an isomorphism of W onto a closed subvariety
v(W) of v(P
n
) P
n;m
.
REMARK 6.21. The Veronese mapping has a very important property. If J is a nonzero
homogeneous form of degree m_1, then V(J) P
n
is called a hypersurface of degree m
and V(J) W is called a hypersurface section of the projective variety W. When m=1,
surface is replaced by plane.
Now let H be the hypersurface in P
n
of degree m
a
i
0
...i
n
X
i
0
0
X
i
n
n
=0,
and let 1 be the hyperplane in P
n;m
dened by
a
i
0
...i
n
X
i
0
...i
n
.
Then v(H) =v(P
n
) 1, i.e.,
H(a) =0 1(v(a)) =0.
128 6. PROJECTIVE VARIETIES
Thus for any closed subvariety W of P
n
, v denes an isomorphism of the hypersurface
section W H of V onto the hyperplane section v(W)1of v(W). This observation often
allows one to reduce questions about hypersurface sections to questions about hyperplane
sections.
As one example of this, note that v maps the complement of a hypersurface section of
W isomorphically onto the complement of a hyperplane section of v(W), which we know
to be afne. Thus the complement of any hypersurface section of a projective variety is an
afne varietywe have proved the statement in (6.14b).
EXAMPLE 6.22. An element =(a
i}
) of GL
n1
denes an automorphism of P
n
:
(.
0
: . . . : .
n
) (. . . :
a
i}
.
}
: . . .):
clearly it is a regular map, and the inverse matrix gives the inverse map. Scalar matrices act
as the identity map.
Let PGL
n1
= GL
n1
,k
a
i
X
i
and 1
2
=
b
}
Y
}
, then the image of the pair is the set of coefcients of the homogeneous form of
degree 2, 1
1
1
2
. From this observation, it is obvious that the map is dened on the whole of
P
n
P
n
(1
1
=0 =1
2
=1
1
1
2
=0) and is injective. On any subset of the formU
i
U
}
it
is dened by polynomials, and so it is regular. Again one can show that it is an isomorphism
onto its image, which is the closed subset of P
nnnn
dened by the equations
n
i}
n
kI
n
iI
n
k}
=0
see Shafarevich 1994, I 5.1. For example, the map
((a
0
: a
1
). (b
0
: b
1
)) (a
0
b
0
: a
0
b
1
: a
1
b
0
: a
1
b
1
): P
1
P
1
P
3
has image the hypersurface
H : W7 =XY.
The map
(n : . : , : :) ((n : ,). (n : .))
is an inverse on the set where it is dened. [Incidentally, P
1
P
1
is not isomorphic to
P
2
, because in the rst variety there are closed curves, e.g., two vertical lines, that dont
intersect.]
If V and W are closed subvarieties of P
n
and P
n
, then the Segre map sends V W
isomorphically onto a closed subvariety of P
nnnn
. Thus products of projective varieties
are projective.
There is an explicit description of the topology on P
n
P
n
: the closed sets are the sets
of common solutions of families of equations
J(X
0
. . . . . X
n
: Y
0
. . . . . Y
n
) =0
with J separately homogeneous in the Xs and in the Y s.
EXAMPLE 6.24. Let 1
1
. . . . . 1
nd
be linearly independent linear forms in n1 variables.
Their zero set 1 in k
n1
has dimension J 1, and so their zero set in P
n
is a Jdimensional
linear space. Dene : P
n
1 P
nd1
by (a) =(1
1
(a) : . . . : 1
nd
(a)); such a map
is called a projection with centre 1. If V is a closed subvariety disjoint from 1, then
denes a regular map V P
nd1
. More generally, if J
1
. . . . . J
i
are homogeneous forms
of the same degree, and 7 =V(J
1
. . . . . J
i
), then a (J
1
(a) : . . . : J
i
(a)) is a morphism
P
n
7 P
i1
.
By carefully choosing the centre 1, it is possible to linearly project any smooth curve
in P
n
isomorphically onto a curve in P
3
, and nonisomorphically (but bijectively on an open
subset) onto a curve in P
2
with only nodes as singularities.
7
For example, suppose we have
7
A nonsingular curve of degree J in P
2
has genus
(d1)(d2)
2
. Thus, if g is not of this form, a curve of
genus g cant be realized as a nonsingular curve in P
2
.
130 6. PROJECTIVE VARIETIES
a nonsingular curve C in P
3
. To project to P
2
we need three linear forms 1
0
, 1
1
, 1
2
and
the centre of the projection is the point 1
0
where all forms are zero. We can think of the
map as projecting from the centre 1
0
onto some (projective) plane by sending the point 1
to the point where 1
0
1 intersects the plane. To project C to a curve with only ordinary
nodes as singularities, one needs to choose 1
0
so that it doesnt lie on any tangent to C, any
trisecant (line crossing the curve in 3 points), or any chord at whose extremities the tangents
are coplanar. See for example Samuel, P., Lectures on Old and New Results on Algebraic
Curves, Tata Notes, 1966.
Projecting a nonsingular variety in P
n
to a lower dimensional projective space usually
introduces singularities; Hironaka proved that every singular variety arises in this way in
characteristic zero.
PROPOSITION 6.25. Every nite set S of points of a quasiprojective variety V is contained
in an open afne subset of V .
PROOF. Regard V as a subvariety of P
n
, let
V be the closure of V in P
n
, and let 7=
V V .
Because S 7 = 0, for each 1 S there exists a homogeneous polynomial J
1
1(7)
such that J
1
(1) =0. We may suppose that the J
1
s have the same degree. An elementary
argument shows that some linear combination J of the J
1
, 1 S, is nonzero at each 1.
Then J is zero on 7, and so
V D(J) is an open afne of V , but J is nonzero at each 1,
and so
V D(J) contains S.
2
Projective space without coordinates
Let 1 be a vector space over k of dimension n. The set P(1) of lines through zero in 1
has a natural structure of an algebraic variety: the choice of a basis for 1 denes a bijection
P(1) P
n
, and the inherited structure of an algebraic variety on P(1) is independent of
the choice of the basis (because the bijections dened by two different bases differ by an au
tomorphism of P
n
). Note that in contrast to P
n
, which has n1 distinguished hyperplanes,
namely, X
0
=0. . . . . X
n
=0, no hyperplane in P(1) is distinguished.
Grassmann varieties
Let 1 be a vector space over k of dimension n, and let G
d
(1) be the set of Jdimensional
subspaces of 1. When J =0 or n, G
d
(1) has a single element, and so from now on we
assume that 0 < J < n. Fix a basis for 1, and let S G
d
(1). The choice of a basis for S
then determines a J n matrix (S) whose rows are the coordinates of the basis elements.
Changing the basis for S multiplies (S) on the left by an invertible J J matrix. Thus, the
family of J J minors of (S) is determined up to multiplication by a nonzero constant,
and so denes a point 1(S) in P
_
n
d
_
1
.
PROPOSITION 6.26. The map S 1(S): G
d
(1) P
_
n
d
_
1
is injective, with image a
closed subset of P
_
n
d
_
1
.
We give the proof below. The maps 1 dened by different bases of 1 differ by an
automorphism of P
_
n
d
_
1
, and so the statement is independent of the choice of the basis
Grassmann varieties 131
later (6.31) we shall give a coordinatefree description of the map. The map realizes
G
d
(1) as a projective algebraic variety called the Grassmann variety of Jdimensional
subspaces of 1.
EXAMPLE 6.27. The afne cone over a line in P
3
is a twodimensional subspace of k
4
.
Thus, G
2
(k
4
) can be identied with the set of lines in P
3
. Let 1 be a line in P
3
, and let
x =(.
0
: .
1
: .
2
: .
3
) and y =(,
0
: ,
1
: ,
2
: ,
3
) be distinct points on 1. Then
1(1) =(
01
:
02
:
03
:
12
:
13
:
23
) P
5
.
i}
def
=
.
i
.
}
,
i
,
}
.
depends only on 1. The map 1 1(1) is a bijection from G
2
(k
4
) onto the quadric
: X
01
X
23
X
02
X
13
X
03
X
12
=0
in P
5
. For a direct elementary proof of this, see (10.20, 10.21) below.
REMARK 6.28. Let S
t
be a subspace of 1 of complementary dimension n J, and let
G
d
(1)
S
0 be the set of S G
d
(V ) such that S S
t
={0]. Fix an S
0
G
d
(1)
S
0 , so that
1 =S
0
S
t
. For any S G
d
(V )
S
0 , the projection S S
0
given by this decomposition is
an isomorphism, and so S is the graph of a homomorphism S
0
S
t
:
s s
t
(s. s
t
) S.
Conversely, the graph of any homomorphism S
0
S
t
lies in G
d
(V )
S
0 . Thus,
G
d
(V )
S
0 ~Hom(S
0
. S
t
) ~Hom(1,S
t
. S
t
). (15)
The isomorphism G
d
(V )
S
0 ~ Hom(1,S
t
. S
t
) depends on the choice of S
0
it is the
element of G
d
(V )
S
0 corresponding to 0 Hom(1,S
t
. S
t
). The decomposition 1 =S
0
S
t
gives a decomposition End(1) =
_
End(S
0
) Hom(S
0
,S
0
)
Hom(S
0
,S
0
) End(S
0
)
_
, and the bijections (15) show
that the group
_
1 0
Hom(S
0
,S
0
) 1
_
acts simply transitively on G
d
(1)
S
0 .
REMARK 6.29. The bijection (15) identies G
d
(1)
S
0 with the afne variety A(Hom(S
0
. S
t
))
dened by the vector space Hom(S
0
. S
t
) (cf. p67). Therefore, the tangent space to G
d
(1)
at S
0
,
T
S
0
(G
d
(1)) .Hom(S
0
. S
t
) .Hom(S
0
. 1,S
0
). (16)
Since the dimension of this space doesnt depend on the choice of S
0
, this shows that
G
d
(1) is nonsingular (5.19).
REMARK 6.30. Let T be the set of all bases of 1. The choice of a basis for 1 identi
es T with GL
n
, which is the principal open subset of A
n
2
where det = 0. In particu
lar, T has a natural structure as an irreducible algebraic variety. The map (e
1
. . . . . e
n
)
(e
1
. . . . . e
d
): T G
d
(1) is a surjective regular map, and so G
d
(1) is also irreducible.
REMARK 6.31. The exterior algebra
_
1 =
d_0
_
d
1 of 1 is the quotient of the tensor
algebra by the ideal generated by all vectors e e, e 1. The elements of
_
d
1 are called
(exterior) Jvectors.The exterior algebra of 1 is a nitedimensional graded algebra over k
132 6. PROJECTIVE VARIETIES
with
_
0
1 =k,
_
1
1 =1; if e
1
. . . . . e
n
form an ordered basis for V , then the
_
n
d
_
wedge
products e
i
1
.. . . .e
i
d
(i
1
< <i
d
) form an ordered basis for
_
d
1. In particular,
_
n
1
has dimension 1. For a subspace S of 1 of dimension J,
_
d
S is the onedimensional
subspace of
_
d
1 spanned by e
1
.. . . .e
d
for any basis e
1
. . . . . e
d
of S. Thus, there is a
welldened map
S
d
S: G
d
(1) P(
d
1) (17)
which the choice of a basis for 1 identies with S 1(S). Note that the subspace spanned
by e
1
. . . . . e
n
can be recovered from the line through e
1
.. . . .e
d
as the space of vectors
such that .e
1
.. . . .e
d
=0 (cf. 6.32 below).
First proof of Proposition 6.26. Fix a basis e
1
. . . . . e
n
of 1, and let S
0
=(e
1
. . . . . e
d
)
and S
t
=(e
d1
. . . . . e
n
). Order the coordinates in P
_
n
d
_
1
so that
1(S) =(a
0
: . . . : a
i}
: . . . : . . .)
where a
0
is the leftmost J J minor of (S), and a
i}
, 1 _ i _ J, J < _ n, is the
minor obtained from the leftmost J J minor by replacing the i th column with the th
column. Let U
0
be the (typical) standard open subset of P
_
n
d
_
1
consisting of the points
with nonzero zeroth coordinate. Clearly,
8
1(S) U
0
if and only if S G
d
(1)
S
0 . We shall
prove the proposition by showing that 1: G
d
(1)
S
0 U
0
is injective with closed image.
For S G
d
(1)
S
0 , the projection S S
0
is bijective. For each i , 1 _i _J, let
e
t
i
=e
i
d~}_n
a
i}
e
}
(18)
denote the unique element of S projecting to e
i
. Then e
t
1
. . . . . e
t
d
is a basis for S. Con
versely, for any (a
i}
) k
d(nd)
, the e
t
i
s dened by (18) span an S G
d
(1)
S
0 and project
to the e
i
s. Therefore, S (a
i}
) gives a onetoone correspondence G
d
(1)
S
0 k
d(nd)
(this is a restatement of (15) in terms of matrices).
Now, if S (a
i}
), then
1(S) =(1 : . . . : a
i}
: . . . : . . . :
k
(a
i}
): . . .)
where
k
(a
i}
) is a polynomial in the a
i}
whose coefcients are independent of S. Thus,
1(S) determines (a
i}
) and hence also S. Moreover, the image of 1: G
d
(1)
S
0 U
0
is the
graph of the regular map
(. . . . a
i}
. . . .) (. . . .
k
(a
i}
). . . .): A
d(nd)
A
_
n
d
_
d(nd)1
.
which is closed (4.26).
Second proof of Proposition 6.26. An exterior Jvector is said to be pure (or de
composable) if there exist vectors e
1
. . . . . e
d
V such that =e
1
.. . . .e
d
. According to
(6.31), the image of G
d
(1) in P(
_
d
1) consists of the lines through the pure Jvectors.
8
If e S
t
S is nonzero, we may choose it to be part of the basis for S, and then the leftmost J J
submatrix of (S) has a row of zeros. Conversely, if the leftmost J J submatrix is singular, we can change
the basis for S so that it has a row of zeros; then the basis element corresponding to the zero row lies in S
t
S.
Grassmann varieties 133
LEMMA 6.32. Let n be a nonzero Jvector and let
M(n) ={ 1 [ .n =0]:
then dim
k
M(n) _J, with equality if and only if n is pure.
PROOF. Let e
1
. . . . . e
n
be a basis of M(n), and extend it to a basis e
1
. . . . . e
n
. . . . . e
n
of V .
Write
n =
1_i
1
~...~i
d
a
i
1
...i
d
e
i
1
.. . . .e
i
d
. a
i
1
...i
d
k.
If there is a nonzero termin this sumin which e
}
does not occur, then e
}
.n=0. Therefore,
each nonzero term in the sum is of the form ae
1
.. . . .e
n
.. . .. It follows that m_J, and
m=J if and only if n =ae
1
.. . . .e
d
with a =0.
2
For a nonzero Jvector n, let n denote the line through n. The lemma shows that
n G
d
(1) if and only if the linear map .n: 1
_
d1
1 has rank _ nJ (in
which case the rank is nJ). Thus G
d
(1) is dened by the vanishing of the minors of
order nJ 1 of this map.
9
Flag varieties
The discussion in the last subsection extends easily to chains of subspaces. Let d =
(J
1
. . . . . J
i
) be a sequence of integers with 0 < J
1
< < J
i
< n, and let G
d
(1) be the
set of ags
J : 1 1
1
1
i
0 (19)
with 1
i
a subspace of 1 of dimension J
i
. The map
G
d
(1)
T(V
i
)
i
G
d
i
(1)
i
P(
_
d
i
1)
realizes G
d
(1) as a closed subset
10
i
G
d
i
(1), and so it is a projective variety, called a
ag variety. The tangent space to G
d
(1) at the ag J consists of the families of homomor
phisms
i
: 1
i
V,1
i
. 1 _i _r. (20)
9
In more detail, the map
n ( .n):
d
1 Hom
k
(1.
d1
1)
is injective and linear, and so denes an injective regular map
P(
d
1) P(Hom
k
(1.
d1
1)).
The condition rank _nJ denes a closed subset W of P(Hom
k
(1.
_
d1
1)) (once a basis has been chosen
for 1, the condition becomes the vanishing of the minors of order nJ 1 of a linear map 1
_
d1
1),
and
G
d
(1) =P(
_
d
1) W.
10
For example, if u
i
is a pure J
i
vector and u
i1
is a pure J
i1
vector, then it follows from (6.32) that
M(u
i
) M(u
i1
) if and only if the map
( .u
i
. .u
i1
): V
d
i
1
V
d
iC1
1
V
has rank _nJ
i
(in which case it has rank nJ
i
). Thus, G
d
(V ) is dened by the vanishing of many minors.
134 6. PROJECTIVE VARIETIES
that are compatible in the sense that
i
[1
i1
i1
mod 1
i1
.
ASIDE 6.33. Abasis e
1
. . . . . e
n
for 1 is adapted to the ag J if it contains a basis e
1
. . . . . e
}
i
for each 1
i
. Clearly, every ag admits such a basis, and the basis then determines the ag.
As in (6.30), this implies that G
d
(1) is irreducible. Because GL(1) acts transitively on
the set of bases for 1, it acts transitively on G
d
(1). For a ag J, the subgroup 1(J)
stabilizing J is an algebraic subgroup of GL(1), and the map
g gJ
0
: GL(1),1(J
0
) G
d
(1)
is an isomorphism of algebraic varieties. Because G
d
(1) is projective, this shows that
1(J
0
) is a parabolic subgroup of GL(V ).
Bezouts theorem
Let V be a hypersurface in P
n
(that is, a closed subvariety of dimension n1). For such
a variety, 1(V ) = (J(X
0
. . . . . X
n
)) with J a homogenous polynomial without repeated
factors. We dene the degree of V to be the degree of J.
The next theorem is one of the oldest, and most famous, in algebraic geometry.
THEOREM 6.34. Let C and D be curves in P
2
of degrees m and n respectively. If C and
D have no irreducible component in common, then they intersect in exactly mn points,
counted with appropriate multiplicities.
PROOF. Decompose C and D into their irreducible components. Clearly it sufces to
prove the theorem for each irreducible component of C and each irreducible component of
D. We can therefore assume that C and D are themselves irreducible.
We know from (2.26) that C D is of dimension zero, and so is nite. After a change
of variables, we can assume that a =0 for all points (a : b : c) C D.
Let J(X. Y. 7) and G(X. Y. 7) be the polynomials dening C and D, and write
J =s
0
7
n
s
1
7
n1
s
n
. G =t
0
7
n
t
1
7
n1
t
n
with s
i
and t
}
polynomials in X and Y of degrees i and respectively. Clearly s
n
=0 =t
n
,
for otherwise J and G would have 7 as a common factor. Let 1 be the resultant of J and
G, regarded as polynomials in 7. It is a homogeneous polynomial of degree mn in X and
Y , or else it is identically zero. If the latter occurs, then for every (a. b) k
2
, J(a. b. 7)
and G(a. b. 7) have a common zero, which contradicts the niteness of C D. Thus 1
is a nonzero polynomial of degree mn. Write 1(X. Y ) = X
nn
1
+
(
Y
A
) where 1
+
(T ) is a
polynomial of degree _mn in T =
Y
A
.
Suppose rst that deg1
+
=mn, and let
1
. . . . .
nn
be the roots of 1
+
(some of them
may be multiple). Each such root can be written
i
=
b
i
o
i
, and 1(a
i
. b
i
) =0. According to
(7.12) this means that the polynomials J(a
i
. b
i
. 7) and G(a
i
. b
i
. 7) have a common root
c
i
. Thus (a
i
: b
i
: c
i
) is a point on C D, and conversely, if (a : b : c) is a point on C D
(so a =0), then
b
o
is a root of 1
+
(T ). Thus we see in this case, that C D has precisely
mn points, provided we take the multiplicity of (a : b : c) to be the multiplicity of
b
o
as a
root of 1
+
.
Hilbert polynomials (sketch) 135
Now suppose that 1
+
has degree r < mn. Then 1(X. Y ) = X
nni
1(X. Y ) where
1(X. Y ) is a homogeneous polynomial of degree r not divisible by X. Obviously 1(0. 1) =
0, and so there is a point (0 : 1 : c) in C D, in contradiction with our assumption.
2
REMARK 6.35. The above proof has the defect that the notion of multiplicity has been too
obviously chosen to make the theoremcome out right. It is possible to showthat the theorem
holds with the following more natural denition of multiplicity. Let 1 be an isolated point
of C D. There will be an afne neighbourhood U of 1 and regular functions and g
on U such that C U = V( ) and DU = V(g). We can regard and g as elements
of the local ring O
1
, and clearly rad(. g) =m, the maximal ideal in O
1
. It follows that
O
1
,(. g) is nitedimensional over k, and we dene the multiplicity of 1 in C D to
be dim
k
(O
1
,(. g)). For example, if C and D cross transversely at 1, then and g will
form a system of local parameters at 1 (. g) =m and so the multiplicity is one.
The attempt to nd good notions of multiplicities in very general situations motivated
much of the most interesting work in commutative algebra in the second half of the twenti
eth century.
Hilbert polynomials (sketch)
Recall that for a projective variety V P
n
,
k
hom
V  =kX
0
. . . . . X
n
,b =k.
0
. . . . . .
n
.
where b = 1(V ). We observed that b is homogeneous, and therefore k
hom
V  is a graded
ring:
k
hom
V  =
n_0
k
hom
V 
n
.
where k
hom
V 
n
is the subspace generated by the monomials in the .
i
of degree m. Clearly
k
hom
V 
n
is a nitedimensional kvector space.
THEOREM 6.36. There is a unique polynomial 1(V. T ) such that 1(V. m) =dim
k
kV 
n
for all m sufciently large.
PROOF. Omitted.
2
EXAMPLE 6.37. For V = P
n
, k
hom
V  = kX
0
. . . . . X
n
, and (see the footnote on page
126), dimk
hom
V 
n
=
_
nn
n
_
=
(nn)(n1)
n
, and so
1(P
n
. T ) =
_
Tn
n
_
=
(T n) (T 1)
n
.
The polynomial 1(V. T ) in the theorem is called the Hilbert polynomial of V . Despite
the notation, it depends not just on V but also on its embedding in projective space.
THEOREM 6.38. Let V be a projective variety of dimension J and degree ; then
1(V. T ) =
J
T
d
terms of lower degree.
PROOF. Omitted.
2
136 6. PROJECTIVE VARIETIES
The degree of a projective variety is the number of points in the intersection of the
variety and of a general linear variety of complementary dimension (see later).
EXAMPLE 6.39. Let V be the image of the Veronese map
(a
0
: a
1
) (a
d
0
: a
d1
0
a
1
: . . . : a
d
1
): P
1
P
d
.
Then k
hom
V 
n
can be identied with the set of homogeneous polynomials of degree m J
in two variables (look at the map A
2
A
d1
given by the same equations), which is a
space of dimension Jm1, and so
1(V. T ) =JT 1.
Thus V has dimension 1 (which we certainly knew) and degree J.
Macaulay knows how to compute Hilbert polynomials.
References: Hartshorne 1977, I.7; Atiyah and Macdonald 1969, Chapter 11; Harris
1992, Lecture 13.
Exercises
61. Show that a point 1 on a projective curve J(X. Y. 7) = 0 is singular if and only if
dJ,dX, dJ,dY , and dJ,d7 are all zero at 1. If 1 is nonsingular, show that the tangent
line at 1 has the (homogeneous) equation
(dJ,dX)
1
X (dJ,dY )
1
Y (dJ,d7)
1
7 =0.
Verify that Y
2
7 =X
3
aX7
2
b7
3
is nonsingular if X
3
aXb has no repeated root,
and nd the tangent line at the point at innity on the curve.
62. Let 1be a line in P
2
and let C be a nonsingular conic in P
2
(i.e., a curve in P
2
dened
by a homogeneous polynomial of degree 2). Show that either
(a) 1 intersects C in exactly 2 points, or
(b) 1 intersects C in exactly 1 point, and it is the tangent at that point.
63. Let V =V(Y X
2
. 7X
3
) A
3
. Prove
(a) 1(V ) =(Y X
2
. 7X
3
).
(b) 7W XY 1(V )
+
kW. X. Y. 7, but 7W XY ((Y X
2
)
+
. (7 X
3
)
+
).
(Thus, if J
1
. . . . . J
i
generate a, it does not follow that J
+
1
. . . . . J
+
i
generate a
+
, even
if a
+
is radical.)
64. Let 1
0
. . . . . 1
i
be points in P
n
. Show that there is a hyperplane H in P
n
passing
through 1
0
but not passing through any of 1
1
. . . . . 1
i
.
65. Is the subset
{(a : b : c) [ a =0. b =0] L{(1 : 0 : 0)]
of P
2
locally closed?
66. Show that the image of the Segre map P
n
P
n
P
nnnn
(see 6.23) is not con
tained in any hyperplane of P
nnnn
.
CHAPTER 7
Complete varieties
Throughout this chapter, k is an algebraically closed eld.
Denition and basic properties
Complete varieties are the analogues in the category of algebraic varieties of compact topo
logical spaces in the category of Hausdorff topological spaces. Recall that the image of a
compact space under a continuous map is compact, and hence is closed if the image space
is Hausdorff. Moreover, a Hausdorff space V is compact if and only if, for all topological
spaces W, the projection q: V W W is closed, i.e., maps closed sets to closed sets (see
Bourbaki, N., General Topology, I, 10.2, Corollary 1 to Theorem 1).
DEFINITION 7.1. An algebraic variety V is said to be complete if for all algebraic varieties
W, the projection q: V W W is closed.
Note that a complete variety is required to be separated we really mean it to be a
variety and not a prevariety.
EXAMPLE 7.2. Consider the projection
(.. ,) ,: A
1
A
1
A
1
This is not closed; for example, the variety V : XY =1 is closed in A
2
but its image in A
1
omits the origin. However, if we replace V with its closure in P
1
A
1
, then its projection
is the whole of A
1
.
PROPOSITION 7.3. Let V be a complete variety.
(a) A closed subvariety of V is complete.
(b) If V
t
is complete, so also is V V
t
.
(c) For any morphism : V W, (V ) is closed and complete; in particular, if V is a
subvariety of W, then it is closed in W.
(d) If V is connected, then any regular map : V P
1
is either constant or onto.
(e) If V is connected, then any regular function on V is constant.
137
138 7. COMPLETE VARIETIES
PROOF. (a) Let 7 be a closed subvariety of a complete variety V . Then for any variety
W, 7W is closed in V W, and so the restriction of the closed map q: V W W to
7W is also closed.
(b) The projection V V
t
W W is the composite of the projections
V V
t
W V
t
W W.
both of which are closed.
(c) Let 1
into W. Since V is
complete, the projection is closed, and so (V ) is closed, and hence is a subvariety of W
(see p76). Consider
1
W (V ) W W.
The variety 1
W W is closed. As 1
i
}
and so we have the notion of a homogeneous ideal b T. It makes sense to speak of the
zero set V(b) W P
n
of such an ideal. For any ideal a , aT is homogeneous, and
V(aT) =V(a) P
n
.
LEMMA 7.9. (a) For each homogeneous ideal b T, the set V(b) is closed, and every
closed subset of W P
n
is of this form.
(b) The set V(b) is empty if and only if rad(b) (X
0
. . . . . X
n
).
(c) If W is irreducible, then W =V(b) for some homogeneous prime ideal b.
PROOF. In the case that = k, we proved this in (6.1) and (6.2), and similar arguments
apply in the present more general situation. For example, to see that V(b) is closed, cover
P
n
with the standard open afnes U
i
and show that V(b) U
i
is closed for all i .
The set V(b) is empty if and only if the cone V
aff
(b) W A
n1
dened by b is
contained in W {origin]. But
a
i
0
...i
n
X
i
0
0
. . . X
i
n
n
. a
i
0
...i
n
kW.
is zero on W {origin] if and only if its constant term is zero, and so
1
aff
(W {origin]) =(X
0
. X
1
. . . . . X
n
).
Thus, the Nullstellensatz shows that V(b) = 0 =rad(b) = (X
0
. . . . . X
n
). Conversely, if
X
1
i
b for all i , then obviously V(b) is empty.
For (c), note that if V(b) is irreducible, then the closure of its inverse image in W
A
n1
is also irreducible, and so 1V(b) is prime.
2
140 7. COMPLETE VARIETIES
PROOF (OF 7.7). Write for the projection W P
n
W. We have to show that 7 closed
in W P
n
implies (7) closed in W. If 7 is empty, this is true, and so we can assume it
to be nonempty. Then 7 is a nite union of irreducible closed subsets 7
i
of W P
n
, and it
sufces to show that each (7
i
) is closed. Thus we may assume that 7 is irreducible, and
hence that 7 =V(b) with b a homogeneous prime ideal in T =X
0
. . . . . X
n
.
If (7) is contained in some closed subvariety W
t
of W, then 7 is contained in W
t
P
n
, and we can replace W with W
t
. This allows us to assume that (7) is dense in W, and
we now have to show that (7) =W.
Because (7) is dense in W, the image of the cone V
aff
(b) under the projection W
A
n1
W is also dense in W, and so (see 3.22a) the map T,b is injective.
Let n W: we shall show that if n (7), i.e., if there does not exist a 1 P
n
such
that (n. 1) 7, then (7) is empty, which is a contradiction.
Let m be the maximal ideal corresponding to n. Then mT b is a homogeneous
ideal, and V(mT b) =V(mT) V(b) =(nP
n
) V(b), and so n will be in the image
of 7 unless V(mT b) = 0. But if V(mT b) = 0, then mT b (X
0
. . . . . X
n
)
1
for
some N (by 7.9b), and so mT b contains the set T
1
of homogeneous polynomials of
degree N. Because mT and b are homogeneous ideals,
T
1
mT b == T
1
=mT
1
T
1
b.
In detail: the rst inclusion says that an T
1
can be written =gh with g mT and
h b. On equating homogeneous components, we nd that
1
= g
1
h
1
. Moreover:
1
= ; if g =
m
i
b
i
, m
i
m, b
i
T, then g
1
=
m
i
b
i1
; and h
1
b because b is
homogeneous. Together these show mT
1
T
1
b.
Let M =T
1
,T
1
b, regarded as an module. The displayed equation says that M =
mM. The argument in the proof of Nakayamas lemma (1.3) shows that (1m)M =0 for
some m m. Because T,b is injective, the image of 1 m in T,b is nonzero. But
M =T
1
,T
1
b T,b, which is an integral domain, and so the equation (1 m)M =
0 implies that M = 0. Hence T
1
b, and so X
1
i
b for all i , which contradicts the
assumption that 7 =V(b) is nonempty.
2
REMARK 7.10. In Example 6.19 above, we showed that every nite set of points in a
projective variety is contained in an open afne subvariety. There is a partial converse to
this statement: let V be a nonsingular complete irreducible variety; if every nite set of
points in V is contained in an open afne subset of V then V is projective. (Conjecture of
Chevalley; proved by Kleiman.
1
)
Elimination theory
We have shown that, for any closed subset 7 of P
n
W, the projection q(7) of 7 in W is
closed. Elimination theory
2
is concerned with providing an algorithm for passing from the
equations dening 7 to the equations dening q(7). We illustrate this in one case.
1
Kleiman, Steven L., Toward a numerical theory of ampleness. Ann. of Math. (2) 84 1966 293344.
See also,
Hartshorne, Robin, Ample subvarieties of algebraic varieties. Lecture Notes in Mathematics, Vol. 156 Springer,
1970, I 9 p45.
2
Elimination theory became unfashionable several decades agoone prominent algebraic geometer went
so far as to announce that Theorem 7.7 eliminated elimination theory from mathematics, provoking Abhyankar,
who prefers equations to abstractions, to start the chant eliminate the eliminators of elimination theory. With
the rise of computers, it has become fashionable again.
Elimination theory 141
Let 1 =s
0
X
n
s
1
X
n1
s
n
and Q=t
0
X
n
t
1
X
n1
t
n
be polynomi
als. The resultant of 1 and Q is dened to be the determinant
s
0
s
1
. . . s
n
s
0
. . . s
n
. . . . . .
t
0
t
1
. . . t
n
t
0
. . . t
n
. . . . . .
n rows
m rows
There are n rows of ss and m rows of t s, so that the matrix is (mn) (mn); all blank
spaces are to be lled with zeros. The resultant is a polynomial in the coefcients of 1 and
Q.
PROPOSITION 7.11. The resultant Res(1. Q) =0 if and only if
(a) both s
0
and t
0
are zero; or
(b) the two polynomials have a common root.
PROOF. If (a) holds, then Res(1. Q) =0 because the rst column is zero. Suppose that
is a common root of 1 and Q, so that there exist polynomials 1
1
and Q
1
of degrees m1
and n1 respectively such that
1(X) =(X )1
1
(X). Q(X) =(X )Q
1
(X).
Using these equalities, we nd that
1(X)Q
1
(X) Q(X)1
1
(X) =0. (21)
On equating the coefcients of X
nn1
. . . . . X. 1 in (21) to zero, we nd that the coef
cients of 1
1
and Q
1
are the solutions of a system of mn linear equations in mn
unknowns. The matrix of coefcients of the system is the transpose of the matrix
_
_
_
_
_
_
_
_
s
0
s
1
. . . s
n
s
0
. . . s
n
. . . . . .
t
0
t
1
. . . t
n
t
0
. . . t
n
. . . . . .
_
_
_
_
_
_
_
_
The existence of the solution shows that this matrix has determinant zero, which implies
that Res(1. Q) =0.
Conversely, suppose that Res(1. Q) =0 but neither s
0
nor t
0
is zero. Because the above
matrix has determinant zero, we can solve the linear equations to nd polynomials 1
1
and
Q
1
satisfying (21). A root of 1 must be also be a root of 1
1
or of Q. If the former,
cancel X from the left hand side of (21), and consider a root of 1
1
,(X ). As
deg1
1
< deg1, this argument eventually leads to a root of 1 that is not a root of 1
1
, and
so must be a root of Q.
2
142 7. COMPLETE VARIETIES
The proposition can be restated in projective terms. We dene the resultant of two
homogeneous polynomials
1(X. Y ) =s
0
X
n
s
1
X
n1
Y s
n
Y
n
. Q(X. Y ) =t
0
X
n
t
n
Y
n
.
exactly as in the nonhomogeneous case.
PROPOSITION 7.12. The resultant Res(1. Q) =0 if and only if 1 and Q have a common
zero in P
1
.
PROOF. The zeros of 1(X. Y ) in P
1
are of the form:
(a) (1 : 0) in the case that s
0
=0;
(b) (a : 1) with a a root of 1(X. 1).
Since a similar statement is true for Q(X. Y ), (7.12) is a restatement of (7.11).
2
Now regard the coefcients of 1 and Q as indeterminates. The pairs of polynomials
(1. Q) are parametrized by the space A
n1
A
n1
=A
nn2
. Consider the closed subset
V(1. Q) in A
nn2
P
1
. The proposition shows that its projection on A
nn2
is the set
dened by Res(1. Q) =0. Thus, not only have we shown that the projection of V(1. Q)
is closed, but we have given an algorithm for passing from the polynomials dening the
closed set to those dening its projection.
Elimination theory does this in general. Given a family of polynomials
1
i
(T
1
. . . . . T
n
: X
0
. . . . . X
n
).
homogeneous in the X
i
, elimination theory gives an algorithm for nding polynomials
1
}
(T
1
. . . . . T
n
) such that the 1
i
(a
1
. . . . . a
n
: X
0
. . . . . X
n
) have a common zero if and only if
1
}
(a
1
. . . . . a
n
) =0 for all . (Theorem 7.7 shows only that the 1
}
exist.) See Cox et al.
1992, Chapter 8, Section 5..
Maple can nd the resultant of two polynomials in one variable: for example, entering
resultant((. a)
5
. (. b)
5
. .) gives the answer (a b)
25
. Explanation: the polyno
mials have a common root if and only if a =b, and this can happen in 25 ways. Macaulay
doesnt seem to know how to do more.
The rigidity theorem
The paucity of maps between complete varieties has some interesting consequences. First
an observation: for any point n W, the projection map V W V denes an isomor
phism V {n] V with inverse (. n): V V W (this map is regular because its
components are).
THEOREM 7.13 (RIGIDITY THEOREM). Let : V W 7 be a regular map, and assume
that V is complete, that V and W are irreducible, and that 7 is separated. If there exist
points
0
V , n
0
W, :
0
7 such that
(V {n
0
]) ={:
0
] =({
0
] W),
then (V W) ={:
0
].
Theorems of Chow 143
PROOF. Because V is complete, the projection map q: V W W is closed. Therefore,
for any open afne neighbourhood U of :
0
,
T =q(
1
(7U))
is closed in W. Note that
W T ={n W [ (V. n) U],
and so n
0
W T . In particular, W T is nonempty, and so it is dense in W. As V {n]
is complete and U is afne, (V {n]) must be a point whenever n W T : in fact,
(V. n) =(
0
. n) ={:
0
].
We have shown that takes the constant value :
0
on the dense subset V (W T ) of
V W, and therefore on the whole of V W.
2
In more colloquial terms, the theorem says that if collapses a vertical and a horizontal
slice to a point, then it collapses the whole of V W to a point, which must therefore be
rigid.
An abelian variety is a complete connected group variety.
COROLLARY 7.14. Every regular map : T of abelian varieties is the composite of a
homomorphism with a translation; in particular, a regular map : T such that (0) =0
is a homomorphism.
PROOF. After composing with a translation, we may suppose that (0) = 0. Consider
the map
: T. (a. a
t
) =(aa
t
) (a) (a
t
).
Then (0) =0 =(0) and so =0. This means that is a homomorphism.
2
COROLLARY 7.15. The group law on an abelian variety is commutative.
PROOF. Commutative groups are distinguished among all groups by the fact that the map
taking an element to its inverse is a homomorphism: if (gh)
1
=g
1
h
1
, then, on taking
inverses, we nd that gh =hg. Since the negative map, a a: , takes the identity
element to itself, the preceding corollary shows that it is a homomorphism.
2
Theorems of Chow
THEOREM 7.16. For every algebraic variety V , there exists a projective algebraic variety
W and a regular map from an open dense subset U of W to V whose graph is closed in
V W; the set U =W if and only if V is complete.
PROOF. To be added.
2
See:
144 7. COMPLETE VARIETIES
Chow, WL., On the projective embedding of homogeneous varieties, Lef
schetzs volume, Princeton 1956.
Serre, JeanPierre. G eom etrie alg ebrique et g eom etrie analytique. Ann.
Inst. Fourier, Grenoble 6 (19551956), 142 (p12).
THEOREM 7.17. For any complete algebraic variety V , there exists a projective algebraic
variety W and a surjective birational map W V .
PROOF. To be added. (See Mumford 1999, p60.)
2
Theorem 7.17 is usually known as Chows Lemma.
Nagatas Embedding Theorem
A necessary condition for a prevariety to be an open subvariety of a complete variety is that
it be separated. A theorem of Nagata says that this condition is also sufcient.
THEOREM 7.18. For every variety V , there exists an open immersion V W with W
complete.
The theorem is important, but the proof is quite difcult, even for varieties. It was
originally proved by Nagata:
Nagata, Masayoshi. Imbedding of an abstract variety in a complete variety. J.
Math. Kyoto Univ. 2 1962 110; A generalization of the imbedding problem of
an abstract variety in a complete variety. J. Math. Kyoto Univ. 3 1963 89102.
For more modern expositions, see:
L utkebohmert, W. On compactication of schemes. Manuscripta Math. 80
(1993), no. 1, 95111.
Deligne, P., Le th eor` eme de plongement de Nagata, personal notes (published
as Kyoto J. Math. 50, Number 4 (2010), 661670.
Conrad, B., Delignes notes on Nagata compactications. J. Ramanujan Math.
Soc. 22 (2007), no. 3, 205257.
Vojta, P., Nagatas embedding theorem, arXiv:0706.1907.
Exercises
71. Identify the set of homogeneous polynomials J(X. Y ) =
a
i}
X
i
Y
}
, 0 _ i. _ m,
with an afne space. Show that the subset of reducible polynomials is closed.
72. Let V and W be complete irreducible varieties, and let be an abelian variety. Let
1 and Q be points of V and W. Show that any regular map h: V W such that
h(1. Q) =0 can be written h = g q where : V and g: W are regular
maps carrying 1 and Q to 0 and and q are the projections V W V. W.
CHAPTER 8
Finite Maps
Throughout this chapter, k is an algebraically closed eld.
Denition and basic properties
Recall that an algebra T is said to be nite if it is nitely generated as an module. This
is equivalent to T being nitely generated as an algebra and integral over . Recall also
that a variety V is afne if and only if 1(V. O
V
) is an afne kalgebra and the canonical
map (V. O
V
) Spm(1(V. O
V
)) is an isomorphism (3.13).
DEFINITION 8.1. A regular map : W V is said to be nite if for all open afne subsets
U of V ,
1
(U) is an afne variety and k
1
(U) is a nite kUalgebra.
For example, suppose W and V are afne and kW is a nite kV algebra. Then is
nite because, for any open afne U in V ,
1
(U) is afne with
k
1
(U) .kW
kV j
kU (22)
(see 4.29, 4.30); in particular, the canonical map
1
(U) Spm(1(
1
(U). O
W
) (23)
is an isomorphism.
PROPOSITION 8.2. It sufces to check the condition in the denition for all subsets in one
open afne covering of V .
Unfortunately, this is not as obvious as it looks. We rst need a lemma.
LEMMA 8.3. Let : W V be a regular map with V afne, and let U be an open afne
in V . There is a canonical isomorphism of kalgebras
1(W. O
W
)
kV j
kU 1(
1
(U). O
W
).
PROOF. Let U
t
=
1
(U). The map is dened by the kV bilinear pairing
(. g) ( [
U
0 . g [
U
0 ): 1(W. O
W
) kU 1(U
t
. O
W
).
145
146 8. FINITE MAPS
When W is also afne, it is the isomorphism (22).
Let W =
_
W
i
be a nite open afne covering of W, and consider the commutative
diagram:
0 1(W. O
W
)
kV j
kU
i
1(W
i
. O
W
)
kV j
kU
i,}
1(W
i}
. O
W
)
kV j
kU
0 1(U
t
. O
W
)
i
1(U
t
W
i
. O
W
)
i,}
1(U W
i}
. O
W
)
Here W
i}
=W
i
W
}
. The bottomrowis exact because O
W
is a sheaf, and the top rowis ex
act because O
W
is a sheaf and kU is at over kV  (see Section 1)
1
. The varieties W
i
and
W
i
W
}
are all afne, and so the two vertical arrows at right are products of isomorphisms
(22). This implies that the rst is also an isomorphism.
2
PROOF (OF THE PROPOSITION). Let V
i
be an open afne covering of V (which we may
suppose to be nite) such that W
i
=
def
1
(V
i
) is an afne subvariety of W for all i and
kW
i
 is a nite kV
i
algebra. Let U be an open afne in V , and let U
t
=
1
(U). Then
1(U
t
. O
W
) is a subalgebra of
i
1(U
t
W
i
. O
W
), and so it is an afne kalgebra nite
over kU.
2
We have a morphism of varieties over V
U
t
Spm(1(U
t
. O
W
))
V
canonical
(24)
which we shall show to be an isomorphism. We know (see (23)) that each of the maps
U
t
W
i
Spm(1(U
t
W
i
. O
W
))
is an isomorphism. But (8.2) shows that Spm(1(U
t
W
i
. O
W
)) is the inverse image of V
i
in Spm(1(U
t
. O
W
)). Therefore the canonical morphism is an isomorphism over each V
i
,
and so it is an isomorphism.
2
PROPOSITION 8.4. (a) For any closed subvariety 7 of V , the inclusion 7 V is nite.
(b) The composite of two nite morphisms is nite.
(c) The product of two nite morphisms is nite.
PROOF. (a) Let U be an open afne subvariety of V . Then 7U is a closed subvariety of
U. It is therefore afne, and the map 7U U corresponds to a map ,a of rings,
which is obviously nite.
1
A sequence 0 M
t
M M
tt
is exact if and only if 0
m
M
t
M
m
M
tt
is exact for all maximal ideals m of . This implies the claim because kU
m
P
.O
U,1
.O
V,1
.kV 
m
P
for all 1 U.
2
Recall that a module over a noetherian ring is noetherian if and only if it is nitely generated, and that
a submodule of a noetherian module is noetherian. Therefore, a submodule of a nitely generated module is
nitely generated.
Denition and basic properties 147
(b) If T is a nite algebra and C is a nite Talgebra, then C is a nite algebra.
To see this, note that if {b
i
] is a set of generators for T as an module, and {c
}
] is a set of
generators for C as a Tmodule, then {b
i
c
}
] is a set of generators for C as an module.
(c) If T and T
t
are respectively nite and
t
algebras, then T
k
T
t
is a nite
k
t
algebra. To see this, note that if {b
i
] is a set of generators for T as an module, and
{b
t
}
] is a set of generators for T
t
as an module, the {b
i
b
t
}
] is a set of generators for
T
T
t
as an module.
2
By way of contrast, an open immersion is rarely nite. For example, the inclusion
A
1
{0] A
1
is not nite because the ring kT. T
1
 is not nitely generated as a kT 
module (any nitely generated kT submodule of kT. T
1
 is contained in T
n
kT  for
some n).
The bres of a regular map : W V are the subvarieties
1
(1) of W for 1 V .
When the bres are all nite, is said to be quasinite.
PROPOSITION 8.5. A nite map : W V is quasinite.
PROOF. Let 1 V ; we wish to show
1
(1) is nite. After replacing V with an afne
neighbourhood of 1, we can suppose that it is afne, and then W will be afne also. The
map then corresponds to a map : T of afne kalgebras, and a point Q of W maps
to 1 if and only
1
(m
Q
) = m
1
. But this holds if and only if m
Q
(m
1
), and so
the points of W mapping to 1 are in onetoone correspondence with the maximal ideals
of T,(m)T. Clearly T,(m)T is generated as a kvector space by the image of any
generating set for T as an module, and the next lemma shows that it has only nitely
many maximal ideals.
2
LEMMA 8.6. A nite kalgebra has only nitely many maximal ideals.
PROOF. Let m
1
. . . . . m
n
be maximal ideals in . They are obviously coprime in pairs, and
so the Chinese Remainder Theorem (1.1) shows that the map
,m
1
,m
n
. a (. . . . a
i
modm
i
. . . .).
is surjective. It follows that dim
k
_
dim
k
(,m
i
) _n (dimensions as kvector spaces).
2
THEOREM 8.7. A nite map : W V is closed.
PROOF. Again we can assume V and W to be afne. Let 7 be a closed subset of W. The
restriction of to 7 is nite (by 8.4a and b), and so we can replace W with 7; we then
have to show that Im() is closed. The map corresponds to a nite map of rings T.
This will factors as ,a T, from which we obtain maps
Spm(T) Spm(,a) Spm().
The second map identies Spm(,a) with the closed subvariety V(a) of Spm(), and so it
remains to show that the rst map is surjective. This is a consequence of the next lemma.
2
LEMMA 8.8 (GOINGUP THEOREM). Let T be rings with T integral over .
148 8. FINITE MAPS
(a) For every prime ideal p of , there is a prime ideal q of T such that q =p.
(b) Let p =q; then p is maximal if and only if q is maximal.
PROOF. (a) If S is a multiplicative subset of a ring , then the prime ideals of S
1
are in onetoone correspondence with the prime ideals of not meeting S (see 1.30). It
therefore sufces to prove (a) after and T have been replaced by S
1
and S
1
T, where
S =p. Thus we may assume that is local, and that p is its unique maximal ideal. In
this case, for all proper ideals b of T, b p (otherwise b 1). To complete the
proof of (a), I shall show that for all maximal ideals n of T, n =p.
Consider T,n ,(n ). Here T,n is a eld, which is integral over its subring
,(n ), and n will be equal to p if and only if ,(n ) is a eld. This follows
from Lemma 8.9 below.
(b) The ring T,q contains ,p, and it is integral over ,p. If q is maximal, then
Lemma 8.9 shows that p is also. For the converse, note that any integral domain integral
over a eld is a eld because it is a union of integral domains nite over the eld, which
are automatically elds (left multiplication by an element is injective, and hence surjective,
being a linear map of a nitedimensional vector space).
2
LEMMA 8.9. Let be a subring of a eld 1. If 1 is integral over , then is also a eld.
PROOF. Let a be a nonzero element of . Then a
1
1, and it is integral over :
(a
1
)
n
a
1
(a
1
)
n1
a
n
=0. a
i
.
On multiplying through by a
n1
, we nd that
a
1
a
1
a
n
a
n1
=0.
from which it follows that a
1
.
2
COROLLARY 8.10. Let : W V be nite; if V is complete, then so also is W.
PROOF. Consider
W T V T T. (n. t ) ((n). t ) t.
Because W T V T is nite (see 8.4c), it is closed, and because V is complete,
V T T is closed. A composite of closed maps is closed, and therefore the projection
W T T is closed.
2
EXAMPLE 8.11. (a) Project XY = 1 onto the X axis. This map is quasinite but not
nite, because kX. X
1
 is not nite over kX.
(b) The map A
2
{origin] A
2
is quasinite but not nite, because the inverse image
of A
2
is not afne (3.21).
(c) Let
V =V(X
n
T
1
X
n1
T
n
) A
n1
.
and consider the projection map
(a
1
. . . . . a
n
. .) (a
1
. . . . . a
n
): V A
n
.
Noether Normalization Theorem 149
The bre over any point (a
1
. . . . . a
n
) A
n
is the set of solutions of
X
n
a
1
X
n1
a
n
=0.
and so it has exactly n points, counted with multiplicities. The map is certainly quasinite;
it is also nite because it corresponds to the nite map of kalgebras,
kT
1
. . . . . T
n
 kT
1
. . . . . T
n
. X,(X
n
T
1
X
n1
T
n
).
(d) Let
V =V(T
0
X
n
T
1
X
n1
T
n
) A
n2
.
The projection
(a
0
. . . . . a
n
. .) (a
1
. . . . . a
n
): V
A
n1
has nite bres except for the bre above o = (0. . . . . 0), which is A
1
. The restriction
[V
1
(o) is quasinite, but not nite. Above points of the form (0. . . . . 0. +. . . . . +)
some of the roots vanish off to o. (Example (a) is a special case of this.)
(e) Let
1(X. Y ) =T
0
X
n
T
1
X
n1
Y ... T
n
Y
n
.
and let V be its zero set in P
1
(A
n1
{o]). In this case, the projection map V
A
n1
{o] is nite. (Prove this directly, or apply 8.23 below.)
(f) The morphism A
1
A
2
, t (t
2
. t
3
) is nite because the image of kX. Y  in kT 
is kT
2
. T
3
, and {1. T ] is a set of generators for kT  over this subring.
(g) The morphism A
1
A
1
, a a
n
is nite (special case of (c)).
(h) The obvious map
(A
1
with the origin doubled ) A
1
is quasinite but not nite (the inverse image of A
1
is not afne).
The Frobenius map t t
]
: A
1
A
1
in characteristic =0 and the map t (t
2
. t
3
): A
1
V(Y
2
X
3
) A
2
from the line to the cuspidal cubic (see 3.18c) are examples of nite bi
jective regular maps that are not isomorphisms.
Noether Normalization Theorem
This theorem sometimes allows us to reduce the proofs of statements about afne varieties
to the case of A
n
.
THEOREM 8.12. For any irreducible afne algebraic variety V of a variety of dimension
J, there is a nite surjective map : V A
d
.
PROOF. This is a geometric restatement of the following theorem.
2
THEOREM 8.13 (NOETHER NORMALIZATION THEOREM). Let be a nitely generated
kalgebra, and assume that is an integral domain. Then there exist elements ,
1
. . . . . ,
d
that are algebraically independent over k and such that is integral over k,
1
. . . . . ,
d
.
150 8. FINITE MAPS
Let .
1
. . . . . .
n
generate as a kalgebra. We can renumber the .
i
so that .
1
. . . . . .
d
are
algebraically independent and .
d1
. . . . . .
n
are algebraically dependent on .
1
. . . . . .
d
(FT,
8.12).
Because .
n
is algebraically dependent on .
1
. . . . . .
d
, there exists a nonzero polynomial
(X
1
. . . . . X
d
. T ) such that (.
1
. . . . . .
d
. .
n
) =0. Write
(X
1
. . . . . X
d
. T ) =a
0
T
n
a
1
T
n1
a
n
with a
i
kX
1
. . . . . X
d
 (~ k.
1
. . . . . .
d
). If a
0
is a nonzero constant, we can divide
through by it, and then .
n
will satisfy a monic polynomial with coefcients in k.
1
. . . . . .
d
,
that is, .
n
will be integral (not merely algebraic) over k.
1
. . . . . .
d
. The next lemma sug
gest how we might achieve this happy state by making a linear change of variables.
LEMMA 8.14. If J(X
1
. . . . . X
d
. T ) is a homogeneous polynomial of degree r, then
J(X
1
z
1
T. . . . . X
d
z
d
T. T ) =J(z
1
. . . . . z
d
. 1)T
i
terms of degree <r in T.
PROOF. The polynomial J(X
1
z
1
T. . . . . X
d
z
d
T. T ) is still homogeneous of degree r
(in X
1
. . . . . X
d
. T ), and the coefcient of the monomial T
i
in it can be obtained by substi
tuting 0 for each X
i
and 1 for T .
2
PROOF (OF THEOREM 8.13). Note that unless J(X
1
. . . . . X
d
. T ) is the zero polynomial,
it will always be possible to choose (z
1
. . . . . z
d
) so that J(z
1
. . . . . z
d
. 1) =0 substitut
ing T =1 merely dehomogenizes the polynomial (no cancellation of terms occurs), and a
nonzero polynomial cant be zero on all of k
n
(Exercise 11).
Let J be the homogeneous part of highest degree of , and choose (z
1
. . . . . z
d
) so that
J(z
1
. . . . . z
d
. 1) =0. The lemma then shows that
(X
1
z
1
T. . . . . X
d
z
d
T. T ) =cT
i
b
1
T
i1
b
0
.
with c =J(z
1
. . . . . z
d
. 1) k
, b
i
kX
1
. . . . . X
d
, degb
i
< r. On substituting .
n
for T
and .
i
z
i
.
n
for X
i
we obtain an equation demonstrating that .
n
is integral over k.
1
z
1
.
n
. . . . . .
d
z
d
.
n
. Put .
t
i
= .
i
z
i
.
n
, 1 _ i _ J. Then .
n
is integral over the ring
k.
t
1
. . . . . .
t
d
, and it follows that is integral over
t
= k.
t
1
. . . . . .
t
d
. .
d1
. . . . . .
n1
.
Repeat the process for
t
, and continue until the theorem is proved.
2
REMARK 8.15. The above proof uses only that k is innite, not that it is algebraically
closed (thats all one needs for a nonzero polynomial not to be zero on all of k
n
). There are
other proofs that work also for nite elds (see CA 5.11), but the above proof is simpler and
gives us the additional information that the ,
i
s can be chosen to be linear combinations of
the .
i
. This has the following geometric interpretation:
let V be a closed subvariety of A
n
of dimension J; then there exists a linear
map A
n
A
d
whose restriction to V is a nite map V A
d
.
Zariskis main theorem
An obvious way to construct a nonnite quasinite map W V is to take a nite map
W
t
V and remove a closed subset of W
t
. Zariskis Main Theorem shows that, when W
and V are separated, every quasinite map arises in this way.
Zariskis main theorem 151
THEOREM 8.16 (ZARISKIS MAIN THEOREM). Any quasinite map of varieties : W
V factors into W
t
W
t
0
V with
t
nite and  an open immersion.
PROOF. Omitted see the references below (152).
2
REMARK 8.17. Assume (for simplicity) that V and W are irreducible and afne. The
proof of the theorem provides the following description of the factorization: it corresponds
to the maps
kV  kW
t
 kW
with kW
t
 the integral closure of kV  in kW.
A regular map : W V of irreducible varieties is said to be birational if it induces
an isomorphism k(V ) k(W) on the elds of rational functions (that is, if it demonstrates
that W and V are birationally equivalent).
REMARK 8.18. One may ask how a birational regular map : W V can fail to be an
isomorphism. Here are three examples.
(a) The inclusion of an open subset into a variety is birational.
(b) The map A
1
C, t (t
2
. t
3
), is birational. Here C is the cubic Y
2
=X
3
, and the
map kC kA
1
 =kT  identies kC with the subring kT
2
. T
3
 of kT . Both
rings have k(T ) as their elds of fractions.
(c) For any smooth variety V and point 1 V , there is a regular birational map : V
t
1
(1) is an isomorphism onto V 1, but
1
(1) is the projective space attached to the vector space T
1
(V ).
The next result says that, if we require the target variety to be normal (thereby excluding
example (b)), and we require the map to be quasinite (thereby excluding example (c)),
then we are left with (a).
COROLLARY 8.19. Let : W V be a birational regular map of irreducible varieties.
Assume
(a) V is normal, and
(b) is quasinite.
Then is an isomorphism of W onto an open subset of V .
PROOF. Factor as in the theorem. For each open afne subset U of V , k
t1
(U) is the
integral closure of kU in k(W). But k(W) = k(V ) (because is birational), and kU
is integrally closed in k(V ) (because V is normal), and so U =
t1
(U) (as varieties). It
follows that W
t
=V .
2
COROLLARY 8.20. Any quasinite regular map : W V with W complete is nite.
PROOF. In this case, : W W
t
must be an isomorphism (7.3).
2
152 8. FINITE MAPS
REMARK 8.21. Let W and V be irreducible varieties, and let : W V be a dominant
map. It induces a map k(V ) k(W), and if dimW =dimV , then k(W) is a nite exten
sion of k(V ). We shall see later that, if n is the separable degree of k(V ) over k(W), then
there is an open subset U of W such that is n : 1 on U, i.e., for 1 (U),
1
(1) has
exactly n points.
Now suppose that is a bijective regular map W V . We shall see later that this
implies that W and V have the same dimension. Assume:
(a) k(W) is separable over k(V );
(b) V is normal.
From (a) and the preceding discussion, we nd that is birational, and from (b) and
the corollary, we nd that is an isomorphism of W onto an open subset of V ; as it is
surjective, it must be an isomorphism of W onto V . We conclude: a bijective regular map
: W V satisfying the conditions (a) and (b) is an isomorphism.
NOTES. The full name of Theorem 8.16 is the main theorem of Zariskis paper Transactions AMS,
53 (1943), 490532. Zariskis original statement is that in (8.19). Grothendieck proved it in the
stronger form (8.16) for all schemes. There is a good discussion of the theorem in Mumford 1999,
III.9. For a proof see Musili, C., Algebraic geometry for beginners. Texts and Readings in Mathe
matics, 20. Hindustan Book Agency, New Delhi, 2001, 65.
The base change of a nite map
Recall that the base change of a regular map : V S is the map
t
in the diagram:
V
S
W
)
0
V
_
0
W
)
S.
PROPOSITION 8.22. The base change of a nite map is nite.
PROOF. We may assume that all the varieties concerned are afne. Then the statement be
comes: if is a nite 1algebra, then
T
T,N is a nite Talgebra, which is obvious.
2
Proper maps
A regular map : V S of varieties is said to be proper if it is universally closed, that is,
if for all maps T S, the base change
t
: V
S
T T of is closed. Note that a variety
V is complete if and only if the map V {point] is proper. From its very denition, it is
clear that the base change of a proper map is proper. In particular, if : V S is proper,
then
1
(1) is a complete variety for all 1 S.
PROPOSITION 8.23. A nite map of varieties is proper.
PROOF. The base change of a nite map is nite, and hence closed.
2
PROPOSITION 8.24. If V S is proper and S is complete, then V is complete.
Proper maps 153
PROOF. Let T be a variety, and consider
V V T
_closed
S S T
_closed
{point] T
As V T .V
S
(S T ) and V S is proper, the map V T S T is closed, and as
S is complete, the map S T T is closed. Therefore, V T T is closed.
2
COROLLARY 8.25. The inverse image of a complete variety under a proper map is com
plete.
PROOF. Let : V S be proper, and let 7 be a complete subvariety of S. Then V
S
7
7 is proper, and V
S
7 .
1
(7).
2
PROPOSITION 8.26. Let : V S be a proper map. The image W of any complete
subvariety W of V is a complete subvariety of S.
PROOF. The image W is certainly closed in S. Let T be a variety, and consider the
diagram
W W T
_
onto
_
onto
W W T
_
T.
Any closed subvariety 7 of W T is the image of a closed subvariety 7
t
of W T
(namely, of its inverse image), and the image of 7
t
in T is closed because W is complete.
2
The next result (whose proof requires Zariskis Main Theorem) gives a purely geometric
criterion for a regular map to be nite.
PROPOSITION 8.27. A proper quasinite map : W V of varieties is nite.
PROOF. Factor into W
t
W
t
W with nite and  an open immersion. Factor  into
W
u(u,tu)
W
V
W
t
(u,u
0
)u
0
W
t
.
The image of the rst map is 1
t
, which is closed because W
t
is a variety (see 4.26; W
t
is
separated because it is nite over a variety exercise). Because is proper, the second
map is closed. Hence  is an open immersion with closed image. It follows that its image is
a connected component of W
t
, and that W is isomorphic to that connected component.
2
154 8. FINITE MAPS
If W and V are curves, then any surjective map W V is closed. Thus it is easy to
give examples of closed surjective quasinite nonnite maps. For example, the map
a a
n
: A
1
{0] A
1
.
which corresponds to the map on rings
kT  kT. T
1
. T T
n
.
is such a map. This doesnt violate the theorem, because the map is only closed, not uni
versally closed.
Exercises
81. Prove that a nite map is an isomorphism if and only if it is bijective and etale. (Cf.
Harris 1992, 14.9.)
82. Give an example of a surjective quasinite regular map that is not nite (different
from any in the notes).
83. Let : V W be a regular map with the property that
1
(U) is an open afne
subset of W whenever U is an open afne subset of V . Show that if V is separated, then so
also is W.
84. For every n _ 1, nd a nite map : W V with the following property: for all
1 _i _n,
V
i
def
={1 V [
1
(1) has _i points]
is a closed subvariety of dimension i .
CHAPTER 9
Dimension Theory
Throughout this chapter, k is an algebraically closed eld. Recall that to an irreducible
variety V , we attach a eld k(V ) it is the eld of fractions of kU for any open afne
subvariety U of V , and also the eld of fractions of O
1
for any point 1 in V . We dened
the dimension of V to be the transcendence degree of k(V ) over k. Note that, directly from
this denition, dimV = dimU for any open subvariety U of V . Also, that if W V is
a nite surjective map, then dimW = dimV (because k(W) is a nite eld extension of
k(V )).
When V is not irreducible, we dened the dimension of V to be the maximum dimen
sion of an irreducible component of V , and we said that V is pure of dimension J if the
dimensions of the irreducible components are all equal to J.
Let W be a subvariety of a variety V . The codimension of W in V is
codim
V
W =dimV dimW.
In 3 and 6 we proved the following results:
9.1. (a) The dimension of a linear subvariety of A
n
(that is, a subvariety dened by
linear equations) has the value predicted by linear algebra (see 2.24b, 5.12). In par
ticular, dimA
n
=n. As a consequence, dimP
n
=n.
(b) Let 7 be a proper closed subset of A
n
; then 7 has pure codimension one in A
n
if
and only if 1(7) is generated by a single nonconstant polynomial. Such a variety is
called an afne hypersurface (see 2.25 and 2.27)
1
.
(c) If V is irreducible and 7 is a proper closed subset of V , then dim7 < dimV (see
2.26).
Afne varieties
The fundamental additional result that we need is that, when we impose additional poly
nomial conditions on an algebraic set, the dimension doesnt go down by more than linear
algebra would suggest.
THEOREM 9.2. Let V be an irreducible afne variety, and let a nonzero regular function.
If has a zero on V , then its zero set is pure of dimension dim(V ) 1.
1
The careful reader will check that we didnt use 5.22 or 5.23 in the proof of 2.27.
155
156 9. DIMENSION THEORY
In other words: let V be a closed subvariety of A
n
and let J kX
1
. . . . . X
n
; then
V V(J) =
_
_
_
V if J is identically zero on V
0 if J has no zeros on V
hypersurface otherwise.
where by hypersurface we mean a closed subvariety of pure codimension 1.
We can also state it in terms of the algebras: let be an afne kalgebra; let
be neither zero nor a unit, and let p be a prime ideal that is minimal among those containing
( ); then
tr deg
k
,p =tr deg
k
1.
LEMMA 9.3. Let be an integral domain, and let 1 be a nite extension of the eld of
fractions 1 of . If 1 is integral over , then so also is Nm
1{1
. Hence, if is
integrally closed (e.g., if is a unique factorization domain), then Nm
1{1
. In this
last case, divides Nm
1{1
in the ring .
PROOF. Let g(X) be the minimum polynomial of over 1.
g(X) =X
i
a
i1
X
i1
a
0
.
In some extension eld 1 of 1, g(X) will split
g(X) =
i
i=1
(X
i
).
1
=.
i
i=1
i
=a
0
.
Because is integral over , each
i
is integral over (see the proof of 1.22), and it follows
that Nm
1{1
FT 5.41
= (
i
i=1
i
)
1:1()j
is integral over (see 1.16).
Now suppose is integrally closed, so that Nm . From the equation
0 =(
i1
a
i1
i2
a
1
) a
0
we see that divides a
0
in , and therefore it also divides Nm =a
n
r
0
.
2
PROOF (OF THEOREM 9.2). We rst show that it sufces to prove the theorem in the case
that V( ) is irreducible. Suppose 7
0
. . . . . 7
n
are the irreducible components of V( ).
There exists a point 1 7
0
that does not lie on any other 7
i
(otherwise the decomposition
V( ) =
_
7
i
would be redundant). As 7
1
. . . . . 7
n
are closed, there is an open neighbour
hood U of 1, which we can take to be afne, that does not meet any 7
i
except 7
0
. Now
V( [U) =7
0
U, which is irreducible.
As V( ) is irreducible, rad( ) is a prime ideal p kV . According to the Noether
normalization theorem (8.13), there is a nite surjective map : V A
d
, which realizes
k(V ) as a nite extension of the eld k(A
d
). We shall show that pkA
d
 =rad(
0
) where
0
=Nm
k(V ){k(A
d
)
. Hence
kA
d
,rad(
0
) kV ,p
is injective. As it is also nite, this shows that dimV( ) =dimV(
0
), and we already know
the theorem for A
d
(9.1b).
Afne varieties 157
By assumption kV  is nite (hence integral) over its subring kA
d
. According to the
lemma,
0
lies in kA
d
, and I claim that p kA
d
 =rad(
0
). The lemma shows that
divides
0
in kV , and so
0
( ) p. Hence (
0
) pkA
d
, which implies
rad(
0
) pkA
d

because p is radical. For the reverse inclusion, let g pkA
d
. Then g rad( ), and so
g
n
= h for some h kV , m N. Taking norms, we nd that
g
ne
=Nm( h) =
0
Nm(h) (
0
).
where e =k(V ) : k(A
n
), which proves the claim.
The inclusion kA
d
 kV  therefore induces an inclusion
kA
d
,rad(
0
) =kA
d
,pkA
d
 kV ,p.
which makes kV ,p into a nite algebra over kA
d
,rad(
0
). Hence
dimV(p) =dimV(
0
).
Clearly =0 =
0
=0, and
0
p =
0
is not a nonzero constant. Therefore dimV(
0
) =
J 1 by (9.1b).
2
COROLLARY 9.4. Let V be an irreducible variety, and let 7 be a maximal proper closed
irreducible subset of V . Then dim(7) =dim(V ) 1.
PROOF. For any open afne subset U of V meeting 7, dimU =dimV and dimU 7 =
dim7. We may therefore assume that V itself is afne. Let be a nonzero regular function
on V vanishing on 7, and let V( ) be the set of zeros of (in V ). Then 7 V( ) V ,
and 7 must be an irreducible component of V( ) for otherwise it wouldnt be maximal in
V . Thus Theorem 9.2 implies that dim7 =dimV 1.
2
COROLLARY 9.5 (TOPOLOGICAL CHARACTERIZATION OF DIMENSION). Suppose V is
irreducible and that
V V
1
V
d
=0
is a maximal chain of distinct closed irreducible subsets of V . Then dim(V ) =J. (Maximal
means that the chain cant be rened.)
PROOF. From (9.4) we nd that
dimV =dimV
1
1 =dimV
2
2 = =dimV
d
J =J.
2
REMARK 9.6. (a) The corollary shows that, when V is afne, dimV =Krull dimkV , but
it shows much more. Note that each V
i
in a maximal chain (as above) has dimension J i ,
and that any closed irreducible subset of V of dimension J i occurs as a V
i
in a maximal
chain. These facts translate into statements about ideals in afne kalgebras that do not hold
for all noetherian rings. For example, if is an afne kalgebra that is an integral domain,
then Krull dim
m
is the same for all maximal ideals of all maximal ideals in have
158 9. DIMENSION THEORY
the same height (we have proved 5.23). Moreover, if p is an ideal in kV  with height i ,
then there is a maximal (i.e., nonrenable) chain of distinct prime ideals
(0) p
1
p
2
p
d
=kV 
with p
i
=p.
(b) Now that we know that the two notions of dimension coincide, we can restate (9.2)
as follows: let be an afne kalgebra; let be neither zero nor a unit, and let p be a
prime ideal that is minimal among those containing ( ); then
Krull dim(,p) =Krull dim() 1.
This statement holds for all noetherian local rings (CA 16.3), and is called Krulls principal
ideal theorem.
COROLLARY 9.7. Let V be an irreducible variety, and let 7 be an irreducible component
of V(
1
. . . .
i
), where the
i
are regular functions on V . Then
codim(7) _r, i.e., dim(7) _dimV r.
For example, if the
i
have no common zero on V , so that V(
1
. . . . .
i
) is empty, then
there are no irreducible components, and the statement is vacuously true.
PROOF. As in the proof of (9.4), we can assume V to be afne. We use induction on
r. Because 7 is a closed irreducible subset of V(
1
. . . .
i1
), it is contained in some
irreducible component 7
t
of V(
1
. . . .
i1
). By induction, codim(7
t
) _ r 1. Also 7 is
an irreducible component of 7
t
V(
i
) because
7 7
t
V(
i
) V(
1
. . . . .
i
)
and 7 is a maximal closed irreducible subset of V(
1
. . . . .
i
). If
i
vanishes identically on
7
t
, then 7 =7
t
and codim(7) =codim(7
t
) _ r 1; otherwise, the theorem shows that
7 has codimension 1 in 7
t
, and codim(7) =codim(7
t
) 1 _r.
2
PROPOSITION 9.8. Let V and W be closed subvarieties of A
n
; for any (nonempty) irre
ducible component 7 of V W,
dim(7) _dim(V ) dim(W) n:
that is,
codim(7) _codim(V ) codim(W).
PROOF. In the course of the proof of (4.27), we showed that V W is isomorphic to z
(V W), and this is dened by the n equations X
i
= Y
i
in V W. Thus the statement
follows from (9.7).
2
REMARK 9.9. (a) The example (in A
3
)
_
X
2
Y
2
=7
2
7 =0
Afne varieties 159
shows that Proposition 9.8 becomes false if one only looks at real points. Also, that the
pictures we draw can mislead.
(b) The statement of (9.8) is false if A
n
is replaced by an arbitrary afne variety. Con
sider for example the afne cone V
X
1
X
4
X
2
X
3
=0.
It contains the planes,
7 : X
2
=0 =X
4
: 7 ={(+. 0. +. 0)]
7
t
: X
1
=0 =X
3
: 7
t
={(0. +. 0. +)]
and 77
t
={(0. 0. 0. 0)]. Because V is a hypersurface in A
4
, it has dimension 3, and each
of 7 and 7
t
has dimension 2. Thus
codim77
t
=3 _11 =codim7codim7
t
.
The proof of (9.8) fails because the diagonal in V V cannot be dened by 3 equations
(it takes the same 4 that dene the diagonal in A
4
) the diagonal is not a settheoretic
complete intersection.
REMARK 9.10. In (9.7), the components of V(
1
. . . . .
i
) need not all have the same di
mension, and it is possible for all of them to have codimension < r without any of the
i
being redundant.
For example, let V be the same afne cone as in the above remark. Note that V(X
1
)V
is a union of the planes:
V(X
1
) V ={(0. 0. +. +)] L{(0. +. 0. +)].
Both of these have codimension 1 in V (as required by (9.2)). Similarly, V(X
2
) V is the
union of two planes,
V(X
2
) V ={(0. 0. +. +)] L{(+. 0. +. 0)].
but V(X
1
. X
2
) V consists of a single plane {(0. 0. +. +)]: it is still of codimension 1 in V ,
but if we drop one of two equations from its dening set, we get a larger set.
PROPOSITION 9.11. Let 7 be a closed irreducible subvariety of codimension r in an afne
variety V . Then there exist regular functions
1
. . . . .
i
on V such that 7 is an irreducible
component of V(
1
. . . . .
i
) and all irreducible components of V(
1
. . . . .
i
) have codimen
sion r.
PROOF. We know that there exists a chain of closed irreducible subsets
V 7
1
7
i
=7
with codim 7
i
= i . We shall show that there exist
1
. . . . .
i
kV  such that, for all
s _ r, 7
x
is an irreducible component of V(
1
. . . . .
x
) and all irreducible components of
V(
1
. . . . .
x
) have codimension s.
160 9. DIMENSION THEORY
We prove this by induction on s. For s = 1, take any
1
1(7
1
),
1
= 0, and apply
Theorem 9.2. Suppose
1
. . . . .
x1
have been chosen, and let Y
1
=7
x1
. . . . . Y
n
, be the
irreducible components of V(
1
. . . . .
x1
). We seek an element
x
that is identically zero
on 7
x
but is not identically zero on any Y
i
for such an
x
, all irreducible components of
Y
i
V(
x
) will have codimension s, and 7
x
will be an irreducible component of Y
1
V(
x
).
But Y
i
_7
x
for any i (7
x
has smaller dimension than Y
i
), and so 1(7
x
) _1(Y
i
). Now the
prime avoidance lemma (see below) tells us that there is an element
x
1(7
x
) such that
x
1(Y
i
) for any i , and this is the function we want.
2
LEMMA 9.12 (PRIME AVOIDANCE LEMMA). If an ideal a of a ring is not contained in
any of the prime ideals p
1
. . . . . p
i
, then it is not contained in their union.
PROOF. We may assume that none of the prime ideals is contained in a second, because
then we could omit it. Fix an i
0
and, for each i = i
0
, choose an
i
p
i
.
i
p
i
0
, and
choose
i
0
a,
i
0
p
i
0
. Then h
i
0
def
=
i
lies in each p
i
with i =i
0
and a, but not in p
i
0
(here we use that p
i
0
is prime). The element
i
i=1
h
i
is therefore in a but not in any p
i
.
2
REMARK 9.13. The proposition shows that for a prime ideal p in an afne kalgebra, if
p has height r, then there exist elements
1
. . . . .
i
such that p is minimal among the
prime ideals containing (
1
. . . . .
i
). This statement is true for all noetherian local rings.
REMARK 9.14. The last proposition shows that a curve C in A
3
is an irreducible com
ponent of V(
1
.
2
) for some
1
,
2
kX. Y. 7. In fact C = V(
1
.
2
.
3
) for suitable
polynomials
1
.
2
, and
3
this is an exercise in Shafarevich 1994 (I 6, Exercise 8; see
also Hartshorne 1977, I, Exercise 2.17). Apparently, it is not known whether two polynomi
als always sufce to dene a curve in A
3
see Kunz 1985, p136. The union of two skew
lines in P
3
cant be dened by two polynomials (ibid. p140), but it is unknown whether
all connected curves in P
3
can be dened by two polynomials. Macaulay (the man, not the
program) showed that for every r _ 1, there is a curve C in A
3
such that 1(C) requires
at least r generators (see the same exercise in Hartshorne for a curve whose ideal cant be
generated by 2 elements).
In general, a closed variety V of codimension r in A
n
(resp. P
n
) is said to be a set
theoretic complete intersection if there exist r polynomials
i
kX
1
. . . . . X
n
 (resp. ho
mogeneous polynomials
i
kX
0
. . . . . X
n
) such that
V =V(
1
. . . . .
i
).
Such a variety is said to be an idealtheoretic complete intersection if the
i
can be chosen
so that 1(V ) =(
1
. . . . .
i
). Chapter V of Kunzs book is concerned with the question of
when a variety is a complete intersection. Obviously there are many idealtheoretic com
plete intersections, but most of the varieties one happens to be interested in turn out not
to be. For example, no abelian variety of dimension > 1 is an idealtheoretic complete
intersection (being an idealtheoretic complete intersection imposes constraints on the co
homology of the variety, which are not fullled in the case of abelian varieties).
2
2
In 1882 Kronecker proved that every algebraic subset in P
n
can be cut out by n1 polynomial equations.
In 1891 Vahlen asserted that the result was best possible by exhibiting a curve in P
3
which he claimed was not
the zero locus of 3 equations. It was only 50 years later, in 1941, that Perron gave 3 equations dening Vahlens
Afne varieties 161
Let 1 be a point on an irreducible variety V A
n
. Then (9.11) shows that there
is a neighbourhood U of 1 in A
n
and functions
1
. . . . .
i
on U such that U V =
V(
1
. . . . .
i
) (zero set in U). Thus U V is a settheoretic complete intersection in U.
One says that V is a local complete intersection at 1 V if there is an open afne neigh
bourhood U of 1 in A
n
such that 1(V U) can be generated by r regular functions on U.
Note that
idealtheoretic complete intersection =local complete intersection at all p.
It is not difcult to show that a variety is a local complete intersection at every nonsingular
point (cf. 5.17).
PROPOSITION 9.15. Let 7 be a closed subvariety of codimension r in variety V , and let
1 be a point of 7 that is nonsingular when regarded both as a point on 7 and as a point on
V . Then there is an open afne neighbourhood U of 1 and regular functions
1
. . . . .
i
on
U such that 7U =V(
1
. . . . .
i
).
PROOF. By assumption
dim
k
T
1
(7) =dim7 =dimV r =dim
k
T
1
(V ) r.
There exist functions
1
. . . . .
i
contained in the ideal of O
1
corresponding to 7 such that
T
1
(7) is the subspace of T
1
(V ) dened by the equations
(J
1
)
1
=0. . . . . (J
i
)
1
=0.
All the
i
will be dened on some open afne neighbourhood U of 1 (in V ), and clearly
7 is the only component of 7
t
def
= V(
1
. . . . .
i
) (zero set in U) passing through 1. Af
ter replacing U by a smaller neighbourhood, we can assume that 7
t
is irreducible. As
1
. . . . .
i
1(7
t
), we must have T
1
(7
t
) T
1
(7), and therefore dim7
t
_ dim7. But
1(7
t
) 1(7U), and so 7
t
7U. These two facts imply that 7
t
=7U.
2
PROPOSITION 9.16. Let V be an afne variety such that kV  is a unique factorization
domain. Then every pure closed subvariety 7 of V of codimension one is principal, i.e.,
1(7) =( ) for some kV .
PROOF. In (2.27) we proved this in the case that V =A
n
, but the argument only used that
kA
n
 is a unique factorization domain.
2
EXAMPLE 9.17. The condition that kV  is a unique factorization domain is denitely
needed. Again let V be the cone
X
1
X
4
X
2
X
3
=0
in A
4
and let 7 and 7
t
be the planes
7 ={(+. 0. +. 0)] 7
t
={(0. +. 0. +)].
curve, thus refuting Vahlens claim which had been accepted for half a century. Finally, in 1973 Eisenbud and
Evans proved that n equations always sufce to describe (settheoretically) any algebraic subset of P
n
(Georges
Elencwajg on mathoverow.net).
162 9. DIMENSION THEORY
Then 77
t
={(0. 0. 0. 0)], which has codimension 2 in 7
t
. If 7 =V( ) for some reg
ular function on V , then V( [7
t
) = {(0. . . . . 0)], which is impossible (because it has
codimension 2, which violates 9.2). Thus 7 is not principal, and so
kX
1
. X
2
. X
3
. X
4
,(X
1
X
4
X
2
X
3
)
is not a unique factorization domain.
Projective varieties
The results for afne varieties extend to projective varieties with one important simplica
tion: if V and W are projective varieties of dimensions r and s in P
n
and r s _ n, then
V W =0.
THEOREM 9.18. Let V = V(a) P
n
be a projective variety of dimension _ 1, and let
kX
0
. . . . . X
n
 be homogeneous, nonconstant, and a; then V V( ) is nonempty
and of pure codimension 1.
PROOF. Since the dimension of a variety is equal to the dimension of any dense open afne
subset, the only part that doesnt follow immediately from (9.2) is the fact that V V( )
is nonempty. Let V
aff
(a) be the zero set of a in A
n1
(that is, the afne cone over V ).
Then V
aff
(a) V
aff
( ) is nonempty (it contains (0. . . . . 0)), and so it has codimension 1 in
V
aff
(a). Clearly V
aff
(a) has dimension _ 2, and so V
aff
(a) V
aff
( ) has dimension _ 1.
This implies that the polynomials in a have a zero in common with other than the origin,
and so V(a) V( ) =0.
2
COROLLARY 9.19. Let
1
. .
i
be homogeneous nonconstant elements of kX
0
. . . . . X
n
;
and let 7 be an irreducible component of V V(
1
. . . .
i
). Then codim(7) _ r, and if
dim(V ) _r, then V V(
1
. . . .
i
) is nonempty.
PROOF. Induction on r, as before.
2
COROLLARY 9.20. Let : P
n
P
n
be regular; if m<n, then is constant.
PROOF.
3
Let : A
n1
{origin] P
n
be the map (a
0
. . . . . a
n
) (a
0
: . . . : a
n
). Then
is regular, and there exist polynomials J
0
. . . . . J
n
kX
0
. . . . . X
n
 such that is
the map
(a
0
. . . . . a
n
) (J
0
(a) : . . . : J
n
(a)).
As factors through P
n
, the J
i
must be homogeneous of the same degree. Note that
(a
0
: . . . : a
n
) =(J
0
(a) : . . . : J
n
(a)).
If m < n and the J
i
are nonconstant, then (9.18) shows they have a common zero and so
is not dened on all of P
n
. Hence the J
i
s must be constant.
2
3
Lars Kindler points out that, in this proof, it is not obvious that the map is given globally by a
system of polynomials (rather than just locally). It is in fact given globally, and this is not too difcult to prove:
a regular map from a variety V to P
n
corresponds to a line bundle on V and a set of global sections, and all line
bundles on A
n
are trivial (see, for example, Hartshorne II 7.1 and II 6.2). I should x this in a future version.
Projective varieties 163
PROPOSITION 9.21. Let 7 be a closed irreducible subvariety of V ; if codim(7) =r, then
there exist homogeneous polynomials
1
. . . . .
i
in kX
0
. . . . . X
n
 such that 7 is an irre
ducible component of V V(
1
. . . . .
i
).
PROOF. Use the same argument as in the proof (9.11).
2
PROPOSITION 9.22. Every pure closed subvariety 7 of P
n
of codimension one is princi
pal, i.e., 1(7) =( ) for some homogeneous element of kX
0
. . . . . X
n
.
PROOF. Follows from the afne case.
2
COROLLARY 9.23. Let V and W be closed subvarieties of P
n
; if dim(V ) dim(W) _ n,
then V W =0, and every irreducible component of it has codim(7) _codim(V )codim(W).
PROOF. Write V = V(a) and W = V(b), and consider the afne cones V
t
= V(a) and
W
t
=V(b) over them. Then
dim(V
t
) dim(W
t
) =dim(V ) 1dim(W) 1 _n2.
As V
t
W
t
= 0, V
t
W
t
has dimension _ 1, and so it contains a point other than the
origin. Therefore V W =0. The rest of the statement follows from the afne case.
2
PROPOSITION 9.24. Let V be a closed subvariety of P
n
of dimension r <n; then there is a
linear projective variety 1 of dimension nr 1 (that is, 1 is dened by r 1 independent
linear forms) such that 1V =0.
PROOF. Induction on r. If r =0, then V is a nite set, and the lemma below shows that
there is a hyperplane in k
n1
not meeting V .
Suppose r > 0, and let V
1
. . . . . V
x
be the irreducible components of V . By assumption,
they all have dimension _ r. The intersection 1
i
of all the linear projective varieties con
taining V
i
is the smallest such variety. The lemma below shows that there is a hyperplane H
containing none of the nonzero 1
i
; consequently, H contains none of the irreducible com
ponents V
i
of V , and so each V
i
H is a pure variety of dimension _ r 1 (or is empty).
By induction, there is an linear subvariety 1
t
not meeting V H. Take 1 =1
t
H.
2
LEMMA 9.25. Let W be a vector space of dimension J over an innite eld k, and let
1
1
. . . . . 1
i
be a nite set of nonzero subspaces of W. Then there is a hyperplane H in W
containing none of the 1
i
.
PROOF. Pass to the dual space V of W. The problem becomes that of showing V is not
a nite union of proper subspaces 1
i
. Replace each 1
i
by a hyperplane H
i
containing
it. Then H
i
is dened by a nonzero linear form 1
i
. We have to show that
1
}
is not
identically zero on V . But this follows from the statement that a polynomial in n variables,
with coefcients not all zero, can not be identically zero on k
n
(Exercise 11).
2
Let V and 1 be as in Proposition 9.24. If 1 is dened by the linear forms 1
0
. . . . . 1
i
then the projection a (1
0
(a) : : 1
i
(a)) denes a map V P
i
. We shall see later
that this map is nite, and so it can be regarded as a projective version of the Noether
normalization theorem.
CHAPTER 10
Regular Maps and Their Fibres
Throughout this chapter, k is an algebraically closed eld.
Consider again the regular map : A
2
A
2
, (.. ,) (.. .,) (Exercise 33). The
image of is
C ={(a. b) A
2
[ a =0 or a =0 =b]
=(A
2
{,axis]) L{(0. 0)].
which is neither open nor closed, and, in fact, is not even locally closed. The bre
1
(a. b) =
_
_
_
{(a. b,a)] if a =0
Y axis if (a. b) =(0. 0)
0 if a =0, b =0.
From this unpromising example, it would appear that it is not possible to say anything about
the image of a regular map, nor about the dimension or number of elements in its bres.
However, it turns out that almost everything that can go wrong already goes wrong for this
map. We shall show:
(a) the image of a regular map is a nite union of locally closed sets;
(b) the dimensions of the bres can jump only over closed subsets;
(c) the number of elements (if nite) in the bres can drop only on closed subsets, pro
vided the map is nite, the target variety is normal, and k has characteristic zero.
Constructible sets
Let W be a topological space. A subset C of W is said to constructible if it is a nite union
of sets of the form U 7 with U open and 7 closed. Obviously, if C is constructible
and V W, then C V is constructible. A constructible set in A
n
is denable by a nite
number of polynomials; more precisely, it is dened by a nite number of statements of the
form
(X
1
. . X
n
) =0. g(X
1
. . X
n
) =0
combined using only and and or (or, better, statements of the form = 0 combined
using and, or, and not). The next proposition shows that a constructible set C that
is dense in an irreducible variety V must contain a nonempty open subset of V . Contrast
, which is dense in 1 (real topology), but does not contain an open subset of 1, or any
innite subset of A
1
that omits an innite set.
165
166 10. REGULAR MAPS AND THEIR FIBRES
PROPOSITION 10.1. Let C be a constructible set whose closure
C is irreducible. Then C
contains a nonempty open subset of
C.
PROOF. We are given that C =
_
(U
i
7
i
) with each U
i
open and each 7
i
closed. We
may assume that each set U
i
7
i
in this decomposition is nonempty. Clearly
C
_
7
i
,
and as
C is irreducible, it must be contained in one of the 7
i
. For this i
C U
i
7
i
U
i
C U
i
C U
i
(U
i
7
i
) =U
i
7
i
.
Thus U
i
7
i
=U
i
C is a nonempty open subset of
C contained in C.
2
THEOREM 10.2. A regular map : W V sends constructible sets to constructible sets.
In particular, if U is a nonempty open subset of W, then (U) contains a nonempty open
subset of its closure in V .
The key result we shall need from commutative algebra is the following. (In the next
two results, and T are arbitrary commutative ringsthey need not be kalgebras.)
PROPOSITION 10.3. Let T be integral domains with T nitely generated as an algebra
over , and let b be a nonzero element of T. Then there exists an element a = 0 in
with the following property: every homomorphism : from into an algebraically
closed eld such that (a) =0 can be extended to a homomorphism : T such that
(b) =0.
Consider, for example, the rings kX kX. X
1
. A homomorphism : kX k
extends to a homomorphism kX. X
1
 k if and only if (X) =0. Therefore, for b =1,
we can take a =X. In the application we make of Proposition 10.3, we only really need the
case b =1, but the more general statement is needed so that we can prove it by induction.
LEMMA 10.4. Let T be integral domains, and assume T =t  ~T ,a. Let c
be the set of leading coefcients of the polynomials in a. Then every homomorphism
: from into an algebraically closed eld such that (c) =0 can be extended
to a homomorphism of T into .
PROOF. Note that c is an ideal in . If a =0, then c =0, and there is nothing to prove (in
fact, every extends). Thus we may assume a =0. Let =a
n
T
n
a
0
be a nonzero
polynomial of minimum degree in a such that (a
n
) = 0. Because T = 0, we have that
m_1.
Extend to a homomorphism : T  T  by sending T to T . The submodule
of T  generated by (a) is an ideal (because T
c
i
(g
i
) =
c
i
(g
i
T )). Therefore,
unless (a) contains a nonzero constant, it generates a proper ideal in T , which will
have a zero c in . The homomorphism
T 
T  . T T c
then factors through T ,a =T and extends .
In the contrary case, a contains a polynomial
g(T ) =b
n
T
n
b
0
. (b
i
) =0 (i >0). (b
0
) =0.
Constructible sets 167
On dividing (T ) into g(T ) we nd that
a
d
n
g(T ) =q(T )(T ) r(T ). J N. q. r T . degr <m.
On applying to this equation, we obtain
(a
n
)
d
(b
0
) = (q) ( ) (r).
Because ( ) has degree m>0, we must have (q) =0, and so (r) is a nonzero constant.
After replacing g(T ) with r(T ), we may assume n <m. If m=1, such a g(T ) cant exist,
and so we may suppose m > 1 and (by induction) that the lemma holds for smaller values
of m.
For h(T ) = c
i
T
i
c
i1
T
i1
c
0
, let h
t
(T ) = c
i
c
0
T
i
. Then the 
module generated by the polynomials T
x
h
t
(T ), s _ 0, h a, is an ideal a
t
in T . More
over, a
t
contains a nonzero constant if and only if a contains a nonzero polynomial cT
i
,
which implies t =0 and =T (since T is an integral domain).
If a
t
does not contain nonzero constants, then set T
t
=T ,a
t
=t
t
. Then a
t
con
tains the polynomial g
t
=b
n
b
0
T
n
, and (b
0
)=0. Because degg
t
< m, the induc
tion hypothesis implies that extends to a homomorphism T
t
. Therefore, there is a
c such that, for all h(T ) =c
i
T
i
c
i1
T
i1
c
0
a,
h
t
(c) =(c
i
) (c
i1
)c c
0
c
i
=0.
On taking h = g, we see that c = 0, and on taking h = , we obtain the contradiction
(a
n
) =0.
2
PROOF (OF 10.3) Suppose that we know the proposition in the case that T is generated
by a single element, and write T =.
1
. . . . . .
n
. Then there exists an element b
n1
such
that any homomorphism: .
1
. . . . . .
n1
 such that (b
n1
) =0 extends to a homo
morphism : T such that (b) =0. Continuing in this fashion, we obtain an element
a with the required property.
Thus we may assume T = .. Let a be the kernel of the homomorphism X .,
X ..
Case (i). The ideal a =(0). Write
b =(.) =a
0
.
n
a
1
.
n1
a
n
. a
i
.
and take a =a
0
. If : is such that (a
0
) =0, then there exists a c such that
(c) =0, and we can take to be the homomorphism
J
i
.
i
(J
i
)c
i
.
Case (ii). The ideal a = (0). Let (T ) = a
n
T
n
, a
n
= 0, be an element of a
of minimum degree. Let h(T ) T  represent b. Since b = 0, h a. Because is
irreducible over the eld of fractions of , it and h are coprime over that eld. Hence there
exist u. T  and c {0] such that
uh =c.
It follows now that ca
n
satises our requirements, for if (ca
n
) = 0, then can be ex
tended to : T by the previous lemma, and (u(.) b) =(c) =0, and so (b) =0.
2
168 10. REGULAR MAPS AND THEIR FIBRES
ASIDE 10.5. In case (ii) of the above proof, both b and b
1
are algebraic over , and so
there exist equations
a
0
b
n
a
n
=0. a
i
. a
0
=0:
a
t
0
b
n
a
t
n =0. a
t
i
. a
t
0
=0.
One can show that a =a
0
a
t
0
has the property required by the Propositionsee Atiyah and
MacDonald, 5.23.
PROOF (OF 10.2) We rst prove the in particular statement of the Theorem. By consid
ering suitable open afne coverings of W and V , one sees that it sufces to prove this in the
case that both W and V are afne. If W
1
. . . . . W
i
are the irreducible components of W, then
the closure of (W) in V , (W)

= (W
1
)

L. . . L(W
i
)

, and so it sufces to prove
the statement in the case that W is irreducible. We may also replace V with (W)

, and so
assume that both W and V are irreducible. Then corresponds to an injective homomor
phism T of afne kalgebras. For some b =0, D(b) U. Choose a as in the lemma.
Then for any point 1 D(a), the homomorphism (1): k extends to a homo
morphism : T k such that (b) =0. The kernel of is a maximal ideal corresponding
to a point Q D(b) lying over 1.
We now prove the theorem. Let W
i
be the irreducible components of W. Then C W
i
is constructible in W
i
, and (W) is the union of the (C W
i
); it is therefore constructible
if the (C W
i
) are. Hence we may assume that W is irreducible. Moreover, C is a
nite union of its irreducible components, and these are closed in C; they are therefore
constructible. We may therefore assume that C also is irreducible;
C is then an irreducible
closed subvariety of W.
We shall prove the theorem by induction on the dimension of W. If dim(W) =0, then
the statement is obvious because W is a point. If
C = W, then dim(
C) < dim(W), and
because C is constructible in
C, we see that (C) is constructible (by induction). We may
therefore assume that
C =W. But then
C contains a nonempty open subset of W, and so
the case just proved shows that (C) contains an nonempty open subset U of its closure.
Replace V be the closure of (C), and write
(C) =U L(C
1
(V U)).
Then
1
(V U) is a proper closed subset of W (the complement of V U is dense
in V and is dominant). As C
1
(V U) is constructible in
1
(V U), the set
(C
1
(V U)) is constructible in V by induction, which completes the proof.
2
Orbits of group actions
Let G be an algebraic group. An action of G on a variety V is a regular map
(g. 1) g1: GV V
such that
(a) 1
G
1 =1, all 1 V ;
(b) g(g
t
1) =(gg
t
)1, all g. g
t
G, 1 V .
Orbits of group actions 169
PROPOSITION 10.6. Let GV V be an action of an algebraic group G on a variety V .
(a) Each orbit of G in X is open in its closure.
(b) There exist closed orbits.
PROOF. (a) Let O be an orbit of G in V and let 1 O. Then g g1: G V is a
regular map with image O, and so O contains a nonempty set U open in
O (10.2). As
O =
_
G(k)
gU, it is open in
O.
(b) Let S =
O be minimal among the closures of orbits. From (a), we know that O is
open in S. Therefore, if S O were nonempty, it would contain the closure of an orbit,
contradicting the minimality of S. Hence S =O.
2
Let G be an algebraic group acting on a variety V . Let G\V denote the quotient
topological space with the sheaf O
G\V
such that 1(U. O
G\V
) =1(
1
U. O
V
)
G
, where
: G G,V is the quotient map. When (G\V. O
G\V
) is a variety, we call it the geometric
quotient of V under the action of G.
PROPOSITION 10.7. Let N be a normal algebraic subgroup of an afne algebraic group
G. Then the geometric quotient of G by N exists, and is an afne algebraic group.
PROOF. Omitted for the present.
2
A connected afne algebraic group G is solvable if there exist connected algebraic
subgroups
G =G
d
G
d1
G
0
={1]
such that G
i
is normal in G
i1
, and G
i
,G
i1
is commutative.
THEOREM 10.8 (BOREL FIXED POINT THEOREM). Aconnected solvable afne algebraic
group G acting on a complete algebraic variety V has at least one xed point.
PROOF. We prove this by induction on the dimG. Assume rst that G is commutative, and
let O =G. be a closed orbit of G in V (see 10.6). Let N be the stabilizer of .. Because G
is commutative, N is normal, and we get a bijection G,N O. As G acts transitively on
G,N and O, the map G,N O is proper (see Exercise 104); as O is complete (7.3a), so
also is G,N (see 8.24), and as it is afne and connected, it consists of a single point (7.5).
Therefore, O consists of a single point, which is a xed point for the action.
By assumption, there exists a closed normal subgroup H of G such that G,H is a
commutative. The set X
1
of xed points of H in X is nonempty (by induction) and closed
(because it is the intersection of the sets
X
h
={. X [ h. =.]
for h H). Because H is normal, X
1
is stable under G, and the action of G on it factors
through G,H. Every xed point of G,H in X
1
is a xed point for G acting on X.
2
170 10. REGULAR MAPS AND THEIR FIBRES
The bres of morphisms
We wish to examine the bres of a regular map : W V . Clearly, we can replace V by
the closure of (W) in V and so assume to be dominant.
THEOREM 10.9. Let : W V be a dominant regular map of irreducible varieties. Then
(a) dim(W) _dim(V );
(b) if 1 (W), then
dim(
1
(1)) _dim(W) dim(V )
for every 1 V , with equality holding exactly on a nonempty open subset U of V .
(c) The sets
V
i
={1 V [ dim(
1
(1)) _i ]
are closed in (W).
EXAMPLE 10.10. Consider the subvariety W V A
n
dened by r linear equations
n
}=1
a
i}
X
}
=0. a
i}
kV . i =1. . . . . r.
and let be the projection W V . For 1 V ,
1
(1) is the set of solutions of
n
}=1
a
i}
(1)X
}
=0. a
i}
(1) k. i =1. . . . . r.
and so its dimension is mrank(a
i}
(1)). Since the rank of the matrix (a
i}
(1)) drops on
closed subsets, the dimension of the bre jumps on closed subsets.
PROOF. (a) Because the map is dominant, there is a homomorphism k(V ) k(W), and
obviously tr deg
k
k(V ) _ tr deg
k
k(W) (an algebraically independent subset of k(V ) re
mains algebraically independent in k(W)).
(b) In proving the rst part of (b), we may replace V by any open neighbourhood of 1.
In particular, we can assume V to be afne. Let m be the dimension of V . From (9.11) we
know that there exist regular functions
1
. . . . .
n
such that 1 is an irreducible component
of V(
1
. . . . .
n
). After replacing V by a smaller neighbourhood of 1, we can suppose that
1 =V(
1
. . . . .
n
). Then
1
(1) is the zero set of the regular functions
1
. . . . .
n
,
and so (if nonempty) has codimension _m in W (see 9.7). Hence
dim
1
(1) _dimW m=dim(W) dim(V ).
In proving the second part of (b), we can replace both W and V with open afne sub
sets. Since is dominant, kV  kW is injective, and we may regard it as an in
clusion (we identify a function . on V with . on W). Then k(V ) k(W). Write
kV  =k.
1
. . . . . .
1
(1) _ 7
i
for any i , but
1
(1) is irreducible and
1
(1) =
_
(
1
(1) 7
i
), and
so this is impossible.
2
172 10. REGULAR MAPS AND THEIR FIBRES
The bres of nite maps
Let : W V be a nite dominant morphism of irreducible varieties. Then dim(W) =
dim(V ), and so k(W) is a nite eld extension of k(V ). Its degree is called the degree of
the map .
THEOREM 10.12. Let : W V be a nite surjective regular map of irreducible varieties,
and assume that V is normal.
(a) For all 1 V , #
1
(1) _deg().
(b) The set of points 1 of V such that #
1
(1) =deg() is an open subset of V , and it
is nonempty if k(W) is separable over k(V ).
Before proving the theorem, we give examples to show that we need W to be separated
and V to be normal in (a), and that we need k(W) to be separable over k(V ) for the second
part of (b).
EXAMPLE 10.13. (a) Consider the map
{A
1
with origin doubled ] A
1
.
The degree is one and that map is onetoone except at the origin where it is twotoone.
(b) Let C be the curve Y
2
=X
3
X
2
, and consider the map
t (t
2
1. t (t
2
1)): A
1
C.
It is onetoone except that the points t =1 both map to 0. On coordinate rings, it corre
sponds to the inclusion
k.. , kT , . T
2
1, , T(T
2
1).
and so is of degree one. The ring k.. , is not integrally closed; in fact kT  is its integral
closure in its eld of fractions.
(c) Consider the Frobenius map : A
n
A
n
, (a
1
. . . . . a
n
) (a
]
1
. . . . . a
]
n
), where =
chark. This map has degree
n
but it is onetoone. The eld extension corresponding to
the map is
k(X
1
. . . . . X
n
) k(X
]
1
. . . . . X
]
n
)
which is purely inseparable.
LEMMA 10.14. Let Q
1
. . . . . Q
i
be distinct points on an afne variety V . Then there is a
regular function on V taking distinct values at the Q
i
.
PROOF. We can embed V as closed subvariety of A
n
, and then it sufces to prove the
statement with V =A
n
almost any linear form will do.
2
PROOF (OF 10.12). In proving (a) of the theorem, we may assume that V and W are afne,
and so the map corresponds to a nite map of kalgebras, kV  kW. Let
1
(1) =
{Q
1
. . . . . Q
i
]. According to the lemma, there exists an kW taking distinct values at
the Q
i
. Let
J(T ) =T
n
a
1
T
n1
a
n
Flat maps 173
be the minimum polynomial of over k(V ). It has degree m _ k(W) : k(V ) = deg,
and it has coefcients in kV  because V is normal (see 1.22). Now J( ) = 0 implies
J((Q
i
)) =0, i.e.,
(Q
i
)
n
a
1
(1) (Q
i
)
n1
a
n
(1) =0.
Therefore the (Q
i
) are all roots of a single polynomial of degree m, and so r _ m _
deg().
In order to prove the rst part of (b), we show that, if there is a point 1 V such that
1
(1) has deg() elements, then the same is true for all points in an open neighbourhood
of 1. Choose as in the last paragraph corresponding to such a 1. Then the polynomial
T
n
a
1
(1) T
n1
a
n
(1) =0 (*)
has r =deg distinct roots, and so m=r. Consider the discriminant discJ of J. Because
(*) has distinct roots, disc(J)(1) =0, and so disc(J) is nonzero on an open neighbourhood
U of 1. The factorization
kV  kV T ,(J)
T(
kW
gives a factorization
W Spm(kV T ,(J)) V.
Each point 1
t
U has exactly m inverse images under the second map, and the rst map is
nite and dominant, and therefore surjective (recall that a nite map is closed). This proves
that
1
(1
t
) has at least deg() points for 1
t
U, and part (a) of the theorem then implies
that it has exactly deg() points.
We now show that if the eld extension is separable, then there exists a point such
that #
1
(1) has deg elements. Because k(W) is separable over k(V ), there exists a
kW such that k(V )  =k(W). Its minimum polynomial J has degree deg() and
its discriminant is a nonzero element of kV . The diagram
W Spm(T ,(J)) V
shows that #
1
(1) _deg() for 1 a point such that disc( )(1) =0.
2
When k(W) is separable over k(V ), then is said to be separable.
REMARK 10.15. Let : W V be as in the theorem, and let V
i
={1 V [ #
1
(1) _i ].
Let J =deg. Part (b) of the theoremstates that V
d1
is closed, and is a proper subset when
is separable. I dont know under what hypotheses all the sets V
i
will closed (and V
i
will
be a proper subset of V
i1
). The obvious induction argument fails because V
i1
may not
be normal.
Flat maps
A regular map : V W is at if for all 1 V , the homomorphismO
(1)
O
1
dened
by is at. If is at, then for every pair U and U
t
of open afnes of V and W such that
(U) U
t
the map 1(U
t
. O
W
) 1(U. O
V
) is at; conversely, if this condition holds for
sufciently many pairs that the Us cover V and the U
t
s cover W, then is at.
174 10. REGULAR MAPS AND THEIR FIBRES
PROPOSITION 10.16. (a) An open immersion is at.
(b) The composite of two at maps is at.
(c) Any base extension of a at map is at.
PROOF. To be added.
2
THEOREM 10.17. A nite map : V W is at if and only if
Q1
dim
k
O
Q
,m
1
O
Q
is independent of 1 W.
PROOF. To be added.
2
THEOREM 10.18. Let V and W be irreducible varieties. If : V W is at, then
dim
1
(Q) =dimV dimW (26)
for all Q W. Conversely, if V and W are nonsingular and (26) holds for all Q W, then
is at.
PROOF. To be added.
2
Lines on surfaces
As an application of some of the above results, we consider the problem of describing the
set of lines on a surface of degree m in P
3
. To avoid possible problems, we assume for the
rest of this chapter that k has characteristic zero.
We rst need a way of describing lines in P
3
. Recall that we can associate with each
projective variety V P
n
an afne cone over
V in k
n1
. This allows us to think of points
in P
3
as being onedimensional subspaces in k
4
, and lines in P
3
as being twodimensional
subspaces in k
4
. To such a subspace W k
4
, we can attach a onedimensional subspace
_
2
W in
_
2
k
4
~k
6
, that is, to each line 1 in P
3
, we can attach point (1) in P
5
. Not
every point in P
5
should be of the form (1)heuristically, the lines in P
3
should form a
fourdimensional set. (Fix two planes in P
3
; giving a line in P
3
corresponds to choosing a
point on each of the planes.) We shall show that there is natural onetoone correspondence
between the set of lines in P
3
and the set of points on a certain hyperspace P
5
. Rather
than using exterior algebras, I shall usually give the oldfashioned proofs.
Let 1 be a line in P
3
and let x =(.
0
: .
1
: .
2
: .
3
) and y =(,
0
: ,
1
: ,
2
: ,
3
) be distinct
points on 1. Then
(1) =(
01
:
02
:
03
:
12
:
13
:
23
) P
5
.
i}
def
=
.
i
.
}
,
i
,
}
.
depends only on 1. The
i}
are called the Pl ucker coordinates of 1, after Pl ucker (1801
1868).
In terms of exterior algebras, write e
0
, e
1
, e
2
, e
3
for the canonical basis for k
4
, so that x,
regarded as a point of k
4
is
.
i
e
i
, and y =
,
i
e
i
; then
_
2
k
4
is a 6dimensional vector
Lines on surfaces 175
space with basis e
i
^e
}
, 0 _i < _3, and .^, =
i}
e
i
^e
}
with
i}
given by the above
formula.
We dene
i}
for all i. , 0 _i. _3 by the same formula thus
i}
=
}i
.
LEMMA 10.19. The line 1 can be recovered from (1) as follows:
1 ={(
}
a
}
0}
:
}
a
}
1}
:
}
a
}
2}
:
}
a
}
3}
) [ (a
0
: a
1
: a
2
: a
3
) P
3
].
PROOF. Let
1 be the cone over 1 in k
4
it is a twodimensional subspace of k
4
and let
x =(.
0
. .
1
. .
2
. .
3
) and y =(,
0
. ,
1
. ,
2
. ,
3
) be two linearly independent vectors in
1. Then
1 ={(y)x(x)y [ : k
4
k linear].
Write =
a
}
X
}
; then
(y)x(x)y =(
a
}
0}
.
a
}
1}
.
a
}
2}
.
a
}
3}
).
2
LEMMA 10.20. The point (1) lies on the quadric P
5
dened by the equation
X
01
X
23
X
02
X
13
X
03
X
12
=0.
PROOF. This can be veried by direct calculation, or by using that
0 =
.
0
.
1
.
2
.
3
,
0
,
1
,
2
,
3
.
0
.
1
.
2
.
3
,
0
,
1
,
2
,
3
=2(
01
23
02
13
03
12
)
(expansion in terms of 22 minors).
2
LEMMA 10.21. Every point of is of the form (1) for a unique line 1.
PROOF. Assume
03
=0; then the line through the points (0 :
01
:
02
:
03
) and (
03
:
13
:
23
: 0) has Pl ucker coordinates
(
01
03
:
02
03
:
2
03
:
01
23
02
13
]
03
]
12
:
03
13
:
03
23
)
=(
01
:
02
:
03
:
12
:
13
:
23
).
A similar construction works when one of the other coordinates is nonzero, and this way
we get inverse maps.
2
Thus we have a canonical onetoone correspondence
{lines in P
3
] {points on ]:
that is, we have identied the set of lines in P
3
with the points of an algebraic variety. We
may now use the methods of algebraic geometry to study the set. (This is a special case of
the Grassmannians discussed in 6.)
We next consider the set of homogeneous polynomials of degree m in 4 variables,
J(X
0
. X
1
. X
2
. X
3
) =
i
0
i
1
i
2
i
3
=n
a
i
0
i
1
i
2
i
3
X
i
0
0
. . . X
i
3
3
.
176 10. REGULAR MAPS AND THEIR FIBRES
LEMMA 10.22. The set of homogeneous polynomials of degree min 4 variables is a vector
space of dimension
_
3n
n
_
PROOF. See the footnote p126.
2
Let v =
_
3n
n
_
1 =
(n1)(n2)(n3)
6
1, and regard P
{surfaces of degree m in P
3
].
(. . . : a
i
0
i
1
i
2
i
3
: . . .) V(J). J =
a
i
0
i
1
i
2
i
3
X
i
0
0
X
i
1
1
X
i
2
2
X
i
3
3
.
The map is not quite injectivefor example, X
2
Y and XY
2
dene the same surface
but nevertheless, we can (somewhat loosely) think of the points of P
as being (possibly
degenerate) surfaces of degree m in P
3
.
Let 1
n
P
P
5
P
; it is therefore a
projective variety. The dimension of 1
n
is
n(n1)(n5)
6
3.
EXAMPLE 10.24. For m=1. 1
n
is the set of pairs consisting of a plane in P
3
and a line
on the plane. The theorem says that the dimension of 1
1
is 5. Since there are o
3
planes in
P
3
, and each has o
2
lines on it, this seems to be correct.
PROOF. We rst show that 1
n
is closed. Let
(1) =(
01
:
02
: . . .) J =
a
i
0
i
1
i
2
i
3
X
i
0
0
X
i
3
3
.
From (10.19) we see that 1 lies on the surface J(X
0
. X
1
. X
2
. X
3
) =0 if and only if
J(
b
}
0}
:
b
}
1}
:
b
}
2}
:
b
}
3}
) =0, all (b
0
. . . . . b
3
) k
4
.
Expand this out as a polynomial in the b
}
s with coefcients polynomials in the a
i
0
i
1
i
2
i
3
and
i}
s. Then J(...) =0 for all b k
4
if and only if the coefcients of the polynomial
are all zero. But each coefcient is of the form
1(. . . . a
i
0
i
1
i
2
i
3
. . . . :
01
.
02
: . . .)
with 1 homogeneous separately in the as and s, and so the set is closed in P
(cf.
the discussion in 7.9).
It remains to compute the dimension of 1
n
. We shall apply Proposition 10.11 to the
projection map
(1. J) 1
n
P
.
PROOF. We have
dim1
n
dimP
=
m(m1)(m5)
6
3
(m1)(m2)(m3)
6
1 =4(m1).
Therefore, if m > 3, then dim1
n
< dimP
, and so [(1
n
) is a proper closed subvariety of
P
1 = set of reguli;
P
9
S = set of nondegenerate quadrics.
The bres of the top map are connected, and of dimension 1 (they are all isomorphic to
P
1
), and the second map is nite and twotoone. Factorizations of this type occur quite
generally (see the Stein factorization theorem (10.30) below).
We now look at the case m=3. Here dim1
3
=19; v =19 : we have a map
[: 1
3
P
19
.
THEOREM 10.27. The set of cubic surfaces containing exactly 27 lines corresponds to an
open subset of P
19
; the remaining surfaces either contain an innite number of lines or a
nonzero nite number _27.
EXAMPLE 10.28. (a) Consider the Fermat surface
X
3
0
X
3
1
X
3
2
X
3
3
=0.
Let be a primitive cube root of one. There are the following lines on the surface, 0 _
i. _2:
_
X
0
i
X
1
=0
X
2
}
X
3
=0
_
X
0
i
X
2
=0
X
1
}
X
3
=0
_
X
0
i
X
3
=0
X
1
}
X
2
=0
There are three sets, each with nine lines, for a total of 27 lines.
Lines on surfaces 179
(b) Consider the surface
X
1
X
2
X
3
=X
3
0
.
In this case, there are exactly three lines. To see this, look rst in the afne space where
X
0
=0here we can take the equation to be X
1
X
2
X
3
=1. A line in A
3
can be written in
parametric form X
i
=a
i
t b
i
, but a direct inspection shows that no such line lies on the
surface. Now look where X
0
= 0, that is, in the plane at innity. The intersection of the
surface with this plane is given by X
1
X
2
X
3
=0 (homogeneous coordinates), which is the
union of three lines, namely,
X
1
=0; X
2
=0; X
3
=0.
Therefore, the surface contains exactly three lines.
(c) Consider the surface
X
3
1
X
3
2
=0.
Here there is a pencil of lines:
_
t
0
X
1
=t
1
X
0
t
0
X
2
=t
1
X
0
.
(In the afne space where X
0
= 0, the equation is X
3
Y
3
= 0, which contains the line
X =t , Y =t , all t.)
We now discuss the proof of Theorem 10.27). If [: 1
3
P
19
were not surjective, then
[(1
3
) would be a proper closed subvariety of P
19
, and the nonempty bres would all have
dimension _1 (by 10.9), which contradicts two of the above examples. Therefore the map
is surjective
1
, and there is an open subset U of P
19
where the bres have dimension 0;
outside U, the bres have dimension >0.
Given that every cubic surface has at least one line, it is not hard to show that there is
an open subset U
t
where the cubics have exactly 27 lines (see Reid, 1988, pp106110); in
fact, U
t
can be taken to be the set of nonsingular cubics. According to (8.23), the restriction
of [ to [
1
(U) is nite, and so we can apply (10.12) to see that all cubics in U U
t
have
fewer than 27 lines.
REMARK 10.29. The twentyseven lines on a cubic surface were discovered in 1849 by
Salmon and Cayley, and have been much studiedsee A. Henderson, The TwentySeven
Lines Upon the Cubic Surface, Cambridge University Press, 1911. For example, it is known
that the group of permutations of the set of 27 lines preserving intersections (that is, such
that 11
t
= 0 o(1) o(1
t
) = 0) is isomorphic to the Weyl group of the root
system of a simple Lie algebra of type 1
6
, and hence has 25920 elements.
It is known that there is a set of 6 skew lines on a nonsingular cubic surface V . Let 1
and 1
t
be two skew lines. Then in general a line joining a point on 1 to a point on 1
t
will
meet the surface in exactly one further point. In this way one obtains an invertible regular
map from an open subset of P
1
P
1
to an open subset of V , and hence V is birationally
equivalent to P
2
.
1
According to Miles Reid (1988, p126) every adult algebraic geometer knows the proof that every cubic
contains a line.
180 10. REGULAR MAPS AND THEIR FIBRES
Stein factorization
The following important theorem shows that the bres of a proper map are disconnected
only because the bres of nite maps are disconnected.
THEOREM 10.30. Let : W V be a proper morphism of varieties. It is possible to factor
into W
1
W
t
2
V with
1
proper with connected bres and
2
nite.
PROOF. This is usually proved at the same time as Zariskis main theorem (if W and V are
irreducible, and V is afne, then W
t
is the afne variety with kW
t
 the integral closure of
kV  in k(W)).
2
Exercises
101. Let G be a connected algebraic group, and consider an action of G on a variety V ,
i.e., a regular map GV V such that (gg
t
) =g(g
t
) for all g. g
t
G and V . Show
that each orbit O =G of G is nonsingular and open in its closure
O, and that
OO is a
union of orbits of strictly lower dimension. Deduce that there is at least one closed orbit.
102. Let G =GL
2
=V , and let G act on V by conjugation. According to the theory of
Jordan canonical forms, the orbits are of three types:
(a) Characteristic polynomial X
2
aX b; distinct roots.
(b) Characteristic polynomial X
2
aX b; minimal polynomial the same; repeated
roots.
(c) Characteristic polynomial X
2
aX b =(X )
2
; minimal polynomial X .
For each type, nd the dimension of the orbit, the equations dening it (as a subvariety of
V ), the closure of the orbit, and which other orbits are contained in the closure.
(You may assume, if you wish, that the characteristic is zero. Also, you may assume the
following (fairly difcult) result: for any closed subgroup H of an algebraic group G, G,H
has a natural structure of an algebraic variety with the following properties: G G,H is
regular, and a map G,H V is regular if the composite G G,H V is regular;
dimG,H =dimGdimH.)
[The enthusiasts may wish to carry out the analysis for GL
n
.]
103. Find 3J
2
lines on the Fermat projective surface
X
d
0
X
d
1
X
d
2
X
d
3
=0. J _3. (. J) =1. the characteristic.
104. (a) Let : W V be a quasinite dominant regular map of irreducible varieties.
Showthat there are open subsets U
t
and U of W and V such that (U
t
) U and : U
t
U
is nite.
(b) Let G be an algebraic group acting transitively on irreducible varieties W and V ,
and let : W V be Gequivariant regular map satisfying the hypotheses in (a). Then is
nite, and hence proper.
CHAPTER 11
Algebraic spaces; geometry over an
arbitrary eld
In this chapter, we explain how to extend the theory of the preceding chapters to a nonalge
braically closed base eld. One major difference is that we need to consider ringed spaces
in which the sheaf of rings is no longer a sheaf of functions on the base space. Once we
allow that degree of extra generality, it is natural to allow the rings to have nilpotents. In
this way we obtain the notion of an algebraic space, which even over an algebraically closed
eld is more general than that of an algebraic variety.
Throughout this chapter, k is a eld and k
al
is an algebraic closure of k.
Preliminaries
Sheaves
A presheaf F on a topological space V is a map assigning to each open subset U of V a
set F(U) and to each inclusion U
t
U a restriction map
a a[U
t
: F(U) F(U
t
):
when U =U
t
the restriction map is required to be the identity map, and if
U
tt
U
t
U.
then the composite of the restriction maps
F(U) F(U
t
) F(U
tt
)
is required to be the restriction map F(U) F(U
tt
). In other words, a presheaf is a
contravariant functor to the category of sets from the category whose objects are the open
subsets of V and whose morphisms are the inclusions. A homomorphism of presheaves
: F F
t
is a family of maps
(U): F(U) F
t
(U)
commuting with the restriction maps, i.e., a morphism of functors.
181
182 11. ALGEBRAIC SPACES; GEOMETRY OVER AN ARBITRARY FIELD
A presheaf F is a sheaf if for every open covering {U
i
] of an open subset U of V
and family of elements a
i
F(U
i
) agreeing on overlaps (that is, such that a
i
[U
i
U
}
=
a
}
[U
i
U
}
for all i. ), there is a unique element a F(U) such that a
i
=a[U
i
for all i .
A homomorphism of sheaves on V is a homomorphism of presheaves.
If the sets F(U) are abelian groups and the restriction maps are homomorphisms, then
the sheaf is a sheaf of abelian groups. Similarly one denes a sheaf of rings, a sheaf of
kalgebras, and a sheaf of modules over a sheaf of rings.
For V , the stalk of a sheaf F (or presheaf) at is
F
=lim
: O
t
()
for all V . A morphism of locally ringed spaces is a morphism of ringed space such
that [
k
k
al
being reduced. Consider for example the algebra = kX. Y ,(X
]
Y
]
a)
where = char(k) and a is not a thpower in k. Then is reduced (even an integral
domain) because X
]
Y
]
a is irreducible in kX. Y , but
k
k
al
.k
al
X. Y ,(X
]
Y
]
a)
=k
al
X. Y ,((X Y )
]
).
]
=a.
which is not reduced because it contains the nonzero element . , whose th power
is zero.
In this subsection, we show that problems of this kind arise only because of insepara
bility. In particular, they dont occur if k is perfect.
Now assume k has characteristic =0, and let be some (large) eld containing k
al
.
Let
k
1
p
={ k
al
[
]
k].
It is a subeld of k
al
, and k
1
p
=k if and only if k is perfect.
Preliminaries 183
DEFINITION 11.1. Subelds 1. 1
t
of containing k are said to be linearly disjoint over
k if the map 1
k
1
t
is injective.
Equivalent conditions:
if e
1
. . . . . e
n
1 are linearly independent over k and e
t
1
. . . . . e
t
n
0
1
t
are linearly
independent over k, then the elements e
1
e
t
1
. e
1
e
t
2
. . . . . e
n
e
t
n
0
of are linearly inde
pendent over k;
if e
1
. . . . . e
n
1 are linearly independent over k, then they are also linearly indepen
dent over 1
t
.
(*) The following statements are easy to prove.
(a) Every purely transcendental extension of k is linearly disjoint from every algebraic
extension of k.
(b) Every separable algebraic extension of k is linearly disjoint from every purely insep
arable algebraic extension of k.
(c) Let 1 k and 1 1 k be subelds of . Then 1 and 1 are linearly disjoint
over k if and only if 1 and 1 are linearly disjoint over k and 11 and 1 are linearly
disjoint over k.
LEMMA 11.2. Let 1 = k(.
1
. . . . . .
d1
) with .
1
. . . . . .
d
algebraically independent
over J, and let kX
1
. . . . . X
d1
 be an irreducible polynomial such that (.
1
. . . . . .
d1
) =
0. If 1 is linearly disjoint from k
1
p
, then kX
]
1
. . . . . X
]
d1
.
PROOF. Suppose otherwise, say, =g(X
]
1
. . . . . X
]
d1
). Let M
1
. . . . . M
i
be the monomi
als in X
1
. . . . . X
d1
that actually occur in g(X
1
. . . . . X
d1
), and let m
i
=M
i
(.
1
. . . . . .
d1
).
Then m
1
. . . . . m
i
are linearly independent over k (because each has degree less than that of
). However, m
]
1
. . . . . m
]
i
are linearly dependent over k, because g(.
]
1
. . . . . .
]
d1
) = 0.
But
a
i
m
]
i
=0 (a
i
k) ==
a
1
p
i
m
i
=0 (a
1
p
i
k
1
p
)
and we have a contradiction.
2
A separating transcendence basis for 1 k is a transcendence basis {.
1
. . . . . .
d
] such
that 1 is separable over k(.
1
. . . . . .
d
). The next proposition improves Theorem 8.21 of FT.
PROPOSITION 11.3. Let 1 be a nitely generated eld extension of k, and let be an
algebraically closed eld containing 1
al
. The following statements are equivalent:
(a) 1,k admits a separating transcendence basis;
(b) for any purely inseparable extension 1 of k in 1, the elds 1 and 1 are linearly
disjoint over k;
(c) the elds 1 and k
1
p
are linearly disjoint over k.
PROOF. (a)=(b). This follows easily from (*).
(b)=(c). Trivial.
(c)=(a). Let 1 = k(.
1
. . . . . .
n
), and let J be the transcendence degree of 1,k. Af
ter renumbering, we may suppose that .
1
. . . . . .
d
are algebraically independent (FT 8.12).
184 11. ALGEBRAIC SPACES; GEOMETRY OVER AN ARBITRARY FIELD
We proceed by induction on n. If n = J there is nothing to prove, and so we may as
sume that n _ J 1. Then (.
1
. . . . . .
d1
) =0 for some nonzero irreducible polynomial
(X
1
. . . . . X
d1
) with coefcients in k. Not all d,dX
i
are zero, for otherwise would
be a polynomial in X
]
1
. . . . . X
]
d1
, which contradicts the lemma. After renumbering again,
we may suppose that d,dX
d1
= 0, and so {.
1
. . . . . .
d
] is a separating transcendence
basis for k(.
1
. . . . . .
d1
) over k, which proves the proposition when n = J 1. In the
general case, k(.
1
. . . . . .
d1
. .
d2
) is algebraic over k(.
1
. . . . . .
d
) and .
d1
is separable
over k(.
1
. . . . . .
d
), and so, by the primitive element theorem (FT 5.1) there is an element
, such that k(.
1
. . . . . .
d2
) =k(.
1
. . . . . .
d
. ,). Thus 1 is generated by the n1 elements
.
1
. . . . .
d
. ,. .
d3
. . . . . .
n
, and we apply induction.
2
A nitely generated eld extension 1 k is said to be regular if it satises the equiv
alent conditions of the proposition.
PROPOSITION 11.4. Let be a reduced nitely generated kalgebra. The following state
ments are equivalent:
(a) k
1
p
k
is reduced;
(b) k
al
k
is reduced;
(c) 1
k
is reduced for all elds 1 k.
When is an integral domain, they are also equivalent to and k
1
p
being linearly disjoint
over k.
PROOF. The implications c==b==a are obvious, and so we only have to prove a ==c.
After localizing at a minimal prime, we may suppose that it is a eld. Let e
1
. . . . . e
n
be
elements of linearly independent over k. If they become linearly dependent over k
1
p
,
then e
]
1
. . . . . e
]
n
are linearly dependent over k, say,
a
i
e
]
i
=0, a
i
k. Now
a
1
p
i
e
i
is
a nonzero element of k
1
p
k
, but
_
a
1
p
i
e
i
_
]
=
a
i
e
]
i
=
1a
i
e
]
i
=1
a
i
e
]
i
=0.
This shows that and k
1
p
are linearly disjoint over k, and so has a separating transcen
dence basis over k. From this it follows that 1
k
is reduced for all elds 1 k.
2
Idempotents
Even when is an integral domain and
k
k
al
is reduced, the latter need not be an
integral domain. Suppose, for example, that is a nite separable eld extension of k.
Then ~kX,((X)) for some irreducible separable polynomial (X), and so
k
k
al
~k
al
X,((X)) =k
al
,(
(X a
i
)) .
k
al
,(X a
i
)
(by the Chinese remainder theorem). This shows that if contains a nite separable eld
extension of k, then
k
k
al
cant be an integral domain. The next proposition provides a
converse.
Afne algebraic spaces 185
PROPOSITION 11.5. Let be a nitely generated kalgebra, and assume that is an inte
gral domain, and that
k
k
al
is reduced. Then
k
k
al
is an integral domain if and only
if k is algebraically closed in (i.e., if a is algebraic over k, then a k).
LEMMA 11.6. Let k be algebraically closed in an extension eld 1, and let a be an element
of 1
al
that is algebraic over k. Then 1 and ka are linearly disjoint over k, and 1a :
1 =ka : k.
PROOF. Let (X) be the minimum polynomial of a over k. Then (X) is irreducible over
1, because the coefcients of any factor of (X) in 1X are algebraic over k, and hence
lie in k. It follows that the map
1
k
ka 1a
is an isomorphism, because both rings are isomorphic to 1X,((X)).
2
PROOF (OF THE PROPOSITION) Let 1 be the eld of fractions of it sufces to show
that 1 is linearly disjoint from 1 where 1 is any nite algebraic extension of k in 1
al
(because then 1
k
1 .11, which is an integral domain). If 1 is separable over k, then
it can be generated by a single element, and so this follows from the lemma. In the general
case, we let 1 be the largest subeld of 1 separable over k. From (*)(c), we see that it
sufces to show that 11 and 1 are linearly disjoint over 1. From (11.4), we see that 1
and k
1{]
are linearly disjoint over k, and so 1 is a regular extension of k (see 11.3). It
follows easily that 11 is a regular extension of 1, and 11 is linearly disjoint from 1 by
(*)(b).
2
After these preliminaries, it is possible rewrite all of the preceding sections with k not
necessarily algebraically closed. I indicate briey how this is done.
Afne algebraic spaces
For a nitely generated kalgebra , we dene spm() to be the set of maximal ideals in
endowed with the topology having as basis the sets D( ), D( ) ={m [ m]. There
is a unique sheaf of kalgebras O on spm() such that 1(D( ). O)) =
(
for all
(recall that
(
is the ring obtained from by inverting ), and we denote the resulting
ringed space by Spm(). The stalk at m V is lim
(
(
.
m
.
Let m be a maximal ideal of . Then k(m)
def
=,m is eld that is nitely generated as
a kalgebra, and is therefore of nite degree over k (Zariskis lemma, 2.7).
The sections of O are no longer functions on V =spm. For m spm() and
we set (m) equal to the image of in k(m). It does make sense to speak of the zero set
of in V , and D( ) ={m[ (m) =0]. For . g ,
(m) =g(m) for all m g is nilpotent.
When k is algebraically closed and is an afne kalgebra, k(m) .k and we recover the
denition of Spm in 3.
186 11. ALGEBRAIC SPACES; GEOMETRY OVER AN ARBITRARY FIELD
An afne algebraic space
1
over k is a ringed space (V. O
V
) such that
1(V. O
V
) is a nitely generated kalgebra,
for each 1 V , 1(1)
def
={ 1(V. O
V
) [ (1) =0] is a maximal ideal in 1(V. O
V
),
and
the map 1 1(1): V Spm(1(V. O
V
)) is an isomorphism of ringed spaces.
For an afne algebraic space, we sometimes denote 1(V. O
V
) by kV . A morphism of
algebraic spaces over k is a morphism of ringed spaces it is automatically a morphism of
locally ringed spaces. An afne algebraic space (V. O
V
) is reduced if 1(V. O
V
) is reduced.
Let : T be a homomorphism of nitely generated kalgebras. For any maximal
ideal m of T, there is an injection of kalgebras ,
1
(m) T,m. As T,m is a eld of
nite degree over k, this shows that
1
(m) is a maximal ideal of . Therefore denes
a map spmT spm, which one shows easily denes a morphism of afne algebraic
kspaces
SpmT Spm .
and this gives a bijection
Hom
kalg
(. T) .Hom
k
(SpmT. Spm).
Therefore Spm() is an equivalence of from the category of nitely generated k
algebras to that of afne algebraic spaces over k; its quasiinverse is V kV 
def
=1(V. O
V
).
Under this correspondence, reduced algebraic spaces correspond to reduced algebras.
Let V be an afne algebraic space over k. For an ideal a in kV ,
spm(,a) .V(a)
def
={1 V [ (1) =0 for all a].
We call V(a) endowed with the ring structure provided by this isomorphism a closed alge
braic subspace of V . Thus, there is a onetoone correspondence between the closed alge
braic subspaces of V and the ideals in kV . Note that if rad(a) =rad(b), then V(a) =V(b)
as topological spaces (but not as algebraic spaces).
Let : Spm(T) Spm() be the map dened by a homomorphism : T.
The image of is dense if and only if the kernel of is nilpotent.
The map denes an isomorphism of Spm(T) with a closed subvariety of Spm()
if and only if is surjective.
Afne algebraic varieties.
An afne kalgebra is a nitely generated kalgebra such that
k
k
al
is reduced. Since
k
k
al
, itself is then reduced. Proposition 11.4 has the following consequences:
if is an afne kalgebra, then
k
1 is reduced for all elds 1 containing k; when k is
perfect, every reduced nitely generated kalgebra is afne.
Let be a nitely generated kalgebra. The choice of a set {.
1
. .... .
n
] of generators
for , determines isomorphisms
k.
1
. .... .
n
 kX
1
. .... X
n
,(
1
. ....
n
).
1
Not to be confused with the algebraic spaces of, for example, of JP. Serre, Espaces Fibr es Alg ebriques,
1958, which are simply algebraic varieties in the sense of these notes, or with the algebraic spaces of M. Artin,
Algebraic Spaces, 1969, which generalize (!) schemes.
Algebraic spaces; algebraic varieties. 187
and
k
k
al
k
al
X
1
. .... X
n
,(
1
. ....
n
).
Thus is an afne algebra if the elements
1
. ....
n
of kX
1
. .... X
n
 generate a radical
ideal when regarded as elements of k
al
X
1
. .... X
n
. From the above remarks, we see that
this condition implies that they generate a radical ideal in kX
1
. .... X
n
, and the converse
implication holds when k is perfect.
An afne algebraic space (V. O
V
) such that 1(V. O
V
) is an afne kalgebra is called an
afne algebraic variety over k. Thus, a ringed space (V. O
V
) is an afne algebraic variety
if 1(V. O
V
) is an afne kalgebra, 1(1) is a maximal ideal in 1(V. O
V
) for each 1 V ,
and 1 1(1): V spm(1(V. O
V
)) is an isomorphism of ringed spaces.
Let
=kX
1
. .... X
n
,a.
T =kY
1
. .... Y
n
,b.
A homomorphism T is determined by a family of polynomials, 1
i
(Y
1
. .... Y
n
), i =
1. .... m; the homomorphism sends .
i
to 1
i
(,
1
. .... ,
n
); in order to dene a homomorphism,
the 1
i
must be such that
J a ==J(1
1
. .... 1
n
) b;
two families 1
1
. .... 1
n
and Q
1
. .... Q
n
determine the same map if and only if 1
i
Q
i
mod b for all i .
Let be a nitely generated kalgebra, and let V = Spm. For any eld 1 k,
1
k
is a nitely generated 1algebra, and hence we get a variety V
1
def
= Spm(1
k
) over 1. We say that V
1
has been obtained from V by extension of scalars or ex
tension of the base eld. Note that if = kX
1
. .... X
n
,(
1
. ....
n
) then
k
1 =
1X
1
. .... X
n
,(
1
. ....
n
). The map V V
1
is a functor from afne varieties over k
to afne varieties over 1.
Let V
0
=Spm(
0
) be an afne variety over k. and let W =V(b) be a closed subvariety
of V
def
= V
0,k
al . Then W arises by extension of scalars from a closed subvariety W
0
of V
0
if and only if the ideal b of
0
k
k
al
is generated by elements of
0
. Except when k is
perfect, this is stronger than saying W is the zero set of a family of elements of .
The denition of the afne space A(1) attached to a vector space 1 works over any
eld.
Algebraic spaces; algebraic varieties.
An algebraic space over k is a ringed space (V. O) for which there exists a nite covering
(U
i
) of V by open subsets such that (U
i
. O[U
i
) is an afne algebraic space over k for all
i . A morphism of algebraic spaces (also called a regular map) over k is a morphism
of locally ringed spaces of kalgebras. An algebraic space is separated if for all pairs of
morphisms of kspaces . : 7 V , the subset of 7 on which and agree is closed.
Similarly, an algebraic prevariety over k is a ringed space (V. O) for which there exists
a nite covering (U
i
) of V by open subsets such that (U
i
. O[U
i
) is an afne algebraic
variety over k for all i . A separated prevariety is called a variety.
With any algebraic space V over k we can associate a reduced algebraic space V
red
such
that
188 11. ALGEBRAIC SPACES; GEOMETRY OVER AN ARBITRARY FIELD
V
red
=V as a topological space,
for all open afnes U V , 1(U. O
V
red
) is the quotient of 1(U. O
V
) by its nilradical.
For example, if V = SpmkX
1
. . . . . X
n
,a, then V
red
= SpmkX
1
. . . . . X
n
,rad(a). The
identity map V
red
V is a regular map. Any closed subset of V can be given a unique
structure of a reduced algebraic space.
Products.
If and T are nitely generated kalgebras, then
k
T is a nitely generated kalgebra,
and Spm(
k
T) is the product of Spm() and Spm(T) in the category of algebraic k
spaces, i.e., it has the correct universal property. This denition of product extends in a
natural way to all algebraic spaces.
The tensor product of two reduced kalgebras may fail to be reduced consider for
example,
=kX. Y ,(X
]
Y
]
a). T =k7,(7
]
a). a k
]
.
However, if and T are afne kalgebras, then
k
T is again an afne kalgebra. To see
this, note that (by denition),
k
k
al
and T
k
k
al
are afne kalgebras, and therefore so
also is their tensor product over k
al
(4.15); but
(
k
k
al
)
k
al (k
al
k
T) .((
k
k
al
)
k
al k
al
)
k
T .(
k
T)
k
k
al
.
Thus, if V and W are algebraic (pre)varieties over k, then so also is their product.
Just as in (4.24, 4.25), the diagonal z is locally closed in V V , and it is closed if and
only if V is separated.
Extension of scalars (extension of the base eld).
Let V be an algebraic space over k, and let 1 be a eld containing k. There is a natural
way of dening an algebraic space V
1
over 1, said to be obtained from V by extension
of scalars (or extension of the base eld): if V is a union of open afnes, V =
_
U
i
, then
V
1
=
_
U
i,1
and the U
i,1
are patched together the same way as the U
i
. If 1 is algebraic
over k, there is a morphism (V
1
. O
V
K
) (V. O
V
) that is universal: for any algebraic 1
space W and morphism (W. O
W
) (V. O
V
), there exists a unique regular map W V
1
giving a commutative diagram,
W V
1
1
V k.
3
The dimension of an algebraic space or variety doesnt change under extension of
scalars.
When V is an algebraic space (or variety) over k
al
obtained from an algebraic space
(or variety) V
0
over k by extension of scalars, we sometimes call V
0
a model for V over k.
More precisely, a model of V over k is an algebraic space (or variety) V
0
over k together
with an isomorphism : V V
0,k
al .
Algebraic spaces; algebraic varieties. 189
Of course, V need not have a model over k for example, an elliptic curve
1 : Y
2
7 =X
3
aX7
2
b7
3
over k
al
will have a model over k k
al
if and only if its invariant (1)
def
=
1T2S(4o)
3
16(4o
3
2Tb
2
)
lies in k. Moreover, when V has a model over k, it will usually have a large number of
them, no two of which are isomorphic over k. Consider, for example, the quadric surface
in P
3
over
al
.
V : X
2
Y
2
7
2
W
2
=0.
The models over V over are dened by equations
aX
2
bY
2
c7
2
JW
2
=0, a, b, c, J .
Classifying the models of V over is equivalent to classifying quadratic forms over in
4 variables. This has been done, but it requires serious number theory. In particular, there
are innitely many (see Chapter VIII of my notes on Class Field Theory).
Let V be an algebraic space over k. When k is perfect, V
red
is an algebraic prevariety
over k, but not necessarily otherwise, i.e., (V
red
)
k
al need not be reduced. This shows that
when k is not perfect, passage to the associated reduced algebraic space does not commute
with extension of the base eld: we may have
(V
red
)
1
=(V
1
)
red
.
PROPOSITION 11.7. Let V be an algebraic space over a eld k. Then V is an algebraic
prevariety if and only if V
k
1
p
is reduced, in which case V
1
is reduced for all elds 1 k.
PROOF. Apply 11.4.
2
Connectedness
A variety V over a eld k is said to be geometrically connected if V
k
al is connected, in
which case, V
,m.
Each ,mis a nite eld extension of k, and it is separable because otherwise (,m)
k
k
al
would not be reduced. This proves (a).
The set S in (b) is V(k
sep
) with the natural action of . We can recover (V. O
V
) from
S as follows: let V be the set \S of orbits endowed with the discrete topology, and, for
e =s \S, let k(e) =(k
sep
)
s
where
x
is the stabilizer of s in ; then, for U V ,
1(U. O
V
) =
eU
k(e).
2
PROPOSITION 11.10. Given a variety V over k, there exists a map : V from V to a
zerodimensional variety such that, for all e , the bre V
e
is a geometrically connected
variety over k(e).
PROOF. Let be the zerodimensional variety whose underlying set is the set of connected
components of V over and such that, for each e =V
i
, k(e) is the algebraic closure of
k in (V
i
). Apply (11.8) to see that the obvious map : V has the desired property.
2
EXAMPLE 11.11. Let V be a connected variety over a k, and let k
t
be the algebraic closure
of k in k(V ). The map : V Spmk realizes V as a geometrically connected variety over
k. Conversely, for a geometrically connected variety : V Spmk
t
over a nite extension
of k, the composite of with Spmk
t
Spmk realizes V as a variety over k (connected,
but not geometrically connected if k
t
=k).
EXAMPLE 11.12. Let : V be as in (11.10). When we regard as a set with an
action of , then its points are in natural onetoone correspondence with the connected
components of V
k
sep and its orbits are in natural onetoone correspondence with the
connected components of V . Let e and let V
t
=
1
k
sep
(e) it is a connected component
of V
k
sep . Let
e
be the stabilizer of e; then V
t
arises from a geometrically connected variety
over k(e)
def
=(k
sep
)
e
.
ASIDE 11.13. Proposition 11.10 is a special case of Stein factorization (10.30).
3
This means that the action factors through the quotient of Gal(
al
,) by an open subgroup (all open
subgroups of Gal(
al
,) are of nite index, but not all subgroups of nite index are open).
4
The cognoscente will recognize this as Grothendiecks way of expressing Galois theory over .
Algebraic spaces; algebraic varieties. 191
Fibred products
Fibred products exist in the category of algebraic spaces. For example, if 1 and
1 T are homomorphisms of nitely generated kalgebras, then
T
T is a nitely
generated kalgebras and
Spm()
Spm(T)
Spm(T) =Spm(
T
T).
For algebraic prevarieties, the situation is less satisfactory. Consider a variety S and two
regular maps V S and W S. Then (V
S
W)
red
is the bred product of V and W over
S in the category of reduced algebraic kspaces. When k is perfect, this is a variety, but
not necessarily otherwise. Even when the bred product exists in the category of algebraic
prevarieties, it is anomolous. The correct object is the bred product in the category of
algebraic spaces which, as we have observed, may no longer be an algebraic variety. This
is one reason for introducing algebraic spaces.
Consider the bred product:
A
1
A
1
A
1 {a]
xx
p
_
A
1
{a]
In the category of algebraic varieties, A
1
A
1 {a] is a single point if a is a th power in k and
is empty otherwise; in the category of algebraic spaces, A
1
A
1 {a] =SpmkT ,(T
]
a),
which can be thought of as a fold point (point with multiplicity ).
The points on an algebraic space
Let V be an algebraic space over k. A point of V with coordinates in k (or a point of V
rational over k, or a kpoint of V ) is a morphism Spmk V . For example, if V is afne,
say V =Spm(), then a point of V with coordinates in k is a khomomorphism k. If
= kX
1
. .... X
n
,(
1
. ....
n
), then to give a khomomorphism k is the same as to
give an ntuple (a
1
. .... a
n
) such that
i
(a
1
. .... a
n
) =0, i =1. .... m.
In other words, if V is the afne algebraic space over k dened by the equations
i
(X
1
. . . . . X
n
) =0. i =1. . . . . m
then a point of V with coordinates in k is a solution to this system of equations in k. We
write V(k) for the points of V with coordinates in k.
We extend this notion to obtain the set of points V(1) of a variety V with coordinates
in any kalgebra 1. For example, when V =Spm(), we set
V(1) =Hom
kalg
(. 1).
Again, if
=kX
1
. .... X
n
,(
1
. ....
n
).
then
V(1) ={(a
1
. .... a
n
) 1
n
[
i
(a
1
. .... a
n
) =0. i =1. 2. .... m].
192 11. ALGEBRAIC SPACES; GEOMETRY OVER AN ARBITRARY FIELD
What is the relation between the elements of V and the elements of V(k)? Suppose V
is afne, say V = Spm(). Let V . Then corresponds to a maximal ideal m
in
(actually, it is a maximal ideal), and we write k() for the residue eld O
,m
. Then k()
is a nite extension of k, and we call the degree of k() over k the degree of . Let 1 be a
eld algebraic over k. To give a point of V with coordinates in 1 is to give a homomorphism
of kalgebras 1. The kernel of such a homomorphism is a maximal ideal m
in ,
and the homomorphisms k with kernel m
m
i
,m
i1
is a polynomial ring in dim1
symbols. Using this, one see that if ., =0 in 1, then one of . or , lies in
_
n
m
n
, which is zero by the Krull
intersection theorem (1.8).
Projective varieties. 193
The converse to (a) of the theorem fails. For example, let k be a eld of characteristic
= 0. 2, and let a be a nonzero element of k that is not a th power. Then (X. Y ) =
Y
2
X
]
a is irreducible, and remains irreducible over k
al
. Therefore,
=kX. Y ,((X. Y )) =k.. ,
is an afne kalgebra, and we let V be the curve Spm. One checks that V is normal, and
hence is regular by Atiyah and MacDonald 1969, 9.2. However,
d
dX
=0.
d
dY
=2Y.
and so (a
1
p
. 0) V
sing
(k
al
): the point 1 in V corresponding to the maximal ideal (,) of
is singular even though O
1
is regular.
The relation between nonsingular and regular is examined in detail in: Zariski,
O., The Concept of a Simple Point of an Abstract Algebraic Variety, Transactions of the
American Mathematical Society, Vol. 62, No. 1. (Jul., 1947), pp. 152.
Separable points
Let V be an algebraic variety over k. Call a point 1 V separable if k(1) is a separable
extension of k.
PROPOSITION 11.15. The separable points are dense in V ; in particular, V(k) is dense in
V if k is separably closed.
PROOF. It sufces to prove this for each irreducible component of V , and we may re
place an irreducible component of V by any variety birationally equivalent with it (4.32).
Therefore, it sufces to prove it for a hypersurface H in A
d1
dened by a polynomial
(X
1
. . . . . X
d1
) that is separable when regarded as a polynomial in X
d1
with coef
cients in k(X
1
. . . . . X
d
) (4.34, 11.3). Then
J(
JA
dC1
=0 (as a polynomial in X
1
. . . . . X
d
), and
on the nonempty open subset D(
J(
JA
dC1
) of A
d
, (a
1
. . . . . a
d
. X
d1
) will be a separable
polynomial. The points of H lying over points of U are separable.
2
Tangent cones
DEFINITION 11.16. The tangent cone at a point 1 on an algebraic space V is Spm(gr(O
1
)).
When V is a variety over an algebraically closed eld, this agrees with the denition in
Chapter 5, except that there we didnt have the correct language to describe it even in
that case, the tangent cone may be an algebraic space (not an algebraic variety).
Projective varieties.
Everything in this chapter holds, essentially unchanged, when k is allowed to be an arbitrary
eld.
If V
k
al is a projective variety, then so also is V . The idea of the proof is the following:
to say that V is projective means that it has an ample divisor; but a divisor D on V is ample
if D
k
al is ample on V
k
al ; by assumption, there is a divisor D on V
k
al that is ample; any
multiple of the sum of the Galois conjugates of D will also be ample, but some such divisor
will arise from a divisor on V .
194 11. ALGEBRAIC SPACES; GEOMETRY OVER AN ARBITRARY FIELD
Complete varieties.
Everything in this chapter holds unchanged when k is allowed to be an arbitrary eld.
Normal varieties; Finite maps.
As noted in (8.15), the Noether normalization theorem requires a different proof when the
eld is nite. Also, as noted earlier in this chapter, one needs to be careful with the denition
of bre. For example, one should dene a regular map : V W to be quasinite if the
bres of the map of sets V(k
al
) W(k
al
) are nite.
Otherwise, k can be allowed to be arbitrary.
Dimension theory
The dimension of a variety V over an arbitrary eld k can be dened as in the case that k
is algebraically closed. The dimension of V is unchanged by extension of the base eld.
Most of the results of this chapter hold for arbitrary base elds.
Regular maps and their bres
Again, the results of this chapter hold for arbitrary elds provided one is careful with the
notion of a bre.
Algebraic groups
We now dene an algebraic group to be an algebraic space G together with regular maps
mult: GG. inverse: G G. e: A
0
G
making G(1) into a group in the usual sense for all kalgebras 1.
THEOREM 11.17. Let G be an algebraic group over k.
(a) If G is connected, then it is geometrically connected.
(b) If G is geometrically reduced (i.e., a variety), then it is nonsingular.
(c) If k is perfect and G is reduced, then it is geometrically reduced.
(d) If k has characteristic zero, then G is geometrically reduced (hence nonsingular).
PROOF. (a) The existence of e shows that k is algebraically closed in k(G). Therefore (a)
follows from (11.8).
(b) It sufces to show that G
k
al is nonsingular, but this we did in (5.20).
(c) As k =k
1
p
, this follows from (11.7).
(d) See my notes, Algebraic Groups..., I, Theorem 6.31 (Cartiers theorem).
2
Exercises
111. Show directly that, up to isomorphism, the curve X
2
Y
2
=1 over C has exactly
two models over 1.
CHAPTER 12
Divisors and Intersection Theory
In this chapter, k is an arbitrary eld.
Divisors
Recall that a normal ring is an integral domain that is integrally closed in its eld of frac
tions, and that a variety V is normal if O
n
i
7
i
. n
i
Z. 7
i
a prime divisor on V.
The support [D[ of D is the union of the 7
i
corresponding to nonzero n
i
s. A divisor is
said to be effective (or positive) if n
i
_0 for all i . We get a partial ordering on the divisors
by dening D _D
t
to mean DD
t
_0.
Because V is normal, there is associated with every prime divisor 7 on V a discrete
valuation ring O
7
. This can be dened, for example, by choosing an open afne subvariety
U of V such that U 7 =0; then U 7 is a maximal proper closed subset of U, and so
the ideal p corresponding to it is minimal among the nonzero ideals of 1 = 1(U. O): so
1
p
is a normal ring with exactly one nonzero prime ideal p1 it is therefore a discrete
valuation ring (Atiyah and MacDonald 9.2), which is dened to be O
7
. More intrinsically
we can dene O
7
to be the set of rational functions on V that are dened an open subset U
of V with U 7 =0.
Let ord
7
be the valuation of k(V )
ord
7
( ) 7.
195
196 12. DIVISORS AND INTERSECTION THEORY
The sum is over all the prime divisors of V , but in fact ord
7
( ) = 0 for all but nitely
many 7s. In proving this, we can assume that V is afne (because it is a nite union of
afnes), say V = Spm(1). Then k(V ) is the eld of fractions of 1, and so we can write
=g,h with g. h 1, and div( ) =div(g) div(h). Therefore, we can assume 1.
The zero set of , V( ) either is empty or is a nite union of prime divisors, V =
_
7
i
(see 9.2) and ord
7
( ) =0 unless 7 is one of the 7
i
.
The map
div( ): k(V )
Div(V )
is a homomorphism. A divisor of the form div( ) is said to be principal, and two divisors
are said to be linearly equivalent, denoted D ~D
t
, if they differ by a principal divisor.
When V is nonsingular, the Picard group Pic(V ) of V is dened to be the group of
divisors on V modulo principal divisors. (Later, we shall dene Pic(V ) for an arbitrary va
riety; when V is singular it will differ from the group of divisors modulo principal divisors,
even when V is normal.)
EXAMPLE 12.2. Let C be a nonsingular afne curve corresponding to the afne kalgebra
1. Because C is nonsingular, 1 is a Dedekind domain. A prime divisor on C can be
identied with a nonzero prime divisor in 1, a divisor on C with a fractional ideal, and
1i c(C) with the ideal class group of 1.
Let U be an open subset of V , and let 7 be a prime divisor of V . Then 7U is either
empty or is a prime divisor of U. We dene the restriction of a divisor D =
n
7
7 on V
to U to be
D[
U
=
7Uy0
n
7
7U.
When V is nonsingular, every divisor D is locally principal, i.e., every point 1 has an
open neighbourhood U such that the restriction of D to U is principal. It sufces to prove
this for a prime divisor 7. If 1 is not in the support of D, we can take =1. The prime
divisors passing through 1 are in onetoone correspondence with the prime ideals p of
height 1 in O
1
, i.e., the minimal nonzero prime ideals. Our assumption implies that O
1
is
a regular local ring. It is a (fairly hard) theorem in commutative algebra that a regular local
ring is a unique factorization domain. It is a (fairly easy) theorem that a noetherian integral
domain is a unique factorization domain if every prime ideal of height 1 is principal (Nagata
1962, 13.1). Thus p is principal in O
p
. and this implies that it is principal in 1(U. O
V
) for
some open afne set U containing 1 (see also 9.13).
If D[
U
=div( ), then we call a local equation for D on U.
Intersection theory.
Fix a nonsingular variety V of dimension n over a eld k, assumed to be perfect. Let
W
1
and W
2
be irreducible closed subsets of V , and let 7 be an irreducible component of
W
1
W
2
. Then intersection theory attaches a multiplicity to 7. We shall only do this in an
easy case.
Intersection theory. 197
Divisors.
Let V be a nonsingular variety of dimension n, and let D
1
. . . . . D
n
be effective divisors on
V . We say that D
1
. . . . . D
n
intersect properly at 1 [D
1
[ . . . [D
n
[ if 1 is an isolated
point of the intersection. In this case, we dene
(D
1
. . . D
n
)
1
=dim
k
O
1
,(
1
. . . . .
n
)
where
i
is a local equation for D
i
near 1. The hypothesis on 1 implies that this is nite.
EXAMPLE 12.3. In all the examples, the ambient variety is a surface.
(a) Let 7
1
be the afne plane curve Y
2
X
3
, let 7
2
be the curve Y = X
2
, and let
1 =(0. 0). Then
(7
1
7
2
)
1
=dimkX. Y 
(A,Y )
,(Y X
2
. Y
2
X
3
) =dimkX,(X
4
X
3
) =3.
(b) If 7
1
and 7
2
are prime divisors, then (7
1
7
2
)
1
= 1 if and only if
1
,
2
are local
uniformizing parameters at 1. Equivalently, (7
1
7
2
)
1
=1 if and only if 7
1
and 7
2
are
transversal at 1, that is, T
7
1
(1) T
7
2
(1) ={0].
(c) Let D
1
be the .axis, and let D
2
be the cuspidal cubic Y
2
X
3
. For 1 = (0. 0),
(D
1
D
2
)
1
=3.
(d) In general, (7
1
7
2
)
1
is the order of contact of the curves 7
1
and 7
2
.
We say that D
1
. . . . . D
n
intersect properly if they do so at every point of intersection of
their supports; equivalently, if [D
1
[ . . . [D
n
[ is a nite set. We then dene the intersec
tion number
(D
1
. . . D
n
) =
1[T
1
[...[T
n
[
(D
1
. . . D
n
)
1
.
EXAMPLE 12.4. Let C be a curve. If D =
n
i
1
i
, then the intersection number
(D) =
n
i
k(1
i
) : k.
By denition, this is the degree of D.
Consider a regular map : W V of connected nonsingular varieties, and let D be a
divisor on V whose support does not contain the image of W. There is then a unique divisor
+
D on W with the following property: if D has local equation on the open subset U
of V , then
+
D has local equation on
1
U. (Use 9.2 to see that this does dene a
divisor on W; if the image of is disjoint from [D[, then
+
D =0.)
EXAMPLE 12.5. Let C be a curve on a surface V , and let : C
t
C be the normalization
of C. For any divisor D on V ,
(C D) =deg
+
D.
LEMMA 12.6 (ADDITIVITY). Let D
1
. . . . . D
n
. D be divisors on V . If (D
1
. . . D
n
)
1
and
(D
1
. . . D)
1
are both dened, then so also is (D
1
. . . D
n
D)
1
. and
(D
1
. . . D
n
D)
1
=(D
1
. . . D
n
)
1
(D
1
. . . D)
1
.
198 12. DIVISORS AND INTERSECTION THEORY
PROOF. One writes some exact sequences. See Shafarevich 1994, IV.1.2.
2
Note that in intersection theory, unlike every other branch of mathematics, we add rst,
and then multiply.
Since every divisor is the difference of two effective divisors, Lemma 12.1 allows us to
extend the denition of (D
1
. . . D
n
) to all divisors intersecting properly (not just effective
divisors).
LEMMA 12.7 (INVARIANCE UNDER LINEAR EQUIVALENCE). Assume V is complete. If
D
n
~D
t
n
, then
(D
1
. . . D
n
) =(D
1
. . . D
t
n
).
PROOF. By additivity, it sufces to show that (D
1
. . . D
n
) =0 if D
n
is a principal divisor.
For n =1, this is just the statement that a function has as many poles as zeros (counted with
multiplicities). Suppose n =2. By additivity, we may assume that D
1
is a curve, and then
the assertion follows from Example 12.5 because
D principal =
+
D principal.
The general case may be reduced to this last case (with some difculty). See Shafare
vich 1994, IV.1.3.
2
LEMMA 12.8. For any n divisors D
1
. . . . . D
n
on an ndimensional variety, there exists n
divisors D
t
1
. . . . . D
t
n
intersect properly.
PROOF. See Shafarevich 1994, IV.1.4.
2
We can use the last two lemmas to dene (D
1
. . . D
n
) for any divisors on a complete
nonsingular variety V : choose D
t
1
. . . . . D
t
n
as in the lemma, and set
(D
1
. . . D
n
) =(D
t
1
. . . D
t
n
).
EXAMPLE 12.9. Let C be a smooth complete curve over C, and let : C C be a regular
map. Then the Lefschetz trace formula states that
(z 1
) =Tr([H
0
(C. )Tr([H
1
(C. )Tr([H
2
(C. ).
In particular, we see that (z z) = 2 2g, which may be negative, even though z is an
effective divisor.
Let : W V be a nite map of irreducible varieties. Then k(W) is a nite extension
of k(V ), and the degree of this extension is called the degree of . If k(W) is separable
over k(V ) and k is algebraically closed, then there is an open subset U of V such that
1
(u) consists exactly J =deg points for all u U. In fact,
1
(u) always consists of
exactly deg points if one counts multiplicities. Number theorists will recognize this as the
formula
e
i
i
=J. Here the
i
are 1 (if we take k to be algebraically closed), and e
i
is
the multiplicity of the i th point lying over the given point.
A nite map : W V is at if every point 1 of V has an open neighbourhood U such
that 1(
1
U. O
W
) is a free 1(U. O
V
)module it is then free of rank deg.
Intersection theory. 199
THEOREM 12.10. Let : W V be a nite map between nonsingular varieties. For any
divisors D
1
. . . . . D
n
on V intersecting properly at a point 1 of V ,
(Q)=1
(
+
D
1
. . .
+
D
n
) =deg (D
1
. . . D
n
)
1
.
PROOF. After replacing V by a sufciently small open afne neighbourhood of 1, we may
assume that corresponds to a map of rings T and that T is free of rank J =deg
as an module. Moreover, we may assume that D
1
. . . . . D
n
are principal with equations
1
. . . . .
n
on V , and that 1 is the only point in [D
1
[ . . . [D
n
[. Then m
1
is the only
ideal of containing a =(
1
. . . . .
n
). Set S =m
1
; then
S
1
,S
1
a =S
1
(,a) =,a
because ,a is already local. Hence
(D
1
. . . D
n
)
1
=dim,(
1
. . . . .
n
).
Similarly,
(
+
D
1
. . .
+
D
n
)
1
=dimT,(
1
. . . . .
n
).
But T is a free module of rank J, and
,(
1
. . . . .
n
)
T =T,(
1
. . . . .
n
).
Therefore, as modules, and hence as kvector spaces,
T,(
1
. . . . .
n
) ~(,(
1
. . . . .
n
))
d
which proves the formula.
2
EXAMPLE 12.11. Assume k is algebraically closed of characteristic =0. Let : A
1
A
1
be the Frobenius map c c
]
. It corresponds to the map kX kX, X X
]
, on
rings. Let D be the divisor c. It has equation X c on A
1
, and
+
D has the equation
X
]
c =(X ;)
]
. Thus
+
D =(;), and so
deg(
+
D) = = deg(D).
The general case.
Let V be a nonsingular connected variety. A cycle of codimension r on V is an element of
the free abelian group C
i
(V ) generated by the prime cycles of codimension r.
Let 7
1
and 7
2
be prime cycles on any nonsingular variety V , and let W be an irre
ducible component of 7
1
7
2
. Then
dim 7
1
dim 7
2
_dim V dim W.
and we say 7
1
and 7
2
intersect properly at W if equality holds.
Dene O
V,W
to be the set of rational functions on V that are dened on some open
subset U of V with U W = 0 it is a local ring. Assume that 7
1
and 7
2
intersect
properly at W, and let p
1
and p
2
be the ideals in O
V,W
corresponding to 7
1
and 7
2
(so
200 12. DIVISORS AND INTERSECTION THEORY
p
i
=(
1
.
2
. ....
i
) if the
}
dene 7
i
in some open subset of V meeting W). The example
of divisors on a surface suggests that we should set
(7
1
7
2
)
W
=dim
k
O
V,W
,(p
1
. p
2
).
but examples show this is not a good denition. Note that
O
V,W
,(p
1
. p
2
) =O
V,W
,p
1
O
V;W
O
V,W
,p
2
.
It turns out that we also need to consider the higher Tor terms. Set
O
(O,p
1
. O,p
2
) =
dimV
i=0
(1)
i
dim
k
(Tor
O
i
(O,p
1
. O,p
2
))
where O =O
V,W
. It is an integer _ 0, and =0 if 7
1
and 7
2
do not intersect properly at
W. When they do intersect properly, we dene
(7
1
7
2
)
W
=mW. m=
O
(O,p
1
. O,p
2
).
When 7
1
and 7
2
are divisors on a surface, the higher Tors vanish, and so this denition
agrees with the previous one.
Now assume that V is projective. It is possible to dene a notion of rational equivalence
for cycles of codimension r: let W be an irreducible subvariety of codimension r 1, and let
k(W)
2
points (counting multiplicities).
Exercises 201
References.
Fulton, W., Introduction to Intersection Theory in Algebraic Geometry, (AMS Publication;
CBMS regional conference series #54.) This is a pleasant introduction.
Fulton, W., Intersection Theory. Springer, 1984. The ultimate source for everything to
do with intersection theory.
Serre: Alg` ebre Locale, Multiplicit es, Springer Lecture Notes, 11, 1957/58 (third edition
1975). This is where the denition in terms of Tors was rst suggested.
Exercises
You may assume the characteristic is zero if you wish.
121. Let V = V(J) P
n
, where J is a homogeneous polynomial of degree without
multiple factors. Show that V has degree according to the denition in the notes.
122. Let C be a curve in A
2
dened by an irreducible polynomial J(X. Y ), and assume C
passes through the origin. Then J =J
n
J
n1
, m_1, with J
n
the homogeneous
part of J of degree m. Let o: W A
2
be the blowup of A
2
at (0. 0), and let C
t
be the
closure of o
1
(C (0. 0)). Let 7 = o
1
(0. 0). Write J
n
=
x
i=1
(a
i
X b
i
Y )
i
i
, with
the (a
i
: b
i
) being distinct points of P
1
, and show that C
t
7 consists of exactly s distinct
points.
123. Find the intersection number of D
1
: Y
2
=X
i
and D
2
: Y
2
=X
x
, r > s > 2, at the
origin.
124. Find Pic(V ) when V is the curve Y
2
=X
3
.
CHAPTER 13
Coherent Sheaves; Invertible
Sheaves
In this chapter, k is an arbitrary eld.
Coherent sheaves
Let V = Spm be an afne variety over k, and let M be a nitely generated module.
There is a unique sheaf of O
V
modules Mon V such that, for all ,
1(D( ). M) =M
(
(=
(
M).
Such an O
V
module Mis said to be coherent. A homomorphism M N of modules
denes a homomorphismMN of O
V
modules, and M Mis a fully faithful functor
from the category of nitely generated modules to the category of coherent O
V
modules,
with quasiinverse M1(V. M).
Now consider a variety V . An O
V
module Mis said to be coherent if, for every open
afne subset U of V , M[U is coherent. It sufces to check this condition for the sets in an
open afne covering of V .
For example, O
n
V
is a coherent O
V
module. An O
V
module Mis said to be locally
free of rank n if it is locally isomorphic to O
n
V
, i.e., if every point 1 V has an open
neighbourhood such that M[U ~O
n
V
. A locally free O
V
module of rank n is coherent.
Let V , and let Mbe a coherent O
V
module. We dene a k()module M() as
follows: after replacing V with an open neighbourhood of , we can assume that it is afne;
hence we may suppose that V =Spm(), that corresponds to a maximal ideal m in (so
that k() =,m), and Mcorresponds to the module M; we then dene
M() =M
k() =M,mM.
It is a nitely generated vector space over k(). Dont confuse M() with the stalk M
of
Mwhich, with the above notations, is M
m
=M
m
. Thus
M() =M
,mM
=k()
m
M
m
.
Nakayamas lemma (1.3) shows that
M() =0 =M
=0.
203
204 13. COHERENT SHEAVES; INVERTIBLE SHEAVES
The support of a coherent sheaf Mis
Supp(M) ={ V [ M() =0] ={ V [ M
=0].
Suppose V is afne, and that Mcorresponds to the module M. Let a be the annihilator
of M:
a ={ [ M =0].
Then M,mM =0 ma (for otherwise ,m contains a nonzero element annihilat
ing M,mM), and so
Supp(M) =V(a).
Thus the support of a coherent module is a closed subset of V .
Note that if Mis locally free of rank n, then M() is a vector space of dimension n for
all . There is a converse of this.
PROPOSITION 13.1. If Mis a coherent O
V
module such that M() has constant dimen
sion n for all V , then Mis a locally free of rank n.
PROOF. We may assume that V is afne, and that Mcorresponds to the nitely generated
module M. Fix a maximal ideal m of , and let .
1
. . . . . .
n
be elements of M whose
images in M,mM form a basis for it over k(). Consider the map
;:
n
M. (a
1
. . . . . a
n
)
a
i
.
i
.
Its cokernel is a nitely generated module whose support does not contain . Therefore
there is an element , m, such that ; denes a surjection
n
(
M
(
. After
replacing with
(
we may assume that ; itself is surjective. For every maximal ideal
n of , the map (,n)
n
M,nM is surjective, and hence (because of the condition on
the dimension of M()) bijective. Therefore, the kernel of ; is contained in n
n
(meaning
n n) for all maximal ideals n in , and the next lemma shows that this implies that
the kernel is zero.
2
LEMMA 13.2. Let be an afne kalgebra. Then
_
m=0 (intersection of all maximal ideals in ).
PROOF. When k is algebraically closed, we showed (4.13) that this follows from the strong
Nullstellensatz. In the general case, consider a maximal ideal m of
k
k
al
. Then
,(m) (
k
k
al
),m=k
al
.
and so ,m is an integral domain. Since it is nitedimensional over k, it is a eld,
and so m is a maximal ideal in . Thus if is in all maximal ideals of , then its
image in k
al
is in all maximal ideals of , and so is zero.
2
For two coherent O
V
modules M and N, there is a unique coherent O
V
module
M
O
V
N such that
1(U. M
O
V
N) =1(U. M)
T(U,O
V
)
1(U. N)
Invertible sheaves. 205
for all open afnes U V . The reader should be careful not to assume that this formula
holds for nonafne open subsets U (see example 13.4 below). For a such a U, one writes
U =
_
U
i
with the U
i
open afnes, and denes 1(U. M
O
V
N) to be the kernel of
i
1(U
i
. M
O
V
N)
i,}
1(U
i}
. M
O
V
N).
Dene Hom(M. N) to be the sheaf on V such that
1(U. Hom(M. N)) =Hom
O
U
(M. N)
(homomorphisms of O
U
modules) for all open U in V . It is easy to see that this is a sheaf.
If the restrictions of Mand N to some open afne U correspond to modules M and N,
then
1(U. Hom(M. N)) =Hom
(M. N).
and so Hom(M. N) is again a coherent O
V
module.
Invertible sheaves.
An invertible sheaf on V is a locally free O
V
module L of rank 1. The tensor product of
two invertible sheaves is again an invertible sheaf. In this way, we get a product structure
on the set of isomorphism classes of invertible sheaves:
L L
t

def
=LL
t
.
The product structure is associative and commutative (because tensor products are associa
tive and commutative, up to isomorphism), and O
V
 is an identity element. Dene
L
=Hom(L. O
V
).
Clearly, L
is free of rank 1 over any open set where L is free of rank 1, and so L
is again
an invertible sheaf. Moreover, the canonical map
L
L O
V
. (. .) (.)
is an isomorphism (because it is an isomorphism over any open subset where L is free).
Thus
L
L =O
V
.
For this reason, we often write L
1
for L
.
From these remarks, we see that the set of isomorphism classes of invertible sheaves on
V is a group it is called the Picard group, Pic(V ), of V .
We say that an invertible sheaf L is trivial if it is isomorphic to O
V
then L represents
the zero element in Pic(V ).
PROPOSITION 13.3. An invertible sheaf L on a complete variety V is trivial if and only if
both it and its dual have nonzero global sections, i.e.,
1(V. L) =0 =1(V. L
).
206 13. COHERENT SHEAVES; INVERTIBLE SHEAVES
PROOF. We may assume that V is irreducible. Note rst that, for any O
V
module Mon
any variety V , the map
Hom(O
V
. M) 1(V. M). (1)
is an isomorphism.
Next recall that the only regular functions on a complete variety are the constant func
tions (see 7.5 in the case that k is algebraically closed), i.e., 1(V. O
V
) = k
t
where k
t
is
the algebraic closure of k in k(V ). Hence Hom(O
V
. O
V
) =k
t
, and so a homomorphism
O
V
O
V
is either 0 or an isomorphism.
We now prove the proposition. The sections dene nonzero homomorphisms
s
1
: O
V
L. s
2
: O
V
L
.
We can take the dual of the second homomorphism, and so obtain nonzero homomorphisms
O
V
x
1
L
x
_
2
O
V
.
The composite is nonzero, and hence an isomorphism, which shows that s
2
is surjective,
and this implies that it is an isomorphism (for any ring , a surjective homomorphism of
modules is bijective because 1 must map to a unit).
2
Invertible sheaves and divisors.
Now assume that V is nonsingular and irreducible. For a divisor D on V , the vector space
1(D) is dened to be
1(D) ={ k(V )
[ div( ) D _0].
We make this denition local: dene L(D) to be the sheaf on V such that, for any open set
U,
1(U. L(D)) ={ k(V )
[ div( ) D _0 on U] L{0].
The condition div( )D_0 on U means that, if D=
n
7
7, then ord
7
( )n
7
_0
for all 7 with 7U =0. Thus, 1(U. L(D)) is a 1(U. O
V
)module, and if U U
t
, then
1(U
t
. L(D)) 1(U. L(D)). We dene the restriction map to be this inclusion. In this
way, L(D) becomes a sheaf of O
V
modules.
Suppose D is principal on an open subset U, say D[U =div(g), g k(V )
. Then
1(U. L(D)) ={ k(V )
[ div(g) _0 on U] L{0].
Therefore,
1(U. L(D)) 1(U. O
V
). g.
is an isomorphism. These isomorphisms clearly commute with the restriction maps for
U
t
U, and so we obtain an isomorphism L(D)[U O
U
. Since every D is locally
principal, this shows that L(D) is locally isomorphic to O
V
, i.e., that it is an invertible
sheaf. If D itself is principal, then L(D) is trivial.
Next we note that the canonical map
L(D) L(D
t
) L(DD
t
). g g
Invertible sheaves and divisors. 207
is an isomorphism on any open set where D and D
t
are principal, and hence it is an isomor
phism globally. Therefore, we have a homomorphism
Div(V ) Pic(V ). D L(D).
which is zero on the principal divisors.
EXAMPLE 13.4. Let V be an elliptic curve, and let 1 be the point at innity. Let D be the
divisor D = 1. Then 1(V. L(D)) = k, the ring of constant functions, but 1(V. L(2D))
contains a nonconstant function .. Therefore,
1(V. L(2D)) =1(V. L(D)) 1(V. L(D)).
in other words, 1(V. L(D) L(D)) =1(V. L(D)) 1(V. L(D)).
PROPOSITION 13.5. For an irreducible nonsingular variety, the map D L(D) denes
an isomorphism
Div(V ),PrinDiv(V ) Pic(V ).
PROOF. (Injectivity). If s is an isomorphism O
V
L(D), then g =s(1) is an element of
k(V )
such that
(a) div(g) D _0 (on the whole of V );
(b) if div( ) D _ 0 on U, that is, if 1(U. L(D)), then = h(g[U) for some
h 1(U. O
V
).
Statement (a) says that D_div(g) (on the whole of V ). Suppose U is such that D[U
admits a local equation = 0. When we apply (b) to , then we see that div( ) _
div(g) on U, so that D[U div(g) _ 0. Since the Us cover V , together with (a) this
implies that D =div(g).
(Surjectivity). Dene
1(U. K) =
_
k(V )
1 is a vector space of
dimension one).
208 13. COHERENT SHEAVES; INVERTIBLE SHEAVES
This can be proved directly. First show that every (fractional) ideal is a projective 
module it is obviously nitely generated of rank one; then show that two ideals are
isomorphic as modules if and only if they differ by a principal divisor; nally, show
that every nitely generated projective module of rank 1 is isomorphic to a fractional
ideal (by assumption M
1 =1, then
M M
W and coherent O
W
module F on
W, there is a canonical isomorphism
()
+
F
~
+
(
+
F).
(b) For any regular map : V W,
+
maps locally free sheaves of rank n to locally free
sheaves of rank n (hence also invertible sheaves to invertible sheaves). It preserves
tensor products, and, for an invertible sheaf L,
+
(L
1
) .(
+
L)
1
.
PROOF. (a) This follows from the fact that, given homomorphisms of rings T T ,
T
B
(T
M) =T
M.
(b) This again follows from wellknown facts about tensor products of rings.
2
See Kleiman.
Principal bundles 209
Principal bundles
To be added.
CHAPTER 14
Differentials (Outline)
In this subsection, we sketch the theory of differentials. We allow k to be an arbitrary eld.
Let be a kalgebra, and let M be an module. Recall (from 5) that a kderivation
is a klinear map D: M satisfying Leibnizs rule:
D(g) = Dg g D. all . g .
A pair (
1
{k
. J) comprising an module
1
{k
and a kderivation J:
1
{k
is called
the module of differential oneforms for over k
al
if it has the following universal prop
erty: for any kderivation D: M, there is a unique klinear map :
1
{k
M such
that D = J,
1
M.
d
T 3 klinear
EXAMPLE 14.1. Let = kX
1
. .... X
n
; then
1
{k
is the free module with basis the
symbols JX
1
. .... JX
n
, and
J =
d
dX
i
JX
i
.
EXAMPLE 14.2. Let =kX
1
. .... X
n
,a; then
1
{k
is the free module with basis the
symbols JX
1
. .... JX
n
modulo the relations:
J =0 for all a.
PROPOSITION 14.3. Let V be a variety. For each n _ 0, there is a unique sheaf of O
V

modules
n
V{k
on V such that
n
V{k
(U) =
_
n
1
{k
whenever U = Spm is an open
afne of V .
PROOF. Omitted.
2
The sheaf
n
V{k
is called the sheaf of differential nforms on V .
211
212 14. DIFFERENTIALS (OUTLINE)
EXAMPLE 14.4. Let 1 be the afne curve
Y
2
=X
3
aX b.
and assume X
3
aX b has no repeated roots (so that 1 is nonsingular). Write . and
, for regular functions on 1 dened by X and Y . On the open set D(,) where , = 0,
let o
1
= J.,,, and on the open set D(3.
2
a), let o
2
= 2J,,(3.
2
a). Since ,
2
=
.
3
a. b,
2,J, =(3.
2
a)J..
and so o
1
and o
2
agree on D(,) D(3.
2
a). Since 1 = D(,) LD(3.
2
a), we see
that there is a differential o on 1 whose restrictions to D(,) and D(3.
2
a) are o
1
and
o
2
respectively. It is an easy exercise in working with projective coordinates to show that
o extends to a differential oneform on the whole projective curve
Y
2
7 =X
3
aX7
2
b7
3
.
In fact,
1
C{k
(C) is a onedimensional vector space over k, with o as basis. Note that o =
J.,, =J.,(.
3
a. b)
1
2
, which cant be integrated in terms of elementary functions.
Its integral is called an elliptic integral (integrals of this form arise when one tries to nd
the arc length of an ellipse). The study of elliptic integrals was one of the starting points for
the study of algebraic curves.
In general, if C is a complete nonsingular absolutely irreducible curve of genus g, then
1
C{k
(C) is a vector space of dimension g over k.
PROPOSITION 14.5. If V is nonsingular, then
1
V{k
is a locally free sheaf of rank dim(V )
(that is, every point 1 of V has a neighbourhood U such that
1
V{k
[U ~(O
V
[U)
dim(V )
).
PROOF. Omitted.
2
Let C be a complete nonsingular absolutely irreducible curve, and let o be a nonzero
element of
1
k(C){k
. We dene the divisor (o) of o as follows: let 1 C; if t is a uni
formizing parameter at 1, then Jt is a basis for
1
k(C){k
as a k(C)vector space, and so we
can write o =Jt , k(V )
; dene orJ
1
(o) =orJ
1
( ), and (o) =
orJ
1
(o)1.
Because k(C) has transcendence degree 1 over k,
1
k(C){k
is a k(C)vector space of di
mension one, and so the divisor (o) is independent of the choice of o up to linear equiv
alence. By an abuse of language, one calls (o) for any nonzero element of
1
k(C){k
a
canonical class 1 on C. For a divisor D on C, let (D) =dim
k
(1(D)).
THEOREM 14.6 (RIEMANNROCH). Let C be a complete nonsingular absolutely irre
ducible curve over k.
(a) The degree of a canonical divisor is 2g 2.
(b) For any divisor D on C,
(D) (1D) =1g deg(D).
213
More generally, if V is a smooth complete variety of dimension J, it is possible to
associate with the sheaf of differential Jforms on V a canonical linear equivalence class
of divisors 1. This divisor class determines a rational map to projective space, called the
canonical map.
References
Shafarevich, 1994, III.5.
Mumford 1999, III.4.
CHAPTER 15
Algebraic Varieties over the
Complex Numbers
This is only an outline.
It is not hard to show that there is a unique way to endow all algebraic varieties over C
with a topology such that:
(a) on A
n
=C
n
it is just the usual complex topology;
(b) on closed subsets of A
n
it is the induced topology;
(c) all morphisms of algebraic varieties are continuous;
(d) it is ner than the Zariski topology.
We call this new topology the complex topology on V . Note that (a), (b), and (c) deter
mine the topology uniquely for afne algebraic varieties ((c) implies that an isomorphism
of algebraic varieties will be a homeomorphism for the complex topology), and (d) then
determines it for all varieties.
Of course, the complex topology is much ner than the Zariski topology this can be
seen even on A
1
. In view of this, the next two propositions are a little surprising.
PROPOSITION 15.1. If a nonsingular variety is connected for the Zariski topology, then it
is connected for the complex topology.
Consider, for example, A
1
. Then, certainly, it is connected for both the Zariski topology
(that for which the nonempty open subsets are those that omit only nitely many points)
and the complex topology (that for which X is homeomorphic to 1
2
). When we remove a
circle from X, it becomes disconnected for the complex topology, but remains connected
for the Zariski topology. This doesnt contradict the proposition, because A
1
C
with a circle
removed is not an algebraic variety.
Let X be a connected nonsingular (hence irreducible) curve. We prove that it is con
nected for the complex topology. Removing or adding a nite number of points to X will
not change whether it is connected for the complex topology, and so we can assume that X
is projective. Suppose X is the disjoint union of two nonempty open (hence closed) sets X
1
and X
2
. According to the RiemannRoch theorem (14.6), there exists a nonconstant rational
function on X having poles only in X
1
. Therefore, its restriction to X
2
is holomorphic.
Because X
2
is compact, is constant on each connected component of X
2
(Cartan 1963
1
,
1
Cartan, H., Elementary Theory of Analytic Functions of One or Several Variables, AddisonWesley, 1963.
215
216 15. ALGEBRAIC VARIETIES OVER THE COMPLEX NUMBERS
VI.4.5) say, (:) =a on some innite connected component. Then (:) a has innitely
many zeros, which contradicts the fact that it is a rational function.
The general case can be proved by induction on the dimension (Shafarevich 1994,
VII.2).
PROPOSITION 15.2. Let V be an algebraic variety over C, and let C be a constructible
subset of V (in the Zariski topology); then the closure of C in the Zariski topology equals
its closure in the complex topology.
PROOF. Mumford 1999, I 10, Corollary 1, p60.
2
For example, if U is an open dense subset of a closed subset 7 of V (for the Zariski
topology), then U is also dense in 7 for the complex topology.
The next result helps explain why completeness is the analogue of compactness for
topological spaces.
PROPOSITION 15.3. Let V be an algebraic variety over C; then V is complete (as an alge
braic variety) if and only if it is compact for the complex topology.
PROOF. Mumford 1999, I 10, Theorem 2, p60.
2
In general, there are many more holomorphic (complex analytic) functions than there
are polynomial functions on a variety over C. For example, by using the exponential func
tion it is possible to construct many holomorphic functions on C that are not polynomials
in :, but all these functions have nasty singularities at the point at innity on the Riemann
sphere. In fact, the only meromorphic functions on the Riemann sphere are the rational
functions. This generalizes.
THEOREM 15.4. Let V be a complete nonsingular variety over C. Then V is, in a natural
way, a complex manifold, and the eld of meromorphic functions on V (as a complex
manifold) is equal to the eld of rational functions on V .
PROOF. Shafarevich 1994, VIII 3.1, Theorem 1.
2
This provides one way of constructing compact complex manifolds that are not alge
braic varieties: nd such a manifold M of dimension n such that the transcendence degree
of the eld of meromorphic functions on M is <n. For a torus C
, of dimension g >1,
this is typically the case. However, when the transcendence degree of the eld of meromor
phic functions is equal to the dimension of manifold, then M can be given the structure, not
necessarily of an algebraic variety, but of something more general, namely, that of an alge
braic space in the sense of Artin.
2
Roughly speaking, an algebraic space is an object that is
locally an afne algebraic variety, where locally means for the etale topology rather than
the Zariski topology.
3
One way to show that a complex manifold is algebraic is to embed it into projective
space.
2
Perhaps these should be called algebraic orbispaces (in analogy with manifolds and orbifolds).
3
Artin, Michael. Algebraic spaces. Whittemore Lectures given at Yale University, 1969. Yale Mathemati
cal Monographs, 3. Yale University Press, New Haven, Conn.London, 1971. vii+39 pp.
Knutson, Donald. Algebraic spaces. Lecture Notes in Mathematics, Vol. 203. SpringerVerlag, BerlinNew
York, 1971. vi+261 pp.
217
THEOREM 15.5. Any closed analytic submanifold of P
n
is algebraic.
PROOF. See Shafarevich 1994, VIII 3.1, in the nonsingular case.
2
COROLLARY 15.6. Any holomorphic map from one projective algebraic variety to a sec
ond projective algebraic variety is algebraic.
PROOF. Let : V W be the map. Then the graph 1
, ( a lattice in C
]
, which has coefcients in k, but
1(V ) =(X Y ) is not generated by polynomials in kX. Y .
Fixed elds
Let k be elds, and let 1 be the group Aut(,k) of automomorphisms of (as an
abstract eld) xing the elements of k. Dene the xed eld
T
of 1 to be
{a [ oa =a for all o 1 ].
PROPOSITION 16.1. The xed eld of 1 equals k in each of the following two cases:
219
220 16. DESCENT THEORY
(a) is a Galois extension of k (possibly innite);
(b) is a separably closed eld and k is perfect.
PROOF. (a) See FT 7.8.
(b) See FT 8.23.
2
REMARK 16.2. (a) The proof of Proposition 16.1 denitely requires the axiom of choice.
For example, it is known that every measurable homomorphism of Lie groups is continuous,
and so any measurable automorphism of C is equal to the identity map or to complex con
jugation. Therefore, without the axiom of choice, 1
def
=Aut(C,) has only two elements,
and C
T
=1.
(b) Suppose that is algebraically closed and k is not perfect. Then k has characteristic
= 0 and contains an element such that k but
]
= a k. As is the unique
root of X
]
a, every automorphism of xing k also xes , and so
T
=k.
The perfect closure of k in is the subeld
k
]
1
={ [
]
n
k for some n]
of . Then k
]
1
is purely inseparable over k, and when is algebraically closed, it is
the smallest perfect subeld of containing k.
COROLLARY 16.3. If is separably closed, then
T
is a purely inseparable algebraic
extension of k.
PROOF. When k has characteristic zero,
T
= k, and there is nothing to prove. Thus,
we may suppose that k has characteristic = 0. Choose an algebraic closure
al
of ,
and let k
]
1
be the perfect closure of k in
al
. As
al
is purely inseparable over ,
every element o of 1 extends uniquely to an automorphism o of
al
: let
al
and let
]
n
; then o() is the unique root of X
]
n
o(
]
n
) in . The action of 1 on
al
identies it with Aut(
al
,k
]
1
). According to the proposition, (
al
)
T
=k
]
1
, and so
k
]
1
T
k.
2
Descending subspaces of vector spaces
In this subsection, k are elds such that k is the xed eld of 1 =Aut(,k).
Let V be a ksubspace of an vector space V() such that the map
c c:
k
V V()
is an isomorphism. Equivalent conditions: V is the kspan of an basis for V(); every
kbasis for V is an basis for V(). The group 1 acts on
k
V through its action on
:
o(
c
i
i
) =
oc
i
i
. o 1. c
i
.
i
V. (27)
Correspondingly, there is a unique action of 1 on V() xing the elements of V and such
that each o 1 acts olinearly:
o(c) =o(c)o() all o 1 , c , V(). (28)
Descending subspaces of vector spaces 221
LEMMA 16.4. The following conditions on a subspace W of V() are equivalent:
(a) W V spans W;
(b) W V contains an basis for W;
(c) the map
k
(W V ) W, c c, is an isomorphism.
PROOF. (a) ==(b,c) A klinearly independent subset of V is linearly independent in
V(). Therefore, if W V spans W, then any kbasis (e
i
)
iJ
for W V will be an 
basis for W. Moreover, (1 e
i
)
iJ
will be an basis for
k
(W V ), and since the
map
k
(W V ) W sends 1e
i
to e
i
, it is an isomorphism.
(c) ==(a), (b) ==(a). Obvious.
2
LEMMA 16.5. For any kvector space V , V =V()
T
.
PROOF. Let (e
i
)
iJ
be a kbasis for V . Then (1 e
i
)
iJ
is an basis for
k
V , and
o 1 acts on =
c
i
e
i
according to the rule (27). Thus, is xed by 1 if and only if
each c
i
is xed by 1 and so lies in k.
2
LEMMA 16.6. Let V be a kvector space, and let W be a subspace of V() stable under
the action of 1 . If W
T
=0, then W =0.
PROOF. Suppose W =0. As V contains an basis for V(), every nonzero element n
of W can be expressed in the form
n =c
1
e
1
c
n
e
n
. c
i
{0]. e
i
V. n _1.
Choose n to be a nonzero element for which n takes its smallest value. After scaling, we
may suppose that c
1
=1. For o 1 , the element
onn =(oc
2
c
2
)e
2
(oc
n
c
n
)e
n
lies in W and has at most n1 nonzero coefcients, and so is zero. Thus, n W
T
={0],
which is a contradiction.
2
PROPOSITION 16.7. Asubspace W of V() is of the formW =W
0
for some ksubspace
W
0
of V if and only if it is stable under the action of 1 .
PROOF. Certainly, if W = W
0
, then it is stable under 1 (and W = (W V )). Con
versely, assume that W is stable under 1 , and let W
t
be a complement to W V in V , so
that
V =(W V ) W
t
.
Then
(W W
t
)
T
=W
T
_
W
t
_
T
=(W V ) W
t
=0,
and so, by (16.6),
W W
t
=0. (29)
As W (W V ) and
V() =(W V ) W
t
,
this implies that W =(W V ).
2
222 16. DESCENT THEORY
Descending subvarieties and morphisms
In this subsection, k are elds such that k is the xed eld of 1 =Aut(,k) and
is separably closed. Recall (11.15) that for any variety V over , V() is Zariski dense in
V . In particular, two regular maps V V
t
coincide if they agree on V().
For any variety V over k, 1 acts on V(). For example, if V is embedded in A
n
or P
n
over k, then 1 acts on the coordinates of a point. If V =Spm, then
V() =Hom
kalgebra
(. ).
and 1 acts through its action on .
PROPOSITION 16.8. Let V be a variety over k, and let W be a closed subvariety of V
such that W() is stable under the action of 1 on V(). Then there is a closed subvariety
W
0
of V such that W =W
0
.
PROOF. Suppose rst that V is afne, and let 1(W) V
 =
k
kV  (see 11). Because W() is
stable under 1 , so also is 1(W), and Proposition 16.7 shows that 1(W) is spanned by
1
0
= 1(W) kV . Therefore, the zero set of 1
0
is a closed subvariety W
0
of V with the
property that W =W
0
.
To deduce the general case, cover V with open afnes V =
_
V
i
. Then W
i
def
=V
i
W
is stable under 1 , and so it arises from a closed subvariety W
i0
of V
i
; a similar statement
holds for W
i}
def
= W
i
W
}
. Dene W
0
to be the variety obtained by patching the varieties
W
i0
along the open subvarieties W
i}0
.
2
PROPOSITION 16.9. Let V and W be varieties over k, and let : V
be a regular
map. If commutes with the actions of 1 on V() and W(), then arises from a
(unique) regular map V W over k.
PROOF. Apply Proposition 16.8 to the graph of , 1
(
(V W)
.
2
COROLLARY 16.10. A variety V over k is uniquely determined (up to a unique isomor
phism) by the variety V
n
(k)
(S) = k, the set Hom
n
(k)
(S
k
V. S
k
W) has the same description, and so
the functor V S
k
V is fully faithful.
The functor V S
k
V sends a vector space V with basis (e
i
)
iJ
to a direct sum of
copies of S indexed by 1. Therefore, to show that the functor is essentially surjective, we
have to prove that every left M
n
(k)module is a direct sum of copies of S.
We rst prove this for M
n
(k) regarded as a left M
n
(k)module. For 1 _i _n, let 1(i )
be the set of matrices in M
n
(k) whose entries are zero except for those in the i th column.
Then 1(i ) is a left ideal in M
n
(k), and 1(i ) is isomorphic to S as an M
n
(k)module.
Hence,
M
n
(k) =
i
1(i ) .S
n
(as a left M
n
(k)module).
We now prove it for an arbitrary left M
n
(k)module M, which we may suppose to be
nonzero. The choice of a set of generators for M realizes it as a quotient of a sum of copies
of M
n
(k), and so M is a sum of copies of S. It remains to show that the sum can be made
direct. Let 1 be the set of submodules of M isomorphic to S, and let S be the set of subsets
J of 1 such that the sum N(J)
def
=
1J
N is direct, i.e., such that for any N
0
J and
nite subset J
0
of J not containing N
0
, N
0
1J
0
N =0. If J
1
J
2
. . . is a chain
of sets in S, then
_
J
i
S, and so Zorns lemma implies that S has maximal elements.
For any maximal J, M = N(J) because otherwise, there exists an element S
t
of 1 not
contained in N(J); because S
t
is simple, S
t
N(J) =0, and it follows that J L{S
t
] S,
contradicting the maximality of J.
2
ASIDE 16.13. Let and T be rings (not necessarily commutative), and let S be T
bimodule (this means that acts on S on the left, T acts on S on the right, and the actions
commute). When the functor M S
B
M: Mod
B
Mod
is an equivalence of cat
egories, and T are said to be Morita equivalent through S. In this terminology, the
lemma says that M
n
(k) and k are Morita equivalent through S.
PROPOSITION 16.14. Let be a nite Galois extension of k with Galois group 1 . The
functor V
k
V from kvector spaces to vector spaces endowed with an action of
1 is an equivalence of categories.
PROOF. Let 1  be the vector space with basis {o 1 ], and make 1  into a k
algebra by setting
_
cT
a
c
o
__
rT
b
r
t
_
=
c,r
(a
c
ob
r
)ot.
224 16. DESCENT THEORY
Then 1  acts klinearly on by the rule
(
cT
a
c
o)c =
cT
a
c
(oc).
and Dedekinds theorem on the independence of characters (FT 5.14) implies that the ho
momorphism
1  End
k
()
dened by this action is injective. By counting dimensions over k, one sees that it is an iso
morphism. Therefore, Lemma 16.12 shows that 1  and k are Morita equivalent through
, i.e., the functor V
k
V from kvector spaces to left 1 modules is an equiva
lence of categories. This is precisely the statement of the lemma.
2
When is an innite Galois extension of k, we endow 1 with the Krull topology, and
we say that an action of 1 on an vector space V is continuous if every element of V is
xed by an open subgroup of 1 , i.e., if
V =
_
z
V
z
(union over the open subgroups z of 1 ).
For example, the action of 1 on is obviously continuous, and it follows that, for any
kvector space V , the action of 1 on
k
V is continuous.
PROPOSITION 16.15. Let be a Galois extension of k (possibly innite) with Galois
group 1 . For any vector space V equipped with a continuous action of 1 , the map
c
i
c
i
i
:
k
V
T
V
is an isomorphism.
PROOF. Suppose rst that 1 is nite. Proposition 16.14 allows us to assume V =
k
W
for some ksubspace W of V . Then V
T
=(
k
W)
T
=W, and so the statement is true.
When 1 is innite, the nite case shows that
k
(V
z
)
T{z
. V
z
for every open
normal subgroup zof 1 . Now pass to the direct limit over z, recalling that tensor products
commute with direct limits (CA 8.1).
2
Descent data
For a homomorphism of elds o: J 1, we sometimes write oV for V
1
(the variety over
1 obtained by base change). For example, if V is embedded in afne or projective space,
then oV is the afne or projective variety obtained by applying o to the coefcients of the
equations dening V .
A regular map : V W denes a regular map
1
: V
1
W
1
which we also denote
o: oV oW. Note that (o)(o7) =o((7)) for any subvariety 7 of V . The map o
is obtained from by applying o to the coefcients of the polynomials dening .
Let k be elds, and let 1 =Aut(,k).
An ,kdescent system on a variety V over is
a family (
c
)
cT
of isomorphisms
c
: oV V
satisfying the following cocycle condition:
c
(o
r
) =
cr
for all o. t 1.
otV oV V
c
1
o for all o xing 1:
oV o(V
0
) =V
0
V.
c
A descent system (
c
)
cT
is said to be continuous if it is split by some model over a
subeld 1 of that is nitely generated over k. A descent datum is a continuous descent
system. A descent datum is effective if it is split by some model over k. In a given situation,
we say that descent is effective or that it is possible to descend the base eld if every descent
datum is effective.
Let V
0
be a variety over k, and let V =V
0
. Then V =oV because the two varieties are
obtained fromV
0
by extension of scalars with respect to the maps k 1and k 1
c
1,
which are equal. Write
c
for the identity map oV V ; then (
c
)
cT
is a descent datum
on V .
Let (
c
)
cT
be an ,k descent system on a variety V , and let 1
t
= Aut(
sep
,k).
Every kautomorphism of extends to a kautomorphism of
sep
, and (
c
)
cT
extends
to the
sep
,k descent system (
t
c
)
cT
0 on V
sep with
t
c
=
_
c[
_
sep
. A model of V
over a subeld 1 of splits (
c
)
cT
if and only if it splits (
t
c
)
cT
0 . This observation
sometimes allows us to assume that is separably closed.
PROPOSITION 16.16. Assume that k is the xed eld of 1 =Aut(,k), and that (V
0
. )
and (V
t
0
.
t
) split descent data (
c
)
cT
and (
t
c
)
cT
on varieties V and V
t
over . To
give a regular map [
0
: V
0
V
t
0
amounts to giving a regular map [: V V
t
such that
[
c
=
t
c
o[ for all o 1 , i.e., such that
oV
_
c)
_
)
oV
t
0
V
t
(31)
commutes for all o 1 .
PROOF. Given [
0
, dene [ to make the right hand square in
oV
c
V
0
_
c)
_
)
0
_
)
oV
t
c
0
V
t
0
0
V
t
commute. The left hand square is obtained from the right hand square by applying o, and
so it also commutes. The outer square is (31).
In proving the converse, we may assume that is separably closed. Given [, use
and
t
to transfer [ to a regular map [
t
: V
0
V
t
0
. Then the hypothesis implies that [
t
commutes with the actions of 1 on V
0
() and V
t
0
(), and so is dened over k (16.9).
2
226 16. DESCENT THEORY
COROLLARY 16.17. Assume that k is the xed eld of 1 =Aut(,k), and that (V
0
. )
splits the descent datum (
c
)
cT
. Let W be a variety over k. To give a regular map
W V
0
(resp. V
0
W) amounts to giving a regular map [: W
V (resp. [: V W
)
compatible with the descent datum, i.e., such that
c
o[ =[ (resp. [
c
=o[).
PROOF. Special case of the proposition in which W
c
(oW). Then the following diagram commutes:
oV
:
V
_
_
_
_
oW
[cW
:
c
W
LEMMA 16.19. The following hold.
(a) For all o. t 1 and W V ,
c
(
r
W) =
cr
W.
(b) Suppose that (
c
)
cT
is split by a model (V
0
. ) of V over k
0
, and let W be a
subvariety of V . If W =
1
(W
0
) for some subvariety W
0
of V
0
, then
c
W =W
for all o 1 ; the converse is true if
T
=k.
PROOF. (a) By denition
c
(
r
W) =
c
(o(
r
(tW)) =(
c
o
r
)(otW) =
cr
(otW) =
cr
W.
In the second equality, we used that (o)(oW) =o(W).
(b) Let W =
1
(W
0
). By hypothesis
c
=
1
o, and so
c
W =(
1
o)(oW) =
1
(o(W)) =
1
(oW
0
) =
1
(W
0
) =W.
Conversely, suppose
c
W =W for all o 1 . Then
(W) =(
c
W) =(o)(oW) =o((W)).
Therefore, (W) is stable under the action of 1 on V
0
, and so is dened over k (see
16.8).
2
For a descent system (
c
)
cT
on V and a regular function on an open subset U of
V , we dene
c
to be the function (o )
1
c
on
c
U, so that
c
(
c
1) =o ((1)) for all
1 U. Then
c
(
r
) =
cr
, and so this denes an action of 1 on the regular functions.
The Krull topology on 1 is that for which the subgroups of 1 xing a subeld of
nitely generated over k form a basis of open neighbourhoods of 1 (see FT Chapter 7). An
action of 1 on an vector space V is continuous if
V =
_
z
V
z
(union over the open subgroups z of 1 ).
For a subeld 1 of containing k, let z
1
=Aut(,1).
Descent data 227
PROPOSITION 16.20. Assume that is separably closed. A descent system (
c
)
cT
on
an afne variety V is continuous if and only if the action of 1 on V  is continuous.
PROOF. If (
c
)
cT
is continuous, it is split by a model of V over a subeld 1 of nitely
generated over k. By denition, z
1
is open, and V 
z
K
contains a set {
1
. . . . .
n
] of
generators for V  as an algebra. Now V  =
_
1
1
. . . . .
n
 where 1 runs over the
subelds of containing 1 and nitely generated over k. As 1
1
. . . . .
n
 = V 
z
L
,
this shows that V  =
_
V 
z
L
.
Conversely, if the action of 1 on V  is continuous, then for some subeld 1 of
nitely generated over k, V 
z
L
will contain a set of generators
1
. . . . .
n
for V  as
an algebra. According to (16.3),
z
L
is a purely inseparable algebraic extension of 1,
and so, after replacing 1 with a nite extension, the embedding V A
n
dened by the
i
will determine a model of V over 1. This model splits (
c
)
cT
, which is therefore
continuous.
2
PROPOSITION 16.21. A descent system (
c
)
cT
on a variety V over is continuous if
there exists a nite set S of points in V() such that
(a) any automorphism of V xing all 1 S is the identity map, and
(b) there exists a subeld 1 of nitely generated over k such that
c
1 = 1 for all
o 1 xing 1.
PROOF. Let (V
0
. ) be a model of V over a subeld 1 of nitely generated over k. After
possibly replacing 1 by a larger nitely generated eld, we may suppose (i) that
c
1 =1
for all o 1 xing 1 and all 1 S (because of (b)) and (ii) that (1) V
0
(1) for all
1 S (because S is nite). Then, for 1 S and every o xing 1,
c
(o1)
def
=
c
1
(i)
=1
(o)(o1) =o(1)
(ii)
=1.
and so
c
and
1
o are isomorphisms oV V sending o1 to 1. Therefore,
c
and
1
o differ by an automorphism of V xing the 1 S, which implies that they are
equal. This says that (V
0
. ) splits (
c
)
cT
.
2
PROPOSITION 16.22. Let V be a variety over whose only automorphism is the identity
map. A descent datum on V is effective if V has a model over k.
PROOF. Let (V. ) be a model of V over k. For o 1 , the maps
c
and
1
o are both
isomorphisms oV V , and so differ by an automorphism of V . Therefore they are equal,
which says that (V. ) splits (
c
)
cT
.
2
Of course, in Proposition 16.21, S doesnt have to be a nite set of points. The propo
sition will hold with S any additional structure on V that rigidies V (i.e., is such that
Aut(V. S) =1) and is such that (V. S) has a model over a nitely generated extension of k.
228 16. DESCENT THEORY
Galois descent of varieties
In this subsection, is a Galois extension of k with Galois group 1 .
THEOREM 16.23. A descent datum (
c
)
cT
on a variety V is effective if V is covered by
open afnes U with the property that
c
U =U for all o 1 .
PROOF. Assume rst that V is afne, and let =kV . A descent datum (
c
)
cT
denes
a continuous action of 1 on (see 16.20). From (16.15), we know that
c a ca:
k
T
(32)
is an isomorphism. Let V
0
=Spm
T
, and let be the isomorphism V V
0
dened by
(32). Then (V
0
. ) splits the descent datum.
In the general case, write V as a nite union of open afnes U
i
such that
c
U
i
=U
i
for
all o 1 . Then V is the variety over obtained by patching the U
i
by means of the maps
U
i
U
i
U
}
U
}
. (33)
Each intersection U
i
U
}
is again afne (4.27), and so the system (33) descends to k. The
variety over k obtained by patching the descended system is a model of V over k splitting
the descent datum.
2
COROLLARY 16.24. If each nite set of points of V(
sep
) is contained in an open afne
subvariety of V
sep
,kdescent datum for V
commutes.
In other words, (V
+
. ) is the 1,kWeil restriction of V if denes an isomorphism
Mor
k
(T. V
+
) Mor
1
(T
1
. V )
(natural in the kvariety T ); in particular,
V
+
() .V(1
k
)
(natural in the afne kalgebra ). If it exists, the 1,kWeil restriction of V is uniquely
determined by its universal property (up to a unique isomorphism).
When (V
+
. ) is the 1,kWeil restriction of V , the variety V
+
is said to have been
obtained from V by (Weil) restriction of scalars or by restriction of the base eld.
PROPOSITION 16.26. If V satises the hypothesis of (16.24) (for example, if V is quasipro
jective) and 1,k is separable, then the 1,kWeil restriction exists.
PROOF. Let be a Galois extension of k large enough to contain all conjugates of 1,
i.e., such that
k
1 .
r:1
t1. Let V
t
=
1
corresponds to
k
1 T
k
1. In the general case, we replace U with any open
afne and write V as a nite union of afnes V =
_
i
V
i
; then V
1
=
_
i
V
i1
.
Let 1 be a eld nitely generated over k, and let V be a variety over 1. For any
irreducible kvariety U with k(U) = 1, there will exist a dominant map : V U with
generic bre V . For example, we can take U =Spm() where is any nitely generated
230 16. DESCENT THEORY
ksubalgebra of 1 containing a set of generators for 1 and containing the coefcients of
some set of polynomials dening V . Let 1 be a point in the image of . Then the bre of
V over 1 is a variety V(1) over k, called the specialization of V at 1.
Similar statements are true for morphisms of varieties.
Rigid descent
LEMMA 16.27. Let V and W be varieties over an algebraically closed eld k. If V and W
become isomorphic over some eld containing k, then they are already isomorphic over k.
PROOF. The hypothesis implies that, for some eld 1 nitely generated over k, there exists
an isomorphism : V
1
W
1
. Let U be an afne kvariety such that k(U) = 1. After
possibly replacing U with an open subset, we can extend to an isomorphism
U
: U V
U W. The bre of
U
at any point of U is an isomorphism V W.
2
Consider elds 1
1
. 1
2
k. Recall (11.1) that 1
1
and 1
2
are said to be linearly
disjoint over k if the homomorphism
a
i
b
i
a
i
b
i
: 1
1
k
1
2
1
1
1
2
is injective.
LEMMA 16.28. Let k be algebraically closed elds, and let V be a variety over . If
there exist models of V over subelds 1
1
. 1
2
of nitely generated over k and linearly
disjoint over k, then there exists a model of V over k.
PROOF. The model of V over 1
1
extends to a model over an irreducible afne variety U
1
with k(U
1
) =1
1
, i.e., there exists a surjective map V
1
U
1
of kvarieties whose generic
bre is a model of V over 1
1
. A similar statement applies to the model over 1
2
. Because
1
1
and 1
2
are linearly disjoint, 1
1
k
1
2
is an integral domain with eld of fractions
k(U
1
U
2
). From the map V
1
U
1
, we get a map V
1
U
2
U
1
U
2
, and similarly for
V
2
.
Assume initially that V
1
U
2
and U
1
V
2
are isomorphic over U
1
U
2
, so that we have
a commutative diagram:
V
1
V
1
U
2
U
1
V
2
V
2
U
1
U
1
U
2
U
2
~
Let 1 be a point of U
1
. When we pull back the triangle to the subvariety 1 U
2
of U
1
U
2
,
we get the diagram at left below. Note that 1 U
2
.U
2
and that 1 .Spmk (because k is
algebraically closed).
V(1) U
2
1 V
2
V(1)
1
2
V
21
2
1 U
2
Spm(1
2
)
~ ~
Rigid descent 231
The generic bre of this diagram is the diagram at right. Here V
1
(1)
1
2
is the variety over
1
2
obtained from V
1
(1) by extension of scalars k 1
2
. As V
21
2
is a model V over 1
2
,
it follows that V
1
(1) is a model of V over k.
We nowprove the general case. The varieties (V
1
U
2
)
k(U
1
U
2
)
and (U
1
V
2
)
k(U
1
U
2
)
become isomorphic over some nite eld extension 1 of k(U
1
U
2
). Let
U be the normal
ization
1
of U
1
U
2
in 1, and let U be a dense open subset of
U such that some isomorphism
of (V
1
U
2
)
1
with (U
1
V
2
)
1
extends to an isomorphism over U. The goingup theorem
(8.8) shows that
U U
1
U
2
is surjective, and so the image U
t
of U in U
1
U
2
contains
a nonempty (hence dense) open subset of U
1
U
2
(see 10.2). In particular, U
t
contains a
subset 1 U
t
2
with U
t
2
a nonempty open subset of U
2
. Now the previous argument gives
us varieties V
1
(1)
1
2
and V
21
2
over 1
2
that become isomorphic over k(U
tt
) where U
tt
is
the inverse image of 1 U
t
2
in
U. As k(U
tt
) is a nite extension of 1
2
, this again shows
that V
1
(1) is a model of V over k.
2
EXAMPLE 16.29. Let 1 be an elliptic curve over with invariant (1). There exists
a model of 1 over a subeld 1 of if and only if (1) 1. If (1) is transcendental,
then any two such elds contain k((1)), and so cant be linearly disjoint. Therefore, the
hypothesis in the proposition implies (1) k, and so 1 has a model over k.
LEMMA 16.30. Let be algebraically closed of innite transcendence degree over k, and
assume that k is algebraically closed in . For any 1 nitely generated over k, there
exists a o Aut(,k) such that 1 and o1 are linearly disjoint over k.
PROOF. Let a
1
. . . . . a
n
be a transcendence basis for 1,k, and extend it to a transcendence
basis a
1
. . . . . a
n
. b
1
. . . . . b
n
. . . . of ,k. Let o be any permutation of the transcendence basis
such that o(a
i
) =b
i
for all i . Then o denes a kautomorphismof k(a
1
. . . . a
n
. b
1
. . . . . b
n
. . . .),
which we extend to an automorphism of .
Let 1
1
= k(a
1
. . . . . a
n
). Then o1
1
= k(b
1
. . . . . b
n
), and certainly 1
1
and o1
1
are
linearly disjoint. In particular, 1
1
k
o1
1
is an integral domain. Because k is algebraically
closed in 1, 1
k
o1 is an integral domain (cf. 11.5). This implies that 1 and o1 are
linearly disjoint.
2
LEMMA 16.31. Let k be algebraically closed elds such that is of innite tran
scendence degree over k, and let V be a variety over . If V is isomorphic to oV for every
o Aut(,k), then V has a model over k.
PROOF. There will exist a model V
0
of V over a subeld 1 of nitely generated over k.
According to Lemma 16.30, there exists a o Aut(,k) such that 1 and o1 are linearly
disjoint. Because V ~oV , oV
0
is a model of V over o1, and we can apply Lemma 16.28.
2
In the next two theorems, k are elds such that the xed eld of 1 =Aut(,k)
is k and is algebraically closed
THEOREM 16.32. Let V be a quasiprojective variety over , and let (
c
)
cT
be a descent
system for V . If the only automorphism of V is the identity map, then V has a model over
k splitting (
c
).
1
Let U
1
U
2
=SpmC; then
U =Spm
C, where
C is the integral closure of C in 1.
232 16. DESCENT THEORY
PROOF. According to Lemma 16.31, V has a model (V
0
. ) over the algebraic closure k
al
of k in , which (see the proof of 16.22) splits (
c
)
cAut({k
al
)
.
Now
t
c
def
=
1
c
o is stable under Aut(,k
al
), and hence is dened over k
al
(16.9). Moreover,
t
c
depends only on the restriction of o to k
al
, and (
t
c
)
cGal(k
al
{k)
is a
descent system for V
0
. It is continuous by (16.21), and so V
0
has a model (V
00
.
t
) over k
splitting (
t
c
)
cGal(k
al
{k)
. Now (V
00
.
t
) splits (
c
)
cAut({k)
.
2
We now consider pairs (V. S) where V is a variety over and S is a family of points
S =(1
i
)
1_i_n
of V indexed by 1. n. A morphism (V. (1
i
)
1_i_n
) (W. (Q
i
)
1_i_n
) is
a regular map : V W such that (1
i
) =Q
i
for all i .
THEOREM 16.33. Let V be a quasiprojective variety over , and let (
c
)
cAut({k)
be a
descent system for V . Let S =(1
i
)
1_i_n
be a nite set of points of V such that
(a) the only automorphism of V xing each 1
i
is the identity map, and
(b) there exists a subeld 1 of nitely generated over k such that
c
1 = 1 for all
o 1 xing 1.
Then V has a model over k splitting (
c
).
PROOF. Lemmas 16.2716.31 all hold for pairs (V. S) (with the same proofs), and so the
proof of Theorem 16.32 applies.
2
EXAMPLE 16.34. Theorem 16.33 can be used to prove that certain abelian varieties at
tached to algebraic varieties in characteristic zero, for example, the generalized Jacobian
varieties, are dened over the same eld as the variety.
2
We illustrate this with the usual
Jacobian variety J of a complete nonsingular curve C. For such a curve C over C, there is
a principally polarized abelian variety J(C) such that, as a complex manifold,
J(C)(C) =1(C.
1
)
,H
1
(C. Z).
The association C J(C) is a functorial, and so a descent datum (
c
)
cAut({k)
on C
denes a descent system on J(C). It is known that if we take S to be the set of points
of order 3 on J(C), then condition (a) of the theorem is satised (see, for example, Milne
1986
3
, 17.5), and condition (b) can be seen to be satised by regarding J(C) as the Picard
variety of C.
Weils descent theorems
THEOREM 16.35. Let k be a nite separable extension of a eld k
0
, and let 1 be the
set of khomomorphisms k k
al
0
. Let V be a quasiprojective variety over k; for each
pair (o. t) of elements of 1, let
r,c
be an isomorphism oV tV (of varieties over k
al
0
).
Then there exists a variety V
0
over k
0
and an isomorphism : V
0k
V such that
r,c
=
t (o)
1
for all o. t 1 if and only if the
r,c
are dened over k
sep
0
and satisfy the
following conditions:
(a)
r,p
=
r,c
c,p
for all j. o. t 1;
2
This was pointed out to me by Niranjan Ramachandran.
3
Milne, J.S., Abelian varieties, in Arithmetic Geometry, Springer, 1986.
Weils descent theorems 233
(b)
ro,co
=o
r,c
for all o. t 1 and all k
0
automorphisms o of k
al
0
over k
0
.
Moreover, when this is so, the pair (V
0
. ) is unique up to isomorphism over k
0
, and V
0
is
quasiprojective or quasiafne if V is.
PROOF. This is Theorem 3 of Weil 1956,
4
p515. It is essentially a restatement of (a) of
Corollary 16.25 (and (V
0
. ) is unique up to a unique isomorphism over k
0
).
2
An extension 1 of a eld k is said to be regular if it is nitely generated, admits a
separating transcendence basis, and k is algebraically closed in 1. These are precisely
the elds that arise as the eld of rational functions on geometrically irreducible algebraic
variety over k.
Let k be a eld, and let k(t ), t = (t
1
. . . . . t
n
). be a regular extension of k (in Weils
terminology, t is a generic point of a variety over k). By k(t
t
) we shall mean a eld
isomorphic to k(t ) by t t
t
, and we write k(t. t
t
) for the eld of fractions of k(t )
k
k(t
t
).
5
When V
t
is a variety over k(t ), we shall write V
t
0 for the variety over k(t
t
) obtained fromV
t
by base change with respect to t t
t
: k(t ) k(t
t
). Similarly, if
t
denotes a regular map
of varieties over k(t ), then
t
0 denotes the regular map over k(t
t
) obtained by base change.
Similarly, k(t
tt
) is a second eld isomorphic to k(t ) by t t
tt
and k(t. t
t
. t
tt
) is the eld of
fractions of k(t )
k
k(t
t
)
k
k(t
tt
).
THEOREM 16.36. With the above notations, let V
t
be a quasiprojective variety over k(t );
for each pair (t. t
t
), let
t
0
,t
be an isomorphism V
t
V
t
0 dened over k(t. t
t
). Then there
exists a variety V dened over k and an isomorphism
t
: V
k(t)
V
t
(of varieties over k(t ))
such that
t
0
,t
=
t
0
1
t
if and only if
t
0
,t
satises the following condition:
t
00
,t
=
t
00
,t
0
t
0
,t
(isomorphism of varieties over k(t. t
t
. t
tt
).
Moreover, when this is so, the pair (V.
t
) is unique up to an isomorphism over k, and V is
quasiprojective or quasiafne if V is.
PROOF. This is Theorem 6 and Theorem 7 of Weil 1956, p522.
2
THEOREM 16.37. Let be an algebraically closed eld of innite transcendence degree
over a perfect eld k. Then descent is effective for quasiprojective varieties over .
PROOF. Let (
c
) be a descent datum on a variety V over . Because (
c
) is continuous, it
is split by a model of V over some subeld 1 of nitely generated over k. Let k
t
be the
algebraic closure of k in 1; then k
t
is a nite extension of k and 1 is a regular extension
of k. Write 1 =k(t ), and let (V
t
.
t
) be a model of V over k(t ) splitting (
c
). According
to Lemma 16.30, there exists a o Aut(,k) such that ok(t ) =k(t
t
) and k(t ) are linearly
disjoint over k. The isomorphism
V
t
0
V
oV
(c
0
)
1
V
t
0
,
4
Weil, Andr e, The eld of denition of a variety. Amer. J. Math. 78 (1956), 509524.
5
If k(t ) and k(t
t
) are linearly disjoint subelds of some large eld , then k(t. t
t
) is the subeld of
generated over k by t and t
t
.
234 16. DESCENT THEORY
is dened over k(t. t
t
) and satises the conditions of Theorem 16.36. Therefore, there exists
a model (W. ) of V over k
t
splitting (
c
)
cAut({k(t)
.
For o. t Aut(,k), let
r,c
be the composite of the isomorphisms
oW
c
oV
tV
r
tW.
Then
r,c
is dened over the algebraic closure of k in and satises the conditions of
Theorem 16.35, which gives a model of W over k splitting (
c
)
cAut({k)
.
2
Restatement in terms of group actions
In this subsection, k are elds such that k =
T
and is algebraically closed. Recall
that for any variety V over k, there is a natural action of 1 on V(). In this subsection, we
describe the essential image of the functor
{quasiprojective varieties over k] {quasiprojective varieties over action of 1 ].
In other words, we determine which pairs (V. +). with V a quasiprojective variety over
and + an action of 1 on V(),
(o. 1) o +1: 1 V() V().
arise from a variety over k. There are two obvious necessary conditions for this.
Regularity condition
Obviously, the action should recognize that V() is not just a set, but rather the set of
points of an algebraic variety. For o 1 , let oV be the variety obtained by applying o to
the coefcients of the equations dening V , and for 1 V() let o1 be the point on oV
obtained by applying o to the coordinates of 1.
DEFINITION 16.38. We say that the action + is regular if the map
o1 o +1: (oV )() V()
is regular isomorphism for all o.
A priori, this is only a map of sets. The condition requires that it be induced by a regular
map
c
: oV V . If V =V
0
for some variety V
0
dened over k, then oV =V , and
c
is the identity map, and so the condition is clearly necessary.
REMARK 16.39. The maps
c
satisfy the cocycle condition
c
o
r
=
cr
. In particular,
c
o
c
1 = id, and so if + is regular, then each
c
is an isomorphism, and the family
(
c
)
cT
is a descent system. Conversely, if (
c
)
cT
is a descent system, then
o +1 =
c
(o1)
denes a regular action of 1 on V(). Note that if + (
c
), then o +1 =
c
1.
Restatement in terms of group actions 235
Continuity condition
DEFINITION 16.40. We say that the action + is continuous if there exists a subeld 1 of
nitely generated over k and a model V
0
of V over 1 such that the action of 1(,1) is
that dened by V
0
.
For an afne variety V , an action of 1 on V gives an action of 1 on V , and one
action is continuous if and only if the other is.
Continuity is obviously necessary. It is easy to write down regular actions that fail it,
and hence dont arise from varieties over k.
EXAMPLE 16.41. The following are examples of actions that fail the continuity condition
((b) and (c) are regular).
(a) Let V =A
1
and let + be the trivial action.
(b) Let ,k =
al
,, and let N be a normal subgroup of nite index in Gal(
al
,)
that is not open,
6
i.e., that xes no extension of of nite degree. Let V be the zero
dimensional variety over
al
with V(
al
) =Gal(
al
,),N with its natural action.
(c) Let k be a nite extension of
]
, and let V = A
1
. The homomorphism k
Gal(k
ab
,k) can be used to twist the natural action of 1 on V().
Restatement of the main theorems
Let k be elds such that k is the xed eld of 1 =Aut(,k) and is algebraically
closed.
THEOREM 16.42. Let V be a quasiprojective variety over , and let + be a regular action
of 1 on V(). Let S =(1
i
)
1_i_n
be a nite set of points of V such that
(a) the only automorphism of V xing each 1
i
is the identity map, and
(b) there exists a subeld 1 of nitely generated over k such that o +1 =1 for all
o 1 xing 1.
Then + arises from a model of V over k.
PROOF. This a restatement of Theorem 16.33.
2
THEOREM 16.43. Let V be a quasiprojective variety over with an action + of 1 . If +
is regular and continuous, then + arises from a model of V over k in each of the following
cases:
(a) is algebraic over k, or
(b) is has innite transcendence degree over k.
PROOF. Restatements of (16.23, 16.25) and of (16.37).
2
The condition quasiprojective is necessary, because otherwise the action may not
stabilize enough open afne subsets to cover V .
6
For a proof that such subgroups exist, see FT 7.25.
236 16. DESCENT THEORY
Faithfully at descent
Recall that a homomorphism : T of rings is at if the functor extension of scalars
M T
M
t
T
M T
M
tt
0
is exact. For a eld k, a homomorphism k is always at (because exact sequences of
kvector spaces are splitexact), and it is faithfully at if =0.
The next theorem and its proof are quintessential Grothendieck.
THEOREM 16.44. If : T is faithfully at, then the sequence
0
(
T
d
0
T
2
T
i
d
r1
T
i1
is exact, where
T
i
=T
T (r times)
J
i1
=
(1)
i
e
i
e
i
(b
0
b
i1
) =b
0
b
i1
1b
i
b
i1
.
PROOF. It is easily checked that J
i
J
i1
=0. We assume rst that admits a section,
i.e., that there is a homomorphism g: T such that g = 1, and we construct a
contracting homotopy k
i
: T
i2
T
i1
. Dene
k
i
(b
0
b
i1
) =g(b
0
)b
1
b
i1
. r _1.
It is easily checked that
k
i1
J
i1
J
i
k
i
=1. r _1,
and this shows that the sequence is exact.
Now let
t
be an algebra. Let T
t
=
t
T and let
t
= 1 :
t
T
t
. The
sequence corresponding to
t
is obtained from the sequence for by tensoring with
t
(because T
i
t
T
t(
etc.). Thus, if
t
is a faithfully at algebra, it sufces to
prove the theorem for
t
. Take
t
=T, and then b
(
b 1: T T
T has a section,
namely, g(b b
t
) =bb
t
, and so the sequence is exact.
2
THEOREM 16.45. If : T is faithfully at and M is an module, then the sequence
0 M
1(
M
T
1d
0
M
T
2
M
B
T
i
1d
r1
T
i1
is exact.
PROOF. As in the above proof, one may assume that has a section, and use it to construct
a contracting homotopy.
2
Faithfully at descent 237
REMARK 16.46. Let : T be a faithfully at homomorphism, and let M be an 
module. Write M
t
for the Tmodule
+
M = T
M. The module e
0+
M
t
= (T
T)
B
M
t
may be identied with T
M
t
where T
T acts by (b
1
b
2
)(b m) =
b
1
b b
2
m, and e
1+
M
t
may be identied with M
t
T where T
T acts by (b
1
b
2
)(mb) = b
1
mb
2
b. There is a canonical isomorphism : e
1+
M
t
e
0+
M
t
arising
from
e
1+
M
t
=(e
1
)
+
M =(e
0
)
+
M =e
0+
M
t
:
explicitly, it is the map
(b m) b
t
b (b
t
m): M
t
T T
M.
Moreover, M can be recovered from the pair (M
t
. ) because
M ={m M
t
[ 1m=(m1)].
Conversely, every pair (M
t
. ) satisfying certain obvious conditions does arise in this way
from an module. Given : M
t
T T
M
t
, dene
1
: T
M
t
T T
M
t
2
: M
t
T T
M
t
.
3
: M
t
T T
M
t
T
by tensoring with id
B
in the rst, second, and third positions respectively. Then a pair
(M
t
. ) arises from an module M as above if and only if
2
=
1
3
. The necessity is
easy to check. For the sufciency, dene
M ={m M
t
[ 1m=(m1)].
There is a canonical map b mbm: T
M M
t
, and it sufces to show that this is
an isomorphism (and that the map arising from M is ). Consider the diagram
M
t
T
1
1
T
M
t
T
1
T
M
t
e
0
1
e
1
1
T
M
t
in which (m) =1m and (m) =(m) 1. As the diagram commutes with either the
upper of the lower horizontal maps (for the lower maps, this uses the relation
2
=
1
3
),
induces an isomorphism on the kernels. But, by dention of M, the kernel of the pair
( 1. 1) is M
+
2
U
t
satisfying
+
31
() =
+
32
()
+
21
().
Here
}i
denotes the projection W W W W W such that
}i
(n
1
. n
2
. n
3
) =
(n
}
. n
i
).
PROOF. When W and V are afne, (16.46) gives a similar statement for modules, hence
for algebras, and hence for algebraic spaces.
2
EXAMPLE 16.48. Let 1 be a nite group, and regard it as an algebraic group of dimension
0. Let V be an algebraic space over k. An algebraic space Galois over V with Galois group
1 is a nite map W V to algebraic space together with a regular map W 1 W such
that
(a) for all kalgebras 1, W(1) 1(1) W(1) is an action of the group 1(1) on the
set W(1) in the usual sense, and the map W(1) V(1) is compatible with the
action of 1(1) on W(1) and its trivial action on V(1), and
(b) the map (n. o) (n. no): W 1 W
V
W is an isomorphism.
Then there is a commutative diagram
7
V W W 1



W 1
2
[[ [[ : :
V W W
V
W



W
V
W
V
W
The vertical isomorphisms are
(n. o) (n. no)
(n. o
1
. o
2
) (n. no
1
. no
1
o
2
).
Therefore, in this case, Theorem 16.47 says that to give an algebraic space afne over V
is the same as to give an algebraic space afne over W together with an action of 1 on it
compatible with that on W. When we take W and V to be the spectra of elds, then this
becomes afne case of Theorem 16.23.
EXAMPLE 16.49. In Theorem 16.47, let be the map corresponding to a regular extension
of elds k k(t ). This case of Theorem 16.47 coincides with the afne case of Theorem
16.36 except that the eld k(t. t
t
) has been replaced by the ring k(t )
k
k(t
t
).
NOTES. The paper of Weil cited in subsection on Weils descent theorems is the rst important
paper in descent theory. Its results havent been superseded by the many results of Grothendieck on
descent. In Milne 1999
8
, Theorem 16.33 was deduced from Weils theorems. The present more ele
mentary proof was suggested by Wolfarts elementary proof of the obvious part of Belyis theorem
(Wolfart 1997
9
; see also Derome 2003
10
).
7
See Milne, J. S., Etale cohomology. Princeton, 1980, p100.
8
Milne, J. S., Descent for Shimura varieties. Michigan Math. J. 46 (1999), no. 1, 203208.
9
Wolfart, J urgen. The obvious part of Belyis theorem and Riemann surfaces with many automorphisms.
Geometric Galois actions, 1, 97112, London Math. Soc. Lecture Note Ser., 242, Cambridge Univ. Press,
Cambridge, 1997.
10
Derome, G., Descente alg ebriquement close, J. Algebra, 266 (2003), 418426.
CHAPTER 17
Lefschetz Pencils (Outline)
In this chapter, we see how to bre a variety over P
1
in such a way that the bres have only
very simple singularities. This result sometimes allows one to prove theorems by induction
on the dimension of the variety. For example, Lefschetz initiated this approach in order to
study the cohomology of varieties over C.
Throughout this chapter, k is an algebraically closed eld.
Denition
A linear form H =
n
i=0
a
i
T
i
denes a hyperplane in P
n
, and two linear forms dene the
same hyperplane if and only if one is a nonzero multiple of the other. Thus the hyperplanes
in P
n
form a projective space, called the dual projective space
`
P
n
.
A line D in
`
P
n
is called a pencil of hyperplanes in P
n
. If H
0
and H
o
are any two
distinct hyperplanes in D, then the pencil consists of all hyperplanes of the form H
0
H
o
with (: ) P
1
(k). If 1 H
0
H
o
, then it lies on every hyperplane in the pencil
the axis of the pencil is dened to be the set of such 1. Thus
=H
0
H
o
=
tT
H
t
.
The axis of the pencil is a linear subvariety of codimension 2 in P
n
, and the hyperplanes of
the pencil are exactly those containing the axis. Through any point in P
n
not on , there
passes exactly one hyperplane in the pencil. Thus, one should imagine the hyperplanes in
the pencil as sweeping out P
n
as they rotate about the axis.
Let V be a nonsingular projective variety of dimension J _ 2, and embed V in some
projective space P
n
. By the square of an embedding, we mean the composite of V P
n
with the Veronese mapping (6.20)
(.
0
: . . . : .
n
) (.
2
0
: . . . : .
i
.
}
: . . . : .
2
n
): P
n
P
.mC2/.mC1/
2
.
DEFINITION 17.1. A line D in
`
P
n
is said to be a Lefschetz pencil for V P
n
if
(a) the axis of the pencil (H
t
)
tT
cuts V transversally;
(b) the hyperplane sections V
t
def
= V H
t
of V are nonsingular for all t in some open
dense subset U of D:
(c) for t U, V
t
has only a single singularity, and the singularity is an ordinary double
point.
239
240 17. LEFSCHETZ PENCILS (OUTLINE)
Condition (a) means that, for every point 1 V , Tgt
1
() Tgt
1
(V ) has codi
mension 2 in Tgt
1
(V ), the tangent space to V at 1.
Condition (b) means that, except for a nite number of t , H
t
cuts V transversally, i.e.,
for every point 1 H
t
V , Tgt
1
(H
t
) Tgt
1
(V ) has codimension 1 in Tgt
1
(V ).
A point 1 on a variety V of dimension J is an ordinary double point if the tangent
cone at 1 is isomorphic to the subvariety of A
d1
dened by a nondegenerate quadratic
form Q(T
1
. . . . . T
d1
), or, equivalently, if
O
V,1
~kT
1
. . . . . T
d1
,(Q(T
1
. . . . . T
d1
)).
THEOREM 17.2. There exists a Lefschetz pencil for V (after possibly replacing the pro
jective embedding of V by its square).
PROOF. (Sketch). Let W V
`
P
n
be the closed variety whose points are the pairs (.. H)
such that H contains the tangent space to V at .. For example, if V has codimension 1 in
P
n
, then (.. H) Y if and only if H is the tangent space at .. In general,
(.. H) W . H and H does not cut V transversally at ..
The image of W in
`
P
n
under the projection V
`
P
n
`
P
n
is called the dual variety
`
V of V .
The bre of W V over . consists of the hyperplanes containing the tangent space at .,
and these hyperplanes form an irreducible subvariety of
`
P
n
of dimension m(dimV 1);
it follows that W is irreducible, complete, and of dimension m1 (see 10.11) and that V
is irreducible, complete, and of codimension _ 1 in
`
P
n
(unless V = P
n
, in which case
it is empty). The map : W
`
V is unramied at (.. H) if and only if . is an ordinary
double point on V H (see SGA 7, XVII 3.7
1
). Either is generically unramied, or it
becomes so when the embedding is replaced by its square (so, instead of hyperplanes, we
are working with quadric hypersurfaces) (ibid. 3.7). We may assume this, and then (ibid.
3.5), one can show that for H
`
V
`
V
sing
, V H has only a single singularity and the
singularity is an ordinary double point. Here
`
V
sing
is the singular locus of
`
V .
By Bertinis theorem (Hartshorne 1977, II 8.18) there exists a hyperplane H
0
such that
H
0
V is irreducible and nonsingular. Since there is an (m1)dimensional space of lines
through H
0
, and at most an (m2)dimensional family will meet V
sing
, we can choose H
o
so that the line D joining H
0
and H
o
does not meet
`
V
sing
. Then D is a Lefschetz pencil
for V.
2
THEOREM 17.3. Let D = (H
t
) be a Lefschetz pencil for V with axis = H
t
. Then
there exists a variety V
+
and maps
V V
+
t
D.
such that:
(a) the map V
+
V is the blowing up of V along V :
1
Groupes de monodromie en g eom etrie alg ebrique. S eminaire de G eom etrie Alg ebrique du BoisMarie
19671969 (SGA 7). Dirig e par A. Grothendieck. Lecture Notes in Mathematics, Vol. 288, 340. Springer
Verlag, BerlinNew York, 1972, 1973.
Denition 241
(b) the bre of V
+
D over t is V
t
=V H
t
.
Moreover, is proper, at, and has a section.
PROOF. (Sketch) Through each point . of V V , there will be exactly one H
x
in D.
The map
: V V D, . H
x
.
is regular. Take the closure of its graph 1
,]
Z
=lim
U7y0
1(U. O
U
). The reverse functor is even easier: simply omit the
nonclosed points from the base space.
1
Every aspiring algebraic and (especially) arithmetic geometer needs to learn the basic
theory of schemes, and for this I recommend reading Chapters II and III of Hartshorne
1997.
1
Some authors call a geometrically reduced scheme of nitetype over a eld a variety. Despite their simi
larity, it is important to distinguish such schemes from varieties (in the sense of these notes). For example, if W
and W
t
are subvarieties of a variety, their intersection in the sense of schemes need not be reduced, and so may
differ from their intersection in the sense of varieties. For example, if W =V(a) A
n
and W
t
=V(a
t
) A
n
0
with a and a
t
radical, then the intersection W and W
t
in the sense of schemes is SpeckX
1
. . . . . X
nn
0 ,(a. a
t
)
while their intersection in the sense of varieties is SpeckX
1
. . . . . X
nn
0 ,rad(a. a
t
) (and their intersection in
the sense of algebraic spaces is SpmkX
1
. . . . . X
nn
0 ,(a. a
t
).
243
APPENDIX A
Solutions to the exercises
11 Use induction on n. For n = 1, use that a nonzero polynomial in one variable has
only nitely many roots (which follows from unique factorization, for example). Now
suppose n >1 and write =
g
i
X
i
n
with each g
i
kX
1
. . . . . X
n1
. If is not the zero
polynomial, then some g
i
is not the zero polynomial. Therefore, by induction, there exist
(a
1
. . . . . a
n1
) k
n1
such that (a
1
. . . . . a
n1
. X
n
) is not the zero polynomial. Now, by
the degreeone case, there exists a b such that (a
1
. . . . . a
n1
. b) =0.
12 (X 2Y. 7); Gaussian elimination (to reduce the matrix of coefcients to row echelon
form); (1), unless the characteristic of k is 2, in which case the ideal is (X 1. 71).
21 W =Y axis, and so 1(W) =(X). Clearly,
(X
2
. XY
2
) (X) rad(X
2
. XY
2
)
and rad((X)) =(X). On taking radicals, we nd that (X) =rad(X
2
. XY
2
).
22 The J J minors of a matrix are polynomials in the entries of the matrix, and the set
of matrices with rank _r is the set where all (r 1) (r 1) minors are zero.
23 Clearly V =V(X
n
X
n
1
. . . . . X
2
X
2
1
). The map
X
i
T
i
: kX
1
. . . . . X
n
 kT 
induces an isomorphism kV  A
1
. [Hence t (t. . . . . t
n
) is an isomorphism of afne
varieties A
1
V .]
24 We use that the prime ideals are in onetoone correspondence with the closed irre
ducible subsets 7 of A
2
. For such a set, 0 _dim7 _2.
Case dim7 =2. Then 7 =A
2
, and the corresponding ideal is (0).
Case dim7 =1. Then 7 =A
2
, and so 1(7) contains a nonzero polynomial (X. Y ).
If 1(7) =( ), then dim7 =0 by (2.25, 2.26). Hence 1(7) =( ).
Case dim7 =0. Then 7 is a point (a. b) (see 2.24c), and so 1(7) =(X a. Y b).
25 The statement Hom
kalgebras
(
k. T
1(V
i
. O
V
i
) to give a regular function on
V
i
is the same
as to give a regular function on each V
i
(this is the obvious ringed space structure).
Thus, if V is afne, it must equal Specm(
i
), where
i
= 1(V
i
. O
V
i
), and so V =
Specm(
i
) (use the description of the ideals in T on p13). Etc..
43 Let H be an algebraic subgroup of G. By denition, H is locally closed, i.e., open in
its Zariski closure
H. Assume rst that H is connected. Then
H is a connected algebraic
group, and it is a disjoint union of the cosets of H. It follows that H =
H. In the general
case, H is a nite disjoint union of its connected components; as one component is closed,
they all are.
51 (b) The singular points are the common solutions to
_
_
_
4X
3
2XY
2
=0 == X =0 or Y
2
=2X
2
4Y
3
2X
2
Y =0 == Y =0 or X
2
=2Y
2
X
4
Y
4
X
2
Y
2
=0.
Thus, only (0. 0) is singular, and the variety is its own tangent cone.
52 Directly from the denition of the tangent space, we have that
T
a
(V H) T
a
(V ) T
a
(H).
As
dimT
a
(V H) _dimV H =dimV 1 =dimT
a
(V ) T
a
(H).
we must have equalities everywhere, which proves that a is nonsingular on V H. (In
particular, it cant lie on more than one irreducible component.)
247
The surface Y
2
=X
2
7 is smooth, but its intersection with the XY plane is singular.
No, 1 neednt be singular on V H if H T
1
(V ) for example, we could have
H V or H could be the tangent line to a curve.
53 We can assume V and W to afne, say
1(V ) =a kX
1
. . . . . X
n

1(W) =b kX
n1
. . . . . X
nn
.
If a = (
1
. . . . .
i
) and b = (g
1
. . . . . g
x
), then 1(V W) = (
1
. . . . .
i
. g
1
. . . . . g
x
). Thus,
T
(a,b)
(V W) is dened by the equations
(J
1
)
a
=0. . . . . (J
i
)
a
=0. (Jg
1
)
b
=0. . . . . (Jg
x
)
b
=0.
which can obviously be identied with T
a
(V ) T
b
(W).
54 Take C to be the union of the coordinate axes in A
n
. (Of course, if you want C to be
irreducible, then this is more difcult. . . )
55 A matrix satises the equations
(1 c)
tr
J (1 c) =1
if and only if
tr
J J =0.
Such an is of the form
_
M N
1 Q
_
with M. N. 1. Q nnmatrices satisfying
N
tr
=N. 1
tr
=1. M
tr
=Q.
The dimension of the space of s is therefore
n(n1)
2
(for N)
n(n1)
2
(for 1) n
2
(for M. Q) =2n
2
n.
56 Let C be the curve Y
2
= X
3
, and consider the map A
1
C, t (t
2
. t
3
). The cor
responding map on rings kX. Y ,(Y
2
) kT  is not an isomorphism, but the map on the
geometric tangent cones is an isomorphism.
57 The singular locus V
sing
has codimension _ 2 in V , and this implies that V is normal.
[Idea of the proof: let k(V ) be integral over kV , kV , =g,h, g. h kV ;
for any 1 V(h) V(g), O
1
is not integrally closed, and so 1 is singular.]
58 No! Let a =(X
2
Y ). Then V(a) is the union of the X and Y axes, and 1V(a) =(XY ).
For a =(a. b),
(JX
2
Y )
a
=2ab(X a) a
2
(Y b)
(JXY )
a
=b(X a) a(Y b).
If a =0 and b =0, then the equations
(JX
2
Y )
a
=a
2
Y =0
(JXY )
a
=aY =0
248 A. SOLUTIONS TO THE EXERCISES
have the same solutions.
61 Let 1 =(a : b : c), and assume c =0. Then the tangent line at 1 =(
o
c
:
b
c
: 1) is
_
dJ
dX
_
1
X
_
dJ
dY
_
1
Y
__
dJ
dX
_
1
_
a
c
_
_
dJ
dY
_
1
_
b
c
__
7 =0.
Now use that, because J is homogeneous,
J(a. b. c) =0 ==
_
dJ
dX
_
1
a
_
dJ
dY
_
1
_
dJ
d7
_
1
c =0.
(This just says that the tangent plane at (a. b. c) to the afne cone J(X. Y. 7) = 0 passes
through the origin.) The point at ois (0 : 1 : 0), and the tangent line is 7 =0, the line at
o. [The line at omeets the cubic curve at only one point instead of the expected 3, and
so the line at otouches the curve, and the point at ois a point of inexion.]
62 The equation dening the conic must be irreducible (otherwise the conic is singular).
After a linear change of variables, the equation will be of the form X
2
Y
2
= 7
2
(this
is proved in calculus courses). The equation of the line in aX bY =c7, and the rest is
easy. [Note that this is a special case of Bezouts theorem (6.34) because the multiplicity is
2 in case (b).]
63 (a) The ring
kX. Y. 7,(Y X
2
. 7X
3
) =k.. ,. : =k. .kX.
which is an integral domain. Therefore, (Y X
2
. 7X
3
) is a radical ideal.
(b) The polynomial J =7XY =(7X
3
) X(Y X
2
) 1(V ) and J
+
=7W
XY . If
7W XY =(Y W X
2
) (7W
2
X
3
)g.
then, on equating terms of degree 2, we would nd
7W XY =a(Y W X
2
).
which is false.
64 Let 1 =(a
0
: . . . : a
n
) and Q=(b
0
: . . . : b
n
) be two points of P
n
, n _ 2. The condition
that the hyperplane 1
c
:
c
i
X
i
=0 pass through 1 and not through Q is that
a
i
c
i
=0.
b
i
c
i
=0.
The (n1)tuples (c
0
. . . . . c
n
) satisfying these conditions form a nonempty open subset of
the hyperplane H:
a
i
X
i
=0 in A
n1
. On applying this remark to the pairs (1
0
. 1
i
), we
nd that the (n1)tuples c =(c
0
. . . . . c
n
) such that 1
0
lies on the hyperplane 1
c
but not
1
1
. . . . . 1
i
form a nonempty open subset of H.
65 The subset
C ={(a : b : c) [ a =0. b =0] L{(1 : 0 : 0)]
of P
2
is not locally closed. Let 1 = (1 : 0 : 0). If the set C were locally closed, then 1
would have an open neighbourhood U in P
2
such that U C is closed. When we look in
U
0
, 1 becomes the origin, and
C U
0
=(A
2
{Xaxis]) L{origin].
249
The open neighbourhoods U of 1 are obtained by removing from A
2
a nite number of
curves not passing through 1. It is not possible to do this in such a way that U C is closed
in U (U C has dimension 2, and so it cant be a proper closed subset of U; we cant have
U C = U because any curve containing all nonzero points on Xaxis also contains the
origin).
72 Dene () = h(. Q) and g(n) = h(1. n), and let = h ( g q). Then
(. Q) = 0 = (1. n), and so the rigidity theorem (7.13) implies that is identically
zero.
66 Let
c
i}
X
i}
=0 be a hyperplane containing the image of the Segre map. We then have
c
i}
a
i
b
}
=0
for all a =(a
0
. . . . . a
n
) k
n1
and b =(b
0
. . . . . b
n
) k
n1
. In other words,
aCb
t
=0
for all a k
n1
and b k
n1
, where C is the matrix (c
i}
). This equation shows that
aC =0 for all a, and this implies that C =0.
82 For example, consider
(A
1
{1]) A
1
xx
n
A
1
for n > 1 an integer prime to the characteristic. The map is obviously quasinite, but it is
not nite because it corresponds to the map of kalgebras
X X
n
: kX kX. (X 1)
1

which is not nite (the elements 1,(X 1)
i
, i _1, are linearly independent over kX, and
so also over kX
n
).
83 Assume that V is separated, and consider two regular maps . g: 7 W. We have to
show that the set on which and g agree is closed in 7. The set where and g agree
is closed in 7, and it contains the set where and g agree. Replace 7 with the set where
and g agree. Let U be an open afne subset of V , and let 7
t
=( )
1
(U) =
( g)
1
(U). Then (7
t
) and g(7
t
) are contained in
1
(U), which is an open afne
subset of W, and is therefore separated. Hence, the subset of 7
t
on which and g agree is
closed. This proves the result.
[Note that the problem implies the following statement: if : W V is a nite regular
map and V is separated, then W is separated.]
84 Let V =A
n
, and let W be the subvariety of A
n
A
1
dened by the polynomial
n
i=1
(X T
i
) =0.
The bre over (t
1
. . . . . t
n
) A
n
is the set of roots of
(X t
i
). Thus, V
n
= A
n
; V
n1
is
the union of the linear subspaces dened by the equations
T
i
=T
}
. 1 _i. _n. i = :
V
n2
is the union of the linear subspaces dened by the equations
T
i
=T
}
=T
k
. 1 _i. . k _n. i. . k distinct,
250 A. SOLUTIONS TO THE EXERCISES
and so on.
101 Consider an orbit O =G. The map g g: G O is regular, and so O contains
an open subset U of
O (10.2). If u U, then gu gU, and gU is also a subset of O
which is open in
O (because 1 g1: V V is an isomorphism). Thus O, regarded as
a topological subspace of
O, contains an open neighbourhood of each of its points, and so
must be open in
O.
We have shown that O is locally closed in V , and so has the structure of a subvariety.
From (5.18), we know that it contains at least one nonsingular point 1. But then g1 is
nonsingular, and every point of O is of this form.
From set theory, it is clear that
O O is a union of orbits. Since
O O is a proper
closed subset of
O, all of its subvarieties must have dimension <dim
O =dimO.
Let O be an orbit of lowest dimension. The last statement implies that O =
O.
102 An orbit of type (a) is closed, because it is dened by the equations
Tr() =a. det() =b.
(as a subvariety of V ). It is of dimension 2, because the centralizer of
_
0
0
_
, =, is
__
+ 0
0 +
__
, which has dimension 2.
An orbit of type (b) is of dimension 2, but is not closed: it is dened by the equations
Tr() =a. det() =b. =
_
0
0
_
. =root of X
2
aX b.
An orbit of type (c) is closed of dimension 0: it is dened by the equation =
_
0
0
_
.
An orbit of type (b) contains an orbit of type (c) in its closure.
103 Let be a primitive Jth root of 1. Then, for each i. , 1 _ i. _ J, the following
equations dene lines on the surface
_
X
0
i
X
1
= 0
X
2
}
X
3
= 0
_
X
0
i
X
2
= 0
X
1
}
X
3
= 0
_
X
0
i
X
3
= 0
X
1
}
X
2
= 0.
There are three sets of lines, each with J
2
lines, for a total of 3J
2
lines.
104 (a) Compare the proof of Theorem 10.9.
(b) Use the transitivity, and apply Proposition 8.23.
121 Let H be a hyperplane in P
n
intersecting V transversally. Then H ~P
n1
and V H
is again dened by a polynomial of degree . Continuing in this fashion, we nd that
V H
1
. . . H
d
is isomorphic to a subset of P
1
dened by a polynomial of degree .
122 We may suppose that X is not a factor of J
n
, and then look only at the afne piece of
the blowup, o: A
2
A
2
, (.. ,) (.. .,). Then o
1
(C (0. 0))is given by equations
X =0. J(X. XY ) =0.
251
But
J(X. XY ) =X
n
(
(a
i
b
i
Y )
i
i
) X
n1
J
n1
(X. Y ) .
and so o
1
(C (0. 0)) is also given by equations
X =0.
(a
i
b
i
Y )
i
i
XJ
n1
(X. Y ) =0.
To nd its closure, drop the condition X = 0. It is now clear that the closure intersects
o
1
(0. 0) (the Y axis) at the s points Y =a
i
,b
i
.
123 We have to nd the dimension of kX. Y 
(A,Y )
,(Y
2
X
i
. Y
2
X
x
). In this ring,
X
i
= X
x
, and so X
x
(X
ix
1) = 0. As X
ix
1 is a unit in the ring, this implies that
X
x
=0, and it follows that Y
2
=0. Thus (Y
2
X
i
. Y
2
X
x
) (Y
2
. X
x
), and in fact the
two ideals are equal in kX. Y 
(A,Y )
. It is now clear that the dimension is 2s.
124 Note that
kV  =kT
2
. T
3
 =
_
a
i
T
i
[ a
i
=0
_
.
For each a k, dene an effective divisor D
o
on V as follows:
D
o
has local equation 1a
2
T
2
on the set where 1aT =0;
D
o
has local equation 1a
3
T
3
on the set where 1aT aT
2
=0.
The equations
(1aT )(1aT ) =1a
2
T
2
. (1aT )(1aT a
2
T
2
) =1a
3
T
3
show that the two divisors agree on the overlap where
(1aT )(1aT aT
2
) =0.
For a =0, D
o
is not principal, essentially because
gcd(1a
2
T
2
. 1a
3
T
3
) =(1aT ) kT
2
. T
3

if D
o
were principal, it would be a divisor of a regular function on V , and that regular
function would have to be 1aT , but this is not allowed.
In fact, one can show that Pic(V ) ~k. Let V
t
=V {(0. 0)], and write 1(+) for the
principal divisors on +. Then Div(V
t
) 1(V ) =Div(V ), and so
Div(V ),1(V ) .Div(V
t
),Div(V
t
) 1(V ) .1(V
t
),1(V
t
) 1(V ) .k.
APPENDIX B
Annotated Bibliography
Apart from Hartshorne 1977, among the books listed below, I especially recommend Sha
farevich 1994 it is very easy to read, and is generally more elementary than these notes,
but covers more ground (being much longer).
Commutative Algebra
Atiyah, M.F and MacDonald, I.G., Introduction to Commutative Algebra, AddisonWesley
1969. This is the most useful short text. It extracts the essence of a good part of Bourbaki
196183.
Bourbaki, N., Alg` ebre Commutative, Chap. 17, Hermann, 196165; Chap 89, Masson,
1983. Very clearly written, but it is a reference book, not a text book.
Eisenbud, D., Commutative Algebra, Springer, 1995. The emphasis is on motivation.
Matsumura, H., Commutative Ring Theory, Cambridge 1986. This is the most useful medium
length text (but read Atiyah and MacDonald or Reid rst).
Nagata, M., Local Rings, Wiley, 1962. Contains much important material, but it is concise to
the point of being almost unreadable.
Reid, M., Undergraduate Commutative Algebra, Cambridge 1995. According to the author, it
covers roughly the same material as Chapters 18 of Atiyah and MacDonald 1969, but is
cheaper, has more pictures, and is considerably more opinionated. (However, Chapters 10
and 11 of Atiyah and MacDonald 1969 contain crucial material.)
Serre: Alg` ebre Locale, Multiplicit es, Lecture Notes in Math. 11, Springer, 1957/58 (third
edition 1975).
Zariski, O., and Samuel, P., Commutative Algebra, Vol. I 1958, Vol II 1960, van Nostrand.
Very detailed and well organized.
Elementary Algebraic Geometry
Abhyankar, S., Algebraic Geometry for Scientists and Engineers, AMS, 1990. Mainly curves,
from a very explicit and downtoearth point of view.
Reid, M., Undergraduate Algebraic Geometry. A brief, elementary introduction. The 
nal chapter contains an interesting, but idiosyncratic, account of algebraic geometry in the
twentieth century.
Smith, Karen E.; Kahanp a a, Lauri; Kek al ainen, Pekka; Traves, William. An invitation to alge
braic geometry. Universitext. SpringerVerlag, New York, 2000. An introductory overview
with few proofs but many pictures.
Computational Algebraic Geometry
253
254 B. ANNOTATED BIBLIOGRAPHY
Cox, D., Little, J., OShea, D., Ideals, Varieties, and Algorithms, Springer, 1992. This gives
an algorithmic approach to algebraic geometry, which makes everything very downtoearth
and computational, but the cost is that the book doesnt get very far in 500pp.
Subvarieties of Projective Space
Harris, Joe: Algebraic Geometry: A rst course, Springer, 1992. The emphasis is on exam
ples.
Musili, C. Algebraic geometry for beginners. Texts and Readings in Mathematics, 20. Hin
dustan Book Agency, New Delhi, 2001.
Shafarevich, I., Basic Algebraic Geometry, Book 1, Springer, 1994. Very easy to read.
Algebraic Geometry over the Complex Numbers
Grifths, P., and Harris, J., Principles of Algebraic Geometry, Wiley, 1978. A comprehensive
study of subvarieties of complex projective space using heavily analytic methods.
Mumford, D., Algebraic Geometry I: Complex Projective Varieties. The approach is mainly
algebraic, but the complex topology is exploited at crucial points.
Shafarevich, I., Basic Algebraic Geometry, Book 3, Springer, 1994.
Abstract Algebraic Varieties
Dieudonn e, J., Cours de G eometrie Alg ebrique, 2, PUF, 1974. A brief introduction to abstract
algebraic varieties over algebraically closed elds.
Kempf, G., Algebraic Varieties, Cambridge, 1993. Similar approach to these notes, but is
more concisely written, and includes two sections on the cohomology of coherent sheaves.
Kunz, E., Introduction to Commutative Algebra and Algebraic Geometry, Birkh auser, 1985.
Similar approach to these notes, but includes more commutative algebra and has a long
chapter discussing how many equations it takes to describe an algebraic variety.
Mumford, D. Introduction to Algebraic Geometry, Harvard notes, 1966. Notes of a course.
Apart from the original treatise (Grothendieck and Dieudonn e 196067), this was the rst
place one could learn the new approach to algebraic geometry. The rst chapter is on
varieties, and last two on schemes.
Mumford, David: The Red Book of Varieties and Schemes, Lecture Notes in Math. 1358,
Springer, 1999. Reprint of Mumford 1966.
Schemes
Eisenbud, D., and Harris, J., Schemes: the language of modern algebraic geometry, Wadsworth,
1992. A brief elementary introduction to scheme theory.
Grothendieck, A., and Dieudonn e, J., El ements de G eom etrie Alg ebrique. Publ. Math. IHES
19601967. This was intended to cover everything in algebraic geometry in 13 massive
books, that is, it was supposed to do for algebraic geometry what Euclids Elements did
for geometry. Unlike the earlier Elements, it was abandoned after 4 books. It is an extremely
useful reference.
Hartshorne, R., Algebraic Geometry, Springer 1977. Chapters II and III give an excellent
account of scheme theory and cohomology, so good in fact, that no one seems willing to
write a competitor. The rst chapter on varieties is very sketchy.
Iitaka, S. Algebraic Geometry: an introduction to birational geometry of algebraic varieties,
Springer, 1982. Not as wellwritten as Hartshorne 1977, but it is more elementary, and it
covers some topics that Hartshorne doesnt.
Shafarevich, I., Basic Algebraic Geometry, Book 2, Springer, 1994. A brief introduction to
schemes and abstract varieties.
History
Dieudonn e, J., History of Algebraic Geometry, Wadsworth, 1985.
Of Historical Interest
255
Hodge, W., and Pedoe, D., Methods of Algebraic Geometry, Cambridge, 194754.
Lang, S., Introduction to Algebraic Geometry, Interscience, 1958. An introduction to Weil
1946.
Weil, A., Foundations of Algebraic Geometry, AMS, 1946; Revised edition 1962. This is
where Weil laid the foundations for his work on abelian varieties and jacobian varieties
over arbitrary elds, and his proof of the analogue of the Riemann hypothesis for curves
and abelian varieties. Unfortunately, not only does its language differ from the current
language of algebraic geometry, but it is incompatible with it.
Index
action
continuous, 224, 235
of a group on a vector space, 223
regular, 234
afne algebra, 186
algebra
nite, 11
nitely generated, 11
algebraic group, 81
algebraic space, 187
in the sense of Artin, 216
axiom
separation, 73
axis
of a pencil, 239
basic open subset, 48
Bezouts Theorem, 200
birationally equivalent, 86
category, 28
Chow group, 200
codimension, 155
complete intersection
idealtheoretic, 160
local, 161
settheoretic, 160
complex topology, 215
cone
afne over a set, 117
content of a polynomial, 17
continuous
descent system, 225
curve
elliptic, 39, 116, 120, 189, 207, 212
cusp, 93
cycle
algebraic, 199
degree
of a hypersurface, 134
of a map, 172, 198
of a point, 192
of a projective variety, 136
total, 18
derivation, 108
descent datum, 225
effective, 225
descent system, 224
Dicksons Lemma, 34
differential, 95
dimension, 85
Krull, 52
of a reducible set, 51
of an irreducible set, 50
pure, 51, 86
division algorithm, 32
divisor, 195
effective, 195
local equation for, 196
locally principal, 196
positive, 195
prime, 195
principal, 196
restriction of, 196
support of, 195
dual projective space, 239
dual variety, 240
element
integral over a ring, 19
irreducible, 16
equivalence of categories, 29
extension
of base eld, 187
of scalars, 187, 188
of the base eld, 188
bre
generic, 229
of a map, 147
eld
xed, 219
eld of rational functions, 50, 85
form
leading, 93
Frobenius map, 65
257
258 INDEX
function
rational, 59
regular, 47, 57, 72
functor, 29
contravariant, 29
essentially surjective, 29
fully faithful, 29
generate, 11
germ
of a function, 56
graph
of a regular map, 82
Groebner basis, see standard basis
group
symplectic, 113
homogeneous, 122
homomorphism
nite, 11
of algebras, 11
of presheaves, 181
of sheaves, 182
hypersurface, 51, 127
hypersurface section, 127
ideal, 11
generated by a subset, 11
homogeneous, 116
maximal, 12
monomial, 33
prime, 12
radical, 44
immersion, 75
closed, 75
open, 75
integral closure, 19
intersect properly, 197, 199
irreducible components, 50
isomorphic
locally, 111
leading coefcient, 32
leading monomial, 32
leading term, 32
Lemma
Gausss, 16
lemma
Nakayamas, 14
prime avoidance, 160
Yoneda, 30
Zariskis, 42
linearly equivalent, 196
local equation
for a divisor, 196
local ring
regular, 15
local system of parameters, 105
manifold
complex, 71
differentiable, 71
topological, 71
map
birational, 151
dominant, 67, 87
dominating, 67
etale, 97, 112
nite, 145
at, 198
proper, 152
quasinite, 147
Segre, 129
separable, 173
Veronese, 126
model, 188
module
of differential oneforms, 211
monomial, 18
Morita equivalent, 223
morphism
of afne algebraic varieties, 60
of functors, 30
of locally ringed spaces, 182
of ringed spaces, 60, 182
multidegree, 32
multiplicity
of a point, 93
neighbourhood
etale, 106
nilpotent, 44
node, 93
nondegenerate quadric, 177
nonsingular, 192
ordering
grevlex, 31
lex, 31
ordinary double point, 240
pencil, 239
Lefschetz, 239
pencil of lines, 178
perfect closure, 220
Picard group, 196, 205
Picard variety, 208
point
multiple, 95
nonsingular, 91, 95
Index 259
ordinary multiple, 93
rational over a eld, 191
singular, 95
smooth, 91, 95
with coordinates in a eld, 191
with coordinates in a ring, 87
polynomial
Hilbert, 135
homogeneous, 115
primitive, 17
presheaf, 181
prevariety, 187
algebraic, 71
separated, 73
principal open subset, 48
product
bred, 84
of algebraic varieties, 80
of objects, 77
tensor, 26
projection with centre, 129
projectively normal, 195
quasiinverse, 29
radical
of an ideal, 43
rationally equivalent, 200
regular eld extension, 184
regular map, 72
regulus, 178
resultant, 141
RiemannRoch Theorem, 212
ring
coordinate, 47
integrally closed, 20
noetherian, 14
normal, 104
of dual numbers, 108
reduced, 44
ringed space, 56, 182
locally, 182
section of a sheaf, 56
semisimple
group, 110
Lie algebra, 111
set
(projective) algebraic, 116
constructible, 165
sheaf, 182
coherent, 203
invertible, 205
locally free, 203
of abelian groups, 182
of algebras, 55
of kalgebras, 182
of rings, 182
support of, 204
singular locus, 92, 192
specialization, 230
splits
a descent system, 225
stalk, 182
standard basis, 35
minimal, 35
reduced, 35
subring, 11
subset
algebraic, 39
multiplicative, 22
subspace
locally closed, 75
subvariety, 75
closed, 65
open afne, 72
tangent cone, 93, 111
geometric, 93, 111, 112
tangent space, 91, 95, 101
theorem
Bezouts , 134
Chinese Remainder, 12
goingup, 147
Hilbert basis, 34, 40
Hilbert Nullstellensatz, 42
Krulls principal ideal, 158
Lefschetz pencils, 240
Lefschetz pencils exist, 240
Noether normalization, 149
Stein factorization, 180
strong Hilbert Nullstellensatz, 44
Zariskis main, 151
topological space
irreducible , 48
noetherian, 47
quasicompact, 47
topology
etale, 106
Krull, 226
Zariski, 41
variety, 187
abelian, 81, 143
afne algebraic, 60
algebraic, 73
complete, 137
ag, 133
Grassmann, 131
260 INDEX
normal, 105, 195
projective, 115
quasiprojective, 115
rational, 86
unirational, 86