Graduate Texts in Mathematics 52

Editorial Board
S. Axler F.W. Gehring K.A. Ribet
Graduate Texts in Mathematics
most recent titles in the GTM series

129 FULTON/HARRIS. Representation Theory: 159 CONWAY. Functions of One

A First Course. Complex Variable II.
Readings in Mathematics 160 LANG. Differential and Riemannian
130 DODSON/PoSTON. Tensor Geometry. Manifolds.
131 LAM. A First Course in Noncommutative 161 BORWEINIERDEL Yl. Polynomials and
Rings. Polynomial Inequalities.
132 BEARDON. Iteration of Rational Functions. 162 ALPERIN/BELL. Groups and
133 HARRIS. Algebraic Geometry: A First Representations.
Course. 163 DIXON/MORTIMER. Permutation
134 RoMAN. Coding and Information Theory. Groups.
135 ROMAN. Advanced Linear Algebra. 164 NATHANSON. Additive Number Theory:
136 ADKINS/WEINTRAUB. Algebra: An The Classical Bases.
Approach via Module Theory. 165 NATHANSON. Additive Number Theory:
137 AxLERIBouRDoN/RAMEY. Harmonic Inverse Problems and the Geometry of
Function Theory. Sumsets.
138 CoHEN. A Course in Computational 166 SHARPE. Differential Geometry: Cartan's
Algebraic Number Theory. Generalization of Klein's Erlangen
139 BREDON. Topology and Geometry. Program.
140 AUBIN. Optima and Equilibria. An 167 MORANDI. Field and Galois Theory.
Introduction to Nonlinear Analysis. 168 EWALD. Combinatorial Convexity and
141 BECKERIWEISPFENNING/KREDEL. Grtibner Algebraic Geometry.
Ba~es. A Computational Approach to 169 BHATIA. Matrix Analysis.
Commutative Algebra. 170 BREDON. Sheaf Theory. 2nd ed.
142 LANG. Real and Functional Analysis. 171 PETERSEN. Riemannian Geometry.
3rd ed. 172 REMMERT. Classical Topics in Complex
143 DOOB. Measure Theory. Function Theory.
144 DENNISIFARB. Noncommutative 173 DIESTEL. Graph Theory.
Algebra. 174 BRIDGES. Foundations of Real and
145 VICK. Homology Theory. An Abstract Analysis.
Introduction to Algebraic Topology. 175 LICKORISH. An Introduction to Knot
2nd ed. Theory.
146 BRIDGES. Computability: A 176 LEE. Riemannian Manifolds.
Mathematical Sketchbook. 177 NEWMAN. Analytic Number Theory.
147 ROSENBERG. Algebraic K-Theory 178 CLARKE/LEDYAEV/STERN/W OLENSKI.
and Its Applications. Nonsmooth Analysis and Control
148 ROTMAN. An Introduction to the Theory.
Theory of Groups. 4th ed. 179 DOUGLAS. Banach Algebra Techniques in
149 RATCLIFFE. Foundations of Operator Theory. 2nd ed.
Hyperbolic Manifolds. 180 SRIVASTAVA. A Course on Borel Sets.
150 EISENBUD. Commutative Algebra 181 KRESS. Numerical Analysis.
with a View Toward Algebraic 182 WALTER. Ordinary Differential
Geometry. Equations.
151 SILVERMAN. Advanced Topics in 183 MEGGINSON. An Introduction to Banach
the Arithmetic of Elliptic Curves. Space Theory.
152 ZIEGLER. Lectures on Polytopes. 184 BOLLOBAS. Modem Graph Theory.
153 FULTON. Algebraic Topology: A 185 COXILITILEIO'SHEA. Using Algebraic
First Course. Geometry.
154 BROWN/PEARCY. An Introduction to 186 RAMAKRISHNAN/V ALENZA. Fourier
Analysis. Analysis on Number Fields.
155 KASSEL. Quantum Groups. 187 HARRIS/MORRISON. Moduli of Curves.
156 KECHRIS. Cla~sical Descriptive Set 188 GOLDBLATT. Lectures on the Hyperreals:
Theory. An Introduction to Nonstandard Analysis.
157 MALLIAVIN. Integration and 189 LAM. Lectures on Modules and Rings.
Probability. 190 ESMONDEIMURTY. Problems in Algebraic
158 ROMAN. Field Theory. Number Theory.
Robin Hartshorne

Algebraic Geometry

~Springer
Robin Hartshorne
Department of Mathematics
University of California
Berkeley, California 94720
USA

Editorial Board
S. Axler K.A. Ribet
Mathematics Department Department of Mathematics
San Francisco State Universi~y University of California at Berkeley
San Francisco, CA 94132 Berkeley, CA 94720-3840
USA USA
ribet@ math. berkeley .edu

Mathematics Subject Classification (2000): 13-xx, 14Al0, 14A15, 14Fxx, 14Hxx, 14Jxx

Library of Congress Cataloging-in-Publication Data

Hartshorne, Robin.
Algebraic geometry.
(Graduate texts in mathematics: 52)
Bibliography: p.
Includes index.
1. Geometry, Algebraic. I. Title II. Series.
QA564.H25 516'.35 77-1177

ISBN 978-1-4419-2807-8 ISBN 978-1-4757-3849-0 (eBook) Printed on acid-free paper.

DOI 10.1007/978-1-4757-3849-0

© 1977 Springer Science+Business Media, Inc.

Softcover reprint of the hardcover I st edition 1977
All rights reserved. This work may not be translated or copied in whole or in part without the
written permission of the publisher (Springer Science+Business Media, Inc., 233 Spring Street, New
York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis.
Use in connection with any form of information storage and retrieval, electronic adaptation, com-
puter software, or by similar or dissimilar methodology now known or hereafter developed is for-
bidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if
they are not identified as such, is not to be taken as an expression of opinion as to whether or not
they are subject to proprietary rights.

(ASC/SBA)

15 14

[Link]
For Edie, Jonathan, and Berifamin
Preface

This book provides an introduction to abstract algebraic geometry using

the methods of schemes and cohomology. The main objects of study are
algebraic varieties in an affine or projective space over an algebraically
closed field; these are introduced in Chapter I, to establish a number of
basic concepts and examples. Then the methods of schemes and
cohomology are developed in Chapters II and III, with emphasis on appli-
cations rather than excessive generality. The last two chapters of the book
(IV and V) use these methods to study topics in the classical theory of
algebraic curves and surfaces.
The prerequisites for this approach to algebraic geometry are results
from commutative algebra, which are stated as needed, and some elemen-
tary topology. No complex analysis or differential geometry is necessary.
There are more than four hundred exercises throughout the book, offering
specific examples as well as more specialized topics not treated in the
main text. Three appendices present brief accounts of some areas of
current research.
This book can be used as a textbook for an introductory course in
algebraic geometry, following a basic graduate course in algebra. I re-
cently taught this material in a five-quarter sequence at Berkeley, with
roughly one chapter per quarter. Or one can use Chapter I alone for a
short course. A third possibility worth considering is to study Chapter I,
and then proceed directly to Chapter IV, picking up only a few definitions
from Chapters II and Ill, and assuming the statement of the Riemann-
Roch theorem for curves. This leads to interesting material quickly, and
may provide better motivation for tackling Chapters II and III later.
The material covered in this book should provide adequate preparation
for reading more advanced works such as Grothendieck [EGA], [SGA],
Hartshorne [5], Mumford [2], [5], or Shafarevich [1].

Vll
Preface

Acknowledgements
In writing this book, I have attempted to present what is essential for a
basic course in algebraic geometry. I wanted to make accessible to the
nonspecialist an area of mathematics whose results up to now have been
widely scattered, and linked only by unpublished "folklore." While I
have reorganized the material and rewritten proofs, the book is mostly a
synthesis of what I have learned from my teachers, my colleagues, and
my students. They have helped in ways too numerous to recount. I owe
especial thanks to Oscar Zariski, J.-P. Serre, David Mumford, and Arthur
Ogus for their support and encouragement.
Aside from the "classical" material, whose origins need a historian to
trace, my greatest intellectual debt is to A. Grothendieck, whose treatise
[EGA] is the authoritative reference for schemes and cohomology. His
results appear without specific attribution throughout Chapters II and III.
Otherwise I have tried to acknowledge sources whenever I was aware of
them.
In the course of writing this book, I have circulated preliminary ver-
sions of the manuscript to many people, and have received valuable
comments from them. To all of these people my thanks, and in particular
to J.-P. Serre, H. Matsumura, and Joe Lipman for their careful reading
and detailed suggestions.
I have taught courses at Harvard and Berkeley based on this material,
and I thank my students for their attention and their stimulating questions.
I thank Richard Bassein, who combined his talents as mathematician
and artist to produce the illustrations for this book.
A few words cannot adequately express the thanks I owe to my wife,
Edie Churchill Hartshorne. While I was engrossed in writing, she created
a warm home for me and our sons Jonathan and Benjamin, and through
her constant support and friendship provided an enriched human context
for my life.
For financial support during the preparation of this book, I thank the
Research Institute for Mathematical Sciences of Kyoto University, the
National Science Foundation, and the University of California at
Berkeley.

August 29, 1977

Berkeley, California ROBIN HARTSHORNE

viii
Contents

Introduction Xlll

CHAPTER I
Varieties 1
I Affine Varieties
2 Projective Varieties 8
3 Morphisms 14
4 Rational Maps 24
5 Nonsingular Varieties 31
6 Nonsingular Curves 39
7 Intersections in Projective Space 47
8 ·what Is Algebraic Geometry? 55

CHAPTER II
Schemes 60
Sheaves 60
2 Schemes 69
3 First Properties of Schemes 82
4 Separated and Proper Morphisms 95
5 Sheaves of Modules 108
6 Divisors 129
7 Projective Morphisms 149
8 Differentials 172
9 Formal Schemes 190

CHAPTER III
Cohomology 201
Derived Functors 202
2 Cohomology of Sheaves 206
3 Cohomology of a Noetherian Affine Scheme 213

ix
Contents

4 Cech Cohomology 218

5 The Cohomology of Projective Space 225
6 Ext Groups and Sheaves 233
7 The Serre Duality Theorem 239
8 Higher Direct Images of Sheaves 250
9 Flat Morphisms 253
10 Smooth Morphisms 268
11 The Theorem on Formal Functions 276
12 The Semicontinuity Theorem 281

CHAPTER IV
Curves 293
Riemann-Roch Theorem 294
2 Hurwitz's Theorem 299
3 Embeddings in Projective Space 307
4 Elliptic Curves 316
5 The Canonical Embedding 340
6 Classification of Curves in P" 349

CHAPTER V
Surfaces 356
1 Geometry on a Surface 357
2 Ruled Surfaces 369
3 Monoidal Transformations 386
4 The Cubic Surface in P" 395
5 Birational Transformations 409
6 Classification of Surfaces 421

APPENDIX A
Intersection Theory 424
I Intersection Theory 425
2 Properties of the Chow Ring 428
3 Chern Classes 429
4 The Riemann-Roch Theorem 431
5 Complements and Generalizations 434

APPENDIX B
Transcendental Methods 438
1 The Associated Complex Analytic Space 438
2 Comparison of the Algebraic and Analytic Categories 440
3 When is a Compact Complex Manifold Algebraic? 441
4 Kahler Manifolds 445
5 The Exponential Sequence 446

X
 Contents

APPENDIX C
The Weil Conjectures 449
I The Zeta Function and the Weil Conjectures 449
2 History of Work on the Weil Conjectures 451
3 The /-adic Cohomology 453
4 Cohomological Interpretation of the Weil Conjectures 454

Bibliography 459

Results from Algebra 470

Glossary of Notations 472

Index 478

XI
Introduction

The author of an introductory book on algebraic geometry has the difficult

task of providing geometrical insight and examples, while at the same
time developing the modem technical language of the subject. For in
algebraic geometry, a great gap appears to separate the intuitive ideas
which form the point of departure from the technical methods used in
current research.
The first question is that of language. Algebraic geometry has
developed in waves, each with its own language and point of view. The
late nineteenth century saw the function-theoretic approach of Riemann,
the more geometric approach of Brill and Noether, and the purely alge-
braic approach of Kronecker, Dedekind, and Weber. The Italian school
followed with Castelnuovo, Enriques, and Severi, culminating in the clas-
sification of algebraic surfaces. Then came the twentieth-century "'Ameri-
can" school of Chow, Wei!, and Zariski, which gave firm algebraic foun-
dations to the Italian intuition. Most recently, Serre and Grothendieck
initiated the French school, which has rewritten the foundations of alge-
braic geometry in terms of schemes and cohomology, and which has an
impressive record of solving old problems with new techniques. Each of
these schools has introduced new concepts and methods. In writing an
introductory book, is it better to use the older language which is closer to
the geometric intuition, or to start at once with the technical language of
current research?
The second question is a conceptual one. Modern mathematics tends to
obliterate history: each new school rewrites the foundations of its subject
in its own language, which makes for fine logic but poor pedagogy. Of
what use is it to know the definition of a scheme if one does not realize
that a ring of integers in an algebraic number field, an algebraic curve, and
a compact Riemann surface are all examples of a ''regular scheme of

xiii
Introduction

dimension one"? How then can the author of an introductory book indi-
cate the inputs to algebraic geometry coming from number theory, com-
mutative algebra, and complex analysis, and also introduce the reader to
the main objects of study, which are algebraic varieties in affine or pro-
jective space, while at the same time developing the modem language of
schemes and cohomology? What choice of topics will convey the meaning
of algebraic geometry, and still serve as a firm foundation for further study
and research?

My own bias is somewhat on the side of classical geometry. I believe

that the most important problems in algebraic geometry are those arising
from old-fashioned varieties in affine or projective spaces. They provide
the geometric intuition which motivates all further developments. In this
book, I begin with a chapter on varieties, to establish many examples and
basic ideas in their simplest form, uncluttered with technical details. Only
after that do I develop systematically the language of schemes, coherent
sheaves, and cohomology, in Chapters II and III. These chapters form the
technical heart of the book. In them I attempt to set forth the most
important results, but without striving for the utmost generality. Thus, for
example, the cohomology theory is developed only for quasi-coherent
sheaves on noetherian schemes, since this is simpler and sufficient for
most applications; the theorem of "coherence of direct image sheaves" is
proved only for projective morphisms, and not for arbitrary proper
morphisms. For the same reasons I do not include the more abstract
notions of representable functors, algebraic spaces, etale cohomology'
sites, and topoi.
The fourth and fifth chapters treat classical material, namely nonsingu-
lar projective curves and surfaces, but they use techniques of schemes
and cohomology. I hope these applications will justify the effort needed to
absorb all the technical apparatus in the two previous chapters.
As the basic language and logical foundation of algebraic geometry, I
have chosen to use commutative algebra. It has the advantage of being
precise. Also, by working over a base field of arbitrary characteristic,
which is necessary in any case for applications to number theory, one
gains new insight into the classical case of base field C. Some years ago,
when Zariski began to prepare a volume on algebraic geometry, he had to
clevelop the necessary algebra as he went. The task grew to such pro-
portions that he produced a book on commutative algebra only. Now we
are fortunate in having a number of excellent books on commutative
algebra: Atiyah-Macdonald [1], Bourbaki [1], Matsumura [2], Nagata [7],
and Zariski-Samuel [1]. My policy is to quote purely algebraic results as
needed, with references to the literature for proof. A list of the results
used appears at the end of the book.
Originally I had planned a whole series of appendices-short expos-
itory accounts of some current research topics, to form a bridge between
the main text of this book and the research literature. Because of limited

xiv
 Introduction

time and space only three survive. I can only express my regret at not
including the others, and refer the reader instead to the Arcata volume
(Hartshorne, ed. [1]) for a series of articles by experts in their fields,
intended for the nonspecialist. Also, for the historical development of
algebraic geometry let me refer to Dieudonne [1]. Since there was not
space to explore the relation of algebraic geometry to neighboring fields as
much as I would have liked, let me refer to the survey article of Cassels [1]
for connections with number theory, and to Shafarevich [2, Part III] for
connections with complex manifolds and topology.
Because I believe strongly in active learning, there are a great many
exercises in this book. Some contain important results not treated in the
main text. Others contain specific examples to illustrate general
phenomena. I believe that the study of particular examples is inseparable
from the development of general theories. The serious student should
attempt as many as possible of these exercises, but should not expect to
solve them immediately. Many will require a real creative effort to under-
stand. An asterisk denotes a more difficult exercise. Two asterisks denote
an unsolved problem.
See (1, §8) for a further introduction to algebraic geometry and this
book.
Terminology
For the most part, the terminology of this book agrees with generally
accepted usage, but there are a few exceptions worth noting. A variety is
always irreducible and is always over an algebraically closed field. In
Chapter I all varieties are quasi-projective. In (Ch. II, §4) the definition is
expanded to include abstract varieties, which are integral separated
schemes of finite type over an algebraically closed field. The words curve,
surface, and 3-fold are used to mean varieties of dimension 1, 2, and 3
respectively. But in Chapter IV, the word curve is used only for a nonsin-
gular projective curve; whereas in Chapter V a curve is any effective
divisor on a nonsingular projective surface. A surface in Chapter V is
always a nonsingular projective surface.
A scheme is what used to be called a prescheme in the first edition of
[EGA], but is called scheme in the new edition of [EGA, Ch. I].
The definitions of a projective morphism and a very ample invertible sheaf
in this book are not equivalent to those in [EGA]-see (II, §4, 5). They are
technically simpler, but have the disadvantage of not being local on the
base.
The word nonsingular applies only to varieties; for more general
schemes, the words regular and smooth are used.

Results from algebra

I assume the reader is familiar with basic results about rin~s. ideals,
modules, noetherian rings, and integral dependence, and is willing to ac-
cept or look up other results, belonging properly to commutative algebra

XV
Introduction

or homological algebra, which will be stated as needed, with references to

the literature. These results will be marked with an A: e.g., Theorem
3.9A, to distinguish them from results proved in the text.
The basic conventions are these: All rings are commutative with iden-
tity element I. All homomorphisms of rings take 1 to 1. In an integral
domain or a field, 0 -=I 1. A prime ideal (respectively, maximal ideal) is an
ideal p in a ring A such that the quotient ring A/p is an integral domain
(respectively, a field). Thus the ring itself is not cc'1sidered to be a prime
ideal or a maximal ideal.
A multiplicative system in a ring A is a subsetS, containing I, and closed
under multiplication. The localizationS ~ 1A is defined to be the ring formed
by equivalence classes of fractions a/s, a EA, s E S, wherea/s and a 'Is' are
said to be equivalent if there is an s" E S such that s"(s 'a -sa') = 0 (see
e.g. Atiyah-Macdonald [I, Ch. 3]). Two special cases which are used
constantly are the following. If p is a prime ideal in A, then S = A - p is a
multiplicative system, and the corresponding localization is denoted by
A,. Iff is an element of A, then S = {I} U {f" In~ I} is a multiplicative
system, and the corresponding localization is denoted by A 1 • (Note for
example that ifjis nilpotent, thenA 1 is the zero ring.)
References
Bibliographical references are given by author, with a number in square
brackets to indicate which work, e.g. Serre, [3, p. 75]. Cross references to
theorems, propositions, lemmas within the same chapter are given by
number in parentheses. e.g. (3.5). Reference to an exercise is given by
(Ex. 3.5). References to results in another chapter are preceded by the
chapter number, e.g. (II, 3.5), or (II, Ex. 3.5).

XVI
CHAPTER I

Varieties

Our purpose in this chapter is to give an introduction to algebraic geometry

with as little machinery as possible. We work over a fixed algebraically
closed field k. We define the main objects of study, which are algebraic
varieties in affine or projective space. We introduce some of the most
important concepts, such as dimension, regular functions, rational maps,
nonsingular varieties, and the degree of a projective variety. And most im-
portant, we give lots of specific examples, in the form of exercises at the end
of each section. The examples have been selected to illustrate many inter-
esting and important phenomena, beyond those mentioned in the text. The
person who studies these examples carefully will not only have a good under-
standing of the basic concepts of algebraic geometry, but he will also have
the background to appreciate some of the more abstract developments of
modern algebraic geometry, and he will have a resource against which to
check his intuition. We will continually refer back to this library of examples
in the rest of the book.
The last section of this chapter is a kind of second introduction to the book.
It contains a discussion of the "classification problem," which has motivated
much of the development of algebraic geometry. It also contains a discussion
of the degree of generality in which one should develop the foundations of
algebraic geometry, and as such provides motivation for the theory of
schemes.

1 Affine Varieties

Let k be a fixed algebraically closed field. We define affine n-space over k,

denoted Ai: or simply An, to be the set of all n-tuples of elements of k. An
element P E An will be called a point, and if P = (ab . .. ,an) with ai E k, then
the ai will be called the coordinates of P.
I Varieties

Let A = k[x 1 , . . . ,xn] be the polynomial ring in n variables over k.

We will interpret the elements of A as functions from the affine n-space
to k, by defining f(P) = f(a 1 , .•• ,an), where f E A and P E An. Thus if
f E A is a polynomial, we can talk about the set of zeros of f, namely
Z(f) = {P E Anlf(P) = 0}. More generally, if T is any subset of A, we
define the zero set of T to be the common zeros of all the elements of T,
namely
Z(T) = {P E Anlf(P) = 0 for all f E T}.

Clearly if a is the ideal of A generated by T, then Z(T) = Z(a). Further-

more, since A is a noetherian ring, any ideal a has a finite set of generators
f 1, . . . ,fr. Thus Z(T) can be expressed as the common zeros of the finite
set of polynomials f1> ... ,fr.

Definition. A subset Y of An is an algebraic set if there exists a subset T ~ A

such that Y = Z(T).

Proposition 1.1. The union of two algebraic sets is an algebraic set. The
intersection of any family of algebraic sets is an algebraic set. The empty
set and the whole space are algebraic sets.

PROOF. If Y1 = Z(T 1 ) and Y2 = Z(T 2 ), then Y1 u Y2 = Z(T 1 T 2 ), where

T 1 T 2 denotes the set of all products of an element of T 1 by an element of
T 2 . Indeed, if P E Y1 u Y2 , then either P E Y1 or P E Y2 , so P is a zero of
every polynomial in T 1 T 2 . Conversely, if P E Z(T 1 T 2 ), and P ¢; Y1 say,
then there is an f E T 1 such that f(P) # 0. Now for any g E T 2 , (fg)(P) = 0
implies that g(P) = 0, so that P E Y2 .
= Z(Ta.) is any family of algebraic sets, nZ(l),=andZ(UTa.),
n If~ then ~ so
also an algebraic set. Finally, the empty set 0
~is = the whole
space An = Z(O).

Definition. We define the Zariski topology on An by taking the open subsets

to be the complements of the algebraic sets. This is a topology, because
according to the proposition, the intersection of two open sets is open,
and the union of any family of open sets is open. Furthermore, the empty
set and the whole space are both open.

Example 1.1.1. Let us consider the Zariski topology on the affine line A1 .
Every ideal in A = k[ x] is principal, so every algebraic set is the set of zeros
of a single polynomial. Since k is algebraically closed, every nonzero poly-
nomial f(x) can be written f(x) = c(x - a 1) · · · (x - an) with c,a 1 , . . . ,an E
k. Then Z(f) = {a 1, . . . ,an}· Thus the algebraic sets in A 1 are just the finite
subsets (including the empty set) and the whole space (corresponding to
f = 0). Thus the open sets are the empty set and the complements of finite
subsets. Notice in particular that this topology is not Hausdorff.

2
 1 Affine Varieties

Definition. A nonempty subset Y of a topological space X is irreducible if

it cannot be expressed as the union Y = Y1 u Y2 of two proper subsets,
each one of which is closed in Y. The empty set is not considered to be
irreducible.

Example 1.1.2. A 1 is irreducible, because its only proper closed subsets are
finite, yet it is infinite (because k is algebraically closed, hence infinite).

Example 1.1.3. Any nonempty open subset of an irreducible space is irre-

ducible and dense.

Example 1.1.4. If Y is an irreducible subset of X, then its closure Y in X is

also irreducible.

Definition. An affine algebraic variety (or simply affine variety) is an irre-

ducible closed subset of An (with the induced topology). An open subset
of an affine variety is a quasi-affine variety.

These affine and quasi-affine varieties are our first objects of study. But
before we can go further, in fact before we can even give any interesting
examples, we need to explore the relationship between subsets of A" and
ideals in A more deeply. So for any subset Y c:; A", let us define the ideal of
Yin A by
I(Y) = {f E Alf(P)_ = 0 for all P E Y}.
Now we have a function Z which maps subsets of A to algebraic sets, and a
function I which maps subsets of A" to ideals. Their properties are sum-
marized in the following proposition.

Proposition 1.2.
(a) If T 1 c:; T 2 are subsets of A, then Z(T 1 ) ::::2 Z(T 2 ).
(b) If ¥1 c:; ¥2 are subsets of An, then I(Yd ::::2 I(¥2 ).
(c) For any two subsets ¥1 , ¥2 of A", we have I(¥1 u Y2 ) = I(Y1 ) n I(¥2 ).
(d) For any ideal a c:; A, I(Z(a)) = JO., the radical of a.
(e) For any subset Y c:; A", Z(J(Y)) = Y, the closure of Y.

PROOF. (a), (b) and (c) are obvious. (d) is a direct consequence of Hilbert's
Nullstellensatz, stated below, since the radical of a is defined as

JO. = {f E Alf' E a for some r > 0}.

To prove (e), we note that Y c:; Z(J(Y) ), which is a closed set, so clearly
Y c:; Z(I(Y) ). On the other hand, let W be any closed set containing Y.
Then W = Z(a) for some ideal a. So Z(a) ::::2 Y, and by (b), IZ(a) c:; I(Y).
But certainly a c:; IZ(a), so by (a) we have W = Z(a) ::::2 ZI(Y). Thus
ZI(Y) = Y.

3
I Vanetles

Theorem 1.3A (Hilbert's Nullstellensatz). Let k be an algebraically closed

field, let a be an ideal in A = k[ x b . . . ,x,], and let f E A be a polynomial
which vanishes at all points of Z(a). Then rEa for some integer r > 0.
PROOF. Llng [2, p. 256] or Atiyah-Macdonald [1, p. 85] or Zariski-Samuel
[1. vol. 2, p. 164].

Corollary 1.4. There is a one-to-one inclusion-reversing correspondence

between algebraic sets in A" and radical ideals (i.e., ideals which are equal
to their own radical) in A, given by Y f-> /(Y) and a f-> Z(a). Furthermore,
an algebraic set is irreducible if and only if its ideal is a prime ideal.
PROOF. Only the last part is new. If Y is irreducible, we show that J(Y) is
prime. Indeed, if fg E l(Y), then Y s Z(fg) = Z(f) u Z(g). Thus Y =
( Y n Z(.j')) u ( Y n Z(g) ), both being closed subsets of Y. Since Y is irre-
ducible, we have either Y = Y n Z(f), in which case Y s Z(.f), or Y s
Z(y). Hence either f E l(Y) or g E l(Y).
Conversely, let p be a prime ideal, and suppose that Z(p) = Y1 u Y2 .
Then p = /(Y1 ) n /(Y2 ), so either p = l(YJ or p = /(Y2 ). Thus Z(p) = Y1
or Y2 , hence it is irreducible.

Example 1.4.1. A" is irreducible, since it corresponds to the zero ideal in A,

which is prime.

Example 1.4.2. Let f be an irreducible polynomial in A = k[x,y]. Then f

generates a prime ideal in A, since A is a unique factorization domain, so
the zero set Y = Z(f) is irreducible. We call it the affine curve defined by
the equationf(x,y) = 0. Iff has degree d, we say that Y is a curve of degree d.

Example 1.4.3. More generally, iff is an irreducible polynomial in A =

k[x 1 , . . . ,x,], we obtain an affine variety Y = Z(f), which is called a surface
if n = 3, or a hyperswface if n > 3.

Example 1.4.4. A maximal ideal m of A = k[ x 1 , . . . ,x,] corresponds to

a minimal irreducible closed subset of A", which must be a point, say
P = (ab ... ,a,). This shows that every maximal ideal of A is of the form
m = (x 1 - a 1 , . . . ,x, - an), for some a 1 , . . . ,a, E k.

Example 1.4.5. If k is not algebraically closed, these results do not hold. For
example, if k = R, the curve x 2 + y 2 + 1 = 0 in Ai has no points. So (1.2d)
is false. See also (Ex. 1.12).

Definition. If Y s A" is an affine algebraic set, we define the affine coordinate

riny A(Y) of Y, to be A/l(Y).

Remark 1.4.6. If Y is an affine variety, then A(Y) is an integral domain.

Furthermore, A( Y) is a finitely generated k-algebra. Conversely, any

4
 1 Affine Varieties

finitely generated k-algebra B which is a domain is the affine coordinate

ring of some affine variety. Indeed, write Bas the quotient of a polynomial
ring A = k[x 1 , . . . ,xn] by an ideal a, and let Y = Z(a).

Next we will study the topology of our varieties. To do so we introduce

an important class of topological spaces which includes all varieties.

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

scending chain condition for closed subsets: for any sequence Y1 ::::::> Y2 ::::::> •••
of closed subsets, there is an integer r such that Y,. = Y,.+ 1 = ...

Example 1.4.7. An is a noetherian topological space. Indeed, if Y1 ::::::> Y2 ::::::> •••

is a descending chain of closed subsets, then I(Y1 ) s;; I(Y2 ) s;; ... is an as-
cending chain of ideals in A = k[x 1 , . . . ,xnJ. Since A is a noetherian ring,
this chain of ideals is eventually stationary. But for each i, Y; = Z(J( Y;) ),
so the chain Y; is also stationary.

Proposition 1.5. In a noetherian topological space X, every nonempty closed

subset Y can be expressed as a finite union Y = Y1 u ... u Y,. of irreducible
closed subsets Y;. If we require that Y; ~ lj for i # j, then the Y; are
uniquely determined. They are called the irreducible components of Y.
PROOF. First we show the existence of such a representation of Y. Let 6
be the set of nonempty closed subsets of X which cannot be written as a
finite union of irreducible closed subsets. If 6 is nonempty, then since X
is noetherian, it must contain a minimal element, say Y. Then Y is not
irreducible, by construction of 6. Thus we can write Y = Y' u Y", where
Y' and Y" are proper closed subsets of Y. By minimality of Y, each of Y'
and Y" can be expressed as a finite union of closed irreducible subsets, hence
Y also, which is a contradiction. We conclude that every closed set Y can
be written as a union Y = Y1 u ... u Y,. of irreducible subsets. By throwing
away a few if necessary, we may assume Y; ~ 1j for i # j.
Now suppose Y = Y~ u ... u Y~ is another such representation. Then
Y~ s;; Y = Y1 u ... u Y,., so Y~ = U(Y~ n Y;). But Y~ is irreducible, so
Y~ s;; Y; for some i, say i = 1. Similarly, Y1 s;; Yj for some j. Then Y~ s;; Yj,
so j = 1, and we find that Y1 = Y~. Now let Z = (Y - ¥1 )-. Then Z =
Y2 u ... u Y,. and also Z = Y2 u ... u Y~. So proceeding by induction on
r, we obtain the uniqueness of the r;.

Corollary 1.6. Every algebraic set in An can be expressed uniquely as a union

of varieties, no one containing another.

Definition. If X is a topological space, we define the dimension of X (denoted

dim X) to be the supremum of all integers n such that there exists a chain
Z 0 c Z 1 c ... c Zn of distinct irreducible closed subsets of X. We
define the dimension of an affine or quasi-affine variety to be its dimen-
sion as a topological space.

5
I Varieties

Example 1.6.1. The dimension of A 1 is 1. Indeed, the only irreducible closed

subsets of A 1 are the whole space and single points.

Definition. In a ring A, the height of a prime ideal p is the supremum of all

integers n such that there exists a chain p 0 c p 1 c ... c Pn = p of
distinct prime ideals. We define the dimension (or Krull dimension) of A
to be the supremum of the heights of all prime ideals.

Proposition 1.7. If Y is an affine algebraic set, then the dimension of Y is

equal to the dimension of its affine coordinate ring A( Y).
PROOF. If Y is an affine algebraic set in An, then the closed irreducible subsets
of Y correspond to prime ideals of A = k[x~o ... ,xn] containing I(Y).
These in turn correspond to prime ideals of A( Y). Hence dim Y is the length
of the longest chain of prime ideals in A( Y), which is its dimension.

This proposition allows us to apply results from the dimension theory of

noetherian rings to algebraic geometry.

Theorem [Link]. Let k be a field, and let B be an integral domain which is a

finitely generated k-algebra. Then:
(a) the dimension of B is equal to the transcendence degree of the
quotient field K(B) of B over k;
(b) For any prime ideal pin B, we have
height p + dim B/p = dim B.
PROOF. Matsumura [2, Ch. 5, §14] or, in the case k is algebraically closed,
Atiyah-Macdonald [ 1, Ch. 11 J

Proposition 1.9. The dimension of An is n.

PROOF. According to (1.7) this says that the dimension of the polynomial
ring k[ x ~o ... ,xn] is n, which follows from part (a) of the theorem.

Proposition 1.10. If Y is a quasi-affine variety, then dim Y = dim Y.

PROOF. If Zo c z1 c ... c zn is a sequence of distinct closed irreducible
subsets of Y, then Z 0 c Z 1 c ... c Zn is a sequence of distinct closed
irreducible subsets of Y (1.1.4), so we have dim Y ~ dim Y. In particular,
dim Y is finite, so we can choose a maximal such chain Z 0 c ... c Zn,
with n = dim Y. In that case Z 0 must be a point P, and the chain P =
Z 0 c ... c Zn will also be maximal (1.1.3). Now P corresponds to a
maximal ideal m of the affine coordinate ring A( f) of Y. The Z; correspond
to prime ideals contained in m, so height m = n. On the other hand, since
Pis a point in affine space, A(f)/m ~ k (1.4.4). Hence by (1.8Ab) we find
that n = dim A( f) = dim Y. Thus dim Y = dim Y.

6
 1 Affine Varieties

Theorem [Link] (Krull's Hauptidealsatz). Let A be a noetherian ring, and let

f E A be an element which is neither a zero divisor nor a unit. Then every
minimal prime ideal p containing f has height 1.

PROOF. Atiyah-Macdonald [1, p. 122].

Proposition 1.12A. A noetherian integral domain A is a unique factorization

domain if and only if every prime ideal of height 1 is principal.
PROOF. Matsumura [2, p. 141], or Bourbaki [1, Ch. 7, §3].

Proposition 1.13. A variety Yin A" has dimension n - 1 if and only if it is

the zero set Z(f) of a single nonconstant irreducible polynomial in A =
k[ x 1 , . . . ,xnJ.
PROOF. Iff is an irreducible polynomial, we have already seen that Z(f) is
a variety. Its ideal is the prime ideal p = (f). By (1.11A), p has height 1,
so by (1.8A), Z(f) has dimension n - 1. Conversely, a variety of dimension
n - 1 corresponds to a prime ideal of height 1. Now the polynomial ring A
is a unique factorization domain, so by (1.12A), p is principal, necessarily
generated by an irreducible polynomial f Hence Y = Z(f).

Remark 1.13.1. A prime ideal of height 2 in a polynomial ring cannot

necessarily be generated by two elements (Ex. 1.11 ).

EXERCISES

1.1. (a) Let Y be the plane curve y = x 2 (i.e., Y is the zero set of the polynomial f =
y - x 2 ). Show that A(Y) is isomorphic to a polynomial ring in one variable
over k.
(b) Let Z be the plane curve xy = 1. Show that A(Z) is not isomorphic to a poly-
nomial ring in one variable over k.
*(c) Let f be any irreducible quadratic polynomial in k[ x,y], and let W be the
conic defined by f Show that A(W) is isomorphic to A(Y) or A(Z). Which one
is it when?
1.2. The Twisted Cubic Curve. Let Y <::::: A 3 be the set Y = { (t,t 2 ,t 3 Jlt E k}. Show that Y
is an affine variety of dimension 1. Find generators for the ideal J(Y). Show that
A(Y) is isomorphic to a polynomial ring in one variable over k. We say that Y
is given by the parametric representation x = t, y = t 2 , z = t 3 •
1.3. Let Y be the algebraic set in A3 defined by the two polynomials x 2 - yz and
xz - x. Show that Y is a union of three irreducible components. Describe them
and find their prime ideals.
1.4. If we identify A2 with A 1 x A1 in the natural way, show that the Zariski topology
on A2 is not the product topology ofthe Zariski topologies on the two copies of A 1 .

7
I Varieties

1.5. Show that a k-algebra B is isomorphic to the affine coordinate ring of some alge-
braic set in A", for some n, if and only if B is a finitely generated k-algebra with no
nilpotent elements.
1.6. Any nonempty open subset of an irreducible topological space is dense and
irreducible. If Y is a subset of a topological space X, which is irreducible in its
induced topology, then the closure Y is also irreducible.
1.7. (a) Show that the following conditions are equivalent for a topological space X:
(i) X is noetherian; (ii) every nonempty family of closed subsets has a minimal
element; (iii) X satisfies the ascending chain condition for open subsets;
(iv) every nonempty family of open subsets has a maximal element.
(b) A noetherian topological space is quasi-compact, i.e., every open cover has a
finite subcover.
(c) Any subset of a noetherian topological space is noetherian in its induced
topology.
(d) A noetherian space which is also Hausdorff must be a finite set with the discrete
topology.
1.8. Let Y be an affine variety of dimension r in A". Let H be a hypersurface in A",
and assume that Y <;/;. H. Then every irreducible component of Y n H has
dimension r - 1. (See (7.1) for a generalization.)
1.9. Let a£; A = k[x1o ... ,xnJ be an ideal which can be generated by r elements.
Then every irreducible component of Z(a) has dimension ;::, n - r.
1.10. (a) IfYisanysubsetofatopologicalsp aceX,thendim Y ~dim X.
(b) If X is a topological space which is covered by a family of open subsets {U;},
then dim X = sup dim U;.
(c) Give an example of a topological space X and a dense open subset U with
dim U <dim X.
(d) If Y is a closed subset of an irreducible finite-dimensional topological space X,
and if dim Y = dim X, then Y = X.
(e) Give an example of a noetherian topological space of infinite dimension.
*1.11. Let Y £; A 3 be the curve given parametrically by x = t 3 , y = t 4 , z = t 5 . Show
that J(Y) is a prime ideal of height 2 in k[x,y,z] which cannot be generated by
2 elements. We say Y is not a local complete intersection-d. (Ex. 2.17).
1.12. Give an example of an irreducible polynomial fER[ x, y], whose zero set
Z(f) in A~ is not irreducible (cf. 1.4.2).

2 Projective Varieties

To define projective varieties, we proceed in a manner analogous to the

definition of affine varieties, except that we work in projective space.
Let k be our fixed algebraically closed field. We defined projective n-space
over k, denoted Pi:, or simply pn, to be the set of equivalence classes of
(n + 1)-tuples (a 0 , .. . ,an) of elements of k, not all zero, under the equiva-
lence relation given by (a 0 , . . . ,an) ~ (.lca 0 , . . . ,.lean) for all A E k, A =f. 0.
Another way of saying this is that pn as a set is the quotient of the set
8
 2 Projective Varieties

An+ 1 - {(0, ... ,0)} under the equivalence relation which identifies points
lying on the same line through the origin.
An element of pn is called a_point. If P is a point, then any (n + 1)-
tuple (a 0 , . . . ,an) in the equivalence class P is called a set of homogeneous
coordinates for P.
LetS be the polynomial ring k[x 0 , . . . ,xnJ. We want to regardS as a
graded ring, so we recall briefly the notion of a graded ring.
A graded ring is a ring S, together with a decomposition S = EBd;.o Sd
of S into a direct sum of abelian groups Sd, such that for any d,e ): 0,
Sd · Se c;:: Sd+e· An element of Sd is called a homogeneous element of degree
d. Thus any element of S can be written uniquely as a (finite) sum of
homogeneous elements. An ideal a c;:: S is a homogeneous ideal if a =
EBd;.o (an Sd). We will need a few basic facts about homogeneous ideals
(see, for example, Matsumura [2, §10] or Zariski-Samuel [1, vol. 2, Ch. VII,
§2]). An ideal is homogeneous if and only if it can be generated by homo-
geneous elements. The sum, product, intersection, and radical of homo-
geneous ideals are homogeneous. To test whether a homogeneous ideal is
prime, it is sufficient to show for any two homogeneous elements f,g, that
fg E a implies f E a or g E a.
We make the polynomial ring S = k[x 0 , . . . ,xn] into a graded ring by
taking Sd to be the set of all linear combinations of monomials of total
weight d in x 0 , . . . ,xn- Iff E S is a polynomial, we cannot use it to define
a function on pn, because of the nonuniqueness of the homogeneous co-
ordinates. However, if f is a homogeneous polynomial of degree d, then
f(Aa 0 , . .. ,Aan) = Adf(a 0 , . .. ,an), so that the property off being zero or
not depends only on the equivalence class of (a 0 , . . . ,an). Thus f gives a
function from pn to {0,1} by f(P) = 0 if f(a 0 , . .. ,an) = 0, and f(P) = 1
if f(a 0 , . . . ,an) =f. 0.
Thus we can talk about the zeros of a homogeneous polynomial, namely
Z(f) = {P E pnlf(P) = 0}. If Tis any set of homogeneous elements of S,
we define the zero set of T to be
Z(T) = {P E pnlf(P) = 0 for all f E T}.
If a is a homogeneous ideal of S, we define Z(a) = Z(T), where Tis the set
of all homogeneous elements in a. Since S is a noetherian ring, any set of
homogeneous elements T has a finite subset f 1 , . •• ,f,. such that Z(T) =
Z(f1 , . . . ,f,.).

Definition. A subset Y of pn is an algebraic set if there exists a set T of ho-

mogeneous elements of S such that Y = Z(T).

Proposition 2.1. The union of two algebraic sets is an algebraic set. The
intersection of any family of algebraic sets is an algebraic set. The empty
set and the whole space are algebraic sets.
PROOF. Left to reader (it is similar to the proof of (1.1) above).
9
1 Varieties

Definition. We define the Zariski topology on pn by taking the open sets

to be the complements of algebraic sets.

Once we have a topological space, the notions of irreducible subset and

the dimension of a subset, which were defined in §1, will apply.

Definition. A projective algebraic variety (or simply projective variety) is an

irreducible algebraic set in pn, with the induced topology. An open
subset of a projective variety is a quasi-projective variety. The dimension
of a projective or quasi-projective variety is its dimension as a topo-
logical space.
If Y is any subset of pn, we define the homogeneous ideal of Y in S,
denoted J(Y), to be the ideal generated by {f E Slf is homogeneous and
f(P) = 0 for all P E Y}. If Y is an algebraic set, we define the homo-
geneous coordinate ring of Y to be S(Y) = Sjl(Y). We refer to (Ex. 2.1 ~
2.7) below for various properties of algebraic sets in projective space
and their homogeneous ideals.

Our next objective is to show that projective n-space has an open covering
by affine n-spaces, and hence that every projective (respectively, quasi-
projective) variety has an open covering by affine (respectively, quasi-affine)
varieties. First we introduce some notation.
If f E S is a linear homogeneous polynomial, then the zero set of f is
called a hyperplane. In particular we denote the zero set of X; by H;, for
i ;= 0, ... ,n. Let U; be the open set pn- H;. Then pn is covered by the
open sets U;, because if P = (a 0 , • •• ,an) is a point, then at least one a; -:f= 0,
hence PE U;. We define a mapping <p;: U;----*An as follows: if P=(a 0 , . . . ,an) E
U;, then <p;(P) = Q, where Q is the point with affine coordinates

with a;/a; omitted. Note that <p; is well-defined since the ratios a)a; are
independent of the choice of homogeneous coordinates.

Proposition 2.2. The map <p; is a homeomorphism of U; with its induced

topology to An with its Zariski topology.

PROOF. <p; is clearly bijective, so it will be sufficient to show that the closed
sets of U; are identified with the closed sets of An by <fJ;· We may assume
i = 0, and we write simply U for U 0 and <p: U----* An for <p 0 .
Let A = k[Yl, . .. ,ynJ. We define a map a from the set Sh of homo-
geneous elements of S to A, and a map f3 from A to Sh. Given f E S\ we
set a(f) = f(l,Yl, . .. ,yn). On the other hand, given g E A of degree e, then

10
 2 Projective Varieties

•.• ,x"jx 0 ) is a homogeneous polynomial of degree e in the x;,

x 0g(xdx 0 ,
which we call f3(g).
Now let Y <;; U be a closed subset. Let Y be its closure in P". This is
an algebraic set, so Y = Z(T) for some subset T <;; Sh. Let T' = cx(T).
Then straightforward checking shows that cp(Y) = Z(T'). Conversely, let
W be a closed subset of A". Then W = Z(T') for some subset T' of A, and
one checks easily that cp - 1(W) = Z(f3(T')) n U. Thus cp and cp - l are both
closed maps, so cp is a homeomorphism.
Corollary 2.3. If Y is a projective (respectively, quasi-projective) variety, then
Y is covered by the open sets Y n U;, i = 0, ... ,n, which are homeomorphic
to affine (respectively, quasi-affine) varieties via the mapping qJ; defined
above.

EXERCISES

2.1. Prove the "homogeneous Nullstellensatz," which says if a <;; S is a homogeneous

ideal, and iff E Sis a homogeneous polynomial with deg f > 0, such that f(P) = 0
for all P E Z( a) in P", then fq E a for some q > 0. [Hint: Interpret the problem in
terms of the affine (n + 1)-space whose affine coordinate ring is S, and use the
usual Nullstellensatz, (1.3A).]
2.2. For a homogeneous ideal a <;; S, show that the following conditions are equi-
valent:
(i) Z(a) = 0 (the empty set);
(ii) .jO. = either s or the ideals+ = ffid>O Sd;
(iii) a ;::> Sd for some d > 0.

2.3. (a) If T 1 <;; T 2 are subsets of Sh, then Z(Td ;::> Z(T2 ).
(b) If Y1 <;; Y2 are subsets ofP", then J(Y1 ) ;::> J(Y2 ).
(c) For any two subsets Y1 ,Y2 ofP", J(Y1 u Y2 ) = J(Ytl n J(Y2 ).
(d) If a <;; Sis a homogeneous ideal with Z(a) #- 0, then J(Z(a)) = .jO..
(e) For any subset Y <;; P", Z(J(Y)) = Y.

2.4. (a) There is a 1-1 inclusion-reversing correspondence between algebraic sets in

P", and homogeneous radical ideals of S not equal to S+, given by Y H J(Y)
and a H Z( a). Nate: Since S + does not occur in this correspondence, it is
sometimes called the irrelevant maximal ideal of S.
(b) An algebraic set Y <;; P" is irreducible if and only if J(Y) is a prime ideal.
(c) Show that P" itself is irreducible.

2.5. (a) P" is a noetherian topological space.

(b) Every algebraic set in P" can be written uniquely as a finite union of irreducible
algebraic sets, no one containing another. These are called its irreducible
components.

2.6. If Y is a projective variety with homogeneous coordinate ring S(Y), show that
dim S(Y) = dim Y + 1. [Hint: Let cp;: U;--> A" be the homeomorphism of(2.2),
let Y; be the affine variety cp;(Y n U;), and let A(Y;) be its affine coordinate ring.

11
I Varieties

Show that A( Y;) can be identified with the sub ring of elements of degree 0 of the
localized ring S(YJx,· Then show that S(Y)x, ~ A(Y;)[x;,X;- 1 ]. Now use (1.7),
(1.8A), and (Ex 1.10), and look at transcendence degrees. Conclude also that
dim Y = dim Y; whenever Y; is nonempty.J
2.7. (a) dim P" = n.
(b) If Y s; P" is a quasi-projective variety, then dim Y = dim f.
[Hint: Use (Ex. 2.6) to reduce to (1.10).]

2.8. A projective variety Y s; P" has dimension n - 1 if and only if it is the zero set of
a single irreducible homogeneous polynomial f of positive degree. Y is called a
hypersurface in P".
2.9. Projective Closure of an Affine Variety. If Y s; A" is an affine variety, we identify
A" with an open set U 0 s; P" by the homeomorphism <p 0 . Then we can speak of
Y, the closure of Yin P", which is called the projective closure of Y.
(a) Show that J(Y) is the ideal generated by f3(I(Y)), using the notation of the
proof of (2.2).
(b) Let Y s; A 3 be the twisted cubic of (Ex. 1.2). Its projective closure Y s; P 3
is called the twisted cubic curve in P 3 . Find generators for J(Y) and J(Y), and
use this example to show that if f 1 , ..• ,f.. generate J(Y), then f3(fd, ... ,{3(!,.)
do not necessarily generate J(Y).
2.10. The Cone Over a Projective Variety (Fig. 1). Let Y s; P" be a nonempty algebraic
set, and let e:A"+ 1 - {(0, ... ,0)}--+ P" be the map which sends the point with
affine coordinates (a 0 , .•• ,an) to the point with homogeneous coordinates
(a 0 , . .. ,anl· We define the affine cone over Y to be

C(Y) = e- 1 (Y) u {(0, ... ,0)}.

(a) Show that C(Y) is an algebraic set in A"+l, whose ideal is equal to J(Y),
considered as an ordinary ideal in k[ x 0 , . . . ,xnJ.
(b) C(Y) is irreducible if and only if Y is.
(c) dim C(Y) = dim Y + 1.
Sometimes we consider the projective closure C( Y) of C( Y) in P" + 1 . This is called
the projective cone over Y.

Figure 1. The cone over a curve in P 2 .

12
 2 Projective Varieties

2.11. Linear Varieties in P". A hypersurface defined by a linear polynomial is called a

hyperplane.
(a) Show that the following two conditions are equivalent for a variety Yin P":
(i) J(Y) can be generated by linear polynomials.
(ii) Y can be written as an intersection of hyperplanes.
In this case we say that Y is a linear variety in P".
(b) If Y is a linear variety of dimension r in P", show that J(Y) is minimally gen-
erated by n - r linear polynomials.
(c) Let Y,Z be linear varieties in P", with dim Y = r, dimZ = s. Ifr + s - n )' 0,
then Y n Z =f. 0. Furthermore, if Y n Z =f. 0, then Y n Z is a linear
variety of dimension )' r + s - n. (Think of A"+ 1 as a vector space over k,
and work with its subspaces.)

2.12. The d-Uple Embedding. For given n,d > 0, let M 0 ,M 1 , . . . ,MN be all the mono-
mials of degree d in the n + 1 variables x 0 , . •. ,x", where N = ("!d) - 1. We
define a mapping Pd: P" --+ pN by sending the point P = (a 0, ... ,a") to the point
pAP) = (M 0 (a), ... ,MN(a)) obtained by substituting the a, in the monomials Mi.
This is called the d-uple embedding ofP" in pN_ For example, ifn = 1, d = 2, then
N = 2, and the image Y of the 2-uple embedding ofP 1 in P 2 is a conic.
(a) Let 8: k[y 0 , . . . ,yNJ --+ k[ x 0 , . . . ,x"J be the homomorphism defined by
sending y, to M,, and let a be the kernel of 8. Then a is a homogeneous prime
ideal, and so Z( a) is a projective variety in pN_
(b) Show that the image of Pd is exactly Z(a). (One inclusion is easy. The other will
require some calculation.)
(c) Now show that Pd is a homeomorphism ofP" onto the projective variety Z(a).
(d) Show that the twisted cubic curve in P 3 (Ex. 2.9) is equal to the 3-uple embed-
ding of P 1 in P 3 , for suitable choice of coordinates.
2.13. Let Y be the image of the 2-uple embedding of P 2 in P 5 . This is the Veronese
surface. If Z £; Y is a closed curve (a curve is a variety of dimension 1), show that
there exists a hypersurface V £; P 5 such that V n Y = Z.
2.14. The Segre Embedding. Let 1/J:P' x P'--+ pN be the map defined by sending the
ordered pair (a 0 , . .. ,a,) x (b 0 , . .. ,b,) to (... ,a,bi, . .. ) in lexicographic order,
where N = rs + r + s. Note that 1/J is well-defined and injective. It is called the
Segre embedding. Show that the image ofljJ is a subvariety ofPN. [Hint: Let the
homogeneous coordinates of pN be {ziili = 0, ... ,r, j = 0, ... ,s}, and let a be
the kernel of the homomorphism k[{zii}] --+ k[x 0 , . . . ,x,y0 , . . . ,y,] which sends
zii to x,yi. Then show that Im 1/1 = Z(a).]
2.15. The Quadric Surface in P 3 (Fig. 2). Consider the surface Q (a swface is a variety of
dimension 2) in P 3 defined by the equation xy - zw = 0.
(a) Show that Q is equal to the Segre embedding of P 1 x P 1 in P 3 , for suitable
choice of coordinates.
(b) Show that Q contains two families of lines (a line is a linear variety of dimen-
sion 1) {L,},{M,}, each parametrized by t E P 1, with the properties that if
L, =f. Lu, then L, n Lu = 0; if M, =f. Mu, M, n Mu = 0. and for all t,u,
L, n Mu = one point.
(c) Show that Q contains other curves besides these lines, and deduce that the
Zariski topology on Q is not homeomorphic via 1/J to the product topology on
P 1 x P 1 (where each P 1 has its Zariski topology).

13
I Varieties

Figure 2. The quadric surface in P 3 .

2.16. (a) The intersection of two varieties need not be a variety. For example, let Q1
and Q2 be the quadric surfaces in P 3 given by the equations x 2 - yw = 0
and x y - zw = 0, respectively. Show that Q1 n Q2 is the union of a twisted
cubic curve and a line.
(b) Even if the intersection of two varieties is a variety, the ideal of the inter-
section may not be the sum of the ideals. For example, let C be the conic in
P 2 given by the equation x 2 - yz = 0. Let L be the line given by y = 0.
Show that C n L consists of one point P, but that J( C) + J(L) f= J(P).
2.17. Complete intersections. A variety Y of dimension r in P" is a (strict) complete
intersection if J(Y) can be generated by n - r elements. Y is a set-theoretic com-
plete intersection if Y can be written as the intersection of n - r hypersurfaces.
(a) Let Y be a variety in P", let Y = Z(a); and suppose that a can be generated
by q elements. Then show that dim Y ;, n - q.
(b) Show that a strict complete intersection is a set-theoretic complete inter-
section.
*(c) The converse of (b) is false. For example let Y be the twisted cubic curve in
P 3 (Ex. 2.9). Show that J(Y) cannot be generated by two elements. On the
other hand, find hypersurfaces H l>H 2 of degrees 2,3 respectively, such that
Y = H 1 n H2 .
**(d) It is an unsolved problem whether every closed irreducible curve in P 3 is
a set-theoretic intersection of two surfaces. See Hartshorne [1 J and Hart-
shorne [5, III, §5] for commentary.

3 Morphisms

So far we have defined affine and projective varieties, but we have not dis-
cussed what mappings are allowed between them. We have not even said
when two are isomorphic. In this section we will discuss the regular func-
tions on a variety, and then define a morphism of varieties. Thus we will
have a good category in which to work.

14
 3 Morphisms

Let Y be a quasi-affine variety in An. We will consider functions f from

Yto k.

Definition. A function f: Y --+ k is regular at a point P E Y if there is an open

neighborhood UwithPE Us; Y,andpolynomialsg,hEA = k[x 1 , . . . ,xn],
such that h is nowhere zero on U, and f = gjh on U. (Here of course we
interpret the polynomials as functions on An, hence on Y.) We say that
f is regular on Y if it is regular at every point of Y.

Lemma 3.1. A regular function is continuous, when k is identified with Al

in its Zariski topology.

PROOF. It is enough to show that f- 1 of a closed set is closed. A closed set

of Al is a finite set of points, so it is sufficient to show that f- 1 (a) =
{P E Ylf(P) = a} is closed for any a E k. This can be checked locally: a
subset Z of a topological space Y is closed if and only if Y can be covered
by open subsets U such that Z n U is closed in U for each U. So let U be
an open set on which f can be represented as gjh, with g,h E A, and h no-
where 0 on U. Thenf- 1(a) n U = {P E Ulg(P)/h(P) =a}. But g(P)/h(P) =
a if and only if (g - ah)(P) = 0. So f- 1(a) n U = Z(g - ah) n U which
is closed. Hence f- 1 (a) is closed in Y.

Now let us consider a quasi-projective variety Y s; pn_

Definition. A function f: Y --+ k is regular at a point P E Y if there is an open

neighborhood U with P E U s; Y, and homogeneous polynomials
g,h E S = k[ x 0 , . . . ,xn], of the same degree, such that h is nowhere zero
on U, and f = gjh on U. (Note that in this case, even though g and h
are not functions on pn, their quotient is a well-defined function whenever
h i= 0, since they are homogeneous of the same degree.) We say that
f is regular on Y if it is regular at every point.

Remark 3.1.1. As in the quasi-affine case, a regular function is necessarily

continuous (proof left to reader). An important consequence of this is the
fact that iff and g are regular functions on a variety X, and iff = g on
some nonempty open subset U s; X, then f = g everywhere. Indeed, the
set of points where f - g = 0 is closed and dense, hence equal to X.

Now we can define the category of varieties.

Definition. Let k be a fixed algebraically closed field. A variety over k (or

simply variety) is any affine, quasi-affine, projective, or quasi-projective
variety as defined above. If X, Yare two varieties, a morphism cp: X --+ Y

15
I Varieties

is a continuous map such that for every open set V ~ Y, and for every
regular functionf: V--+ k, the function! o cp:cp- 1 (V)--+ k is regular.

Clearly the composition of two morphisms is a morphism, so we have a

category. In particular, we have the notion of isomorphism: an isomorphism
cp: X --+ Y of two varieties is a morphism which admits an inverse morphism
ljJ: Y --+ X with ljJ o cp = idx and cp o 1jJ = idy. Note that an isomorphism is
necessarily bijective and bicontinuous, but a bijective bicontinuous mor-
phism need not be an isomorphism (Ex. 3.2).

Now we introduce some rings of functions associated with any variety.

Definition. Let Y be a variety. We denote by @(Y) the ring of all regular

functions on Y. If P is a point of Y, we define the local ring of P on Y,
@P.Y (or simply @p) to be the ring of germs of regular functions on Y
near P. In other words, an element of @p is a pair <U,f) where U is an
open subset of Y containing P, and f is a regular function on U, and
where we identify two such pairs (U,f) and (V,g) iff = g on U n V.
(Use (3.1.1) to verify that this is an equivalence relation!)

Note that @P is indeed a local ring: its maximal ideal m is the set of germs
of regular functions which vanish at P. For if f(P) =1= 0, then 1/f is regular
in some neighborhood of P. The residue field @pjm is isomorphic to k.

Definition. If Yis a variety, we define the function field K(Y) of Y as follows:

an element of K(Y) is an equivalence class of pairs (U,f) where U is a
nonempty open subset of Y, f is a regular function on U, and where
we identify two pairs (U,f) and (V,g) iff= g on U n V. The elements
of K ( Y) are called rational functions on Y.

Note that K(Y) is in fact a field. Since Y is irreducible, any two non-
empty open sets have a nonempty intersection. Hence we can define addition
and multiplication in K~ Y), making it a ring. Then if <U,.f) E K( Y) with
f =I= 0, we can restrict f to the open set V = U - U n Z(f) where it never
vanishes, so that 1/f is regular on V, hence (V,1/f) is an inverse for (U,f).
Now we have defined, for any variety Y, the ring of global functions @( Y),
the local ring @P at a point of Y, and the function field K( Y). By restricting
functions we obtain natural maps @(Y)--+ @p --+ K(Y) which in fact are
injective by (3.1.1). Hence we will usually treat @(Y) and @pas subrings of
K(Y).
If we replace Y by an isomorphic variety, then the corresponding rings are
isomorphic. Thus we can say that @(Y), @p, and K(Y) are invariants of the
variety Y (and the point P) up to isomorphism.
Our next task is to relate @( Y), (l] p, and K ( Y) to the affine coordinate
ring A(Y) of an affine variety, and the homogeneous coordinate ring S(Y)

16
 3 Morphisms

of a projective variety, which were introduced earlier. We will find that for
an affine variety Y, A( Y) = (()( Y), so it is an invariant up to isomorphism.
However, for a projective variety Y, S( Y) is not an invariant: it depends on
the embedding of Y in projective space (Ex. 3.9).

Theorem 3.2. Let Y c:; An be an affine variety with affine coordinate ring
A(Y). Then:
(a) @(Y) ~ A(Y);
(b) for each point P E Y, let mp c:; A(Y) be the ideal of functions
vanishing at P. Then P 1--> mp gives a 1-1 correspondence between the
points of Y and the maximal ideals of A( Y);
(c) for each P, @p ~ A(Y)"'P' and dim @p = dim Y;
(d) K(Y) is isomorphic to the quotient field of A(Y), and hence K(Y)
is [Link] generated extension field of k, of transcendence degree = dim Y.

PROOF. We will proceed in several steps. First we define a map a:A(Y)-->

(()( Y). Every polynomial f E A = k[ x 1, ... ,xn] defines a regular function
on An and hence on Y. Thus we have a homomorphism A --> @(Y). Its
kernel is just l(Y), so we obtain an injective homomorphism a: A( Y) --> @(Y).
From (1.4) we know there is a 1-1 correspondence between points of Y
(which are the minimal algebraic subsets of Y) and maximal ideals of A
containing J(Y). Passing to the quotient by l(Y), these correspond to the
maximal ideals of A(Y). Furthermore, using a to identify elements of A(Y)
with regular functions on Y, the maximal ideal corresponding to P is just
mp = {f E A(Y)if(P) = 0}. This proves (b).
For each P there is a natural map A( Y)mp --> @p. It is injective because a
is injective, and it is surjective by definition of a regular function! This
shows that (OP ~ A(Y)mp· Now dim (()P =height mp. Since A(Y)/mp ~ k,
we conclude from (1.7) and ([Link]) that dim @p = dim Y.
From (c) it follows that the quotient field of A(Y) is isomorphic to the
quotient field of (OP for every P, and this is equal to K(Y), because every
rational function is actually in some @p. Now A(Y) is a finitely generated
k-algebra, so K(Y) is a finitely generated field extension of k. Furthermore,
the transcendence degree of K(Y)/k is equal to dim Y by (1.7) and ([Link]).
This proves (d).
To prove (a) we note that (O(Y) c:; nPEY @p, where all our rings are re-
garded as subrings of K ( Y).
Using (b) and (c) we have

A(Y) c:; @(Y) c:; n A(Y)"''

"'
where m runs over all the maximal ideals of A(Y). The equality now follows
from the simple algebraic fact that if B is an integral domain, then B is
equal to the intersection (inside its quotient field) of its localizations at all
maximal ideals.

17
I Varieties

Propositi' , 3.3. Let Ui s P" be the open set defined by the equation xi #- 0.
Then tne mapping cpi: Ui --+ A" of (2.2) above is an isomorphism of varieties.

PROOF. We have already shown that it is a homeomorphism, so we need

only check that the regular functions are the same on any open set. On Ui
the regular functions are locally quotients of homogeneous polynomials in
x 0 , . .. ,x" of the same degree. On A" the regular functions are locally
quotients of polynomials in y 1 , . . . ,Yn· One can check easily that these two
concepts are identified by the maps ex and f3 of the proof of (2.2).

Before stating the next result, we introduce some notation. If S is a

graded ring, and p a homogeneous prime ideal in S, then we denote by S<Pl
the subring of elements of degree 0 in the localization of S with respect to
the multiplicative subset T consisting of the homogeneous elements of S
not in p. Note that T- 1 S has a natural grading given by deg(f/g) = deg f -
deg g for f homogeneous in S and g E T. S<Pl is a local ring, with maximal
ideal (p · T- 1S) n S<vJ· In particular, if Sis a·domain, then for p = (0) we
obtain a field s((O))• Similarly, iff E s is a homogeneous element, we denote
by s(f) the sub ring of elements of degree 0 in the localized ring sf•

Theorem 3.4. Let Y s P" be a projective variety with homogeneous co-

ordinate ring S( Y). Then:
(a) (O(Y) = k;
(b) for any point P E Y, let mp S S(Y) be the ideal generated by the
set of homogeneous f E S(Y) such that f(P) = 0. Then @p = S(Y)(mp);
(c) K(Y) ~ S(Y)<<OJJ·
PROOF. To begin with, let Ui s P" be the open set xi #- 0, and let Y; =
Y n Ui. Then Ui is isomorphic to A" by the isomorphism cpi of (3.3), so we
can consider Y; as an affine variety. There is a natural isomorphism cpf
of the affine coordinate ring A( Y;) with the localization S( Y)<xil of the homo-
geneous coordinate ring of Y. We first make an isomorphism of k[ y 1 , . . . , Yn]
with k[x 0 , ••• ,xnJx;) by sending f(y 1 , •.• ,yn) to f(x 0 jxi> ... ,xn/xJ, leaving
out xdxi> as in the proof of (2.2). This isomorphism sends /( }/) to /( Y)S<x.J
(cf. Ex. 2.6), so passing to the quotient, we obtain the desired isomorphism
cpf:A(Y;) ~ S(Y)<x.J·
Now to prove (b), let P E Y be any point, and choose i so that P E Y;.
Then by (3.2), @p ~ A(}/)111 " ' where m~ is the maximal ideal of A(Y;) corre-
sponding toP. One checks easily that cpf(m~) = mp · S(Y)<x.J· Now xi f! mp,
and localization is transitive, so we find that A( Y;)mJ. ~ S( Y)(mp)' which
proves (b).
To prove (c), we use (3.2) again to see that K(Y), which is equal to K(Y;),
is the quotient field of A(}/). But by cpf, this is isomorphic to S( Y)<<on·
To prove (a), Jet f E (0( Y) be a global regular function. Then for each i,
f is regular on Y;, so by (3.2), f E A( Y;). But we have just seen that A( Y;) ~
S(Y)<x,J• so we conclude that f can be written as gdxf' where gi E S(Y) is

18
 3 Morphisms

homogeneous of degree N;. Thinking of (O(Y), K(Y) and S(Y) all as sub-
rings of the quotient field L of S(Y), this means that xf'f E S(Y)N,, for each i.
Now choose N ~ L,N;. Then S(Y)N is spanned as a k-vector space by
monomials of degree N in x 0 , . . . ,xn, and in any such monomial, at least
one X; occurs to a power ~N;. Thus we have S(Y)N · f s:;; S(Y)N. Iterating,
we have S(Y)N ° r s:;; S(Y)N for all q > 0. In particular, x~r E S(Y) for
all q > 0. This shows that the sub ring S( Y)[f] of Lis contained in x 0 N S( Y),
which is a finitely generated S(Y)-module. Since S(Y) is a noetherian ring,
S(Y)[f] is a finitely generated S(Y)-module, and therefore f is integral
over S(Y) (see, e.g., Atiyah-Macdonald [1, p. 59]). This means that there
are elements ab ... ,am E S(Y) such that
fm + alfm-1 + 0 0 0 +am= 0.

Since f has degree 0, we can replace the a; by their homogeneous components

of degree 0, and still have a valid equation. But S(Y) 0 = k, so the a; E k,
and f is algebraic over k. But k is algebraically closed, so f E k, which
completes the proof.

Our next result shows that if X and Y are affine varieties, then X is iso-
morphic to Y if and only if A( X) is isomorphic to A( Y) as a k-algebra.
Actually the proof gives more, so we state the stronger result.

Proposition 3.5. Let X be any variety and let Y be an affine variety. Then
there is a natural bijective mapping of sets

o::Hom(X,Y) ~ Hom(A(Y),(O(X))

where the left Hom means morphisms of varieties, and the right Hom
means homomorphisms of k-algebras.
PROOF. Given a morphism q>: X ----> Y, q> carries regular functions on Y to
regular functions on X. Hence q> induces a map (O(Y) to (O(X), which is
clearly a homomorphism of k-algebras. But we have seen (3.2) that (O(Y) ~
A(Y), so we get a homomorphism A(Y)----> (O(X). This defines o:.
Conversely, suppose given a homomorphism h: A( Y)----> (O(X) of k-algebras.
Suppose that Y is a closed subset of A", so that A(Y) = k[x 1 , . . . ,xn]/I(Y).
Let :X; be the image of X; in A(Y), and consider the elements~; = h(xJ E (O(X).
These are global functions on X, so we can use them to define a mapping
1/J:X----> A" by 1/J(P) = (~ 1 (P), ... '~"(P)) for P EX.
We show next that the image of 1jJ is contained in Y. Since Y = Z(J(Y) ),
it is sufficient to show that for any P EX and any f E I(Y),f(lji(P)) = 0. But

Now f is a polynomial, and h is a homomorphism of k-algebras, so we have

19
I Varieties

since f E J(Y). So 1/1 defines a map from X to Y, which induces the given
homomorphism h.
To complete the proof, we must show that 1jJ is a morphism. This is a
consequence of the following lemma.

Lemma 3.6. Let X be any variety, and let Y £ An be an affine variety. A

map of sets 1/1: X --+ Y is a morphism if and only if X; o 1/J is a regular
function on X for each i, where x 1, . . . ,xn are the coordinate functions
on An.

PROOF. If 1jJ is a morphism, the X; o 1jJ must be regular functions, by definition

of a morphism. Conversely, suppose the x; o 1jJ are regular. Then for any
polynomial f = f(x 1 , . . . ,xn), f o 1/J is also regular on X. Since the closed
sets of Y are defined by the vanishing of polynomial functions, and since
regular functions are continuous, we see that 1/J- 1 takes closed sets to closed
sets, so 1/J is continuous. Finally, since regular functions on open subsets of
Y are locally quotients of polynomials, g o 1jJ is regular for any regular
function g on any open subset of Y. Hence 1jJ is a morphism.

Corollary 3.7. If X, Y are two affine varieties, then X and Y are isomorphic
if and only if A(X) and A(Y) are isomorphic ask-algebras.
PROOF. Immediate from the proposition.

In the language of categories, we can express the above result as follows:

Corollary 3.8. The functor X 1---> A( X) induces an arrow-reversing equivalence
of categories between the category of affine varieties over k and the category
of finitely generated integral domains over k.

We include here an algebraic result which will be used in the exercises.

Theorem 3.9A (Finiteness of Integral Closure). Let A be an integral domain

which is a finitely generated algebra over a field k. Let K be the quotient
field of A, and let L be a finite algebraic extension of K. Then the integral
closure A' of A in L is a finitely generated A -module, and is also a finitely
generated k-algebra.
PROOF. Zariski-Samuel [1, vol. 1, Ch. V., Thm. 9, p. 267.]

EXERCISES

3.1. (a) Show that any conic in A2 is isomorphic either to A 1 or A 1 - { 0} (cf. Ex. 1.1 ).
(b) Show that A 1 is not isomorphic to any proper open subset of itself. (This result
is generalized by (Ex. 6.7) below.)
(c) Any conic in P 2 is isomorphic to P 1 .
(d) We will see later (Ex. 4.8) that any two curves are homeomorphic. But show
now that A2 is not even homeomorphic to P 2 •

20
 3 Morphisms

(e) If an affine variety is isomorphic to a projective variety, then it consists of only

one point.
3.2. A morphism whose underlying map on the topological spaces is a homeomor-
phism need not be an isomorphism.
(a) For example, let <p:A 1 --> A 2 be defined by t f--> (t 2 ,t 3 ). Show that <p defines a
bijective bicontinuous morphism of A 1 onto the curve y 2 = x 3 , but that <p is
not an isomorphism.
(b) For another example, let the characteristic of the base field k be p > 0, and
define a map <p :A 1 --> A 1 by t f--> tP. Show that <p is bijective and bicontinuous
but not an isomorphism. This is called the Frobenius morphism.
3.3. (a) Let <p:X--> Y be a morphism. Then for each P EX, <p induces a homomor-
phism of local rings <pt:(!J<PiP).Y--> (!JP.x·
(b) Show that a morphism <p is an isomorphism if and only if <p is a homeomor-
phism, and the induced map <pp on local rings is an isomorphism, for all P EX.
(c) Show that if <p(X) is dense in Y, then the map <pp is injective for all P EX.
3.4. Show that the d-uple embedding of P" (Ex. 2.12) is an isomorphism onto its
image.
3.5. By abuse of language, we will say that a variety "is affine" if it is isomorphic to
an affine variety. If H <;::; P" is any hypersurface, show that P" - H is affine.
[Hint: Let H have degree d. Then consider the d-uple embedding of P" in pN
and use the fact that pN minus a hyperplane is affine.
3.6. There are quasi-affine varieties which are not affine. For example, show that
X = A 2 - {(0,0)} is not affine. [Hint: Show that (!)(X) ~ k[x,y] and use (3.5).
See (Ill, Ex. 4.3) for another proof.]
3.7. (a) Show that any two curves in P 2 have a nonempty intersection.
(b) More generally, show that if Y <;::; P" is a projective variety of dimension ;, 1,
and if H is a hypersurface, then Y n H =f. 0. [Hint: Use (Ex. 3.5) and (Ex.
3.1e). See (7.2) for a generalization.]
3.8. Let H; and Hi be the hyperplanes in P" defined by X; = 0 and x1 = 0, with i =f. j.
Show that any regular function on P" - (H; n Hi) is constant. (This gives an
alternate proof of (3.4a) in the case Y = P".)
3.9. The homogeneous coordinate ring of a projective variety is not invariant under
isomorphism. For example, let X = P 1 , and let Y be the 2-uple embedding of
P 1 in P 2 . Then X ~ Y (Ex. 3.4). But show that S(X) "1. S( Y).

3.10. Subvarieties. A subset of a topological space is locally closed if it is an open

subset of its closure, or, equivalently, if it is the intersection of an open set with
a closed set.
If X is a quasi-affine or quasi-projective variety and Y is an irreducible locally
closed subset, then Y is also a quasi-affine (respectively, quasi-projective) variety,
by virtue of being a locally closed subset of the same affine or projective space.
We call this the induced structure on Y, and we call Y a subvariety of X.
Now let <p:X--> Y be a morphism, let X' s::; X and Y' s::; Y be irreducible
locally closed subsets such that <p(X') <;::; Y'. Show that <fJix·: X' --> Y' is a mor-
phism.

21
I Varieties

3.11. Let X be any variety and let P E X. Show there is a 1-1 correspondence between
the prime ideals of the local ring {!)P and the closed subvarieties of X containing P.
3.12. If P is a point on a variety X, then dim {!)P = dim X. [Hint:Reduce to the
affine case and use (3.2c).]
3.13. The Local Ring of a Subvariety. Let Y <:; X be a subvariety. Let {!)Y.x be the set
<
of equivalence classes U,f) where U <:; X is open, U n Y =/= 0, and f is a
regular function on U. We say <UJ) is equivalent to <V,g), iff= g on U n V.
Show that {!)Y.x is a local ring, with residue field K(Y) and dimension = dim X --
dim Y. It is the local ring of Yon X. Note if Y = Pis a point we get {!)p, and if
Y = X we get K(X). Note also that if Y is not a point, then K(Y) is not alge-
braically closed, so in this way we get local rings whose residue fields are not
algebraically closed.
3.14. Projection from a Point. Let P" be a hyperplane in pn+ 1 and let P E pn+ 1 - P".
Define a mapping cp :P"+ 1 - { P}--> P" by cp(Q) =the intersection of the unique
line containing P and Q with P"-
(a) Show that cp is a morphism.
(b) Let Y <:; P 3 be the twisted cubic curve which is the image of the 3-uple em-
bedding of P 1 (Ex. 2.12). If t,u are the homogeneous coordinates on PI, we
say that Y is the curve given parametrically by (x,y,z,w) = (t 3 ,t 2 u,tu 2 ,u 3 ). Let
P = (0,0,1,0), and let P 2 be the hyperplane z = 0. Show that the projection of
Y from P is a cuspidal cubic curve in the plane, and find its equation.

3.15. Products of Affine Varieties. Let X <:; A" and Y <:; Am be affine varieties.
(a) Show that X x Y <:; An+m with its induced topology is irreducible. [Hint:
Suppose that X x Y is a union of two closed subsets Z 1 u Z 2 . Let X; =
{xEXIx X y <:; Z;}, i = 1,2. Show that X= XI u Xz and xl,x2 are
closed. Then X= X 1 or X 2 so X x Y = Z 1 or Z 2 .] The affine variety
X x Y is called the product of X and Y. Note that its topology is in general
not equal to the product topology (Ex. 1.4).
(b) Show that A(X x Y) ~ A(X) @k A(Y).
(c) Show that X x Y is a product in the category of varieties, i.e., show (i) the
projections X x Y --> X and X x Y --> Y are morphisms, and (ii) given a
variety Z, and the morphisms Z --> X, Z --> Y, there is a unique morphism
Z --> X x Y making a commutative diagram
z ---------+ X X y

V><~
~Y. X
(d) Show that dim X x Y = dim X + dim Y.
3.16. Products of Quasi-Projective Varieties. Use the Segre embedding (Ex. 2.14) to
identify P" x pm with its image and hence give it a structure of projective variety.
Now for any two quasi-projective varieties X <:; P" and Y <:; pm, consider
X X Y <:; P" X pm_
(a) Show that X x Y is a quasi-projective variety.
(b) If X, Y are both projective, show that X x Y is projective.
*(c) Show that X x Y is a product in the category of varieties.

22
 3 Morphisms

3.17. Normal Varieties. A variety Y is normal at a point P E Y if @p is an integrally

closed ring. Y is normal if it is normal at every point.
(a) Show that every conic in P 2 is normal.
(b) Show that the quadric surfaces Q 1,Q 2 in P 3 given by equations Q 1 :xy = zw;
Q2 :xy = z2 are normal (cf. (II. Ex. 6.4) for the latter.)
(c) Show that the cuspidal cubic y 2 = x 3 in A2 is not normal.
(d) If Y is affine, then Y is normal= A(Y) is integrally closed.
(e) Let Y be an affine variety. Show that there is a normal affine variety Y, and a
morphism n: Y--+ Y, with the property that whenever Z is a normal variety,
and cp:Z--+ Y is a dominant morphism (i.e., cp(Z) is dense in Y), then there is
a unique morphism e:z--+ Y such that cp = no e. Y is called the normaliza-
tion of Y. You will need (3.9A) above.
3.18. Projectively Normal Varieties. A projective variety Y s; P" is projectively normal
(with respect to the given embedding) if its homogeneous coordinate ring S(Y)
is integrally closed.
(a) If Y is projectively normal, then Y is normal.
(b) There are normal varieties in projective space which are not projectively
normal. For example, let Y be the twisted quartic curve in P 3 given para-
metrically by (x,y,z,w) = (t 4 ,t 3 u,tu 3 ,u4 ). Then Yis normal but not projectively
normal. See (III, Ex. 5.6) for more examples.
(c) Show that the twisted quartic curve Y above is isomorphic to P 1 , which is
projectively normal. Thus projective normality depends on the embedding.
3.19. Automorphisms of A". Let cp:A"--+ A" be a morphism of A" to A" given by n
polynomials f 1 , . . . ,fn of n variables x 1 , . . . ,x"" Let J = detliJ/;/iJxjl be the
Jacobian polynomial of cp.
(a) If cp is an isomorphism (in which case we call cp an automorphism of A") show
that J is a nonzero constant polynomial.
**(b) The converse of (a) is an unsolved problem, even for n = 2. See, for example.
Vitushkin [1].
3.20. Let Y be a variety of dimension ~ 2, and let P E Y be a normal point. Let f be
a regular function on Y - P.
(a) Show that f extends to a regular function on Y.
(b) Show this would be false for dim Y = 1.
See (III, Ex. 3.5) for generalization.
3.21. Group Varieties. A group variety consists of a variety Y together with a morphism
J1: Y x Y -+ Y, such that the set of points of Y with the operation given by J1 is a
group, and such that the inverse map y --+ y- 1 is also a morphism of Y--+ Y.
(a) The additive group Ga is given by the variety A 1 and the morphism J1: A2 --+ A 1
defined by Jl(a,b) = a + b. Show it is a group variety.
(b) The multiplicative qroup Gm is given by the variety A 1 - {(0)} and the mor-
phism !liu.h) = ah. Show it is a group variety.
(c) If G is a group variety, and X is any variety, show that the set Hom(X,G) has a
natural group structure.
(d) For any variety X, show that Hom(X,Ga) is isomorphic to (1J(X) as a group
under addition.
(e) For any variety X, show that Hom(X,Gml is isomorphic to the group of units
in @(X), under multiplication.

23
I Varieties

4 Rational Maps

In this section we introduce the notions of rational map and birational

equivalence, which are important for the classification of varieties. A rational
map is a morphism which is only defined on some open subset. Since an
open subset of a variety is dense, this already carries a lot of information.
In this respect algebraic geometry is more "rigid" than differential geometry
or topology. In particular, the concept of birational equivalence is unique
to algebraic geometry.

Lemma 4.1. Let X and Y be varieties, let cp and ljJ be two morphisms fi'om
X to Y, and suppose there is a nonempty open subset U s; X such that
({Jiu = 1/Jiu· Then cp = 1/J.
PROOF. We may assume that Y s; P" for some n. Then by composing with
the inclusion morphism Y --> P", we reduce to the case Y = P". We consider
the product P" x P", which has a structure of projective variety given by its
Segre embedding (Ex. 3.16). The morphisms cp and ljJ determine a map
cp x 1/f:X __. P" x P", which in fact is a morphism (Ex. 3.16c). Let .d =
{P x PIP E P"} be the diagonal subset of P" x P". It is defined by the
equations {xd'j = xjy;li,j = 0,1, ... ,n} and so is a closed subset ofP" x P".
By hypothesis cp x 1/J(U) s; .d. But U is dense in X, and .d is closed, so
cp x 1/f(X) s; .d. This says that cp = 1/J.

Definition. Let X, Y be varieties. A rational map cp: X --> Y is an equivalence

class of pairs <U,cpu) where U is a nonempty open subset of X, ({Ju is a
morphism of U to Y, and where <U,cpu) and <V,<pv) are equivalent if
({Ju and ({Jv agree on U n V. The rational map cp is dominant if for some
<
(and hence every) pair U ,({Ju), the image of ({Ju is dense in Y.

Note that the lemma implies that the relation on pairs <U,cpu) just
described is an equivalence relation. Note also that a rational map cp: X --> Y
is not in general a map of the set X to Y. Clearly one can compose dvminant
rational maps, so we can consider the category of varieties and dominant
rational maps. An "isomorphism" in this category is called a birational map:

Definition. A birational map cp: X --> Y is a rational map which admits an

inverse, namely a rational map ljJ: Y --> X such that ljJ a cp = idx and
cp o ljJ = idy as rational maps. If there is a birational map from X to Y,
we say that X and Yare birationally equivalent, or simply birational.

The main result of this section is that the category of varieties and domi-
nant rational maps is equivalent to the category of finitely generated field

24
 4 Rational Maps

extensions of k, with the arrows reversed. Before giving this result, we need
a couple of lemmas which show that on any variety, the open affine subsets
form a base for the topology. We say loosely that a variety is affine if it is
isomorphic to an affine variety.

Lemma 4.2. Let Y be a hypersurface in An given by the equationf(x 1 , •.. ,xn) =

0. Then An - Y is isomorphic to the hypersurface H in An+ 1 given by
xn + d = 1. In particular, An - Y is affine, and its affine ring is
k[xl, ... ,xn]f·
PROOF. For P = (a~o ... ,an+ 1 ) E H, let cp(P) = (a 1 , . . • ,an). Then clearly cp
is a morphism from H to An, corresponding to the homomorphism of rings
A --> A 1 , where A = k[ x 1 , . . . ,xn]. It is also clear that cp gives a bijective
mapping of H onto its image, which is An - Y To show that cp is an isomor-
phism, it is sufficient to show that cp- 1 is a morphism. But cp- 1 (a 1 , ••• ,an) =
(a 1 , . • . ,an,l/f(a 1 , . . . ,an)), so the fact that cp- 1 is a morphism on An - Y
follows from (3.6).

Proposition 4.3. On any variety Y, there is a base for the topology consisting
of open affine subsets.
PROOF. We must show for any point P E Y and any open set U containing P,
that there exists an open affine set V with P E V <:; U. First, since U is also
a variety, we may assume U = Y Secondly, since any variety is covered by
quasi-affine varieties (2.3), we may assume that Y is quasi-affine in An.
Let Z = Y - Y, which is a closed set in An, and let a <:; A = k[x 1 , . . . ,xn]
be the ideal of Z. Then, since Z is closed, and P ¢ Z, we can find a polynomial
f E a such that f(P) =f. 0. Let H be the hypersurface f = 0 in An. Then
Z <:; H but P ¢ H. Thus P E Y - Y n H, which is an open subset of
Y Furthermore, Y - Y n H is a closed subset of An - H, which is affine
by (4.2), hence Y - Y n H is affine. This is the required affine neighbor-
hood of P.

Now we come to the main result of this section. Let cp: X --> Y be a
<
dominant rational map, represented by U,ffJu ). Let f E K( Y) be a rational
function, represented by <V,f) where Vis an open set in Y, and f is a regular
function on V. Since ffJu(U) is dense in Y, cp{) 1(V) is a nonempty open subset
of X, sofa ffJu is a regular function on cp{) 1 (V). This gives us a rational
function on X, and in this manner we have defined a homomorphism of
k-algebras from K(Y) to K(X).

Theorem 4.4. For any two varieties X and Y, the above construction gives a
bijection between
(i) the set of dominant rational maps from X to Y, and
(ii) the set of k-algebra homomorphisms from K( Y) to K(X).

25
I Varieties

Furthermore, this correspondence gives an arrow-reversing equivalence of

categories of the category of varieties and dominant rational maps with the
category of finitely generated field extensions of k.
PROOF. We will construct an inverse to the mapping given by the construction
.tbove. Let B:K(Y)--+ K(X) be a homomorphism of k-algebras. We wish
to define a rational map from X to Y. By (4.3), Y is covered by affine varieties,
so we may assume Y is affine. Let A(Y) be its affine coordinate ring, and let
y 1 , . . . ,y" be generators for A(Y) as a k-algebra. Then B(y 1 ), . . . ,B(y") are
rational functions on X. We can find an open set U s; X such that the func-
tions B(y;) are all regular on U. Then e defines an injective homomorphism
of k-algebras A( Y) --+ (?)( U). By (3.5) this corresponds to a morphism
q>: U --+ Y, which gives us a dominant rational map from X to Y. It is easy
to see that this gives a map of sets (ii) --+ (i) which is inverse to the one defined
above.
To see that we have an equivalence of categories as stated, we need only
check that for any variety Y, K(Y) is finitely generated over k, and conversely,
if K/k is a finitely generated field extension, then K = K(Y) for some Y.
If Y is a variety, then K( Y) = K(U) for any open affine subset, so we may
assume Y affine. Then by (3.2d), K(Y) is a finitely generated field extension
of k. On the other hand, let K be a finitely generated field extension of k.
Let y 1 , . • . ,y" E K be a set of generators, and let B be the sub-k-algebra of K
generated by y 1 , . . . ,y.. Then B is a quotient of the polynomial ring
A = k[x 1 , . . . ,x"]' soB ~ A(Y) for some variety Yin A". Then K ~ K(Y)
so we are done.

Corollary 4.5. For any two varieties X,Y the following conditions are equiv-
ulent:
(i) X and Yare birationally equivalent;
(ii) there are open subsets U s; X and V s; Y with U isomorphic to V,
(iii) K(X) ~ K( Y) us k-algebrus.

PROOF.
(i) = (ii). Let q>: X --+ Y and lj;: Y --+ X be rational maps which are inverse
to each other. Let q> be represented by <U,q>) and let lj; be represented by
<V,l/J). Then lj; a q> is represented by <<r>- 1 (V),lj; o q>), and since lj; o q> = idx
as a rational map, lj; o q> is the identity on q> - 1 ( V). Similarly q> o lj; is the
identity on lj;- 1(U). We now take q>- 1(lj;- 1(U)) as our open set in X, and
lj;- 1 ( q>- 1 ( V)) as our open set in Y. It follows from the construction that these
two open sets are isomorphic via q> and lj;.
(ii) =(iii) follows from the definition of function field.
(iii) = (i) follows from the theorem.

As an illustration of the notion of birational correspondence, we will use

some algebraic results on field extensions to show that every variety is bi-

26
 4 Rational Maps

rational to a hypersurface. We assume familiarity with the notion of sepa-

rable algebraic field extensions, and the notions of transcendence base and
transcendence degree for infinite field extensions (see, e.g., Zariski-Samuel
[1, Ch. II]).

Theorem 4.6A (Theorem of the Primitive Element). Let L be a finite separable

extension field of afield K. Then there is an element r:t.. E L which generates
L as an extension field of K. Furthermore, if {3 1 , . . . ,f3n is any set of
generators of L over K, and if K is infinite, then r:t.. can be taken to be a
linear combination r:t.. = c 1 {3 1 + ... + cnf3n of the f3i with coefficients
C;EK.

PROOF. Zariski-Samuel [1, Ch. II, Theorem 19, p. 84]. The second statement
follows from the proof given there.

Definition. A field extension Kjk is separably generated if there is a tran-

scendence base {x;} for Kjk such that K is a separable algebraic extension
of k( {x;} ). Such a transcendence base is called a separating transcendence
base.

Theorem 4.7A. If a field extension Kjk is finitely generated and separably

generated, then any set of generators contains a subset which is a separating
transcendence base.

PROOF. Zariski-Samuel [1, Ch. II, Theorem 30, p. 104].

Theorem 4.8A. If k is a perfect field (hence in particular if k is algebraically

closed), any .finitely generated field extension Kjk is separably generated.

PROOF. Zariski-Samuel [1, Ch. II, Theorem 31, p. 105], or Matsumura

[2, Ch. 10, Corollary, p. 194].

Proposition 4.9. Any variety X of dimension r is birational to a hypersurface

Yin pr+ I.

PROOF. The function field K of X is a finitely generated extension field of k.

By (4.8A), K is separably generated over k. Hence we can find a transcendence
base x 1, ... ,xr E K such that K is a finite separable extension of k(x 1 , . •• ,xrl·
Then by (4.6A) we can find one further element y E K such that
K = k(x~> . .. ,x.,y). Now y is algebraic over k(x~> . .. ,xr), so it satisfies a
polynomial equation with coefficients which are rational functions in
x~> ... ,xr. Clearing denominators, we get an irreducible polynomial
f(x~> . .. ,x.,y) = 0. This defines a hypersurface in Ar+ 1 with function
field K, which, according to (4.5}, is birational to X. Its projective closure
(Ex. 2.9) is the required hypersurface Y c:; pr+ 1 .

27
I Varieties

Blowing Up
As another example of a birational map, we will now construct the blowing-
up of a variety at a point. This important construction is the main tool in
the resolution of singularities of an algebraic variety.
First we will construct the blowing-up of A" at the point 0 = (0, ... ,0).
Consider the product A" x P"- 1, which is a quasi-projective variety
(Ex. 3.16). If x 1 , . . . ,x" are the affine coordinates of A", and if y 1 , . . . ,y"
are the homogeneous coordinates of P"- 1 (observe the unusual notation!),
then the closed subsets of A" x P"- 1 are defined by polynomials in the
X;,yi, which are homogeneous with respect to the Yi·
We now define the blowing-up of A" at the point 0 to be the closed subset
X of A" x P"- 1 defined by the equations {x;Yi = XiY;Ii,j = 1, ... ,n}.

X A" X pn-1

We have a natural morphism cp: X --+ A" obtained by restricting the pro-
jection map of A" x P"- 1 onto the first factor. We will now study the
properties of X.
(1) If PEA", P # 0, then cp- 1(P) consists of a single point. In fact,
cp gives an isomorphism of X- cp- 1 (0) onto A"- 0. Indeed, let P =
(ab ... ,an), with some a; # 0. Now if P x (Yb ... ,y") E cp- 1 (P), then for
each j, Yi = (aia;)y;, so (y 1, . . . ,yn) is uniquely determined as a point in
P"- 1 . In fact, setting Y; = a;, we can take (yb ... ,yn) = (a 1 , . . . ,an). Thus
cp - 1 (P) consists of a single point. Furthermore, for PEA" - 0, setting
t/J(P) = (ab . .. ,an) x (a 1 , . . . ,an) defines an inverse morphism to cp,
showingX- cp- 1 (0)isisomorphictoA"- 0.
(2) cp- 1 ( 0) ~ P"- 1 . Indeed, cp- 1 ( 0) consists of all points 0 x Q, with
Q = (y 1 , . . . ,yn) E P"-1, subject to no restriction.
(3) The points of cp - 1 ( 0) are in 1-1 correspondence with the set of lines
through 0 in A". Indeed, a line L through 0 in A" can be given by para-
metric equations X; = a;t, i = 1, ... ,n, where a; E k are not all zero, and
tEA 1. Now consider the line L' = cp - 1(L - 0) in X - cp - 1( 0). It is given
parametrically by X; = a;t, Y; = a;t, with t E A 1 - 0. But the Y; are homo-
geneous coordinates in P"- 1 , so we can equally well describe L' be the
equations X; = a;t,y; = a;, for tEA 1 - 0. These equations make sense
also for t = 0, and give the closure L' of L' in X. Now L' meets cp - 1 ( 0) in
the point Q = (a 1 , . . . ,an) E pn- \ so we see that sending L to Q gives a
1-1 correspondence between lines through 0 in A" and points of cp- 1 ( 0).
(4) X is irreducible. Indeed, X is the union of X - cp - 1( 0) and cp - 1( 0).
The first piece is isomorphic to A" - 0, hence irreducible. On the other
hand, we have just seen that every point of cp- 1 ( 0) is in the closure of some
28
 4 Rational Maps

uo
Figure 3. Blowing up.

subset (the line L') of X - <p- 1( 0). Hence X - <p- 1( 0) is dense in X, and
X is irreducible.
Definition. If Y is a closed subvariety of A" passing through 0, we define
the blowing-up of Y at the point 0 to be Y = (<p - 1 ( Y - 0) )-, where
<p:X--> A" is the blowing-up of A" at the point 0 described above. We
denote also by <p: Y --> Y the morphism obtained by restricting <p :X --> A"
to Y. To blow up at any other point P of A", make a linear change of
coordinates sending P to 0.
Note that <p induces an isomorphism of Y- <p- 1 (0) to Y- 0, so that
<pis a birational morphism of Y to Y. Note also that this definition apparently
depends on the embedding of Y in A", but in fact, we will see later that
blowing-up is intrinsic (II, 7.15.1).
The effect of blowing up a point of Y is to "pull apart" Y near 0 according
to the different directions of lines through 0. We will illustrate this with
an example.
Example 4.9.1. Let Y be the plane cubic curve given by the equation y 2 =
x 2 (x + 1). We will blow up Y at 0 (Fig. 3). Let t,u be homogeneous co-
ordinates for P 1 . Then X, the blowing-up of A2 at 0, is defined by the
equation xu= ty inside A 2 x P 1 . It looks like A2 , except that the point 0
has been replaced by a P 1 corresponding to the slopes of lines through 0.
We will call this P 1 the exceptional curve, and denote it by E.
We obtain the total inverse image of Yin X by considering the equations
y 2 = x 2 (x + 1) and xu = ty in A2 x P 1 . Now P 1 is covered by the open
sets t i= 0 and u i= 0, which we consider separately. If t i= 0, we can set
t = 1, and use u as an affine parameter. Then we have the equations

y2 = x 2 (x + 1)
y = xu

29
I Varieties

in A 3 with coordinates x, y,u. Substituting, we get x 2 u 2 - x 2 (x + 1) = 0,

which factors. Thus we obtain two irreducible components, one defined by
x = 0, y = 0, u arbitrary, which is E, and the other defined by u 2 = x + 1,
y = xu. This is Y. Note that Y meets Eat the points u = ±1. These points
correspond to the slopes of the two branches of Y at 0.
Similarly one can check that the total inverse image of the x-axis consists
of E and one other irreducible curve, which we call the strict transform of the
x-axis (it is the curve L' described earlier corresponding to the line L = x-axis).
This strict transform meets E at the point u = 0. By considering the other
open set u =1= 0 in A 2 x P 1 , one sees that the strict transform of the y-axis
meets E at the point t = 0, u = 1.
These conclusions are summarized in Figure 3. The effect of blowing up
is thus to separate out branches of curves passing through 0 according to
their slopes. If the slopes are different, their strict transforms no longer meet
in X. Instead, they meet E at points corresponding to the different slopes.

EXERCISES

4.1. Iff and g are regular functions on open subsets U and V of a variety X, and if
f = g on U n V, show that the function which is f on U and g on V is a regular
function on U u V. Conclude that iff is a rational function on X, then there is
a largest open subset U of X on which f is represented by a regular function.
We say that f is defined at the points of U.
4.2. Same problem for rational maps. If [Link] is a rational map of X to Y, show there
is a largest open set on which [Link] is represented by a morphism. We say the ra-
tional map is defined at the points of that open set.
4.3. (a) Letfbe the rational function on P 2 given by f = xdx 0 . Find the set of points
where f is defined and describe the corresponding regular function.
(b) Now think of this function as a rational map from P 2 to A 1 • Embed A 1 in P 1 ,
and let I.{J:P 2 --> P 1 be the resulting rational map. Find the set of points where
[Link] is defined, and describe the corresponding morphism.

4.4. A variety Y is rational if it is birationally equivalent toP" for some n (or, equiva-
lently by (4.5), if K(Y) is a pure transcendental extension of k).
(a) Any conic in P 2 is a rational curve.
(b) The cuspidal cubic y 2 = x 3 is a rational curve.
(c) Let Y be the nodal cubic curve y 2 z = x 2 (x + z) in P 2 . Show that the pro-
jection [Link] from the point P = (0,0,1) to the line z = 0 (Ex. 3.14) induces a
birational map from Y to P 1 . Thus Y is a rational curve.

4.5. Show that the quadric surface Q:xy = zw in P 3 is birational to P 2 , but not
isomorphic to P 2 (cf. Ex. 2.15).

4.6. Plane Cremona Tramiformations. A birational map of P 2 into itself is called a

plane Cremona transformation. We give an example, called a quadratic transfor-
mation. It is the rational map I.{J:P 2 --> P 2 given by (a 0 ,a 1 ,a 2 )--> (a 1a 2 ,a 0 a 2 ,a 0 a 1 )
when no two of a 0 ,a 1 ,a 2 are 0.

30
 5 Nonsingular Varieties

(a) Show that <p is birational, and is its own inverse.

(b) Find open sets U,V s; P 2 such that <p: U ---> Vis an isomorphism.
(c) Find the open sets where <p and <p- 1 are defined, and describe the correspond-
ing morphisms. See also (V, 4.2.3).
4.7. Let X and Y be two varieties. Suppose there are points P EX and Q E Y such
that the local rings ~P.x and (r;;Q.Y are isomorphic as k-algebras. Then show
that there are open sets P E U s; X and Q E V s; Y and an isomorphism of
U to V which sends P to Q.

4.8. (a) Show that any variety of positive dimension over k has the same cardinality as
k. [Hints: Do A" and P" first. Then for any X, use induction on the dimension
11. Use (4.9) to make X birational to a hypersurface H <;: pn+ 1 . Use (Ex. 3.7)

to show that the projection of H to P" from a point not on H is finite-to-one

and surjective.]
(b) Deduce that any two cwTes over k are homeomorphic (cf. Ex. 3.1).
4.9. Let X be a projective variety of dimension r in P". with 11 ? r + 2. Show that
for suitable choice of P ¢X, and a linear P"- 1 s; P". the projection from P to
pn-t (Ex. 3.14) induces a hirotionol morphism of)( onto its image x· s; P"- 1 •
You will need to use (4.6Al. (4.7A). and (4.8A). This shows in particular that the
birational map of (4.9) can be obtained by a finite number of su~.:h proje~.:tions.
4.10. Let r be the cusp ida! cubic curve _r 2 = x 3 in A2 Blow up the point 0 = (0,0).
let E be the exceptional curve, and let Y be the strict transform of Y. Show that
E meets Y in one point, and that Y ::; A 1 • In this case the morphism <p: Y ---> Y
is bijective and bicontinuous, but it is not an isomorphism.

5 Nonsingular Varieties

The notion of nonsingular variety in algebraic geometry corresponds to the

notion of manifold in topology. Over the complex numbers, for example,
the nonsingular varieties are those which in the "usual" topology are complex
manifolds. Accordingly, the most natural (and historically first) definition
of nonsingularity uses the derivatives of the functions defining the variety:

Definition. Let Y ~ A" be an affine variety, and let j 1 , ••• ,J; E A =

k [ x 1 , . . . ,x"] be a set of generators for the ideal of Y. Y is nonsingular at a
point P E Y if the rank of the matrix IIU)Rx)(Plll is 11 - r, where r is
the dimension of Y. Y is nonsingular if it is nonsingular at every point.

A few comments are in order. In the first place, the notion of partial
derivative of a polynomial with respect to one of its variables makes sense
over any field. One just applies the usual rules for differentiation. Thus no
limiting process is needed. But funny things can happen in characteristic
p > 0. For example, if f(x) = xP, then df/dx = pxp-J = 0, since p = 0 ink.
In any case, iff E A is a polynomial, then for each i, of/ex; is a polynomial.

31
I Varieties

The matrix ll(o/;/oxi)(P)II is called the Jacobian matrix at P. One can show
easily that this definition of nonsingularity is independent of the set of
generators of the ideal of Y chosen.
One drawback of our definition is that it apparently depends on the
embedding of Y in affine space. However, it was shown in a fundamental
paper, Zariski [1 ], that nonsingularity could be described intrinsically in
terms of the local rings. In our case the result is this.

Definition. Let A be a noetherian local ring with maximal ideal m and residue
field k = A/m. A is a regular local ring if dimk m/m 2 = dim A.

Theorem 5.1. Let Y <;; An be an affine variety. Let P E Y be a point. Then Y

is nonsingular m P if and only if the local ring (!) P,Y is a regular local ring.
PROOF. Let p be the point (ab ... ,an) in An, and let Op = (xl - ({1, ... ,Xn- an)
be the corresponding maximal ideal in A = k[x 1 , . . • ,xnJ. We define a
linear map 0: A --+ kn by

of (P), ... , -0cf (P) )

O(f) = \ -~
cx 1 Xn
for any f E A. Now it is clear that O(x; - a;) for i = 1, ... ,n form a basis of
kn, and that O(a~) = 0. Thus 0 induces an isomorphism O':ap/a~--+ kn.
Now let b be the ideal of Yin A, and let f 1 , ••. ,fr be a set of generators of b.
Then the rank of the Jacobian matrix J = li(o/;/oxi)(P)II is just the dimension
of O(b) as a subspace of kn. Using the isomorphism 0', this is the same as the
dimension of the subspace (b + a~)/a~ of ap/a~. On the other hand, the
local ring(!) P of P on Y is obtained from A by dividing by b and localizing at
the maximal ideal ap. Thus if m is the maximal ideal of (!)p, we have
m/m 2 ~ ap/(b + a~).
Counting dimensions of vector spaces, we have dim m/m 2 + rank J = n.
Now let dim Y = r. Then (!)Pis a local ring of dimension r (3.2), so (!)Pis
regular if and only if dimk mjm 2 = r. But this is equivalent to rank J = n - r,
which says that P is a nonsingular point of Y.
Note. Later we will give another characterization of nons in gular points in
terms of the sheaf of differential forms on Y (II, 8.15).
Now that we know the concept of nonsingularity is intrinsic, we can extend
the definition to arbitrary varieties.

Definition. Let Y be any variety. Y is nonsingular at a point P E Y if the local

ring (! P.r is a regular local ring. Y is nonsingulur if it is nonsingular at
every point. Y is sinyular if it is not nonsingular.
Our next objective is to show that most points of a variety are nonsingular.
We need an algebraic preliminary.

32
 5 Nonsingular Varieties

Proposition 5.2A. If A is a noetherian local ring with maximal ideal m and

residue field k, then dimk m/m 2 ~ dim A.
PROOF. Atiyah-Macdonald [1, Cor. 11.15, p. 121] or Matsumura [2,
p. 78].

Theorem 5.3. Let Y be a variety. Then the set Sing Y of singular points of Y
is a proper closed subset of Y.

PROOF. (See also II, 8.16.) First we show Sing Y is a closed subset. It is
sufficient to show for some open covering Y = U Y; of Y, that Sing Y; is
closed for each i. Hence by (4.3) we may assume that Y is affine. By (5.2) and
the proof of (5.1) we know that the rank of the Jacobian matrix is always
~ n - r. Hence the set of singular points is the set of points where the rank is
< n - r. Thus Sing Y is the algebraic set defined by the ideal generated by
I( Y) together with all determinants of (n - r) x (n - r) submatrices of the
matrix II8N8xill· Hence Sing Y is closed.
To show that Sing Y is a proper subset of Y, we first apply (4.9) to get Y
birational to a hypersurface in Pn. Since birational varieties have isomorphic
open subsets, we reduce to the case of a hypersurface. It is enough to consider
any open affine subset of Y, so we may assume that Y is a hypersurface in An,
defined by a single irreducible polynomial f(xb . .. ,xn) = 0.
Now Sing Y is the set of points P E Y such that (8f/8x;)(P) = 0 for i =
1, ... ,n. If Sing Y = Y, then the functions 8fj8x; are zero on Y, and hence
8fj8x; E I(Y) for each i. But I(Y) is the principal ideal generated by f, and
deg(8f/8x;) ~ deg f - 1 for each i, so we must have 8f/8x; = 0 for each i.
In characteristic 0 this is already impossible, because if X; occurs in f,
then 8f/8x; i= 0. So we must have char k = p > 0, and then the fact that
8f/8x; = 0 implies that f is actually a polynomial in xf. This is true for each
i, so by taking pth roots of the coefficients (possible since k is algebraically
closed), we get a polynomial g(x 1 , . . . ,xn) such that f = gP. But th:s
contradicts the hypothesis that f was irreducible, so we conclude that
Sing Y < Y.

Completion
For the local analysis of singularities we will now describe the technique of
completion. Let A be a local ring with maximal ideal m. The powers of m
define a topology on A, called the m-adic topology. By completing with
respect to this topology, one defines the completion of A, denoted A. Alter-
natively, one can define A as the inverse limit !!!!! A/mn. See Atiyah-
Macdonald [1, Ch. 10], Matsumura [2, Ch. 9], or Zariski-Samuel [1, vol. 2,
Ch. VIII] for general information on completions.
The significance of completion in algebraic geometry is that by passing
to the completion &P of the local ring of a point P on a variety X, one can
study the very local behavior of X near P. We have seen (Ex. 4.7) that if

33
I Varieties

points P EX and Q E Y have isomorphic local rings, then already P and Q

have isomorphic neighborhoods, so in particular X and Y are birational.
Thus the ordinary local ring (np carries information about almost all of X.
However, the completion r!p, as we will see, carries much more local in-
formation, closer to our intuition of what "local" means in topology or
differential geometry.
We will recall some of the algebraic properties of completion and then
give some examples.

Theorem 5.4A. Let A be a noetherian local ring with maximal ideal m, and
let A be its completion.
(a) A is a local ring, with maximal ideal m = mA, and there is a natural
injective homomorphism A --> A.
(b) If M is a finitely generated A-module, its completion M with respect
to its m-adic topology is isomorphic toM ®A A.
(c) dim A = dim A.
(d) A is regular if and only if A is regular.
PROOF. See Atiyah~Macdonald [1, Ch. 10, 11] or Zariski~Samuel [1, vol. 2,
Ch. Vlll].

Theorem 5.5A (Cohen Structure Theorem). If A is a complete regular local

ring of dimension n contain in?] some field, then A ~ k[[x 1 , •.• ,xn]J, the
ring of formal power series over the residue field k of A.
PROOF. Matsumura [2, Cor. 2, p. 206] or Zariski~Samuel [1, vol. 2, Cor.,
p. 307].

Definition. We say twu points P EX and Q E Y are analytically isomorphic

if there is an isomorphism &P ~ &Q as k-algebras.

Example 5.6.1. If P EX and Q E Y are analytically isomorphic, then

dim X = dim Y. This follows from (5.4A) and the fact that any local ring
of a point on a variety has the same dimension as the variety (Ex. 3.12).

Example 5.6.2. If P E X and Q E Y are nonsingular points on varieties of the

same dimension, then P and Q are analytically isomorphic. This follows
from (5.4A) and (5.5A). This example is the algebraic analogue of the fact
that any two manifolds (topological, differentiable, or complex) of the same
dimension are locally isomorphic.

Example 5.6.3. Let X be the plane nodal cubic curve given by the equation
y 2 = x 2 (x + 1). Let Y be the algebraic set in A 2 defined by the equation
x.r = 0. We will show that the point 0 = (0,0) on X is analytically iso-
morphic to the point 0 on Y. (Since we haven't yet developed the general
theory of local rings of points on reducible algebraic sets, we use an ad hoc

34
 5 Nonsingular Varieties

definition @0 ,Y = (k[x,y]/(xy))(x,y)· Thus &o,Y ~ k[[x,y]]/(xy).) This ex-

ample corresponds to the geometric fact that near 0, X looks like two lines
crossing.
To prove this result, we consider the completion &o,x which is isomorphic
to k[[x,y]]/(y 2 - x 2 - x 3 ). The key point is that the leading form of the
equation, namely y 2 - x 2 , factors into two distinct factors y + x andy - x
(we assume char k # 2). I claim there are formal power series
g = Y+ x + gz + g3 + · · ·
h = y - x + h2 + h 3 + ...
in k[[ x, y]], where g;,h; are homogeneous of degree i, such that y 2 - x 2 -
x 3 = gh. We construct g and h step by step. To determine g 2 and h2 , we
need to have
(y - x)gz + (y + x)h 2 = -x 3 .
This is possible, because y - x and y + x generate the maximal ideal of
k[[x,y]]. To determine g 3 and h 3, we need
(y - x)g 3 + (y + x)h 3 = -g 2 h 2
which is again possible, and so on.
Thus &o,x = k[[ x,y ]]/(gh). Since g and h begin with linearly independent
linear terms, there is an automorphism of k[[ x, y]] sending g and h to x
andy, respectively. This shows that &o,x ~ k[[ x,y]]/(xy) as required.
Note in this example that @0 ,x is an integral domain, but its completion
is not.

We state here an algebraic result which will be used in (Ex. 5.15) below.

Theorem 5.7A (Elimination Theory). Let f 1 , . . . ,f,. be homogeneous polyno-

mials in x 0 , . . . ,x", having indeterminate coefficients aii. Then there is a
set g 1 , . . . ,g 1 of polynomials in the a;i, with integer coefficients, which are
homogeneous in the coefficients of each J; separately, with the following
property: for any field k, and for any set of special values of the aii E k,
a necessary and sufficient condition for the J; to have a common zero different
from (0,0, ... ,0) is that the aii are a common zero of the polynomials gi.
PROOF. Vander Waerden [1, vol. II, §80, p. 8].

EXERCISES

5.1. Locate the singular points and sketch the following curves in A2 (assume char
k =1- 2). Which is which in Figure 4?
(a) x2 = + .l:
x4
(b) xy = x 6 + y 6;
(c) x3 = )'2 + x4 + )'4;
(d) x2y + xi = x4 + y4.

35
I Varieties

Node Triple point Cusp Tacnode

Figure 4. Singularities of plane curves.

5.2. Locate the singular points and describe the singularities of the following sur-
faces in A3 (assume char k # 2). Which is which in Figure 5?
(a) xy 2 = z 2 ;
(b) xz + yz = zz;
(c) xy + x 3 + y 3 = 0.

Conical double point Double line Pinch point

Figure 5. Surface singularities.

5.3. Multiplicities. Let Y ~ A 2 be a curve defined by the equation f(x,y) = 0. Let

P = (a,b) be a point of A2 . Make a linear change of coordinates so that P be-
comes the point (0,0). Then write f as a sum f = fo + f 1 + ... + f:J, where
j; is a homogeneous polynomial of degree i in x and y. Then we define the multi-
plicity of P on Y, denoted ,Up(Y), to be the least r such that fr # 0. (Note that
P E Y = ,Up(Y) > 0.) The linear factors of fr are called the tangent directions
at P.
(a) Show that ,Up(Y) = 1 =Pis a nonsingular point of Y.
(b) Find the multiplicity of each of the singular points in (Ex. 5.1) above.
5.4. Intersection Multiplicity. If Y,Z ~ A2 are two distinct curves, given by equations
f = 0, g = 0, and if P E Y n Z, we define the intersection multiplicity ( Y · Z)p
of Y and Z at P to be the length of the C9p-module C9p j(f,g).
(a) Show that (Y · Z)p is finite, and (Y · Z)p ~ flp(Y) · ,Up(Z).
(b) If P E Y, show that for almost all lines L through P (i.e., all but a finite number),
(L . Y)p = ,Up(Y).
(c) If Y is a curve of degree din P 2 , and if Lis a line in P 2 , L # Y, show that
(L · Y) = d. Here we define (L · Y) = L)L · Y)p taken over all points P E
L n Y, where (L · Y)p is defined using a suitable affine cover ofP 2 .

36
 5 N onsingular Varieties

5.5. For every degree d > 0, and every p = 0 or a prime number, give the equation
of a nonsingular curve of degree d in P 2 over a field k of characteristic p.

5.6. Blowing Up Curve Singularities.

(a) Let Y be the cusp or node of (Ex. 5.1). Show that the curve Y. obtained by
blowing up Yat 0 = (0,0) is nonsingular (cf. (4.9.1) and (Ex. 4.10)).
(b) We define a node (also called ordinary double point) to be a double point
(i.e., a point of multiplicity 2) of a plane curve with distinct tangent directions
(Ex. 5.3). If P is a node on a plane curve Y, show that rp - 1 (P) consists of two
distinct nonsingular points on the blown-up curve Y. We say that "blowing
up P resolves the singularity at P".
(c) Let P E Y be the tacnode of(Ex. 5.1). If rp: Y--> Y is the blowing-up at P, show
that rp - 1(P) is a node. Using (b) we see that the tacnode can be resolved by
two successive blowings-up.
(d) Let Y be the plane curve y 3 = x 5 , which has a "higher order cusp" at 0. Show
that 0 is a triple point; that blowing up 0 gives rise to a double point (what
kind?) and that one further blowing up resolves the singularity.
Note: We will see later (V, 3.8) that any singular point of a plane curve can be
resolved by a finite sequence of successive blowings-up.

5.7. Let Y <::::: P 2 be a nonsingular plane curve of degree > 1, defined by the equation
f(x,y,z) = 0. Let X <::::: A 3 be the affine variety defined by f (this is the cone
over Y; see (Ex. 2.1 0) ). Let P be the point (0,0,0), which is the vertex of the cone.
Let rp:X--> X be the blowing-up of X at P.
(a) Show that X has just one singular point, namely P.
(b) Show that X is nonsingular (cover it with open affines).
(c) Show that rp - 1(P) is isomorphic to Y.
5.8. Let Y <::::: P" be a projective variety of dimension r. Let f 1, ... ,j, E S =
k[ x 0 , . •• ,x"] be homogeneous polynomials which generate the ideal of Y. Let
P E Y be a point, with homogeneous coordinates P = (a 0 , . .• ,a"). Show that
Pis nonsingular on Y if and only if the rank of the matrix jj(8};/8xj)(a 0 , . . . ,anlll
is n - r. [Hint: (a) Show that this rank is independent of the homogeneous
coordinates chosen for P; (b) pass to an open affine U, <::::: P" containing P and
use the affine Jacobian matrix; (c) you will need Euler's lemma, which says that
iff is a homogeneous polynomial of degree d, then I,x;(3fj3x;) = d · f]
5.9. Let f E k[x,y,z] be a homogeneous polynomial, let Y = Z(f).::::: P 2 be the
algebraic set defined by f, and suppose that for every P E Y, at least one of
w;ax)(P), W/3y)(P), w;az)(P) is nonzero. Show that f is irreducible (and hence
that Y is a nonsingular variety). [Hint: Use (Ex. 3.7).]
5.10. For a point P on a variety X, let m be the maximal ideal of the local ring {!)P·
We define the Zariski tangent space Tp(X) of X at P to be the dual k-vector space
ofm/m 2 .
(a) For any point P EX, dim Tp(X) ;::, dim X, with equality if and only if Pis
nonsingular.
(b) For any morphism rp:X--> Y, there is a natural induced k-linear map Tp(rp):
T p(X) --> T cp<Pl( Y).
(c) If rp is the vertical projection of the parabola x = y 2 onto the x-axis, show that
the induced map T 0 (rp) of tangent spaces at the origin is the zero map.

37
I Varieties

5.11. The Elliptic Quartic Curve in P 3 . Let Y be the algebraic set in P 3 defined by the
equations x 2 - xz - yw = 0 and yz - xw - zw = 0. Let P be the point
(x,y,z,w) = (0,0,0,1), and let cp denote the projection from P to the plane w = 0.
Show that cp induces an isomorphism of Y - P with the plane cubic curve
y 2 z - x 3 + xz 2 = 0 minus the point (1,0,-1). Then show that Y is an irre-
ducible nonsingular curve. It is called the elliptic quartic curve in P 3 • Since it
is defined by two equations it is another example of a complete intersection
(Ex. 2.17).

5.12. Quadric Hypersurfaces. Assume char k i= 2, and let f be a homogeneous poly-

nomial of degree 2 in x 0 , •.. ,x•.
(a) Show that after a suitable linear change of variables,! can be brought into the
"form f = x6 + ... + x; for some 0 ~ r ~ n.
(b) Show that f is irreducible if and only if r ~ 2.
(c) Assume r ~ 2, and let Q be the quadric hypersurface in P" defined by f Show
that the singular locus Z = Sing Q of Q is a linear variety (Ex. 2.11) of dimen-
sion n - r - 1. In particular, Q is nonsingular if and only if r = n.
(d) In case r < n, show that Q is a cone with axis Z over a nonsingular quadric
hypersurface Q' s:::: P'. (This notion of cone generalizes the one defined in
(Ex. 2.10). If Y is a closed subset of P', and if Z is a linear subspace of dimen-
sion n - r - 1 in P", we embed P' in P" so that P' n Z = 0, and define
the cone over Y with axis Z to be the union of all lines joining a point of Y
to a point of Z.)
5.13. It is a fact that any regular local ring is an integrally closed domain (Matsumura
[2, Th. 36, p. 121 ]). Thus we see from (5.3) that any variety has a nonempty
open subset of normal points (Ex. 3.17). In this exercise, show directly (without
using (5.3)) that the set of nonnormal points of a variety is a proper closed sub-
set (you will need the finiteness of integral closure: see (3.9A) ).
5.14. Analytically Isomorphic Singularities.
(a) If P E Y and Q E Z are analytically isomorphic plane C'lrve singularities, show
that the multiplicities Jlp(Y) and Jlq(Z) are the same (Ex. 5.3).
(b) Generalize the example in the text (5.6.3) to show that iff = f.. + f..+ 1 + ... E
k[[ x, y]], and if the leading form f.. off factors as f.. = gsh,, where g.,h, are
homogeneous of degrees s and t respectively, and have no common linear
factor, then there are formal power series

g = Os + 9s+ 1 + · · ·
h=h,+ht+1+ ...
ink[[ x,y ]] such that f = gh.
(c) Let Ybedefinedbytheequationf(x,y) = OinA 2 ,andletP = (O,O)beapoint
of multiplicity r on Y, so that when f is expanded as a polynomial in x and y,
we have f = f.. + higher terms. We say that Pis an ordinary r-fold point if
f.. is a product of r distinct linear factors. Show that any two ordinary double
points are analytically isomorphic. Ditto for ordinary triple points. But show
that there is a one-parameter family of mutually nonisomorphic ordinary
4-fold points.
*(d) Assume char k i= 2. Show that any double point of a plane curve is analy-
tically isomorphic to the singularity at (0,0) of the curve i = x', for a uniquely

38
 6 Nonsingular Curves

determined r ~ 2. If r = 2 it is a node (Ex. 5.6). If r = 3 we call it a cusp;

if r = 4 a tacnode. See (V, 3.9.5) for further discussion.
5.15. Families of Plane Curves. A homogeneous polynomial f of degree d in three
variables x,y,z has (di 2 ) coefficients. Let these coefficients represent a point in
PN, where N = (di 2) - 1 = ±d(d + 3).
(a) Show that this gives a correspondence between points of pN and algebraic
sets in P 2 which can be defined by an equation of degree d. The correspondence
is 1-1 except in some cases where f has a multiple factor.
(b) Show under this correspondence that the (irreducible) nonsingular curves of
degree d correspond 1-1 to the points of a nonempty Zariski-open subset of
PN. [Hints: (1) Use elimination theory (5.7A) applied to the homogeneous
polynomials 8f/8x 0 , . • . ,8f/8x"; (2) use the previous (Ex. 5.5, 5.8, 5.9) above.]

6 Nonsingular Curves

In considering the problem of classification of algebraic varieties, we can

formulate several subproblems, based on the idea that a nonsingular pro-
jective variety is the best kind: (a) classify varieties up to birational equiva-
lence; (b) within each birational equivalence class, find a nonsingular
projective variety; (c) classify the nonsingular projective varieties in a given
birational equivalence class.
In general, all three problems are very difficult. However, in the case of
curves, the situation is much simpler. In this section we will answer problems
(b) and (c) by showing that in each birational equivalence class, there is a
unique nonsingular projective curve. We will also give an example to show
that not all curves are birationally equivalent to each other (Ex. 6.2). Thus
for a given finitely generated extension field K of k of transcendence degree 1
(which we will call a function field of dimension 1) we can talk about the
nonsingular projective curve CK with function field equal to K. We will see
also that if K 1 ,K 2 are two function fields of dimension 1, then any k-homo-
morphism K 2 ~ K 1 is represented by a morphism of CK, to CKz·
We will begin our study in an oblique manner by defining the notion of an
"abstract nonsingular curve" associated with a given function field. It will
not be clear a priori that this is a variety. However, we will see in retrospect
that we have defined nothing new.
First we have to recall some basic facts about valuation rings and Dede-
kind domains.

Definition. Let K be a field and let G be a totally ordered abelian group. A

valuation of K with values in G is a map v:K - {0} ~ G such that for all
x,y E K, x,y i= 0, we have:

(1) v(xy) = v(x) + v(y);

(2) v(x + y) ;;:;, min(v(x),v(y) ).

39
I Varieties

Ifv is a valuation, then the set R = {xEKJv(x) ~ 0} u {0} is a subring of K,

which we call the valuation ring ofv. The subset m = {xEKJv(x) > 0} u
{0} is an ideal in R, and R,m is a local ring. A valuation ring is an integral
domain which is the valuation ring of some valuation of its quotient field.
If R is a valuation ring with quotient field K, we say that R is a valuation
ring of K. If k is a subfield of K such that v(x) = 0 for all x E k - {0},
then we say vis a valuation of Kjk, and R is a valuation ring of Kjk. (Note
that valuation rings are not in general noetherian!)

Definition. If A,B are local rings contained in a .field K, we say that B dominates
A if A S:::: Band m 8 n A = rnA-

Theorem 6.1A. Let K be afield. A local ring R contained inK is a valuation

ring of K if and only if it is a maximal element of the set of local rings con-
tained in K, with respect to the relation of domination. Every local ring
contained in K is dominated by some valuation ring of K.
PROOF. Bourbaki [2, Ch. VI, §1, 3] or Atiyah-Macdonald [1, Ch. 5, p. 65,
and exercises, p. 72].

Definition. A valuation v is discrete if its value group G is the integers. The

corresponding valuation ring is called a discrete valuation ring.

Theorem 6.2A. Let A be a noetherian local domain of dimension one, with

maximal ideal m. Then the following conditions are equivalent:
(i) A is a discrete valuation ring;
(ii) A is integrally closed;
(iii) A is a regular local ring;
(iv) m is a principal ideal.
PROOF. Atiyah-Macdonald [1, Prop. 9.2, p. 94].

Definition. A Dedekind domain is an integrally closed noetherian domain of

dimension one.

Because integral closure is a local property (Atiyah-Macdonald [1, Prop.

5.13, p. 63]), every localization of a Dedekind domain at a nonzero prime
ideal is a discrete valuation ring.

Theorem 6.3A. The integral closure of a De de kind domain in a finite extension

field of its quotient field is again a Dedekind domain.
PROOF. Zariski-Samuel [1, vol. 1, Th. 19, p. 281].

We now turn to the case of a function field K of dimension 1 over k, where

k is our fixed algebraically closed base field. We wish to establish a connection

40
 6 Nonsingular Curves

between non-singular curves with function field K and the set of discrete
valuation rings of Kjk. If Pis a point on a nonsingular curve Y, then by (5.1)
the local ring (!JP is a regular local ring of dimension one, and so by (6.2A) it
is a discrete valuation ring. Its quotient field is the function field K of Y,
and since k ~ (!Jp, it is a valuation ring of Kjk. Thus the local rings of Y
define a subset of the set CK of all discrete valuation rings of Kjk. This
motivates the definition of an abstract nonsingular curve below. But first
we need a few more preliminaries.

Lemma 6.4. Let Y be a quasi-projective variety, let P,Q E Y, and suppose that
(!JQ ~ (!JP as subrings of K(Y). Then P = Q.
PROOF. Embed Y in pn for some n. Replacing Y by its closure, we may
assume Y is projective. After a suitable linear change of coordinates in pn,
we may assume that neither P nor Q is in the hyperplane H 0 defined by
x 0 = 0. Thus P,Q E Y n (Pn - H 0 ) which is affine, so we may assume that
Y is an affine variety.
Let A be the affine ring of Y. Then there are maximal ideals m,n ~ A
such that (!JP = A"' and (!JQ = Alt. If (!JQ ~ (!Jp, we must have m ~ n. But
m is a maximal ideal, so m = n, hence P = Q, by (3.2b).

Lemma 6.5. Let K be a function field of dimension one over k, and let x E K.
Then {R E CKix ¢ R} is a finite set.
PROOF. If R is a valuation ring, then x ¢ R if and only if 1/x E mR. So letting
y = 1/x, we have to show that if y E K, y # 0, then {R E CKIY E mR} is a
finite set. If y E k, there are no such R, so let us assume y ¢ k.
We consider the sub ring k[ yJof K generated by y. Since k is algebraically
closed, y is transcendental over k, hence k[y] is a polynomial ring. Further-
more, since K is finitely generated and of transcendence degree 1 over k,
K is a finite field extension of k(y). Now let B be the integral closure of
k[y] inK. Then by (6.3A), B is a Dedekind domain, and it is also a finitely
generated k-algebra (3.9A).
Now if y is contained in a discrete valuation ring R of Kjk, then k[y J ~ R,
and since R is integrally closed in K, we have B ~ R. Let n = mR n B.
Then n is a maximal ideal of B, and B is dominated by R. But Bit is also a
discrete valuation ring of Kjk, hence Bit = R by the maximality of valuation
rings (6.1A).
If furthermore y E mR, then yEn. Now B is the affine coordinate ring
of some affine variety Y (1.4.6). Since B is a Dedekind domain, Y has di-
mension one and is nonsingular. To say that yEn says that y, as a regular
function on Y, vanishes at the point of Y corresponding to n. But y # 0,
so it vanishes only at a finite set of points; these are in 1-1 correspondence
with the maximal ideals of B by (3.2), and R = Bit is determined by the
maximal ideal n. Hence we conclude that y E mR for only finitely many
RECK, as required.

41
I Varieties

Corollary 6.6. Any discrete valuation ring of Kjk is isomorphic to the local
ring of a point on some nonsingular affine curve.
PROOF. Given R, let y E R - k. Then the construction used in the proof
of (6.5) gives such a curve.

We now come to the definition of an abstract nonsingular curve. Let

K be a function field of dimension 1 over k (i.e., a finitely generated exten-
sion field of transcendence degree 1). Let CK be the set of all discrete valua-
tion rings of Kjk. We will sometimes call the elements of CK points, and
write P E CK, where P stands for the valuation ring Rp. Note that the set
CK is infinite, because it contains all the local rings of any nonsingular
curve with function field K; those local rings are all distinct (6.4), and there
are infinitely many of them (Ex. 4.8). We make CK into a topological space
by taking the closed sets to be the finite subsets and the whole space. If
U c:; C K is an open subset of C K• we define the ring of regular functions
on u to be (!J(U) = nPeU Rp. An element/ E (!J(U) defines a function from
U to k by taking f(P) to be the residue off modulo the maximal ideal of
Rp. (Note by (6.6) that for any RECK, the residue field of R is k.) If two
elements f,g E @(U) define the same function, then f - g E mp for infi-
nitely many PECK, so by (6.5) and its proof, f = g. Thus we can identify
the elements of @(U) with functions from U to k. Note also by (6.5) that
any f E K is a regular function on some open set U. Thus the function
field of CK, defined as in §3, is just K.

Definition. An abstract nonsingular curve is an open subset U c:; CK, where

K is a function field of dimension 1 over k, with the induced topology,
and the induced notion of regular functions on its open subsets.

Note that it is not clear a priori that such an abstract curve is a variety.
So we will enlarge the category of varieties by adjoining the abstract curves:

Definition. A morphism q>: X ~ Y between abstract nonsingular curves or

varieties is a continuous mapping such that for every open set V c:; Y,
and every regular function f: V ~ k, f o q> is a regular function on
q> -l(V).

Now that we have apparently enlarged our category, our task will be
to show that every nonsingular quasi-projective curve is isomorphic to an
abstract nonsingular curve, and conversely. In particular, we will show
that CK itself is isomorphic to a nonsingular projective curve.

Proposition 6.7. Every nonsingular quasi-projective curve Y is isomorphic

to an abstract nonsingular curve.

42
 6 Nonsingular Curves

PROOF. Let K be the function field of Y. Then each local ring (!JP of a point
P E Y is a discrete valuation ring of Kjk, by (5.1) and (6.2A). Furthermore,
by (6.4), distinct points give rise to distinct sub rings of K. So let U s; CK
be the set of local rings of Y, and let <p: Y ---> U be the bijective map defined
by <p(P) = @p.
First, we need to show that U is an open subset of CK. Because open
sets are complements of finite sets, it is sufficient to show that U contains a
nonempty open set. Thus, by (4.3), we may assume Y is affine, with affine
ring A. Then A is a finitely generated k-algebra, and by (3.2), K is the
quotient field of A, and U is the set of localizations of A at its maximal
ideals. Since these local rings are all discrete valuation rings, U consists in
fact of all discrete valuation rings of Kjk containing A. Now let xt> ... ,x"

n
be a set of generators of A over k. Then A s; Rp if and only if x 1 , . . . ,
Xn E Rp. Thus u = U;, where U; = {P E CKixi E Rp }. But by (6.5),
{P E CKix; ¢ Rp} is a finite set. Therefore· each U; and hence also U is open.
So we have shown that the U defined above is an abstract nonsingular
curve. To show that <p is an isomorphism, we need only check that the
regular functions on any open set are the same. But this follows from the
definition of the regular functions on U and the fact that for any open set
v s; Y, G(V) = nPeV GP,Y·

Now we need a result about extensions of morphisms from curves to

projective varieties, which is interesting in its own right.

Proposition 6.8. Let X be an abstract nonsingular curve, let P E X, let Y be

a projective variety, and let <p:X - P---> Y be a morphism. Then there
exists a unique morphism q5: X ---> Y extending <p.

PROOF. Embed Y as a closed subset of P" for some n. Then it will be suffi-
cient to show that <p extends to a morphism of X into P", because if it does,
the image is necessarily contained in Y. Thus we reduce to the case Y = P".
Let pn have homogeneous coordinates x 0 , . . . ,x", and let U be the open
set where x 0 , . . . ,xn are all nonzero. By using induction on n, we may
assume that <p(X - P) n U =/= 0. Because if <p(X - P) n U = 0. then
<p(X - P) s; P" - U. But pn - U is the union of the hyperplanes H;
defined by X; = 0. Since <p(X - P) is irreducible, it must be contained in
H; for some i. Now H; ~ pn-1, so the result would follow by induction.
So we will assume that <p(X - P) n U =1- 0.
For each i,j, x;/xi is a regular function on U. Pulling it back by <p, we
obtain a regular function fii on an open subset of X, which we view as a
rational function on X, i.e., fii E K, where K is the function field of X.
Let v be the valuation of K associated with the valuation ring Rp. Let
r; = v(f;o), i = 0,1, ... ,n, r; E Z. Then since x;/xi = (x;/x 0 )/(x)x 0 ), we have

i,j = 0, ... ,n.

43
I Varieties

Choose k such that rk is minimal among r 0 , . .. ,rn- Then v(};d ~ 0 for all i,
hence j~b ... Jnk E Rp. Now define ip(P) = (f0k(P), ... ,j,k(P) ), and ip(Q) =
cp(Q) for Q -=f. P. I claim that ip is a morphism of X to pn which extends cp,
and that ip is unique. The uniqueness is clear by construction (it also follows
from (4.1) ). To show that ip is a morphism, it will be sufficient to show that
regular functions in a neighborhood of ip(P) pull back to regular functions
on X. Let uk <:; pn be the open set where xk -=f. 0. Then ip(P) E ub since
hk(P) = 1. Now Uk is affine, with affine coordinate ring equal to

k[ x 0/xb ... ,xn/xk].

These functions pull back to fob ... ,j,k which are regular at P by con-
struction. It follows immediately that for any smaller neighborhood ip(P) E
V <:; Ub regular functions on V pull back to regular functions on X. Hence
ip is a morphism, which completes the proof.

Now we come to our main result.

Theorem 6.9. Let K be a function field of dimension 1 over k. Then the

abstract nonsingular curve CK defined above is isomorphic to a nonsingular
projective curve.

PROOF. The idea of the proof is this: we first cover C = CK with open
subsets Ui which are isomorphic to nonsingular affine curves. Let Y; be
the projective closure of this affine curve. Then we use (6.8) to define a
morphism <pi: C --> Y;. Next, we consider the product mapping cp: C --> flY;,
and let Y be the closure of the image of C. Then Y is a projective curve,
and we show that cp is an isomorphism of C onto Y.
To begin with, let P E C be any point. Then by (6.6) there is a nonsingular
affine curve V and a point Q E V with Rp ~ (IJQ· It follows that the function
field of Vis K, and then by (6.7}, Vis isomorphic to an open subset of C.
Thus we have shown that every point P E C has an open neighborhood
which is isomorphic to an affine variety.
Since C is quasi-compact, we can cover it with a finite number of open
subsets Ui, each of which is isomorphic to an affine variety V;. Embed
V; <:; A"', think of An' as an open subset of pn,, and let Y; be the closure of
V; in P"'. Then Y; is a projective variety, and we have a morphism tp;: U; --> Y;
which is an isomorphism of Ui onto its image.
By (6.8) applied to the finite set of points C - U;, we can find a morphism
i[5;: C --> Y; extending <pi. Let f1 Y; be the product of the projective varieties
Y; (Ex. 3.16). Then f1Y; is also a projective variety. Let cp:C--> f1Y; be the
"diagonal" map cp(P) = f1ip;(P), and let Y be the closure of the image of
<p. Then Y is a projective variety, and cp: C --> Y is a morphism whose
image is dense in Y. (It follows that Y is a curve.)
Now we must show that cp is an isomorphism. For any point P E C, we
have P E U; for some i. There is a commutative diagram

44
 6 Nonsingular Cunt

c y

J
of dominant morphisms, where n is the projection map onto the ith factor.
Thus we have inclusions of local rings
(!) rp;(P),Y; ~ (!) rp(P),Y ~ (!) P,C

by (Ex. 3.3). The two outside ones are isomorphic, so the middle one is
also. Thus we see that for any P E C, the map cpp: (!) rp(P),Y --+ (!) P,c is an
isomorphism.
Next, let Q be any point of Y. Then (!)Q is dominated by some discrete
valuation ring R of Kjk (take for example a localization of the integral
closure of (!)Qat a maximal ideal). But R = Rp for some P E C, and (l)rp(Pl ~
R, so by (6.4) we must have Q = cp(P). This shows that cp is surjective.
But cp is clearly injective, because distinct points of C correspond to distinct
subrings of K.
Thus cp is a bijective morphism of C to Y, and for every P E C, cpp is an
isomorphism, so by (Ex. 3.3b), cp is an isomorphism.

Corollary 6.10. Every abstract nonsingular curve is isomorphic to a quasi-

projective curve. Every nonsingular quasi-projective curve is isomorphic
to an open subset of a nonsingular projective curve.

Corollary 6.11. Every curve is birationally equivalent to a nonsingular pro-

jective curve.
PROOF. Indeed, if Y is any curve, with function field K, then Y is birationally
equivalent to CK which is nonsingular and projective.

Corollary 6.12. The following three categories are equivalent:

(i) nonsingular projective curves, and dominant morphisms;
(ii) quasi-projective curves, and dominant rational maps;
(iii) function fields of dimension 1 over k, and k-homomorphisms.
PROOF. We have an obvious functor from (i) to (ii). We have the functor
Y --+ K( Y) from (ii) to (iii), which induces an equivalence of categories by
(4.4). To complete the cycle, we need a functor from (iii) to (i).
To a function field K, associate the curve CK, which by the theorem is a
projective nonsingular curve. If K 2 --+ K 1 is a homomorphism, then by
(ii) ~ (iii), it induces a rational map of the corresponding curves. This can
be represented by a morphism cp: U --+ CK 2 , where U ~ CK, is an open
subset. By (6.8) cp extends to a morphism cp:CK,--+ CK 2 • If K 3 --+ K 2 --+ K 1

45
I Varieties

are two homomorphisms, it follows from the uniqueness part of (6.8) that
the corresponding morphisms C 1 -+ C 2 -+ C 3 and C 1 -+ C 3 are compatible.
Hence K ~ CK is a functor from (iii) -+ (i). It is clearly inverse to the given
functor (i) -+ (ii) -+ (iii), so we have an equivalence of categories.

EXERCISES

6.1. Recall that a curve is rational if it is birationally equivalent to P 1 (Ex. 4.4). Let Y
be a nonsingular rational curve which is not isomorphic to P 1 .
(a) Show that Y is isomorphic to an open subset of A 1 .
(b) Show that Y is affine.
(c) Show that A(Y) is a unique factorization domain.
6.2. An Elliptic Curve. Let Y be the curve y 2 = x 3 - x in A 2 , and assume that the
characteristic of the base field k is =1= 2. In this exercise we will show that Y is not a
rational curve, and hence K(Y) is not a pure transcendental extension of k.
(a) Show that Yisnonsingular,anddeducethatA = A(Y):::::: k[x,y]j(y 2 - x 3 + x)
is an integrally closed domain.
(b) Let k[x] be the subring of K = K(Y) generated by the image of x in A. Show
that k[ x] is a polynomial ring, and that A is the integral closure of k[ x] in K.
(c) Show that there is an automorphism u: A ->A which sends y to - y and leaves
x fixed. For any a E A, define the norm of a to be N(a) = a· u(a). Show that
N(a) E k[x], N(l) = 1, and N(ab) = N(a) · N(b) for any a,b EA.
(d) Using the norm, show that the units in A are precisely the nonzero elements of
k. Show that x and y are irreducible elements of A. Show that A is not a
unique factorization domain.
(e) Prove that Y is not a rational curve (Ex. 6.1). See (II, 8.20.3) and (III, Ex. 5.3)
for other proofs of this important result.

6.3. Show by example that the result of (6.8) is false if either (a) dim X :;:, 2, or (b) Y is
not projective.
6.4. Let Y be a nonsingular projective curve. Show that every nonconstant rational
function f on Y defines a surjective morphism tp: Y -> P 1, and that for every P E P 1,
tp - 1 (P) is a finite set of points.

6.5. Let X be a nonsingular projective curve. Suppose that X is a (locally closed)

subvariety of a variety Y (Ex. 3.10). Show that X is in fact a closed subset of Y.
See (II, Ex. 4.4) for generalization.
6.6. Automorphisms ofP 1 • Think of P 1 as A 1 u { oo }. Then we define a fractional
linear transformation of P 1 by sending x f-> (ax + b)j(cx + d), for a,b,c,d E k,
ad - be =1= 0.
(a) Show that a fractional linear transformation induces an automorphism of P 1
(i.e., an isomorphism of P 1 with itself). We denote the group of all these
fractional linear transformations by PGL(l).
(b) Let Aut P 1 denote the group of all automorphisms of P 1 . Show that Aut P 1 ::::::
Aut k(x), the group of k-automorphisms of the field k(x).
(c) Now show that every automorphism of k(x) is a fractional linear transforma-
tion, and deduce that PGL(l)-> Aut P 1 is an isomorphism.

46
 7 Intersections in Projective Space

Note: We will see later (II, 7.1.1) that a similar result holds for P": every automor-
phism is given by a linear transformation of the homogeneous coordinates.
6.7. LetP 1 , . . . ,P, Q1 , . . . ,Qs be distinct points of A 1 . IfA 1 - {P 1 , . . . ,P,} is isomor-
phic to A 1 - {Q 1 , . . . ,Qs}, show that r = s. Is the converse true? Cf. (Ex. 3.1).

7 Intersections in Projective Space

The purpose of this section is to study the intersection of varieties in a

projective space. If Y, Z are varieties in pn, what can one say about Y n Z?
We have already seen (Ex. 2.16) that Y n Z need not be a variety. But it
is an algebraic set, and we can ask first about the dimensions of its irreducible
components. We take our cue from the theory of vector spaces: if U,V are
subspaces of dimensions r,s of a vector space W of dimension n, then
U n V is a subspace of dimension ~ r + s - n. Furthermore, if U and V
are in sufficiently general position, the dimension of U n V is equal to
r + s - n (provided r + s - n ~ 0). This result on vector spaces imme-
diately implies the analogous result for linear subspaces of pn (Ex. 2.11).
Our first result in this section will be to prove that if Y,Z are subvarieties of
dimensions r,s ofPn, then every irreducible component of Y n Z has dimen-
sion ~ r + s - n. Furthermore, if r + s - n ~ 0, then Y n Z is nonempty.
Knowing something about the dimension of Y n Z, we can ask for more
precise information. Suppose for example that r + s = n, and that Y n Z
is a finite set of points. Then we can ask, how many points are there? Let
us look at a special case. If Y is a curve of degree din P 2 , and if Z is a line
in P 2 , then Y n Z consists of at most d points, and the number comes to d
exactly if we count them with appropriate multiplicities (Ex. 5.4). This
result generalizes to the well-known theorem of Bezout, which says that if
Y,Z are plane curves of degrees d,e, with Y -:f. Z, then Y n Z consists of
de points, counted with multiplicities. We will prove Bezout's theorem
later in this section (7.8).
The ideal generalization of Bezout's theorem to pn would be this. First,
define the degree of any projective variety. Let Y,Z be varieties of dimen-
sions r,s, and of degrees d,e in pn. Assume that Y and Z are in a sufficiently
general position so that all irreducible components of Y n Z have di-
mension = r + s - n, and assume that r + s - n ~ 0. For each ir-
reducible component W of Y n Z, define the intersection multiplicity
i(Y,Z; W) of Y and Z along W. Then we should have

l)(Y,Z; W) · deg W = de,

where the sum is taken over all irreducible components of Y n Z.

The hardest part of this generalization is the correct definition of the
intersection multiplicity. (And, by the way, historically it took many at-
tempts before a satisfactory treatment was given by Severi [3] geometrically

47
I Varieties

and by Chevalley [1] and Wei! [1] algebraically). We will define the inter-
section multiplicity only in the case where Z is a hypersurface. See Appendix
A for the general case.
Our main task in this section will be the definition of the degree of a
variety Y of dimension r in P". Classically, the degree of Y is defined as the
number of points of intersection of Y with a sufficiently general linear space
L of dimension 11 - r. However, this definition is difficult to use. Cutting
Y successively with r sufficiently general hyperplanes, one can find a
linear space L of dimension 11 - r which meets Y in a finite number of
points (Ex. 1.8). But the number of intersection points may depend on L,
and it is hard to make precise the notion "sufficiently general."
Therefore we will give a purely algebraic definition of degree, using the
Hilbert polynomial of a projective variety. This definition is less geo-
metrically motivated, but it has the advantage of being precise. In an
exercise we show that it agrees with the classical definition in a special case
(Ex. 7.4).

Proposition 7.1 (Affine Dimension Theorem). Let Y,Z be varieties of dimen-

sions r,s in A". Then every irreducible component W of Y n Z has
dimension ;?! r + s - n.
PROOF. We proceed in several steps. First, suppose that Z is a hypersurface,
defined by an equation f = 0. If Y [Link] Z, there is nothing to prove. If
Y ¢. Z, we must show that each irreducible component W of Y n Z has
dimension r - 1. Let A( Y) be the affine coordinate ring of Y Then the
irreducible components of Y n Z correspond to the minimal prime ideals
p of the principal ideal (f) in A(Y). Now by Krull's Hauptidealsatz (1.11A),
each such p has height one, so by the dimension theorem (1.8A), A( Y)/p
has dimension r - 1. By (1.7) this shows that each irreducible component
W has dimension r - 1.
Now for the general case. We consider the product Y x Z [Link] A 2 ",
which is a variety of dimension r + s (Ex. 3.15). Let L1 be the diagonal
{P x PIPE A"} [Link] A 2 ". ThenA"[Link] ~ P x P,
and under this isomorphism, Y n Z corresponds to ( Y x Z) n .d. Since
~ has dimension n, and since r + s - n = (r + s) + n - 2n, we reduce
to proving the result for the two varieties Y x Z and L1 in A2 ". Now L1 is
an intersection of exactly n hypersurfaces, namely, x 1 - .h = 0, ... ,x" -
Yn = 0, where xb . . . ,xn, y 1 , . . . ,yn are the coordinates of A 2 ". Now ap-
plying the special case above n times, we have the result.

Theorem 7.2 (Projective Dimension Theorem). Let Y,Z be varieties of dimen-

sions r,s in P". Then every irreducible component of Y n Z has dimension
;?! r + s - n. Furthermore, if r + s - n ;?! 0, then Y n Z is nonempty.

PROOF. The first statement follows from the previous result, since P" is
covered by affine n-spaces. For the second result, let C(Y) and C(Z) be the

48
 7 Intersections in Projective Space

cones over Y,Z in A"+ 1 (Ex. 2.10). Then C(Y), C(Z) have dimensions r + 1,
s + 1, respectively. Furthermore, C(Y) n C(Z) # 0, because both contain
the origin P = (0, ... ,0). By the affine dimension theorem, C(Y) n C(Z) has
dimension ~(r + 1) + (s + 1) - (n + 1) = r + s - n + 1 > 0. Hence
C(Y) n C(Z) contains some point Q # P, and soY n Z # 0.

Next, we come to the definition of the Hilbert polynomial of a projective

variety. The idea is to associate to each projective variety Y s;: Pi: a poly-
nomial Py E Q[ z] from which we can obtain various numerical invariants
of Y. We will define Pr starting from the homogeneous coordinate ring S(Y).
In fact, more generally, we will define a Hilbert polynomial for any graded
S-module, where S = k[x 0 , . . . ,xnJ. Although the next few results are almost
pure algebra, we include their proofs, for lack of a suitable reference.

Definition. A numerical polynomial is a polynomial P(z) E Q[ z] such that

P(n) E Z for all n » 0, n E Z.

Proposition 7.3.
(a) If P E Q[ z] is a numerical polynomial, then there are integers
c0 ,ct. ... ,c, such that

where

(z)
r
= J_ z(z
r!
- 1) · · · (z - r + 1)

is the binomial coefficient function. In particular P(n) E Z for all n E Z.

(b) If f:Z --+ Z is any function, and if there exists a numerical poly-
nomial Q(z) such that the difference function LJf = f(n + 1) - f(n) is equal
to Q(n) for all n » 0, then there exists a numerical polynomial P(z) such
that f(n) = P(n) for all n » 0.

PROOF.
(a) By induction on the degree of P, the case of degree 0 being obvious.
Since (;) = z'/r! + ... , we can express any polynomial P E Q[ z] of degree r
in the above form, with c0 , ... ,c, E Q. For any polynomial P we define the
difference polynomialLJP by LJP(z) = P(z + 1) - P(z). Since L1 (~) = {r..: d,

L1P = c (r ~ 1) + (r ~ 2) + ... +
0 C1 cr-1·

By induction, c0 , ... ,c,_ 1 E Z. But then c, E Z since P(n) E Z for n » 0.

(b) Write

Q = c0 (:) + ... + c,

49
I Varieties

with c0 , ... ,c, E Z. Let

Then L1P = Q, so L1(f - P)(n) = 0 for all n » 0, so (f - P)(n) =

constant c,+ 1 for all n » 0, so
f(n) = P(n) + c,+ 1
for all n » 0, as required.

Next, we need some preparations about graded modules. Let S be a

graded ring (cf. §2). A graded S-module is an S-module M, together with a
decomposition M = (£ldEZ Ma, such that Sa· Me ~ Md+e· For any graded
S-module M, and for any IE Z, we define the twisted module M(l) by shifting
l places to the left, i.e., M(/)d = Md+l· If M is a graded S-module, we define
the annihilator of M, Ann M = {s E Sis· M = 0} .. This is a homogeneous
ideal inS.
The next result is the analogue for graded modules of a well-known result
for modules of finite type over a noetherian ring (Bourbaki [1, Ch. IV,
§1, no. 4] or Matsumura [2, p. 51]). Again, we include the proof for lack
of an adequate reference.

Proposition 7.4. Let M be [Link] generated graded module over a noetherian

graded ringS. Then there exists a .filtration 0 = M 0 ~ M 1 ~ . . . ~ M' =
M by graded submodules, such that for each i, Mi/Mi- 1 ~ (S/p;)(l;),
where Pi is a homogeneous prime ideal of S, and liE Z. The filtration is
not unique, but for any such .filtration we do have:
(a) if p is a homogeneous prime ideal of S, then .p ~ Ann M <o> .p ~ .Pi
for some i. In particular, the minimal elements of the set {.p b . . . ,.p,} are
just the minimal primes of M, i.e., the primes which are minimal containing
AnnM;
(b) for each minimal prime of M, the number of times which .p occurs
in the set {p 1 , . . . ,p,} is equal to the length of Mp over the local ring Sp
(and hence is independent of the filtration).
PROOF. For the existence of the filtration, we consider the set of graded
submodules of M which admit such a filtration. Clearly, the zero module
does, so the set is nonempty. M is a noetherian module, so there is a maximal
such submodule M' ~ M. Now consider M" = MjM'. If M" = 0, we are
done. If not, we consider the set of ideals -3 = {Im = Ann(m)lm EM" is a
homogeneous element, m #- 0}. Each Im is a homogeneous ideal, and Jm #- S.
Since S is a noetherian ring, we can find an element mE M", m #- 0, such
that Im is a maximal element of the set -3. I claim that Im is a prime ideal.
Let a,b E S. Suppose that abE Jm, but b ¢ Im. We wish to show a E Jm. By
splitting into homogeneous components, we may assume that a,b are homo-
geneous elements. Now consider the element bm EM". Since b ¢ Im,

50
 7 Intersections in Projective Space

bm =1- 0. We have Im ~ Ibm' so by maximality of lm, Im = Ibm· But abE Im,

so abm = 0, so a E Ibm = Im as required. Thus Im is a homogeneous prime
ideal of S. Call it p. Let m have degree l. Then the module N ~ M" generated
by m is isomorphic to (S/p )( -1). Let N' ~ M be the inverse image of N in M.
Then M' ~ N', and N'/M' ~ (S/p)( -1). So N' also has a filtration of the
type required. This contradicts the maximality of M'. We conclude that M'
was equal to M, which proves the existence of the filtration.
Now suppose given such a filtration of M. Then it is clear that p 2
Ann M <o:> p 2 Ann(M;/M;- 1 ) for some i. But Ann((S/p;)(l)) = p; so this
proves (a).
To prove (b) we localize at a minimal prime p. Since p is minimal in the
set {PI> ... ,p,}, after localization, we will have M!, = M/,- 1 except in the
cases where P; = p. And in those cases M~/M/,- 1 ~ (S/p)p = k(p), the
quotient field of Sjp (we forget the grading). This shows that MP is an
Sp-module of finite length equal to the number of times p occurs in the set
{p1, ... ,p,}.

Definition. If p is a minimal prime of a graded S-module M, we define the

multiplicity of M at p, denoted flp(M), to be the length of MP over Sp.
Now we can define the Hilbert polynomial of a graded module Mover the
polynomial ring S = k[ x 0 , . . . ,xnJ. First, we define the Hilbert function
CfJM of M, given by

for each l E Z.

Theorem 7.5. (Hilbert-Serre). Let M be a finitely generated graded S =

k[ x 0 , . . . ,xnJ-module. Then there is a unique polynomial PM(z) E Q[ z]
such that cpM(l) = PM(l) for all l » 0. Furthermore, deg PM(z) =
dim Z(Ann M), where Z denotes the zero set in pn of a homogeneous
ideal (cf. §2).
PROOF. If 0---+ M' ---+ M ---+ M" ---+ 0 is a short exact sequence, then CfJM =
CfJM' + CfJM"' and Z(Ann M) = Z(Ann M') u Z(Ann M"), so if the theorem
is true forM' and M", it is also true forM. By (7.4), M has a filtration with
quotients of the form (Sjp)(l) where p is a homogeneous prime ideal, and
l E Z. So we reduce to M ~ (S/p)(l), The shift l corresponds to a change
of variables z f--+ z + l, so it is sufficient to consider the case M = Sjp. If
p = (x 0 , . , . ,xn), then CfJM(l) = 0 for l > 0, so PM = 0 is the corresponding
polynomial, and deg PM = dim Z(p), where we make the convention that
the zero polynomial has degree -1, and the empty set has dimension - 1.
If p =1- (x 0 , . . . ,xn), choose X; ¢ p, and consider the exact sequence
0---+ M ~ M---+ M" ---+ 0, where M" = M/x;M, Then CfJM"(l) = CfJM(l) -
CfJM(l- 1) = (LJcpM)(l- 1). On the other hand, Z(Ann M") = Z(p) n H,
where His the hyperplane X; = 0, and Z(p) '*' H by choice of X;, so by (7.2),
dim Z(Ann M") = dim Z(p) - 1. Now using induction on dim Z(Ann M),

51
I Varieties

we may assume that CfJM" is a polynomial function, corresponding to a poly-

nomial PM" of degree = dim Z(Ann M"). Now, by (7.3), it follows that CfJM
is a polynomial function, corresponding to a polynomial of degree =
dim Z(p). The uniqueness of PM is clear.

Definition. The polynomial PM of the theorem is the Hilbert polynomial of M.

Definition. If Y <:; P" is an algebraic set of dimension r, we define the Hilbert

polynomial of Y to be the Hilbert polynomial Py of its homogeneous
coordinate ring S(Y). (By the theorem, it is a polynomial of degree r.)
We define the degree of Y to be r! times the leading coefficient of Py.

Proposition 7.6.
(a) If Y <:; P", Y =I= 0, then the degree of Y is a positive integer.
(b) Let Y = Y1 u Y2 , where Y1 and Y2 have the same dimension r, and
where dim(Y1 n Y2 ) < r. Then deg Y = deg Y1 + deg Y2 •
(c) deg P" = 1.
(d) If H <:; P" is a hypersurface whose ideal is generated by a homo-
geneous polynomial of degree d, then deg H = d. (In other words, this
definition of degree is consistent with the degree of a hyperswface as defined
earlier (1.4.2).)
PROOF.
(a) Since Y =I= 0, Py is a nonzero polynomial of degree r = dim Y. By
(7.3a), deg Y = c0 , which is an integer. It is a positive integer because for
l » 0, Py(l) = CfJs1Al) ~ 0.
(b) Let It. I 2 be the ideals of Y1 and Y2 . Then I = I 1 n I 2 is the ideal of
Y. We have an exact sequence
0---> Sji ~ S/I 1 EB S/I 2 ---> S/(1 1 + I 2 )---> 0.
NowZ(I 1 + I 2 ) = Y1 n Y2 ,whichhassmallerdimension . HencePs;u,+J 2 l
has degree <r. So the leading coefficient of Ps 11 is the sum of the leading
coefficients of Ps 11 , and Ps;lo'
(c) We calculate the Hilbert polynomial of Pn. It is the polynomial Ps,
where s = k[xo, 0 ,xnJ. For l > 0, C(Js(l) = e~"), soPs = (Z~"). In partic-
0 0

ular, its leading coefficient is 1/n !, so deg P" = 1.

(d) Iff E Sis homogeneous of degree d, then we have an exact sequence
of graded S-modules
J
0 ---> S(- d) ---> S ---> S/(f) ---> 0.
Hence
CfJs;(fj(/) = CfJs(l) - CfJsU - d).
Therefore we can find the Hilbert polynomial of H, as

PH(z) = (z: n)- (z- ~ + n) = (n ~ 1)! zn-1 + ....

Thus deg H = d.

52
 7 Intersections in Projective Space

Now we come to our main result about the intersection of a projective

variety with a hypersurface, which is a partial gen-eralization of Bezout's
theorem to higher projective spaces. Let Y c:; P" be a projective variety
of dimension r. Let H be a hypersurface not containing Y Then, by (7.2),
Y n H = Z 1 u ... u Z, where Zi are varieties of dimension r - 1. Let
Pi be the homogeneous prime ideal of Zi. We define the intersection multi-
plicity of Yand H along Zi to be i(Y,H; Z) = [Link],(Sf(/y + lu)). Here lrJH
are the homogeneous ideals of Y and H. The module M = Sj(/y + /H) has
annihilator /y + IH, and Z(/y + /H) = Y n H, so pi is a minimal prime of
M, and f.1 is the multiplicity introduced above.

Theorem 7.7. Let Y be a variety of dimension ~ 1 in P", and let H be a hyper-

surface not containing Y. Let Z 1, ... ,Zs be the irreducible components
of Yn H. Then
s
L i(Y,H; Z). deg zj = (deg Y)(deg H).
j= 1

[Link] H be defined by the homogeneous polynomial f of degree d.

We consider the exact sequence of graded S-modules
0--> (Sjly)( -d)!.. Sflr--> M--> 0,

where M = Sj(I y + I H). Taking Hilbert polynomials, we find that

PM(z) = Py(z) - Py(z - d).

Our result comes from comparing the leading coefficients of both sides
of this equation. Let Y have dimension r and degree e. Then Py(z) =
(ejr!)z' + ... so on the right we have

(ejr!)z' + ... - [(ejr!)(z - d)' + ... ] = (dej(r - l)!)z'- 1 + ....

Now consider the module M. By (7.4), M has a filtration 0 = M 0 c:; M 1 c:;
... c:; Mq = M, whose quotients Mi/Mi- 1 are of the form (S/q;)(l;). Hence
PM = L1= 1 P;, where P; is the Hilbert polynomial of (S/q;)(l;). If Z(q;) is
a projective variety of dimension r; and degree/;, then
P; = (/;/r; !)z'' + ....
Note that the shift I; does not affect the leading coefficient of P;. Since we
are interested only in the leading coefficient of P;, we can ignore those P;
of degree < r - 1. We are left with those P;, where q; is a minimal prime of
M, namely, one of the primes p 1 , . . . ,ps corresponding to the Zi. Each one
of these occurs f.1p1 (M) times, so the leading coefficient of PM is

(t 1
i(Y,H; Zi) · deg zi)/(r- 1)!

Comparing with the above, we have our result.

53
I Varieties

Corollary 7.8 (Bezout's Theorem). Let Y,Z be distinct curves in P 2 , having

degrees d,e. Let Y n Z = {P 1 , . . . ,P5 }. Then
2)( Y,Z; Pi) = de.
PROOF. We have only to observe that a point has Hilbert polynomial 1,
hence degree 1. See (V, 1.4.2) for another proof.

Remark 7.8.1. Our definition of intersection multiplicity in terms of the

homogeneous coordinate ring is different from the local definition given
earlier (Ex. 5.4). However, it is easy to show that they coincide in the case of
intersections of plane curves.

Remark 7.8.2. The proof of (7.8) extends easily to the case where Y and Z are
"reducible curves," i.e., algebraic sets of dimension 1 in P 2 , provided they
have no irreducible component in common.

EXERCISES

7.1. (a) Find the degree of the d-uple embedding ofP" in pN (Ex. 2.12). [Answer: d"]
(b) Find the degree of the Segre embedding ofP' x ps in pN (Ex. 2.14). [Answer:
('~s)J

7.2. Let Y be a variety of dimension r in P", with Hilbert polynomial Py. We define
the arithmetic genus of Y to be p.(Y) = ( -1)'(Py(0) - 1). This is an important
invariant which (as we will see later in (Ill, Ex. 5.3)) is independent of the projective
embedding of Y
(a) Show that p.(P") = 0.
(b) If Y is a plane curve of degree d, show that p.(Y) = !(d - 1)(d - 2).
(c) More generally, if His a hypersurface of degree din P", then p.(H) = (d~ 1 ).
(d) If Y is a complete intersection (Ex. 2.17) of surfaces of degrees a,b in P 3 , then
p.(Y) = !ab(a + b - 4) + 1.
(e) Let Y' c:; P", zs c:; pm be projective varieties, and embed Y x Z c:; P" x
pm ..... pN by the Segre embedding. Show that

7.3. The Dual Curve. Let Y c:; P 2 be a curve. We regard the set oflines in P 2 as another
projective space, (P 2 )*, by taking (a 0 ,a 1 ,a 2 ) as homogeneous coordinates of the
line L:a 0 x 0 + a 1 x 1 + a2 x 2 = 0. For each nonsingular point P E Y, show that
there is a unique line Tp(Y) whose intersection multiplicity with Y at Pis > 1.
This is the tangent line to Y at P. Show that the mapping P r--> Tp(Y) defines a
morphism of Reg Y (the set of nonsingular points of Y) into (P 2 )*. The closure of
the image of this morphism is called the dual curve Y* c:; (P 2 )* of Y
7.4. Given a curve Y of degree din P 2 , show that there is a nonempty open subset U of
(P 2 )* in its Zariski topology such that for each L E U,L meets Yin exactly d points.
[Hint: Show that the set of lines in (P 2 )* which are either tangent to Y or pass
through a singular point of Y is contained in a proper closed subset.] This result
shows that we could have defined the degree of Y to be the number d such that
almost all lines in P 2 meet Y in d points, where "almost all" refers to a nonempty

54
 8 What Is Algebraic Geometry?

open set of the set of lines, when this set is identified with the dual projective space
(P2)*.

7.5. (a) Show that an irreducible curve Y of degree d > 1 in P 2 cannot have a point of
multiplicity ~ d (Ex. 5.3).
(b) If Y is an irreducible curve of degree d > 1 having a point of multiplicity
d - 1, then Y is a rational curve (Ex. 6.1).

7.6. Linear Varieties. Show that an algebraic set Y of pure dimension r (i.e., every
irreducible component of Y has dimension r) has degree 1 if and only if Y is a
linear variety (Ex. 2.11). [Hint: First, use (7.7) and treat the case dim Y = 1. Then
do the general case by cutting with a hyperplane and using induction.]
7.7. Let Ybe a variety of dimension rand degree d > 1 in P". Let P E Ybe a nonsingular
point. Define X to be the closure of the union of all lines PQ, where Q E Y, Q #- P.
(a) Show that X is a variety of dimension r + 1.
(b) Show that deg X < d. [Hint: Use induction on dim Y.J
7.8. Let Y' <;; P" be a variety of degree 2. Show that Y is contained in a linear subspace
L of dimension r + 1 in P". Thus Y is isomorphic to a quadric hypersurface in
pr+ 1 (Ex. 5.12).

8 What Is Algebraic Geometry?

Now that we have met some algebraic varieties, and have encountered some
of the main concepts about them, it is appropriate to ask, what is this subject
all about? What are the important problems in the field, and where is it
going?
To define algebraic geometry, we could say that it is the study of the
solutions of systems of polynomial equations in an affine or projective
n-space. In other words, it is the study of algebraic varieties.
In any branch of mathematics, there are usually guiding problems, which
are so difficult that one never expects to solve them completely, yet which
provide stimulus for a great amount of work, and which serve as yardsticks for
measuring progress in the field. In algebraic geometry such a problem is the
classification problem. In its strongest form, the problem is to classify all
algebraic varieties up to isomorphism. We can divide the problem into
parts. The first part is to classify varieties up to birational equivalence. As
we have seen, this is equivalent to the question of classifying function fields
(finitely generated extension fields) over k up to isomorphism. The second
part is to identify a good subset of a birational equivalence class, such as the
nonsingular projective varieties, and classify them up to isomorphism. The
third part is to study how far an arbitrary variety is from one of the good
ones considered above. In particular, we want to know (a) how much do you
have to add to a nonprojective variety to get a projective variety, and (b)
what is the structure of singularities, and how can they be resolved to give a
nonsingular variety?

55
I Varieties

Typically, the answer to any classification problem in algebraic geometry

consists of a discrete part and a continuous part. So we can rephrase the
problem as follows: define numerical invariants and continuous invariants
of algebraic varieties, which allow one to distinguish among nonisomorphic
varieties. Another special feature of the classification problem is that often
when there is a continuous family of nonisomorphic objects, the parameter
space can itself be given a structure of algebraic variety. This is a very power-
ful method, because then all the techniques of the subject can be applied to
the study of the parameter space as well as to the original varieties.

Let us illustrate these ideas by describing what is known about the classi-
fication of algebraic curves (over a fixed algebraically closed field k). First,
the birational classification. There is an invariant called the genus of a
curve, which is a birational invariant, and which takes on all nonnegative
values g ~ 0. For g = 0 there is exactly one birational equivalence class,
namely, that of the rational curves (i.e., those curves which are birationally
equivalent to P 1 ). For each g > 0 there is a continuous family of birational
equivalence classes, which can be parametrized by an irreducible algebraic
variety 9Jl9 , called the variety of moduli of curves of genus g, which has di-
mension 1 if g = 1, and dimension 3g - 3 if g ~ 2. Curves with g = 1 are
called elliptic curves. Thus for curves, the birational classification question
is answered by giving the genus, which is a discrete invariant, and a point on
the variety of moduli, which is a continuous invariant. See Chapter IV for
more details.
The second question for curves, namely, to describe all nonsingular pro-
jective curves in a given birational equivalence class, has a simple answer, as
we have seen, since there is exactly one.
For the third question, we know that any curve can be completed to a
projective curve by adding a finite number of points, so there is not much
more to say there. As for the classification of singularities of curves, see
(V, 3.9.4).

While we are discussing the classification problem, I would like to describe

another special case where a satisfactory answer is known, namely, the
classification of nonsingular projective surfaces within a given birational
equivalence class. In this case one knows that (1) every birational equivalence
class of surfaces has a nonsingular projective surface in it, (2) the set of
nonsingular projective surfaces with a given function field Kjk is a partially
ordered set under the relation given by the existence of a birational mor-
phism, (3) any birational morphism f: X --+ Y can be factored into a finite
number of steps, each of which is a blowing-up of a point, and (4) unless K is
rational (i.e., equal to K(P 2 )) or ruled (i.e., K is the function field of a product
P 1 x C, where C is a curve), there is a unique minimal element of this
partially ordered set, which is called the minimal model of the function field K.
(In the rational and ruled cases, there are infinitely many minimal elements,

56
 8 What Is Algebraic Geometry?

and their structure is also well-known.) The theory of minimal models is a

very beautiful branch of the theory of surfaces. The results were known to the
Italians, but were first proved in all characteristics by Zariski [ 5], [ 6]. See
Chapter V for more details.

From these remarks it should be clear that the classification problem is a

very fruitful problem to keep in mind while studying algebraic geometry.
This leads us to the next question: how does one go about defining invariants
of an algebraic variety? So far, we have defined the dimension, and for
projective varieties we have defined the Hilbert polynomial, and hence the
degree and the arithmetic genus Pa· Of course the dimension is a birational
invariant. But the degree and the Hilbert polynomial depend on the em-
bedding in projective space, so they are not even invariants under isomor-
phism of varieties. Now it happens that the arithmetic genus is an invariant
under isomorphism (III, Ex. 5.3), and is even a birational invariant in most
cases (curves, surfaces, nonsingular varieties in characteristic 0; see (V, 5.6.1) ),
but this is not at all apparent from our definition.
To go further, we must study the intrinsic geometry on a variety, which we
have not done at all yet. So, for example, we will study divisors on a variety X.
A divisor is an element ofthe free abelian group generated by the subvarieties
of codimension one. We will define linear equivalence of divisors, and then
we can form the group of divisors modulo linear equivalence, called the
Picard group of X. This is an intrinsic invariant of X. Another very important
notion is that of a differential form on a variety X. Using differential forms,
one can give an intrinsic definition of the tangent bundle and cotangent
bundle on an algebraic variety. Then one can carry over many constructions
from differential geometry to define numerical invariants. For example,
one can define the genus of a curve as the dimension of the vector space of
global differential forms on the nonsingular projective model. From this
definition it is clear that it is a birational invariant. See (II, §6,7,8).
Perhaps the most important modern technique for defining numerical
invariants is by cohomology. There are many cohomology theories, but we
will be principally concerned in this book with the cohomology of coherent
sheaves, which was introduced by Serre [3]. Cohomology is an extremely
powerful and versatile tool. Not only can it be used to define numerical
invariants (for example, the genus of a curve X can be defined as dim
H 1 (X,(()x) ), but it can be used to prove many important results which do not
apparently have any connection with cohomology, such as "Zariski's main
theorem," which has to do with the structure of birational transformations.
To set up a cohomology theory requires a lot of work, but I believe it is well
worth the effort. We will devote a whole chapter to cohomology later in the
book (Chapter III). Cohomology is also a useful vehicle for understanding
and expressing important results such as the Riemann-Roch theorem. This
theorem was known classically for curves and surfaces, but it was by using
cohomology that Hirzebruch [1 J and Grothendieck (see Borel and Serre

57
I Varieties

[1 ]) were able to clarify and generalize it to varieties of any dimension

(Appendix A).

Now that we have seen a little bit of what algebraic geometry is about, we
should discuss the degree of generality in which to develop the foundations
of the subject. In this chapter we have worked over an algebraically closed
field, because that is the simplest case. But there are good reasons for allowing
fields which are not algebraically closed. One reason is that the local ring
of a subvariety on a variety has a residue field which is not algebraically
closed (Ex. 3.13), and at times it is desirable to give a unified treatment of
properties which hold along a subvariety and properties which hold at a
point. Another strong reason for allowing non-algebraically closed fields
is that many problems in algebraic geometry are motivated by number
theory, and in number theory one is primarily concerned with solutions of
equations over finite fields or number fields. For example, Fermat's problem
is equivalent to the question, does the curve x" + y" = z" in P 2 for n ~ 3
have any points rational over Q (i.e., points whose coordinates are in Q),
with x,y,z # 0.
The need to work over arbitrary ground fields was recognized by Zariski
and Weil. In fact, perhaps one of the principal contributions of Weil's
"Foundations" [1] was to provide a systematic framework for studying
varieties over arbitrary fields, and the various phenomena which occur
with change of ground field. Nagata [2] went further by developing the
foundations of algebraic geometry over Dedekind domains.
Another direction in which we need to expand our foundations is to define
some kind of abstract variety which does not a priori have an embedding in an
affine or projective space. This is especially necessary in problems such as the
construction of a variety of moduli, because there one may be able to make
the construction locally, without knowing anything about a global em-
bedding. In §6 we gave a definition of an abstract curve. In higher dimen-
sions that method does not work, because there is no unique nonsingular
model of a given function field. However, we can define an abstract variety
by starting from the observation that any variety has an open covering by
affine varieties. Thus one can define an abstract variety as a topological
space X, with an open cover Ui, plus for each Ui a structure of affine variety,
such that on each intersection Ui n Ui the induced variety structures are
isomorphic. It turns out that this generalization of the notion of variety is
not illusory, because in dimension ~ 2 there are abstract varieties which
are not isomorphic to any quasi-projective variety (II, 4.10.2).
There is a third direction in which it is useful to expand our notion of
algebraic variety. In this chapter we have defined a variety as an irreducible
algebraic set in affine or projective space. But it is often convenient to allow
reducible algebraic sets, or even algebraic sets with multiple components.
For example, this is suggested by what we have seen of intersection theory
in §7, since the intersection of two varieties may be reducible, and the sum

58
 8 What Is Algebraic Geometry?

of the ideals of the two varieties may not be the ideal of the intersection. So
one might be tempted to define a "generalized projective variety" in P" to
be an ordered pair <V,J), where V is an algebraic set in P", and I s; S =
k[x 0 , . •• ,xnJ is any ideal such that V = Z(J). This is not in fact what we will
do, but it gives the general idea.
All three generalizations of the notion of variety suggested above are
contained in Grothendieck's definition of a scheme. He starts from the
observation that an affine variety corresponds to a finitely generated integral
domain over a field (3.8). But why restrict one's attention to such a special
class of rings? So for any commutative ring A, he defines a topological space
Spec A, and a sheaf of rings on Spec A, which generalizes the ring of regular
functions on an affine variety, and he calls this an affine scheme. An arbitrary
scheme is then defined by glueing together affine schemes, thus generalizing
the notion of abstract variety we suggested above.
One caution about working in extreme generality. There are many ad-
vantages to developing a theory in the most general context possible. In
the case of algebraic geometry there is no doubt that the introduction of
schemes has revolutionized the subject and has made possible tremendous
advances. On the other hand, the person who works with schemes has to
carry a considerable load of technical baggage with him: sheaves, abelian
categories, cohomology, spectral sequences, and so forth. Another more
serious difficulty is that some things which are always true for varieties may
no longer be true. For example, an affine scheme need not have finite di-
mension, even if its ring is noetherian. So our intuition must be supported
by a good knowledge of commutative algebra.
In this book we will develop the foundations of algebraic geometry using
the language of schemes, starting with the next chapter.

59
CHAPTER II

Schemes

This chapter and the next form the technical heart of this book. In this
chapter we develop the basic theory of schemes, following Grothendieck
[EGA]. Sections 1 to 5 are fundamental. They contain a review of sheaf
theory (necessary even to define a scheme), then the basic definitions of
schemes, morphisms, and coherent sheaves. This is the language that we use
for the rest of the book.
Then in Sections 6, 7, 8, we treat some topics which could have been done
in the language of varieties, but which are already more convenient to discuss
using schemes. For example, the notion of Cartier divisor, and of an in-
vertible sheaf, which belong to the new language, greatly clarify the dis-
cussion ofWeil divisors and linear systems, which belong to the old language.
Then in §8, the systematic use of nonclosed scheme points gives much more
flexibility in the discussion of sheaves of differentials and nonsingular
varieties, improving the treatment of (1, §5).
In §9 we give the definition of a formal scheme, which did not have an
analogue in the theory of varieties. It was invented by Grothendieck as a
good way of dealing with Zariski's theory of"holomorphic functions," which
Zariski regarded as an analogue in abstract algebraic geometry of the
holomorphic functions in a neighborhood of a subvariety in the classical case.

1 Sheaves

The concept of a sheaf provides a systematic way of keeping track of local

algebraic data on a topological space. For example, the regular functions
on open subsets of a variety, introduced in Chapter I, form a sheaf, as we will
see shortly. Sheaves are essential in the study of schemes. In fact, we cannot

60
 1 Sheaves

even define a scheme without using sheaves. So we begin this chapter with
sheaves. For additional information, see the book of Godement [1].

Definition. Let X be a topological space. A presheaf :Ji' of abelian groups on

X consists of the data
(a) for every open subset U s; X, an abelian group :Ji'(U), and
(b) for every inclusion V s; U of open subsets of X, a morphism of
abelian groups Puv::!i'(U)--+ :!i'(V),
subject to the conditions
(0) :!i'(0) = 0, where 0 is the empty set,
(1) Puu is the identity map :Ji'(U) --+ :Ji'(U), and
(2) if W s; V s; U are three open subsets, then Puw = Pvw o Puv·

The reader who likes the language of categories may rephrase this defi-
nition as follows. For any topological space X, we define a category 'Iop(X),
whose objects are the open subsets of X, and where the only morphisms are
the inclusion maps. Thus Hom(V,U) is empty if V <J._ U, and Hom(V,U)
has just one element if V s; U. Now a presheaf is just a contravariant
functor from the category 'Iop(X) to the category 'lib of abelian groups.
We define a presheaf of rings, a presheaf of sets, or a presheaf with values
in any fixed category <I, by replacing the words "abelian group" in the
definition by "ring", "set", or "object of <I" respectively. We will stick to
the case of abelian groups in this section, and let the reader make the necessary
modifications for the case of rings, sets, etc.
As a matter of terminology, if :Ji' is a presheaf on X, we refer to :Ji' ( U) as
the sections of the presheaf :Ji' over the open set U, and we sometimes use
the notation r(U,:Ji') to denote the group :Ji'(U). We call the maps Puv
restriction maps, and we sometimes write siv instead of Puv(s), if s E :Ji'(U).

A sheaf is roughly speaking a presheaf whose sections are determined by

local data. To be precise, we give the following definition.

Definition. A presheaf :Ji' on a topological space X is a sheaf if it satisfies

the following supplementary conditions:
(3) if U is an open set, if {V;} is an open covering of U, and if s E :Ji'(U) is
an element such that siv, = 0 for all i, then s = 0;
(4) if U is an open set, if {V;} is an open covering of U, and if we have
elements si E :Ji'( V;) for each i, with the property that for each i,j, s;jv,n vj =
sjiv,nvj' then there is an elements E :Ji'(U) such that siv, = si for each i.
(Note condition (3) implies that sis unique.)

Nate. According to our definition, a sheaf is a presheaf satisfying certain

extra conditions. This is equivalent to the definition found in some other

61
II Schemes

books, of a sheaf as a topological space over X with certain properties

(Ex. 1.13).

Example 1.0.1. Let X be a variety over the field k. For each open set U s; X,
let @(U) be the ring of regular functions from U to k, and for each V s; U, let
Puv:l9(U)--+ l9(V) be the restriction map (in the usual sense). Then (9 is a
sheaf of rings on X. It is clear that it is a presheaf of rings. To verify the
conditions (3) and (4), we note that a function which is 0 locally is 0, and a
function which is regular locally is regular, because of the definition of regular
function (1, §3). We call (9 the sheaf of regular functions on X.

Example 1.0.2. In the same way, one can define the sheaf of continuous real-
valued functions on any topological space, or the sheaf of differentiable
functions on a differentiable manifold, or the sheaf of holomorphic functions
on a complex manifold.

Example 1.0.3. Let X be a topological space, and A an abelian group. We

define the constant sheaf d on X determined by A as follows. Give A the
discrete topology, and for any open set U s; X, let d ( U) be the group of all
continuous maps of U into A. Then with the usual restriction maps, we
obtain a sheaf d. Note that for every connected open set U, d (U) ~ A,
whence the name "constant sheaf." If U is an open set whose connected
components are open (which is always true on a locally connected topological
space), then d ( U) is a direct product of copies of A, one for each connected
component of U.

Definition. If§' is a presheaf on X, and if P is a point of X, we define the

stalk ff'p of ff' at P to be the direct limit of the groups %( U) for all open
sets U containing P, via the restriction maps p.

Thus an element of ff'p is represented by a pair< U,s), where U is an open

neighborhood of P, and sis an element of ff'(U). Two such pairs <U,s) and
<V,t) define the same element of ff'p if and only ifthere is an open neighbor-
hood W of P with W s; U n V, such that siw = tlw· Thus we may speak of
elements of the stalk ff'p as germs of sections of§' at the point P. In the case
of a variety X and its sheaf of regular functions (9, the stalk (9 P at a point P
is just the local ring of P on X, which was defined in (1, §3).

Definition. If ff' and<§ are presheaves on X, a morphism cp:ff'--+ <§consists

of a morphism of abelian groups cp(U): ff'(U)--+ <§(U) for each open set
U, such that whenever V s; U is an inclusion, the diagram

ff'(U) _cp'--"(_U"-)___. <§(U)

]Puv jPuv
%( V) ------'cp_,_(V--')----+ <§ ( V)
62
 1 Sheaves

is commutative, where p and p' are the restriction maps in ff and '!J. If
ff and 'fJ are sheaves on X, we use the same definition for a morphism
of sheaves. An isomorphism is a morphism which has a two-sided inverse.

Note that a morphism cp: ff ~ 'fJ of presheaves on X induces a morphism

q>p:ffp ~ '!Jp on the stalks, for any point P EX. The following proposition
(which would be false for presheaves) illustrates the local nature of a sheaf.

Proposition 1.1. Let cp: ff ~ 'fJ be a morphism of sheaves on a topological

space X. Then cp is an isomorphism if and only if the induced map on the
stalk q>p:ffp ~ '!Jp is an isomorphism for every P EX.
PROOF. If cp is an isomorphism it is clear that each q>p is an isomorphism.
Conversely, assume q>p is an isomorphism for all P EX. To show that cp is
an isomorphism, it will be sufficient to show that cp(U):ff(U) ~ '!J(U) is
an isomorphism for all U, because then we can define an inverse morphism
tjJ by t/J(U) = cp(U)- 1 for each U. First we show cp(U) is injective. Let
s E ff(U), and suppose cp(s) E '!J(U) is 0. Then for every point P E U, the
image cp(s)p of cp(s) in the stalk '!Jp is 0. Since q>p is injective for each P, we
deduce that Sp = 0 in ffp for each P E U. To say that sp = 0 means that s
and 0 have the same image in ffp, which means that there is an open neigh-
borhood Wp of P, with Wp s; U, such that siwp = 0. Now U is covered by
the neighborhoods Wp of all its points, so by the sheaf property (3), s is 0
on U. Thus cp(U) is injective.
Next, we show that cp(U) is surjective. Suppose we have a section t E '!J(U).
For each P E U, let tp E '!Jp be its germ at P. Since q>p is surjective, we can
find Sp E ff P such that q>p(sp) = tp. Let sp be represented by a section s(P)
on a neighborhood Vp of P. Then cp(s(P)) and tlvp are two elements of
'!J(Vp), whose germs at P are the same. Hence, replacing Vp by a smaller
neighborhood of P if necessary, we may assume that cp(s(P)) = tlvp in
'!J(Vp). Now U is covered by the open sets Vp, and on each Vp we have a
section s(P) E ff(Vp). If P,Q are two points, then s(P)ivpnVQ and s(Q)IvpnVQ
are two sections of ff(Vp n VQ), which are both sent by cp to tlvpn vQ· Hence
by the injectivity of cp proved above, they are equal. Then by the sheaf
property (4), there is a section s E ff(U) such that sivp = s(P) for each P.
Finally, we have to check that cp(s) = t. Indeed, cp(s), t are two sections of
'!J(U), and for each P, cp(s)ivp = tlvP' hence by the sheaf property (3) applied
to cp(s) - t, we conclude that cp(s) = t.

Our next task is to define kernels, cokernels and images of morphisms

of sheaves.

Definition. Let cp:ff ~ 'fJ be a morphism of presheaves. We define the

presheaf kernel of cp, presheaf cokernel of cp, and presheaf image of cp to
be the presheaves given by U r-+ ker( cp( U) ), U r-+ coker( cp( U) ), and
U r-+ im(cp(U)) respectively.
63
II Schemes

Note that if qJ:ff---+ '§is a morphism of sheaves, then the presheafkernel

of qJ is a sheaf, but the presheaf cokernel and presheaf image of qJ are in
general not sheaves. This leads us to the notion of a sheaf associated to a
presheaf.

Proposition-Definition 1.2. Given a presheaf JF, there is a sheaf JF+ and a

morphism e:JF---+ JF+, with the property that for any sheaf'§, and any
morphism qJ:Ji!---+ '§, there is a unique morphism t/f:JF+ ---+ '§ such that
qJ = t/J a e. Furthermore the pair (JF+ ,e) is unique up to unique isomorphism.
JF+ is called the sheaf associated to the presheaf JF.
PROOF. We construct the sheaf ff+ as follows. For any open set U, let JF+(U)
be the set of functions s from U to the union UPE u Jl!p of the stalks of JF
over points of U, such that
(1) for each P E U, s(P) E Jl!p, and
(2) for each P E U, there is a neighborhood V of P, contained in U, and an
element t E JF(V), such that for all Q E V, the germ tQ oft at Q is equal
to s(Q).
Now one can verify immediately ( !) that ff+ with the natural restriction
maps is a sheaf, that there is a natural morphism e:JF---+ JF+, and that it
has the universal property described. The uniqueness of JF+ is a formal
consequence of the universal property. Note that for any point P, Ji!p = JF~.
Note also that if JF itself was a sheaf, then ff+ is isomorphic to JF via
e.

Definition. A subsheaf of a sheaf JF is a sheaf ff' such that for every open set
U s X, ff'( U) is a subgroup of ff( U), and the restriction maps of the
sheaf ff' are induced by those of JF. It follows that for any point P, the
stalk ff~ is a subgroup of Ji!p.
If qJ: JF ---+ '§ is a morphism of sheaves, we define the kernel of qJ,
denoted ker qJ, to be the presheaf kernel of qJ (which is a sheaf). Thus
ker qJ is a subsheaf of JF.
We say that a morphism of sheaves qJ:ff---+ '§is injective ifker qJ = 0.
Thus qJ is injective if and only if the induced map qJ(U): ff( U) ---+ '§(U) is
injective for every open set of X.
If qJ: JF ---+ '§ is a morphism of sheaves, we define the image of qJ,
denoted im qJ, to be the sheaf associated to the presheaf image of qJ. By
the universal property of the sheaf associated to a presheaf, there is a
natural map im qJ ---+ '§. In fact this map is injective (see Ex. 1.4), and thus
im qJ can be identified with a subsheaf of'§.
We say that a morphism qJ:Ji!---+ '§of sheaves is surjective ifim qJ = '§.

We say that a sequence ... ---+ Jl!i- 1 ~ Jl!i ~ Jl!i+ 1 ---+ ••• of sheaves
and morphisms is exact if at each stage ker qJi = im qJi- 1 . Thus a sequence

64
 1 Sheaves

0 --+ fi' ~ ':§ is exact if and only if q; is injective, and fi' ~ ':§ --+ 0 is exact
if and only if q; is surjective.
Now let ff'' be a subsheaf of a sheaf ff'. We define the quotient sheaf
fi' /fi'' to be the sheaf associated to the presheaf U --+ fi'(U)/ff'(U). It
follows that for any point P, the stalk (ff' /fi'')p is the quotient Ji'pjfi'~.
If q;: fi' --+ ':§ is a morphism of sheaves, we define the co kernel of q;,
denoted coker q;, to be the sheaf associated to the presheaf cokernel of q;.
Caution 1.2.1. We saw that a morphism q;:ff'--+ ':§of sheaves is injective if
and only if the map on sections q;(U):fi'(U)--+ ':§(U) is injective for each
U. The corresponding statement for surjective morphisms is not true: if
q;:ff'--+ ':§is surjective, the maps q;(U):fi'(U)--+ ':§(U) on sections need not
be surjective. However, we can say that q; is surjective if and only if the maps
q; P: ffp --+ ':§ P on stalks are surjective for each P. More generally, a sequence
of sheaves and morphisms is exact if and only if it is exact on stalks (Ex. 1.2).
This again illustrates the local nature of sheaves.

So far we have talked only about sheaves on a single topological space.

Now we define some operations on sheaves, associated with a continuous
map from one topological space to another.

Definition. Let f: X --+ Y be a continuous map of topological spaces. For

any sheaf fi' on X, we define the direct image sheaf f*Ji' on Y by
(f*fi')(V) = fi'(f- 1 (V)) for any open set V s; Y. For any sheaf ':§ on
Y, we define the inverse image sheaf f- 1 ':§ on X to be the sheaf associated
to the presheaf U ~---+ limv 2 f<UJ ':§(V), where U is any open set in X, and
the limit is taken over all open sets V of Y containingf(U). Do not confuse
f- 1':§ with the sheaf j*':§ which will be defined later for a morphism of
ringed spaces (§5).

Note that f* is a functor from the category ~b(X) of sheaves on X to

the category ~b(Y) of sheaves on Y. Similarly,f- 1 is a functor from ~b(Y)
to ~b(X).

Definition. If Z is a subset of X, regarded as a topological subspace with the

induced topology, if i: Z --+ X is the inclusion map, and if fi' is a sheaf
on X, then we call i- 1 fi' the restriction of fi' to Z, and we often denote
it by ff'lz· Note that the stalk of ff'lz at any point P E Z is just ffp.

EXERCISES

1.1. Let A be an abelian group, and define the constant presheaf associated to A on
the topological space X to be the presheaf U 1--+ A for all U i= 0, with restriction
maps the identity. Show that the constant sheaf sf defined in the text is the sheaf
associated to this presheaf.

65
II Schemes

1.2. (a) Foranymorphismofsheavescp:ff--> ~.showthatforeachpointP,(kercp)p =

ker(cpp) and (im cp)p = im(cpp).
(b) Show that cp is injective (respectively, surjective) if and only if the induced map
on the stalks qJp is injective (respectively, surjective) for all P.
(c) Show that a sequence ... g;•- 1 S ff' ~ g;i+ 1 --> ... of sheaves and mor-
phisms is exact if and only if for each P E X the corresponding sequence of
stalks is exact as a sequence of abelian groups.
1.3. (a) Let cp: ff --> ~ be a morphism of sheaves on X. Show that cp is surjective if
and only if the following condition holds: for every open set U c;; X, and for
every s E ~(U), there is a covering {U;} of U, and there are elements tiE ff(Ui),
such that cp(t;) = siu, for all i.
(b) Give an example of a surjective morphism of sheaves cp:ff--> ~. and an
open set U such that cp(U):ff(U)--> ~(U) is not surjective.

1.4. (a) Let cp:ff--> ~be a morphism ofpresheaves such that cp(U):ff(U)--> ~(U)
is injective for each U. Show that the induced map cp + : ff + --> ~ + of asso-
ciated sheaves is injective.
(b) Use part (a) to show that if cp.:ff--> ~is a morphism of sheaves, then im cp
can be naturally identified with a subsheaf of~. as mentioned in the text.
1.5. Show that a morphism of sheaves is an isomorphism if and only if it is both
injective and surjective.
1.6. (a) Let ff' be a subsheaf of a sheaf ff. Show that the natural map of ff to the
quotient sheaf ff Iff' is surjective, and has kernel ff'. Thus there is an exact
sequence
o ...... ff' ...... g; ...... ff Iff' ...... o.
(b) Conversely, if 0 --> ff' --> ff ...... ff" ...... 0 is an exact sequence, show that ff'
is isomorphic to a subsheaf of ff, and that ff" is isomorphic to the quotient of
ff by this subsheaf.

1.7. Let cp: ff --> ~ be a morphism of sheaves.

(a) Show that im cp ~ ff jker cp.
(b) Show that coker cp ~ ~lim cp.
1.8. For any open subset U <;; X, show that the functor r(U, ·) from sheaves on X to
abelian groups is a left exact functor, i.e., if 0 --> ff' --> ff --> ff" is an exact
sequence of sheaves, then 0 ...... r(U,ff') ...... r(U,ff) ...... r(U,ff") is an exact
sequence of groups. The functor r(U,·) need not be exact; see (Ex. 1.21) below.

1.9. Direct Sum. Let ff and~ be sheaves on X. Show that the presheaf U H ff( U) EB
~(U) is a sheaf. It is called the direct sum of ff and~. and is denoted by ff EB ~.
Show that it plays the role of direct sum and of direct product in the category of
sheaves of abelian groups on X.

1.10. Direct Limit. Let {ff;} be a direct system of sheaves and morphisms on X. We
define the direct limit of the system {ff,}, denoted lim
----+
ffi, to be the sheaf associated
to the presheaf U H ~ ffi(U). Show that this is a direct limit in the category
of sheaves on X, i.e., that it has the following universal property: given a sheaf~.
and a collection of morphisms ffi --> ~. compatible with the maps of the direct

66
 Sheaves

system, then there exists a unique map !i!!!

ff; ---+ '!J such that for each i, the original
map :Y'; ---+ '!J is obtained by composing the maps ff; ---+ !i!!!:Y'; ---+ '!J.
[Link]. Let {ff;} be a direct system of sheaves on a noetherian topological space X. In
this case show that the presheaf U f---+ lim :Y';(U) is already a sheaf. In particular,
T(X,lim :Y';) = lim T(X,:Y';). ----->
-----> ----->

1.12. lnrerse Limit. Let [ff;} be an inverse system of sheaves on X. Show that the pre-
sheaf U f---+ lim ff;(U) is a sheaf. It is called the inverse limit of the system {:Y';},
and is denoted by lim :Y';. Show that it has the universal property of an inverse
limit in the categorYof sheaves.
1.13. Espace Etale of a Presheaf. (This exercise is included only to establish the con-
nection between our definition of a sheaf and another definition often found in
the literature. See for example Godement [1, Ch. II, §1.2].) Given a presheaf §'
on X, we define a topological space SpeC~). called the espace hale of §', as
follows. As a set, Spe(.~) = UPEX :i'p. We define a projection map n:Spe(:Y')---+ X
by sending s E :Y'p to P. For each open set U ~ X and each sections E :Y'( U), we
obtain a maps: U ---+ Spe(:Y') by sending P f---+ sp, its germ at P. This map has the
property that n c s = idu, in other words, it is a "section" of n over U. We now
make SpeC~) into a topological space by giving it the strongest topology such that
all the maps s: U ---+ Spe(.¥) for all U, and all s E :Y'( U), are continuous. Now
show that the sheaf y;+ associated to F can be described as follows: for any
open set U ~ X, §' + ( U) is the set of continuous sections of Spe(.~) over U. In
particular, the original presheaf §'was a sheaf if and only if for each U, :Y'(U) is
equal to the set of all continuous sections of Spe(.~) over U.
1.14. Support. Let ff be a sheaf on X, and lets E ff(U) be a section over an open set U.
The support of s, denoted Supp s, is defined to be {P E Uisp ¥= 0}, where sp
denotes the germ of sin the stalk ffp. Show that Supp sis a closed subset of U.
We define the support of :Y', Supp ff, to be {P E Xiffp ¥= 0}. It need not be a
closed subset.
1.15. Sheaf X om. Let§', '!J be sheaves of abelian groups on X. For any open set U ~ X,
show that the set Hom(fflu,'!ilu) of morphisms of the restricted sheaves has a
natural structure of abelian group. Show that the presheaf U f---+ Hom(fflu,'!Jiu)
is a sheaf. It is called the sheaf of local morphisms of ff into '!J, "sheaf hom" for
short, and is denoted Xom(ff,'!J).
1.16. Flasque Sheaves. A sheaf ff on a topological space X is jlasque if for every in-
clusion V ~ U of open sets, the restriction map ff(U) ---+ :Y'(V) is surjective.
(a) Show that a constant sheaf on an irreducible topological space is flasque. See
(I, §I) for irreducible topological spaces.
(b) If 0 ---+ ff' ---+ ff ---+ §'" ---+ 0 is an exact sequence of sheaves, and if §'' is
ftasque, then for any open set U, the sequence 0---+ ff'(U)---+ ff(U)---+
§'"( U) ---+ 0 of abelian groups is also exact.
(c) If 0 ---+ ff' ---+ ff ---+ .~" ---+ 0 is an exact sequence of sheaves, and if ff' and§'
are ftasque, then §'" is ftasque.
(d) If f:X ---+ Y is a continuous map, and if ff is a ftasque sheaf on X, then f*:Y'
is a ftasque sheaf on Y.
(e) Let ff be any sheaf on X. We define a new sheaf'!J, called the sheaf of discon-
tinuous sections of ff as follows. For each open set U ~ X, '!J(U) is the set of

67
II Schemes

maps s: U ---> UPeu :!'p such that for each P E U, s(P) E ffp. Show that '§
is a fiasque sheaf, and that there is a natural injective morphism of:!' to '§.

1.17. Skyscraper Sheaves. Let X be a topological space, let P be a point, and let A be an
abelian group. Define a sheaf ip(A) on X as follows: ip(A)(U) = A if P E U, 0
otherwise. Verify that the stalk of ip(A) is A at every point Q E { P}-, and 0
elsewhere, where {P}- denotes the closure of the set consisting of the point P.
Hence the name "skyscraper sheaf." Show that this sheaf could also be described
as i*(A), where A denotes the constant sheaf A on the closed subspace {P}-, and
i: {P}- --->X is the inclusion.
1.18. Adjoint Property off- 1. Let f: X ---> Y be a continuous map of topological spaces.
Show that for any sheaf:!' on X there is a natural map f- 1f*:!' ---> :!', and for
any sheaf'§ on Y there is a natural map'§ ---> f*f- 1'§. Use these maps to show
that there is a natural bijection of sets, for any sheaves :!' on X and '§ on Y,
Homx(F 1 '§,:!') = Homy('§,f*ff).
Hence we say that f- 1 is a left adjoint off*' and that f* is a right adjoint of f- 1 •
1.19. Extending a Sheaf by Zero. Let X be a topological space, let Z be a closed subset,
let i:Z---> X be the inclusion, let U = X - Z be the complementary open subset,
and letj: U---> X be its inclusion.
(a) Let:!' be a sheaf on Z. Show that the stalk (i*ff)p of the direct image sheaf on
X is :l'p if P E Z, 0 if P r/= Z. Hence we call i*ff the sheaf obtained by extending
:!' by zero outside Z. By abuse of notation we will sometimes write :!' instead
of i*ff, and say "consider:!' as a sheaf on X," when we mean "consider i*ff."
(b) Now let:!' be a sheaf on U. LetNff) be the sheaf on X associated to the pre-
sheaf V c-+ ff(V) if V <;::: U, V c-+ 0 otherwise. Show that the stalk (j.(ff) )p is
equal to :l'p if P E U, 0 if P r/= U, and show thatj,:!' is the only sheaf on X which
has this property, and whose restriction to U is :!'. We call}!:!' the sheaf
obtained by extending :!' by zero outside U.
(c) Now let:!' be a sheaf on X. Show that there is an exact sequence of sheaves
on X,
0---> j,(fflul---> :!'---> i*(fflzl---> 0.
1.20. Subsheaf with Supports. Let Z be a closed subset of X, and let:!' be a sheaf on X.
We definerz(X,ff) to be the subgroup of T(X,ff) consisting of all sections whose
support (Ex. 1.14) is contained in Z.
(a) Show that the presheaf V c-+ r z n v([Link]) is a sheaf. It is called the subsheaf
of§' with supports in Z, and is denoted by J'f~(:!').
(b) Let U = X - Z, and let}: U ---> X be the inclusion. Show there is an exact
sequence of sheaves on X
0---> J'f~(:!')---> :!' ---> j*(fflul·
Furthermore, if:!' is fiasque, the map:!' ---> j*(fflul is surjective.
1.21. Some Examples of Sheaves on Varieties. Let X be a variety over an algebraically
closed field k, as in Ch. I. Let (Qx be the sheaf of regular functions on X (1.0.1).
(a) Let Y be a closed subset of X. For each open set U <;:::X, let §y( U) be
the ideal in the ring (1)x(U) consisting of those regular functions which vanish

68
 2 Schemes

at all points of Y n U. Show that the presheaf U H Jy(U) is a sheaf. It is

called the sheaf of ideals Jy of Y, and it is a subsheaf of the sheaf of rings {!) x·
(b) If Y is a subvariety, then the quotient sheaf I!Jx/ .Yy is isomorphic to i*(I!Jy),
where i: Y-> X is the inclusion, and I!Jy is the sheaf of regular functions on Y.
(c) Now let X= P 1 , and let Y be the union of two distinct points P,Q EX. Then
there is an exact sequence of sheaves on X, where:?= i*I!Jp EB i*I!JQ,
0-> fy ->~X -> :ff7-> 0.

Show however that the induced map on global sections r(X,I!Jx)-> r(X,$7)
is not surjective. This shows that the global section functor r(X, ·)is not exact
(cf. (Ex. 1.8) which shows that it is left exact).
(d) Again let X = P 1, and let{!) be the sheaf of regular functions. Let:£ be the
constant sheaf on X associated to the function field K of X. Show that there
is a natural injection 0 -> .ff. Show that the quotient sheaf:£ ji!J is isomorphic
to the direct sum of sheaves LPEX ip(Ip), where lp is the group Kji!Jp, and
ip(I p) denotes the skyscraper sheaf (Ex. 1.17) given by I P at the point P.
(e) Finally show that in the case of (d) the sequence
0 -> T(X,I!J) -> r(X,ff) -> T(X,ff j(!J) -> 0

is exact. (This is an analogue of what is called the "first Cousin problem" in

several complex variables. See Gunning and Rossi [1, p. 248].)

1.22. Glueing Sheaves. Let X be a topological space, let U = { Ui} be an open cover of
X, and suppose we are given for each i a sheaf :?i on Ui, and for each i,j an iso-
morphism <flii::?ilu,nu,.:::.. :?ilu,nuj such that (1) for each i, <flii = id, and (2) for
each i,j,k, <flik = <flik <flii on Ui n Ui n U k· Then there exists a unique sheaf
:? on X, together with isomorphisms 1/Ji::?iu, .:::.. :?i such that for each i,j, 1/Ji =
cpii o 1/Ji on Ui n Ui. We say loosely that.'#' is obtained by glueing the sheaves :?i
via the isomorphisms <flii·

2 Schemes

In this section we will define the notion of a scheme. First we define affine
schemes: to any ring A (recall our conventions about rings made in the
Introduction!) we associate a topological space together with a sheaf of
rings on it, called Spec A. This construction parallels the construction of
affine varieties (I, §1) except that the points of Spec A correspond to all prime
ideals of A, not just the maximal ideals. Then we define an arbitrary scheme
to be something which locally looks like an affine scheme. This definition
has no parallel in Chapter I. An important class of schemes is given by the
construction of the scheme Proj S associated to any graded ring S. This
construction parallels the construction of projective varieties in (1, §2).
Finally, we will show that the varieties of Chapter I, after a slight modification,
can be regarded as schemes. Thus the category of schemes is an enlargement
of the category of varieties.

69
II Schemes

Now we will construct the space Spec A associated to a ring A. As a set,

we define Spec A to be the set of all prime ideals of A. If a is any ideal of A,
we define the subset V(a) <:::::: Spec A to be the set of all prime ideals which
contain a.

Lemma 2.1.
(a) If a and bare two ideals of A, then V(ab) = V(a) u V(b).
(b) If {aJ is any set of ideals of A, then V(l:aJ = nv(aJ
(c) If a and b are two ideals, V(a) <:::::: V(b) if and only if JO. 2 Jh.
PROOF.
(a) Certainly if p 2 a or p 2 b, then p 2 ab. Conversely, if p 2 ab, and
if p "/2 b for example, then there is a bE b such that b ¢; p. Now for any
a E a, ab E p, so we must have a E p since p is a prime ideal. Thus p 2 a.
(b) p contains Ia; if and only if p contains each a;, simply because Ia;
is the smallest ideal containing all of the ideals a;.
(c) The radical of a is the intersection ofthe set of all prime ideals contain-
ing a. So JO. Jb
2 if and only if V( a) <:::::: V(b ).

Now we define a topology on Spec A by taking the subsets of the form

V(a) to be the closed subsets. Note that V(A) = 0; V((O)) =Spec A; and
the lemma shows that finite unions and arbitrary intersections of sets of the
form V(a) are again of that form. Hence they do form the set of closed sets
for a topology on Spec A.
Next we will define a sheaf of rings (!) on Spec A. For each prime ideal
p <:::::: A, let AP be the localization of A at p. For an open set U <:::::: Spec A,
we define (!)(U) to be the set of functions s: u ~ upEU Ap, such that s(p) E Ap
for each p, and such that sis locally a quotient of elements of A: to be precise,
we require that for each p E U, there is a neighborhood V of p, contained in
U, and elements a,f E A, such that for each q E V, f ¢; q, and s(q) = ajf in
Aq. (Note the similarity with the definition of the regular functions on a
variety. The difference is that we consider functions into the various local
rings, instead of to a field.)
Now it is clear that sums and products of such functions are again such,
and that the element 1 which gives 1 in each AP is an identity. Thus (!)(U) is a
commutative ring with identity. If V <:::::: U are two open sets, the natural
restriction map (!)(U) ~ (!)(V) is a homomorphism of rings. It is then clear
that(!) is a presheaf. Finally, it is clear from the local nature of the definition
that (!) is a sheaf.

Definition. Let A be a ring. The spectrum of A is the pair consisting of the

topological space Spec A together with the sheaf of rings (!) defined above.

Let us establish some basic properties of the sheaf(!) on Spec A. For any
element f E A, we denote by D(f) the open complement of V( (f)). Note
70
 2 Schemes

that open sets of the form D(f) form a base for the topology of Spec A.
Indeed, if V(a) is a closed set, and p ~ V(a), then p f2. a, so there is an f E a,
f ~ p. Then p E D(f) and D(f) n V(a) = 0-

Proposition 2.2. Let A be a ring, and (Spec A, @) its spectrum.

(a) For any p E Spec A, the stalk @P of the sheaf (1) is isomorphic to the
local ring AP.
(b) For any element f E A, the ring @(D(f)) is isomorphic to the localized
ring A 1 .
(c) In particular, r{Spec A,@) ~ A.
PROOF.
(a) First we define a homomorphism from (1)P to AP by sending any local
section s in a neighborhood of p to its value s(p) E AP. This gives a well-
defined homomorphism cp from (1) P to AP. The map cp is surjective, because
any element of AP can be represented as a quotient a/f, with a,f E A, f ~ p.
Then D(f) will be an open neighborhood of p, and a/f defines a section of (1)
over D(f) whose value at p is the given element. To show that cp is injective,
let U be a neighborhood ofp, and let s,t E @(U) be elements having the same
value s(p) = t(p) at p. By shrinking U if necessary, we may assume that
s = ajf, and t = bjg on U, where a,b,f,g E A, andf,g ~ p. Since a/f and bjg
have the same image in AP, it follows from the definition of localization that
there is an h ~ p such that h(ga - fb) = 0 in A. Therefore alf = bjg in every
local ring Aq such that f,g,h ~ q. But the set of such q is the open set D(f) n
D(g) n D(h), which contains p. Hence s = t in a whole neighborhood of p,
so they have the same stalk at p. So cp is an isomorphism, which proves (a).
(b) and (c). Note that (c) is the special case of (b) when f = 1, and D(f)
is the whole space. So it is sufficient to prove (b). We oefine a homomorphism
1/1: A J ~ @(D(f)) by sending a/f" to the section s E @(D(f)) which assigns to
each p the image of ajf" in Ap.
First we show t/J is injective. If 1/J(a/f") = t/J(b/fm), then for every p E
D(f), ajf" and b/fm have the same image in AP. Hence there is an element
h ~ p such that h(fma - f"b) = 0 in A. Let a be the annihilator of fma -
f"b. Then h E a, and h ~ p, so a rj;. p. This holds for any p E D(f), so we
conclude that V(a) n D(f) = 0. Therefore f E JO, so some power PEa,
so PUma - f"b) = 0, which shows that ajf" = b/fm in A 1 . Hence t/1 is
injective.
The hard part is to show that t/J is surjective. So let s E @(D(f) ). Then
by definition of@, we can cover D(f) with open sets v;, on which sis repre-
sented by a quotient a;/g;, with g; ~ p for all p E v;, in other words, Vi £ D(g;).
Now the open sets of the form D(h) form a base for the topology, so we may
assume that v; = D(h;) for some h;. Since D(h;) £ D(g;), we have V( (h;)) 2
V( (g;) ), hence by (2.1c), J(hJ £ J(iJ, and in particular, h? E (g;) for some n.
So h? = cg;, so a;/g; = cajh?. Replacing h; by h? (since D(h;) = D(h7)) and
a; by ca;, we may assume that D(f) is covered by the open subsets D(h;),
and that sis represented by a;/h; on D(h;).
71
II Schemes

Next we observe that D(f) can be covered by a finite number of the D(h;).
Indeed, D(f) [Link] UD(hJ if and only if V((f)) 2 n
V((hJ) = V(~)h;)).
By ([Link]) again, this is equivalent to saying f E ~.DhJ, or r E L,(hJ for
some n. This means that r can be expressed as a finite sum r = L,b;h;,
b; E A. Hence a finite subset of the h; will do. So from now on we fix a
finite set h 1, . . . A such that D(f) [Link] D(h 1 ) u ... u D(h,).
For the next step, note that on D(hJ n D(hj) = D(h;hj) we have two
elements of Ah,h1 , namely a;/h; and a)hj both of which represent s. Hence,
according to the injectivity of ljJ proved above, applied to D(h;h), we must
have a;/h; = a)hj in Ah,h1 • Hence for some n,
(h;h;t(h;a; - h;aj) = 0.
Since there are only finitely many indices involved, we may pick n so large
that it works for all i,j at once. Rewrite this equation as
h~+ 1 (h~a.)- hn+ 1 (h~a.) = 0
} l l
l J } •

Then replace each h; by h?+ 1 , and a; by h?a;. Then we still haves represented
on D(h;) by a;/h;, and furthermore, we have hjai = h;aj for all i,j.
Now write r = l);h; as above, which is possible for some n since the
D(hJ cover D(f). Let a = l);a;. Then for each j we have
hja = L: b;a;hj = I b;h;aj = raj.
i i

This says that afr = a)hj on D(h). So ljJ(a/r) = s everywhere, which

shows that ljJ is surjective, hence an isomorphism.

To each ring A we have now associated its spectrum (Spec A,(()). We

would like to say that this correspondence is functorial. For that we need a
suitable category of spaces with sheaves of rings on them. The appropriate
notion is the category of locally ringed spaces.

Definition. A ringed space is a pair (X,(Ox) consisting of a topological space

X and a sheaf of rings (Ox on X. A morphism of ringed spaces from (X,(()x)
to (Y,(()y) is a pair (f,f#) of a continuous map f:X--+ Y and a map
f# :(()y--+ j*(()x of sheaves of rings on Y. The ringed space (X,(Ox) is a
locally ringed space if for each point P EX, the stalk mx,P is a local ring.
A morphism of locally ringed spaces is a morphism (f,f#) of ringed
spaces, such that for each point P EX, the induced map (see below) of
local ringsfff:(OY,J(P)--+ mx,P is a local homomorphism oflocal rings. We
explain this last condition. First of all, given a point P E X, the morphism
of sheaves f# :(()y--+ j*(()x induces a homomorphism of rings (()y(V)--+
(Ox(j- 1 V), for every open set Vin Y. As Vranges over all open neighbor-
hoods of f(P), f- 1 (V) ranges over a subset of the neighborhoods of P.
Taking direct limits, we obtain a map
(()Y,J(P) = lim (()y(V) --+ lim mxu- 1 V),
~ ~

72
 2 Schemes

and the latter limit maps to the stalk mx,P· Thus we have an induced
homomorphism f: :(!)Y,J(P)--> mx,P· We require that this be a local
homomorphism: If A and Bare local rings with maximal ideals rnA and
mB respectively, a homomorphism cp:A --> B is called a local homo-
morphism if cp- 1 (mB) = mk
An isomorphism of locally ringed spaces is a morphism with a two-
sided inverse. Thus a morphism (f,f #) is an isomorphism if and only
iff is a homeomorphism of the underlying topological spaces, and f#
is an isomorphism of sheaves.

Proposition 2.3.
(a) If A is a ring, then (Spec A, (!)) is a locally ringed space.
(b) If cp:A --> B is a homomorphism of rings, then cp induces a nat-
ural morphism of locally ringed spaces
(f,f #):(Spec B, (!)Spec a) --> (Spec A, (!)Spec A).
(c) If A and B are rings, then any morphism of locally ringed spaces
from Spec B to Spec A is induced by a homomorphism of rings cp: A --> B
as in (b).
PROOF.
(a) This follows from (2.2a).
(b) Given a homomorphism cp:A --> B, we define a map f: Spec B-->
Spec A by f(p) = cp- 1(p) for any p E Spec B. If a is an ideal of A, then it is
immediate that f- 1(V(a)) = V(cp(a)), so f is continuous. For each p E
Spec B, we can localize cp to obtain a local homomorphism of local rings
cpP:A"'-'<Pl--> BP. Now for any open set V ~ Spec A we obtain a homo-
morphism of rings f#:(!JspecA(V)--> (!)specB(f- 1 (V)) by the definition of(!),
composing with the maps f and <fJp· This gives the morphism of sheaves
f # :@spec A --> f* ((!)spec B). The induced maps f # on the stalks are just the
local homomorphisms <pp, so (f,f #) is a morphism of locally ringed spaces.
(c) Conversely, suppose given a morphism of locally ringedspaces (f,f#)
from Spec B to Spec A. Taking global sections, f # induces a homomorphism
of rings cp: r{Spec A, (!)Spec A) --> T(Spec B, (!)Spec B)· By (2.2c), these rings
are A and B, respectively, so we have a homomorphism cp:A--> B. For any
p E Spec B, we have an induced local homomorphism on the stalks,
(!)spec A.f(P) --> (!)spec B,p or A f<Pl --> BP, which must be compatible with the
map cp on global sections and the localization homomorphisms. In other
words, we have a commutative diagram

A B

l
Af<pJ

73
II Schemes

Since f # is a local homomorphism, it follows that cp- 1 ( p) = f( p), which

shows that f coincides with the map Spec B ---+ Spec A induced by cp. Now
it is immediate that f# also is induced by cp, so that the morphism (f,f#)
of locally ringed spaces does indeed come from the homomorphism of
rmgs cp.

Caution 2.3.0. Statement (c) of the proposition would be false, if in the

definition of a morphism of locally ringed spaces, we did not insist that the
induced maps on the stalks be local homomorphisms of local rings (see
(2.3.2) below).

Now we come to the definition of a scheme.

Definition. An affine scheme is a locally ringed space (X,@x) which is iso-

morphic (as a locally ringed space) to the spectrum of some ring. A
scheme is a locally ringed space (X,@x) in which every point has an open
neighborhood U such that the topological space U, together with the
restricted sheaf (Dxlu, is an affine scheme. We call X the underlying topo-
logical space of the scheme (X,@x), and (Dx its structure sheaf By abuse
of notation we will often write simply X for the scheme (X,@x). If we
wish to refer to the underlying topological space without its scheme
structure, we write sp(X), read "space of X." A morphism of schemes is
a morphism as locally ringed spaces. An isomorphism is a morphism
with a two-sided inverse.

Example 2.3.1. If k is a field, Spec k is an affine scheme whose topological

space consists of one point, and whose structure sheaf consists of the field k.

Example 2.3.2. If R is a discrete valuation ring, then T = Spec R is an

affine scheme whose topological space consists of two points. One point
t 0 is closed, with local ring R; the other point t 1 is open and dense, with
local ring equal to K, the quotient field of R. The inclusion map R ---+ K
corresponds to the morphism Spec K ---+ T which sends the unique point
of Spec K to t 1 . There is another morphism of ringed spaces Spec K ---+ T
which sends the unique point of SpecK to t 0 , and uses the inclusion R ---+ K
to define the associated map f# on structure sheaves. This morphism is
not induced by any homomorphism R ---+ K as in (2.3b,c), since it is not a
morphism of locally ringed spaces.

Example 2.3.3. If k is a field, we define the affine line over k, Ai, to be

Spec k[ x]. It has a point ~' corresponding to the zero ideal, whose closure
is the whole space. This is called a generic point. The other points, which
correspond to the maximal ideals in k[ x ], are all closed points. They are

74
 2 Schemes

in one-to-one correspondence with the nonconstant monic irreducible poly-

nomials in x. In particular, if k is algebraically closed, the closed points of
Al are in one-to-one correspondence with elements of k.
Example 2.3.4. Let k be an algebraically closed field, and consider the
affine plane over k, defined as A~ = Spec k[x,y J (Fig. 6). The closed points
of A~ are in one-to-one correspondence with ordered pairs of elements of k.
Furthermore, the set of all closed points of M, with the induced topology,
is homeomorphic to the variety called A2 in Chapter I. In addition to the
closed points, there is a generic point ~' corresponding to the zero ideal of
k[x,y], whose closure is the whole space. Also, for each irreducible poly-
nomial f(x,y), there is a point 1J whose closure consists of 1J together with
all closed points (a,b) for which f(a,b) = 0. We say that 1J is a generic point
of the curve f(x,y) = 0.

~ generic
point of
curve
X

closed
points

Figure 6. Spec k [ x, y].

Example 2.3.5. Let X 1 and X 2 be schemes, let U 1 ~ X 1 and U 2 ~ X 2 be

open subsets, and let <p:(Ubmx,[u,)---> (U 2 ,(i)x 2 [u 2 ) be an isomorphism of
locally ringed spaces. Then we can define a scheme X, obtained by glueing
X 1 and X 2 along U 1 and U 2 via the isomorphism <p. The topological space
of X is the quotient of the disjoint union X 1 u X 2 by the equivalence
relation x 1 ~ <p(x 1 ) for each x 1 E U b with the quotient topology. Thus
there are maps i 1 : X 1 ---> X and i 2 : X 2 ---> X, and a subset V ~ X is open
if and only if ij 1 (V) is open in X 1 and i2 1 (V) is open in X 2 . The structure
sheaf mx is defined as follows: for any open set V ~ X,
(l)x(V) = { <s 1 ,sz)[s 1 E (17x,(ij 1 (V)) and s2 E (i)x 2 (i2 1 (V)) and
<p(sl[i~ 1 (V) n u,) = s2fi2 1 (V) n uJ·
Now it is clear that (l)x is a sheaf, and that (X,(I)x) is a locally ringed space.
Furthermore, since X 1 and X 2 are schemes, it is clear that every point of X
has a neighborhood which is affine, hence X is a scheme.

Example 2.3.6. As an example of glueing, let k be a field, let X 1 = X 2 =

AL let U 1 = U 2 = Al - {P}, where P is the point corresponding to the
75
II Schemes

maximal ideal (x), and let qJ: U 1 ~ U 2 be the identity map. Let X be ob-
tained by glueing X 1 and X 2 along U 1 and U 2 via ([J. We get an "affine
line with the point P doubled."

This is an example of a scheme which is not an affine scheme ( !). It is also

an example of a nonseparated scheme, as we will see later (4.0.1).

Next we will define an important class of schemes, constructed from

graded rings, which are analogous to projective varieties.
Let S be a graded ring. See (I, §2) for our conventions about graded
rings. We denote by s+ the ideal EBd>O Sd.
We define the set Proj S to be the set of all homogeneous prime ideals p,
which do not contain all of S +. If a is a homogeneous ideal of S, we define
the subset V(a) = {p E Proj Sip 2 a}.

Lemma 2.4.
(a) If a and bare homogeneous ideals inS, then V(ab) = V(a) u V(b).
(b) If {a;} is any family of homogeneous ideals of S, then V(Lai) =
nv(aJ
PROOF. The proofs are the same as for (2.1a,b), taking into account the
fact that a homogeneous ideal p is prime if and only if for any two homo-
geneous elements a,b E S, ab E p implies a E p or b E p.

Because of the lemma we can define a topology on Proj S by taking the

closed subsets to be the subsets of the form V( a).
Next we will define a sheaf of rings(!) on Proj S. For each p E Proj S, we
consider the ring S<vl of elements of degree zero in the localized ring T- 1 S,
where T is the multiplicative system consisting of all homogeneous elements
of S which are not in p. For any open subset U <;; Proj S, we define (!)(U)
to be the set of functions s: U ~ il S(p) such that for each p E U, s(.p) E S(p)•
and such that sis locally a quotient of elements of S: for each p E U, there
exists a neighborhood V of p in U, and homogeneous elements a,f in S,
of the same degree, such that for all q E V, f ¢ q, and s(q) = a/fin S(q)· Now
it is clear that (!) is a presheaf of rings, with the natural restrictions, and it is
also clear from the local nature of the definition that (!) is a sheaf.

Definition. If Sis any graded ring, we define (Proj S,(!)) to be the topological
space together with the sheaf of rings constructed above.

Proposition 2.5. LetS be a graded ring.

(a) For any p E Proj S, the stalk (!)v is isomorphic to the local ring S<vl·
(b) For any homogeneous fES+, let D+(f) = {.pEProj Slf¢p}.

76
 2 Schemes

Then D +(f) is open in Proj S. Furthermore, these open sets cover Proj S,
and for each such open set, we have an isomorphism of locally ringed
spaces

where s(f) is the sub ring of elements of degree 0 in the localized ring sJ·
(c) Proj Sis a scheme.

PROOF. Note first that (a) says that Proj S is a locally ringed space, and
(b) tells us it is covered by open affine schemes, so (c) is a consequence of
(a) and (b).
The proof of (a) is practically identical to the proof of (2.2a) above, so
is left to the reader.
To prove (b), first note that D +(f) = Proj S - V( (f)), so it is open.
Since the elements of Proj S are those homogeneous prime ideals p of S
which do not contain all of S +, it follows that the open sets D +(f) for homo-
geneous! E s+ cover Proj S. Now fix a homogeneous! E S+. We will define
an isomorphism (cp,cp#) of locally ringed spaces from D+(f) to Spec S<n·
There is a natural homomorphism of rings S -+ S I' and S<n is a subring of
S J· For any homogeneous ideal a <:; S, let cp(a) = (aS I) n S(f). In partic-
ular, if p E D+(f), then cp(p) E Spec S<n' so this gives the map cp as sets.
The properties of localization show that cp is bijective as a map from D +(f)
to Spec S(f). Furthermore, if a is a homogeneous ideal of S, then p 2 a
if and only if cp(p) 2 cp(a). Hence cp is a homeomorphism. Note also if
p ED +(f), then the local rings S<P> and (S<n),<P> are naturally isomorphic.
These isomorphisms and the homeomorphism cp induce a natural map of
sheaves cp# :(Ospecsul-+ cp*((OProjslv+(fl) which one recognizes immediately to
be an isomorphism. Hence (cp,cp#) is an isomorphism of locally ringed
spaces, as required.

Example 2.5.1. If A is a ring, we define projective n-space over A to be the

scheme PA. = Proj A[ x 0 , . . . ,xnJ. In particular, if A is an algebraically
closed field k, then P~ is a scheme whose subspace of closed points is naturally
homeomorphic to the variety called projective n-space-see (Ex. 2.14d)
below.

Next we will show that the notion of scheme does in fact generalize the
notion of variety. It is not quite true that a variety is a scheme. As we have
already seen in the examples above, the underlying topological space of a
scheme such as At or A~ has more points than the corresponding variety.
However, we will show that there is a natural way of adding generic points
(Ex. 2.9) for every irreducible subset of a variety so that the variety becomes
a scheme.
To state our result, we need a definition.

77
II Schemes

Definition. LetS be a fixed scheme. A scheme overS is a scheme X, together

with a morphism X --4 S. If X and Y are schemes over S, a morphism
of X to Y as schemes over S, (also called an S-morphism) is a morphism
f: X --4 Y which is compatible with the given morphisms to S. We denote
by 6c()(S) the category of schemes over S. If A is a ring, then by abuse of
notation we write 6c()(A) for the category of schemes over Spec A.

Proposition 2.6. Let k be an algebraically closed field. There is a natural

fully faithful functor t: IBar(k) --4 6c()(k) from the category of varieties over
k to schemes over k. For any variety V, its topological space is homeo-
morphic to the set of closed points of sp(t(V) ), and its sheaf of regular
functions is obtained by restricting the structure sheaf of t( V) via this
homeomorphism.
PROOF. To begin with, let X be any topological space, and let t(X) be the
set of (nonempty) irreducible closed subsets of X. If Y is a closed subset of
X, then t( Y) <;; t(X). Furthermore, t( Y1 u Y2 ) = t( Y1 ) u t( Y2 ) and t(n Y;) =
nt(Y;). So we can define a topology on t(X) by taking as closed sets the
subsets of the form t( Y), where Y is a closed subset of X. Iff: X 1 --4 X 2 is a
continuous map, then we obtain a map t(f): t(X 1 ) --4 t(X 2 ) by sending an
irreducible closed subset to the closure of its image. Thus t is a functor on
topological spaces. Furthermore, one can define a continuous map rx: X --4
t(X) by rx(P) = {P}-. Note that rx induces a bijection between the set of
open subsets of X and the set of open subsets of t(X).
Now let k be an algebraically closed field. Let V be a variety over
k, and let 0v be its sheaf of regular functions (1.0.1). We will show that
(t(V),rx*(Gv)) is a scheme over k. Since any variety can be covered by open
affine subvarieties (1, 4.3), it will be sufficient to show that if V is affine,
then (t(V),rx*(Gv)) is a scheme. So let V be an affine variety with affine
coordinate ring A. We define a morphism of locally ringed spaces
f3:(V, Gv) --4 X = Spec A
as follows. For each point P E V, let f3(P) = mp, the ideal of A consisting
of all regular functions which vanish at P. Then by (1, 3.2b), f3 is a bijection
of V onto the set of closed points of X. It is easy to see that f3 is a homeo-
morphism onto its image. Now for any open set U <;; X, we will define a
homomorphism of rings Gx( U) --4 {3*( Gv )( U) = Gv(/3- 1 U). Given a section
s E Gx(U), and given a point pEr 1(U), we define s(P) by taking the image
of sin the stalk Gx,p(PJ• which is isomorphic to the local ring A"'P' and then
passing to the quotient ring A"'Pjmp which is isomorphic to the field k. Thus
s gives a function from f3- 1 ( U) to k. It is easy to see that this is a regular
function, and that this map gives an isomorphism Gx( U) ~ Gv(/3- 1 U).
Finally, since the prime ideals of A are in 1-1 correspondence with the irre-
ducible closed subsets of V (see (1, 1.4) and proof), these remarks show that
(X,(Cx) is isomorphic to (t(V), rx*Gv), so the latter is indeed an affine scheme.

78
 2 Schemes

To give a morphism of (t(V),cx*(i]v) to Spec k, we have only to give a

homomorphism of rings k----> T(t(V),cx*CDv) = r(V, CDvl· We send I.E k to
the constant function I. on V. Thus t(V) becomes a scheme over k. Finally,
if V and Ware two varieties, then one can check (Ex. 2.15) that the natural
map
Homtlnr(ki(V,W)----> Hom 2 rt,(kl(t(V),t(W))
is bijective. This shows that the functor t:llJnr(k)----> Scl)(k) is fully faithful.
In particular it implies that t( V) is isomorphic to t( W) if and only if V is
isomorphic to W
It is clear from the construction that ex: V---+ t(V) induces a homeo-
morphism from V onto the set of closed points of t( V ), with the induced
topology.

Note. We will see later (4.10) what the image of the functor tis.

EXERCISES

2.1. Let A be a ring, let X = Spec A, let f E A and let D(f) <;; X be the open comple-
ment of V( (f)). Show that the locally ringed space (D(f), (f'xiD<fl) is isomorphic
to Spec A 1 .

2.2. Let (X,0x) be a scheme, and let U <;; X be any open subset. Show that (U,0xluJ
is a scheme. We call this the induced scheme structure on the open set U, and we
refer to (U,0xlul as an open subscheme of X.

2.3. Reduced Schemes. A scheme (X,(f;x) is reduced if for every open set U <;; X, the
ring (!) x( U) has no nilpotent elements.
(a) Show that (X,0x) is reduced if and only if for every P EX, the local ring crx.P
has no nilpotent elements.
(b) Let (X,0x) be a scheme. Let ((l:x)ced be the sheaf associated to the presheaf
U 1--+ (l)x(U),.d, where for any ring A, we denote by A,.d the quotient of A
by its ideal of nilpotent elements. Show that (X,((f x )"J) is a scheme. We call
it the reduced scheme associated to X, and denote it by X"J· Show that there is
a morphism of schemes X"d --->X, which is a homeomorphism on the under-
lying topological spaces.
(c) Letf: X ---> Y be a morphism of schemes, and assume that X is reduced. Show
that there is a unique morphism{]: X ---> Y,,d such that f is obtained by com-
posing {] with the natural map Y,,d ---> Y.

2.4. Let A be a ring and let (X,lPx) be a scheme. Given a morphism f:X---> Spec A,
we have an associated map on sheaves f #: cr:srccA ---> j*(!;x· Taking global sections
we obtain a homomorphism A ---> r(X,6x ). Thus there is a natural map
:x: Hom 2 ,,(X,Spec A)---> Hom"",, . (A,r(X,(' xl).
Show that :x is bijective (cf. (1, 3.5) for an analogous statement about varieties).

2.5. Describe Spec Z, and show that it is a final object for the category of schemes,
i.e., each scheme X admits a unique morphism to Spec Z.

79
II Schemes

2.6. Describe the spectrum of the zero ring, and show that it is an initial object for
the category of schemes. (According to our conventions, all ring homomorphisms
must take 1 to 1. Since 0 = 1 in the zero ring, we see that each ring R admits a
unique homomorphism to the zero ring, but that there is no homomorphism
from the zero ring toR unless 0 = 1 in R.)
2.7. Let X be a scheme. For any x EX, let (!)x be the local ring at x, and mx its maximal
ideal. We define the residue field of x on X to be the field k(x) = (!)x/mx. Now
let K be any field. Show that to give a morphism of SpecK to X it is equivalent
to give a point x EX and an inclusion map k(x)--+ K.
2.8. Let X be a scheme. For any point x EX, we define the Zariski tangent space Tx
to X at x to be the dual of the k(x)-vector space mx/m;. Now assume that X is
a scheme over a field k, and let k[ t:]/t: 2 be the ring of dual numbers over k. Show
that to give a k-morphism of Spec k[t:]/t: 2 to X is equivalent to giving a point
x EX, rational over k (i.e., such that k(x) = k), and an element of Tx.
2.9. If X is a topological space, and Z an irreducible closed subset of X, a generic
point for Z is a point ( such that Z = {0-. If X is a scheme, show that every
(nonempty) irreducible closed subset has a unique generic point.
2.10. Describe Spec R[ x]. How does its topological space compare to the set R? To C?
2.11. Let k = F P be the finite field with p elements. Describe Spec k[ x]. What are
the residue fields of its points? How many points are there with a given residue
field?
2.12. Clueing Lemma. Generalize the glueing procedure described in the text (2.3.5) as
follows. Let {X;} be a family of schemes (possible infinite). For each i of j,
suppose given an open subset Uij <;; Xi, and let it have the induced scheme
structure (Ex. 2.2). Suppose also given for each i of j an isomorphism of schemes
cp;i: Uu --+ Uii such that (1) for each i,j, cpii = ({J;j 1, and (2) for each i,j,k,
CfJu(U;i n U;k) = Uii n Uik• and CfJ;k = cpik o CfJu on Uii n Uik· Then show that
there is a scheme X, together with morphisms 1/J;:X;--+ X for each i, such that
(1) 1/Ji is an isomorphism of X; onto an open subscheme of X, (2) the 1/J;(X;) cover
X, (3) 1/J;(Ui) = 1/J;(X;) n 1/Ji(X) and (4) 1/J; = 1/Ji o CfJu on Uii. We say that X is
obtained by glueing the schemes Xi along the isomorphisms CfJu· An interesting
special case is when the family X; is arbitrary, but the Uii and cpii are all empty.
Then the scheme X is called the disjoint union of the X;, and is denoted UX;.
2.13. A topological space is quasi-compact if every open cover has a finite subcover.
(a) Show that a topological space is noetherian (I, §1) if and only if every open
subset is quasi-compact.
(b) If X is an affine scheme, show that sp(X) is quasi-compact, but not in general
noetherian. We say a scheme X is quasi-compact if sp(X) is.
(c) If A is a noetherian ring, show that sp(Spec A)is a noetherian topological space.
(d) Give an example to show that sp(Spec A) can be noetherian even when A is not.
2.14. (a) LetS be a graded ring. Show that Proj S = ¢ if and only if every element of
S + is nilpotent.
(b) Let cp: S --+ T be a graded homomorphism of graded rings (preserving degrees).
Let U = {p E Proj Tlpi12 q>(S+)}. Show that U is an open subset ofProj T,
and show that cp determines a natural morphism f: U --+ Proj S.

80
 2 Schemes

(c) The morphism f can be an isomorphism even when cp is not. For example,
suppose that cpd:Sd-> Td is an isomorphism for all d ;::, d 0 , where d0 is an
integer. Then show that U = Proj T and the morphism f: Proj T -> Proj S
is an isomorphism.
(d) Let V be a projective variety with homogeneous coordinate ringS (I, §2). Show
that t(V) ~ Proj S.

2.15. (a) Let V be a variety over the algebraically closed field k. Show that a point
P E t( V) is a closed point if and only if its residue field is k.
(b) If f:X-> Y is a morphism of schemes over k, and if P EX is a point with
residue field k, then f(P) E Y also has residue field k.
(c) Now show that if V,W are any two varieties over k, then the natural map

is bijective. (Injectivity is easy. The hard part is to show it is surjective.)

2.16. Let X be a scheme, let f E T(X,(IJx), and define X 1 to be the subset of points
x E X such that the stalk fx off at x is not contained in the maximal ideal mx
of the local ring (I) x·
(a) If U = Spec B is an open affine subscheme of X, and if J E B = r(U,(I)xlul is
the restriction off, show that U n X 1 = D(J). Conclude that X 1 is an open
subset of X.
(b) Assume that X is quasi-compact. Let A = r(X,(I)x), and let a E A be an
element whose restriction to X 1 is 0. Show that for some n > 0, f"a = 0.
[Hint:Use an open affine cover of X.]
(c) Now assume that X has a finite cover by open affines U; such that each inter-
section U; n Ui is quasi-compact. (This hypothesis is satisfied, for example,
if sp(X) is noetherian.) Let bE r(X1 ,(1)xJ Show that for some n > 0, f"b is
the restriction of an element of A.
(d) With the hypothesis of(c), conclude that r(X1 ,(1Jx1 ) ~ A 1 .

2.17. A Criterion for Affineness.

(a) Let f: X -> Y be a morphism of schemes, and suppose that Y can be covered
by open subsets U;, such that for each i, the induced map f- 1 (U;) -> U; is an
isomorphism. Then f is an isomorphism.
(b) A scheme X is affine if and only if there is a finite set of elements f 1 , •• . ,f,. E
A = r(X,(I)x), such that the open subsets X 1 , are affine, andf1 , ... ,f,. generate
the unit ideal in A. [Hint: Use (Ex. 2.4) and (Ex. 2.16d) above.]
2.18. In this exercise, we compare some properties of a ring homomorphism to the
induced morphism of the spectra of the rings.
(a) Let A be a ring, X = Spec A, andf EA. Show thatf is nilpotent if and only if
D(f) is empty.
(b) Let cp: A -> B be a homomorphism of rings, and let f: Y = Spec B -> X =
Spec A be the induced morphism of affine schemes. Show that cp is injective if
and only if the map of sheaves J# :(IJx -> f*(IJY is injective. Show furthermore
in that case f is dominant, i.e., f(Y) is dense in X.
(c) With the same notation, show that if cp is surjective, then f is a homeomor-
phism of Yonto a closed subset of X, andf# :(IJx-> j*(IJY is surjective.

81
II Schemes

(d) Prove the converse to (c), namely, iff: Y--> X is a homeomorphism onto a
closed subset, and f# :@x--> f*(r)Y is surjective, then qJ is surjective. [Hint:
Consider X' = Spec(A/ker ({J) and use (b) and (c).]
2.19. Let A be a ring. Show that the following conditions are equivalent:
(i) Spec A is disconnected;
(ii) there exist nonzero elements e~oe 2 E A such that e 1 e2 = 0, ei = e 1 , e~ = e2 ,
e 1 + e2 = 1 (these elements are called orthogonal idempotents);
(iii) A is isomorphic to a direct product A 1 x A 2 of two nonzero rings.

3 First Properties of Schemes

In this section we will give some of the first properties of schemes. In particu-
lar we will discuss open and closed subschemes, and products of schemes. In
the exercises we introduce the notion of constructible subsets, and study the
dimension of the fibres of a morphism.

Definition. A scheme is connected if its topological space is connected. A

scheme is irreducible if its topological space is irreducible.

Definition. A scheme X is reduced if for every open set U, the ring (l)x(U) has
no nilpotent elements. Equivalently (Ex. 2.3), X is reduced if and only if
the local rings @p, for all P EX, have no nilpotent elements.

Definition. A scheme X is integral if for every open set U <:; X, the ring
(l)x(U) is an integral domain.

Example 3.0.1. If X = Spec A is an affine scheme, then X is irreducible if

and only if the nilradical nil A of A is prime; X is reduced if and only if
nil A = 0; and X is integral if and only if A is an integral domain.

Proposition 3.1. A scheme is integral if and only if it is both reduced and ir-
reducible.
PROOF. Clearly an integral scheme is reduced. If X is not irreducible, then
one can find two nonempty disjoint open subsets U 1 and U 2 . Then
@(U 1 u U 2 ) = @(U 1 ) x @(U 2 ) which is not an integral domain. Thus
integral implies irreducible.
Conversely, suppose that X is reduced and irreducible. Let U <:; X be an
open subset, and suppose that there are elements f,g E @(U) with fg = 0.
Let Y = {x E Uifx E mx}, and let Z = {x E Ulgx E mx}· Then Y and Z are
closed subsets (Ex. 2.16a), and Y u Z = U. But X is irreducible, so U is
irreducible, so one of Y or Z is equal to U, say Y = U. But then the restric-
tion off to any open affine subset of U will be nilpotent (Ex. 2.18a), hence
zero, so f is zero. This shows that X is integral.

82
 3 First Properties of Schemes

Definition. A scheme X is locally noetherian if it can be covered by open affine

subsets Spec A;, where each A; is a noetherian ring. X is noetherian if
it is locally noetherian and quasi-compact. Equivalently, X is noetherian
if it can be covered by a finite number of open affine subsets Spec A;,
with each A; a noetherian ring.

Caution 3.1.1. If X is a noetherian scheme, then sp(X) is a noetherian topo-

logical space, but not conversely (Ex. 2.13) and (Ex. 3.17).

Note that in this definition we do not require that every open affine
subset be the spectrum of a noetherian ring. So while it is obvious from the
definition that the spectrum of a noetherian ring is a noetherian scheme, the
converse is not obvious. It is a question of showing that the noetherian
property is a "local property". We will often encounter similar situations
later in defining properties of a scheme or of a morphism of schemes, so we
will give a careful statement and proof of the local nature of the noetherian
property, to illustrate this type of situation.

Proposition 3.2. A scheme X is locally noetherian if and only if for every open
affine subset U = Spec A, A is a noetherian ring. In particular, an affine
scheme X = Spec A is a noetherian scheme if and only if the ring A is a
noetherian ring.
PROOF. The "if" part follows from the definition, so we have to show if X
is locally noetherian, and if U = Spec A is an open affine subset, then A is a
noetherian ring. First note that if B is a noetherian ring, so is any localization
B J· The open subsets D(f) ~ Spec B f form a base for the topology of Spec B.
Hence on a locally noetherian scheme X there is a base for the topology con-
sisting of the spectra of noetherian rings. In particular, our open set U can
be covered by spectra of noetherian rings.
So we have reduced to proving the following statement: let X = Spec A
be an affine scheme, which can be covered by open subsets which are spectra
of noetherian rings. Then A is noetherian. Let U = Spec B be an open
subset of X, with B noetherian. Then for some f E A, D(f) ~ U. Let
J be the image off in B. Then A f ~ B 1 , hence A f is noetherian. So we
can cover X by open subsets D(f) ~ Spec A J with A J noetherian. Since X
is quasi-compact, a finite number will do.
So now we have reduced to a purely algebraic problem: A is a ring,
/ 1 , . . . ,f.. are a finite number of elements of A, which generate the unit ideal,
and each localization AJ, is noetherian. We have to show A is noetherian.
First we establish a lemma. Let a ~ A be an ideal, and let ({J;: A --+ A J, be
the localization map, i = 1, ... ,r. Then
a= n({J;- 1 (cp;(a) · AJJ
The inclusion ~ is obvious. Conversely, given an element b E A contained

83
II Schemes

in this intersection, we can write cp;(b) = a;/fi' in AJ, for each i, where
a; E a, and n; > 0. Increasing the n; if necessary, we can make them all
equal to a fixed n. This means that in A we have
f'[''(fib - a;) = 0
for some m;. And as before, we can make all them; = m. Thus f'['+nb E a for
each i. Since f 1 , . . . ,f.. generate the unit ideal, the same is true of their Nth
powers for any N. Take N = n + m. Then we have 1 = '[Link] for suitable
c; EA. Hence
b = '[Link] E a
as required.
Now we can easily show that A is noetherian. Let a 1 s; a 2 s; ... be an
ascending chain of ideals in A. Then for each i,
qJ;(a 1 )·AJ, s; o/;(a 2 )·AJ, s; ...

is an ascending chain of ideals in A f•' which must become stationary because

AJ, is noetherian. There are only finitely many AJ,, so from the lemma we
conclude that the original chain is eventually stationary, and hence A is
noetherian.

Definition. A morphism f: X ---+ Y of schemes is locally offinite type ifthere

exists a covering of Y by open affine subsets V; = Spec B;, such that for
each i, f- 1(V;) can be covered by open affine subsets U;i = Spec Aii, where
each A;i is a finitely generated B;-algebra. The morphism f is of finite
type if in addition each f- 1 (V;) can be covered by a finite number of the
Uii.

Definition. A morphism f: X ---+ Y is a finite morphism if there exists a

covering of Y by open affine subsets V; = Spec B;, such that for each i,
f- 1 (V;) is affine, equal to Spec A;, where A; is a B;-algebra which is a
finitely generated B;-module.

Note in each of these definitions that a property of a morphism f: X ---+ Y

is defined by the existence of an open affine cover of Y with certain properties.
In fact in each case it is equivalent to require the given property for every
open affine subset of Y (Ex. 3.1-3.4).

Example 3.2.1. If V is a variety over an algebraically closed field k, then the

assoc1ated scheme t(V) (see (2.6)) is an integral noetherian scheme of finite
type over k. Indeed, V can be covered by a finite number of open affine
subvarieties (I, 4.3), so t(V) can be covered by a finite number of open affines
of the form Spec A;, where each A; is an integral domain which is a finitely
generated k-algebra and hence noetherian.

84
 3 First Properties of Schemes

Example 3.2.2. If P is a point of a variety V, with local ring (!) p, then Spec(!) P
is an integral noetherian scheme, which is not in general of finite type over k.

Next we come to open and closed subschemes.

Definition. An open subscheme of a scheme X is a scheme U, whose topological

space is an open subset of X, and whose structure sheaf (!)u is isomorphic
to the restriction (!)xlu of the structure sheaf of X. An open immersion is a
morphism f:X--> Y which induces an isomorphism of X with an open
subscheme of Y.

Note that every open subset of a scheme carries a unique structure of

open subscheme (Ex. 2.2).

Definition. A closed immersion is a morphism f: Y--> X of schemes such that

f induces a homeomorphism of sp(Y) onto a closed subset of sp(X),
and furthermore the induced map f #: (!)x --> f * (!)Y of sheaves on X is
surjective. A closed subscheme of a scheme X is an equivalence class of
closed immersions, where we say f: Y--> X and f': Y' --> X are equi-
valent if there is an isomorphism i: Y'--> Y such that f' = f o i.

Example 3.2.3. Let A be a ring, and let a be an ideal of A. Let X = Spec A

and let Y = Spec A/a. Then the ring homomorphism A --> A/a induces a
morphism of schemes f: Y-> X which is a closed immersion. The map fis
a homeomorphism of Y onto the closed subset V(a) of X, and the map of
structure sheaves (!)x--> j*(!)Y is surjective because it is surjective on the
stalks, which are localizations of A and A/a,respectively (Ex. 2.18).
Thus for any ideal a s; A we obtain a structure of closed subscheme on the
closed set V(a) s; X. In particular, every closed subset Y of X has many closed
subscheme structures, corresponding to all the ideals a for which V(a) = Y.
In fact, every closed subscheme structure on a closed subset Y of an affine
scheme X arises from an ideal in this way (Ex. 3.11 b) or (5.10).

Example 3.2.4. For some more specific examples, let A = k[ x,y ], where k is
a field. Then Spec A = M is the affine plane over k. The ideal a = (xy)
gives a reducible subscheme, consisting of the union of the x and y axes. The
ideal a = (x 2 ) gives a subscheme structure with nilpotents on the y-axis.
The ideal a = (x 2 ,xy) gives another subscheme structure on they-axis, this
one having nilpotents only in the local ring at the origin. We say the origin
is an embedded point for this subscheme.

Example 3.2.5. Let V be an affine variety over the field k, and let W be a
closed subvariety. Then W corresponds to a prime ideal p in the affine co-
ordinate ring A of V (1, §1). Let X = t(V) and Y = t(W) be the associated
schemes. Then X = Spec A and Y is the closed subscheme defined by p.
For each n ;?: 1 let Y, be the closed subscheme of X corresponding to the

85
II Schemes

ideal pn. Then Y1 = Y, but for each n > 1, Y, is a nonreduced scheme struc-
ture on the closed set Y, which does not correspond to any subvariety of V.
We call Y, the nth irifinitesimal neighborhood of Yin X. The schemes Y, reflect
properties of the embedding of Yin X. Later (§9) we will study the "formal
completion" of Yin X, which is roughly the limit of the schemes Y, as n --+ oo.

Example 3.2.6. Let X be a scheme, and let Y be a closed subset. In general Y

will have many possible closed subscheme structures. However, there is one
which is "smaller" than any other, called the reduced induced closed subscheme
structure, which we now describe.
First let X = Spec A be an affine scheme, and let Y be a closed subset.
Let a <;; A be the ideal obtained by intersecting all the prime ideals in Y. This
is the largest ideal for which V(a) = Y. Then we take the reduced induced
structure on Y to be the one defined by a.
Now let X be any scheme, and let Y be a closed subset. For each open
affine subset Ui <;; X, consider the closed subset Y; = Y n Ui of Ui, and give
it the reduced induced structure just defined for affines (which may depend
on UJ I claim that for any i,j, the restrictions to Y; n lj of the two structure
sheaves just defined on Y; and lj are isomorphic, and furthermore, that the
three such isomorphisms on Y; n lj n 1k are compatible for all i,j,k. One
reduces easily to showing that if U = Spec A is an open affine, and iff E A,
and if V = D(f) = Spec A 1 , then the reduced induced structure on Y n U
obtained from A when restricted to Y n V agrees with the one obtained
from A 1 . This corresponds to the algebraic fact that if a is the intersection
of those prime ideals of A which are in Y, then aA 1 is the intersection of those
prime ideals of A 1 which are in Y n D(f).
So now we can glue the sheaves defined on the Y; to obtain a sheaf on Y
(Ex. 1.22}, which gives us the desired reduced induced subscheme structure
on Y. See (Ex. 3.11) below for a universal property of the reduced induced
subscheme structure.

Definition. The dimension of a scheme X, denoted dim X, is its dimension as a

topological space (I, §1). If Z is an irreducible closed subset of X, then the
codimension of Z in X, denoted codim(Z,X) is the supremum of integers n
such that there exists a chain
Z = Z 0 < Z 1 < ... < Zn
of distinct closed irreducible subsets of X, beginning with Z. If Y is any
closed subset of X, we define
codim(Y,X) = inf codim(Z,X)
Z"Y

where the infimum is taken over all closed irreducible subsets of Y.

Example 3.2.7. If X = Spec A is an affine scheme, then the dimension of X

is the same as the Krull dimension of A (1, §1).

86
 3 First Properties of Schemes

Caution 3.2.8. Be careful in applying the concepts of dimension and codi-

mension to arbitrary schemes. Our intuition is derived from working with
schemes of finite type over a field, where these notions are well-behaved.
For example, if X is an affine integral scheme of finite type over a field k,
and if Y c:; X is any closed irreducible subset, then (1, 1.8A) implies that
dim Y + codim( Y,X) = dim X. But on arbitrary (even noetherian) schemes,
funny things can happen. See (Ex. 3.20-3.22), and also Nagata [7], and
Grothendieck [EGA IV, §5].

Definition. Let S be a scheme, and let X, Y be schemes over S, i.e., schemes

with morphisms to S. We define the .fibred product of X and Y overS,
denoted X x s Y, to be a scheme, together with morphisms p 1 :X x s
Y --+ X and p 2 : X x s Y --+ Y, which make a commutative diagram with
the given morphisms X --+ S and Y --+ S, such that given any scheme Z
over S, and given morphisms f:Z--+ X and g:Z--+ Y which make a
commutative diagram with the given morphisms X --+ S and Y --+ S, then
there exists a unique morphism e: z--+ X X s y such that f = p 1 0 e, and
g = p 2 o e. The morphisms p 1 and p 2 are called the projection morphisms
of the fibred product onto its factors.

z ________ .. X X s y

~y
~/
s
If X and Y are schemes given without reference to any base scheme S,
we take S = Spec Z (Ex. 2.5) and define the product of X and Y, denoted
X X Y, to be X Xspec z Y.

Theorem 3.3. For any two schemes X and Y over a schemeS, the .fibred product
X x s Y exists, and is unique up to unique isomorphism.

PROOF. The idea is first to construct products for affine schemes and then
glue. We proceed in seven steps.
Step 1. Let X = Spec A, Y = Spec B, S = Spec R all be affine. Then
A and BareR-algebras, and I claim that Spec (A ® R B) is a product for X and
Y overS. Indeed, for any scheme Z, to give a morphism of Z to Spec (A ® R B)
is the same as to give a homomorphism of the ring A ® R B into the ring
r(Z,(!J 2 ), by (Ex. 2.4). But to give a homomorphism of A ®R B into any ring
is the sallie as to give homomorphisms of A and B into that ring, inducing
the same homomorphism on R. Applying (Ex. 2.4) again, we see that to give
a morphism of Z into Spec (A ® R B) is the same as giving morphisms of Z
into X and into Y, which give rise to the same morphism of Z into S. Thus
Spec (A ® R B) is the desired product.

87
II Schemes

Step 2. It follows immediately from the universal property of the product

that it is unique up to unique isomorphism, if it exists. We will need this
uniqueness for those products already constructed, as we go along.
Step 3. Glueing morphisms. We have already seen how to glue sheaves
(Ex. 1.22) and how to glue schemes (Ex. 2.12). Now we glue morphisms. If X
and Yare schemes, then to give a morphism f from X to Y, it is equivalent to
give an open cover { U;} of X, together with morphisms _[;: U; --+ Y, where U;
has the induced open subscheme structure, such that the restrictions of J;
and jj to U; n Ui are the same, for each i,j. The proof is straightforward.
Step 4. If X, Yare schemes over a schemeS, if U <:::; X is an open subset,
and if the product X x s Y exists, then p1 1( U) <:::; X x s Y is a product for U
and Y over S. Indeed, given a scheme Z, and morphisms f: Z --+ U and
g:Z--+ Y, f determines a map of Z to X by composing with the inclusion
U <:::; X. Hence there is a map B:Z --+X x s Y compatible with f,g and the
projections. But since f(Z) <:::; U, we have 8(Z) <:::; p1 1 ( U). So 8 can be
regarded as a morphism Z --+ p1 1(U). It is clearly unique, so p1 1 (U) is a
product U x s Y.
Step 5. Suppose given X,Y schemes over S, suppose {X;} is an open
covering of X, and suppose that for each i, X; x s Y exists. Then X x s Y
exists. Indeed, for each i,j, let Uii .:::; X; x s Y be p1 1 (X;i), where X;i =
X; n Xi. Then by Step 4, Uii is a product for Xii andY overS. Hence by the
uniqueness of products there are (unique) isomorphisms ({J;/ uij --+ uji for
each i,j compatible with all the projections. Furthermore, these isomor-
phisms are compatible with each other for each i,j,k, in the sense of(Ex. 2.12).
Thus we are in a position to glue the schemes X; x s Yvia the isomorphisms
({J;i· We obtain by (Ex. 2.12) a scheme X x s Y which I claim is a product for
X and Y overS. The projection morphisms p 1 and p 2 are defined by glueing
the projections from the pieces X; x s Y (Step 3). Given a scheme Z and
morphisms f:Z--+ X, g:Z--+ Y, let Z; = f~ 1 (X;). Then we get maps
8;: Z; --+ X; x s Y, hence by composition with the inclusions X; x s Y <:::;
X x s Y we get maps 8;: Z; --+ X x s Y. One verifies that these maps agree on
Z; n Zi, so we can glue the morphisms (Step 3) to obtain a morphism 8: Z --+
X x s Y, compatible with the projections and f and g. The uniqueness of 8
can be checked locally.
Step 6. We know from Step 1 that if X, Y, S are all affine, then X x s Y
exists. Thus using Step 5 we conclude that for any X, but Y, S affine, the
product exists. Using Step 5 again, with X and Y interchanged, we find that
the product exists for any X and any Y over an affine S.
Step 7. Given arbitrary X,Y, S, let q:X--+ S and r: Y--+ S. be the given
morphisms. Let S; be an open affine cover of S. Let X; = q~ 1 (S;) and let
li = r~ 1(S;). Then by Step 6, X; x s; li exists. Note that this same scheme is
a product for X; and Y overS. Indeed, given morphisms f:Z--+ X; and
g:Z--+ Y over S, the image of g must land inside l;. Thus X; x s Y exists
for each i, and one more application of Step 5 gives us X x s Y. This completes
the proof.

88
 3 First Properties of Schemes

Perhaps this is a good place to make some general remarks on the im-
portance and uses offibred products. To begin with, we can define the fibres
of a morphism.

Definition. Let f: X ---+ Y be a morphism of schemes, and let y E Y be a

point. Let k(y) be the residue field of y, and let Spec k(y) ---+ Y be the
natural morphism (Ex. 2.7). Then we define the fibre of the morphism
f over the point y to be the scheme
xy =X X y Spec k(y).

The fibre XY is a scheme over k(y), and one can show that its underlying
topological space is homeomorphic to the subset f- 1 (y) of X (Ex. 3.10).
The notion of the fibre of a morphism allows us to regard a morphism
as a family of schemes (namely its fibres) parametrized by the points of the
image scheme. Conversely, this notion of family is a good way of making
sense of the idea of a family of schemes varying algebraically. For example,
given a scheme X 0 over a field k, we define a family of deformations of X 0
to be a morphism f:X ---+ Y with Y connected, together with a point y 0 E Y,
such that k(y 0 ) = k, and XYo ~ X 0 . The other fibres XY off are called
deformations of X 0 .
An interesting kind offamily arises when we have a scheme X over Spec Z.
In this case, taking the fibre over the generic point gives a scheme XQ over
Q, while taking the fibre over a closed point, corresponding to a prime
number p, gives a scheme X P over the finite field F p· We say that X P arises
by reduction mod p of the scheme X.

Another important application of fib red products is to the notion of base

extension. Let S be a fixed scheme which we think of as a base scheme,
meaning that we are interested in the category of schemes over S. For
example, think of S = Spec k, where k is a field. If S' is another base scheme,
and if S' ----> S is a morphism, then for any scheme X over S, we let X' =
X x s S', which will be a scheme overS'. We say that X' is obtained from X
by making a base extension S' ---+ S. For example, think of S' = Speck'
where k' is an extension field of k. Note, by the way, that base extension is a
transitive operation: if S" ---+ S' ---+ S are two morphisms, then (X x s S') x s·
S" ~X Xs S".
This ties in with a general philosophy, emphasized by Grothendieck in
his "Elements de Geometrie Algebrique" ([EGA]), that one should try to
develop all concepts of algebraic geometry in a relative context. Instead of
always working over a fixed base field, and considering properties of one
variety at a time, one should consider a morphism of schemes f: X ---+ S,
and study properties of the morphism. It then becomes important to study
the behavior of properties off under base extension, and in particular, to
relate properties off to properties of the fibres off For example, iff: X ---+ S

89
II Schemes

is a morphism of finite type, and if S'----+ Sis any base extension, thenf' :X'----+ S'
is also a morphism of finite type, where X' = X x s S'. Hence we say the
property of a morphism f being of finite type is stable under base extension.
On the other hand, if for example f: X ----+ Sis a morphism of integral schemes,
the fibres off may be neither irreducible nor reduced. So the property of a
scheme being integral is not stable under base extension.

Example 3.3.1. Let k be an algebraically closed field, let

X = Spec k[x,y,t]/(ty- x 2 ),
let Y = Spec k[t], and let f: X ----+ Y be the morphism determined by the
natural homomorphism k[t] ----+ k[x,y,t]/(ty - x 2 ). Then X and Y are
integral schemes of finite type over k, and f is a surjective morphism. We
identify the closed points of Y with elements of k. For a E k, a i= 0, the fibre
Xa is the plane curve ay = x 2 in A~, which is an irreducible, reduced curve.
But for a = 0, the fibre X 0 is the nonreduced scheme given by x 2 = 0 in A 2 .
Thus we have a family (Fig. 7) in which most members are irreducible curves,
but one is nonreduced. This shows how nonreduced schemes occur naturally
even if one is primarily interested in varieties. We can say that the nonreduced
scheme x 2 = 0 in A2 is a deformation of the irreducible parabola ay = x 2
as a ----+ 0.

Figure 7. An algebraic family of schemes.

Example 3.3.2. Similarly, if X = Spec k[x,y,t]/(xy - t), we get a family

whose general member Xa is an irreducible hyperbola xy = a, when a i= 0,
but whose special member X 0 is the reducible scheme xy = 0 consisting
of two lines.

EXERCISES

3.1. Show that a morphism f:X-> Y is locally of finite type if and only if for every
open affine subset V = Spec B of Y, f- 1 (V) can be covered by open affine subsets
Ui = Spec Ai, where each Ai is a finitely generated B-algebra.

90
 3 First Properties of Schemes

3.2. A morphism f: X --> Y of schemes is quasi-compact if there is a cover of Y by open

affines V; such that f~ 1 (V;) is quasi-compact for each i. Show that f is quasi-
compact if and only if for every open affine subset V s; Y, f~ 1(V) is quasi-compact.
3.3. (a) Show that a morphism f: X --> Y is of finite type if and only if it is locally of
finite type and quasi-compact.
(b) Conclude from this that f is of finite type if and only if for er•ery open affine
subset V = Spec B of Y, f~ 1 ( V) can be covered by a finite number of open
affines U 1 = Spec A 1, where each A1 is a finitely generated B-algebra.
(c) Show also iff is of finite type, then for every open affine subset V = Spec B s;
Y, and for every open affine subset U = Spec A s; f~ 1 (V), A is a finitely gener-
ated B-algebra.
3.4. Show that a morphism f: X --> Y is finite if and only if for erery open affine subset
V = Spec B of Y, f~ 1(V) is affine, equal to Spec A, where A is a finite B-module.
3.5. A morphism f: X --> Y is quasi-finite if for every point y E Y, f~ 1(y) is a finite set.
(a) Show that a finite morphism is quasi-finite.
(b) Show that a finite morphism is closed, i.e., the image of any closed subset is
closed.
(c) Show by example that a surjective, finite-type, quasi-finite morphism need not
be finite.
3.6. Let X be an integral scheme. Show that the local ring @~ of the generic point ~
of X is a field. It is called the function field of X, and is denoted by K(X). Show
also that if U = Spec A is any open affine subset of X, then K(X) is isomorphic
to the quotient field of A.
3.7. A morphism f:X --> Y, with Y irreducible, is generically finite iff~ 1 (1'/) is a finite
set, where t) is the generic point of Y. A morphism f: X --> Y is dominant if f(X)
is dense in Y. Now let f:X--> Y be a dominant, generically finite morphism of
finite type of integral schemes. Show that there is an open dense subset U <;::: Y
such that the induced morphism f~ 1 ( U) --> U is finite. [Hint: First show that the
function field of X is a finite field extension of the function field of Y.J
3.8. Normalization. A scheme is normal if all of its local rings are integrally closed
domains. Let X be an integral scheme. For each open affine subset U = Spec A
of X, let A be the integral closure of A in its quotient field, and let U = Spec A.
Show that one can glue the schemes U to obtain a normal integral scheme X,
called the normalization of X. Show also that there is a morphism X--> X, having
the following universal property: for every normal integral scheme Z, and for every
dominant morphism f: Z--> X, f factors uniquely through X. If X is of finite type
over a field k, then the morphism X --> X is a finite morphism. This generalizes
(I, Ex. 3.1 7).

3.9. The Topological Space of a Product. Recall that in the category of varieties, the
Zariski topology on the product of two varieties is not equal to the product
topology (I, Ex. 1.4). Now we see that in the category of schemes, the underlying
point set of a product of schemes is not even the product set.
(a) Let k be a field, and let Ai = Spec k[ x J be the affine line over k. Show that
At x Speck Ai ~ Af, and show that the underlying point set of the product is
not the product of the underlying point sets of the factors (even if k is algebrai-
cally closed).

91
II Schemes

(b) Let k be a field, lets and t be indeterminates over k. Then Spec k(s), Spec k(t),
and Spec k are all one-point spaces. Describe the product scheme Spec
k(s) X Speck Spec k(t).
3.10. Fibres of a Morphism.
(a) Iff:X -+ Yis a morphism, andy E Ya point, show that sp(Xy) is homeomor-
phic tor 1 (y) with the induced topology.
(b) LetX = Speck[s,t]/(s- t2 ),let Y = Speck[s],andletf:X-+ Ybethemor-
phism defined by sending s -+ s. If y E Y is the point a E k with a =1= 0, show
that the fibre X Y consists of two points, with residue field k. If y E Y cor-
responds to 0 E k, show that the fibre X Y is a nonreduced one-point scheme.
If '1 is the generic point of Y, show that X q is a one-point scheme, whose residue
field is an extension of degree two of the residue field of '1· (Assume k alge-
braically closed.)
3.11. Closed Subschemes.
(a) Closed immersions are stable under base extension: iff: Y-+ X is a closed
immersion, and if X' -+ X is any morphism, then f': Y x x X' -+ X' is also a
closed immersion.
*(b) If Y is a closed subscheme of an affine scheme X = Spec A, then Y is also
affine, and in fact Y is the closed subscheme determined by a suitable ideal
a c;; A as the image of the closed immersion Spec A/a -+ Spec A. [Hints: First
show that Y can be covered by a finite number of open affine subsets of the
form D(h) n Y, with J; EA. By adding some more J; with D(J;) n Y = 0,
if necessary, show that we may assume that the D(J;) cover X. Next show that
f 1, . . . ,f.. generate the unit ideal of A. Then use (Ex. 2.17b) to show that Y
is affine, and (Ex. 2.18d) to show that Y comes from an ideal a c;; A.] Note: We
will give another proof of this result using sheaves of ideals later (5.10).
(c) Let Y be a closed subset of a scheme X, and give Y the reduced induced sub-
scheme structure. If Y' is any other closed subscheme of X with the same
underlying topological space, show that the closed immersion Y -+ X factors
through Y'. We express this property by saying that the reduced induced
structure is the smallest subscheme structure on a closed subset.
(d) Let f:Z-+ X be a morphism. Then there is a unique closed subscheme Y of
X with the following property: the morphism f factors through Y, and if Y'
is any other closed subscheme of X through which f factors, then Y -+ X
factors through Y' also. We call Y the scheme-theoretic image off If Z is a
reduced scheme, then Y is just the reduced induced structure on the closure of
the image f(Z).
3.12. Closed Subschemes of Proj S.
(a) Let cp: S -+ T be a surjective homomorphism of graded rings, preserving
degrees. Show that the open set U of (Ex. 2.14) is equal to Proj T, and the
morphism f: Proj T -+ Proj S is a closed immersion.
(b) If I c;; S is a homogeneous ideal, take T = Sjl and let Y be the closed sub-
scheme of X = Proj S defined as image of the closed immersion Proj Sjl-+ X.
Show that different homogeneous ideals can give rise to the same closed sub-
scheme. For example, let d0 be an integer, and let I' = EBd"'do !d. Show that
I and I' determine the same closed subscheme.
We will see later (5.16) that every closed subscheme of X comes from a ho-
mogeneous ideal I of S (at least in the case where Sis a polynomial ring over S0 ).

92
 3 First Properties of Schemes

3.13. Properties of Morphisms of Finite Type.

(a) A closed immersion is a morphism of finite type.
(b) A quasi-compact open immersion (Ex. 3.2) is of finite type.
(c) A composition of two morphisms of finite type is of finite type.
(d) Morphisms of finite type are stable under base extension.
(e) If X and Yare schemes of finite type overS, then X x s Y is of finite type over
S.
(f) If X !... Y .!!... Z are two morphisms, and iff is quasi-compact, and g c f is of
finite type, then f is of finite type.
(g) Iff: X --> Y is a morphism of finite type, and if Y is noetherian, then X is
noetherian.
3.14. If X is a scheme of finite type over a field, show that the closed points of X are
dense. Give an example to show that this is not true for arbitrary schemes.
3.15. Let X be a scheme of finite type over a field k (not necessarily algebraically closed).
(a) Show that the following three conditions are equivalent (in which case we say
that X is geometrically irreducible).
(i) X x k k is irreducible, where k denotes the algebraic closure of k. (By
abuse of notation, we write X x k k to denote X x Speck Speck.)
(ii) X x k ks is irreducible, where ks denotes the separable closure of k.
(iii) X x k K is irreducible for every extension field K of k.
(h) Show that the following three conditions are equivalent (in which case we say X
is geometrically reduced).
(i) X x k k is reduced.
(ii) X x k kP is reduced, where kP denotes the perfect closure of k.
(iii) X x k K is reduced for all extension fields K of k.
(c) We say that X is geometrically integral if X x k k is integral. Give examples of
integral schemes which are neither geometrically irreducible nor geometrically
reduced.
3.16. Noetherian Induction. Let X be a noetherian topological space, and let fJ' be a
property of closed subsets of X. Assume that for any closed subset Y of X, iffY
holds for every proper closed subset of Y, then fY holds for Y. (In particular, fY
must hold for the empty set.) Then fY holds for X.
3.17. Zariski Spaces. A topological space X is a Zariski space if it is noetherian and
every (nonempty) closed irreducible subset has a unique generic point (Ex. 2.9).
For example, let R be a discrete valuation ring, and let T = sp(Spec R). Then
T consists of two points t 0 = the maximal ideal, t 1 = the zero ideal. The open
subsets are 0. {ti}, and T. This is an irreducible Zariski space with generic point
t 1·
(a) Show that if X is a noetherian scheme, then sp(X) is a Zariski space.
(b) Show that any minimal nonempty closed subset of a Zariski space consists of
one point. We call these closed points.
(c) Show that a Zariski space X satisfies the axiom T 0 :given any two distinct
points of X, there is an open set containing one but not the other.
(d) If X is an irreducible Zariski space, then its generic point is contained in every
nonempty open subset of X.
(e) If x 0 ,x 1 are points of a topological space X, and if x 0 E {xd -, then we say
that x 1 specializes to x 0 , written x 1 ~"W+x 0 . We also say x 0 is a specialization

93
II Schemes

of x 1 , or that x 1 is a generization of x 0 . Now let X be a Zariski space. Show

that the minimal points, for the partial ordering determined by x 1 > x 0 if x 1 Nv->
x 0 , are the closed points, and the maximal points are the generic points of the
irreducible components of X. Show also that a closed subset contains every
specialization of any of its points. (We say closed subsets are stable under
specialization.) Similarly, open subsets are stable under generization.
(f) Let t be the functor on topological spaces introduced in the proof of (2.6).
If X is a noetherian topological space, show that t(X) is a Zariski space.
Furthermore X itself is a Zariski space if and only if the map r:x: X --+ t(X) is
a homeomorphism.

3.18. Constructible Sets. Let X be a Zariski topological space. A constructible subset

of X is a subset which belongs to the smallest family 0: of subsets such that (1) every
open subset is in 0:, (2) a finite intersection of elements of 0: is in 0:, and (3) the
complement of an element of 0: is in 0:.
(a) A subset of X is locally closed if it is the intersection of an open subset with a
closed subset. Show that a subset of X is constructible if and only if it can be
written as a finite disjoint union of locally closed subsets.
(b) Show that a constructible subset of an irreducible Zariski space X is dense if
and only if it contains the generic point. Furthermore, in that case it contains
a nonempty open subset.
(c) A subset S of X is closed if and only if it is constructible and stable under
specialization. Similarly, a subset T of X is open if and only if it is constructible
and stable under generization.
(d) If f:X --+ Y is a continuous map of Zariski spaces, then the inverse image of
any constructible subset of Y is a constructible subset of X.
3.19. The real importance of the notion of constructible subsets derives from the follow-
ing theorem of Chevalley-see Cartan and Chevalley [1, expose 7] and see also
Matsumura [2, Ch. 2, §6]: let f: X --+ Y be a morphism of finite type of noetherian
schemes. Then the image of any constructible subset of X is a constructible
subset of Y. In particular, f(X), which need not be either open or closed, is a
constructible subset of Y. Prove this theorem in the following steps.
(a) Reduce to showing that f(X) itself is constructible, in the case where X and Y
are affine, integral noetherian schemes, and f is a dominant morphism.
*(b) In that case, show that f(X) contains a nonempty open subset of Y by using
the following result from commutative algebra: let A <:; B be an inclusion of
noetherian integral domains, such that B is a finitely generated A-algebra.
Then given a nonzero element b E B, there is a nonzero element a E A with
the following property: if <p: A --+ K is any homomorphism of A to an algebrai-
cally closed field K, such that rp(a) oft 0, then <p extends to a homomorphism
<p' of B into K, such that <p'(b) oft 0. [Hint: Prove this algebraic result by
induction on the number of generators of B over A. For the case of one
generator, prove the result directly. In the application, take b = 1.]
(c) Now use noetherian induction on Y to complete the proof.
(d) Give some examples of morphisms f: X --+ Y of varieties over an algebraically
closed field k, to show that f(X) need not be either open or closed.
3.20. Dimension. Let X be an integral scheme of finite type over a field k (not necessarily
algebraically closed). Use appropriate results from (I, §1) to prove the following.

94
 4 Separated and Proper Morphisms

(a) For any closed point P EX, dim X = dim((} p, where for rings, we always mean
the Krull dimension.
(b) Let K(X) be the function field of X (Ex. 3.6). Then dim X = tr.d. K(X)jk.
(c) If Y is a closed subset of X, then codim(Y,X) = inf{ dim (I)[Link] E Y}.
(d) If Y is a closed subset of X, then dim Y + codim(Y,X) = dim X.
(e) If U is a nonempty open subset of X, then dim U = dim X.
(f) If k,;; k' is a field extension, then every irreducible component of X'= X xk k'
has dimension = dim X.
3.21. Let R be a discrete valuation ring containing its residue field k. Let X =
Spec R[t] be the affine line over Spec R. Show that statements (a), (d), (e) of
(Ex. 3.20) are false for X.
*3.22. Dimension of the Fibres of a Morphism. Let f: X --+ Y be a dominant morphism
of integral schemes of finite type over a field k.
(a) Let Y' be a closed irreducible subset of Y, whose generic point rJ' is contained
in f(X). Let Z be any irreducible component of f- 1 ( Y'), such that IJ' E f(Z),
and show that codim(Z,X) ~ codim(Y',Y).
(b) Let e = dim X - dim Y be the relative dimension of X over Y. For any point
y E f(X), show that every irreducible component of the fibre Xy has dimen-
sion ~e. [Hint: Let Y' = {y}-, and use (a) and (Ex. 3.20b).]
(c) Show that there is a dense open subset U ,;; X, such that for any y E f(U),
dim UY = e. [Hint: First reduce to the case where X and Yare affine, say
X = Spec A and Y = Spec B. Then A is a finitely generated B-algebra.
Take t 1 , . . • ,teE A which form a transcendence base of K(X) over K(Y), and
let X 1 =Spec B[t 1 , •.• ,teJ. Then X 1 is isomorphic to affine e-space over Y,
and the morphism X --+ X 1 is generically finite. Now use (Ex. 3.7) above.]
(d) Going back to our original morphism f: X --+ Y, for any integer h, let Eh be
the set of points x EX such that, letting y = f(x), there is an irreducible com-
ponent Z of the fibre Xy, containing x, and having dim Z ~ h. Show that
(1) Ee = X (use (b) above); (2) if h > e, then Eh is not dense in X (use (c)
above); and (3) Eh is closed, for all h (use induction on dim X).
(e) Prove the following theorem of Chevalley-see Cartan and Chevalley [1,
expose 8]. For each integer h, let Ch be the set of points y E Y such that dim
XY = h. Then the subsets Ch are constructible, and Ce contains an open
dense subset of Y.
3.23. If V, W are two varieties over an algebraically closed field k, and if V x W is
their product, as defined in (1, Ex. 3.15, 3.16), and if t is the functor of (2.6),
then t(V X W) = t(V) X Speck t(W).

4 Separated and Proper Morphisms

We now come to two properties of schemes, or rather of morphisms between
schemes, which correspond to well-known properties of ordinary topological
spaces. Separatedness corresponds to the Hausdorff axiom for a topological
space. Properness corresponds to the usual notion of properness, namely
that the inverse image of a compact subset is compact. However, the usual
definitions are not suitable in abstract algebraic geometry, because the Zariski
topology is never Hausdorff, and the underlying topological space of a scheme

95
II Schemes

does not accurately reflect all of its properties. So instead we will use def-
initions which reflect the functorial behavior of the morphism within the
category of schemes. For schemes of finite type over C, one can show that
these notions, defined abstractly, are in fact the same as the usual notions if
we consider those schemes as complex analytic spaces in the ordinary
topology (Appendix B).
In this section we will define separated and proper morphisms. We will
give criteria for a morphism to be separated or proper using valuation rings.
Then we will show that projective space over any scheme is proper.

Definition. Let f: X -+ Y be a morphism of schemes. The diagonal morphism

is the unique morphism L1 :X -+ X x r X whose composition with both
projection maps p 1 ,p 2 :X x r X-+ X is the identity map of X-+ X. We
say that the morphism f is separated if the diagonal morphism L1 is a
closed immersion. In that case we also say X is separated over Y. A scheme
X is separated if it is separated over Spec Z.

Example 4.0.1. Let k be a field, and let X be the affine line with the origin
doubled (2.3.6). Then X is not separated over k. Indeed, X x k X is the
affine plane with doubled axes and four origins. The image of Ll is the usual
diagonal, with two of those origins. This is not closed, because all four
origins are in the closure of L1(X).

Example 4.0.2. We will see later (4.10) that if Vis any variety over an alge-
braically closed field k, then the associated scheme t(V) is separated over k.

Proposition 4.1. Iff: X -+ Y is any morphism of affine schemes, then f is

separated.
PROOF. Let X = Spec A, Y =.Spec B. Then A is a B-algebra, and X x r X
is also affine, given by Spec A @ 8 A. The diagonal morphism L1 comes from
the diagonal homomorphism A @ 8 A -+ A defined by a @a' -+ aa'. This is
a surjective homomorphism of rings, hence Ll is a closed immersion.

Corollary 4.2. An arbitrary morphism f: X -+ Y is separated if and only if

the image of the diagonal morphism is a closed subset of X x r X.
PROOF. One implication is obvious, so we have only to prove that if Ll(X) is
a closed subset, then Ll :X -+ X x r X is a closed immersion. In other words,
we have to check that L1: X -+ L1(X) is a homeomorphism, and that the
morphism of sheaves (!)XxyX-+ L1*(!)x is surjective. Let p 1 :X x r X-+ X be
the first projection. Since p 1 o L1 = idx, it follows immediately that L1 gives a
homeomorphism onto L1(X). To see that the map of sheaves (!)x x yX -+ Ll*(!)x
is surjective is a local question. For any point P EX, let U be an open affine

96
 4 Separated and Proper Morphisms

neighborhood of P which is small enough so that f(U) is contained m an

open affine subset V of Y. Then U x v U is an open affine neighborhood of
L1(P), and by the proposition, L1: U --+ U x v U is a closed immersion. So
our map of sheaves is surjective in a neighborhood of P, which completes
the proof.

Next we will discuss the valuative criterion of separatedness. The rough

idea is that in order for a scheme X to be separated, it should not contain
any subscheme which looks like a curve with a doubled point, as in the
example above. Another way of saying this is that if C is a curve, and P a
point of C, then given any morphism of C - P into X, it should admit at
most one extension to a morphism of all of C into X. (Compare (1, 6.8) where
we showed that a projective variety has this property.)
In practice, this rough idea has to be modified. The question is local, so
we replace the curve by its local ring at P, which is a discrete valuation ring.
Then since our schemes may be quite general, we must consider arbitrary
(not necessarily discrete) valuation rings. Finally, we make the criterion
relative over the image scheme Y of a morphism.
See (1, §6) for the definition and basic properties of valuation rings.

Theorem 4.3 (Valuative Criterion of Separatedness). Let f: X --+ Y be a mor-

phism of schemes, and assume that X is noetherian. Then f is separated if
and only if the following condition holds. For any field K, and for any
valuation ring R with quotient field K, let T = Spec R, let U = Spec K,
and let i: U --+ T be the morphism induced by the inclusion R [Link] K. Given
a morphism of T to Y, and given a morphism of U to X which makes a
commutative diagram

T Y,
there is at most one morphism of T to X making the whole diagram com-
mutative.

We will need two lemmas.

Lemma 4.4. Let R be a valuation ring of a field K. Let T = Spec R and let
U = SpecK. To give a morphism of U to a scheme X is equivalent to
giving a point x 1 EX and an inclusion of fields k(x 1) [Link] K. To give a
morphism of T to X is equivalent to giving two points x 0 ,x 1 in X, with x 0
a specialization (see Ex. 3.17e) ofx 1 , and an inclusion of fields k(xd [Link] K,

97
II Schemes

such that R dominates the local ring @ of x 0 on the sub scheme Z = { x 1 }-

of X with its reduced induced structure.
PROOF. U is a one-point scheme, with structure sheaf K. To give a local
homomorphism @x,.x --+ K is the same as giving an inclusion of k(x 1) <;; K,
so the first part is obvious. For the second part, let t 0 = mR be the closed
point ofT, and let t 1 = (0) be the generic point ofT. Given a morphism of
T to X, let x 0 and x 1 be the images of t 0 and t 1 . Since Tis reduced, the
morphism T--+ X factors through Z (Ex. 3.11). Furthermore, k(x 1 ) is the
function field of Z. So we have a local homomorphism of@ = @xo,z to R
compatible with the inclusion k(xd <;; K. In other words R dominates @.
Conversely, given the data consisting ofx 0 ,xt. and the inclusion k(x 1 ) <;; K
such that R dominates@, the inclusion@ --+ R gives a morphism T--+ Spec@,
which composed with the natural map Spec @ --+ X gives the desired
morphism T --+ X.
Lemma 4.5. Let f: X --+ Y be a quasi-compact morphism of schemes (see
Ex. 3.2). Then the subset f(X) of Y is closed if and only if it is stable under
specialization (Ex. 3.17e).
PROOF. One implication is obvious, so we have only to show that if f(X) is
stable under specialization, then it is closed. Clearly we may assume that
X and Y are both reduced, and that f(X)- = Y (replace Y by the reduced
induced structure on f(X)-). So let y E Y be a point. We wish to show that
y E f(X). Now we can replace Y by an affine neighborhood of y, and so
assume that Y is affine. Then since f is quasi-compact, X will be a finite
union of open affines Xi. We know that y E f(X)-. Hence y E f(XJ- for
some i. Let Y; = f(XJ- with the reduced induced structure. Then Y;
is also affine, and we will consider the dominant morphism Xi --+ Y; of
reduced affine schemes. Let Xi = Spec A and Y; = Spec B. Then the cor-
responding ring homomorphism B --+ A is injective, because the morphism
is dominant. The point y E Y; corresponds to a prime ideal p <;; B. Let
p' <;; p be a minimal prime ideal of B contained in p. (Minimal prime ideals
exist, by Zorn's lemma, because the intersection of any family of prime ideals,
totally ordered by inclusion, is again a prime ideal!) Then p' corresponds to
a point y' of Y; which specializes to y. I claim y' E f(XJ Indeed, let u~
localize A and B at p'. Localization is an exact functor, so BP' <;; A @ BP'·
Now Bp' is a field. Let q~ be any prime ideal of A @ BP'· Then q~ n BP' = (0).
Let q' <;; A be the inverse image of q~ under the localization map A --+ A @ BP'·
Then q' n B = p'. So q' corresponds to a point x' E Xi withf(x') = y'. Now
go back to the morphismf:X--+ Y. We have x' E X,f(x') = y', soy' E f(X).
But f(X) is stable under specialization by hypothesis, andy' /\1'-+ y, soy E f(X),
which is what we wanted to prove.
PROOF OF THEOREM 4.3. First suppose f is separated, and suppose given a
diagram as above where there are two morphisms h,h' of T to X making the
whole diagram commutative.

98
 4 Separated and Proper Morphisms

Then we obtain a morphism h": T --+ X x r X. Since the restrictions of h

and h' to U are the same, the generic point t 1 ofT has image in the diagonal
Ll(X). Since Ll(X) is closed, the image of t 0 is also in the diagonal. Therefore
h and h' both send the points t 0 ,t 1 to the same points x 0 ,x 1 of X. Since the
inclusions of k(x d. s K induced by h and h' are also the same, it follows
from (4.4) that hand h' are equal.
Conversely, let us suppose the condition of the theorem satisfied. To
show that f is separated, it is sufficient by (4.2) to show that Ll(X) is a closed
subset of X x r X. And since we have assumed that X is noetherian, the
morphism Ll is quasi-compact, so by (4.5) it will be sufficient to show that
Ll(X) is stable under specialization. So let ~ 1 E Ll(X) be a point, and let
~ 1 1V'-+ ~ 0 be a specialization. Let K = k(~ d and let (!) be the local ring of ~ 0
on the subscheme {~ 1 } - with its reduced induced structure. Then (!) is a
local ring contained in K, so by (I, 6.1A) there is a valuation ring R of K
which dominates(!). Now by (4.4) we obtain a morphism ofT = Spec R to
X x r X sending t 0 and t 1 to ~ 0 and ~ 1 . Composing with the projections
p 1 ,p 2 gives two morphisms ofT to X, which give the same morphism to Y,
and whose restrictions to U = Spec K are the same, since ~ 1 E Ll(X). So
by the condition, these two morphisms ofT to X must be the same. Therefore
the morphism T --+ X x r X factors through the diagonal morphism
Ll :X--+ X x r X, and so ~ 0 E Ll(X). This completes the proof. Note in the
last step it would not be sufficient to know only that p 1 (~ 0 ) = p 2 (~ 0 ). For in
general if~ EX x r X then pd~) = p 2(() does not imply¢ E Ll(X).

Corollary 4.6. Assume that all schemes are noetherian in the following state-
ments.
(a) Open and closed immersions are separated.
(b) A composition of two separated morphisms is separated.
(c) Separated morphisms are stable under base extension.
(d) If f:X--+ Y and f':X'--+ Y' are separated morphisms of schemes over
a base scheme S, then the product morphism f x f': X x s X'--+ Y x s Y'
is also separated.
(e) Iff: X --+ Y and g: Y --+ Z are two morphisms and if g o f is separated,
then f is separated.
(f) A morphism f:X--+ Y is separated if and only if Y can be covered by
open subsets V; such that f- 1 ( V;) --+ V; is separated for each i.

PROOF. These statements all follow immediately from the condition of the
theorem. We will give the proof of (c) to illustrate the method. Let f: X --+ Y

99
II Schemes

be a separated morphism, let Y' --+ Y be any morphism, and let X' =
X x y Y' be obtained by base extension. We must show that f':X'--+ Y'
is separated. So suppose we are given morphisms of T to Y' and U to X' as
in the theorem, and two morphisms of T to X' making the diagram
------+X

T ---------> Y' y
commutative. Composing with the map X' --+ X, we obtain two morphisms
of T to X. Since f is separated, these are the same. But X' is the fibred
product of X and Y' over Y, so by the universal property of the fib red prod-
uct, the two maps of T to X' are the same. Hence f' is separated.

Note on Noetherian Hypotheses. You have probably noticed that in order

to apply the theorem, it is not necessary to assume that all the schemes
mentioned in the corollary are noetherian. In fact, even in the theorem
itself, you can get by with assuming something less than X noetherian (see
Grothendieck [EGA I, new ed., 5.5.4]). My feeling is that if a noetherian
hypothesis will make statements and proofs substantially simpler, then I
will make that hypothesis, even though it may not be necessary. My justi-
fication for this attitude is that most of the motivation and examples in
algebraic geometry come from schemes of finite type over a field, and
constructions made from them, and practically all the schemes encountered
in this way are noetherian. This attitude will prevail in Chapter III, where
noetherian hypotheses are built into the very foundations of our treatment
of cohomology. The reader who wishes to avoid noetherian hypotheses is
advised to read [EGA], especially [EGA IV, §8].

Definition. A morphism f: X --+ Y is proper if it is separated, of finite type,

and universally closed. Here we say that a morphism is closed if the
image of any closed subset is closed. A morphism f: X --+ Y is universally
closed if it is closed, and for any morphism Y' --+ Y, the corresponding
morphism f': X' --+ Y' obtained by base extension is also closed.

Example 4.6.1. Let k be a field and let X be the affine line over k. Then X
is separated and of finite type over k, but it is not proper over k. Indeed,
take the base extension X --+ k. The map X x k X --+ X we obtain is the
projection map of the affine plane onto the affine line. This is not a closed
map. For example, the hyperbola given by the equation xy = 1 is a closed
subset of the plane, but its image under projection consists of the affine line
minus the origin, which is not closed.
Of course it is clear that what is missing in this example is the point at
infinity on the hyperbola. This suggests that the projective line would be

100
 4 Separated and Proper Morphisms

proper over k. In fact, we will see later (4.9) that any projective variety over
a field is proper.

Theorem 4.7 (Valuative Criterion of Properness). Let f:X--+ Y be a mor-

phism of finite type, with X noetherian. Then f is proper if and only if
for every valuation ring R and for every morphism of U to X and T to Y
forming a commutative diagram

u X

i j ..................................../jf
T y

(using the notation of (4.3) }, there exists a unique morphism T--+ X making
the whole diagram commutative.
PROOF. First assume that f is proper. Then by definition f is separated,
so the uniqueness of the morphism T--+ X will follow from (4.3), once we
know it exists. For the existence, we consider the base extension T--+ Y,
and let X r = X x y T. We get a map U --+ X T from the given maps U --+ X
and U--+ T.
U----~ Xr - - - - - - X

T y

Let ~ 1 EX r be the image of the unique point t 1 of U. Let Z = g d-.

Then Z is a closed subset of X T· Since f is proper, it is universally closed,
so the morphism f': X T --+ T must be closed, so f'(Z) is a closed subset
ofT. But f'(~ 1 } = tt> which is the generic point ofT, so in fact f'(Z) = T.
Hence there is a point ~ 0 E Z with f'(~ 0 } = t 0 . So we get a local homo-
morphism of local rings R --+ (l)~o.z corresponding to the morphism f'.
Now the function field of Z is k(~ 1 }, which is contained inK, by construc-
tion of~ 1 . By (1, 6.1A), R is maximal for the relation of domination between
local subrings of K. Hence R is isomorphic to (!) ~ 0 .z, and in particular R
dominates it. Hence by (4.4) we obtain a morphism ofT to Xr sending
t 0 ,t 1 to ~ 0 ,~ 1 . Composing with the map X r --+ X gives the desired morphism
ofT to X.
Conversely, suppose the condition of the theorem holds. To show f is
proper, we have only to show that it is universally closed, since it is of finite
type by hypothesis, and it is separated by (4.3). So let Y' --+ Y be any mor-
phism, and let f':X'--+ Y' be the morphism obtained from f by base ex-
tension. Let Z be a closed subset of X', and give it the reduced induced
structure.

101
II Schemes

Z s;: X' -----~ X

Y' - - - - - - - - + y

We need to show that f'(Z) is closed in Y'. Since f is of finite type, so is f'
and so is the restriction off' to Z (Ex. 3.13). In particular, the morphism
f':Z--+ Y' is quasi-compact, so by (4.5) we have only to show that f(Z) is
stable under specialization. So let z1 E Z be a point, let Y1 = f'(zd, and let
y 1 ~ y 0 be a specialization. Let (!) be the local ring of y 0 on { y 1 }- with its
reduced induced structure. Then the quotient field of(!) is k(y 1 ), which is a
subfield of k(z 1 ). Let K = k(z d, and let R be a valuation ring of K which
dominates(!) (which exists by (I, 6.1A) ).
From this data, by (4.4) we obtain morphisms U--+ Z and T--+ Y'
forming a commutative diagram
u --------+ z

j
T -------+ Y'.

Composing with the morphisms Z --+ X' --+ X and Y' --+ Y, we get mor-
phisms U --+ X and T --+ Y to which we can apply the condition of the
theorem. So there is a morphism of T --+ X making the diagram commute.
Since X' is a fibred product, it lifts to give a morphism T --+ X'. And since
Z is closed, and the generic point of T goes to z 1 E Z, this morphism factors
to give a morphism T--+ Z. Now let z 0 be the image of t0 . Then f'(z 0 ) =
y 0 , so Yo E f'(Z). This completes the proof.

Corollary 4.8. In the following statements, we take all schemes to be noetherian.

(a) A closed immersion is proper.
(b) A composition of proper morphisms is proper.
(c) Proper morphisms are stable under base extension.
(d) Products of proper morphisms are proper as in (4.6d).
(e) If f:X--+ Y and g: Y--+ Z are two morphisms, if go f is proper,
and if g is separated, then f is proper.
(f) Properness is local on the base as in (4.6£).
PROOF. These results follow immediately from the condition of the theorem,
taking into account (Ex. 3.13) which deals with the finite type property,
and (4.6). We will give the proof of (e) to illustrate the method. Assume
go f is proper and g is separated. Then f is of finite type by (Ex. 3.13). (We
have assumed that X is noetherian, so f is automatically quasi-compact.)
Also f is separated by (4.6). So we have to show that given a valuation
ring R, and morphisms U --+ X and T --+ Y making a commutative diagram,

102
 4 Separated and Proper Morphisms

----------~
T y

then there exists a morphism of T to X making the diagram commutative.

Let T ~ Z be the composed map. Then since g o f is proper, there is a
map of T to X commuting with the map of T ~ Z. By composing with f,
we get a second map of T to Y. But now since g is separated, the two maps
of T to Y are the same, so we are done.

Our next objective is to define projective morphisms and to show that

any projective morphism is proper. Recall that in Section 2 we defined
projective n-space PA over any ring A to be Proj A[ x0 , . . . ,xnJ. Note that
if A ~ B is a homomorphism of rings, and Spec B ~ Spec A is the corre-
sponding morphism of affine schemes, then P~ ;::;: PA x spec A Spec B. In
particular, for any ring A, we have PA ;::;: P~ x Spec z Spec A. This motivates
the following definition for any scheme Y.

Definition. If Y is any scheme, we define projective n-space over Y, denoted

P~, to be P~ x spec z Y. A morphism f:X ~ Y of schemes is projective
if it factors into a closed immersion i:X ~ P~ for some n, followed by
the projection P~ ~ Y. A morphism f:X ~ Y is quasi-projective if it
factors into an open immersionj:X ~X' followed by a projective mor-
phism g: X' ~ Y. (This definition of projective morphism is slightly
different from the one in Grothendieck [EGA II, 5.5]. The two definitions
are equivalent in case Y itself is quasi-projective over an affine scheme.)

Example 4.8.1. Let A be a ring, letS be a graded ring with S 0 = A, which

is finitely generated as an A-algebra by S 1 . Then the natural map Proj S ~
Spec A is a projective morphism. Indeed, by hypothesis S is a quotient of
a polynomial ring S' = A[x0 , . . . ,xnJ. The surjective homomorphism of
graded rings S' ~ S gives rise to a closed immersion Proj S ~ Proj S' =
PA, which shows that Proj Sis projective over A (Ex. 3.12).
Theorem 4.9. A projective morphism of noetherian schemes is proper. A quasi-
projective morphism of noetherian schemes is of finite type and separated.
PROOF. Taking into account the results of (Ex. 3.13) and (4.6) and (4.8), it
will be sufficient to show t~at X = P~ is proper over Spec Z. Recall by
(2.5) that X is a union of open affine subsets v; = D + (x;), and that v; is

103
II Schemes

isomorphic to Spec Z[ x 0/x;, . .. ,xn/xJ. Thus X is of finite type. To show

that X is proper, we will use the criterion of (4.7) and imitate the proof
of (I, 6.8). So suppose given a valuation ring R and morphisms U --+ X,
T --+ Spec Z as shown:

T - - - - - - - - + Spec Z.

Let ~ 1 EX be the image of the unique point of U. Using induction on n,

we may assume that ~ 1 is not contained in any of the hyperplanes X - V;,
which are each isomorphic to P"- 1 . In other words, we may assume that
~1 En J.'i, and hence all of the functions x;/xj are invertible elements of the
local ring @~~.
We have an inclusion k(~ 1 ) s;: K given by the morphism U --+X. Let
fu E K be the image of x;/xi. Then the fii are nonzero elements of K, and
hk = hi · jjk for all i,j,k. Let v: K --+ G be the valuation associated to the
valuation ring R. Let g; = v(/; 0 ) for i = 0, ... ,n. Choose k such that gk
is minimal among the set {g 0 , . . . ,gn}, for the ordering of G. Then for
each i we have

hence hk E R for i = 0, ... ,n. Then we can define a homomorphism

cp: Z[ x 0 /xk, ... ,xn/xk] -4 R
by sending x;/xk to hk· It is compatible with the given field inclusion k(~ d s;:
K. This homomorphism cp gives a morphism T --+ ~' and hence a mor-
phism ofT to X which is the one required. The uniqueness of this morphism
follows from the construction and the way the V; patch together.

Proposition 4.10. Let k be an algebraically closed field. The image of the

functor t: l!Jar(k) --+ 6cl)(k) of (2.6) is exactly the set of quasi-projective
integral schemes over k. The image of the set of projective varieties is the
set of projective integral schemes. In particular, for any variety V, t(V)
is an integral, separated scheme of finite type over k.
PROOF. We have already seen in Section 3 that for any variety V, the asso-
ciated scheme t( V) is integral and of finite type over k. Since varieties were
defined as locally closed subsets of projective space (I, §3), it is clear that
t(V) is also quasi-projective.
For the converse, it will be sufficient to show that any projective integral
scheme Y over k is in the image of t. Let Y be a closed subscheme of PZ,
and let V be the set of closed points of Y. Then V is a closed subset of the
variety P". Since V is dense in Y (Ex. 3.14) we see that V is irreducible, so
V is a projective variety, and we see also that t(V) and Y have the same

104
 4 Separated and Proper Morphisms

underlying topological space. But they are both reduced closed subschemes
of PZ, so they are isomorphic (Ex. 3.11).

Definition. An abstract variety is an integral separated scheme of finite type

over an algebraically closed field k. If it is proper over k, we will also
say it is complete.

Remark 4.10.1. From now on we will use the word "variety" to mean
"abstract variety" in the sense just defined. We will identify the varieties
of Chapter I with their associated schemes, and refer to them as quasi-
projective varieties. We will use the words "curve," "surface," "three-fold,"
etc., to mean an abstract variety of dimension 1, 2, 3, etc.

Remark 4.10.2. The concept of an abstract variety was invented by Weil

[1]. He needed it to provide a purely algebraic construction of the Jacobian
variety of a curve, which at first appeared only as an abstract variety
(Weil [2]). Then Chow [3] gave a different construction of the Jacobian
variety showing that it was in fact a projective variety. Later Weil [6]
himself showed that all abelian varieties were projective.
Meanwhile Nagata [1] found an example of a complete abstract non-
projective variety, showing that in fact the new class of abstract varieties
is larger than the class of projective varieties.
We can sum up the present state of knowledge of this subject as follows.
(a) Every complete curve is projective (III, Ex. 5.8).
(b) Every nonsingular complete surface is projective (Zariski [5]). See also
Hartshorne [ 5, 11.4.2].
(c) There exist singular nonprojective complete surfaces (Nagata [3]). See
also (Ex. 7.13) and (III, Ex. 5.9).
(d) There exist nonsingular complete nonprojective three-folds (Nagata
[ 4], Hironaka [2], and (Appendix B)).
(e) Every variety can be embedded as an open dense subset of a complete
variety (Nagata [6]).

The following algebraic result will be used in (Ex. 4.6).

Theorem 4.11A. If A is a subring of a field K, then the integral closure of A

in K is the intersection of all valuation rings of K which contain A.
PROOF. Bourbaki [1, Ch. VI, §1, no. 3, Thm. 3, p. 92].

EXERCISES

4.1. Show that a finite morphism is proper.

4.2. Let S be a scheme, let X be a reduced scheme over S, and let Y be a separated
scheme over S. Let f and g be two S-morphisms of X to Y which agree on an
open dense subset of X. Show that f = g. Give examples to show that this

105
II Schemes

result fails if either (a) X is nonreduced, or (b) Y is nonseparated. [Hint: Consider

the map h:X-> Y x s Y obtained from f and g.]
4.3. Let X be a separated scheme over an affine scheme S. Let U and V be open
affine subsets of X. Then U n Vis also affine. Give an example to show that this
fails if X is not separated.
4.4. Let f: X -> Y be a morphism of separated schemes of finite type over a noetherian
scheme S. Let Z be a closed subscheme of X which is proper over S. Show that
f(Z) is closed in Y, and that f(Z) with its image subscheme structure (Ex. 3.11d)
is proper over S. We refer to this result by saying that "the image of a proper
scheme is proper." [Hint: Factor f into the graph morphism r 1 :X-> X Xs Y
followed by the second projection p 2 , and show that r 1 is a closed immersion.]
4.5. Let X be an integral scheme of finite type over a field k, having function field K.
We say that a valuation of Kjk (see I, §6) has center x on X if its valuation ring R
dominates the local ring (!Jx,x·
(a) If X is separated over k, then the center of any valuation of Kjk on X (if it
exists) is unique.
(b) If X is proper over k, then every valuation of Kjk has a unique center on X.
*(c) Prove the converses of(a) and (b). [Hint: While parts (a) and (b) follow quite
easily from (4.3) and (4.7), their converses will require some comparison of
valuations in different fields.]
(d) If X is proper over k, and if k is algebraically closed, show that r(X,(!Jx) = k.
This result generalizes (1, 3.4a). [Hint: Let a E r(X,(!Jx), with a¢ k. Show that
there is a valuation ring R of Kjk with a- 1 E mR. Then use (b) to get a con-
tradiction. J
Note. If X is a variety over k, the criterion of (b) is sometimes taken as the de-
finition of a complete variety.
4.6. Let f: X -> Y be a proper morphism of affine varieties over k. Then f is a finite
morphism. [Hint: Use (4.11A).]
4.7. Schemes Over R. For any scheme X 0 over R, let X= X 0 x R C. Let ocC-> C be
complex conjugation, and let a: X -> X be the automorphism obtained by keeping
X 0 fixed and applying r:t. to C. Then X is a scheme over C, and a is a semi-linear
automorphism, in the sense that we have a commutative diagram
X _______a____~
X

l
Spec C
l
Spec C.

Since a 2 = id, we call a an involution.

(a) Now let X be a separated scheme of finite type over C, let a be a semilinear
involution on X, and assume that for any two points xl>x 2 EX, there is an
open affine subset containing both of them. (This last condition is satisfied
for example if X is quasi-projective.) Show that there is a unique separated
scheme X 0 of finite type over R, such that X 0 x R C ~ X, and such that this
isomorphism identifies the given involution of X with the one on X 0 x R C
described above.

106
 4 Separated and Proper Morphisms

For the following statements, X 0 will denote a separated scheme of finite

type over R, and X,a will denote the corresponding scheme with involution
over C.
(b) Show that X 0 is affine if and only if X is.
(c) If X 0 ,Y0 are two such schemes over R, then to give a morphism f 0 :X 0 -+ Y0
is equivalent to giving a morphism f:X-+ Y which commutes with the in-
volutions, i.e., f o ax = ay of
(d) If X~ A~, then X 0 ~ A~.
(e) If X ~ P~, then either X 0 ~ P~, or X 0 is isomorphic to the conic in Pi given
by the homogeneous equation x~ + xf + x~ = 0.
4.8. Let [l}J be a property of morphisms of schemes such that:
(a) a closed immersion has&;
(b) a composition of two morphisms having [l}J has&;
(c) [l}J is stable under base extension.
Then show that:
(d) a product of morphisms having & has&;
(e) if f:X-+ Y and g: Y-+ Z are two morphisms, and if go f has & and g is
separated, then f has&;
(f) If f:X -+ Y has&, then feed :X"d -+ Yced has&.
[Hint: For (e), consider the graph morphism r 1 :X-+ X x z Y and note that
it is obtained by base extension from the diagonal morphism Ll: Y -+ Y x z Y.]
4.9. Show that a composition of projective morphisms is projective. [Hint: Use the
Segre embedding defined in (1, Ex. 2.14) and show that it gives a closed immersion
P' x ps -+ prs+r+s.J Conclude that projective morphisms have properties
(a)-(f) of (Ex. 4.8) above.
*4.10. Chow's Lemma. This result says that proper morphisms are fairly close to pro-
jective morphisms. Let X be proper over a noetherian scheme S. Then there is
a scheme X' and a morphism g: X' -+ X such that X' is projective over S, and
there is an open dense subset U ~ X such that g induces an isomorphism of
g- 1 (U) to U. Prove this result in the following steps.
(a) Reduce to the case X irreducible.
(b) Show that X can be covered by a finite number of open subsets V;, i = 1, . .. ,n,
each of which is quasi-projective over S. Let Vi -+ Pi be an open immersion
of Vi into a scheme Pi which is projective overS.
n
(c) Let V = Vi, and consider the map
f:V-+X x 5 P 1 x 5 • .. x 5 Pn
deduced from the given maps V -+ X and V -+ Pi. Let X' be the closed image
subscheme structure (Ex. 3.1ld) f( V) -. Let g: X' -+ X be the projection onto
the first factor, and let h: X' -+ P = P 1 x s . . . x s P" be the projection onto
the product of the remaining factors. Show that h is a closed immersion,
hence X' is projective over S.
(d) Show that g- 1(V) -+ Vis an isomorphism, thus completing the proof.
4.11. If you are willing to do some harder commutative algebra, and stick to noetherian
schemes, then we can express the valuative criteria of separatedness and properness
using only discrete valuation rings.
(a) If (IJ,m is a noetherian local domain with quotient field K, and if Lis a finitely
generated field extension of K, then there exists a discrete valuation ring R of

107
II Schemes

L dominating @. Prove this in the following steps. By taking a polynomial

ring over (1), reduce to the case where Lis a finite extension field of K. Then
show that for a suitable choice of generators x 1, . . . ,x" of m, the ideal a = (x 1 )
in (!!' = (I [x 2 /x 1, ••. ,x"/x 1] is not equal to the unit ideal. Then let p be a
minimal prime ideal of a, and let (!!'~'be the localization of(!!' at p. This is a
noetherian local domain of dimension 1 dominating@. Let lP'~'be the integral
closure of (!!'~'in L. Use the theorem of Krull-Akizuki (see Nagata [7, p. 115])
to show that if·., is noetherian of dimension 1. Finally, take R to be a local-
ization of if'., at one of its maximal ideals.
(b) Let f: X -+ Y be a morphism of finite type of noetherian schemes. Show that
f is separated (respectively, proper) if and only if the criterion of (4.3) (respec-
tively, (4.7)) holds for all discrete valuation rings.
4.12. Examples of Valuation Rings. Let k be an algebraically closed field.
(a) If K is a function field of dimension I over k (I, §6), then every valuation ring
of Kjk (except for K itself) is discrete. Thus the set of all of them is just the
abstract nonsingular curve CK of (1, §6).
(b) If K/k is a function field of dimension two, there are several different kinds of
valuations. Suppose that X is a complete·nonsingular surface with function
field K.
(I) If Y is an irreducible curve on X, with generic point X~o then the local ring
R = (!! x,.x is a discrete valuation ring of Kjk with center at the (nonclosed)
point x 1 on X.
(2) Iff: X' -+ X is a birational morphism, and if Y' is an irreducible curve in
X' whose image in X is a single closed point x 0 , then the local ring R of
the generic point of Y' on X' is a discrete valuation ring of Kjk with center
at the closed point x 0 on X.
(3) Let x 0 E X be a closed point. Let f: X 1 -+ X be the blowing-up of x 0
(I, §4) and let E 1 = f- \x 0 ) be the exceptional curve. Choose a closed
point x 1 E £ 1 , let / 2 : X 2 -+ X 1 be the blowing-up of X~o and let £ 2 =
/2 1(x 1 ) be the exceptional curve. Repeat. In this manner we obtain a
sequence of varieties X; with closed points X; chosen on them, and for
each i, the local ring (lx,+,.x,., dominates (!ix,,x,. Let R 0 = Ui~o (!ix,.x,·
Then R 0 is a local ring, so it is dominated by some valuation ring R of
Kjk by (I, 6.1A). Show that R is a valuation ring of Kjk, and that it has
center x 0 on X. When is R a discrete valuation ring?
Note. We will see later (V, Ex. 5.6) that in fact the R 0 of(3) is already a valuation
ring itself, so R 0 = R. Furthermore, every valuation ring of Kjk (except for K
itself) is one of the three kinds just described.

5 Sheaves of Modules
So far we have discussed schemes and morphisms between them without
mentioning any sheaves other than the structure sheaves. We can increase
the flexibility of our technique enormously by considering sheaves of modules
on a given scheme. Especially important are quasi-coherent and coherent
sheaves, which play the role of modules (respectively, finitely generated
modules) over a ring.

108
 5 Sheaves of Modules

In this section we will develop the basic properties of quasi-coherent and

coherent sheaves. In particular we will introduce the important "twisting
sheaf" @(1) of Serre on a projective scheme.
We will start by defining sheaves of modules on a ringed space.

Definitions. Let (X,@x) be a ringed space (see §2). A sheaf of @x-modules

(or simply an @x-module) is a sheaf !F on X, such that for each open set
U s;:: X, the group !F(U) is an @x(U)-module, and for each inclusion of
open sets V s;:: U, the restriction homomorphism !F(U) --+ !F(V) is com-
patible with the module structures via the ring homomorphism @x(U) --+
@x(V). A morphism !F --+ ':§ of sheaves of @x-modules is a morphism of
sheaves, such that for each open set U s;:: X, the map !F(U) --+ ':#(U) is a
homomorphism of @x(U)-modules.
Note that the kernel, co kernel, and image of a morphism of @x-modules
is again an @x-module. If !F' is a subsheaf of @x-modules of an @x-module
!F, then the quotient sheaf !Fj!F' is an @x-module. Any direct sum,
direct product, direct limit, or inverse limit of (9 x-modules is an (9 x-module.
If !F and':§ are two @x-modules, we denote the group ofmorphisms from
!F to ':§ by Homlllx(!F,':#), or sometimes Homx(ff,':#) or Hom(!#','§) if no
confusion can arise. A sequence of @x-modules and morphisms is exact
if it is exact as a sequence of sheaves of abelian groups.
If U is an open subset of X, and if !F is an @x-module, then :Flu is an
@xlu-module. If !F and ':§ are two @x-modules, the presheaf
U f--+ Homlllxlu(!Fiu,':#lu)
is a sheaf, which we call the sheaf J'fom (Ex. 1.15), and denote by
J'fomlllx(!F,':#). It is also an @x-module.
We define the tensor product !F ®lllx ':§ of two @x-modules to be the
sheaf associated to the presheaf U f--+ !F(U) ®lllx(UJ ':#(U). We will often
write simply !F 0 ':#,with (Qx understood.
An @x-module !F is free if it is isomorphic to a direct sum of copies of
@x. It is locally free if X can be covered by open sets U for which :Flu
is a free @xlu-module. In that case the rank of !F on such an open set is
the number of copies of the structure sheaf needed (finite or infinite).
If X is connected, the rank of a locally free sheaf is the same everywhere.
A locally free sheaf of rank 1 is also called an invertible sheaf
A sheaf of ideals on X is a sheaf of modules f which is a subsheaf
of (Qx· In other words, for every open set U, f(U) is an ideal in @x(U).
Let f:(X,@x)--+ (Y,@y} be a morphism of ringed spaces (see §2). If
!F is an @x-module, then f*!F is an f*@x-module. Since we have the
morphism f# :@y--+ f*(Qx of sheaves of rings on Y, this gives f*!F a
natural structure of @y-module. We call it the direct image of !F by the
morphism!
Now let':§ be a sheaf of @y-modules. Then f- 1 '§ is an f- 1 @y-module.
Because of the adjoint property off - l (Ex. 1.18) we have a morphism

109
II Schemes

f- 1 (!Jy--+ (!Jx of sheaves of rings on X. We define f*<§ to be the tensor

product
f-l<g ®J-•i!!y (!Jx·
Thus f*<§ is an @x-module. We call it the inverse image of <§ by the
morphism!
As in (Ex. 1.18) one can show that f* and f* are adjoint functors
between the category of (!Jx-modules and the category of @y-modules.
To be precise, for any @x-module ff and any @y-module <§, there is a
natural isomorphism of groups
Homl!!x(f*<§,ff) ~ Homi!Jy(<§,f*ff).
Now that we have the general notion of a sheaf of modules on a ringed
space, we specialize to the case of schemes. We start by defining the sheaf
of modules M on Spec A associated to a module M over a ring A.

Definition. Let A be a ring and let M be an A-module. We define the sheaf

associated to M on Spec A, denoted by M, as follows. For each prime
ideal p s;:: A, let MP be the localization of M at p. For any open set
U s;:: Spec A we define the group M(U) to be the set of functions s: U --+
Upe u MP such that for each p E U, s(p) E MP, and such that sis locally
a fraction m/f with mE M and f EA. To be precise, we require that for
each p E U, there is a neighborhood V of p in U, and there are elements
mE M and f E A, such that for each q E V, f 4 q, and s(q) = m/f in Mq.
We make Minto a sheaf by using the obvious restriction maps.

Proposition 5.1. Let A be a ring, let M be an A-module, and let M be the

sheaf on X = Spec A associated toM. Then:
(a) M is an (!Jx-module;
(b) for each p EX, the stalk (M)v of the sheaf M at p is isomorphic to
the localized module Mv;
(c) for any f E A, the A rmodule M(D(f)) is isomorphic to the localized
module M 1 ;
(d) in particular, r(X,M) = M.
PROOF. Recalling the construction of the structure sheaf (!Jx from §2, it is
clear that M is an @x-module. The proofs of (b), (c), (d) are identical to the
proofs of (a), (b), (c) of (2.2), replacing A by Mat appropriate places.

Proposition 5.2. Let A be a ring and let X = Spec A. Also let A --+ B be a
ring homomorphism, and let f: Spec B --+ Spec A be the corresponding
morphism of spectra. Then:
(a) the map M --+ M gives an exact, fully faithful functor from the
category of A-modules to the category of @x-modules;
(b) if M and N are two A-modules, then (M ®A Nf ~ M ®l!!x N;
(c) if {M;} is any family of A-modules, then (ffiM;f ~ ffiMi;

110
 5 Sheaves of Modules

(d) for any B-module N we have f*(il) ~ (AN)-, where AN means N

considered as an A -module;
(e) for any A-module M we have f*(M) ~ (M ®A B)-.
PROOF. The map M ...... M is clearly functorial. It is exact, because localiza-
tion is exact, and exactness of sheaves can be measured at the stalks (use
(Ex. 1.2) and ([Link])). It commutes with direct sum and tensor product,
because these commute with localization. To say it is fully faithful means
that for any A-modules M and N, we have HomA(M,N) = Hom(l)x(M,N).
The functor ~ gives a natural map HomA(M,N) ...... Hom(l)x(M,N). Applying
r and using ([Link]) gives a map the other way. These two maps are clearly
inverse to each other, hence isomorphisms. The last statements about f*
and f* follow directly from the definitions.

These sheaves of the form M on affine schemes are our models for quasi-
coherent sheaves. A quasi-coherent sheaf on a scheme X will be an mx-
module which is locally of the form M. In the next few lemmas and propo-
sitions, we will show that this is a local property, and we will establish some
facts about quasi-coherent and coherent sheaves.

Definition. Let (X,(!)x) be a scheme. A sheaf of mx-modules ~ is quasi-

coherent if X can be covered by open affine subsets U; = Spec A;, such
that for each i there is an A;-module M; with ~lu, ~ M;. We say that
~ is coherent if furthermore each M; can be taken to be a finitely gen-
erated A;-module.

Although we have just defined the notion of quasi-coherent and coherent

sheaves on an arbitrary scheme, we will normally not mention coherent
sheaves unless the scheme is noetherian. This is because the notion of
coherence is not at all well-behaved on a nonnoetherian scheme.

Example 5.2.1. On any scheme X, the structure sheaf (!)x is quasi-coherent

(and in fact coherent).

Example 5.2.2. If X = Spec A is an affine [Link], if Y ~ X is the closed

subscheme defined by an ideal a s; A (3.2.3), and if i: Y ...... X is the in-
clusion morphism, then i*(!)y is a quasi-coherent (in fact coherent) mx-
module. Indeed, it is isomorphic to (A/ar.

Example 5.2.3. If U is an open subscheme of a scheme X, with inclusion

map j: U ...... X, then the sheaf j!((!)u) obtained by extending (!)u by zero
outside of U (Ex. 1.19), is an mx-module, but it is not in general quasi-
coherent. For example, suppose X is integral, and V = Spec A is any
open affine subset of X, not contained in U. Then j!((!)u )lv has no global

111
II Schemes

sections over V, and yet it is not the zero sheaf. Hence it cannot be of the
form M for any A-module M.

Example 5.2.4. If Y is a closed subscheme of a scheme X, then the sheaf

(()xlr is not in general quasi-coherent on Y. In fact, it is not even an @y-
module in general.

Example 5.2.5. Let X be an integral noetherian scheme, and let X be the

constant sheaf with group K equal to the function field of X (Ex. 3.6). Then
X is a quasi-coherent (()x-module, but it is not coherent unless X is reduced
to a point.

Lemma 5.3. Let X ;= Spec A be an affine scheme, let f E A, let D(f) s;: X
be the corresponding open set, and let §' be a quasi-coherent sheaf on X.
(a) If s E r(X,ff') is a global section of§' whose restriction to D(f)
is 0, then for some n > 0, rs = 0.
(b) Given a section t E §'(D(f)) of §' over the open set D(f), then
for some n > 0, rt extends to a global section of §' over X.
PROOF. First we note that since §' is quasi-coherent, X can be covered by
open affine subsets of the form V = Spec B, such that ff'lv ~ M for some
B-module M. Now the open sets ofthe form D(g) form a base for the topology
of X (see §2), so we can cover V by open sets of the form D(g), for various
g EA. An inclusion D(g) s;: V corresponds to a ring homomorphism
B---+ A 9 by (2.3). Hence 9""lv<o> ~ (M 0B A 9 ) - by (5.2). Thus we have shown
that if§' is quasi-coherent on X, then X can be covered by open sets of the
form D(g;) where for each i, ff'lv(g;) ~ Mi for some module Mi over the ring
A 9 ,. Since X is quasi-compact, a finite number of these open sets will do.
(a) Now suppose givens E r(X,9"") with siv<n = 0. For each i, s restricts
to give a section si of§' over D(g;), in other words, an element si E Mi (using
([Link]) ). Now D(f) n D(g;) = D(fg;), so 9""lv(Jgd = (MJj using ([Link]). Thus
the image of si in (MJ1 is zero, so by the definition of localization, rsi = 0
for some n. This n may depend on i, but since there are only finitely many i,
we can pick n large enough to work for them all. Then since the D(g;) cover
X, we have rs = 0.
(b) Given an element t E :#'(D(f) ), we restrict it for each ito get an element
t of :#'(D(fg;)) = (M;) 1 . Then by the definition of localization, for some
n > 0 there is an element ti E Mi = :#'(D(g;)) which restricts to rt on
D(fgJ The integer n may depend on i, but again we take one large enough
to work for all i. Now on the intersection D(g;) n D(gj) = D(gigj) we have
two sections ti and tj of §', which agree on D(fgigj) where they are both
equal to rt. Hence by part (a) above, there is an integer m > 0 such that
fm(ti - tj) = 0 on D(gigj). This m depends on i and j, but we take one m
large enough for all. Now the local sections fmti of§' on D(gJ glue together
to give a global sections of:#', whose restriction to D(f) is r+mt.

112
 5 Sheaves of Modules

Proposition 5.4. Let X be a scheme. Then an (l}x-module ff is quasi-coherent if

and only if for every open affine subset U = Spec A of X, there is an A-
module M such that fflu ~ M. If X is noetherian, then ff is coherent if
and only if the same is true, with the extra condition that M be a finitely
generated A-module.
PROOF. Let ff be quasi-coherent on X, and let U = Spec A be an open
affine. As in the proof of the lemma, there is a base for the topology consisting
of open affines for which the restriction of ff is the sheaf associated to a
module. It follows [Link] is quasi-coherent, so we can reduce to the case
X affine = Spec A. Let M = r(X,ff). Then in any case there is a natural
map rx: M --. ff (Ex. 5.3). Since ff is quasi-coherent, X can be covered by
open sets D(g;) with fflv(g;) ~ M; for some A9 ,-module M;. Now the lemma,
applied to the open set D(g;), tells us exactly that ff(D(g;)) ~ M 9 ,, so
M; = M 9 ,. It follows that the map rx, restricted to D(g;), is an isomorphism.
The D(g;) cover X, so rx is an isomorphism.
Now suppose that X is noetherian, and ff coherent. Then, using the
above notation, we have the additional information that each M 9 , is a
finitely generated A9 ,-module, and we want to prove that M is finitely
generated. Since the rings A and A 9 , are noetherian, the modules M 9 , are
noetherian, and we have to prove that M is noetherian. For this we just use
the proof of (3.2) with A replaced by M in appropriate places.

Corollary 5.5. Let A be a ring and let X = Spec A. The functor M r--+ M
gives an equivalence of categories between the category of A-modules and
the category of quasi-coherent (!} x-modules. Its inverse is the functor
ff r--+ r(X,ff). If A is noetherian, the same functor also gives an equiv-
alence of categories between the category of finitely generated A-modules
and the category of coherent (l}x-modules.
PROOF. The only new information here is that ff is quasi-coherent on X if
and only if it is of the form M, and in that case M = r(X,ff). This follows
from (5.4).

Proposition 5.6. Let X be an affine scheme, let 0 --. ff' --. ff --. ff" --. 0 be
an exact sequence of (l}x-modules, and assume that ff' is quasi-coherent.
Then the sequence
0 --. r(X,ff') --. r(X,ff) --. r(X,ff") --. 0
is exact.
PROOF. We know already that r is a left-exact functor (Ex. 1.8) so we have
only to show that the last map is surjective. Let s E r(X,ff") be a global
section of ff". Since the map of sheaves ff --. ff" is surjective, for any
x E X there is an open neighborhood D(f) of x, such that sfv<n lifts to a
section t E ff(D(f)) (Ex. 1.3). I claim that for some n > 0, f"s lifts to a
global section of ff. Indeed, we can cover X with a finite number of open

I I3
II Schemes

sets D(g;), such that for each i, siv<9 il lifts to a section t; E ff(D(g;) ). On
D(f) n D(g;) = D(fg;), we have two sections t,t; E ff(D(fg;)) both lifting s.
Therefore t - t; E ff'(D(fg;) ). Since.?!'' is quasi-coherent, by (5.3b) for some
n > 0, r(t - t;) extends to a section U; E ff'(D(g;) ). As usual, we pick one
n to work for all i. Let t; = rti + U;. Then t; is a lifting of rs on D(g;),
and furthermore andt; rr agree on D(fg;). Now on D(g;gj) we have two
sections ti and tj of.?!', both of which lift rs, so ti - tj E ff'(D(g;g) ). Further-
more, t; and tj are equal on D(fg;gj), so by (5.3a) we have fm(t; - tj) = 0 for
some m > 0, which we may take independent of i and j. Now the sections
fmt; of.?!' glue to give a global section t" of.?!' over X, which lifts r+ms.
This proves the claim.
Now cover X by a finite number of open sets D(/;), i = 1, ... ,r, such that
slvuil lifts to a section of.?!' over D(/;) for each i. Then by the claim, we can
find an integer n (one for all i) and global sections t; E r(X,ff) suchthat t;
is a lifting of fis. Now the open sets D(/;) cover X, so the ideal (f~, . .. ,f~)
is the unit ideal of A, and we can write 1 = Lt=
1 aJi, with a; EA. Let
t = Ia;t;. Then t is a global section of .?!' whose image in r(X,ff") is
IaJis = s. This completes the proof.

Remark 5.6.1. When we have developed the techniques of cohomology, we

will see that this proposition is an immediate consequence of the fact that
H 1 (X,ff') = 0 for any quasi-coherent sheaf .?!'' on an affine scheme X
(III, 3.5).

Proposition 5.7. Let X be a scheme. The kernel, cokernel, and image of any
morphism of quasi-coherent sheaves are quasi-coherent. Any extension of
quasi-coherent sheaves is quasi-coherent. If X is noetherian, the same is
true for coherent sheaves.
PROOF. The question is local, so we may assume X is affine. The statement
about kernels, cokernels and images follows from the fact that the functor
M r--+ M is exact and fully faithful from A-modules to quasi-coherent sheaves
(5.2a and 5.5). The only nontrivial part is to show that an extension of quasi-
coherent sheaves is quasi-coherent. So let 0 --+ .?!'' --+ .?!' --+ :JP' --+ 0 be an
exact sequence of l'Dx-modules, with :F and $'" quasi-coherent. By (5.6),
the corresponding sequence of global sections over X is exact, say
0 --+ M' --+ M --+ M" --+ 0. Applying the functor "', we get an exact com-
mutative diagram

0 --+ $'' --+ $' --+ $'" --+ 0.

The two outside arrows are isomorphisms, since §'' and §'" are quasi-
coherent. So by the 5-lemma, the middle one is also, showing that .?!' is
quasi-coherent.

114
 5 Sheaves of Modules

In the noetherian case, if :F' and :#'" are coherent, then M' and· M"
are finitely generated, so M is also finitely generated, and hence :F is
coherent.

Proposition 5.8. Let f: X --+ Y be a morphism of schemes.

(a) If '!I is a quasi-coherent sheaf of @y-modules, then f*'!l is a quasi-
coherent sheaf of (!) x-modules.
(b) If X andY are noetherian, and if '!I is coherent, then f*'!l is coherent.
(c) Assume that either X is noetherian, or f is quasi-compact (Ex. 3.2)
and separated. Then if :F is a quasi-coherent sheaf of {!}x-modules, f*:F
is a quasi-coherent sheaf of @y-modules.
PROOF.
(a) The question is local on both X and Y, so we can assume X and Y both
affine. In this case the result follows from (5.5) and (5.2e).
(b) In the noetherian case, the same proof works for coherent sheaves.
(c) Here the question is local on Y only, so we may assume that Y is affine.
Then X is quasi-compact (under either hypothesis) so we can cover X with
a finite number of open affine subsets U;. In the separated case, U; n Uj is
again affine (Ex. 4.3). Call it Uijk· In the noetherian case, U; n Uj is at least
quasi-compact, so we can cover it with a finite number of open affine subsets
Uijk· Now for any open subset V of Y, giving a section s of :F over f- 1 Vis
the same thing as giving a collection of sections S; of :F over u-
1 V) (\ U;

whose restrictions to the open subsets f- (V) n Uijk are all equal. This is
1

just the sheaf property (§1). Therefore, there is an exact sequence of sheaves
on Y,
0 --+ f*:F --+ EB f*(:Fiu) --+ EB f*(:Fiu, J,
1
i i,j,k

where by abuse of notation we denote also by f the induced morphisms

U; --+ Y and Uijk --+ Y. Now !*(:Flu) and f*(:FiuijJ are quasi-coherent by
(5.2d). Thus f*:F is quasi-coherent by (5.7).

Caution 5.8.1. If X and Yare noetherian, it is not true in general that f* of

a coherent sheaf is coherent (Ex. 5.5). However, it is true iff is a finite
morphism (Ex. 5.5) or a projective morphism (5.20) or (III, 8.8), or more
generally, a proper morphism: see Grothendieck [EGA III, 3.2.1].

As a first application of these concepts, we will discuss the sheaf of ideals

of a closed subscheme.

Definition. Let Y be a closed subscheme of a scheme X, and let i: Y --+ X be

the inclusion morphism. We define the ideal sheaf of Y, denoted § y, to
be the kernel of the morphism i# :(!)x--+ i*@y.

115
II Schemes

Proposition 5.9. Let X be a scheme. For any closed subscheme Y of X, the

corresponding ideal sheaf 5 y is a quasi-coherent sheaf of ideals on X. If
X is noetherian, it is coherent. Conversely, any quasi-coherent sheaf of
ideals on X is the ideal sheaf of a uniquely determined closed subscheme
of X.
-
PROOF. If Yis a closed subscheme of X, then the inclusion morphism i: Y -+ X
is quasi-compact (obvious) and separated (4.6), so by (5.8), i*(!)Y is quasi-
coherent on X. Hence 5 y, being the kernel of a morphism of quasi-coherent
sheaves, is also quasi-coherent. If X is noetherian, then for any open affine
subset U = Spec A of X, the ring A is noetherian, so the ideal I = r(U,5 rlu ),
is finitely generated, so 5 y is coherent.
Conversely, given a scheme X and a quasi-coherent sheaf of ideals /,
let Y be the support of the quotient sheaf(!) xl f. Then Y is a subspace of X,
and (I;(!)x//) is the unique closed subscheme of X with ideal sheaf f. The
unicity is clear, so we have only to check that ( Y, (!)x/f) is a closed sub-
scheme. This is a local question, so we may assume X = Spec A is affine.
Since/ is quasi-coherent, cf = a for some ideal a <;; A. Then (Y,(Ilx//) is
just the closed subscheme of X determined by the ideal a (3.2.3).

Corollary 5.10. If X = Spec A is an affine scheme, there is a 1-1 correspon-

dence between ideals a in A and closed subschemes Y of X, given by a~
image of Spec A/a in X (3.2.3). In particular, every closed subscheme of
an affine scheme is affine.
PROOF. By (5.5) the quasi-coherent sheaves of ideals on X are in 1-1 corre-
spondence with the ideals of A.

Our next concern is to study quasi-coherent sheaves on the Proj of a

graded ring. As in the case of Spec, there is a connection between modules
over the ring and sheaves on the space, but it is more complicated.

Definition. Let S be a graded ring and let M be a graded S-module. (See

(1, §7) for generalities on graded modules.) We define the sheaf associated
toM on Proj S, denoted by M, as follows. For each p E Proj S, let M(p)
be the group of elements of degree 0 in the localization y-t M, where T
is the multiplicative system of homogeneous elements of S not in p
(cf. definition of Proj in §2). For any open subset U <;; Proj S we define
M(U) to be the set of functions s from u to upEU M(p) which are locally
fractions. This means that for every p E U, there is a neighborhood V of
p in U, and homogeneous elements m E M and f E S of the same degree,
such that for every q E V, we have f ¢ q, and s(q) = m/f in M(q)· We make
M into a sheaf with the obvious restriction maps.
Proposition 5.11. Let S be a graded ring, and M a graded S-module. Let
X= Proj S.

116
 5 Sheaves of Modules

(a) For any p EX, the stalk (M)v = M<vJ· _

(b) For any homogeneous f E S+, we have Mln+<fl ~ (Mur via the
isomorphism of D +(f) with Spec S<n (see (2.5b) ), where M<n denotes the
group o[ elements of degree 0 in the localized module M J·
(c) M is a quasi-coherent {!}x-module. If S is noetherian and M is
finitely generated, then M is coherent.

PROOF. For (a) and (b), just repeat the proof of (2.5), with M in place of S.
Then (c) follows from (b).

Definition. LetS be a graded ring, and let X = Proj S. For any n E Z, we

define the sheaf (!Jx(n) to be S(nr. We call (!Jx(l) the twisting sheaf of
Serre. For any sheaf of {!}x-modules, :F, we denote by :F(n) the twisted
sheaf :F ®mx (!Jx(n).

Proposition 5.12. Let S be a graded ring and let X = Proj S. Assume that S
is generated by S 1 as an S 0 -algebra.
(a) The sheaf(!Jx(n) is an invertible sheaf on X.
(b) For any graded S-module M, M(n) ~ (M(n) r.
In particular,
(!Jx(n) ® (!Jx(m) ~ (!Jx(n + m).
(c) Let T be another graded ring, generated by T 1 as a T 0 -algebra,
let q>: S --+ T be a homomorphism preserving degrees, and let U <;; Y =
Proj T and f: U --+X be the morphism determined by q> (Ex. 2.14). Then
f*((!Jx(n)) ~ (!Jy(n)lu and f*((!Jy(n)iu) ~ (f*(!Ju)(n).
PROOF.
(a) Recall that invertible means locally free of rank 1. Let f E S 1, and
consider the restriction (!Jx(n)ln+ <fl· By the previous proposition this is
isomorphic to S(n)(jl on Spec S<n· We will show that this restriction is free
of rank 1. Indeed, S(n)<n is a free S<n-module of rank 1. For S<n is the group
of elements of degree 0 in S1 , and S(n)<n is the group of elements of degree n
in S1 . We obtain an isomorphism of one to the other by sending s tors.
This makes sense, for any n E Z, because f is invertible inS1 . Now since S
is generated by S 1 as an S0 -algebra, X is covered by the open sets D +(f)
for f E S 1. Hence (!J(n) is invertible.
(b) This follows from the fact that (M ®s Nr ~ M ®mx N for any two
graded S-modules M and N, when Sis generated by S 1 . Indeed, for any
f E S 1 we have (M ®s N)<n = M<n ®s(f> N<n·
(c) More generally, for any graded S-module M, f*(M) ~ (M ®s Tr lu
and for any graded T-module N, f*(Niu) ~ (8 N)-. Furthermore, the sheaf
T on X is just f*( (!Ju ). The proofs are straightforward (cf. (5.2) for the affine
case).

The twisting operation allows us to define a graded S-module associated

to any sheaf of modules on X = Proj S.

117
II Schemes

Definition. Let S be a graded ring, let X = Proj S, and let !!1' be a sheaf of
(()x-modules. We define the graded S-module associated to !!1' as a group, to
be r *(.~) = ffin E z T(X,!!i'(n) ). We give it a structure of graded S-module
as follows. If s E Sd, then s determines in a natural way a global section
s E T(X,(()x(d) ). Then for any t E T(X,!!i'(n)) we define the products· tin
T(X,!!i'(n + d)) by taking the tensor products ® t and using the natural
map !!i'(n) ® (()x(d) ~ !!i'(n + d).

Proposition 5.13. Let A be a ring, let S = A[ x 0 , . .. ,xr], r ~ 1, and let

X = Proj S. (This is just projective r-space over A.) Then T *((()x) ~ S.
PROOF. We cover X with the open sets D+(x;). Then to give a section
t E T(X,Gx(n)) is the same as giving sections t; E (()x(n)(D +(x;)) for each i,
which agree on the intersections D +(x;xi). Now t; is just a homogeneous
element of degree n in the localization Sx,, and its restriction to D+(x;x) is
just the image of that element in Sx,x1 . Summing over all n, we see that
r *(0x) can be identified with the set of (r + 1)-tuples (t 0 , . . . ,t,) where for
each i, t; E S,,, and for each i,j, the images oft; and ti in Sx,x1 are the same.
Now the X; are not zero divisors in S, so the localization maps S ~ Sx,
and Sx, ~ Sx,x1 are all injective, and these rings are all subrings of S' =
Sxo···xr· Hence T*(lDx) is the intersection nsx, taken insideS'. Now
any homogeneous element of S' can be written uniquely as a product
x~ · · · x~f(x 0 , . . . ,x,), where the ii E Z, and f is a homogeneous polynomial
not divisible by any X;. This element will be in Sx, if and only if ii ~ 0 for
j =1 i. It follows that the intersection of all the Sx, (in fact the intersection of
any two of them) is exactly S.

Caution 5.13.1. If S is a graded ring which is not a polynomial ring, then it

is not true in general that r *(lDx) = S (Ex. 5.14).

Lemma 5.14. Let X be a scheme, let 2! be an invertible sheaf on X, let f E

T(X,ff!), let X f be the open set of points x E X where fx ¢; mxff! x• and let
'"W be a quasi-coherent sheaf on X.
(a) Suppose that X is quasi-compact, and let s E T(X,!!i') be a global
section of !!1' whose restriction to X f is 0. Then for some n > 0, we have
fns = 0, where f"s is considered as a global section of !!1' ® ff!!?;~n_
(b) Suppose furthermore that X has a finite covering by open affine
subsets U;, such that 2iu, is free for each i, and such that U; n Ui is quasi-
compact for each i,j. Given a section t E T(X 1 ,!!1'), then for some n > 0,
the section f"t E T(X 1 ,!!1' ® ff!®n) extends to a global section of !!1' ® ff!®n.
PROOF. This lemma is a direct generalization of (5.3), with an extra twist due
to the presence of the invertible sheaf 2!. It also generalizes (Ex. 2.16). To
prove (a), we first cover X with a finite number (possible since X is quasi-
compact) of open affines U = Spec A such that 2lu is free. Let 1/f: 2lu ~ lDu
be an isomorphism expressing the freeness of 2lu· Since !!1' is quasi-coherent,

118
 5 Sheaves of Modules

by (5.4) there is an A-module M with ~lu ~ M. Our section s E F(X,~)

restricts to give an elements EM. On the other hand, our section/ E F(X,!f)
restricts to give a section of !flu, which in turn gives rise to an element
g = t/J(f) EA. Clearly X 1 n U = D(g). Now sixr is zero, so g"s = 0 in M
for some n > 0, just as in the proof of (5.3). Using the isomorphism
id x t/1°":~ ® !E"Iu ~ ~lu.
we conclude that f"s E F(U,~ ® !£") is zero. This statement is intrinsic
(i.e., independent of t/1). So now we do this for each open set of the covering,
pick one n large enough to work for all the sets of the covering, and we
find f"s = 0 on X.
To prove (b), we proceed as in the proof of(5.3), keeping track of the twist
due to !f as above. The hypothesis Vi n Vi quasi-compact is used to be
able to apply part (a) there.

Remark 5.14.1. The hypotheses on X made in the statements (a) and (b)
above are satisfied either if X is noetherian (in which case every open set is
quasi-compact) or if X is quasi-compact and separated (in which case the
intersection of two open affine subsets is again affine, hence quasi-compact).

Proposition 5.15. Let S be_ a graded ring, which is finitely generated by S 1 as

an S 0 -algebra. Let X = Proj S, and let~ be a quasi-coherent sheaf on X.
Then there is a natural isomorphism {J: r *(~)- --+ ~-
PROOF. First let us define the morphism {J for any (!) x-module ~- Let f E S 1 .
Since r *(~)- is quasi-coherent in any case, to define {J, it is enough to give
the image of a section of r *(~)- over D +(f) (see Ex. 5.3). Such a section is
represented by a fraction m/fd, where mE r(X,~(d) ), for some d ~ 0. We
can think of f-d as a section of (!)x(- d), defined over D +(f). Taking their
tensor product, we obtain m ® f-d as a section of ~ over D +(f). This
defines {J.
Now let~ be quasi-coherent. To show that {J is an isomorphism we have
to identify the module r *(~)<n with the sections of~ over D +(f). We apply
(5.14), considering f as a global section of the invertible sheaf !f = (!)(1).
Since we have assumed that Sis finitely generated by S 1 as an S0 -algebra, we
can find finitely many elements / 0 , . . . ,fr E S 1 such that X is covered by the
open affine subsets D+(JJ The intersections D+(JJ n D+(fJ) are also affine,
and !Eiv+ <JJ is free for each i, so the hypotheses of (5.14) are satisfied. The
conclusion of(5.14) tells us that ~(D+(f)) ~ r*(~)<n• which is just what
we wanted.

Corollary 5.16. Let A be a ring.

(a) If Y is a closed subscheme of P~, then there is a homogeneous ideal
I c;: S = A[ x 0 , . .. ,x,] such that Y is the closed subscheme determined by I
(Ex. 3.12).

119
II Schemes

(b) A scheme Y over Spec A is projective if and only if it is isomorphic to

Proj S for some graded ring S, where S 0 = A, and S is finitely generated by
S 1 as an S 0 -algebra.
PROOF.
(a) Let f y be the ideal sheaf of Y on X = PA. Now f y is a subsheaf of
lD x; the twisting functor is exact; the global section functor r is left exact;
hence r*(fy) is a submodule of r*(lDx). But by (5.13), r*(lDx) = S. Hence
r *(f y) is a homogeneous ideal of S, which we will call I. Now I determines
a closed subscheme of X (Ex. 3.12), whose sheaf of ideals will be I. Since f y
is quasi-coherent by (5.9), we have f y ~ l by (5.15), and hence Y is the sub-
scheme determined by I. In fact, r *(f y) is the largest ideal in S defining Y
(Ex. 5.10).
(b) Recall that by definition Y is projective over Spec A if it is isomorphic
to a closed subscheme ofPA for some r (§4). By part (a), any such Y is isomor-
phicto Proj Sji, and we can take I to be contained inS+ = ffid >O Sd (Ex. 3.12),
so that (S//) 0 = A. Conversely, any such graded ring S is a quotient of a
polynomial ring, so Proj S is projective.

Definition. For any scheme Y, we define the twisting sheaf lD(1) on P~ to be

g*(lD(1) ), where g:P~ -+ Pz is the natural map (recall that P~ was defined
as Pz X z Y).
Note that if Y = Spec A, this is the same as the lD(1) already defined on
PA = Proj A[ x 0 , ..• ,x,], by (5.12c).

Definition. If X is any scheme over Y, an invertible sheaf !l' on X is very ample

relative to Y, if there is an immersion i:X -+ P~ for some r, such that
i*(lD(1)) ~ !l'. We say that a morphism i:X-+ Z is an immersion if it
gives an isomorphism of X with an open subscheme of a closed subscheme
of Z. (This definition of very ample differs slightly from the one in
Grothendieck [EGA II, 4.4.2].)

Remark 5.16.1. Let Y be a noetherian scheme. Then a scheme X over Y is

projective if and only if it is proper, and there exists a very ample sheaf on X
relative to Y. Indeed, if X is projective over Y, then X is proper by (4.9). On
the other hand, there is a closed immersion i:X -+ P~ for some r, so i*lD(1)
is a very ample invertible sheaf on X. Conversely, if X is proper over Y, and
!l' is a very ample invertible sheaf, then fil ~ i*(lD(1)) for some immersion
i:X -+ P~. But by (Ex. 4.4) the image of X is closed, so in facti is a closed
immersion, so X is projective over Y.
Note however that there may be several nonisomorphic very ample
sheaves on a projective scheme X over Y. The sheaf !l' depends on the
embedding of X into P~ (Ex. 5.12). If Y = Spec A, and if X = Proj S, where S
is a graded ring as in (5.16b), then the sheaf lD(1) on X defined earlier is a very
ample sheaf on X. However, there may be nonisomorphic graded rings
having the same Proj and the same very ample sheaf lD(1) (Ex. 2.14).

120
 5 Sheaves of Modules

We end this section with some special results about sheaves on a projective
scheme over a noetherian ring.

Definition. Let X be a scheme, and let ff' be a sheaf of C9x-modules. We say

that ff' is generated by global sections if there is a family of global sections
{sJiEI• s; E r(X,ff'), such that for each x EX, the images of s; in the stalk
ff'x generate that stalk as an {9x-module.
Note that ff' is generated by global sections if and only if ff' can be
written as a quotient of a free sheaf. Indeed, the generating sections
{s;};EI define a surjective morphism of sheaves EBiEI C9x---> ff', and
conversely.

Example 5.16.2. Any quasi-coherent sheaf on an affine scheme is generated

by global sections. Indeed, if ff' = M on Spec A, any set of generators for M
as an A-module will do.

Example 5.16.3. Let X = Proj S, where Sis a graded ring which is generated
by S1 as an S0 -algebra. Then the elements of S1 give global sections of (9 x(l)
which generate it.

Theorem 5.17 (Serre). Let X be a projective scheme over a noetherian ring A,

let (9(1) be a very ample invertible sheaf on X, and let ff' be a coherent
C9x-module. Then there is an integer n0 such that for all n ~ n0 , the sheaf
ff'(n) can be generated by a .finite number of global sections.
PROOF. Let i:X---> P~ be a closed immersion of X into a projective space over
A, such that i*(CD(l)) = C9x(l). Then i*ff' is coherent on P~ (Ex. 5.5), and
i*(JF(n)) = (i*JF)(n) (5.12) or (Ex. 5.1d), and JF(n) is generated by global
sections if and only if i*(ff'(n)) is (in fact, their global sections are the same),
so we reduce to the case X = P~ = Proj A[ x 0 , . . . ,xrJ.
Now cover X with the open sets D +(x;), i = 0, ... ,r. Since ff' is coherent,
for each i there is a finitely generated module M; over B; =A[ x 0 /x;, ... ,xn/x;]
such that ff'lv+ <xil ~ M;. For each i, take a finite number of elements
sii EM; which generate this module. By (5.14) there is an integer n such that
x7s;j extends to a global section t;j of ff'(n). As usual, we take one n to work
for all i,j. Now ff'(n) corresponds to a B;-module Mi on D +(x;), and the map
x7:ff'---> ff'(n) induces an isomorphism of M; to Mi. So the sections x7s;j
generate M;, and hence the global sections tij E F(X,ff'(n)) generate the sheaf
ff'(n) everywhere.

Corollary 5.18. Let X be projective over a noetherian ring A. Then any

coherent sheaf ff' on X can be written as a quotient of a sheaf <ff, where <ff
is a finite direct sum of twisted structure sheaves CD(n;) for various integers
n;.

121
II Schemes

PROOF. Let .?(n) be generated by a finite number of global sections. Then

we have a surjection EBf= 1 (!Jx---+ .?(n)---+ 0. Tensoring with (!Jx(-n) we
obtain a surjection EBf= 1 (!Jx( -n)---+.?---+ 0 as required.

Theorem 5.19. Let k be afield, let A be a finitely generated k-algebra, let X be

a projective scheme over A, and let .? be a coherent {!}x-module. Then
r(X,.?) is a finitely generated A-module. In particular, if A = k, r(X,.?)
is a finite-dimensional k-vector space.
PROOF. First we write X = Proj S, where Sis a graded ring with S0 = A which
is finitely generated by S 1 as an S 0 -algebra (5.16b). Let M be the graded S-
module r *(.?). Then by (5.15) we have M ~ .?. On the other hand, by (5.17),
for n sufficiently large, .?(n) is generated by a finite number of global sections
in r(X,.?(n) ). Let M' be the submodule of M generated by these sections.
Then M' is a finitely generated S-module. Furthermore, the inclusion
M' ~ M induces an inclusion of sheaves M' ~ M = .?. Twisting by n we
have an inclusion M'(n) ~ .?(n) which is actually an isomorphism, because
.?(n) is generated by global sections in M'. Twisting by - n we find that
M' ~ .?. Thus .? is the sheaf associated to a finitely generated S-module,
and so we have reduced to showing that if M is a finitely generated S-module,
then r(X,M) is a finitely generated A-module.
Now by (1, 7.4), there is a finite filtration
0 = M0 ~ M1 ~ ... ~ Mr = M
of M by graded submodules, where for each i, Mi/Mi- 1 ~ (S/p;}(n;) for some
homogeneous prime ideal Pi ~ S, and some integer ni. This filtration gives
a filtration of M, and the short exact sequences
o ---+ Mi- 1 ---+ Mi ---+ Mi I Mi- 1 ---+ o
give rise to left-exact sequences
o---+ r(X,Mi- 1 )---+ r(X,Mi)---+ r(X,Mi/Mi- 1 ).
Thus to show that r(X,M) is finitely generated over A, it will be sufficient to
show that r(X,(S/p)-(n)) is finitely generated, for each p and n. Thus we have
reduced to the following special case: Let S be a graded integral domain,
finitely generated by S 1 as an S0 -algebra, where S 0 = A is a finitely generated
integral domain over k. Then r(X,(!Jx(n)) is a finitely generated A-module,
for any n E Z.
Let x 0 , . .. ,xr E S 1 be a set of generators of S 1 as an A-module. Since S
is an integral domain, multiplication by x 0 gives an injection S(n) ---+ S(n + 1)
for any n. Hence there is an injection r(X,(!Jx(n)) ---+ r(X,(!Jx(n + 1)) for
any n. Thus it is sufficient to prove r(X,(!Jx(n)) finitely generated for all
sufficiently large n, say n ~ 0.
LetS'= ffin;.o r(X,(!Jx(n) ). Then S' is a ring, containing S, and contained
in the intersection nsx, of the localizations of Sat the elements Xo, . .. ,Xr.

122
 5 Sheaves of Modules

(Use the same argument as in the proof of (5.13).) We will show that S' is
integral over S.
Let s' E S' be homogeneous of degree d ~ 0. Since s' E Sx, for each i, we
can find an integer n such that x7s' E S. Choose one n that works for all i.
Since the X; generate s 1' the monomials in the X; of degree m generate sm for
any m. So by taking a larger n, we may assume that ys' E S for ally E S". In fact,
since s' has positive degree, we can say that for any yES ;,n = EBe;,n Se,
ys' E S;,n· Now it follows inductively, for any q ~ 1 that y · (s')q E S;,n for
any y E S;,"' Take for example y = x~. Then for every q ~ 1 we have
(s')q E (1/x~)S. This is a finitely generated sub-S-module of the quotient field
of S'. It follows by a well-known criterion for integral dependence (Atiyah-
Macdonald [1, p. 59]), that s' is integral overS. Thus S' is contained in the
integral closure of Sin its quotient field.
To complete the proof, we apply the theorem of finiteness of integral
closure (I, 3.9A). Since Sis a finitely generated k-algebra, S' will be a finitely
generated S-modqle. It follows that for every n, S~ is a finitely generated
S0 -module, which is what we wanted to prove. In fact, our proof shows that
S~ = S" for all sufficiently large n (Ex. 5.9) and (Ex. 5.14).

Remark 5.19.1. This proof is a generalization of the proof of (1, 3.4a). We

will give another proof of this theorem later, using cohomology (III, 5.2.1).

Remark 5.19.2. The hypothesis "A is a finitely generated k-algebra" is used

only to be able to apply (1, 3.9A). Thus it would be sufficient to assume only
that A is a "Nagata ring" in the sense of Matsumura [2, p. 231 ]-see also
[Joe. cit., Th. 72, p. 240].

Corollary 5.20. Let f: X ---+ Y be a projective morphism of schemes of finite

type over afield k. Let:#' be a coherent sheaf on X. Then f*:F is coherent
on Y.
PROOF. The question is local on Y, so we may assume Y = Spec A, where A
is a finitely generated k-algebra. Then in any case, f*:F is quasi-coherent
(5.8c), so f*:F = T( Y,f*:F)- = T(X,:Fr. But T(X,ff) is a finitely generated
A-module by the theorem, so f*:F is coherent. See (III, 8.8) for another proof
and generalization.

EXERCISES

5.1. Let (X,((x) be a ringed space, and let tC be a locally free 0x-module of finite rank.
We define the dual of tff, denoted i, to be the sheaf Jffom(l)x(tff,C9x).
(a) Show that (ir ~ tff.
(b) For any C9x-module :#', Jffomf'x(tff,:J') ~ i ®(l)x :F.
(c) For any crx-modules .'#','§, Hom~x(tff ® :#','§) ~ Hom(l)x(:J',Jffom(')x(tff,'§) ).

123
II Schemes

(d) (Projection Formula). If f:(X/!Jx)-+ (Y,@y) is a morphism of ringed spaces, if

:F is an @x-module, and if<! is a locally free @y-module of finite rank, then there
is a natural isomorphism f*(:F ®<'!x f*<!) ~ f*(:F) ®<'!y <!.
5.2. Let R be a discrete valuation ring with quotient field K, and let X= Spec R.
(a) To give an @x-module is equivalent to giving an R-module M, a K-vector
space L, and a homomorphism p:M ®R K ..... L.
(b) That @x-module is quasi-coherent if and only if pis an isomorphism.
5.3. Let X = Spec A be an affine scheme. Show that the functors -and rare adjoint,
in the following sense: for any A-module M, and for any sheaf of @x-modules :F,
there is a natural isomorphism

5.4. Show that a sheaf of @x-modules :F on a scheme X is quasi-coherent if and only

if every point of X has a neighborhood U, such that :Flu is isomorphic to a
cokernel of a morphism of free sheaves on U. If X is noetherian, then :F is co-
herent if and only if it is locally a co kernel of a morphism of free sheaves of finite
rank. (These properties were originally the definition of quasi-coherent and
coherent sheaves.)
5.5. Let f: X ..... Y be a morphism of schemes.
(a) Show by example that if :F is coherent on X, then f*:F need not be coherent
on Y, even if X and Y are varieties over a field k.
(b) Show that a closed immersion is a finite morphism (§3).
(c) Iff is a finite morphism of noetherian schemes, and if :F is coherent on X,
then f*:F is coherent on Y.
5.6. Support. Recall the notions of support of a section of a sheaf, support of a sheaf,
and subsheafwith supports from (Ex. 1.14) and (Ex. 1.20).
(a) Let A be a ring, let M be an A-module, let X = Spec A, and let :F = M.
For any mE M = r(X,:F), show that Supp m = V(Ann m), where Ann m is
the annihilator ofm = {a E Alam = 0}.
(b) Now suppose that A is noetherian, and M finitely generated. Show that
Supp :F = V(Ann M).
(c) The support of a coherent sheaf on a noetherian scheme is closed.
(d) For any ideal a <;; A, we define a submodule T0 (M) of M by T.,(M) =
{mE Mla"m = 0 for some n > 0}. Assume that A is noetherian, and Many
A-module. Show that r.,(Mr ~ £'~(/F), where Z = V(a) and :F = M.
[Hint: Use (Ex. 1.20) and (5.8) to show a priori that £'~(/F) is quasi-coherent.
Then show that Fa(M) ~ Tz(ff).]
(e) Let X be a noetherian scheme, and let Z be a closed subset. If :F is a quasi-
coherent (respectively, coherent) @x-module, then £'~(/F) is also quasi-
coherent (respectively, coherent).
5.7. Let X be a noetherian scheme, and let :F be a coherent sheaf.
(a) If the stalk ffx is a free ((!x-module for some point x EX, then there is a neigh-
borhood U of x such that :Flu is free.
(b) :F is locally free if and only if its stalks .'f'x are free @x-modules for all x EX.
(c) :F is invertible (i.e., locally free of rank 1) if and only if there is a coherent sheaf
'!J such that :F ® '!J ~ ((!x· (This justifies the terminology invertible: it means

124
 5 Sheaves of Modules

that :F is an invertible element of the monoid of coherent sheaves under the

operation ®.)
5.8. Again let X be a noetherian scheme, and :F a coherent sheaf on X. We will
consider the function
rp(x) = dimk(x) ffx ®mx k(x),
where k(x) = (l)xfmx is the residue field at the point x. Use Nakayama's lemma
to prove the following results.
(a) The function rp is upper semi-continuous, i.e., for any nEZ, the set {xEXIrp(x) ;;> n}
is closed.
(b) If :F is locally free, and X is connected, then rp is a constant function.
(c) Conversely, if X is reduced, and rp is constant, then :F is locally free.
5.9. Let S be a graded ring, generated by S 1 as an S0 -algebra, let M be a graded S-
module, and let X = Proj S.
(a) Show that there is a natural homomorphism cx:M---> r *(M).
(b) Assume now that S 0 = A is a finitely generated k-algebra for some field k,
that S 1 is a finitely generated A-module, and that M is a finitely generated
S-module. Show that the map ex is an isomorphism in all large enough
degrees, i.e., there is a d0 E Z such that for all d ;;> d0 , cxd:Md---> T(X,M(d))
is an isomorphism. [Hint: Use the methods of the proof of (5.19).]
(c) With the same hypotheses, we define an equivalence relation ~ on graded
S-modules by saying M ~ M' if there is an integer d such that M :3d ~ M':3d·
Here M :3d = ffin:3d Mn. We will say that a graded S-module M is quasi-
finitely generated if it is equivalent to a finitely generated module. Now show
that the functors - and r * induce an equivalence of categories between the
category of quasi-finitely generated graded S-modules modulo the equivalence
relation ~, and the category of coherent (l)x-modules.
5.10. Let A be a ring, letS = A[x 0 , . . . ,x,] and let X = Proj S. We have seen that a
homogeneous ideal I in S defines a closed subscheme of X (Ex. 3.12), and that
conversely every closed subscheme of X arises in this way (5.16).
(a) For any homogeneous ideal I c;; S, we define the saturation 1 of I to be
{s E Slfor each i = 0, ... ,r, there is ann such that x7s E I}. We say that I is
saturated if I = 1. Show that 1 is a homogeneous ideal of S.
(b) Two homogeneous ideals I 1 and I 2 of S define the same closed subscheme of
X if and only if they have the same saturation.
(c) If Y is any closed subscheme of X, then the ideal r *( J\) is saturated. Hence
it is the largest homogeneous ideal defining the subscheme Y.
(d) There is a 1-1 correspondence between saturated ideals of Sand closed sub-
schemes of X.
5.11. Let S and T be two graded rings with S 0 = T 0 = A. We define the Cartesian
product s X A T to be the graded ring EBd:30 sd ®A Td. If X = Proj s and
Y = Proj T, show that Proj(S x A T) ~ X x A Y, and show that the sheaf (1)(1)
on Proj(S x A T) is isomorphic to the sheaf pf((l)x(1)) ® p!((l)y(1)) on X x Y.
The Cartesian product of rings is related to the Segre embedding of projective
spaces (1, Ex. 2.14) in the following way. If x 0 , . . . ,x, is a set of generators for S 1
over A, corresponding to a projective embedding X 4 PA., and if y 0 , . .• ,y, is
a set of generators for T 1 , corresponding to a projective embedding Y 4 P~,
then {xi® yj} is a set of generators for (S x A T) 1, and hence defines a projective

125
II Schemes

embedding Proj(S x A T) c. P~, with N = rs + r + s. This is just the image

of X x Y c:; P' x P' in its Segre embedding.
5.12. (a) Let X be a scheme over a scheme Y, and let 2', A be two very ample invertible
sheaves on X. Show that fi' ®A is also very ample. [Hint: Use a Segre
embedding.]
(b) Let f:X -+ Y and g: Y-+ Z be two morphisms of schemes. Let fi' be a very
ample invertible sheaf on X relative to Y, and let A be a very ample invertible
sheaf on Y relative to Z. Show that fi' ® f*jt is a very ample invertible sheaf
on X relative to Z.
5.13. Let S be a graded ring, generated by S 1 as an S0 -algebra. For any integer d > 0,
let s<d) be the graded ring EBn:.
0 s~d) where s~d) = snd· Let X = Proj S. Show

that Proj s<dl ~ X, and that the sheaf CD(l) on Proj s<dl corresponds via this
isomorphism to CDx(d).
This construction is related to the d-uple embedding (1, Ex. 2.12) in the fol-
lowing way. If x 0 , . . . ,x, is a set of generators for S 1 , corresponding to an em-
bedding X c. PA, then the set of monomials of degree d in the xi is a set of
generators for S\dl = Sd. These define a projective embedding of Proj s<dl which
is none other than the image of X under the d-uple embedding of P~.
5.14. Let A be a ring, and let X be a closed subscheme of P~. We define the homo-
geneous coordinate ring S(X) of X for the given embedding to be A[x 0 , . . . ,x,]/1,
where I is the ideal r *(.I x) constructed in the proof of (5.16). (Of course if A is
a field and X a variety, this coincides with the definition given in (1, §2) !) Recall
that a scheme X is normal if its local rings are integrally closed domains. A closed
subscheme X c:; P~ is projectively normal for the given embedding, if its homo-
geneous coordinate ring S(X) is an integrally closed domain (cf. (1, Ex. 3.18) ).
Now assume that k is an algebraically closed field, and that X is a connected,
normal closed subscheme ofP~. Show that for some d > 0, the d-uple embedding
of X is projectively normal, as follows.
(a) LetS be the homogeneous coordinate ring of X, and letS' = EBn:.o T(X,CDx(n) ).
Show that S is a domain, and that S' is its integral closure. [Hint: First show
that X is integral. Then regard S' as the global sections of the sheaf of rings
!/' = EBn:.o CDx(n) on X, and show that !/' is a sheaf of integrally closed
domains.]
(b) Use (Ex. 5.9) to show that Sd = S~ for all sufficiently large d.
(c) Show that s<dl is integrally closed for sufficiently large d, and hence conclude
that the d-uple embedding of X is projectively normal.
(d) As a corollary of (a), show that a closed subscheme X c:; P~ is projectively
normal if and only if it is normal, and for every n ~ 0 the natural map
T(P',@p,(n))-+ r(X,CDx(n)) is surjective.

5.15. Extension of Coherent Sheaves. We will prove the following theorem in several
steps: Let X be a noetherian scheme, let U be an open subset, and let.? be a
coherent sheaf on U. Then there is a coherent sheaf.?' on X such that .?'lu ~ .?.
(a) On a noetherian affine scheme, every quasi-coherent sheaf is the union of its
coherent subsheaves. We say a sheaf.? is the union of its subsheaves Sf
if for every open set U, the group .?(V) is the union of the subgroups Sf (U).
(b) Let X be an affine noetherian scheme, U an open subset, and.? coherent on
U. Then there exists a coherent sheaf.?' on X with .?'lu ~ .?. [Hint: Let
i: U -+ X be the inclusion map. Show that i*.? is quasi-coherent, then use(a).]

126
 5 Sheaves of Modules

(c) With X,U,:F as in (b), suppose furthermore we are given a quasi-coherent

sheaf'§ on X such that :F ~ '§lu· Show that we can find ff' a coherent sub-
sheaf of'§, with :F'lu ~ :F. [Hint: Use the same method, but replace i*ff
by p- 1 (i*ff), where pis the natural map'§-.. i*('§lul-]
(d) Now let X be any noetherian scheme, U an open subset, :Fa coherent sheaf
on U, and'§ a quasi-coherent sheaf on X such that :F ~ '§lu· Show that there
is a coherent subsheaf :F ~ '§on X with :F'lu ~ :F. Taking'§ = i*:F proves
the result announced at the beginning. [Hint: Cover X with open affines, and
extend over one of them at a time.]
(e) As an extra corollary, show that on a noetherian scheme, any quasi-coherent
sheaf :F is the union of its coherent subsheaves. [Hint: If sis a section of :F
over an open set U, apply (d) to the subsheaf of :Flu generated by s.]

5.16. Tensor Operations on Sheaves. First we recall the definitions of various tensor
operations on a module. Let A be a ring, and let M be an A-module. Let T"(M)
be the tensor product M ® ... ® M of M with itself n times, for n ;, 1. For
n = 0 we put T 0 (M) = A. Then T(M) = EB.~o T"(M) is a (noncommutative)
A-algebra, which we call the tensor algebra of M. We define the symmetric
algebra S(M) = ffi.~o S"(M) of M to be the quotient of T(M) by the two-sided
ideal generated by all expressions x ® y - y ® x, for all x,y EM. Then S(M)
is a commutative A-algebra. Its component S"(M) in degree n is called the nth
symmetric product of M. We denote the image of x ® y in S(M) by xy, for any
x,y EM. As an example, note that if M is a free A-module of rank r, then S(M) ~
A[x 1, ••• ,x,].
We define the exterior algebra 1\(M) = EB.~o /\"(M) of M to be the quo-
tient of T(M) by the two-sided ideal generated by all expressions x ® x for
x E M. Note that this ideal contains all expressions of the form x ® y + y ® x,
so that /\(M) is a skew commutative graded A-algebra. This means that if u E
/\'(M) and v E f\•(M), then u 1\ v = ( -l)"v 1\ u (here we denote by 1\ the
multiplication in this algebra; so the image of x ®yin f\ 2 (M) is denoted by
x 1\ y). The nth component /\"(M) is called the nth exterior power of M.
Now let (X,(!Jx) be a ringed space, and let :F be a sheaf of (!Jx-modules. We
define the tensor algebra, symmetric algebra, and exterior algebra of :F by taking
the sheaves associated to the presheaf, which to each open ·set U assigns the
corresponding tensor operation applied to ff(U) as an (!Jx(U)-module. The
results are (!Jx-algebras, and their components in each degree are (!Jx-modules.
(a) Suppose that :F is locally free of rank n. Then T'(ff), S'(ff), and /\'(ff) are
also locally free, of ranks n', ("~.':! 1 ), and G) respectively.
(b) Again let :F be locally free of rank n. Then the multiplication map/\':F ®
1\" -,:F -.. 1\ ".~ is a perfect pairing for any r, i.t,., it induces an isomorphism
of /\' :F with ( /\"- ':F r ® 1\" :F. As a special case, note if :F has rank 2,
then :F ~ :F- ® 1\1 :?.
(c) Let 0 -.. :F' -.. :F -.. :F" -.. 0 be an exact sequence of locally free sheaves.
Then for any r there is a finite filtration of S'(ff),

S'(ff) = F0 2 F 1 2 ... 2 F' 2 pr+ 1 = 0

with quotients

for each p.

127
II Schemes

(d) Same statement as (c), with exterior powers instead of symmetric powers. In
particular, if ff',ff,ff" have ranks n',n,n" respectively, there is an isomorphism
1\"ff ~ 1\"'ff' ® 1\""ff".
(e) Let f:X --+ Y be a morphism of ringed spaces, and let ff be an (l)y-module.
Then f* commutes with all the tensor operations on ff, i.e., f*(S"(ff)) =
S"(f* ff) etc.
5.17. Affine M orphisms. A morphism f: X --+ Y of schemes is affine ifthere is an open
affine cover {V;} of Y such that f- 1(V;) is affine for each i.
(a) Show that f: X --+ Y is an affine morphism if and only if for every open affine
V <;::; Y,f- 1 (V) is affine. [Hint: Reduce to the case Yaffine, and use (Ex. 2.17).]
(b) An affine morphism is quasi-compact and separated. Any finite morphism is
affine.
(c) Let Y be a scheme, and let d be a quasi-coherent sheaf of (l)y-algebras (i.e., a
sheaf of rings which is at the same time a quasi-coherent sheaf of (l)y-modules).
Show that there is a unique scheme X, and a morphism f:X --+ Y, such that
for every open affine V <;::; Y, f- 1(V) ~ Spec d(V), and for every inclusion
U 4 V of open affines of Y, the morphism f- 1 (U) 4 f- 1 (V) corresponds to
the restriction homomorphism d(V)--+ d(U). The scheme X is called
Spec d. [Hint: Construct X by glueing together the schemes Spec d(V),
for V open affine in Y.J
(d) If d is a quasi-coherent (l)y-algebra, then f:X = Spec d--+ Y is an affine
morphism, and d ~ j*(l)x· Conversely, if f:X--+ Y is an affine morphism,
then d = j*(l)x is a quasi-coherent sheaf of (l)y-algebras, and X ~ Spec d.
(e) Letf:X--+ Ybe an affine morphism, and let d = j*(l)x· Show thatf* induces
an equivalence of categories from the category of quasi-coherent (l)x-modules
to the category of quasi-coherent d-modules (i.e., quasi-coherent (l)y-modules
having a structure of d-module). [Hint: For any quasi-coherent d-module
.A, construct a quasi-coherent (l)x-module .ii, and show that the functors f*
and- are inverse to each other.
5.18. Vector Bundles. Let Y be a scheme. A (geometric) vector bundle of rank n over
Y is a scheme X and a morphism f:X--+ Y, together with additional data con-
sisting of an open covering {U;} of Y, and isomorphisms lj;i:f- 1 (U;)--+ Au,,
such that for any i,j, and for any open affine subset V = Spec A <;::; Ui n Ui,
the automorphism lj; = lj;ioi/Ji 1 of A~= SpecA[x 1 , . . . ,x"] is given by a
linear automorphism() of A[ x 1 , . . . ,x"], i.e., O(a) = a for any a E A, and O(x;) =
[Link] for suitable aii EA.
An isomorphism g:(X,f,{Ui},{lj;i})--+ (X',f',{U;},{tf;;}) of one vector bundle
of rank n to another one is an isomorphism g:X--+ X' of the underlying schemes,
such that f = f' o g, and such that X,f, together with the covering of Y con-
sisting of all the Ui and u;, and the isomorphisms lj;i and tf;; o g, is also a vector
bundle structure on X.
(a) Let@' be a locally free sheaf of rank non a scheme Y. Let S(t&') be the symmetric
algebra on t&', and let X = Spec S(t&'), with projection morphism f:X--+ Y.
For each open affine subset U <;::; Y for which t&'lu is free, choose a basis of@',
and let lj;:f- 1 (U)--+ Au be the isomorphism resulting from the identification
of S(t&'(U)) with (I)(U)[x 1, . . . ,x"]. Then (X,f,{U},{Ij;}) is a vector bundle of
rank n over Y, which (up to isomorphism) does not depend on the bases of
@' u chosen. We call it the geometric vector bundle associated to@', and denote
it by V(t&').

128
 6 Divisors

(b) For any morphism f: X --+ Y, a section off over an open set U s Y is a mor-
phism s: U--+ X such that f o s = idu. It is clear how to restrict sections to
smaller open sets, or how to glue them together, so we see that the presheaf
U H {set of sections off over U} is a sheaf of sets on Y, which we denote by
Y'(X /Y). Show that iff: X --+ Y is a vector bundle of rank n, then the sheaf
of sections Y'(X/Y) has a natural structure of @y-module, which makes it a
locally free @y-module of rank n. [Hint: It is enough to define the module
structure locally, so we can assume Y = Spec A is affine, and X = A~. Then a
sections: Y--+ X comes from an A-algebra homomorphism !:I:A[x 1 , •.. ,x.]--+
<
A, which in turn determines an ordered n-tuple !:l(x 1 ), . . . , !:l(x.)) of elements
of A. Use this correspondence between sections s and ordered n-tuples of
elements of A to define the module structure.]
(c) Again let tff be a locally free sheaf of rank non Y, let X = V(tff), and let Y' =
Y'(X/Y) be the sheaf of sections of X over Y. Show that Y' ~ tff~, as follows.
Given a sections E r(V,tff~) over any open set V, we think of s as an element of
Hom(tff!v,@v). So s determines an @v-algebra homomorphism S(t!lvl--+ @y.
This determines a morphism of spectra V = Spec (Dv--+ Spec S(t!lvl =
f- 1(V), which is a section of X /Y. Show that this construction gives an iso-
morphism of ,g~ to Y'.
(d) Summing up, show that we have established a one-to-one correspondence
between isomorphism classes of locally free sheaves of rank n on Y, and iso-
morphism classes of vector bundles of rank n over Y. Because of this, we
sometimes use the words "locally free sheaf" and "vector bundle" inter-
changeably, if no confusion seems likely to result.

6 Divisors
The notion of divisor forms an important tool for studying the intrinsic
geometry on a variety or scheme. In this section we will introduce divisors,
linear equivalence and the divisor class group. The divisor class group is
an abelian group which is an interesting and subtle invariant of a variety.
In §7 we will see that divisors are also important for studying maps from a
given variety to a projective space.
There are several different ways of defining divisors, depending on the
context. We will begin with Weil divisors, which are easiest to understand
geometrically, but which are only defined on certain noetherian integral
schemes. For more general schemes there is the notion of Cartier divisor
which we treat next. Then we will explain the connection between Weil
divisors, Cartier divisors, and invertible sheaves.

We start with an informal example. Let C be a nonsingular projective

curve in Pf, the projective plane over an algebraically closed field k. For
each line Lin P 2 , we consider L n C, which is a finite set of points on C.
If C is a curve of degree d, and if we count the points with proper multi-
plicity, then L n C will consist of exactly d points (I, Ex. 5.4). We write
L n C = l,n;P;, where P; E Care the points, and n; the multiplicities, and
we call this formal sum a divisor on C. As L varies, we obtain a family of
divisors on C, parametrized by the set of all lines in P 2 , which is the dual

129
II Schemes

projective space (P 2 )*. We call this set of divisors a linear system of divisors
on C. Note that the embedding of C in P 2 can be recovered just from
knowing this linear system: if Pis a point of C, we consider the set of divisors
in the linear system which contain P. They correspond to the lines L E (P 2 )*
passing through P, and this set of lines determines P uniquely as a point
of P 2 . This connection between linear systems and embeddings in pro-
jective space will be studied in detail in §7.
This example should already serve to illustrate the importance of divi-
sors. To see the relation among the different divisors in the linear system,
let Land L' be two lines in P 2 , and let D = L n C and D' = L' n C be the
corresponding divisors. If L and L' are defined by linear homogeneous
equations f = 0 and f' = 0 in P 2 , then f/f' gives a rational function on
P 2 , which restricts to a rational function g on C. Now by construction, g
has zeros at the points of D, and poles at the points of D', counted with
multiplicities, in a sense which will be made precise below. We say that
D and D' are linearly equivalent, and the existence of such a rational func-
tion can be taken as an intrinsic definition of the linear equivalence. We
will make these concepts more precise in our formal discussion, starting now.

Wei! Divisors
Definition. We say a scheme X is regular in codimension one (or sometimes
nonsingular in codimension one) if every local ring (iJ x of X of dimension
one is regular.

The most important examples of such schemes are nonsingular varieties

over a field (1, §5) and noetherian normal schemes. On a nonsingular
variety the local ring of every closed point is regular (1, 5.1), hence all the
local rings are regular, since they are localizations of the local rings of
closed points. On a noetherian normal scheme, any local ring of dimen-
sion one is an integrally closed domain, hence is regular (1, 6.2A).
In this section we will consider schemes satisfying the following condition:
(*)X is a noetherian integral separated scheme which is regular in
codimension one.

Definition. Let X satisfy (*). A prime divisor on X is a closed integral sub-

scheme Y of codimension one. A Weil divisor is an element of the free
abelian group Div X generated by the prime divisors. We write a divisor
as D = L\ }i, where the 1i are prime divisors, the n; are integers, and
only finitely many n; are different from zero. If all the n; ~ 0, we say
that D is effective.
If Y is a prime divisor on X, let lJ E Y be its generic point. Then the
local ring (iJ~.x is a discrete valuation ring with quotient field K, the
function field of X. We call the corresponding discrete valuation vy
the valuation of Y. Note that since X is separated, Y is uniquely deter-

130
 6 Divisors

mined by its valuation (Ex. 4.5). Now let f E K* be any nonzero rational
function on X. Then vy(f} is an integer. If it is positive, we say f has
a zero along Y, of that order; if it is negative, we say f has a pole along Y,
of order - Vy(f).

Lemma 6.1. Let X satisfy (*), and let f E K* be a nonzero function on X.

Then vy(f) = 0 for all except finitely many prime divisors Y.
PROOF. Let U = Spec A be an open affine subset of X on which f is regular.
Then Z = X - U is a proper closed subset of X. Since X is noetherian,
Z can contain at most finitely many prime divisors of X; all the others
must meet U. Thus it will be sufficient to show that there are only finitely
many prime divisors Y of U for which vy(f) # 0. Since f is regular on U,
we have vy(f} :;?: 0 in any case. And vy(f) > 0 if and only if Y is contained
in the closed subset of U defined by the ideal Af in A. Since f # 0, this is
a proper closed subset, hence contains only finitely many closed irreducible
subsets of codimension one of U.

Definition. Let X satisfy (*) and let f E K*. We define the divisor off,
denoted (f), by
(f) = Ivy(f) . Y,
where the sum is taken over all prime divisors of X. By the lemma, this
is a finite sum, hence it is a divisor. Any divisor which is equal to the
divisor of a function is called a principal divisor.

Note that if f,g E K*, then (fjg) = (f) - (g) because of the properties
of valuations. Therefore sending a function f to its divisor (f) gives a
homomorphism of the multiplicative group K* to the additive group
Div X, and the image, which consists of the principal divisors, is a sub-
group of Div X.

Definition. Let X satisfy(*). Two divisors D and D' are said to be linearly
equivalent, written D ,...., D', if D - D' is a principal divisor. The group
Div X of all divisors divided by the subgroup of principal divisors is
called the divisor class group of X, and is denoted by Cl X.

The divisor class group of a scheme is a very interesting invariant. In

general it is not easy to calculate. However, in the following propositions
and examples we will calculate a number of special cases to give some
idea of what it is like.

Proposition 6.2. Let A be a noetherian domain. Then A is a unique factor-

ization domain if and only if X = Spec A is normal and Cl X = 0.

131
II Schemes

PROOF. (See also Bourbaki [1, Ch. 7, §3]). It is well-known that a UFD
is integrally closed, so X will be normal. On the other hand, A is a UFD if
and only if every prime ideal of height 1 is principal (I, 1.12A). So what we
must show is that if A is an integrally closed domain, then every prime
ideal of height 1 is principal if and only if Cl(Spec A) = 0.
One way is easy: if every prime ideal of height 1 is principal, consider a
prime divisor Y <;; X = Spec A. Y corresponds to a prime ideal p of
height 1. If p is generated by an element f E A, then clearly the divisor
off is 1 · Y. Thus every prime divisor is principal, so Cl X = 0.
For the converse, suppose Cl X = 0. Let p be a prime ideal of height 1,
and let Y be the corresponding prime divisor. Then there is an f E K, the
quotient field of A, with (f) = Y. We will show that in fact f E A and
f generates p. Since vy(f) = 1, we have f E AP, and f generates pAP. If
p' <;; A is any other prime ideal of height 1, then p' corresponds to a prime
divisor Y' of X, and Vy·(f) = 0, so f E Ap'· Now the algebraic result (6.3A)
below implies that f EA. In fact, f E A n pAv = p. Now to show that f
generates p, let g be any other element of p. Then Vy(g) ? 1 and Vy·(g) ? 0
for all Y' i= Y. Hence vy.(g/f) ? 0 for all prime divisors Y' (including Y).
Thus gjf E Ap' for all p' ofheight 1, so by(6.3A) again,g/f EA. In other words,
g E Af, which shows that p is a principal ideal, generated by f

Proposition 6.3A. Let A be an integrally closed noetherian domain. Then

A= n Av
ht p =1
where the intersection is taken over all prime ideals of height 1.
PROOF. Matsumura [2, Th. 38, p. 124].

Example 6.3.1. If X is affine n-space AJ: over a field k, then Cl X = 0.

Indeed, X = Spec k[x 1 , ... ,xn], and the polynomial ring is a UFD.

Example 6.3.2. If A is a Dedekind domain, then Cl(Spec A) is just the ideal

class group of A, as defined in algebraic number theory. Thus (6.2) generalizes
the fact that A is a UFD if and only if its ideal class group is 0.

Proposition 6.4. Let X be the projective space PJ: over a field k. For any
divisor D = In;¥;, define the degree of D by deg D = In; deg Y;, where
deg Y; is the degree of the hypersurface ¥;. Let H be the hyperplane x 0 =
0. Then:
(a) if D is any divisor of degree d, then D ~ dH;
(b) for any f E K*, deg(f) = 0;
(c) the degree function gives an isomorphism deg:Cl X ---+ Z.
PROOF. Let S = k[ x 0 , .•. ,xn] be the homogeneous coordinate ring of X.
If g is a homogeneous element of degree d, we can factor it into irreducible

132
 6 Divisors

polynomials g = g1' · · · g~r. Then g; defines a hypersurface Y; of degree

d; = deg g;, and we can define the divisor of g to be (g) = In; Y;. Then
deg(g) = d. Now a rational function f on X is a quotient gjh of homo-
geneous polynomials of the same degree. Clearly (f) = (g) - (h), so we
see that deg(f) = 0, which proves (b).
If D is any divisor of degree d, we can write it as a difference D 1 - D 2
of effective divisors of degrees d 1 ,d2 with d 1 - d2 = d. Let D 1 = (g 1) and
D 2 = (g 2 ). This is possible, because an irreducible hypersurface in pn
corresponds to a homogeneous prime ideal of height 1 in S, which is prin-
cipal. Taking power products we can get any effective divisor as (g) for
some homogeneous g. Now D - dH = (f) where f = gtfx~g 2 is a ra-
tional function on X. This proves (a). Statement (c) follows from (a), (b),
and the fact that deg H = 1.

Proposition 6.5. Let X satisfy(*), let Z be a proper closed subset of X, and

let U = X - Z. Then:
(a) there is a surjective homomorphism Cl X -> Cl U defined by D =
In; Y; f---+ In;( Y; n U), where we ignore those Y; n U which are empty;
(b) if codim(Z,X) ;?; 2, then Cl X -> Cl U is an isomorphism;
(c) if Z is an irreducible subset of codimension 1, then there is an exact
sequence
Z -> Cl X -> Cl U -> 0,
where the first map is defined by 1 f---+ 1 · Z.
PROOF.
(a) If Y is a prime divisor on X, then Y n U is either empty or a prime
divisor on U. Iff E K*, and (f) = In;Y;, then considering f as a rational
function on U, we have (f)u = In;(Y; n U), so indeed we have a homo-
morphism Cl X -> Cl U. It is surjective because every prime divisor of U
is the restriction of its closure in X.
(b) The groups Div X and Cl X depend only on subsets of codimension
1, so removing a closed subset Z of codimension ;?; 2 doesn't change anyr l~ing.
(c) The kernel of Cl X -> Cl U consists of divisors whose support is
contained in Z. If Z is irreducible, the kernel is just the subgroup of Cl X
generated by 1 · Z.

Example 6.5.1. Let Y be an irreducible curve of degree d in Pl. Then

Cl(P 2 - Y) = Z/dZ. This follows immediately from (6.4) and (6.5).

Example 6.5.2. Let k be a field, let A = k[x,y,z]/(xy - z 2 ), and let X =

Spec A. Then X is an affine quadric cone in A~. We will show that Cl X =
Zj2Z, and that it is generated by a ruling of the cone, say Y:y = z = 0
(Fig. 8).
First note that Y is a prime divisor, so by (6.5) we have an exact sequence
Z -> Cl X -> Cl(X - Y) -> 0,

133
II Schemes

Figure 8. A ruling on the quadric cone.

where the first map sends 1~---> 1 · Y. Now Y can be cut out set-theoretically
by the function y. In fact, the divisor of y is 2 · Y, because y = 0 =z 2 = 0,
and z generates the maximal ideal of the local ring at the generic point of Y.
Hence X - Y = Spec AY. NowAY = k[x,y,y-l,z]/(xy - z2 ). In this ring
x = y- 1 z 2 , so we can eliminate x, and find AY ~ k[y,y- 1 ,z]. This is a UFD,
so by (6.2), Cl(X - Y) = 0.
Thus we see that Cl X is generated by Y, and that 2 · Y = 0. It remains
to show that Y itself is not a principal divisor. Since A is integrally closed
(Ex. 6.4), it is equivalent to show that the prime ideal of Y, namely p =
(y,z), is not principal (cf. proof of (6.2) ). Let m = (x,y,z), and note that
m/m 2 is a 3-dimensional vector space over k generated by x,y;z, the images
of x,y,z. Now p ~ m, and the image of pin m/m 2 contains y and z. Hence
p cannot be a principal ideal.

Proposition 6.6. Let X satisfy (*). Then X x A 1 (=X x specz Spec Z[t])
also satisfies (*), and Cl X ~ Cl(X x A 1 ).
PROOF. Clearly X x A 1 is noetherian, integral, and separated. To see that
it is regular in codimension one, we note that there are two kinds of points
of codimension one on X x A 1 . Type 1 is a point x whose image in X
is a pointy of codimension one. In this case xis the generic point of n- 1(y),
where n:X x A 1 --+ X is the projection. Its local ring is (!)x ~ (!)y[t]my'
which is clearly a discrete valuation ring, since (!) Y is. The corresponding
prime divisor {x}- isjustn- 1 ({y}-).
Type 2 is a point x E X x A 1 of codimension one, whose image in X
is the generic point of X. In this case (!)xis a localization of K[t] at some
maximal ideal, where K is the function field of X. It is a discrete valuation
ring because K[t] is a principal ideal domain. Thus X x A 1 also satisfies(*).
We define a map Cl X--+ Cl(X x A 1 ) by D =In)';~---> n*D = Inin- 1(¥;).
Iff E K*, then n*( (f)) is the divisor off considered as an element of K(t),
the function field of X x A 1 . Thus we have a homomorphism n*: Cl X --+
Cl(X X A 1).

134
 6 Divisors

To show n* is injective, suppose D E Div X, and n* D = (f) for some

f EK(t). Since n*D involves only prime divisors of type 1, f must be inK.
For otherwise we could write f = g/h, with g,h E K[t], relatively prime.
If g,h are not both in K, then (f) will involve some prime divisor of type 2
on X x A 1 . Now iff E K, it is clear that D = (f), son* is injective.
To show that n* is surjective, it will be sufficient to show that any prime
divisor of type 2 on X x A 1 is linearly equivalent to a linear combination
of prime divisors of type 1. So let Z s; X x A 1 be a prime divisor of type 2.
Localizing at the generic point of X, we get a prime divisor in Spec K[t],
which corresponds to a prime ideal p s; K[t]. This is principal, so let f
be a generator. Then f E K(t), and the divisor off consists of Z plus perhaps
something purely of type 1. It cannot involve any other prime divisors of
type 2. Thus Z is linearly equivalent to a divisor purely of type 1. This
completes the proof.

Example 6.6.1. Let Q be the nonsingular quadric surface xy = zw in Pi.

We will show that Cl Q ~ Z EB Z. We use the fact that Q is isomorphic
to Pl x k Pl (1, Ex. 2.15). Let p 1 and p 2 be the projections of Q onto the
two factors. Then as in the proof of (6.6) we obtain homomorphisms
Pi ,p~: Cl P 1 - t Cl Q. First we show that Pi and p~ are injective. Let Y =
pt x P 1 . Then Q - Y = A 1 x P 1 , and the composition
Cl P 1 ~ Cl Q - t Cl(A 1 X P 1)
is the isomorphism of (6.6). Hence p~ (and similarly pi) is injective.
Now consider the exact sequence of(6.5) for Y:
z -t Cl Q - t Cl(A 1 X P 1) -t 0.
In this sequence the first map sends 1 to Y. But if we identify Cl P 1 with
Z by letting 1 be the class of a point, then this first map is just Pi, hence is
injective. Since the image of p~ goes isomorphically to Cl(A 1 x P 1 ) as we
have just seen, we conclude that Cl Q ~ Im Pi EB Im p~ = Z EB Z. If D
is any divisor on Q, let (a,b) be the ordered pair of integers in Z EB Z cor-
responding to the class of D under this isomorphism. Then we say D is
of type (a,b) on Q.

Example 6.6.2. Continuing with the quadric surface Q s; P 3 , we will show

that the embedding induces a homomorphism Cl P 3 - t Cl Q, and that the
image of a hyperplane H, which generates Cl P 3 , is the element (1,1) in Cl Q =
Z EB Z. Let Y be any irreducible hypersurface of P 3 which does not con-
tain Q. Then we can assign multiplicities to the irreducible components
of Y n Q so as to obtain a divisor Y · Q on Q. Indeed, on each standard
open set U; of P 3 , Y is defined by a single function f; we can take the value
of this function (restricted to Q) for each valuation of a prime divisor of Q
to define the divisor Y · Q. By linearity we extend this map to define a

135
II Schemes

divisor D · Q on Q, for each divisor D = In;¥; on P 3 , such that no ¥; con-

tains Q. Clearly linearly equivalent divisors restrict to linearly equivalent
divisors. Since any divisor on P 3 is linearly equivalent to one whose prime
divisors don't contain Q by (6.4), we obtain a well-defined homomorphism
Cl P 3 ---+ Cl Q. Now if His the hyperplane w = 0, then H n Q is the divisor
consisting of the two lines x = w = 0 and y = w = 0. One is in each
family (I, Ex. 2.15) so H n Q is of type (1,1) in Cl Q = Z 6::> Z. Note that
the two families of lines correspond to pt x P 1 and P 1 x pt, so they are
of type (1,0) and (0,1).

Example 6.6.3. Carrying this example one step further, let C be the twisted
cubic curve x = t 3 , y = u 3 , z = t 2 u, w = tu 2 which lies on Q. If Y is the
quadric cone yz = w 2 , then Y n Q = C u L where Lis the line y = w = 0.
Since Y ~ 2H on P 3 , Y n Q is a divisor of type (2,2). The line L has type
(1,0), so C is of type (1,2). It follows that there does not exist any surface
Y s; P\ not containing Q, such that Y n Q = C, even set-theoretically!
For in that case the divisor Y n Q would be rC for some integer r > 0.
This is a divisor of type (r,2r) in Cl Q. But if Y is a surface of degree d, then
Y n Q is of type (d,d), which can never equal (r,2r). Thus Y does not exist.

Example 6.6.4. We will see later (V, 4.8) that if X is a nonsingular cubic
surface in P 3 , then Cl X ~ Z 7 .

Divisors on Curves
We will illustrate the notion of the divisor class group further by paying
special attention to the case of divisors on curves. We will define the degree
of a divisor on a curve, and we will show that on a complete nonsingular
curve, the degree is stable under linear equivalence. Further study of
divisors on curves will be found in Chapter IV.
To begin with, we need some preliminary information about curves and
morphisms of curves. Recall our conventions about terminology from the
end of Section 4:

Definition. Let k be an algebraically closed field. A curve over k is an

integral separated scheme X of finite type over k, of dimension one.
If X is proper over k, we say that X is complete. If all the local rings
of X are regular local rings, we say that X is nonsingular.

Proposition 6.7. Let X be a nonsingular curve over k with function field K.

Then the following conditions are equivalent:
(i) X is projective;
(ii) X is complete;
(iii) X ~ t(Cx), where Cx is the abstract nonsingular curve of (1, §6),
and t is the functor from varieties to schemes of (2.6).

136
 6 Divisors

PROOF.
(i) => (ii) follows from (4.9).
(ii) => (iii). If X is complete, then every discrete valuation ring of Kjk
has a unique center on X (Ex. 4.5). Since the local rings of X at the closed
points are all discrete valuation rings, this implies that the closed points
of X are in 1-1 correspondence with the discrete valuation rings of Kjk,
namely the points of Cx. Thus it is clear that X ~ t(Cx).
(iii) => (i) follows from (1, 6.9).

Proposition 6.8. Let X be a complete nonsingular curve over k, let Y be any

curve over k, and let f:X ~ Y be a morphism. Then either (1) f(X) =
a point, or (2) f(X) = Y. In case (2), K(X) is a finite extension field of
K(Y), f is a finite morphism, and Y is also complete.
PROOF. Since X is complete, f(X) must be closed in Y, and proper over
Spec k (Ex. 4.4). On the other hand, f(X) is irreducible. Thus either
(1) f(X) = pt, or (2) f(X) = Y, and in case (2), Y is also complete.
In case (2), f is dominant, so it induces an inclusion K(Y) ~ K(X) of
function fields. Since both fields are finitely generated extension fields of
transcendence degree 1 of k, K(X) must be a finite algebraic extension of
K(Y). To show that f is a finite morphism, let V = Spec B be any open
affine subset of Y. Let A be the integral closure of B in K(X). Then A is a
finite B-module (I, 3.9A), and Spec A is isomorphic to an open subset U of
X (I, 6. 7). Clearly U = f- 1 V, so this shows that f is a finite morphism.

Definition. Iff: X~ Y is a finite morphism of curves, we define the degree

off to be the degree of the field extension [ K(X):K(Y)].

Now we come to the study of divisors on curves. If X is a nonsingular

curve, then X satisfies the condition (*) used above, so we can talk about
divisors on X. A prime divisor is just a closed point, so an arbitrary divisor
can be written D = I,n;P;, where the P; are closed points, and n; E Z. We
define the degree of D to be :Ln;.

Definition. If f: X ~ Y is a finite morphism of nonsingular curves, we

define a homomorphism f*: Div Y ~ Div X as follows. For any point
Q E Y, let t E (!)Q be a local parameter at Q, i.e., tis an element of K(Y)
with vQ(t) = 1, where vQ is the valuation corresponding to the discrete
valuation ring (!)Q· We define f*Q = LJ<PJ;Q vp(t) · P. Since f is a
finite morphism, this is a finite sum, so we get a divisor on X. Note
that f*Q is independent of the choice of the local parameter t. Indeed,
if t' is another local parameter at Q, then t' = ut where u is a unit in
(!)Q· For any point P EX with f(P) = Q, u will be a unit in @p, so vp(t) =
vp(t'). We extend the definition by linearity to all divisors on Y. One
sees easily that f* preserves linear equivalence, so it induces a homo-
morphism f*: Cl Y ~ Cl X.

137
II Schemes

Proposition 6.9. Let f: X --+ Y be a finite morphism of nonsingular curves.

Then for any divisor Don Y we have deg f*D = deg f · deg D.
PROOF. It will be sufficient to show that for any closed point Q E Y we have
deg f*Q = deg f Let V = Spec B be an open affine subset of Y containing
Q. Let A be the integral closure of B in K(X). Then, as in the proof of
(6.8), U =Spec A is the open subset f- 1 v of X. Let mQ be the maximal
ideal of Q in B. We localize both Band A with respect to the multiplicative
system S = B - mQ, and we obtain a ring extension (DQ 4 A', where A'
is a finitely generated {DQ-module. Now A' is torsion-free, and has rank
equal to r = [K(X):K(Y)], so A' is a free {DQ-module of rank r = deg f
If t is a local parameter at Q, it follows that A'/tA' is a k-vector space of
dimension r.
On the other hand, the points Pi of X such that f(P;) = Q are in 1-1
correspondence with the maximal ideals mi of A', and for each i, A;" =
(DP,· Clearly tA' = ni(tA;,, 11 A'), so by the Chinese remainder theor~m,
dimk A'/tA' = L dimk A'/(tA;n, 11 A').
i
But
A'/(tA;", 11 A') ~ A;"jtA;", = (Dpjt@p,,
so the dimensions in the sum above are just equal to vp.(t). But f*Q =
l:vp,(t) · Pi, so we have shown that deg f*Q = deg f as required.

Corollary 6.10. A principal divisor on a complete nonsingular curve X has

degree zero. Consequently the degree function induces a surjective homo-
morphism deg: Cl X --+ Z.
PROOF. Let f E K(X)*. Iff E k, then (f) = 0, so there is nothing to prove.
Iff ¢= k, then the inclusion of fields k(f) <;; K(X) induces a finite morphism
cp:X --+ P 1 . It is a morphism by (1, 6.12), and it is finite by (6.8). Now (f) =
cp*( {0} - {oo }). Since {0} - {oo} is a divisor of degree 0 on P\ we conclude
that (f) has degree 0 on X.
Thus the degree of a divisor on X depends only on its linear equivalence
class, and we obtain a homomorphism Cl X --+ Z as stated. It is surjective,
because the degree of a single point is 1.

Example 6.10.1. A complete nonsingular curve X is rational if and only if

there exist two distinct points P,Q E X with P "' Q. Recall that rational
means birational to P 1 . If X is rational, then in fact it is isomorphic to P 1
by (6.7). And on P 1 we have already seen that any two points are linearly
equivalent (6.4). Conversely, suppose X has two points P =1- Q with P "' Q.
Then there is a rational function f E K(X) with (f) = P - Q. Consider
the morphism cp: X --+ P 1 determined by f as in the proof of (6.10). We
have cp*( {0}) = P, so cp must be a morphism of degree 1. In other words,
cp is birational, so X is rational.

138
 6 Divisors

Example 6.10.2. Let X be the nonsingular cubic curve y 2 z = x 3 - xz 2 in

P~, with char k =!= 2. We have already seen that X is not rational (I, Ex. 6.2).
Let Cl 0 X be the kernel of the degree map Cl X ~ Z. Then from the previous
example we know that Cl 0 X =!= 0. We will show in fact that there is a
natural 1-1 correspondence between the set of closed points of X and the
elements of the group Clo X. On the one hand this elucidates the structure
of the group Clo X. On the other hand it gives us a group structure on the
set of closed points of X, which makes X into a group variety (Fig. 9).

Figure 9. The group law on a cubic curve.

Let P 0 be the point (0,1,0) on X. It is an inflection point, so the tangent

line z = 0 at that point meets the curve in the divisor 3P 0 . If Lis any other
line in P 2 , meeting X in three points P,Q,R (which may coincide), then since L
is linearly equivalent to the line z = 0 in P 2 , we have P + Q + R ~ 3P 0
on X, as in (6.6.2) above.
Now to any closed point P EX, we associate the divisor P - P0 E Cl" X.
This map is injective, because if P - P0 ~ Q - P 0 , then P ~ Q, and X
would be rational by the previous example, which is impossible.
To show that the map from the closed points of X to Clo X is surjective,
we proceed in several steps. Let D E Clo X. Then D = [Link], with [Link] = 0.
Hence we can also write D = [Link](Pi - P 0 ). Now for any point R, let the
line P 0 R meet X further in the point T (always counting intersections with
multiplicities-for example, if R = P 0 , we take the line P 0 R to be the tangent
line at P 0 , and then the third intersection Tis alsoP 0 ). Then P 0 + R + T ~
3P 0 , so R - P 0 ~ -(T - P 0 ). If i is an index such that ni < 0 in D, we
take Pi = R. Then replacing Pi by T, we get a linearly equivalent divisor
with the ith coefficient - ni > 0. Repeating this process, we may assume that
D = I.n;(P; - P 0 ) with all n; > 0. We now show by induction on [Link] that
D ~ P - P 0 for some point P. If In; = 1, there is nothing to prove. So
suppose In; ~ 2, and let P,Q be two of the points P; (maybe the same) which
occur in D. Let the line PQ meet X in R, and let the line P 0 R meet X in T.

139
II Schemes

Then we have
P + Q+ R ~ 3P0 and P0 +R + T ~ 3P 0
so
(P- P 0 ) + (Q - P0 ) ~ (T- P 0 ).
Replacing P and Q by T, we get D linearly equivalent to another divisor of the
same form whose In; is one less, so by induction D ~ P - P 0 for some P.
Thus we have shown that the group Clo X is in 1-1 correspondence with
the set of closed points of X. One can show directly that the addition law
determines a morphism of X x X --+X, and the inverse law determines a
morphism X--+ X (see for example Olson [1]). Thus X is a group variety
in the sense of (1, Ex. 3.21). See (IV, 1.3.7) for a generalization.

Remark 6.10.3. This example of the cubic curve illustrates the general fact
that the divisor class group of a variety has a discrete component (in this
case Z) and a continuous component (in this case Clo X) which itself has the
structure of an algebraic variety.
More specifically, if X is any complete nonsingular curve, then the group
Clo X is isomorphic to the group of closed points of an abelian variety called
the Jacobian variety of X. An abelian variety is a complete group variety
over k. The dimension of the Jacobian variety is the genus of the curve.
Thus the whole divisor class group of X is an extension of Z by the group
of closed points of the Jacobian variety of X.
If X is a nonsingular projective variety of dimension ~ 2, then one can
define a subgroup Clo X of Cl X, namely the subgroup of divisor classes
algebraically equivalent to zero, such that Cl X /Clo X is a finitely generated
abelian group, called the Neron-Severi group of X, and CloX is isomorphic
to the group of closed points of an abelian variety called the Picard variety
of X.
Unfortunately we do not have space in this book to develop the theory
of abelian varieties and to study the Jacobian and Picard varieties of a
given variety. For more information and further references on this beautiful
subject, see Lang [1], Mumford [2], Mumford [5], and Hartshorne [6].
See also (IV, §4), (V, Ex. 1.7), and Appendix B.

Cartier Divisors
Now we want to extend the notion of divisor to an arbitrary scheme. It
turns out that using the irreducible subvarieties of codimension one doesn't
work very well. So instead, we take as our point of departure the idea that
a divisor should be something which locally looks like the divisor of a
rational function. This is not exactly a generalization of the Weil divisors
(as we will see), but it gives a good notion to use on arbitrary schemes.

Definition. Let X be a scheme. For each open affine subset U = Spec A,

letS be the set of elements of A which are not zero divisors, and let K ( U) be

140
 6 Divisors

the localization of A by the multiplicative systemS. We call K(U) the total

quotient ring of A. For each open set U, let S(U) denote the set of elements of
r(U,f!'x) which are not zero divisors in each local ring {!!x for x E U. Then the
rings S(U)- 1 r(U,(!Jx) form a presheaf, whose associated sheaf of rings X we
call the sheaf of total quotient rings of(!). On an arbitrary scheme, the sheaf
f replaces the concept of function field of an integral scheme. We denote
by f* the sheaf (of multiplicative groups) of invertible elements in the
sheaf of rings f. Similarly {!!* is the sheaf of invertible elements in (!).

Definition. A Cartier divisor on a scheme X is a global section of the sheaf

f* j(!J*. Thinking of the properties of quotient sheaves, we see that a
Cartier divisor on X can be described by giving an open cover {U;} of X,
and for each i an element J; E r(U;,f*), such that for each i,j, J;!jj E
r( U; n Ui,{!!*). A Cartier divisor is principal if it is in the image of the
natural map r(X,f*) --> T(X,f* j{!!*). Two Cartier divisors are linearly
equivalent if their difference is principal. (Although the group operation
on x* j{!!* is multiplication, we will use the language of additive groups
when speaking of Cartier divisors, so as to preserve the analogy with
Wei! divisors.)

Proposition 6.11. Let X be an integral, separated noetherian scheme, all of

whose local rings are unique factorization domains (in which case we say
X is locally factorial). Then the group Div X of Wei! divisors on X is
isomorphic to the yroup of Cartier divisors r(X,f* j(!J*), and furthermore,
the principal Wei/ divisors correspond to the principal Cartier divisors
under this isomorphism.
PROOF. First note that X is normal, hence satisfies(*), since a UFD is inte-
grally closed. Thus it makes sense to talk about Weil divisors. Since X is
integral, the sheaf ff is just the constant sheaf corresponding to the function
field K of X. Now let a Cartier divisor be given by {(U;,J;)} where {U;} is
an open cover of X, and J; E r(U;,f*) = K*. We define the associated
Weil divisor as follows. For each prime divisor Y, take the coefficient of Y
to be ry(J;), where i is any index for which Y n U; i= 0. If j is another
such index, then J;/jj is invertible on U; n Ui, so vy(J;/jj) = 0 and vy(f) =
t'r(jj). Thus we obtain a well-defined Weil divisor D = Ivr(J;)Y on X.
(The sum is finite because X is noetherian!)
Conversely, if D is a Wei! divisor on X, let x EX be any point. Then D
induces a Weil divisor Dx on the local scheme Spec {!!x· Since {!!xis a UFD,
Dx is a principal divisor, by (6.2), so let Dx = Ux) for some fx E K. Now
the principal divisor Ux) on X has the same restriction to Spec {!!x as D,
hence they differ only at prime divisors which do not pass through x. There
are only finitely many of these which have a non-zero coefficient in D or
Ux), so there is an open neighborhood U x of x such that D and Ux) have the
same restriction to U x· Covering X with such open sets U x• the functions
fx give a Cartier divisor on X. Note that if f,f' give the same Weil divisor

141
II Schemes

on an open set U, then f/f' E r(U,lD*), since X is normal (cf. proof of (6.2) ).
Thus we have a well-defined Cartier divisor.
These two constructions are inverse to each other, so we see that the
groups of Wei! divisors and Cartier divisors are isomorphic. Furthermore
it is clear that the principal divisors correspond to each other.
Remark 6.11.1A. Since a regular local ring is UFO (Matsumura [2, Th. 48,
p. 142]), this proposition applies in particular to any regular integral sepa-
rated noetherian scheme. A scheme is regular if all of its local rings are
regular local rings.

Remark 6.11.2. If X is a normal scheme, which is not necessarily locally

factorial, we can define a subgroup of Div X consisting of the locally prin-
cipal Wei! divisors: Dis locally principal if X can be covered by open sets
U such that Diu is principal for each U. Then the above proof shows that
the Cartier divisors are the same as the locally principal Wei! divisors.

Example 6.11.3. Let X be the affine quadric cone Spec k[ x,y,z ]/(xy - z 2 )
treated above (6.5.2). The ruling Y is a Wei! divisor which is not locally
principal in the neighborhood of the vertex of the cone. Indeed, our earlier
proof shows that its prime ideal pAm is not a principal ideal even in the
local ring Am. Thus Y does not correspond to a Cartier divisor. On the
other hand 2 Y is locally principal, and in fact principal. So in this case the
group of Cartier divisors modulo principal divisors is 0, whereas Cl X ~
Z/2Z.
Example 6.11.4. Let X be the cuspidal cubic curve y 2 z = x 3 in P~, with
char k # 2. In this case X does not satisfy (*), so we cannot talk about
Wei! divisors on X. However, we can talk about the group CaCl X of
Cartier divisor classes modulo principal divisors. Imitating the case of the
nonsingular cubic curve (6.10.2) we will show:
(a) there is a surjective degree homomorphism deg:CaCl X --+ Z;
(b) there is a 1-1 correspondence between the set of nonsingular closed
points of X and the kernel CaClo X of the degree map, which makes it into
a group variety; and in fact
(c) there is a natural isomorphism of group varieties between CaCloX
and the additive group Ga of the field k (1, Ex. 3.21a).
To define the degree of a Cartier divisor on X, note that any Cartier
divisor is linearly equivalent to one whose local function is invertible in some
neighborhood of the singular point Z = (0,0,1). Then this Cartier divisor
corresponds to a Weil divisor D = 'IniPi on X - Z, and we define the
degree of the original divisor to be deg D = 'Ini. The proof of(6.10) shows
that iff E K is invertible at Z, then the principal divisor (f) on X - Z has
degree 0. Thus the degree of a Cartier divisor on X is well-defined, and it
passes to linear equivalence classes to give a surjective homomorphism
deg: CaCl X --+ Z.

142
 6 Divisors

Now let P 0 be the point (0,1,0) as in the case of the nonsingular cubic
curve. To each closed point P EX - Z, we associate the Cartier divisor
Dp which is 1 in a neighborhood of Z, and which corresponds to the Wei!
divisor P - P 0 on X - Z. First note this map is injective: if P # Q are
two points in X - Z, and if Dp - DQ, then there is an f E K*, which is
invertible at Z, and such that (f) = P - Q on X - Z. Then f gives a
morphism of X to P 1 , which must be birational. But then the local ring of
Z on X would dominate some discrete valuation ring of P 1 , and this is
impossible, because Z is a singular point.
To show that every divisor in CaClo X is linearly equivalent to Dp for
some closed point P E X - Z, we proceed exactly as in the case of the non-
singular cubic curve above. The only difference is to note that the geometric
constructions R H T and P,Q H R, T described above remain inside of
X - Z. Thus the group CaCloX is in 1-1 correspondence with the set of
closed points of X - Z, making it into a group variety.
In this case we are able to identify the group variety as G.. Of course,
we know that X is a rational curve, and so X - Z ~ Af (1, Ex. 3.2). But
in fact, if we use the right parametrization, the group law corresponds. So
define a morphism of G. = Spec k[t] to X - Z by t H (t,1,t 3 ). This is
clearly an isomorphism of varieties. Using a little elementary analytic
geometry (left to reader!) one shows that if P = (t,1,t 3 ) and if Q = (u,1,u 3 ),
then the point T constructed above is just (t + u,1,(t + u) 3 ). So we have an
isomorphism of group varieties of G. to X - Z with the group structure
ofCaCl 0 X.

Invertible Sheaves
Recall that an invertible sheaf on a ringed space X is defined to be a locally
free C9x-module of rank 1. We will see now that invertible sheaves on a
scheme are closely related to divisor classes modulo linear equivalence.

Proposition 6.12. If f£ and .A are invertible sheaves on a ringed space X,

so is f£ ® .A. Iff£ is any invertible sheaf on X, then there exists an
invertible sheaf y- 1 on X such that f£ ® y- 1 ~ C9x.
PROOF. The first statement is clear, since f£ and .A are both locally free of
rank 1, and (!Jx ® (!Jx ~ (!Jx· For the second statement, let f£ be any in-
vertible sheaf, and take f£- 1 to be the dual sheaf y~ = Yfom(fi',(!Jx). Then
!£~ ® f£ ~ Yfom(f£,!£) = (!Jx by (Ex. 5.1).

Definition. For any ringed space X, we define the Picard group of X, Pic X,
to be the group of isomorphism classes of invertible sheaves on X, under
the operation ®. The proposition shows that in fact it is a group.

Remark 6.12.1. We will see later (III, Ex. 4.5) that Pic X can be expressed as
the cohomology group H 1 (X,(!Jk).

143
II Schemes

Definition. Let D be a Cartier divisor on a scheme X, represented by {(Ui,.t;)}

as above. We define a subsheaf Y(D) of the sheaf of total quotient rings
%by taking Y(D) to be the sub-lDx-module of% generated by fi- 1 on
Ui. This is well-defined, since.f;/./j is invertible on Ui n Ui, so fi- 1 andfj 1
generate the same lDx-module. We callY(D) the sheaf associated to D.

Proposition 6.13. Let X be a scheme. Then:

(a) for any Cartier divisor D, Y(D) is an invertible sheaf on X. The
map D ~ Y(D) gives a 1-1 correspondence between Cartier divisors on X
and invertible sub sheaves of %;
(b) 2(D 1 - D 2 ) ~ 2(D 1 ) ® 2(D 2 )- 1 ;
(c) D 1 ~ D 2 if and only if 2(D 1) ~ 2(D 2 ) as abstract invertible
sheaves (i.e., disregarding the embedding in %).
PROOF.
(a) Since each/; E r(Ui,%*), the map lDu,-+ Y(D)Iu, defined by 1 ~ fi- 1
is an isomorphism. Thus Y(D) is an invertible sheaf. The Cartier divisor D
can be recovered from Y(D) together with its embedding in %, by taking
/; on Ui to be the inverse of a local generator of Y(D). For any invertible
subsheaf of %, this construction gives a Cartier divisor, so we have a 1-1
correspondence as claimed.
(b) If D 1 is locally defined by /; and D 2 is locally defined by gi, then
2(D 1 - D 2 ) is locally generated by fi 1 gi, so 2(D 1 - D 2 ) = 2(D 1 ) · 2(D 2 )- 1
as subsheaves of %. This product is clearly isomorphic to the abstract
tensor product 2'(D 1 ) ® 2'(D 2 )- 1 .
(c) Using (b), it will be sufficient to show that D = D 1 - D 2 is principal
if and only if Y(D) ~ lDx. If Dis principal, defined by f E F(X,%*), then
Y(D) is globally generated by f- 1 , so sending 1 ~ f- 1 gives an isomor-
phism lDx ~ Y(D). Conversely, given such an isomorphism, the image of 1
gives an element of r(X,%*) whose inverse will define D as a principal
divisor.

Corollary 6.14. On any scheme X, the map D ~ Y(D) gives an injective

homomorphism of the group CaCl X of Cartier divisors modulo linear
equivalence to Pic X.

Remark 6.14.1. The map CaCl X -+ Pic X may not be surjective, because
there may be invertible sheaves on X which are not isomorphic to any
invertible subsheaf of %. For an example of Kleiman, see Hartshorne
[5, 1.1.3, p. 9]. On the other hand, this map is an isomorphism in most
common situations. Nakai [2, p. 301] has shown that it is an isomorphism
whenever X is projective over a field. We will show now that it is an isomor-
phism if X is integral.

144
 6 Divisors

Proposition 6.15. If X is an integral scheme, the homomorphism CaCl X ~

Pic X of (6.14) is an isomorphism.
PROOF. We have only to show that every invertible sheaf is isomorphic to a
subsheaf of x, which in this case is the constant sheaf K, where K is the
function field of X. So let 2 be any invertible sheaf, and consider the sheaf
2 ®(/)x X. On any open set U where 2 ~ (!Jx, we have 2 @ X ~ X, so
it is a constant sheaf on U. Now because X is irreducible, it follows that any
sheaf whose restriction to each open set of a covering of X is constant, is
in fact a constant sheaf. Thus 2 @ X is isomorphic to the constant sheaf
X, and the natural map 2 ~ 2 @X ~ X expresses 2 as a subsheaf
of X.

Corollary 6.16. If X is a noetherian, integral, separated locally factorial

scheme, then there is a natural isomorphism Cl X ~ Pic X.
PROOF. This follows from (6.11) and (6.15).

Corollary 6.17. If X = PI: for some field k, then every invertible sheaf on X
is isomorphic to (!)(I) for some l E Z.
PROOF. By (6.4), Cl X ~ Z, so by (6.16), Pic X ~ Z. Furthermore the gen-
erator of Cl X is a hyperplane, which corresponds to the invertible sheaf
(!)(1). Hence Pic X is the free group generated by (!)(1), and any invertible
sheaf 2 is isomorphic to (!J(l) for some l E Z.

We conclude this section with some remarks about closed subschemes of

codimension one of a scheme X.

Definition. A Cartier divisor on a scheme X is effective if it can be repre-

sented by {(Ui,.J;)} where all the .J; E r(Ui,(!Ju). In that case we define
the associated subscheme of codimension 1, Y, to be the closed subscheme
defined by the sheaf of ideals § which is locally generated by .J;.

Remark 6.17.1. Clearly this gives a 1-1 correspondence between effective

Cartier divisors on X and locally principal closed subschemes Y, i.e., sub-
schemes whose sheaf of ideals is locally generated by a single element. Note
also that if X is an integral separated noetherian locally factorial scheme,
so that the Cartier divisors correspond to Weil divisors by (6.11 ), then the
effective Cartier divisors correspond exactly to the effective Weil divisors.

Proposition 6.18. Let D be an effective Cartier divisor on a scheme X, and let

Y be the associated locally principal closed subscheme. Then§ y ~ 2(- D).

145
II Schemes

PROOF. 2(- D) is the subsheaf of :ff generated locally by J;. Since D is

effective, this is actually a subsheaf of (!Jx, which is none other than the ideal
sheaf§ y of Y.

EXERCISES

6.1. Let X be a scheme satisfying (*). Then X x P" also satisfies (*),and Cl(X x P") ~
(ClX) X Z.

*6.2. Varieties in Projective Space. Let k be an algebraically closed field, and let X
be a closed subvariety of P~ which is nonsingular in codimension one (hence
satisfies(*)). For any divisor D = IniY; on X, we define the degree of D to be
Ini deg 1;, where deg 1; is the degree of 1;, considered as a projective variety
itself (I, §7).
(a) Let V be an irreducible hypersurface in P" which does not contain X, and let
1; be the irreducible components of V n X. They all have codimension 1 by
(I, Ex. 1.8). For each i, let J; be a local equation for Von some open set Ui of
P" for which Y; n Ui =1= 0, and let ni = Vy (hJ, where J: is the restriction of
J; to ui " X. Then we define the divisor v.x to be Ini Y;. Extend by linearity,
and show that this gives a well-defined homomorphism from the subgroup of
Div P" consisting of divisors, none of whose components contain X, to Div X.
(b) If Dis a principal divisor on P", for which D.X is defined as in (a), show that D.X
is principal on X. Thus we get a homomorphism Cl P" --> Cl X.
(c) Show that the integer ni defined in (a) is the same as the intersection multiplicity
i(X,V; Y;) defined in (I, §7). Then use the generalized Bezout theorem (I, 7.7)
to show that for any divisor Don P", none of whose components contain X,
deg(D.X) = (deg D)· (deg X).
(d) If Dis a principal divisor on X, show that there is a rational function f on P"
such that D = (f).X. Conclude that deg D = 0. Thus the degree function
defines a homomorphism deg: Cl X --> Z. (This gives another proof of (6.10),
since any complete nonsingular curve is projective.) Finally, there is