LONDON MATHEMATICAL SOCIETY LECTURE NOTE SERIES

Managing Editor: PROFESSOR 1. M. JAMES, Mathematical Institute,

24-29 St Giles, Oxford
Already published in this series
1. General cohomology theory and K-theory, PETER HILTON.
4. Algebraic topology: A student's guide, J. F. ADAMS.
5. Commutative algebra, J. T. KNIGHT.
7. Introduction to combinatory logic, J. R. HINDLEY, B. LERCHER,
and J. P. SELDIN.
8. Integration and harmonic analysis on compact groups, R. E.
EDWARDS.
9. Elliptic functions and elliptic curves, PATRICK DU VAL.
10. Numerical ranges II, F. F. BONSALL and J. DUNCAN.
11. New developments in topology, G. SEGAL (ed.).
12. Symposium on complex analysis Canterbury, 1973, J. CLUNIE and
W. K. HAYMAN (eds. ).
13. Combinatorics, Proceedings of the British combinatorial con-
ference 1973, T. P. McDONOUGH and V. C. MAVRON (eds.).
14. Analytic theory of abelian varieties, H. P. F. SWINNERTON-
DYER.
15. An introduction to topological groups, P. J. HIGGINS.
16. Topics in finite groups, TERENCE M. GAGEN.
17. Differentiable germs and catastrophes, THEOOOR BROCKER and
L. LANDER.
18. A geometric approach to homology theory, S. BUONCRISTIANO,
C. P. ROURKE and B. J. SANDERSON.
19. Graph theory, coding theory and block designs, P. J. CAMERON
and J. H. VAN LINT.
20. Sheaf theory, B. R. TENNISON.
21. Automatic continuity of linear operators, ALLAN M. SINCLAIR.
22. Presentations of groups, D. L. JOHNSON.
23. Parallelisms of complete designs, PETER J. CAMERON.
24. The topology of Stiefel manifolds, I. M. JAMES.
25. Lie groups and compact groups, J. F. PRICE.
26. Transformation groups: Proceedings of the conference in the
University of Newcastle upon Tyne, August 1976, CZES
KOSNIOWSKI.
27. Skew field constructions, P. M. COHN.
28. Brownian motion, Hardy spaces and bounded mean oscillation,
K. E. PETERSEN.
29. Pontryagin duality and the structure of locally compact abelian
groups, SIDNEY A. MORRIS.
30. Interaction models, N. L. BIGGS.
31. Continuous crossed products and type III von Neumann algebras,
A. VAN DAELE.
32. Uniform algebras and Jensen measures, T. W. GAMELIN.
33. Permutation groups and combinatorial structures, N. L. BIGGS and
A. T. WHITE.
34. Representation theory of Lie groups, M. F. A TIYAH.
35. Trace ideals and their applications, BARRY SIMON.
36. Homological group theory, edited by C. T. C. WALL.
37. Partially ordered rings and semi-algebraic geometry, GREGORY
W. BRUMFIEL.
38. Surveys in combinatorics, edited by B. BOLLOBAS.
London Mathematical Society Lecture Note Series. 39

Affine Sets and Affine Groups

D. G. Northcott
Professor of Pure Mathematics
University of Sheffield

CAMBRIDGE UNIVERSITY PRESS

CAMBRIDGE
LONDON NEW YORK NEW ROCHELLE
MELBOURNE SYDNEY
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore,
São Paulo, Delhi, Dubai, Tokyo, Mexico City

Cambridge University Press

The Edinburgh Building, Cambridge cb2 8ru, UK

Published in the United States of America by Cambridge University Press, New York

www.cambridge.org
Information on this title: www.cambridge.org/9780521229098

© Cambridge University Press 1980

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.

First published 1980

Re-issued 2010

A catalogue record for this publication is available from the British Library

Library of Congress Cataloguing in Publication Data

Northcott, Douglas Geoffrey

Affine sets and affine groups

(London Mathematical Society lecture note series; 39

ISSN 0076-0552)
1. Geometry, Affline. 2. Set theory.
I. Title. II. Series.
516’4 QA477 79-41595

isbn 978-0-521-22909-8 Paperback

Cambridge University Press has no responsibility for the persistence or

accuracy of URLs for external or third-party internet websites referred to in
this publication, and does not guarantee that any content on such websites is,
or will remain, accurate or appropriate.
Contents
page
Introduction ix

PART I. AFFINE SETS

Chapter 1 Preliminaries concerning algebras 3

1.1 Algebras 3
1.2 Subalgebras and factor algebras 4
1.3 Examples of K-algebras 5
1.4 Tensor products of vector spaces 6
1.5 Tensor products of algebras 7
1.6 Enlargement of the ground field 9

Chapter 2 Affine sets 12

2.1 Rational maximal ideals 12
2.2 Function algebras 17
2.3 Loci and the associated topology 19
2.4 Affine sets 24
2.5 Examples of affine sets 28
2.6 Morphisms of affine sets 33
2. 7 Products of function algebras 37
2.8 Products of affine sets 40
2.9 Some standard morphisms 42
2.10 Enlargement of the ground field 46
2.11 Generalized points and generic points 55

Chapter 3 Irreducible affine sets 60

3.1 Irreducible spaces 60
3.2 Irreducible affine sets 63
3.3 Localization 70
3.4 Rational functions and dimension 74
3.5 Enlargement of the ground field 79
3.6 Almost surjective K-morphisms 84

Chapter 4 Derivations and tangent spaces 95

4.1 Derivations in algebras 95
4.2 Examples of derivations 101
4.3 Derivations in fields 104
v
 4.4 Tangent spaces and simple points 114
4.5 Tangent spaces and products 124
4.6 Differentials 125

PART II. AFFINE GROUPS

Chapter 5 Affine groups 133

5.1 Affine groups 133
5.2 Components of an affine group 143
5.3 Examples of affine groups 148
5.4 K-homomorphisms of affine groups 159
5.5 K- morphic actions on an affine set 162
5.6 G-modules 166
5. 7 General rational G- modules 173
5.8 Linearly reductive affine groups 175
5.9 Characters and semi-invariants 181
5.10 Linearly reductive affine groups and invariant
theory 189
5.11 Quotients with respect to linearly reductive
groups 196
5.12 Quotients with respect to finite groups 202
5.13 Quotient groups of affine groups 206

Chapter 6 The associated Lie algebra 222

6.1 General K-algebras 222
6.2 Lie algebras 224
6.3 The Lie algebra of an affine group 227
6.4 Extension of the ground field 241
6.5 A basic example 243
6.6 Further examples 250
6. 7 Adjoint representations 255
Chapter 7 Power series and exponentials 263
7.1 Rings of formal power series 263
7.2 Modules over a power series ring 267
7.3 Exponentials 271
7.4 Applications to affine groups 279
References 283
Index 284
vi
Introduction

The topics treated in the following pages were largely covered in

two seminars, both given at Sheffield University, one during the session
1976/7 and the other during 1978/9. I had noted sometime earlier that
M. Hochster and J. A. Eagon had established a connection between
Determinantal Ideals and Invariant Theory. However in order to under-
stand what was involved I had first to acquaint myself with the relevant
parts of the theory of Algebraic Groups. With this in mind, I began to
read J. Fogarty's book on Invariant Theory.
Almost at once my interest broadened. It had been my experience
to see Commutative Algebra develop out of attempts to provide classical
Algebraic Geometry with proper foundations, but it had been a matter of
regret that the algebraic machinery created for this purpose tended to
conceal the origins of the subject. Fogarty's book helped me to see how
one could look at Geometry from a readily accessible modern standpoint
that was still not too far r~moved from the kind of Coordinate Geometry
which now belongs to the classical period of the subject.
When my own ideas had reached a sufficiently advanced stage I
decided to try and develop them further by committing myself to giving a
seminar. With a subject such as this, and in circumstances where the
time available was very limited, it was necessary to assume a certain
amount of background knowledge. Indeed most accounts of aspects of the
theory of Algebraic Groups assume a very great deal in the way of
prerequisites. In my case the audience could be assumed to be know-
ledgeable about Commutative Noetherian rings and I planned the lectures
with this in mind. The outcome is that the following treatment is very
nearly self-contained if one presupposes a knowledge of field theory,
tensor products and the more familiar parts of Commutative Algebra
including, of course, the famous Basis Theorem and Zeros Theorem of
Hilbert. It is true that there are a few places where additional background

vii
knowledge is required, but the reader who knows the topics mentioned
above will find that other results are rarely used, and that where they
are it will usually require little time and effort to fill in the gaps. To
help him I have provided suitable references wherever they are likely
to be needed
The book falls naturally into two parts with Chapters 1-4 forming
the first part. Here the aim is to show how those loci in classical
analytical geometry that are defined by the solutions of simultaneous
algebraic equations (together with the appropriate transformations of one
such object into another) can be turned into a category. In this context
the ambient affine space which makes geometrical thinking possible has
to be removed from the picture, but not so far that it cannot be brought
back readily when geometrical insight into a situation is needed.
Turning now to points of detail, Chapter 1 is used to explain certain
matters that have to do with terminology and notation. It is also used to
give a brief survey of the properties of tensor products over a field. The
latter enables the discussion of (i) products of affine algebraic sets, and
(ii) the consequences of enlarging the ground field, to take place later
without an interruption to explain technicalities of a purely algebraic nature
The development of geometrical ideas begins with Chapter 2. Beside
describing affine sets, their products and their morphisms, I have also
revived the theory of specializations because, it seems to me, it provides
techniques that are both interesting and highly effective. Chapter 3
introduces the concept of the irreducible components of an affine set, the
idea of dimension, and the very important topic of almost surjective
morphisms. The last of the chapters devoted to Geometry deals with the
subject of tangent spaces and simple points.
Part 1 contains all the Geometry that is needed for the reader to be
able to understand the rest of the book. But although it was planned with
the idea of supporting an introduction to Algebraic Groups it is the author's
hope that it will be of interest to those whose main concern is to get an
insight into the foundations of Algebraic Geometry. If the first four
chapters are read with this more limited end in view, then I would suggest
ending with section (4. 5) because the last section of the chapter consists
of technical material used later to study the connection between an Affine

viii
Group and its associated Lie Algebra. A natural continuation of the
geometrical sections would be the theory of tangent bundles.
The second part moves fairly quickly into the study of Affine Groups.
These are groups which also have an affine structure and where the group
operations are compatible with this structure. After the definitions and
a discussion of the relation between connectedness and irreducibility,
such topics as rational representations, linearly reductive groups, and
the beginnings of invariant theory are considered. Chapter 5 ends with a
comparatively elementary proof that every factor group of an affine group,
with respect to a closed normal subgroup, has itself a natural.structure
as an affine group. This is one place where the account draws more
heavily on the reader's knowledge of Commutative Algebra than it does
elsewhere, but the topic seemed of sufficient importance to justify a
departure from the guide lines which I had set myself for the book as a
whole. It may help the reader to know that Chapter 5 makes very little
use of Chapter 4. If therefore he wishes to get to the second part as
quickly as possible, he may prefer to begin Chapter 5 after completing
Chapter 3, and then to return to Chapter 4 to fill in certain gaps at a
later stage. This in fact is what I did in my seminars.
The theory of Lie Algebras is not introduced until Chapter 6. Here
it is shown that with each Affine Group there is associated a Lie Algebra
and a detailed study is made of some of the most important examples. It
is possible to exploit this connection very effectively in the case of a
ground field of characteristic zero. The final chapter provides the theory
which in this case enables properties of the associated Lie Algebras to be
transferred back to the Affine Groups. It is here that the account stops.
To continue further it would be necessary to include a short course on
Lie Algebras after which an account of classical affine groups could be
given.
I have, of course, made use of the writings of other mathematicians.
I have already mentioned the special debt I owe to J. Fogarty's book on
Invariant Theory. I also made considerable use of C. Chevalley's first
account of Algebraic Groups, and the presentation, by H. Bass, of a
course given by A. Borel on Linear Algebraic Groups. In the case of
Geometry, it was a pleasure to re-read part of A. Weil's book on the

ix
Foundations of Algebraic Geometry. So far as Commutative Algebra is
concerned I have relied principally on the books of O. Zariski, P. Samuel
M. Nagata, and my own writings. Detailed references will be found in the
text and in the booklist given at the end.
On a more personal level I would like to thank those who came to
my seminars and encouraged me by the interest they showed, and once
again it is a pleasure to thank Mrs. E. Benson for doing all the typing
with such excellent results.
Finally I would like to express my appreciation to the London
Mathematical Society and the Syndics of the Cambridge University Press
for agreeing to let this appear in their series of Lecture Notes.

Sheffield D. G. Northcott
February 1979

x
Part I. Affine Sets
1 . Preliminaries concerning algebras

General remarks

In order to describe the structure of affine algebraic sets we need

to use the theory of algebras, so this initial chapter will be devoted to a
survey of some of the basic properties of algebras that will be needed
shortly. However for the present it will suffice to restrict our attention
to those algebras which are associative and have identity elements. These
will be referred to as unitary, associative algebras.
There is one convention we shall use to which it is necessary to draw
attention. Normally any ring we consider will have an identity element.
When dealing with such rings the term ring-homomorphism is always
understood to mean a mapping which is compatible with addition and
multiplication and which also takes identity element into identity element.
In other terms our ring-homomorphisms are assumed to be unitary.
Throughout Chapter 1 K will be used to denote a field. It is not assumed
that K is algebraically closed nor even that it contains infinitely many
elements.

1. 1 Algebras

Let R be a ring which is also a vector space over the field K. R

is said to be an associative K-algebra if

(ka)b = a(kb) = k(ab)

for all a, b E Rand k E K. If in addition R possesses an identity

element 1R , then R is described as a unitary, associative K-algebra.
Suppose that R is a unitary, associative K-algebra. Then the
mapping

3
defined by w(k) = k1R is a (unitary) ring-homomorphism which maps K
into the centre of R. This ring-homomorphism is referred to as the
structural homomorphism of the algebra. If R is non-trivial, that is
if its identity element is not zero, this allows us to identify K with a
subring of the centre of R.
Let us reverse the procedure. Suppose that R is a ring with
identity element and that we are given a ring-homomorphism w: K ~ R
which maps K into the centre of R. If for a E Rand k E K we put
ka = w(k)a, then R becomes a unitary, associative K-algebra having w
as its structural homomorphism. Both ways of looking at these algebras
will be useful.
Let Rand R' be unitary, associative K-algebras and ¢ : R ~ R'
a (unitary) ring-homomorphism. We say that ¢ is a homomorphism of
K-algebras if it is also a K-linear mapping of R into R'. Thus if
w : K ~ R and w' : K ~ R I are the structural homomorphisms, then ¢
is a homomorphism of K-algebras if and only if </> 0 w = w'. In the case
where Rand R' are both non-trivial we may regard K as being em-
bedded in each of the two rings. In this situation ¢ is a homomorphism of
K-algebras when and only when it leaves the elements of K fixed.
An isomorphism of K-algebras is, of course, a homomorphism of
K-algebras which is also a bijection. Every isomorphism has an inverse
which is also an isomorphism.

1.2 Subalgebras and factor algebras

Let R be a unitary, associative K-algebra and let S be a subring

of R. (It is understood that the identity element of R is shared with all
its subrings.) If S is also a K-subspace of R, then S has the structure
of a unitary, associative K-algebra and the inclusion mapping j : S ~ R
is a homomorphism of K-algebras. We describe this situation by saying
that S is a subalgebra of the K-algebra R. Evidently the intersection of
any family of subalgebras of R is again a subalgebra.
Let A be a set of elements of R. There will be a smallest sub-
algebra S (say) which contains A. We say that S is generated by A.
The typical element of S is a linear combination (with coefficients in K)

4
of finite products a a ... a (n:::: 0), where a. EA. Naturally the empty
1 2 n 1
product is understood to have the value lR' If S = R, then A is said
to be a system of generators for R. If R can be generated by a finite
set, then it is said to be finitely generated as a K-algebra.
Now let ~ be a two-sided ideal of R. Then ~ is a K-subspace of
R and the ring R/~ has a unique structure as a unitary, associative
K-algebra subject to the requirement that the natural mapping of R on to
R/~ shall be a homomorphism of K-algebras. Whenever we refer to
R/~ as a K-algebra it is always this structure that we have in mind.

1. 3 Examples of K-algebras

All the examples of K-algebras given in section (1. 3) are unitary

and associative.

Example 1. K itself is a K-algebra the structural homomorphism

in this case being the identity mapping. It follows that if R is any unitary,
associative K-algebra, then the structural homomorphism w: K ... R is
a homomorphism of K-algebras.

Example 2. Any ring R with identity element which contains K

as a subring of its centre is, of course, a K-algebra. In particular any
extension field of K is a K-algebra.

Example~. Let V be a vector space over K. (The dimension of

V need not be finite.) The endomorphisms of K form a K-algebra which
we denote by En%(V). Here the structural homomorphism is obtained
by mapping the element k, of K, into the endomorphism that consists in
multiplication by k.

Example 4. Let n be a positive integer and denote by M (K) the

n
set of all n x n matrices with entries in K. These form a K-algebra.
To obtain the structural homomorphism we map the element of k, of K,
into the corresponding scalar matrix.
Let V be an n-dimensional vector space over K. Then M (K) and
n
En%(V) are isomorphic K-algebras.

5
 Example 5. Let X , X , •.• , X be indeterminates. Then the
1 2 n
polynomial ring K[X 1 , X 2 , ... , Xn] is a commutative K-algebra.

Example 6. Let V be a non-empty set and denote by (J K(V) the

set of all mappings of V into K. If f, g belong to (J K(V) we define
their sum and product by

(f + g)(v) = f(v) + g(v)

and

(fg)(v) = f(v)g(v).

This turns (J K(V) into a commutative K-algebra where the structural

homomorphism identifies k E K with the corresponding constant function.

1. 4 Tensor products of vector spaces

Before we examine the important concept of the tensor product of

two K-algebras we shall say a few words about tensor products of vector
spaces.
Throughout section (1. 4) U, V and W will denote vector spaces
over K. (Their dimensions need not be finite.) All tensor products will
be taken over K so we shall use ® instead of the more ornate ®K'
As is well known U ® V is a vector space over K its typical
element being a finite sum of tensor products u ® v, where u E U and
v E V. A fundamental property of U ® V is the following: given a bi-
linear mapping

w:UXV-+W,

there is a unique K-linear mapping w : U ® V -+ W such that

w(u ® v) = w(u, v) for all u in U and v in V.
Let {ui }iEI be a base for the space U and {vj } j EJ a base for V.
Then the elements u. ® v. (i E I, j E J) form a base for U ® V. Thus
1 ]
if A is a linearly independent subset of U and B a linearly independent
subset of V, then the elements a ® b (a E A, b E B) form a linearly

6
independent subset of U ® V. In particular if u E U, U *- 0 and v E V,
v*-o, then u ® v *- O.
The following result is often useful.

Theorem 1. Let u I ' u 2 ' ••• , un belong to U and vI' v 2 ' ••• , vn
to V. Suppose that u I ' u 2 ' ••• , un are linearly independent over K
and that

u ®v
1 1
+ u2 ®v
2
+... + un ®v
n
= O.
Then v. = 0 -for i = I, 2, ••• , n.
-- 1

Remark. Of course the roles of U and V may be interchanged.

Proof. The theorem becomes obvious as soon as one expresses

VI' v 2 ' ••• , vn in terms of a base for V over K.
Now let q,: U - U' and l/I: V - V' be homomorphisms of vector
spaces. These give rise to a homomorphism

q, ® l/I : U ® V - U' ® V'

in which u ® v is mapped into q,(u) ® l/I(v). Obviously if q, and l/I are

surjections, then q, ® l/I is a surjection. But K is a field. Consequently
if q, and l/I are injections, then q, ® l/I is also an injection.
Let A be a subspace of U and B a subspace of V. The inclusion
mappings A - U and B - V induce an injection A ® B - U ® V. This
enables us to regard A ® B as a subspace of U ® V. Next the natural
homomorphisms U ... U/A and V - V /B give rise to a surjective homo-
morphism

U ®V ... (U/A) ® fY /B)

and it is readily checked that

Ker{U ® V - (U/A) ® (V /B)} = (A ® V) + (U ® B). (1. 4. 1)

1. 5 Tensor products of algebras

In this section the letters Rand S will be used to denote unitary,

7
associative K-algebras. As in the last section we shall use ® to stand
for a tensor product taken over K.
If we form R ® 8, then this is not simply a K-space. It has in fact
the structure of a unitary, associative K-algebra in which multiplication
satisfies

(r ® s )(r ® s ) = r r ® s s , (1. 5. 1)
1 12 2 12 12

where the notation is self-explanatory. The identity element of this

algebra is lR ® 18 , Furthermore R ® 8 and 8 ® R are isomorphic
K-algebras under an isomorphism which matches r ® s with s ® r.
Consider the mapping R .... R ® 8 in which r
r i&I 18, This is a 1-+

homomorphism of K-algebras and in the case where 8 is non-trivial (and

therefore 18 '* 08) it is an injection. Accordingly when 8 is non-trivial
we may regard R as a subalgebra of the K-algebra R i&I 8. Naturally
there is a corresponding homomorphism 8'" R ® 8 to which similar
observations apply.
K itself plays a special role in the theory of tensor products. This
manifests itself through the fact that we always have an isomorphism of
K-algebras

(1. 5. 2)

in which r ® k is matched with kr.

Now assume that «P : R .... R' and l/I : 8 .... 8' are homomorphisms of
K-algebras. Then

«P ® l/I : R i8l 8 .... R' ® 8'

is a homomorphism of K-algebras. Let III be a two-sided ideal of R,

~ a two-sided ideal of 8, and let us apply the above observation to the
natural mappings R .... R/l1l and 8 .... 8/~. 8ince these are surjective
homomorphisms of K-algebras, they give rise to

R i8l 8 .... (R/I1I) i8l (8/~) (1. 5. 3)

which is itself a surjective homomorphism of K-algebras. Put

8
 [SU, ~ ] = SU ® S + R ® ~ (1. 5. 4)

this being regarded in the first instance as a subspace of R ® S. By

(1.4. 1) it is none other than the kernel of (1. 5. 3). It follows therefore
that [SU, ~] is a two-sided ideal of R i8l S and moreover we have an iso-
morphism

(R ® S)/[SU, ~] "" (R/SU) i8l (S/~) (1. 5. 5)

of K-algebras.

Lemma 1. Let the notation be as above and suppose that SU '" R,

~ '" S. Then the inverse images of [SU,~] with respect to the canonical
homomorphisms R .... R ® S and S - R ® S are SU and ~ respectively.

This is clear because, as previously observed, [SU, ~] is the kernel

of (1. 5. 3).
Finally we observe that if {a.}. rT is a system of generators for
1 1~

the K-algebra Rand {b.}. J is a system of generators for the K-

J JE
algebra S, then the elements

where i E I and j E J, form a system of generators for R ® S. In par-

ticular!! R and S are finitely generated K-algebras so too is R ® S.

1.6 Enlargement of the ground field

Throughout section (1. 6) we shall assume that L is an extension

!ield of K.
First suppose that V is a vector space over K and put

(1. 6. 1)

L
Then V has a natural structure as a vector space over L in which

~'(v ® ~) = v ® ~'~ (1. 6. 2)

L
it being understood that v E V and A, ~' E L. Naturally V can also be

9
regarded as a K-space.
Consider the mapping V .... V L in which v 1-+ v ® l L . This is K-
linear and it is also an injection. Consequently we may (and usually do)
regard V as a K-subspace of VL . On this understanding any base for
V over K is a base for VL over L; in particular, if the elements
v , v , ... , v of V are linearly independent over K, then they are
1 2 n
also linearly independent over L. Furthermore to a certain extent the
roles of V and L can be interchanged. To be precise let {A.}. I be a
L 1 IE
base for Lover K. If now W E V , then there exists a unique family
{w. L I of elements of V such that (i) only a finite number of the w.
1 IE 1
are non zero, and (ii) W = ~ A.W
1 1
..
Next let f : U .... V be a homomorphism of K-spaces. This will have
a unique extension to a homomorphism U .... V
L L of L-spaces. In fact the
extension of f is just f ® L.
Now suppose that Rand S are unitary, associative K-algebras.
Then RL =R ® L is both a vector space over L and at the same time
a ring with an identity element. Indeed RL is a unitary, associative L-
L
algebra in which the structural homomorphism L .... R maps the element
A, of L, into lR ® A.
Our earlier remarks show that we can regard R as being embedded
in RL. If this is done, then R is a subalgebra of RL when the latter is
regarded as a K-algebra. Moreover any set of elements of R which
generates R as a K-algebra will also generate RL as an L-algebra. In
particular, when R is a finitely generated K-algebra, RL is a finitely
generated L-algebra.
Once again let {A.}. I be a base for Lover K and suppose that
1 IE
{3 E RL. Our previous observations show that there exists a unique family
{b. }. rT of elements of R such that {3 = ~ A.b.. Hence if III is a two-
1 It:J. 1 1
sided ideal of Rand'll R L is the two-sided ideal it generates in R L,
then {3 E'llRL if and only if b. E III for all i E I. It follows that
1

(1. 6. 3)

Also if {'lI.}. J is a family of two-sided ideals of R, then

J JE

10
 n (1. 6. 4)
jEJ

Finally assume that cp: R ~ S is a homomorphism of K-algebras.

Then the homomorphism RL ~ SL, of L-spaces, which it induces is
actually a homomorphism of L-algebras. For example, suppose that ~[

is a two-sided ideal of R. Then the natural homomorphism R - R/~[

induces a surjective homomorphism RL - (R/~)L of L-algebras, that

is a homomorphism R ® L - (R/~) ® L. By (1. 4. 1) the kernel of this is
~ ® L =~ RL and therefore we have an isomorphism

(1. 6. 5)

of L-algebras.
Although considerably more could be said about algebras and tensor
products in the general context of this chapter, we have now reached a
convenient point at which to begin concentrating the discussion on the
situations that are particularly relevant to the treatment of affine alge-
braic sets.

11
2 . Affine sets

General remarks

In this chapter the conventions and notation remain the same as

those introduced in the introductory paragraphs to Chapter 1. In particulal
K will denote a completely arbitrary field and the symbol i8l will be used
to denote a tensor product taken over K when there is no possibility of
a misunderstanding. In other situations, for example when it is neces-
sary to form a tensor product over a different field, a subscript will be
attached to i8l in order to make the intention quite clear.
We shall now place further restrictions on the algebras we study.
Whereas Chapter 1 was prinCipally concerned with K-algebras that were
unitary and aSSOCiative, throughout this chapter those that will occupy
our attention will usually be commutative as well.
One further point. We shall have occasion to consider polynomials
in indeterminates X I , X 2 , ... , Xn with coefficients in K. By the ---
null
polynomial will be meant the polynomial all of whose coefficients are
zero. Now if f(X I , X , ... , X ) belongs to the polynomial ring
2 n
K[X , X , ... , X ], then it determines a function, f say, from the n-
1 2 n
fold product K x K x ... x K to K. We shall call f a polynomial
function. It is necessary to bear in mind that f may be the zero function
(that is all of its values may be zero) even though f(X , X , ... , X )
1 2 n
is not the null polynomial.

2. 1 Rational maximal ideals

Throughout section (2.1) Rand S will denote K-algebras which

are unitary, associative and commutative. Naturally the tensor product
R i8l S, taken over K, inherits the same three properties.
Let M be a maximal ideal of R. Then RIM is a field containing
K as a subfie ld.

12
 Definition. M will be said to be a 'rational maximal ideal' or
(more precisely) to be 'K-rational' if RIM =K or (equivalently) if
RIM and K are isomorphic K-algebras.

The set of K-rational maximal ideals of R will be denoted by

W1 K (R).

Theorem 1. Let cp : S ~ R be a homomorphism of K-algebras

and M a rational maximal ideal of R. Then cp -1 (M) is a rational
maximal ideal of S.

Proof. We may regard S/cp-l(M) as a subalgebra of the non-

trivial K-algebra RIM, and then

The theorem follows.

In the case where cp is surjective the relation between the rational
maximal ideals of R and those of S is very simple. We record the facts
in the following

Corollary. Let cp: S ~ R be a surjective homomorphism of

K-algebras and let M be a maximal ideal of R. Then cp -1 (M) is a
maximal ideal of S. Furthermore if either M or cp -1 (M) is K- rational,
then so is the other.

Proof. We have an isomorphism S/CP-l(M) '" RIM of K-algebras

and all the assertions follow from this.

Example. Let Xl' X 2 , ••• , Xn be indeterminates, where n 2: 1,

and put

Here it is understood that the right hand side is an ordinary cartesian

product and there are n factors. Thus a typical element of Kn is a
sequence (aI' a 2, ••. , an)' where a i E K, and it is clear that we have
a bijection between

13
which matches (a , a , ..• , a ) with the rational maximal ideal
1 2 n
(Xl-aI' X 2 -a 2 , ... , Xn-a n)·
We mention, in passing, that when K is algebraically closed, ever;
maximal ideal of K[X I , X 2 ' ••• , Xn] is K-rational. t This is an impor-
tant result with far-reaching consequences, but they will concern us only
to a small extent in Chapter 2.
Suppose, by way of contrast, that K is a finite field with q ele-
ments. Then the non-zero elements of K form a multiplicative group
of order q - 1 and therefore a q = a for all a E K. In particular the
polynomial X; - Xl belongs to all the rational maximal ideals of
K[Xl' X 2 ' ••• , Xn] and, in consequence, the intersection of these maxi-
mal ideals is not zero. However, when K is an infinite field the situation
is quite different as we shall now show.

Theorem 2. Let U be an infinite subset of K, and let

cp(X , X , .•• , X ) be non-null and belong to K[X , X , ..• , X]. Then
1 2 n I 2 n
there exist WI' W 2' ••• , Wn in 0 with the property that
cp( wI' W 2' ••• , W n) '* O.

Proof. We use induction on n and begin by observing that the

assertion is certainly true when n = 1. Now assume that n > 1 and that
the theorem has been proved for polynomials involving only n - 1 vari-
ables.
We can write
m
cp(X 1 , X 2 ' ... , Xn) =.L cp].(X I , X 2 ' ... , X n-1 )Xnm - j ' (2. 1. 1)
]=0
where cp.(X , X , •.. , X 1)
K[X , X , •.• , Xn_ 1] and
E
] I 2 n- I 2
cp (Xl' X , ... , X 1) is not null. The inductive hypothesis shows that
a 2 n-
there exist W , W , ••• , W 1 in 0 for which cp (w ,w , ••• , W 1) '* 0;
1 2 n- a 1 2 n-
and then the case of a single variable shows that we can find W , in 0,
n

t See [(9) Theorem 6, p. 285]. The number in round brackets refers to

the list of references at the end.

14
 m
so that cp(w 1 , ••• , wn _1' W ) = L cp.(w , ... , W
n-l
)w m - j is different
n
n j=O] 1
from zero.

Corollary. If K is an infinite field and n 2:: 1, then the inter-

section of all the K-rational maximal ideals of K[X 1 , X 2 , ••• , Xn] is
zero.

At this stage it is convenient to introduce a new definition. We shall

say that the K-algebra R is rationally reduced with respect to K if

n M = (0).
MEmK(R)

This enables us to reorganize some of the information derived above.

Theorem 3. Suppose that n 2:: 1. Then the polynomial ring

K[X1 , X 2 , ••• , Xn] is rationally reduced with respect to K if and only if
K is infinite.

We shall now take a quick look at what happens to R and its rational
maximal ideals if we enlarge the ground field from K to L (say).
Certain general comments on this operation were made in section (1. 6).
These will be used here. Note that RL is now a commutative L-algebra.

Theorem 4. Let M be a rational maximal ideal of the K-algebra

R, and let L be an extension field of K. Then MRL is a rational maxi-
mal ideal of the L-algebra RL.

Proof. By (1. 6. 5) and (1.5.2), we have isomorphisms

RL jMRL .., (RiM) ® L'" K ® L'" L

K K
of L-algebras and from these the theorem follows.

Corollary 1. If R is rationally reduced with respect to K, then

RL is rationally reduced with respect to L.

Proof. If M denotes a typical rational maximal ideal of R, then

the intersection of all the M's is zero. It follows, from (1. 6.4), that
the RL -ideals MRL have zero as their intersection, so a fortiori the full

15
set of L-rational maximal ideals also has intersection zero.
From section (1. 6) we know that a homomorphism ¢ : S ... R of
K-algebras induces a homomorphism SL ... RL of L-algebras.

Corollary 2. Let cp: S ... R be a homomorphism of K-algebras

and M ~ K-rational maximal ideal of R. Then the inverse image of
MRL with respect to the induced homomorphism SL ... RL is
(CP-1(M))SL. This is an L-rational maximal ideal of SL.

Proof. By Theorems 1 and 4, (cp

-1
(M))S
L is an L-rational maxi-
mal ideal of SL and it is obviously mapped into MRL. The corollary
follows.
We next turn our attention to rational maximal ideals in a tensor
product. To this end let M be a rational maximal ideal of R, N a
rational maximal ideal of S, and define [M, N] as in (1. 5.4). Then
[M, N] is an ideal of R ® S and, by (1. 5. 5), we have isomorphisms

(R ® S)![M, N]:::; (RiM) ® (SIN) :::; K ® K :::; K

of K-algebras. It follows that [M, N] is a rational maximal ideal of

R ® S. Thus there is a mapping

in which (M, N) is mapped into [M, N]. Moreover, by Chapter 1

Lemma 1, the mapping is an injection.
Now let I be a rational maximal ideal of R ® S and denote by M'
respectively N' the inverse image of I with respect to the canonical
homomorphism R'" R ® S respectively S ... R ® S. It follows (Theorem 1
that M' is a rational maximal ideal of Rand N' a rational maximal
ideal of S. Accordingly [M', N'] is a maximal ideal of R ® Sand
since [M', N'] ~ I we have [M', N'] = 1. This proves

Theorem 5. There is a bijection between 9J( K(R) x 9.R K(S) and

9J(K(R ®K S) which matches (M, N) of the former with [M, N] of the
latter.

16
2. 2 Function algebras

Let V be a set and as in section (1.3) denote by lJ K(V) the com-

mutative K-algebra formed by all the mappings of V into K. Now let
R be a subalgebra of the K-algebra lJ K(V), Thus R is a K-algebra
whose elements are K-valued functions on V. To describe this situation
we shall say that (V, R) is a function algebra over K. Note that such a
function algebra is a unitary, associative and commutative K-algebra,
and that it contains all constant K-valued functions with domain V.
For the rest of section (2. 2) it will be assumed that (V, R) is a
given function algebra over K. Let x E V. Then there exists a sur-
jective homomorphism R" K of K-algebras in which f .... f(x). Put

M = {f If E Rand f(x)
x
= 0 }, (2. 2. 1)

Then Mx is the kernel of the homomorphism and therefore it is a ---

maxi-
mal ideal of R. Indeed from the isomorphism RiMx :>: K we see that
---
Mx is a rational maximal ideal. Furthermore any f which belongs to
all the M (x E V) must vanish everywhere, and therefore it must be the
x
zero element of the algebra. This proves

Theorem 6. Every function algebra over K is rationally reduced

with respect to K.

We return to the consideration of (V, R). If x E V and fER, then

f - f(x) belongs to M I and therefore we have
x

f(x) = the M -residue of f. (2. 2. 2)

x
Next we have a mapping V .. 9Jl K(R) given by x .... Mx' Usually this is
neither an injection nor a surjection. However we do have

Lemma 1. Let x, y EV. Then the following statements are equi-

valent:
(a) M =F M .
x y'
(b) there exists fER such that f(x) = 0 and f(y) =F 0;
(c) there exists g E R such that g(y) =0 and g(x) =F O.

17
 Proof. Assume (a). Then M is not contained in M. Con-
x y
sequently there exists f M such that f t M. This shows that (a)
E
x y
implies (b) and similarly it implies (c). Obviously each of (b) and (c)
implies (a) so the proof is complete.
FunCtion algebras arise quite naturally. Suppose that S is some
given unitary, associative, and commutative K-algebra. For each I/J E S
we can define a mapping

by

¢(M) = the M-residue of I/J. (2.2.3)

The set of all I/J's forms a subalgebra S (say) of !J K(WlK(S)); that is to

say (WlK(S), S) is a function algebra over K. We shall call it the
derived function algebra of S. Note that we have a surjective homomor-
phism S - S of K-algebras in which I/J'" ¢. We shall speak of this as the
canonical homomorphism of S on to its derived function algebra.

Lemma 2. The kernel of the canonical homomorphism S - S is the

intersection of all the rational maximal ideals of S. Consequently the
canonical homomorphism is an isomorphism of K-algebras if and only if
S is rationally reduced.

This is clear because ¢ = 0 if and only if I/J E M for every M in

WlK(S). Note ~t if M is a rational maximal ideal of S, then MS is a
rational maximal ideal of S. In fact we have the

Corollary. There is a bijection between WlK(S) and Wl K(S) in

which M in Wl K(S) is matched with MS in Wl K(S)'

Note that when (WlK(S), S) is regarded as a function algebra and

M E WlK(S), then the corresponding rational maximal ideal of S in the
sense of (2.2.1) is MS. Thus, in this particular instance, we get a bi-
jection between the domain of the function algebra and the set of its rational
maximal ideals.

18
 Example. Consider the derived function algebra of
K[X , X , .•. , X ], where X , X , ... , X are indeterminates and
1 2 n 1 2 n
n ~
1. We have already seen, in section (2. 1), that there is a natural
bijection between Kn and the set of rational maximal ideals. If we now
identify these, then the derived function algebra may be written as
(K n, K[X 1 , ••• , Xn ])·
Suppose next that q,{X 1 , X 2 ' ••• , Xn) belongs to K[X 1 , X 2 '''' , Xn]
and let ~ be its image in K[X , X , •.. , X]. Then ¢: Kn .... K and,
1 2 n
by (2. 2. 3),

Thu~ K[X 1 , X 2 ' ••• , Xn] consists of the members of K[X 1 , X 2 ' ••• ,Xn ]
now regarded as functions from Kn to K. Observe that, by Lemma 2
and Theorem 3, the canonical homomorphism

is an isomorphism when and only when the field K is infinite.

2. 3 Loci and the associated topology

Throughout section (2. 3) (V, R) denotes a given function algebra

over K. As before we set

M
x
{f If E Rand f{x) = 0}
so that M is a K-rational maximal ideal of R. Now suppose that A is
x
a subset of R and put

CV(A) = {x Ix E V and f{x) =0 whenever f E A}. (2. 3. 1)

Thus CV{A) consists of the common zeros of the functions which make up
A. We call CV{A) the locus of the set A. Note that if AR denotes the
ideal generated by A, then

(2. 3. 2)

19
Also if A and A are subsets of R, then
1 2

A !:: A implies CV(A ) !:: CV(A ). (2. 3. 3)

1 2 2 1

Suppose that fER. It will be convenient to write CV(f) rather

than CV ( {f}). Thus

CV(f) = {xix EV and f(x) = 0 l. (2. 3.4)

A locus of this kind is sometimes called a principal locus. Since

CV(A) = n CV(f) (2. 3. 5)

fEA
it follows that every locus is an intersection of principal loci and vice
versa.
Let ~,lB be ideals of R and, in addition, let {~.}.
1 1E
I be a
family of ideals of the algebra. It is easily verified that

CV(O) = V and CV(l) = ¢

CV(~lB) = CV(~) U CV(lB), (2. 3. 6)

CV ( L ~ .) = n CV(~')'
iEI 1 iEI 1

This shows that V can be endowed with a topology in which the closed
sets are precisely the sets CV(~)' i. e. the closed sets in the topology
are the various loci. We shall refer to this as the R-topology on V.
Let ~ be an ideal of R. In preparation for the next set of obser-
vations we define Rady(~) as the intersection of all the rational maxima
ideals of the form M , where x E V, that contain ~. (If there are no
x
such maximal ideals containing ~, then it is to be understood that
Rady(~) = R.) Evidently

(2. 3. 7)

and

(2. 3. 8)

20
Also if 21 and 2I are ideals, then
I 2

2I I~2
2I implies RacL_(2I
-VI
) ~ RacL_(2I
-V2
). (2. 3. 9)

Lemma 3. Let 2I be an ideal of R. Then 21 = Ra<ly('lI) if and

only if some subfamily of {M } V has 2I as its intersection.
x XE -

This is clear from the definition of Ra<ly(2I) when it is taken in

conjunction with (2. 3. 7).
Once again let 2I be an ideal. Then

(2.3.10)

which shows at once that

(2. 3. 11)

Theorem 7. Let 2I and !B be ideals of R. Then CV (2I) = CV(!B)

if and only if Ra<ly(2I) = Ra<ly(!B).

If CV (2I) = CV(!B), then the relation Ra<ly(2I) = Ra<ly(!B)

Proof.
follows from (2.3.10). The converse is a consequence of (2.3.11). Note
that (2.3. ll) and Theorem 7 together show that Ra<ly(2I) is the largest
ideal with the same locus as 2I .
We shall now consider subsets of V and seek to find a useful criteri-
on for determining when such a subset is a locus. To this end suppose that
U ~ V and put

ly(U) = {f If E Rand f(x) = a whenever x E U }. (2.3.12)

Evidently ly(U) is an ideal of R. (If U is empty we put ly(U) equal

to R.) It will be called the associated ideal of the subset U. Now

ly(U) = n Mx
XEU
(2.3.13)

and therefore, by Lemma 3,

(2.3.14)

21
Again, by (2.3.13) and (2.3.10),

(2. 3. 15)

whenever III is an ideal of R, and from this follows

Theorem 8. Let III be an ideal of R. Then III = Ra<ly(Ill) if and

only if 1V(CV (Ill» =Ill.

Before proceeding to the next theorem let us note for future refer-
ence that if U ,U are subsets of V, then
I 2

U !:: U implies L(U ) !:: L(U ). (2.3. 16)

I2 -V2-VI

Theorem 9. Let V be endowed with the R-topology, let U be a

subset of V, and let U be the closure of U in V. Then

and therefore U is closed if and only if Cv (1V(U» = U. In any case

1V(U) = Iy(U).

Proof. Since Cv (1V(U» is closed and contains U, we have

U !:: Cv (1V(U». Further, because U is closed, U = CV(Ill) for some
ideal Ill. Now f(U) = {O} and hence f(U) = {O} for every f E Ill. Thus
III !:: 1V(U) and therefore

as required.
It remains for us to prove the final assertion. However
1V(U) = Iy(CV (ly(U» = Ra<ly(ly(U» by (2.3.15), and so 1V(U) = ly(U)
by (2. 3. 14).

Corollary. Let UI and U 2 be subsets of V. -

Then L(U )=L(U )
- - - y I -y 2
if and only if U = U •
I 2

The next theorem summarizes some of the most important con-

sequences of our discussion so far.

22
 Theorem 10. Let U denote a typical closed subset of V and III
a typical ideal of R satisfying III = Rady(Ill). Then there is a natural
bijection connecting the set of U's and the set of Ill's. This is such
that if U is matched with Ill, then ly(U) = III and CV(Ill} = U. More-
over the bijection reverses inclusion relations.

Example. Let X be an indeterminate. Then the derived function

algebra of K[X] is (K, K[X]) , where the notation is the same as that
employed in section (2. 2). The typical member of K[X] is a polynomial
g(X) considered as a function defined on K. Accordingly the correspon-
ding principal locus consists of the zeros of g(X) in K. This can be an
arbitrary finite set or, in the case where g(X) is the null polynomial,
the whole of X.
If S is any set, then there is a topology on S in which the closed
subsets are S itself and all its finite subsets. Let us call this the finite
sets topology on S. Since every locus is an intersection of prinCipal
loci, the remarks of the last paragraph show that the K[X]- topology on
K is preCisely the finite sets topology.
Once again suppose that (V, R) is a function algebra over K. Let
fER and let {k, k , ... , k } be a finite subset of K. Put
1 2 S
g = (f-k )(f-k ) ... (f-k). Then g E R and the inverse image of
1 2 s
{k , k , ... , k } with respect to f is the principal locus CV(g). This
1 2 s
is a closed subset of V when V has the R- topology.

Theorem 11. Let K be endowed with the finite sets topology.

Then the R-topology on V is the weakest topology in which the mappings
f : V .... K, that constitute R, are continuous.

Proof. Assume V has a topology in which every member of R

is continuous. The theorem will be proved if we show that the various
loci in V are closed in the given topology, and this will follow if we
establish that the typical principal locus CV(f), where fER, is closed.
But this is clear because CV(f) is the inverse image of the finite set
whose only member is the zero element of K.

23
2. 4 Affine sets

Let (V, R) be a function algebra over K, let x E V, and as

previously put M = {f /f
x
E Rand f(x) = oL We shall say that (V, R)
is an algebraic affine set defined over K or, when there is no need to
be quite so explicit, an affine set provided both the following conditions
are satisfied:
(a) the mapping V ~ 9J1 K (R) given by x 1-+ Mx is a bijection,
(b) as a K-algebra, R is finitely generated.
For the moment we shall postpone the justification of the choice of
language, and instead make an adjustment to the terminology. Suppose
that we have the above situation. We may then use R to impose a certain
structure on V. More precisely R may be employed to derive certain
distinguished subsets of V, and in the case where we have a second affine
set (W, S) to determine the significant mappings of V into W. In
practice we shall modify the language and say that V itself is an affine
set and we shall refer to R as its coordinate ring. The elements of V
are usually referred to as points and from now on the coordinate ring of
V will be denoted by K[V). Note that, as x varies in V, M ranges
x
over all the rational maximal ideals of K[V) giving each one once and
once only. The K[V)-topology on V will be called the affine topology.

Theorem 12. Let V be an affine set and let x E V. Then {x}

is a closed subset of V. Also if x, Y E V, then x"* y if and only if there
exists f E K[V) such that f(x) = 0, f(y) "* O.

Proof. By Lemma 1, CV(M x ) = {x}. This establishes the first

assertion and the second follows from the same lemma.
We add some further simple consequences of the definition of an
affine set.

Lemma 4. Let V be an affine set. Then there exists a finite set

u1 ' u 2 ' ••• , un of elements of the coordinate ring K[V) such that
K[V) = K[ u 1 , u 2 , ••. , u).
n
Hence there is a surjective homomorphism

24
of K-algebras, where Xl' X 2 , ••• , Xn are indeterminates that are
mapped into u , u , ... , u respectively.
1 2 n

Proof. By the definition of an affine set, K[V] is a finitely genera-

ted K-algebra. The lemma follows by taking u 1 ' u 2 ' ••• , un to be any
system of generators.

Theorem 13. Let V be an affine set and let f1 , f 2' ••• , fr belong
to K[V]. Then the following statements are equivalent:
(a) f, f , ... , f are linearly independent over K;
1 2 r
(b) in V ~su.::.c.::.h_th,-a_t_t-=h-,-e_d,-,e_te-,-r_m,---ina_n_t
there exist y 1 , y 2 , .•. , yr - - .
If. I
(y.) is non-zero.
1 J

Proof. Assume (a). We can certainly find y 1 so that f1 (y 1) *- o.

We can then find y with the property that
2

f1 (y 1) f1 (y 2)

f 2(Y1) f 2(y)

And so on. Thus (a) implies (b) and the converse is obvious.
Let V be an affine set with coordinate ring K[V]. If now III is
an ideal of K[V], then Rarly (Ill) , as defined in section (2.3), is the inter-
section of all the rational maximal ideals of K[V] that contain III .
This is a convenient point at which to draw attention to what happens
if K should be algebraically closed. Suppose that this is the case and let
X1 , X 2 , •.. , Xn be indeterminates. We have already observed that in
the situation under consideration every maximal ideal of K[X 1, X 2 , ••• ,X n]
is K-rational. But, whatever the field K, the ring K[X , X , ... , X ]
1 2 n
has the property that for any ideal III the intersection of all the maximal
ideals containing III is Rad III , t i. e. the set of elements for which some
power belongs to the ideal. A ring with this property is sometimes called
a Hilbert ring because of the intimate connection with Hilbert's Zeros
Theorem. The same rings are also known as Jacobson rings because they
are characterized by the fact that in every homomorphic image the nil-

t See, for example, [(9) Theorem 1, Cor. 1, p. 279].

25
radical is the same as the Jacobson radical. Because of the geometrical
association we shall use the former name. In fact the general theory of
these rings was discovered independently by O. Goldman (5) and W. Krull
(7).
It is clear that any homomorphic image of a Hilbert ring is itself
a Hilbert ring. Consequently, in view of Lemma 4, if V is an affine set
defined over an algebraically closed field. K, then K[V] is a Hilbert ring
and all its maximal ideals are K-rational. Consequently for any ideal III ,
of K[V], Ra~(Ill) = Rad(Ill). The next theorem also illustrates a special
feature pertaining to the case where the ground field is algebraically
closed.

Theorem 14. Let V be an affine set defined over an algebraically

closed field K and let f E K[V]. Then f is a unit in the algebra K[V]
if and only if f does not vanish at any point of V.

Proof. To say that f does not vanish at any point of V is equi-

valent to saying that f does not belong to any of the ideals M (x E V).
x
But, because K is algebraically closed, this is equivalent to saying that
f is not in any maximal ideal and so the theorem follows.
For the remainder of Chapter 2 we revert to our original assumption
namely that K is an arbitrary field. In the next theorem S denotes a
unitary, associative and commutative K-algebra.

Theorem 15. Suppose that S is finitely generated as a K-algebra

and let (9Jl K (S), S) be its derived function algebra. Then 9Jl K (S) is an
affine set (defined over K) having S as its coordinate ring.

Proof. Since S is a homomorphic image of S it is finitely genera-

ted as a K-algebra. Consequently we have only to establish that there is
a bijection, of the right kind, between 9Jl K(S) and the set of K-rational
maximal ideals of S. However this has already been done in the discussiOl
which follows Lemma 2.
We shall now characterize (up to isomorphism) the K-algebras which
are the coordinate rings of affine sets. To this end we make the

26
 Definition. An 'affine K-algebra' is a K-algebra that is isomorphic
to the coordinate ring of some affine set defined over K.

Theorem 16. Let S be a unitary, associative and commutative

K-algebra. Then the following statements are equivalent:
(a) S is finitely generated (as a K-algebra) and is rationally reduced;
(b) S is an affine K-algebra.

Proof. Assume (a) and let us use the notation of Theorem 15. By
Lemma 2, Sand S are isomorphic K-algebras and, by Theorem 15, S
is an affine algebra. This showS that (a) implies (b). The converse, of
course, is trivial.

Corollary. The polynomial ring K[X 1 , X 2 , ••• , Xn] is an affine

K-algebra if and only if K is an infinite field.

This is now immediate in view of Theorem 3.

Let Y be an affine set and U an arbitrary subset of Y. By re-
striction each f in K[Y] induces a mapping f* : U - K, and the set of
all these restrictions forms a K-algebra S say. In fact (U, S) is a
function algebra over K, the mapping K[Y] - S in which f ~ f* is a
surjective homomorphism of K-algebras, and the kernel of the homomor-
phism is ly(U).
Let y E Y. By (2.3.10), ly(U) !: My if a~ only if y E Cy(ly(U));
also, by Theorem 9, Cy(ly(U)) is the closure, U, of U in Y. It follows
as y varies in U, M S ranges over the K-rational maximal ideals
that ~y
of S giving rise to each one exactly once.
Since S is a homomorphic image of K[Y], it is a finitely generated
K-algebra.Let x E U. Then

{f*/f EK[Y] and f*(x) = 0)

is just M S. Thus different points of U give rise to different K-rational
x
maximal ideals of S, but unless U = U not all the rational maximal
ideals of S are obtained. This proves

27
 Theorem 17. Let V be an affine set and U a subset of V. Let
S be the K-algebra consisting of the restrictions, to U, of the functions
which make up K[V]. Then in order that U should be an affine set with
S as its coordinate ring it is necessary and sufficient that U be a closed
subset of V.

From now on, whenever we have a closed subset of an affine set we

shall regard the subset itself as being an affine set with the coordinate
ring as described in Theorem 17.

Corollary. Let U be a closed subset of the affine set V. Then U

can be regarded as an affine set and as such it has an affine topology. This
coincides with the subspace topology.

The assertion is clear because the principal loci on U are simply

the contractions of the principal loci on V.

2. 5 Examples of affine sets

In this section we assemble some examples of affine sets. They will

all be defined over the field K. We recall that if V is a set, then
trK(V) denotes the K-algebra formed by all K-valued functions on V.

Example 1. Let V be an affine set with only a finite number of

points. Let these be denoted by x , x , .•. , x . Suppose 2::5 i ::5 m.
12m
By Theorem 12, there exists f.1 E K[Vl- such that f.1(x 1 ) '* 0, f.1 (x.)
1
= o.
Put g = cf f .•. f ,where c E K. Then if c is suitably chosen we
1 23m
have gl (xl) =1 and gl (Xi) =0 for 2::5 i ~ m. Of course gl E K[V].
In fact we can find g , g , •.. , g in K[V] so that g.(x.) = 0 ..•
12m 1] 1]
Finally let k 1 , k 2 , ••• , k m belong to K and put
g = k g + k g + ... + k g . Then g(x.) = k. for 1 ~ i ~ m and we
11 22 mm 1 1
have proved

Theorem 18. Let V be a finite affine set. Then K[V] = tr K(V).

Conversely if V consists of m distinct elements x , x , ... , X
12m'
then V is an affine set with tr K(V) as its coordinate ring. For put

28
 M.= {flfEjJK(V) and f(x.)=O}
1 1

Then M , M , ... , M are distinct rational maximal ideals of jJ K(V),

12m
Further if M is a rational maximal ideal, then, because M1 M2 ••• Mm=O,
we have M = Mi for some i. Of course, jJ K(V) is a finitely generated
K-algebra.

Example 2. Let V be an n-dimensional vector space over K and

let V* = HomK(V, K). Thus V* is the K-space of linear forms on V or
(as it is sometimes known) the dual space. Let ~ E V*, i. e. let ~ be a
typical linear form. Since ~: V - K it belongs to jJ K(V), Denote by
K[V] the subalgebra of the K-algebra jJ K(V) which the linear forms
generate. We claim that V is an affine set with coordinate ring K[V].
To see this let e , e , ..• , e be a base for the K-space V. Then
1 2 n
V* has a base ~ , ~ , .•• , ~ ,where Ue.) = 0... This is the so-
l 2 n 1 J 1J
called dual base. Evidently K[V] = K[ ~ , ~ , •.. , ~ ].
1 2 n
Let X , X , ..• , X be indeterminates. We have a bijection
1 2 n
between Kn and V in which (a , a , ••• , a ) in Kn is matched with
1 2 n
a e
11
+ a 22
e + ... + a e. Let cp(X , X , •.. , X ) belong to
nn 1 2 n
K[X 1 , X 2 , ••• , Xn] and put ~ = CP(~l' ~2' ••• , ~n) so that ~ is a typical
member of K[V]. If now v = a 1 e 1 + a 2 e 2 + ... + ane n belongs to V
and we identify V with Kn, then ~(v) = cp(a , a , ••• , a). Thus
1 2 n
~ : V - K is just cp(X 1 , X 2 , •.• , Xn ) considered as a function from
n
K to V. Accordingly (V, K[V]) has been identified with the derived
function algebra of K[X , X , ••• , X]. Consequently our claim follows
1 2 n
from Theorem 15.
The example at the end of section (2. 2) shows that the mapping

in which cp(X , ... , X ) 1-+ cp( ~1' ••• , ~ ) is an isomorphism of K-

1 n n
algebras if and only if K is infinite. We therefore obtain

Theorem 19. Let V be an n-dimensional vector space over K

and ~ , •.• , ~n
, ~any base of V* = Hoffi...(V, K). -
Then-V-is -
an
--12 · K
affine set with K[ ~ , ~ , ••• , ~ ] as its coordinate ring. Moreover
1 2 n

29
~ , ~ , ... , ~ are algebraically independent over K if and only if K
1 2 n
is infinite. If K is finite, then V is a finite set and
K[ ~l' ~2' ••• , ~n] = 'd K(V).
The members of K[V] are often referred to as polynomial functions
on V.

Example 3.
This is essentially the same as Example 2 but the
restatement involved helps to explain our terminology. Kn is an n-
dimensional vector space over K and it has the natural base
e , e , ..• , e , where e. = (0, ••. , 1, ••. , 0) with the 1 occurring
1 2 n 1
in the i-th position. Let X. : Kn ... K be the i-th coordinate function so
1
that

x.(a , a , .•• , a )
1 1 2 n
= a 1..
Then X , X , ••. , X is the base of HomK(Kn , K) that is dual to
1 2 n
e , e , ... , e . Accordingly Kn is an affine set with K[X ,X , •.. , X ]
1 2 n 1 2 n
as its coordinate ring. We refer to this affine set as affine n-space. Of
course

where the right hand side consists of the members of K[X 1 , X 2 , ••• ,Xn ]
regarded as functions from Kn to K, and the natural homomorphism

K[X 1 , X 2 , ••• , X ] ... K[X , X , ... , X ]

n 1 2 n

maps X. into X ..
I 1
Let </l(X 1 , X 2 ' ••• , Xn) f K[X1 , X 2 ' ••• , Xn] and let ¢ be the
corresponding member of the coordinate ring of Kn. The principal locus
determined by ¢ is composed of the points (a , a , .•• , a ) of Kn
1 2 n
that are zeros of </l(X , X , •.• , X). It follows that the closed subsets
1 2 n
of affine n-space are just those loci that are defined by algebraic equations
in the manner that is familiar from elementary coordinate geometry. Of
course when K is infinite we can regard K[X , X , •.. , X ] itself as
1 2 n
the coordinate ring of Kn.

30
 Example 4. This is an important example of a way in which affine
sets often arise. Let V be a given affine set and let f E K[V]. Put

U = {xIx EV and f(x) *- oJ,

and observe that, because U is the complement of a principal locus, it

is an open subset of V.
Each g E K[V] determines by restriction a mapping g*: U ~ K.
If we denote the set of these restrictions by S, then S is a subalgebra of
the K-algebra ~ K(U), and the mapping g 1-+ g* is a surjective homomor-
phism K[V] ~ S of K-algebras. Evidently (U, S) is a function algebra
over K but, in view 01 Theorem 17, it need not be an affine set even
though, as is clear, S is a finitely generated K-algebra.
The restriction f*, of f, vanishes nowhere on U and so we have
a mapping U ~ K in which x 1-+ l/f(x). This mapping will be denoted by
1 /f*. The smallest subalgebra of ~ K(U) containing S and this new
function is S[l/f*]. This too is finitely generated as a K-algebra. Also
(U, S[l/f*]) is a function algebra over K.

Theorem 20. Let the notation be as above. Then U is an affine

set with S[l/f*] as its coordinate ring.

Proof. Let x, y E U with x *- y. By Theorem 12, there exists

g E K[V] such that g(x) = 0, g(y) *- O. The restriction g*, of g to U,
belongs to S[l/f*] and, of course, g*(x) = 0, g*(y) *- O. This shows that
different elements of U give rise to different maximal ideals of S[l/f*].
Let N be a rational maximal ideal of S[l/f*]. It follows (Theorem
1) that N n S is a rational maximal ideal of S and therefore there exists
z E V such that M S = N n S. Since f* is a unit in S[l/f*], we see that
z
f* ¢ N and therefore f ¢ M • This shows that Z E U.
z
Now assume that l/I E N. For some v > 0 we have (f*) l/I EMS. " Z
Consequently l/I(z) = 0 and therefore

N k {w IW E S[l/f*] and w(z) = 0 }.

However there must be equality because N is a maximal ideal. Thus we

have the desired bijection between U and the set of rational maximal

31
ideals of 8[I/f*]. The theorem is therefore proved.
An important simplification takes place when f is a non-zero-
divisor in K[V]. Assume that

g E Ker(K[V] ... 8).

Then fg vanishes everywhere on V and hence g = O. Thus K[V]'" 8

is an isomorphism of K-algebras and it leads to an isomorphism

K[V] UJ ~ 8 U~]
on the basis of which we can make the identification

K[U] = K[V] UJ . (2. 5. 1)

It is clear how the members of the right hand side are to be regarded as
K-valued functions defined on U. We also see that if K[V] is an integral
domain and f *- 0, then K[U] is also an integral domain.

Example 5. Let V be an n-dimensional vector space over K and

put E = En%(V) so that E consists of all the K-linear mappings of V
into itself. t E is a unitary, associative K algebra. In particular it is
a vector space over K and as such its dimension is n 2 • Accordingly,
by Example 2, there is a natural way in which E can be regarded as an
affine set.
We now define a mapping

(2. 5. 2)

by means of

D(g) = the determinant of g (2. 5. 3)

for each gEE. Clearly D belongs to the coordinate ring, K[E], of E.

We shall apply the construction described in Example 4 to this situation.
To this end we put

t 8ee section (1. 3) Example 3.

32
 GL(V) = {g Ig E E and D(g)"* 0 l. (2. 5. 4)

Then GL(V) consists of the units of the algebra En~(V) and therefore
it has a group structure. It is usually referred to as the general linear
group of V.
For the moment our main concern is to note that the considerations
put forward under Example 4 enable us to regard- GL(V) as an affine set.
In fact if S is the K-algebra obtained by restricting the members of
K[E] to GL(V), then the coordinate ring of GL(V) is given by

K[GL(V)] = S [~~ (2. 5. 5)

where D* denotes the restriction of D.

For the remainder of section (2.5) we shall assume that K is an
infinite field. Theorem 19 then shows that K[E] is an integral domain
and therefore, by (2.5.1), we may write

K[GL(V)] = K[E] [~J . (2. 5. 6)

Indeed we can find a base f.. (1:s i :s n, 1 :S j :S n) for E with the follow-
1]
ing property: if f = ~~a. J .. , where a .. E K, then D(f) is the deter-
1~ IJ IJ -- .
minant of the matrix II a .. II. This enables us to identify K[E] with
IJ
K[X ,X
, •.. , X ], where the X .. are indeterminates, and then D
11 12 nn IJ
in K[E] becomes identified with the polynomial Det Ilx .. 11. Thus to sum
IJ
up: when K is an infinite field

K[GL(V)]= K[X 1 l' X12 , ... , Xnn ] [ Det Xij IIJ II (2. 5. 7)

In this context if f = ~~a. J.. belongs to GL(V) and we wish to evaluate

1) IJ
a member of the right hand side of (2.5. 7) at f, then we do this by
assigning to Xij the value air

2.6 Morphisms of affine sets

Throughout section (2. 6) we shall use V and W to denote given

affine sets that are defined over K. Let x EV and yEW. It will be
convenient to use M respectively N to denote the corresponding
x y

33
rational maximal ideal of K[V] respectively K[W].
Suppose that we have a mapping cp: V - W. For g E K[W] put

cp*(g) = g 0 cpo (2. 6. 1)

Then cp*(g) : V - K and, with a self-explanatory notation,

CP*(gl + g2) = cp*(g) + CP*(g2)'

CP*(gl g) = CP*(gl).p*(g2)'
cp*(kg) = klJl*(g).

Of course ¢* preserves constant mappings.

Definition. The mapping 1Jl: V -W is called a 'K-morphism' of

affine sets if cp*(g) E K[V] for all g E K[W].

Suppose that cp: V -W is a K-morphism. This induces a homo-

morphism

1Jl* : K[W] - K[V] (2. 6. 2)

of K-algebras in which g 1-+ g 0 cpo It follows that if x E V, then

(2. 6. 3)

if cp ,cp : V -Ware K-morphisms and cp* = cp*, then

Consequently - 12 - 12--
CP1 = CP2'
On the other hand suppose that we start with a homomorphism
w : K[W] - K[V] of K-algebras. By Theorem I and the definition of an
affine set, there is a uniquely determined mapping 1/1: V - W such that

for all x E V. If now gK[W], then w(g - (w(g))(x)) belongs to M and

E
x
therefore g - (w(g))(x) is in NI/I(x)' Thus

(I/I*(g)(x) = g(l/I(x)) = (w(g)(x)

from which it follows that I/I*(g) = w(g) E K[V]. We now see that 1/1: V-W

34
is a K- morphism and 1/1* = w. Moreover by combining our observations
we obtain the important

Theorem 21. There is a natural bijection between the K-morphisms

V - W and the homomorphisms K[W] - K[V] of K-algebras. If
!/J : V -W is a K-morphism of affine sets, then the algebra homomorphism
associated with it is !/J*: K[W] .... K[V], where this is defined in the
manner explained above.

We shall make a fairly deep study of K-morphisms at a later stage.

For the moment we shall content ourselves with some simple observations.
First of all the identity mapping of V is a K- morphism and it is associa-
ted with the identity homomorphism of K[V]. Next suppose that U is a
third affine set and that If> : V - Wand X: W - U are K-morphisms. We
observe, for future reference, that X 0 !/J : V - U is a K-morphism and
that

(X 0 !/J) * = !/J* 0 X*. (2.6.4)

Our aim is to study affine sets and their morphisms. It will now be
clear that this is virtually equivalent to studying affine algebras and their
homomorphisnis. Thus we could, if we wished, restrict our attention to
the study of algebras. This indeed would have substantial advantages in-
cluding the possibility of achieving great generality. However we shall
not do this because it would sever the connections of the subject with
geometry and, in a certain sense, it would stand the subject on its head.
During the twentieth century Geometry has been the victim of a takeover
bid by Algebra and a very useful way of looking at certain questions has
been lost to some mathematicians. For once generality will be asked to
give way to historical considerations.
Returning to the main discussion, let !/J: V -W be a K-morphism
of affine sets.

Definition. We say that !/J is a 'K-isomorphism' of V on to W if

it is a bijection and has the property that !/J -1 : W - V is als 0 a K-
morphism.

35
 Evidently 1/>: V - W is a K-isomorphism if and only if
1/>* : K[W] - K[V] is an isomorphism of K-algebras. Furthermore
when this is the case we have

Example. It is important to note that a K-morphism can be a

bijection without being a K-isomorphism. The following example illus-
trates this point.
Let K be an algebraically closed field of characteristic p, where
p is a prime. Since K is infinite, if we regard K itself as affine
I-space, then its coordinate ring may be taken to be K[X], where X is
an indeterminate. Define I/> : K - K by I/>(a) = a P• This is a K-morphism
and 1/>* : K[X] - K[X] maps g(X) E K[X] into g(XP). Consequently I/>
is not a K-isomorphism. However it certainly is bijective.
The next theorem provides the justification for the name affine set.
Before proceeding to prove it we observe that any bijection between two
finite affine sets is a K-isomorphism. This follows from Theorem 18.

Theorem 22. Let V be an affine set defined over K. If n is a

sufficiently large positive integer, then V is K- isomorphic to a closed
subset of Kn.

Proof. By Lemma 4, we can construct a surjective homomorphism

(2. 6. 5)

of K-algebras, where Xl' X , ..• , X are indeterminates. If K is a

2 n
finite field, then K[X , X , ••• , X ], and hence also K[V], has only a
I 2 n
finite number of rational maximal ideals, and therefore V has only a
finite number of points. But in this case what we have to prove is trivial.
From here on we assume that K is infinite. Let III be the kernel
of (2. 6. 5) and U the locust of Ill. Then U is a closed subset of Kn.
Because K[V] is rationally reduced, III is the intersection of all the

t We regard K[X , X , ... , X ] as the coordinate ring of Kn.

I 2 n

36
rational maximal ideals that contain it. Accordingly, by (2.3.15),
I (U) = III and therefore
Kn

these being isomorphisms of K-algebras. It follows U and V are K-

isomorphic.

Corollary. t .!!. K[V] can be generated, as a K-algebra, by

n 2: 1 elements, then n is a large enough integer for the purposes of
the theorem.

The next result shows that K-morphisms are continuous.

Theorem 23. Let cp: V -+ W be a K-morphism of affine sets and

let each of V and W be endowed with its affine topology. Then cp is
continuous.

Proof. Let g E K[W] and put

C = {y lYE Wand g(y) = 0 }.

Since every closed subset of W is an intersection of principal loci, it

will suffice to show that cp-l(C) is a principal locus in V. But this is
clear because, if f =g 0 cp, then f E K[V] and

¢-l(C) = {xix E V and f(x) = O}.

2. 7 Products of function algebras

Throughout section (2. 7) we shall use (V, R) and rN, S) to denote

function algebras over K. Let fER, g E S and define

fvg:VXW-+K (2. 7. 1)

by

(f v g) (x, y) = f(x)g(y). (2. 7. 2)

t This holds for arbitrary K.

37
The functions f v g will generate a subalgebra of the K-algebra
!j K(V x W). This subalgebra will be denoted by R x K S. Accordingly
(V x W, R x K S) is a new function algebra over K. It will be called the
product of (V, R) and 0N, S). Note that a typical element of R XKS
can be written in the form

f v g + f v g + •.• +f vg,
1 1 2 2 IP m

where f. E Rand g. E S. Note also that

1 1

(f v g)(f v g ) = f f v g g . (2. 7. 3)
1 12 2 12 12

Again it is easy to check that there is a K-linear mapping

in which f ® g ~ f v g, and then a further simple verification shows this

to be a surjective homomorphism of K-algebras. However much more is
true as is shown by

Theorem 24. The K-linear mapping R ®K S -+ R x K S, in which

f ® g is mapped into f v g, is an isomorphism of K-algebras.

Proof. Let f ® g + f ® g +... + f ® g belong to the kerne

1 1 22m m
of the mapping. We have to show that this is zero. Without loss of
generality we may assume that gl' g2' .•• , gm are linearly independent
over K.
Let x V and put g = f (x)g + f (x)g +... + f (x)g • Then g
E
1 1 22m m
vanishes identically on Wand therefore .

It follows that f.(x)

1
=0 for i = 1, 2, ••• , m. This shows that
f , f , ••. , f are all zero and the theorem is established.
12m
When it is convenient we shall use Theorem 24 to identify R ®K S
with R x K S. Let III be an ideal of Rand 5B an ideal of S. Then t

t See (1. 5.4).

38
[Ill, Q3] is an ideal of R ®K S = R xK S. In fact a typical element of
[Ill, Q3] can be written in the form

f v g + f v g + ... +f vg,
1 1 22m m

where for each i (1 ::::: i ::::: m) either f. E III and g. E S or f. E Rand

1 1 1
gi EQ3.
Let x E V and yEW. The former of these determines a rational
maximal ideal M of R and the latter a rational maximal ideal N of
x Y
S. By Theorem 5, [M , N ] is a rational maximal ideal of
x y
R ®K S = R x K S. Suppose that I/> E [Mx' Ny] k R ®K S. The remarks
of the last paragraph show that I/> (x, y) = O. Accordingly we obtain

Lemma 5. Let the notation be as above. Then

[M , N ] = {I/> II/> E R x K Sand I/> (x, y) =0}

x y

for all x E V and yEW.

We concluje this section by applying our observations to prove

Theorem 25. Let R' and S' be rationally reduced K-algebras. t

Then R' ®K S' is also a rationally reduced K-algebra.

Proof. Let (V, R) respectively f:N, S) be the derived function

algebra of R' respectively S'. By Lemma 2, R' is isomorphic to R
and S' is isomorphic to S. Consequently, by Theorem 24, we have
isomorphisms

R' ®K
S' '" RK Rx S
® S""K

of K-algebras, and R xK S is rationally reduced because (V x W, R xK S)

is a function algebra (see Theorem 6).

t It is assumed that R' and S' are unitary, associative, and commuta-
tive.

39
2. 8 Products of affine sets

Throughout section (2. 8) the letters V and W denote given affine

sets defined over K. Also if x E V and yEW, then M
x and Ny
denote the corresponding rational maximal ideals of K[V] and K[W]
respective ly.
By Theorem 24, there is an isomorphism

K[V] ®K K[W] "" K[V) x K K[W) (2. 8. 1)

of K-algebras in which f ® g is matched with f v g. Let us use (2. 8. 1)

to identify the two algebras. Then, by Lemma 5,

[M , N ] = {eplep E K[V] x K K[W) and I/> (x, y) = O}. (2. 8. 2)

x y
Also, by section (1. 5), K[V) x K K[W] is a finitely generated K-algebra.
The next result is now an immediate consequence of Theorem 5 and the
fact that V and Ware affine sets.

Theorem 26. Let V and W be affine sets defined over K. Then

the cartesian product V x W has a natural structure as an affine set with
K[V x W] = K[V] x K K[W].

Since V x W is an affine set it has an affine topology. It is im-

portant to note that this is not the product of the affine topology of V with
that of W. This is illustrated by the following example.
Let K be an infinite field and X, Y indeterminates. Take V = K,
W = K and identify their coordinate rings with K[X) and K[Y). The
coordinate ring of K x K is then K[X) ® K[Y) = K[X, Y) which shows
that the product of K with itself is the same as affine 2-space. Put

I
A = {(x, y) (x, y) E K x K and x = y }.

Then .Il is a closed subset of K x K when K x K has its affine topology.

Assume that 0 and 0 are non-empty open subsets of K. Then,
1 2
as we saw in section (2.3), their complements are finite. Consequently
there exists a E K such that (a, a) EO X O. Accordingly when K x K
1 2
is endowed with the product topology every non-empty open set meets A.

40
Hence the product topology on K x K is different from the affine
topology.
We continue now with the general discussion. Let A be a cbsed
subset of V and B a closed subset of W. By identifying K[V x W]
with K[V] ® K[W] we can regard [Iy(A), lw{B)] as an ideal of the
former. The locus of this ideal is A x B and therefore A x B is a_
closed subset of V x W. Note that if P and Q are open subsets of V
and W respectively, then P x Wand V x Q are open subsets of
V x W because they are the complements of (V \P) x Wand V x (W \Q)
respectively. Accordingly (P x W) n (V x Q), that is to say P x Q, is
an open subset of V x W.
Since A x B is a closed subset of V x W we have two ways of
regarding A x B as an affine set. First of all we may regard A and
B as affine sets and then A x B is an affine set in its own right. Secondly
A x B is a closed subset of V x Wand as such inherits an affine
structure. However, as the next theorem shows, there is no danger of
ambiguity.

Theorem 27. Let A be a closed subset of the affine set V and B

a closed subset of the affine set W. Then the two ways of re~ding
A x B as an affine set (see above) yield the same coordinate ring on
A x B.

Proof. Let f i , f 2, .•. , fs belong to K[V] and gi' g2' ... , gs

to K[W]. Denote by f~ respectively g~ the restriction of f. respec-
1 1 1
tively g. to A respectively B. The theorem now follows from the
1
observation that

v g* + f* v g* +
f*1 1 2 2 · · · + f*S v g*S

is the restriction of f v g +f v g + ... + f v g to A x B.

1 1 2 2 S S
So far we have only considered products of two affine sets but the
concept extends easily to arbitrary finite products. Since no essentially
new ideas are involved, we shall deal with the extension rather briefly.
Let V , V , ... , V be affine sets defined over K and let
1 2 S
f.
1
€ K[V.] for i
1
= 1, 2, ... , s. From these we construct a function

41
 v1 xv 2 x ... xv s -+K
which is denoted by f1 v f2 v ••• v fs and which is such that

(x , x , ... , x ) 1-+ f (x )f (x ) . .. f (x ).
12 S 1122 SS

These functions generate a subalgebra of the K-algebra

trK(V xV x... x V). Denote the subalgebra by
1 2 S

An easy argument using induction now shows that V x V x... x V

1 2 S
may be regarded as an affine set whose coordinate ring is given by

It should be noted that, when convenient,

VI XV 2X... xV and (V xV x ... xV ) x (V xV x XV)

S 12m m+l m+2'" s
may be identified in the obvious way. Also if h, i , ... , i } is a
1 2 S
permutation of {I, 2, ... , s}, then the canonical bijection

V1 X V 2 X ... x Vs =.. ViI x Vi2 X .•• x Vis

is a K-isomorphism.

2.9 Some standard morphisms

Certain operations with affine sets lead to mappings and it is im-

portant to know when these are morphisms. In this section we shall make
a useful collection of results concerning such situations. To avoid repe-
tition we record at the outset that all affine sets considered in section
(2. 9) are to be understood as being defined over K. It will be remembered
from section (2.6) that an identity mapping of an affine set is a K-isomor-
phism and that the result of applying two K-morphisms in succession is
another K-morphism.

42
 Lemma 6. Let V and W be affine sets and suppose that V is
finite. Then every mapping !fi: V ... W is a K-morphism.

This follows from Theorem 18.

Lemma 7. Let A be a closed subset of the affine set V and let

j : A - V be the inclusion mapping. Then j is a K-morphism and the
associated homomorphism j* : K[V] ... K[A] ~ K-algebras is surjective.

This is clear because we obtain K[A] by taking the functions which

make up K[V] and restricting their common domain to A.

Lemma 8. Let!fi: V"'W be a K-morphism of affine sets and let

B be a closed subset of W. Suppose that !fi(V) ~ B. Then the induced
mapping V'" B is also a K-morphism of affine sets.

For if f E K[W] restricts to f* on B, then the mapping x 1-+ f*(!fi(x))

of V into K is just f 0 !fi and this belongs to K[V].
Next suppose that !fi : V ... V' and 1/1: W ... W' are mappings and
define

!fi x 1/1 : V x W ... V' x W' (2. 9. 1)

by

(!fi x I/I)(x, y) = (!fi(x), 1/1 (y)). (2. 9. 2)

Lemma 9. Let!fi: V "'V' and 1/1 : W -W' be K-morphisms of

affine sets. Then

!fi x 1/1 : V x W ... V' x W'

is a K-morphism as well.

This is clear because if f' E K[V'] and g' E K[W'], then

(f' v g') 0 (!fi x 1/1) = (f' 0 !fi) v (g' 0 1/1).

Lemma 10. Let V and W be affine sets. Then the projections

1TI : V x W- V -
and
- 1T 2 : V x W- W

43
ar~ K-morphisms of affine sets.

Proof. We need only consider 11 1 • Let f E K[V]. Then

f 0 11 = f v I E K[V x w]
1

and therefore 111 is a K- morphism.

For Lemma 11 it will be assumed that V, W, W' are affine sets
and that ¢: V -W, ¢' : V -W' are mappings. Define Ij;: V -W x W'
by Ij;(x) = (¢(x), I/>'(x)).

Lemma 11. With the above notation 1/1: V -W x W' is a K-

morphism if and only if ¢ and ¢' are K-morphisms.

Proof. Suppose that I/> and ¢' are K-morphisms and assume that
g E K[W], g' E K[W']. The fact that 1/1 is a K-morphism follows from the
relation

(g v g') 0 1/1 = (g 0 ¢)(g' 0 1/>').

On the other hand if Ij; is a K-morphism, then, by combining Ij; with

the projections V x W - V and V x W -+ Wand using Lemma 10, we
see that ¢ and cp' are K-morphisms.
Next suppose that x E V, yEW and define

a :V-VxW (2. 9. 3)
y
T :W-VxW (2.9.4)
x

by a (x) = (x, y) and T (y) = (x, y).

y x
Lemma 12. Let V and W be affine sets and suppose that x E V,
YEW. Then,- with the above notation,- ayand Tare K-morphisms.
-- x-

Proof. Let w E K[V x W] say

w =1 f v g
1
+f
2
v g2
+· •• +fvg
S s'

where f. E K[V] and g. E K[W]. Then

1 1

44
 w 0 T = f (x)g + f (x)g + ... + f (x)g E K[W]
XII 2 2 S S

which shows that T is a K-morphism. The proof that a is a K-

x y
morphism is similar.
Let x belong to the affine set V and y to the affine set W. Then
there is a bijection V'" V x {y} in which ~ in V is matched with
(~, y) in V x {y} and a bijection W", {x} x W which arises similarly.
By Theorem 12, {x} is a closed subset of V and {y} a closed sub-
set of W. Moreover V x {y} and {x} x Ware affine sets. t

Theorem 28. Let V and W be affine sets and suppose that

x E V, yEW. Then the natural bijections V'" V x {y} and
W ", {x} x W are K-isomorphisms of affine sets.

Proof. By Lemma 12, a : V -V x W is a K-morphism and, of

y
course, V x {y} is a closed subset of V x W. Lemma 8 therefore shows
that the bijection V - V x {y} is a K-morphism. However the inverse
mapping V x {y} - V is a K-morphism because it is a projection.
Accordingly V -V x {y} is a K-isomorphism. The other assertion is
proved similarly.

Corollary. Let V, W be affine sets and let A respectively B be

a subset of V respectively W. If now A, B are the closures of A, B
in V, W respectively, then the closure, A x B say, of Ax B in
V x W is A x B.

Proof. Let b E B. By Theorem 28, A x {b} is the closure of

A x {b} in V x {b}, Since (V x {b}) nA x B contains A x {b},
we see that A x {b} ~ A x B and therefore A x B <;;; A x B. Accordingly
A x B = A x B and likewise A x 13 = A x B. In the last equation replace
A by A. This yields

t In this connection the reader should recall Theorem 27.

45
as required.
We mention, in passing, a K-morphism which expresses the fact
that an affine set V is, to a certain extent, separated.
Define I): V -V x V by I)(x) = (x, x). This is a K-morphism by
Lemma 11. The image of I) is a subset, d say, of V x V called its
diagonal. By Theorem 12, this subset consists of the common zeros of
the functions f v 1- 1 v f obtained by varying f in K[V]. Thus d

is a closed subset of V xV. Also I) induces a K-isomorphism, of V

on to d, whose inverse is obtained by combining the inclusion mapping
d - V x V with a projection.
The next result completes our collection of useful morphisms. It
will be recalled that, in section (2. 5), we showed that if V is a vector
space over K of finite dimension, then V has a natural structure as
an affine set.

Lemma 13. Let V and V' be finite-dimensional vector spaces

over K and let 1/>: V - V' be a K-linear mapping. If now V and V'
are regarded as affine sets, then I/> is also a K- morphism of such sets.

This is clear in view of the way in which polynomial functions, on V

and V', are defined.

2. 10 Enlargement of the ground field

In this section we shall examine the effect on an affine set of enlar-

ging the ground field. Accordingly, throughout section (2.10), V and W
will denote affine sets defined over K, and L will denote an extension
field of K.
Suppose that f E K[V]. Then f : V - K and therefore f determines
a mapping of V into L. (This mapping will be denoted by the same
letter.) Thus K[V] ~ jjL (V). Denote by S the subalgebra of the L-
algebra jj L (V) that K[V] generates. Then (V , S) is a function algebra
over L. We must now identify S.
Let {'\.}. I be a base for Lover K and suppose that W E S.
1 IE
Then W can be represented in the form

46
 w =L A.f., (2.10.1)
iEI 1 1

where f. E K[V] and only finitely many of the f. are non-zero. Let
1 1
x E V. Then w(x) = 0 if and only if f.(x) = 0 for all i, and therefore
1
w = 0 if and only if every f. is zero. This shows that the representation
1
of a member of 8 in the form (2. 10. 1) is unique and now it follows that

8 = K[V]L, (2.10.2)

where the notation is the same as that employed in section (2. 6). The
results of that section show that 8 is a finitely generated L-algebra,
whereas Theorem 4 Cor. 1 enables us to conclude that it is rationally
reduced. Accordingly 8 = K[V]L is an affine L-algebra. Note that !!
f , f , ... , f belong to K[V] and are linearly independent over K,
12m
then they are linearly independent over L when considered as elements
of 8.
Let w E 8 and let w be represented as in (2.10. 1). Further let
x E V and denote by M the corresponding maximal ideal of K[V]. Then
x
w(x) = 0 if and only if f. E M
1 X
for i E I, that is if and only if w E M 8.
X
Thus

M 8 = {w Iw E 8 and w(x) = 0 } (2.10. 3)

x

and

w(x) = M 8-residue of w. (2.10.4)

x

8ince M 8 n K[V]
x = Mx, we conclude from (2. 10. 3) that different points
of V give rise to different L-rational maximal ideals of 8. Of course,
usually not every L-rational maximal ideal of 8 will arise from a point
of V. To these observations we add the obvious one that if w(x) =0
for all x E V, then w = O.
We now enlarge V to a new set V(L) which is put into a one-one
correspondence with mL (8). At the same time we construct the bijection
V(L) .... mL (8) in such a way that x, in V, is matched with Mx8. Finally
for w E 8 and z E V(L) we define w(z) to be the residue of w with
respect to the member of m L (8) that corresponds to z. The outcome is

47
that y(L) becomes an affine set, defined over L, whose coordinate ring
may be identified with S. Thus

(2.10.5)

Of course y(L) is (essentially) uniquely determined by Y and L. Also

we obtain the isomorphism L[y(L)] '" S of (2. 10. 5) by taking the L-
valued functions that make up L[y(L)] and in each case restricting the
domain to y.

Definition. y(L) is called the 'affine set obtained from Y by

enlarging the ground field from K to L'.

Note that if f E K[Y] ~ K[y]L, then there is a unique member of

L[y(L)] which becomes f when its domain is restricted to y. We call
this the natural prolongation of f to y(L). Frequently the same letter
is used for a member of K[Y] and its prolongation. Each member of
L[y(L)] is a linear combination (with coefficients in L) of such prolong-
ations.

Theorem 29. Let the notation be as above. Then the affine topology
on V is the same as the topology induced on it by the affine topology of
V(L). Moreover the closure of y, in V(L), is y(L) itself.

Proof. Let W E L[V(L)] and write W in the form

w=Af +Af + ••• +A f ,

11 22 mm

where A , A , •.• , A belong to L and are linearly independent over

12m
K, and each f.1 is (the prolongation of) a member of K[Y]. Then the
principal locus

{z Iz E y(L) and w(z) = O}

intersects Y in the locus defined by the equations

f. (x) = 0 (i = 1, 2, •.. , n).

1

On the other hand, any principal locus on Y is the trace on Y of a

48
principal locus on y(L). The first assertion follows from these observa-
tions. The second assertion is clear because if w vanishes at every
point of y, then w = o.

Corollary. Let w ,w
-- 1 2
E L[y(L)]. Then w
1
= w2 if and only if
W
1
(x) = w2 (x) for all x E y.

The following lemma, which is too obvious to require proof, is a

useful aid in identifying the results of extending the ground field.

Lemma 14. Let Y be an affine set defined over K, y* an affine

set defined over L, and suppose that Y is a subset of Y*. Denote by
S the L-algebra obtained by restricting the functions that make up L[Y*]
to Y and suppose that the following two conditions are satisfied:
(a) S = K[y]L,
(b) the homomorphism L[Y*] -+ S of L-algebras induced by
restriction is an isomorphism.
Then the obvious bijection between y(L) and y* is an L-isomorphism.
(Hence y(L) may be identifiedwith Y*.)

We give an application of this lemma.

Theorem 30. Let U be a closed subset of Y. Then the closure

of U in y(L) is U(L).

Remark. It should be noted that if U is a principal locus on Y it

does not follow that U(L) will be a principal locus on y(L).

Proof. Let {~.}.

I IE
I be a base for Lover K and let w E L[y(L)].
Then w has a unique representation in the form

w= L ~.f., (2.10.6)
i€! I I

where f. is (the prolongation of) a member of K[Y], and only finitely

I
many f. are non-zero. It follows that w vanishes on U if and only if
I
fi € Iy(U) for all i. Consequently w(U) = {o} precisely when
W € \r(U)L[y(L)]. Denote by U* the locus of the ideal Iy(U)L[y(L)] in
y(L). Then Theorem 9 shows that U* is the closure of U in y(L).

49
 Let S* respectively S be the L-algebra that one obtains from
L[y(L)] by restricting the domain of the functions to U* respectively
U. Then U* is an affine set (defined over L) whose coordinate ring
is S* and we have a natural homomorphism S* -+ S of L-algebras.
This indeed is an isomorphism because if w vanishes on U, then it must
vanish on U*.
Let W E L[y(L)] and be represented as in (2.10.6). The restriction
of w to U is then

L >..f,
if! 11

where f. is the restriction of f. to U. It follows that

1 1

and now the theorem is a consequence of Lemma 14.

Corollary 1. Let U be a closed subset of the affine set Y and let

U(L) be regarded as a closed subset of y(L). Then the ideal of L[y(L)]
that is associated with U(L) (in the sense of Theorem 10) is Iy(U)L[y(L)].

Proof. Let W E L[v(L)]. The proof just given shows that w

vanishes on U(L) if and only if it vanishes on U. It also shows that w
vanishes on U if and only if W E Iy(U)L[y(L)].

Corollary 2. Let U be a closed subset of the affine set Y and let

U(L) be regarded as a closed subset of y(L). Then U(L) n Y = U.

Proof. From Theorems 29 and 30 we see that U(L) n Y is the

closure of U in y. Consequently U(L) n Y = U.
Corollary 3. Let UI
--
' U 2 , .•• , Um be closed subsets of the affine
set Y and let L be an extension field of K. Then, as subsets of y(L),
we have

50
 We shall now examine the effect of enlarging K to L on K-
morphisms. To this end let !/J: V ... W be a K-morphism of affine sets
and 1/1: V(L) ... W(L) an L-morphism of affine sets. We recall that !/J
induces a homomorphism !/J*: K[W] ... K[V] of K-algebras. Likewise
1/1 induces a homomorphism

of L-algebras. As usual, if x EV and yEW, then M and N will

x y
denote the corresponding rational maximal ideals in K[V] and K[W].

Lemma 15. Let the notation be as above. Then 1/1 extends

!/J : V ... W if and only if 1/1* extends !/J*: K[W] ... K[V].

Proof. Assume first that 1/1* extends !/J*. Let x EV. By (2.6.3),

whence

and therefore 1/I(x) = !/J(x). Thus 1/1 extends !/J.

Now assume that 1/1 extends !/J and suppose that 8 EK[W] ~ L[W(L)].
Then !/J*(8) E K[V] ~ L[V(L)] and it will suffice to show that !/J*(8) and
1/1*(8) coincide as members of L[V(L)]. Let x EV. Then, at x, !/J*(8)
and 1/1*(8) both take the value 8(!/J(x». That !/J*(8) = 1/1*(8) now follows
from Theorem 29 Cor.
We know, from section (1.6), that there is a unique homomorphism

of L-algebras which extends !/J*. This yields the

Corollary. There is exactly one L-morphism V(L) ... W(L) which

extends a given K-morphism !/J : V "'W.

In view of this result we introduce the

51
 Notation. The unique extension of a K-morphism rf>: V -W to an
L-morphism V(L) _W(L) will be denoted by rf>(L).

Obviously if rf> is an identity K-morphism, then rf>(L) is an identity

L-morphism. Also if rf>l : V .... V' and rf>2 : V' .... V" are K-morphisms,
then

(rf> 0 rf> )(L) = rf>(L) 0 rf>(L). (2.10. 7)

2 1 2 1

Lemma 16. Let rf> : V .... W be a K-morphism. Then rf> is a K-

isomorphism if and only if rf>(L) : V(L) .... W(L) is an L-isomorphism.

Proof. Put 1/1 = rf>(L). It is trivial that when rf> is a K-isomor-

phism 1/1 is an L-isomorphism. Assume that 1/1 is an L-isomorphism.
Then, with the notation of Lemma 15, 1/1* : L[W(L)] - L[V(L)] is an iso-
morphism of L-algebras extending rf>* : K[W] - K[V] and we have to show
that rf>* is an isomorphism. Obviously rf>* is an injection.
Let {>t;} i EI be a base for Lover K. Since the elements of
K[W] span L[W(L)] as an L-space, and since 1/1* is surjective, we
see that each W E L[V(L)] can be expressed in the form

W = b >t.f.
iEI 11

where f. E rf>*(K[W]) and only finitely many of the f. are non-zero. But
L[V(L)] ~ K[V]L and now it is clear that rf>*(K[W]) ~ K[V]. This completes
the proof.
Let us now consider the effect of the extension of the ground field
on the product V x W. Suppose therefore that {f.}. I respectively.
1 IE
{g.}. J is a base for K[V] respectively K[W] over K; furthermore let
J JE
F. respectively G. denote the prolongation of f. respectively g. to
1 J (L) 1 (L) J
V (L) respectively W • Then {F.}. I is a base for L[V ] over L
(L) 1 IE
and {G. }. J a base for L[W ] over L. Accordingly the functions
J JE (L) (L)
F. v G. form a base for L[V x W ] over L. On the other hand, the
1 J
f. v g. form a base for K[V x W] over K and therefore they form a
1 J L
base for K[V x W] over L.
Let S be the L-algebra formed by the restrictions of the members
of L[V(L) x W(L)] to V x Wand consider the homomorphism

52
 (2.10.8)

of L-algebras to which the process of restriction gives rise. Since

F, v G, maps into f, v g, it follows that
1 J 1 J

S = K[V x W]L.
Also, because a base of L[V(L) x W(L)] over L is mapped into a base
of S over L, (2.10.8) is an isomorphism. We may therefore conclude,
by appealing to Lemma 14, that

(2.10.9)

We illustrate the consequences of enlarging the ground field by

considering two special cases.

Theorem 31. If the affine set V is finite, then V(L) = V.

L
Proof. By Theorem 18, we have K[V] = \j L (V). However V
can be regarded as an affine set defined over L with \j L (V) as its
coordinate ring. Thus, in this case, making the ground field bigger
produces no new points.
For our second example, we let V be an n-dimensional vector
space over K. Then, as we saw in section (2. 5), V has a natural
structure as an affine set. Let ~ , ~ , ... , ~ be a base for HoIn.p(V, K)
1 2 n .l\.
over K. Then, by Theorem 19, K[V] = K[ ~ , ~ , ... , ~]. Con-
L 1 2 n
sequently K[V]
L
= L[ ~ 1 ,
2
~ , ... ,
n
~ ].
Define V as in section (1.6.1). This is an n-dimensional vector
space over L, each ~,: V ... K has a unique extension to an L-linear
,A L 1 A A A, L
mappmg ~i: V ... L, and ~l' ~2' ••• , ~n IS a base for Hom L (V , L).
Accordingly, by Theorem 19,
LA..... ,.,
L[V ] = L[~l' ~2' ••• , ~n]·

Let S be the L-algebra obtained by restricting the domain of the

functions forming L[V L ] to V. The natural surjective homomorphism

53
of L-algebras which results is such that ~i 1-+ ~r Consequently

L
~ ] = K[Y] .
n

Theorem 32. Let Y be an n-dimensional vector space over K.

Then Y can be regarded as an affine set defined over K and yL as an
~ set defined over L. If the field K is infinit~ then yL = y~
Remark. Theorem 31 shows that the requirement that K be
infinite cannot be left out.

Proof. We use the same notation as that employed in the intro-

duction to the theorem, and note at once that, because of Lemma 14, it
will suffice to prove that ~ ,
are algebraically independent
~ , ... , ~
1 2 v v IJ n
over L. But, by Theorem 19, the power products ~ 1 ~ 2 • •• ~ n are
1 2 n
linearly independent over K and so, when considered as belonging to
K[y]L, they will be linearly independent over L. The theorem follows.
We add a few general comments. Let If>: Y -'W be a K-morphism
of affine sets. The closure If>(Y) , of I/> (V) in W, is an affine set and
¢(L) : y(L) -'W(L) is an L-morphism. Because I/> can be factored
through I/>(y) , it follows that

and therefore, with a self-explanatory notation,

On the other hand I/>(V) ~ I/>(L)(y(L» and hence

Consequently, by Theorem 30,

and thus we have

54
 (2.10.10)

One final observation. In addition to assuming that L is an

extension field of K, let L' be an extension field of L. Then

(2.10.11)

This follows from the relation

(K[V]L)L' = K[V]L'.

Of course V(K) = V.

2. 11 Generalized points and generic pOints

Let V be an affine set defined over K and let L be an extension

field of K. A point ~ of V(L) is called a generalized point of V.
Suppose that ~ is such a point. Then there is a homomorphism

K[V] - L (2.11.1)

of K-algebras in which f E K[V] is mapped into f(~). The image of

K[V] under (2.11.1) is denoted by K[~]. We use K(~) to denote the
quotient field of K[ n
Conversely let

(J : K[V] - L (2. 11. 2)

be a homomorphism of K-algebras. Then there exists a homomorphism

K[V] (gk L - L, of L-algebras, in which f i&l ~ 1-+ ~(J(f); that is we have a
homomorphism

(2.11. 3)

of L-algebras, which on K[V] reduces to (J. The kernel of (2.11.3)

corresponds to a point ~ E V(L) and (J itself is the mapping K[V] - L
in which f 1-+ f(~).

55
 Definition. Let ~ € V(L). We say that ~ is a 'generic point' of
V if the homomorphism (2. 11. 1) is an injection.

Thus a generalized point ~, of V, is a generic point if and only if

it has the following property: whenever f € K[V] and f(~) = 0, then
f = O. Evidently if ~ is a generic pOint of V, then K[V] and K[~]

are isomorphic K-algebras. In particular K[V] is an integral domain.

Theorem 33. If V has a generic point, then K[V] is an integral

domain. If 'K[V] is an integral domain and La is an extension field of
K, then there exists an extension field L, of L , such that V(L) con-
- a
tains a generic point of V.

Proof. We need only establish the final assertion. Let K[V] be

an integral domain. We can certainly find an extension field L, of L ,
a
and a homomorphism a: K[V] - L of K-algebras which is an injection.
As we saw above, there exists ~ € V(L) such that a(f) = f(~) for all
f € K[V]. Thus ~ is a generic point of V.
Now suppose that each of Land Q is an extension field of K. Let
~ € V(L), 71 € v(Q) so that ~, 71 are generalized points of V.

Definition. We say that '71 is a specialization of ~'if whenever

f € K[V] and f(~) = 0, then f(71) = o.
Accordingly 71 is a specialization of ~ if and only if there is a
homomorphism K[~] - [71], of K-algebras, which makes

K[V]

/~
K[ ~] - - - - - K[71]

a commutative diagram. It follows that if ~ is a generic point of V,

then every point of V (generalized or ordinary) is a specialization of ~.

Suppose now that ~, 71, ~ are generalized points of V, that 71 is

a specialization of ~, and ~ is a specialization of 71. Then ~ is a
specialization of ~. Again, if each of ~, 71 is a specialization of the
other, then the homomorphism K[~] -K[71] of K-algebras is an isomor-

56
phism and K[ 1/] .... K[ 1;] is its inverse. Hence if 1/ is a specialization of
I; and 1/ is a generic point of V, then I; is also a generic point of V.

Theorem 34. Let V be an affine set defined over K, let L be an

extension field of K, and let I; E V(L). If now U is a closed subset of
V, then the following two statements are equivalent:
(a) I; E u(L\
(b) f(i;) = 0 whenever f E Iy(U).

Proof. Assume (a) and let f E Iy(U). Then f, considered as a

member of L[V(L)], vanishes on the closure of U in V(L), that is to say
it vanishes on U(L). In particular f(i;) = O.
Next suppose that (b) holds. Then I; is a common zero of the
members of Iy(U)L[V(L)]. But, by Theorem 30 Cor. 1, Iy(U)L[V(L)] is
the ideal of L[V(L)] that is associated with u(L) and therefore its zeros
are the points of U(L). It follows that I; belongs to U(L) and so the
proof is complete.

Corollary 1. Let 1;, 1;' belong to V(L), V(n) respectively t and

suppose that i;' is a specialization of I; when these are considered as
generalized pOints of V. Let U be a closed subset of V. If now
I; E U(L), then i;' E U(n) and 1;' is a specialization of I; when they are
considered as generalized points of U.

Proof. Suppose that f E K[V] and let g be its contraction to U.

Then f( 1;) = g( 1;). The corollary follows from this observation.

Corollary 2. Let U be a closed subset of V and I; a generalized

point of U and hence also of V. Then the specializations of I; when it
is considered as a generalized point of V are the same as its specializa-
~ions when considered as a generalized point of U.

Corollary 3. Let U be a closed subset of V and I; a generalized

point of U and hence also of V. Then K[ 1;] is the same whether I; ~

regarded as belonging to U or as belonging to v.

t It is understood that n, as well as L, is an extension field of K.

57
 Suppose next that V, Ware affine sets defined over K. (As before
L denotes an extension field of K.) Let ~, ~' E V(L) and TJ, TJ' EW(L).
By (2.10. 9), (~, TJ) and (~', TJ') belong to (V x W)(L). Evidently the
K-algebra K[(~, TJ)] contains K[~] and K[TJ] as subalgebras and the
smallest subring of K[ (~, TJ)] that contains both K[~] and K[ TJ] is
K[(~, TJ)] itself. Also if (~', TJ') is a specialization of (~, TJ), then ~'

is a specialization of ~ and TJ' is a specialization of 7J.

Consider a K-morphism cp: V -+W and let ~ EV(L), where L
is an extension field of K. By section (2.10), cp determines an L-
morphism cp(L) : V(L) -+W(L) which extends cpo The point cp(L)W is
usually denoted by cp( ~). Thus cp(~) is a generalized point of Wand if
g E K[W], then

g(cp(~)) = fW, (2.11.4)

where f = g 0 cp E K[V].

Lemma 17. Let cp: V -+ W be a K- morphism of affine sets and

~ a generalized point of V. Then K[ cpW] <: K[ ~].

Proof. Let g E K[W]. Then g(cp(~)) E K[~] by (2.11. 4) and the

lemma follows.
In the next theorem, Land Q denote extension fields of K.

Theorem 35. Let cp: V -+W be a K-morphism of affine sets and

suppose that ~ E V(L~ E V(Q). ~~s a specialization of ~, then
cp(TJ) is a specialization of cpW.

Proof. Let g E K[W] and put f = g 0 cp so that f E K[V]. Assume

that g(cpW) = O. By (2.11.4) we have f(~) = 0 and therefore f(TJ) = O.
But then, again by (2.11.4), g(cp(TJ)) = 0 and the theorem is proved.
Sometimes after the ground field has been enlarged to achieve one
particular end, we later wish to contract it partially in order to achieve
another. In this connection the following lemma can be useful.

Lemma 18. Let V be an affine set defined over K, let L be an

extension field of K, and L' and extension field of L. If now ~'E V(L')

58
 Proof. Let w: L'[V(L')] ... L' be the homomorphism of L'-
algebras which maps each function into its value at ~'. For f E K[V],
we have f( ~') E K[ ~'] k: L. It follows that restriction leads to a homo-
morphism L[V(L)] ... L of L-algebras and so there exists ~ E V(L) such
that fW = f(~') for all f E K[V]. But then OW = O(~') for all 0 in
L'[V(L')] and hence ~ = ~'.

59
3 . Irreducible affine sets

General remarks

The' next stage in the development of the theory of affine sets will
be the application of the theory of irreducible topological spaces. Since
the properties of such spaces may be unfamiliar to the reader, all the
results that we shall need will be established in a short preliminary
section.
Once again K denotes a field concerning which no special assump-
tions are made unless these are stated explicitly in connection with par-
ticular situations.

3. 1 Irreducible spaces

Throughout section (3. 1) X denotes a topological space. If U is

a subset of X, then U denotes its closure.

Definition. The space X is called 'irreducible' if (i) it is not

empty, and (ii) whenever X is expressed in the form X = FlU F 2'
where F l' F 2 are closed subsets, either F 1 =X or F 2 = X.
Theorem 1. An irreducible topological space is connected. Also
any continuous image of an irreducible space is itself irreducible.

These assertions follow immediately from the definition.

Theorem 2. Let X be irreducible and let 0 be a non-empty open

subset. Then X = O.

This is clear because X is the union of the closed sets 0 and

X\O.

60
 Theorem 3.
Let X be irreducible and let 0 1 , O 2 be non-empty
open subsets. Then 0 nO is non-empty.
1 2

By Theorem 2, O2 = X. Let x £ 0 1 • Then x belongs to

Proof.
the closure of 0 and 0 is an open neighbourhood of x. Consequently
2 1
0 1 and O 2 intersect.
Let Y be a subset of X. Then Y can be regarded as a subspace.

Definition. We say that Y is an 'irreducible subset' of X if,

when regarded as a subspace, Y is an irreducible topological space.

For example, if x £ X, then {x} is an irreducible subset of X.

Lemma 1. Let Y be a non-empty subset of X. Then the following

statements are equivalent:
(a) Y is an irreducible subset of X;
(b) if Y
-
~ F 1 u F 2' where F 1 , F2 are closed in X, then either
Y ~ F or Y ~ F •
1 - 2

This is no more than a restatement of the definition of an irreducible

subset.

Lemma 2. Let Y be an irreducible subset of X and suppose that

Y ~ F 1 U F 2 U • •• u F m' where F l ' F 2' ••• , F m are closed subsets of
X. Then Y ~ F.1for ----
some i.

Lemma 2 follows at once from Lemma 1.

Theorem 4. Let Y be an irreducible subset of X. Then Y is

also an irreducible subset of X.

This also follows from Lemma 1.

Lemma 3. Let {y.}. I be a family of irreducible subsets of X,

- 11£
where (i) I is not empty, and (ii) Y. ~ Y. or Y. ~ Y. whenever
--- 11 12 - 12 11

i 1 , i2 £ I. Put

Y =U Y .•
iEl 1

Then Y is an irreducible subset of X.

61
 Proof. Suppose that Y ~ Fl U F , where F ,F are closed sub-
212
sets of X, and assume that Y ~ and Y ~ F. We must derive
F a
1 2 '
contradiction.
There exist i , i E I such that Y. ~ F 1 and Y. ~ F. Now
1 2 11 12 2

either Y. ~ Y. or Y. ~ Y. . We shall suppose, for definiteness, that

11 12 12 11
Yi1 ~Yi2· Since Yi2 ~ Fl U F 2, Yi2 is irreducible, and Yi2 ~ F 2, it
follows that Y. ~ F. But then Y. ~ F and we have obtained the
1 1 1 1
2 1
desired contradiction.
Lemma 3 shows that the irreducible subsets of X, when partially
ordered by inclusion, form an inductive system.

Definition. The maximal irreducible subsets of X are called the

'irreducible components' of X.

Theorem 5. Each irreducible subset of X is contained in an

irreducible component of X. The irreducible components of X are
closed subsets, and X itself is their union.

Proof. Let Y be an irreducible subset of X. The irreducible

subsets of X that contain Y form a non-empty inductive system and
each maximal member of this inductive system will be an irreducible
component of X. This proves the first assertion. The second assertion
follows from Theorem 4. Finally if x E X, then, by the first part, there
is an irreducible component of X containing {x}. This completes the
proof.
Let us suppose that we have a representation of X in the form

X=X uX U ••• UX, (3. 1. 1)

12m

where each X. is a closed irreducible subset of X. Then the irreducible

1
components of X are the maximal members of the set {X, X , ••• , X }
12m
and, in particular, the number of irreducible components in this case is
finite. Moreover if the representation (3. 1. 1) is irredundant, so that for
each val ue of i

X '* Xu
1
••• u X.1- 1 U X.+
1 1
U ••• u X m ,

62
then X , X , ... , X are precisely the irreducible components of X.
12m
Conversely if X has only a finite number of irreducible components,
then not only is X their union but also the union is irredundant.
We now introduce a condition on X that ensures that the number
of its irreducible components is finite.

Definition. The space X is said to be 'Noetherian' if it satisfies

the minimal condition for closed subsets.

Thus if X is Noetherian, then every non-empty collection of

closed subsets contains at least one minimal member. Further any sub-
space formed by a closed subset will also be Noetherian.

Theorem 6. Let X be a Noetherian topological space. Then X

has only a finite number of irreducible components.

Proof. Let Y be a closed subset of X. It will be shown that Y

is the union of a finite number of closed irreducible subsets of X. This
will establish the theorem.
Assume the contrary. Then among the closed subsets Y which do
not have a representation of the right kind there will exist one, Yo say,
which is minimal with respect to inclusion. Now Yo is not empty nor
= F u F , where F "* Y , F "* Y
can it be irreducible. Hence Y 0 12 102 a
and each of F l' F 2 is closed in X. But, by the minimality of Yo' F 1 '
and likewise F 2' is a finite union of closed irreducible subsets of X.
But this implies that F u F = Y is also a finite union of closed
1 2 a
irreducible subsets and now we have the contradiction that we were
seeking.

3. 2 Irreducible affine sets

Throughout section (3. 2) V will denote an affine set defined over K.

Since V is a topological space by virtue of its affine topology, the results
of section (3.1) may be applied to it. First, however, it will be convenient
to recall some basic results from the theory of commutative rings.
To this end let R be a commutative ring with an identity element.

63
It is known t that the following statements concerning R are equivalent:
(a) every ideal is finitely generated;
(b) every infinite increasing sequence of ideals becomes constant;
(c) every non-empty set of ideals contains at least one maximal
member.
Should it happen that R possesses these properties, then it is said to be
a Noetherian ring. Clearly any homomorphic image of a Noetherian ring
is itself Noetherian.
Next suppose that X , X , •..
1 2
,.xn are indeterminates. The
celebrated Basis Theorem t states that if R is a Noetherian ring, then
the polynomial ring R[X1 , X , ..• , X ] is also Noetherian. Since the
2 n
field K is trivially Noetherian, it follows that K[X , X , .•. , X ] is
1 2 n
Noetherian. But Chapter 2 Lemma 4 shows that K[V] is a homomorphic
image of such a ring. Consequently we have proved

Theorem 7. The coordinate ring of an affine set V is a Noetherian

Next, by Chapter 2 Theorem 10, there is a bijection between the

set of closed subsets of V on the one hand and the set formed by certain
ideals of K[V] on the other; furthermore the bijection described in the
theorem reverses inclusion relations. Hence, in view of Theorem 7, we
obtain

Theorem 8. With respect to its affine topology, the affine set V

is a Noetherian space.

It now follows, from Theorem 6, that V has only a finite number

of irreducible components. We say that V is an irreducible affine set if
it has precisely one irreducible component. In other words V is an
irreducible affine set if and only if it is irreducible as a topological space.

t See, for example, [(9) Proposition 9, p. 22].

t For a proof see [(9) Theorem 1, p. 179].

64
 Theorem 9. The following statements, concerning the affine set
V, are equivalent:
(i) V is irreducible;
(ii) K[V] is an integral domain;
(iii) V possesses a generic point.

Proof. We already know, from Chapter 2 Theorem 33, that (ii)

and (iii) are equivalent. In what follows we demonstrate the equivalence
of (i) and (ii).
First assume (i). Suppose that f, g E K[V] and fg = O. If we put

F
1
= {xix EV and f(x) = 0),
F
2
= {xix EV and g(x) = 0),
then F . F
l' 2
are closed subsets of V and Fl U F
2
= V. Thus either
F
1
=V or F
2
= V, that is to say either f =0 or g = O. Accordingly
(i) implies (ii).
Now assume (ii). We suppose that V = FlU F 2' where F l' F 2
are closed subsets of V. It will suffice to prove that F 1 =V or
F
2
= V. We assume the contrary and seek a contradiction.
Since F 1 *- V, it follows (see Chapter 2 Theorem 10) that
Iy(F 1) *- (0). Likewise Iy(F 2) *- (0). On the other hand

and therefore Iy(F 1 )ly(F) = (0). But this is contrary to our assumption
that K[V] is an integral domain.

Corollary 1. Let U be a closed subset of the affine set V. Then

U is irreducible if and only if Iy(U) is a prime ideal.

This follows from Theorem 9 because K[U] is isomorphic to

K[V]!Iy(U).

Corollary 2. Let!/l: V -W be a K-morphism of affine sets,

where V is irreducible. Then the closure </l{V}, of </l(V) in W, is
irreducible Moreover if ~ is a generic point of V, then </l(~) is a

65
generic point of cp(V).

Proof. The first assertion follows from Theorems 1 and 4. Next,

by (2.10.10), CPW is a generalized point of cp(V). Put

~= {flfEK[W) and f(CP(~))=O}

and let x E V. Then, by Chapter 2 The()rem 35, CP(x) is a specialization

of CPW and therefore CP(x) E CW(~)' It follows that ¢(V) k CW(~)'
Firullly let g E K[ I/l(V)) be the restriction, to CP(V), of f E K[W)
and suppose that g(I/l(~)) = O. Then f E ~ and hence it vanishes on
I/l(V). Thus g = 0 and the proof is complete.
At this point we interrupt the development of the main theory to
mention two examples.

Theorem 10. If V is a finite affine set, then its irreducible com-

ponents are its individual points.

This is obvious.

Theorem 11. Let V be a finite-dimensional vector space over K.

If K is an infinite field and V is considered as an affine set, then V
is irreducible.

Proof. Let X , X , ••• , X be indeterminates, where n is the

1 2 n
dimension of V as a vector space. Then K[V) and K[X , X , ... , X ]
1 2 n
are isomorphic K-algebras. Consequently K[V) is an integral domain.
We now return to the general theory.

Theorem 12. There is a natural bijection between the closed

irreducible subsets of V and the prime ideals P, of K[V), which satisfy
Rady(P) = P. If the closed irreducible subset U is associated with the
prime ideal P (in this bijection), then U = CV(P) and P = ly(U).
This follows by combining Theorem 9 Cor. 1 with Theorem 10 of
Chapter 2.

Corollary. If K is algebraically closed, the bijection described

in Theorem 12 is between the set of closed irreducible subsets of V and

66
the set of all prime ideals of K[V].

Proof. Since K is algebraically closed, every prime ideal P,

of K[V], satisfies t Rady(P) = P.
Let U be a closed subset of V. From Theorems 6 and 8 we see
that U has only a finite number of irreducible components. Denote these
by UI , U2 , ••• , Um . Then

(3. 2. 1)

Put

(3. 2. 2)

Since U is an irreducible closed subset of V, Po is a prime ideal of

0

1 1
K[V] and Rad._(Po) = Po. Also if i"* j, then Po \l Po.
-V 1 1 1 J
From (3. 2. 1) we conclude that

that is to say

PIn P 2 n • .. n Pm = ly(U). (3. 2. 3)

Let P be any prime ideal of K[V] that contains ly(U). Then

PI P •.• P k P and therefore Po k P for some value of i. Thus
2 m 1
P , P , ... , P are the minimal members of the set of all prime
12m
ideals that contain \r(U), that is to say they are the minimal prime ideals
of \r(U), These observations add up to a proof of

Theorem 13. Let U be a closed subset of V. Then ly(U) has

only a finite number of minimal prime ideals. Let these be P , P , ..• ,P •
12m
Then Rad._(P
---- ·V 10) = Po1 for each value of i, the irreducible affine sets

are the irreducible components of U, and

t See the remarks follOWing Theorem 13 of Chapter 2.

67
 L(U)
-v = P1
n P n... n P .
2m

The case U = V deserves a special mention.

Corollary. The irreducible components of V correspond to the

minimal prime ideals of the zero ideal of K[V].

Theorem 14. Let V be an irreducible affine set, let f E K[V]

and suppose that f vanishes on some non-empty open subset N of V.

Then f = O.

Proof. The principal locus

{x Ix E V and f(x) = O}
is closed and it contains N. It therefore contains the closure 'iii' of N.
However, by Theorem 2, 'iii' = V.

Theorem 15. Let x E V and f E K[V]. Suppose that f vanishes on

some open set N, where x E N. Then there· exists g E K[V] such that
g(x) *- 0 and fg = O.

Proof. Let V , V , ••• , V be the irreducible components of V

1 2 S
and put P.
1
= L(V.).
-V 1
Further let M
X
be the rational maximal ideal of
K[V] that corresponds to x. Without loss of generality we may suppose
that

and

P II +1 <;;; MX , P II +2 <;;; MX , ... , P S <;;; MX .

Suppose that 11+ 1 :5 j :5 s. Then x

V. and f vanishes on V. n N. E
J J
Consequently, by Theorem 14, f vanishes on the whole of V. and there-
J
fore f E P .. Choose g in P n P n... n P so that g f M. Thus
J 1 2 II X
g(x) *- 0 and fg belongs to

P n P n •.• n P = L(V) = (0),

1 2 s-Y

68
i. e. fg = O. This completes the proof.
The next theorem provides a good example of the power of topologi-
cal methods. In it V and W denote affine sets defined over K. Let
x € V and denote by

T :W-+vxW
x

the mapping in which (y) = (x, y). We know, from Chapter 2 Lemma 12,
T
x
that T is a K- morphism.
x

Theorem 16. Let V and W be irreducible affine sets. Then

V x W is also an irreducible affine set.

Proof. We must show that, when V x W is endowed with the

affine topology, it is an irreducible space. Suppose therefore that
V xW = T u T', where T and T' are closed subsets of V x W. It
will suffice to show that either V x W ~ T or V x W ~ T'.
Let y € W. By Chapter 2 Theorem 28, V and V x {y} are K-
isomorphic and therefore they are homeomorphic topological spaces.
Consequently V x {y J is irreducible and now it follows that either
V x {y J ~ T or V x {y J ~ T'.
Define subsets Wand W , of W, by
1 2

W = {y /y €W and V x {y} ~ T}
1

and

W = {y /y € Wand V x {y} ~ T' J.

2

Then W = W1 u W 2 , by virtue of the observations made in the last para-

graph, and it will now suffice to show that either W
1
=W or W
2
=W.
Moreover this will follow if we show that Wand Ware closed in W.
1 2
But

W = n T-1(T)
1 X€V x
and this is closed because each T is continuous. For similar reasons
x
W 2 is closed. The theorem follows.

69
 Corollary. Let R and S be affine algebras over K and suppose
that R and S are integral domains. Then R ®K S is an integral domain

As the next result shows, Theorem 16 is easily generalized.

Theorem 17. Let V , V , ... , V be the irreducible components

- 1 2 r
of the affine set V -
and
-
W1 , W 2 , ... , Ws the irreducible components
of the affine set W. Then the sets V. x W., where 1 ~ i ~ r, 1 :S j :S S,
1 J ---
are the irreducible components of V x W.

Proof. The V. x W. are closed subsets of V x Wand, by

1 ]
Theorem 16, they are irreducible. Furthermore V x W is their union.
Since none of V1 x Wl' V1 X W 2' ••• , Vr x Ws is contained by any of
the others, the theorem follows.

3. 3 Localization

Let V be an affine set defined over K and let x € V. Suppose

that f, g € K[V] and g(x)"* O. Then

N = {y y I € V and g(y) "* 0 }

is an open subset of V and x € N. Furthermore we can define a mapping

(3. 3. 1)

by
e(y) = ili)
f(y) . (3. 3. 2)

Now keep V and x fixed and let f and g vary. In this way we
obtain a collection, --V,x
A__ say, of K-valued functions. Suppose that e, e'
are in --V,x
A__ • We introduce an equivalence relation on !IV
,x by regar-
ding e and e' as equivalent if they agree on some non-empty open sub-
set of V that contains x. Denote the set of equivalence classes by
QV and let 8 € QV • Then the functions which make up 8 all take
,x ,x
the same value at x. If this common value is denoted by 8(x), then we
have a mapping

70
 Q -+K (3.3.3)
V,x

in which e 1-+ e(x).

Next assume that f , g , f ,g all belong to K[V] and that
1 1 2 2
g (x) if' 0, g (x) if' O. The pair f., g. determines a function, e. say, that
1 2 1 1 1
belongs to J\V . If N denotes the open set where g g does not vanish,
,x 1 2
then x E N and both eland e2 are defined on N. Furthermore
e1 + e2 and e1 e2 , considered as K-valued functions defined on N, belong
to AV ,x.
It will now be clear how addition and multiplication may be defined
on ~,x and it is also clear that ~,x has a natural structure as a
K-space. In fact with this additional structure QV
,x is a K-algebra t
and (3. 3. 3) is a surjective homomorphism of K-algebras.

Definition. The K-algebra Q_- is called the 'local ring' of V

""'V,x
at x.

If f E K[V], then f and 1 together determine a function (defined on

V) which belongs to flV
,x. In this way we can associate with each
f E K[V] an element of QV
,x and thus arrive at a homomorphism
K[V] -+QV (3. 3. 4)
,x
of K-algebras. It is customary to refer to (3.3.4) as the canonical
homomorphism of K[V] into QV •
,x
The above approach to localization is based on functions. We shall
now give an alternative procedure that is more algebraic. First, however,
we need

Theorem 18. Let f , g , f ,g belong to K[V] and assume that

- 1 1 2 2
gl (x) if' 0, g2 (x) if' O. Further let the pair fi' gi determine the function
e.1 -in -V,x
~- . Then each of the follOWing statements implies the other:

(0 e1 and
--
e2 are equivalent members of J\V
,x
;
(ii) there exists h E K[V] such that h(x) if' 0 and hg f
-- 21
=hg12
f •

t This algebra is unitary, associative and commutative.

71
 Proof. If (ii) holds, then 81 and 8 2 agree on the open set where
hg g does not vanish. Thus (ii) implies (i).
1 2
Now suppose that (i) holds. Then f g - f g vanishes on some
1 2 2 1
open set containing x. That there exists h E K[V] with the required
properties is therefore a consequence of Theorem 15. This completes
the proof.
At this point we make a fresh start. Let x
E V and denote by M
x
the corresponding K-rational maximal ideal of K[V]. We now consider
formal fractions fig, where f, g E K[V) and g ¢ M. These are just
x
ordered pairs (written in a certain way) of elements of K[V), where the
second component (the denominator) satisfies g(x) * O. Two such formal
fractions, say fig and f' Ig', are to be considered as equivalent if there
exists h K[V], such that (i) h ¢ M , and (ii) hg'f = hgf'. (This is an
E
x
equivalence relation in the abstract sense.) The set of equivalence classes
will be denoted by (K[V])M and we use fig not only to denote the formal
x
fraction determined by f and g, but also the equivalence class to which
the fraction belongs.
Suppose that f/g 1 and f/g 2 belong to (K[V])M and that k E K.
x
It is easily verified that (K[V])M constitutes an associative, unitary,
x
and commutative K-algebra in which

f1 (kf 1 )
k(-) = - .
gl gl
Next let fig be a formal fraction and the member of --v,
f) A.._ x which it
determines. Then, by Theorem 18, fractions equivalent to fig determine
functions equivalent to 8 and the outcome is that we obtain an isomor-
phism

(3. 3. 5)

72
of K-algebras. This isomorphism is frequently used to identify the two
sides of (3. 3. 5). If this is done, then the homomorphism (3. J. 3) maps
fig into f(x)/g(x) whereas (3.3.4) takes f, of K[Y], into fll.
Suppose that fig f Qy ,where f, g EK[Y] and g(x)
,x *D. Obviously
fig is a unit in Qy
,x if and only if f(x) O. * Consequently fig is a non-
unit if and only if f EM. Denote by M Qy the ideal, of Qy ,which
x x ,x ,x
is generated by the image of M under the canonical homomorphism
x
K[Y] ... Qy . In terms of this notation our remarks prove
,x
Theorem 19. The non-units of Qy form the ideal M Qy .
,x x ,x
This ideal is the kernel of the homomorphism (3.3. 3) and therefore it is

-a K- rational maximal ideal. It is the only maximal ideal of Qy ,x

rational or otherwise.

We next examine briefly the connection between K-morphisms and

localization. To this end we assume that rp: Y ... W is a K-morphism of
affine sets; and for x f y, Y f W we use M respectively N to denote
x y
the corresponding K-rational maximal ideal of K[Y] respectively K[W].
The K-morphism I/J induces a homomorphism I/J*: K[W] - K[Y] of
K-algebras. Since, by (2.6. 3),

(3. 3. 6)

rp* induces a homomorphism

(3. 3. 7)

of K-algebras. To see how the homomorphism operates, let (J' = f'lg',

where f', g' E K[W] and g' ¢ Nrp(x)' Then

* (J' = I/J*{g'}
¢x()
rp*(f') =!~
g' I/J . 0
(3. 3. 8)

Accordingly

(¢*((J')(x) = (J'(rp(x)) (3. 3. 9)

x
and

(3.3.10)

73
In particular the inverse image, in (3.3. 7), of the maximal ideal of
QV, x is the maximal ideal of ~,cp(x)' a fact which is sometimes des-
cribed by saying that cp* is a local homomorphism.
x
Clearly if cp : V ..... W is a K-isomorphism of affine sets, then cp*
x
is an isomorphism of K-algebras. Also if cp: V ..... Wand 1/1: W ..... Tare
both of them K-morphisms of affine sets, then

(1/1 0 cp)*x = cp*x 1/1*cp(x) •

0 (3. 3. 11)

Finally let U be a closed subset of V and let x E U. The inclusion

K- morphism j: U ..... V induces a surjective homomorphism j*: K[V].....K[U]
of K-algebras whose kernel is !y(U). It follows that

J.*.~
x . , x .... QU, x
is surjective and its kernel is the ideal !y(U)QV, x that is generated by
the image of !y(U) under the canonical homomorphism K[V] .... QV, x'
Accordingly we obtain

Theorem 2 O. Let U be a closed subset of the affine set V and let

x E U. Then the K-algebras QU and Qy IL(U)Qy are naturally
,x ,x' -V ,x
isomorphic.

3. 4 Rational functions and dimension

Throughout section (3.4), V will denote an affine set defined over K.

First suppose that V is irreducible. By Theorem 9, K[V] is an
integral domain and therefore it has a quotient field. The quotient field
will be denoted by K(V). It is, of course, an extension field of K.
Let e E K(V) and suppose that x E V. We say that e is defined at
x if there exist f, g E K[V] such that e= fig and g(x) * D. If N
denotes the set where e is defined, then N is a union of open sets and
therefore it is itself an open set.
Suppose now that x E N so that there exist f, g E K[V] such that
e = fig and g(x) * D. We may put e(x) = f(x)/g(x) because f(x)/g(x)
does not depend on the choice of f and g. In this way, we derive, from

74
e, a function whose domain is N and which takes values in K. By an
abuse of language a function obtained in this way is called a rational
function on V.

Theorem 21. Let V be an irreducible affine set and let el' e2

belong to K(V). Then
--
e1 = e2 if and only if there is a non-empty open
subset N, of V, on which the corresponding rational functions are defined
and equal.

Proof. Assume that e1 (y) = e2 (y) for all YEN, where N is a

non-empty open subset of V. Choose x EN. Then there exist f 1 , gl'
f ,g in K[V] such that g (x) "* 0, g (x) "* 0 and f.jg. = e.. Next
22 1 2 111
there is a non-empty open subset, N' say, of V on which fl g2 - f 2g1
vanishes, and now it follows,, by Theorem 14, that fIg 2 - f 2 g 1 = O.
Thus e1 = e.2 This proves part of the theorem and the rest is trivial.
Theorem 21 shows that there is a natural bijection between K(V)
and the set of rational functions on V. Indeed on the basis of this we may
regard the rational functions, on the irreducible affine set V, as forming
an extension field of K. In practice we often speak of the elements of
K(V) as though they were rational functions thereby making the obvious
identification.
Still assuming that V is irredUCible, we note that the rational
functions that are defined at the point x form a subalgebra of K(V), and
that this subalgebra is none other than the local ring Q_- of V at x.
~,x
Weare now ready to introduce the notion of dimension.

Definition. The dimension, Dim V, of the irreducible affine set V

is the transcendence degree of the field K(V), of rational functions on V,
over the ground field K.

Now suppose that V is not necessarily irreducible and let

V , V , ••• , V be its irreducible components. We extend the above
d:finit~on by pufting t
Dim V = max {Dim VI' Dim V 2' ••• , Dim Vp }. (3.4. 1)

t The empty affine set is given the conventional dimension -1.

75
 The next result is a crucial lemma in dimension theory. In order to
prevent the statement of it from becoming too long, we preface it by some
remarks which describe the general setting.
Suppose that Rand S are finitely generated K-algebras, which are
also integral domains. Denote by L the quotient field of R and by Q

that of S. Then L has transcendence degree, p say, over K and Q

has transcendence degree, q say, also over K. The result in question

can now be stated as

Lemma 4. Let the situation be as described above and let I/J: R - S

~~ a surjective homomorphism of K-algebras. Then p?: q. ~ P = q,
then cfJ is an isomorphism.

Proof. Let R = K[~ , ~ , ... , ~ ] and put 17. = cfJ(~.) so that

1 2 n 1 1
S = K[ 17 , 17 , ... , 17 ]. Without loss of generality we may suppose that
1 2 n
TJ , 17 , ..• , 17 are algebraically independent over K. Then
1 2 q
~ , ~ , ..• , ~ are also algebraically independent over K and therefore
1 2 q
P ?: q. From here on we assume that p = q.
Suppose that z E Rand cfJ(z) = O. It will suffice to show that z is
zero. Let Z, Xl' X 2 , ••• , Xq be indeterminates. Since z is algebraic
over K(~ , ~ , ... , ~), we can find a non-null member
1 2 q
F(Z, Xl' X 2 , . . . , Xq) of K[Z, Xl' X 2 , . . . , Xq] suchthat
F(z, ~ , ~ , ... , ~ ) = 0 and, moreover, we can arrange that
1 2 q
F(Z, Xl' X 2 , ... , Xq) is irreducible. Next F(O, 1}l' 17 2 , ... , 17 q )=O
and therefore F(O, Xi' X 2 , ••• , Xq) = O. Accordingly F(Z, X l ,X 2 , ... ,Xq )
has Z as a factor and thus we see that

for some c '* 0 belonging to K. This shows that z = 0 and completes

the proof.
Weare now ready to prove

Theorem 22. Let V be an irreducible affine set and let U be a

closed subset of V. If now U '* V, then Dim U < Dim V.

76
 Proof. Clearly we may assume that U, as well as V, is
irreducible. Consider the surjective homomorphism K[V] -+ K[U] that
arises from the inclusion mapping of U into V. This has kernel ly(U)
and, since U"* V, ly(U) "* O. The theorem therefore follows from
Lemma 4.

Corollary. Let U be a closed subset of the affine set V. Then

Dim U =s Dim V.

Proof. The assertion is clear as soon as one notes that each com-
ponent of U is contained in a component of V.
A second application of Lemma 4 is provided by

Theorem 23. Let V be an irreducible affine set and ~ a general-

ized point of V. Then the transcendence degree of K(~) over K is at
most Dim V. Furthermore there is equality if and only if ~ is a generic
point of V.

Proof. If ~ is a generic point of V, then K(V) and K(~) are iso-

morphic and therefore the transcendence degree of K(~) over K is
Dim V.
Now assume that ~ is not a generic point of V. By (2.11. 1), there
is a surjective homomorphism K[V] - K[~] of K-algebras and, by
hypothesis, this is not an isomorphism. Thus, by Lemma 4, the trans-
cendence degree of K(~) over K is smaller than Dim V and so the
theorem is proved.
We take this opportunity to identify those affine sets that have
dimension zero.

Theorem 24. The affine set V is irreducible and of dimension

zero if and only if it consists of a single point. When this is the case
K[V] = K.
Proof. Assume that V is irreducible and that Dim V = O. Since
K[V] is a ring between K and K(V), and K(V) is an algebraic extension
of K, K[V] itself must be a field and therefore it has exactly one maximal
ideal namely (0). But K[V] has at least one K-rational maximal ideal.

77
This shows that K[V] = K and that V has only one point. The remaining
assertions of the theorem are trivial.

Corollary. Let V be a non-empty affine set. Then Dim V = 0

if and only if V is a finite set.

There is another special case which it is convenient to mention

at this point. Suppose that V is an n":dimensional vector space over K.
If K is a finite field, then Theorem 24 Cor. shows that, as an affine set,
V has dimension zero. On the other hand, if K is infinite, then K[V]
is isomorphic to the polynomial ring K[X , X , ... , X ] and therefore
1 2 n
the dimension of V, as an affine set, is also n.
We next consider the dimension of the product of two affine sets.
To this end let V, W be non-empty affine sets defined over K and, on
the basis of (2.8.1), let us make the identification

K[V x W] = K[V] ®K K[W]. (3. 4. 2)

The discussion given in section (1. 5) shows that K[V], K[W] may be
regarded as subalgebras of K[V] ®K K[W] and hence of K[V x W]. On
this understanding a typical element of K[V x W] may be written in the
form

fg+fg+ ... + f g , (3. 4. 3)

11 22 mm

where fi E K[V] and gi E K[W]. It follows that if K[V]=K[~1'~2'"'' ~p]

and K[W] = K[17 1 , 17 2 , ••• , 17q]' then

K[V X W] = K[~l' ... , ~p' 17 1, ... , 17q]' (3.4. 4)

Lemma 5. Suppose that V, W are non-empty affine sets. Let

f , f , ... ,f belong to K[V] and g, g , ... , g to K[W]. If now
12m -- 1 2 m -
f , f , •.• ,f are linearly independent over K and
12m
f11
g + f 22g + •.. + fmm
g = 0, - then
-1
g. = 0 -for i = 1, 2, ••• , m.

Proof. In view of (3.4. 2) this is simply a special case of Chapter 1

Theorem 1.

78
 After these preliminaries let us assume that V and Ware
irreducible. By Theorem.s 9 and 16, each of K[V], K[W], and
K[V x W] is an integral domain. Also, using our previous notation,
K(V) = K(~l' ~2' ... , ~), K(W) = K(1]l' 1]2' ... , 1]q)' and

K(V x W) = K(~l' ••• , ~p' 1]1' ... , 1]q) (3.4. 5)

Let Dim V =r and Dim W = s. We may suppose, without loss of

generality, that ~1' ~ , ... , ~ are algebraically independent over K
2 r
and 1] , 1] , ••• , 1] are also algebraically independent over K. This
1 2 S
ensures that K(V x W) is an algebraic extension of K( ~ , ... , ,1] ,... ,1] ) ~
r 1 s 1
and so it follows that Dim(V x W) ::;: r + s. Now the set formed by the
v v v
power products ~ 1 ~ 2 .•• ~ r is linearly independent over K and thus
1 2 r
we see, from Lemma 5, that it is also linearly independent over K(W).
In other words, ~ , ~ , ••• , ~ are algebraically independent over
1 2 r
K(1] , 1] , ••• , 1] ) and therefore Dim(V x W) 2: r + s. Our inequalities
1 2 q
can now be combined to yield Dim(V x W) = r + s.

Theorem 25. Let V and W be non-empty affine sets defined over

K. Then

Dim(V x W) = Dim V + Dim W.

Proof. In view of Theorem 17 we may assume that V and Ware

irreducible. However in that case the desired result follows from the
preceding discussion.

3. 5 Enlargement of the ground field

In section (3. 5) we shall consider an affine set V, defined over K,

along with an extension field L of K. As in section (2.10), this allows
us to construct, in a natural manner, an affine set V(L), defined over L
and containing V as a subset. Each member of K[V] has a natural
prolongation to a member of L[V(L)] and, by identifying each member of
K[V] with its prolongation, we may embed K[V] in L[V(L)]. If this is
done, then

79
 (3. 5. 1)

where the right-hand side is to be understood in the sense explained in

section (1. 6).

Theorem 26. If V is irreducible, then so too is V(L). Further-

more V and V(L) have the same dimension.

Proof. Suppose that w, w' E L[V(L)] and ww' = D. We shall

show that either w = 0 or w' = O. Now there exist ~ , ~ , ••. , ~ ,
1 2 S
in L and linearly independent over K, so that

w = \ fl + ~2f2 + . .. + ~sfs'
w' =~ 1 f'1 + ~ f' + ... +
2 2
~
S
f'
s'

where f. and f! belong to K[V]. Denote by U respectively U' the

1 1
closed subset of V where f 1 , f 2, ... , f respectively f', f' , ... ,f'
s 1 2 S
all vanish. Then when x E V either

~ f (x) +
11
~ f (x) + . .. + ~ f (x)
22 SS
=0
or

~ f' (x) + ~ f' (x) + . .. + ~ f' (x) = O.

11 22 SS

This shows that x E U or x E U'. Thus V = U U U' whence U = V or

U' = V. If U = V, then all the f. are zero and therefore w = O. If
U' = V, then w' = O. AccOrdingl~ V(L) is irreducible.
Next let K[V] = K[ ~1' ~2' .•. , \,] and suppose that Dim V = r.
Without loss of generality we can assume that ~1' ~ , ••• , ~ are
2 r
algebraically independent over K. By (3. 5. 1), L[V(L)]
(
= L[ ~1 ,~ 2 , ..• , ~P]
and this shows immediately that Dim V L) ~ r. On the other hand, the
/.I /.I /.I
power products ~11 ~ 2 •.• ~ r are linearly independent over K and
2 r
therefore by (3.5.1), linearly independent over L. Consequently we also
have Dim V(L) ~ r and now the proof is complete.

Theorem 27. Let VI' V , ... , V be the irreducible components

of the affine set V. Then V(L): V(L), ••r., V(L) are the irreducible com-
-- 1 2 r

80
ponents of V(L).

Remark. Here we use Chapter 2 Theorem 30. This allows us to

regard V~L) as a closed subset of V(L).
J
Proof. By Theorem 26, V~L) is irreducible and, by Chapter 2
J
Theorem 30 Cor. 3,

V(L) = V1(L) u V(L) u • •• u V(L).

2 r

Moreover Chapter 2 Theorem 30 Cor. 2 shows that none of

V(L), V(L), ..• , V(L) is contained in any of the others. The theorem
1 2 r
follows.

Corollary 1. The relation Dim V = Dim V(L) holds even if the

affine set V is reducible.

Corollary 2. Let ~ be a generalized point of V. Then the

transcendence degree of K(~) over K is at most Dim V.

This corollary is obtained by combining Theorem 27 with Theorem

23.
At this point we shall interrupt the discussion of affine sets in order
to introduce certain ideas that belong to the theory of fields.

Lemma 6. Let L 1 ,L 2 be subfields of a field n and let K be a

--
common subfield of Land L. Assume that every set of elements of
1 -- 2
L1 which is linearly independent over K is also linearly independent
over L. Then every set of elements of L which is linearly independent
--- 2 2
over K is linearly independent over L .
--- 1

Proof. Let R be the subring of n that is generated by L1 and

L. Then R is a K-algebra and from our assumption it follows that we
2
have an isomorphism L ®K L '" R which matches A ® A with A A •
1 2 1 2 12
(Here A. belongs to L .. ) The lemma therefore follows from Chapter 1
1 1
Theorem 1.
It will be noticed that, in the situation described in Lemma 6, the
relation between Land L is a symmetrical one. This relation is
1 2

81
described by saying that Ll and L2 are linearly disjoint over K.
Let Q be an extension field of K and let Q be the algebraic
closure of Q. Then 11 contains the algebraic closure K of K.

Definition. Suppose that, in the above Situation, Q and K are

linearly disjoint over K. Then Q is said to be a 'regular extension' of
K.
Later we shall be concerned with the properties of regular extensions
For the moment, however, we merely note the definition and show that
regular extensions arise naturally in connection with irreducible affine
sets.
Suppose that V is irreducible. Then not only is K[V] an integral
domain but, by Theorem 26, L[V(L)] is an integral domain as well. Now
K[V] is a subring of L[V(L)]. Consequently K(V) and L have L(V(L))
as a common extension field. They also have K as a common subfield

Theorem 28. Let V be irreducible. Then K(V) ~ L, con-

sidered as subfields of L(V(L\ are linearly disjoint over K.

Proof. Let f1 , f2, ..•m

, f belong to K(V) and be linearly
independent over K. Further, suppose that A f + A f +... + A f = 0
11 22 mm
where A. E L. We have to show that A , A , ••• , A are all zero. By
1 12m
multiplying f1, f2, ...m, f by a suitable non-zero element of K[V], we
may suppose that f. E K[V] for all i. But now what we have to prove is
1 (L) L
clear because, by (3. 5. 1), L[V ] = K[V] .

Theorem 29. Let the affine set V be irreducible. Then K(V) is

a regular extension of K.

This follows from Theorem 28 by taking L to be the algebraic

closure of K.
We next examine the effect produced by an extension of the ground
field on rational functions. As a preliminary to this examination we prove

Lemma 7. Let V be an irreducible affine set and 0 a non-empty

open subset of V. ~en the closure of 0 in V(L) is V(L) itself.

82
 Proof. Theorem 2 shows that the closure of 0 in V is just V
and we know, from Chapter 2 Theorem 29, that the closure of V in V(L)
is V(L). The lemma follows from these facts.

Theorem 30. Let the affine set V be irreducible and let e l' e 2
be rational functions on V(L). If
-
e1 and
--
e2 are defined and equal on
some non-empty open subset of V, -
then
-
e1 = e 2.
This is an easy consequence of Lemma 7.
Once again let V be irreducible. We know that K(V) is a sub-
field of L(V(L». Consequently each rational function on V gives rise to
a rational function on V(L) or, as we shall say, each rational function on
V has a natural prolongation to a rational function on V(L). Let e be a
rational function on V and denote by e its prolongation to V(L). Con-
sidered as functions, e is defined on an open subset of V and e on an
open subset of V(L). Now assume that x E V. At first sight it appears
possible that e could be defined at x even though e was not. However,
as the following theorem shows, such inconvenient behaviour cannot, in
fact, occur.

Theorem 31. Suppose that the affine set V is irreducible. Let e

be a rational function on V and denote by e its natural prolongation to a
rational function on V(L). Finally let x be a point of V. Then e is
defined at x if and only if e is defined at x.·

Proof. It will be assumed that e is defined at x and we shall

deduce that e is defined at x. The converse is trivial.
By hypothesis there exist f, g E K[V] and F, G E L[V(L)] such that
g * 0, G(x) * 0, F /G = e, and e= fig. Thus gF = fG. Choose
AI' A , •.• , A in L, linearly independent over K, so that
2 n

F=Af +Af + ... +Af,

11 22 nn
G=Ag +Ag + •.• +Ag,
11 22 nn

where the f. and g. belong to K[V]. Then

1 1

83
 n
L A.(gf. - fg.) = 0
i=1 1 1 1

and therefore, because L[V (L)] = K[V]L, gf. = fg. for all i. But
1 1
G(x) *0 and therefore we can choose i so that g.(x)
1
* O. Since e= f./g.
1 1
this proves that e is defined at x.

3.6 Almost surjective K-morphisms

A number of results in section (3. 6) require that the ground field K

be ~lgebraically closed. Whenever this is the case the fact is always made
quite explicit in the statement of the result in question.

Definition. A K-morphism I/J: V ... W of affine sets is said to be

'almost surjective' if the closure of ¢(V) in W is W itself.

Theorem 32. The K-morphism ¢: V ... W is almost surjective if

and only if the associated homomorphism ¢*: K[W] ... K[V] of K-algebras
is an injection.

Proof. If y E K[W] and vanishes on I/J(V) , then it vanishes on the

closure of cp(V). Consequently when cp is almost surjective cp* must be
an injection. We shall therefore assume that ¢ is not almost surjective
and deduce that cp* is not an injection.
In the situation under discussion the closure ¢(V), of ¢(V), is a
proper closed subset of W. Hence there exists g E K[W] such that the
principal locus CW(g) contains cp(V) but is not the whole of W. But
then

¢*(g) =g 0 rfi =0
whereas g * O. Thus the theorem is proved.
We now need some lemmas concerning commutative, finitely genera-
ted K-algebras, and for these it will be convenient to use a common nota-
tion. This will now be explained.
For Lemmas 8 to 10 inclusive R will denote a finitely generated
K-algebra which is also an integral domain; and S will be a finitely
generated subalgebra. By Dim R will be meant the transcendence degree

84
of the quotient field of Rover K, and if II is a prime ideal of R we
shall put

dim II = Dim(R;TI).

Similar definitions apply to S and its prime ideals. Note that

Dim S :5 Dim R.
Since R is finitely generated as a K-algebra, we can certainly find
~ , ~ , •.. , ~ in R so that R = S[ ~ , ~ , .•• , ~]. Put S = Sand
12 n 12 n a
S. = S[~ , ~ , •.• ,
1 1 2 1
n
so that S = Rand S. = S. l[t]. Of course,
n 1 1- 1
each S. is a finitely generated subalgebra of R. We use 9J1 (S.) to denote
1 1
the set of all maximal ideals of Sr We recall that if K is algebraically
closed, then 9J1 (S.)
--- 1
is the same as the set of all K-rational maximal
ideals of S ..
1

Lemma 8. Suppose that fER, f "* O. Then there exists g E S,

g"* 0 with the following property: given a prime ideal P, in S, such that
g f P there can be found a prime ideal II in R for which (i) II n S = P,
(ii) f f II, and (iii) dim II = dim P + Dim R - Dim S. Such a g has the
further property that if N E 9]1(S) and g f N, then there exists M E 9J1(R)
satisfying f f M and M n S = N.

Proof. For the time being we shall ignore the final assertion.
Suppose that for each i (1:5 i :5 n) we can establish the other assertions
when Rand S are replaced by S. and S. 1 respectively. Evidently
1 1-
this part of the lemma will then follow. Accordingly we shall suppose that
R = S[ n The element f can now be expressed in the form

f = d 0"tt + d1 .,tt-1 + ••• + dt ,

where d. E Sand d "* O.

1 a
First assume that ~ is transcendental over the quotient field of S
so that Dim R - Dim S = 1. In this instance we may take g = d.
a
For
if P is a prime ideal of S not containing d , then II = PS[~] is a prime
a
ideal of R with the required properties.
Next assume that ~ is algebraic over the quotient field of Sand
therefore Dim R= Dim S. There exists b E S, b"* 0 such that b~ = Tf

85
(say) is integral over S. Choose an integer m > 0 so that bmf =h
(say) belongs to S[ 11]. Then we have a relation

hq + c hq- 1 + c hq - 2 + ... + cq = 0,
1 2

where c , c , .•• , c belong to S. Let us arrange that q is as small

1 q
2
as possible. Then, because h 0, we have cq o. '* '*
'*
Put g = c b. Then g € S, g O. Now suppose that g 1 P, where
P is a prime id~al of S. Because t S[ 11] is an integral extension of S,
there exists a prime ideal Q in S[ 11] such that Q n S = P. Since
S[lI]!Q is an integral extension of SiP, we have dim Q = dim P. Also
b 1 Q and c 1 Q. From c 1 Q it follows that h 1 Q.
q q
Denote by T the subring of the quotient field of S[ 11] formed by
II
all elements Alb , where A € S[ 11] and II ~ 0 is an integer. Then

and QT is a prime ideal of T which contracts to Q in S[ 11 ]. Put

II = QT n R. This is a prime ideal of R which contracts to P in S.
Also f 1 II because h 1 II. Next

and the quotient field of S[ 11 ];Q is the same as the quotient field of T;QT.
It follows that dim II = dim Q = dim P. Consequently II has all the
required properties, and (as already explained) the first part of the lemma
follows in full generality.
Finally assume that N € 9Jl (S) and g 1 N. We know there exists a
prime ideal II, of R, such that f 1 II and II n S = N. But II is the
intersection of all the members of 9Jl (R) that contain it. Thus there
exists M € 9Jl(R) for which II ~ M and f 1 M. But now M n S = N
because N is a maximal ideal of S. This completes the proof.

Lemma 9. Let II be a prime ideal of R and P a prime ideal

of S. Suppose that II n S = P and that II is a minimal prime ideal of

t See [(9) Theorem 8, p. 90].

86
PRo Then dim II 2: dim P + Dim R - Dim S.

Proof. Put q = Dim S and v = dim P. We can choose

xl' x 2 ' ••• , Xv in S so that their natural images in SiP are alge-
braically independent over K. Then X , x , ... , x are themselves
I 2 V
algebraically independent over K and, by forming fractions with respect
to the non-zero elements of K[x , x , ... , x ] and replacing K by
I 2 v
K(x , x , ••. , x ), we can reduce the problem to the case where dim P=O.
I 2 V
Accordingly in what follows it will be assumed that P is a maximal ideal.
By the normalization theorem t there exist q = Dim S elements
fJ , fJ , ••• , 1J , in S, which are algebraically independent over K and
I 2 q
are such that S is an integral extension of K[ 1J , 1J , ••• , 1J ]. Then
I 2 q
P n K[ 1J I' 1J 2' ••• , 71 q ] is a maximal ideal of K[ 1J I' 1J 2' ••• , 1J q ] and
therefore it is possible to find q elements, say ~l' ~2' ••• , ~q' which
generate it. ~ We claim that II is a minimal prime ideal of
(~ , ~ , ••• , ~)R. For let II be a prime ideal of R which satisfies
I 2 q 0

Then 110 n S and II n S are prime ideals of S which have the same
contraction in K[ 1J I' 1J 2' ••• , 1J q] and the former is contained in the
latter. But this implies that

II nS=lIns=p
o

and hence that PR ~ II ~ II. Thus II = II and our claim is established

o § 0
But what we have just proved shows that

dim II 2: Dim R - q

and so the lemma follows.

t See [(13) Vol. 2 Theorem 25, p. 200].

~ See [(9) Theorem 3, p. 281].
§ By [(9) Theorem 22, p. 217] we have rank 1I:s q, and
rank II + dim II = Dim R by [(13) Vol. 2 Theorem 20, p. 193].

87
 Lemma 10. There exists fER, f"* 0 with the following property:
if II is a prime ideal of R and f I II, then dim II ::s dim(II n 8) +
Dim R - Dim 8.

Proof. We continue to use the notation introduced just before

Lemma 8 so that, in particular, R = 8[ ~l' ~ , ••• , ~ ]
2 n
and
8.
1
= 8[ ~ 1 , ~
2
, .•. , ~.]. We use I to denote the set of integers i,
1
between 1 and n, for which ~. is algebraic over the quotient field of
1
8.1- l'
Let i E I. There exists a non-null polynomial p.(X) with coefficient
1
in 8. 1 and such that p. (~.)
1- 1 1
= O. We let f. denote the leading coefficient
1
of p. (X). Also if 1::s j ::s nand j I I we put f. = 1. Thus f = f f ... f
1 J 1 2 n
belongs to Rand f"* O.
Now assume that II is a prime ideal of R for which f I II and
put lIt = II n 8t . Then 8elI t = (8t _l ;11t_l)[1t ], where ~t is the residue
of ~t modulo lIt' and therefore

dim lIt ::s dim II t _ l + 1.

Now suppose that i E I. In this case the leading coefficient of p. (X) is

1
not in II. 1 and therefore ~. is algebraic over the quotient field of
1- 1
8.~ 1;11.~ l' Accordingly dim II.1 = dim II.~ l' Thus to sum up

if tEl,
dim lIt - dim II t _ l =
if tn.
The inequality dim II - dim(II n 8) ::s Dim R - Dim 8 follows by addition.
We shall now reformulate the substance of these lemmas in the
language of affine sets. Note that if ¢: V - W is an almost surjective
K-morphism of affine sets and V is irreducible, then W is irreducible
as well. This follows from Theorems 1 and 4 and the fact that ¢ is con-
tinuous.

Theorem 33. Let ¢: V-W be an almost surjective K-morphism of

affine sets and suppose that V is irreducible and K is algebraically
closed. If now U is a non- empty open subset of V, then ¢(U) contains

88
a non-empty open subset of W.

Proof. By Theorem 32, the associated homomorphism

¢* : K[W] - K[V], of K-algebras, is an injection. Hence if we put
R = K[V], S = K[W], then R is an integral domain and S is a subalgebra.
Also Dim R = Dim V and Dim S = Dim W. Next, because every closed
subset of V is an intersection of principal loci, 'we can find fER, f "* 0
such that

U = {xix EV and f(x)"* oj

o
is a non-empty open subset of V contained in U. Choose g E S, g"* 0
as in Lemma 8. Then

{y Iy EWand g(y) "* 0 J

is a non-empty open subset of Wand, by the final assertion of Lemma 8,

it is contained in ¢(U 0). Since I/>(U 0) h ¢(U), the theorem is proved.

Corol1ary. Suppose that f E K[V], where V is an affine set and K

is algebraically closed. If f takes infinitely many different values on V,
then K\f(V) if a finite set.

Proof. By replacing V by a suitable irreducible component, we

may suppose that V itself is irreducible. Let us regard K as being
affine I-space. Then f: V - K is a K-morphism and, because f(V) is
infinite, it is almost surjective. By Theorem 33, f(V) contains a non-
empty open subset of K, and now the corollary follows.

Theorem 34. Suppose that I/> : V -W is almost surjective, that

V is irreducible and that K is algebraically closed. Suppose also that U
is a non-empty open subset of V. In these circumstances there exists a
non- empty open subset T, ~ W, with the following pr operty: given any
irreducible closed subset Y, ~ W, which meets T there is a closed
irreducible subset X, of V, such that X meets U, I/>(X) h Y, and
Dim X = Dim Y + Dim V - Dim W.

89
 Proof. Put R = K[V], S = K[W] and regard S as a subalgebra
of R. Choose f E K[V], f *- 0 so that

uo = {x Ix E V and f(x) *- 0 }

is contained in U. Without loss of generality we may suppose that U =U I

in the subsequent discussion.
Choose g E S, g*-O as in Lemma 8 and put

T = {y lYE Wand g(y) *- 0 J.

Then T is a non-empty open subset of W. Now assume that Y is an

irreducible closed subset of W which meets T.
To Y corresponds a prime ideal P in S and we have g f- P
because Y meets T. Consequently, by Lemma 8, there exists a prime
ideal II, of R, such that f f- II, II n S = P, and

dim II = dim P + Dim V - Dim W.

Since K is algebraically closed, there is a closed irreducible subset

X, of V, 'which corresponds to II. This has the required properties.

Theorem 35. Let I/J: V -W be an almost surjective K-morphism,

where V is irreducible and K is algebraically closed. Let X be an
irreducible closed subset of V and Y an irreducible closed subset of W.
If now X is a component of ¢-l(y) and ¢(X) = Y, then

Dim X :::: Dim Y + Dim V - Dim W.

Proof. Once again put R = K[V], S = K[W] and regard S as a

subalgebra of R. To X there will correspond a prime ideal II in R
and to Y a prime ideal P in S. Put P' = II n S. Then, because
¢(X) ~ Y, we have P ~ P'. Let Y' be the irreducible closed subset of
W that corresponds to P'. Then ¢(X) ~ Y' ~ Y, whence Y' =Y because
cp(X) = Y. Thus II n S = P and hence PR ~ II.
Suppose that II' is a prime ideal of R satisfying PR ~ II' ~ II
and let X' be the irreducible closed subset of V that corresponds to II'.
Then X ~ X' and I/J(X') ~ Y because II' n S = P. It follows that X = X'

90
and therefore II' = II. Accordingly II is a minimal prime ideal of PR
and so

dim II ~ dim P + Dim R - Dim S

by Lemma 9. But this is equivalent to

Dim X ~ Dim Y + Dim V - Dim W

and now the theorem is proved.

Theorem 36. Let cp: V"'W be an almost surjective K-morphism,

where V is irreducible. Then there exists a non-empty open subset U,
of V, such that if X is any irreducible closed subset of V which meets
U, then Dim X ::5 Dim( cp(X)) + Dim V - Dim W.

Proof. We put R = K[V], S=K[W] and regard S as a subalgebra

of R. Choose fER, f,* 0 as in Lemma 10 and set U = {xix E V and
f(x) '* 0 }.
Now suppose that X is an irreducible closed subset of V that meets
U. To X there will correspond a prime ideal II of R and, since X
meets U, we have f t II. Next, to ,/>{X) corresponds a prime ideal P,
of S, and each rational maximal ideal of R that contains II contracts
to a rational maximal ideal of S that contains P. But (Theorem 12) II
is the intersection of all the rational maximal ideals that contain it.
Consequently P ~ II n S and therefore, by Lemma 10,

dim II ::5 dim(II n S) + Dim R - Dim S

::5 dim P + Dim V - Dim W

and so the theorem is proved.

Theorem 37. Let cp: V"'W be an almost surjective K-morphism

of affine sets, where V is irreducible and K is algebraically closed.
In these circumstances there exists a non-empty open subset T, ~ W,
with the following property: for any closed irreducible subset Y (of W)
which meets T, ,/>-1 (Y) has a component whose dimens ion is equal to

91
Dim Y + Dim V - Dim W.

Proof. Choose a non-empty open subset U, of V, as in Theorem

36 and then a non-empty open subset T, of W, as in Theorem 34. Now
suppose that Y is a closed irreducible subset of W which meets T.
By Theorem 34, there exists a closed irreducible subset X, of V,
which meets U, is contained in 1/>-1 (Y), and satisfies

Dim X = Dim Y + Dim V - Dim W.

Choose a component X' of 1/>-1 (Y) so that X ~ X'. Then X' meets U
and therefore, by Theorem 36,

Dim X' :::: Dim(I/>(X') + Dim V - Dim W

:::: Dim Y + Dim V - Dim W
= Dim X.

Accordingly, by Theorem 22, X = X' and the theorem is proved.

Corollary. Suppose that yET. Then 1/>-1 ( {y }) has a component

whose dimension is equal to Dim V - Dim W.

It will be recalled, from section (2.6), that a K-morphism can be a

bijection without being a K-isomorphism. Theorem 33 will now be used
to throw some light on this situation.

Lemma n. Let cfJ : V .... W be an almost surjective K-morphism

which is also an injection. Assume that V ~ W are irreducible and
that K is algebraically closed. Then Dim V = Dim W.

Proof. By hypothesis K[V] is an integral domain and K[W] may

be regarded as a subalgebra. We wish to show that K(V) is an algebraic
extension of K(W). We shall assume that it is not and seek a contradiction.
Our assumptions ensure that there exists T E K(V) which is trans-
cendental over K(W); indeed we can choose T so that it lies in K[V].
Then

K[W] ~ (K[W])[T] ~ K[V].

92
Next (K[W])[T] may be regarded as the coordinate ring of W x K and
then the inclusion homomorphisms K[W] - (K[W])[T] and
(K[W])[T] - K[V] correspond respectively to the projection 11:W x K-W
and a certain K-morphism l/I: V -W x K. Furthermore I/J = 11 0 l/I.
By Theorem 32, l/I is almost surjective and therefore, by Theorem
33, l/I(V) contains a non-empty open subset U of W x K. Let (w , k )
o 0
belong to U and consider the K-morphism K- -W x K in which
k 1-+ (w 0' k). The inverse image of U is a non-empty open subset of K
and therefore it contains infinitely many elements. In particular we can
find k
1
E K so that k
10
*k and (w , k )
01
E U. Thus (w , k ) and
00
(w , k ) are distinct points of l/I(V) having the same projection on to W.
o 1
This, however, contradicts our original assumption that I/J = 11 0 l/I is an
injection.
The hypotheses of the lemma can be relaxed a little. This is shown
by

Theorem 38. Let ¢: V -W be an almost surjective K-morphislll

of affine sets which is also an injection. If now K is algebraically closed,
then Dim V = Dim W.

Proof. Let VI' V 2' ••• , Vr be the irreducible components of V

(we may assume that V is non-empty) and let I/J(V.) denote the closure
J
of ¢(V.) in W. Then C/J induces an almost surjective K-morphism
J

¢j : Vj ... ¢jWji
which is also an injection. Consequently, by Lemma 11,

Dim V. = Dim (I/J (V .)).

J J
Now ¢(V 1 ), I/J{V ), ... , ""f(\C) are closed irreducible subsets of Wand,
2 r
because ¢(V) = W, their union is W. Accordingly the irreducible com-
ponents of Ware the maximal members of

{~(V~"), ¢(V), ... , I/J{V r) }

and therefore it follows, in view of Theorem 22, that

93
 Dim W = max Dim{¢(V.))
l:sj:sr J
= max Dim V.
J
= Dim V.

94
4 . Derivations and tangent spaces

General remarks

In this chapter the concept of a derivation is introduced. This will

provide an important new tool which is useful in studying local properties
of affine sets.
As before K denotes a field. Unless there is an explicit statement
to the contrary, no special assumptions are made concerning K.
Suppose that V is an affine set, defined over K, and that x is a
point of V. As in section (3.3) we use QV
,x to denote the local ring of
V at x. By Chapter 3 Theorem 19, CL has only one maximal ideal.
-VV,x
In this chapter the maximal ideal is denoted by ---Y,
~-
x.

4. 1 Derivations in algebras

Throughout section (4. 1) we use R, S, T to denote K-algebras. It

will be assumed that each of them is unitary, associative and commutative.
Let q,: R - S be a homomorphism of K-algebras. By a K-deriva-
tion of type q" of R into S, we shall understand a K-linearmapping
D : R - S with the property that

D(XY) = (Dx).p(y) + .p(x)(Dy) (4. 1. 1)

for all x, y in R. The set of all such derivations of R into S will be

denoted by DerK(R, S, q,).
There are various situations where the notation can conveniently be
simplified. For example, if R is a subalgebra of S we put

(4. 1. 2)

where j: R - S is the inclusion homomorphism. We also set

95
 (4.1.3)

where, this time, j : R - R is the identity homomorphism.

Another case where the notation can be simplified arises as follows.
Let Q be a K-algebra (unitary, associative, and commutative) which has
exactly one K-rational maximal ideal. In this case there is only a single
possible homomorphism Q - K of K-algebras and therefore we can write
DerK(Q, K) without risk of ambiguity.
Let us return to the general situation. If Dl' D 2 belong to
DerK(R, S, 1/1) and k E K, and if we define Dr + D2 and kDl in the
obvious manner, then these too belong to DerK(R, S, cp). In fact we have

Lemma 1. Let Ho~(R, S) be regarded as a vector space over K

Then DerK(R, S, 1/1) is a subspace.

Assume that D E DerK(R, S, 1/1) and s E S. Define

sD: R-S (4. 1. 4)

by

(sD)x = s(Dx). (4.1.5)

Then sD E DerK(R, S, 1/1). Bearing this in mind we at once obtain

Lemma 2. Let the notation be as above. Then (4.1.4) endows

DerK(R, S, 1/1) with a natural structure as an S-module.

It should be noted that the S-module structure of DerK(R, S, 1/1) is

compatible with its structure as a K-space.

Lemma 3. If D E DerK(R, S, 1/1), then D(k1 R ) = 0 for all k E K.

Proof. It is enough to show that D1R = O. ThiS, however, follows

from

DIR = D(l R 1R )
= l/I(lR )(D1 R) + (D1 R)I/I(lR )
= D1R + Dl R •

96
 When R is non-trivial K may be regarded as a subfield of R.
Lemma 3 then states that every K- derivation of R into S vanishes on K.
In this connection we note

Lemma 4. Suppose that R is non-trivial and that D: R ..... S is a

mapping which satisfies

D(x + y) = Dx + Dy,
D(XY) = (Dx)¢(y) + ¢(x)(Dy),

for all x, y E R. Then D E DerK(R, S, ¢) if and only if D vanishes

on K.

Proof. Assume that D(k) = 0 for all k E K. Then, for x E R,

D(kx) = ¢(k)(Dx) + (Dk)¢(x) = k(Dx)

and therefore D E DerK(R, S, ¢). The converse follows from Lemma 3.
It is sometimes necessary to decide when two derivations coincide.
For this the following result is often useful.

Theorem 1. Let D, D' E DerK(R, S, ¢) and let A be a subset of

R which generates R as a K-algebra. If now Da = D'a for all a E A,
then D = D'.

Proof. ... , a n belong to A and let k E K. It is

clear that

D(ka1a •.• a ) = D'(ka a ••• a ).

2 n 1 2 n

Since a typical element of R is a sum of elements such as kala 2••• an'

the theorem follows.
The next lemma embodies the same idea as Theorem 1 but it is
more explicit.

Lemma 5. Let R = K[ ~l' ~2' ••• , ~n] and suppose that

z = f(~l' ~2' ••• , ~ ), where
n --
f(X, X 2 ,
1
••• , X ) belongs to the poly-
n
nomial ring K[X , X , ••. , X]. Then
1 2 n----

97
 Dz =.~ ~ [a~.\ (DU
1
1=1 \ I}
for any D in DerK(R, S, ~).

This is obvious.
It will now be shown how, from given derivations, we can sometimes
derive new ones. To this end let .p: R ... Sand 1/1: S ... T be homomor-
phisms of K-algebras and let D € DerK(R, S, ¢). It is easily verified
that 1/1 0 D belongs to DerK(R, T, 1/1 0 CP).

Theorem 2. Let the situation be as described above. Then the

mapping

in which D t-+ 1/1 0 D, is K-linear.

We leave the verification to the reader.

Let .p: R ... Sand 1/1 : S ... T be as before and suppose that
Ll € Der K(S, T, 1/1). It is eal'lily checked that (i) Llo ~ belongs to
DerK(R, T, 1/1 0 CP), (ii) the new derivation vanishes on Ker t/J, and
(iii) the mapping

(4. 1. 6)

in which Ll t-+ Ll 0 .p, is K-linear.

Theorem 3. Let the situation be as described in the preceding

paragraph and suppose that the homomorphism ~: R ....S is surjective.
Then the K-linear mapping

described in (4.1.6) establishes a bijection between DerK(S, T, 1/1) and

the subspace of DerK(R, T, 1/1 0 ~) formed by the derivations which vanish
on Ker cpo

98
 Proof. The fact that cp is surjective obviously ensures that
(4.1.6) is an injection. Now suppose that D belongs to DerK(R, T, l/J 0 cp)
and that it vanishes on Ker cpo Then D gives rise to a K-linear mapping
L1 : S - T which is such that L1(cp(r)) = Dr for all r E R. A simple veri-
fication shows that L1 is in DerK(S, T, 1/1) and, by construction,
D = L1 0 cpo Accordingly (4. 1. 6) is also a surjection and the proof is
complete.
Our next result has to do with the formation of fractions. For the
moment we assume that our K-algebras R, S are integral domains; in
addition we suppose that ~ is a non-empty multiplicatively closed subset
of R not containing zero, and that Q is a non-empty multiplicatively
closed subset of S also not containing the zero element. Denote by R~

the set of all elements belonging to the quotient field of R that can be
expressed in the form r la, where r E R and a E ~; and define SfG
similarly. Evidently R~ and SQ may be regarded as K-algebras. More-
over R is a subalgebra of R~ and S a subalgebra of SQ'
Now suppose that cp: R - S is a homomorphism of K-algebras and
that it satisfies

(4.1.7)

Then cp extends to a homomorphism

(4. 1. 8)

of K-algebras in which, with a self-explanatory notation,

;j;(.!:) = cp(r) (4. 1. 9)

'I' a' cp{a)'

Next if D E DerK(R, S, cp), then it is possible to define a mapping

(4.1. 10)

by means of the formula

D(~ = l/>(a)(Dr) - l/>(r)(Da) (4. 1. 11)

[cp(a)]2

99
and a straightforward verification shows not only that 15 EDerK(R~, S{2' ~),
but also. that 15 agrees with D on R and is the only member of
DerK(R~, S,,' ~) to do so. For future reference we record

Lemma 6. Let R, S, </>, ~ and ~ be as above. Then the

mapping

which results from extending D in DerK(R, S, </» by means of the

formula (4.1. 11), is K-linear and an injection.

Corollary. In the special case where S is a field the mapping

described in the lemma, is an isomorphism of K-spaces.

It should be noted that in this instance S~ coincides with S.

Once again let </> : R .... S be a homomorphism of K-algebras, but
now let L be an extension field of K. Then, as in section (1.6),
q,L : RL .... SL is a homomorphism of L-algebras. Also if D : R .... S is
a K-linear mapping, then DL : RL .... SL is L-linear. Note that if
a E R and A E L, then q,L(aA) = Aq,(a) and DL(Aa) = A(Da).

Lemma 7. Suppose that D belongs to DerK(R, S, </». Then DL

L
belongs to DerL(R , S , q, ).
L L

L
Proof. Let x, y E R • Then we can write

X=Aa +Aa + ••• +A a ,

11 22 mm
y = A'b + A'b + ••• + A'b
1 1 2 2 n n'
where A., A! ELand a., b. E R. Next
1 J 1 J

DL(xy) = LL A.A!D(a.b.)
ij IJ IJ
= LL A.(Da.)A!q,(b.) + LL A.q,(a.)A!(Db.)
ij 1 1 J J ij 1 1 J J
= (DLx)q,L(y) + q,L(x)(DLy).

100
Since DL is L-linear, this completes the proof.

4.2 Examples of derivations

We pause, fora moment, to illustrate some of the general ideas

introduced in the last section. As before R, S denote K-algebras which
are assumed to be unitary, associative, and commutative.

Example 1. Let Xl' X2 ' ••• , Xn be indeterminates. Then, for

each i (1::::: i ::::: n),

a
ax. : K[Xl , X2 ' ••• , Xn ] ... K[Xl' X 2 ' ••• , Xn]
1

belongs to DerK(K[X l , X 2 ' ••• , Xn]).

Example 2. Let Xl' X 2 ' ••• , Xn be indeterminates and suppose

that

is a homomorphism of K-algebras. By Example 1 and Theorem 2,

a
'\ = t/I ax.
0 (4.2.1)
1

is a member of Der K(K[X l , X2 ' ••• , Xn], S, t/I). Next let aI' a 2' ••• , an
belong to S. Then

belongs to DerK(K[X l , ••. , Xn ], S, t/I) and it is characterized by the

property that

(4. 2. 2)

for i = 1, 2, ... , n. This establishes

Theorem 4. The S-module DerK(K[X1 , Xn ], S, t/I) is free and

••• ,

if '\ is defined as in (4.2.1), then AI' A 2 , ... , An form a base of this

module.

101
 As a matter of notation we recall that if l/I(X.) = TJ., then it is usual
I I
to write

a a (4.2.3)
l/I 0 oX i = oTJ i
Thus 0/01/ , 0/01/ , ••• , 0/01/ is a base for the S-module
I 2 n
DerK(K[XI , ••• , Xn ], S, l/I).

Example 3. We next consider the ease of a finitely generated

K-algebra. Suppose then that

(4. 2. 4)

Let X , X , •.. , X be indeterminates and denote by f) the surje.ctive

I 2 n
homomorphism

(4.2.5)

of K-algebras in which O(X.) = ~.. Put

I I

~= Ker O. (4. 2. 6)

Then, since K[X I , X 2 ' ••• , Xn] is a Noetherian ring, we can find a
finite set of polynomials, say

F. (X , X , •.. , X ) where 1 ~ j ~ m,
J I 2 n

which generates ~.

Now suppose that I/J : R - S is a homomorphism of K-algebras and

put

I/J(~.) = x .• (4.2. 7)
I . I

Then Example 2 shows that

is a free S-module having o/Ox I ' 0/Ox 2 ' ••• , o/Oxn as a base. Hence
if A belongs to

102
 DerK(K[X1 , x 2 , ... , x n ], s, cp 0 e),

then there exist unique elements a l' a 2' ••• , an in S such that

+a -
o (4.2. 8)
nOx
n
and, for every F in K[X , X , ••. , Xn ],
1 2
of of
~F = a1 Ox
1
+a
2
ax2
+

Consider the equations

of of
_la +_la
Ox 1 1 Ox 2 2
of of
_2a +_2a (4.2.9)
Ox 1 Ox 2
1 2

of of of
~a
ox1 1 + ':>.~2 Vl\.
a2 + ... + -ax-
ma - 0
n-
n

Lemma 8. Suppose that

f:j.=a - + a - +
o 0
+a -
o
lOx 1 20x 2 nOx n
where aI' a 2' ••• , an belong to S. Then f:j. vanishes on III if and only
if the equations (4.2.9) are satisfied.

Proof. Each F in III can be expressed in the form

where G.
J
E K[X , X , .•• , X]. Since F .(x , x , .•• , x )
12 n J12 n
= 0, we see
that f:j.F =0 for all F E III if and only if f:j.F. = 0 for j = 1, 2, ••• , m.
J
The lemma follows.
In the next theorem aI' a 2' ••• , an continue to denote elements of
S.

103
 Theorem 5. There is a bijection between the set of solutions of
the equations (4. 2. 9) and DerK(R, S, cp). This is such that if
D E: DerK(R, S, cp), then the corresponding solution of the equations
(4.2.9) is given by D~. = a. (1 5 i 5 n).
1 1

In view of our previous remarks, this is an immediate consequence

of Theorem 3.
These observations can be taken a stage further. We know that
DerK(R, S, cp) is an S-module. Now the solutions, in S, of the equations
(4. 2. 9) also form an S-module. Let us call this the solution module of
the equations. We then have

Theorem 6. The bijection of Theorem 5 is an isomorphism between

the S-module DerK(R, S, cp) and the solution module of the equations
(4. 2. 9) when these are regarded as equations to be solved in S.

Example 4. Let X , X , •.. , X continue to denote indeterminates

1 2 n
and let L be an extension field of the quotient field K(X , X , ... , X )
1 2 n
of K[X , X , ... , X ]. Then K[X , X , ... , X ] is a subalgebra of L
1 2 n 1 2 n
and, by Theorem 4,

is a vector space over L and as such it has the n derivations a/ax. as

1
a base.
Next our discussion of derivations in relation to fractions shows that
each member of DerK(K[X 1 , Xn ], L) has a unique extension to a
••• ,

member of DerK(K(X , •.. , X ), L). The extension of a/ax. is denoted

1 n 1
by the same symbol. We now see, from Lemma 6 Cor., that these n
extended derivations form a base for DerK(K(X , ..• , X ), L) considered
1 n
as a vector space over L.

4. 3 Derivations in fields

We must now make a special study of derivations in the situation where

the algebras involved are fields. Accordingly throughout section (4. 3) L
will denote an extension field of K.

104
 Suppose that K* is a field between K and L and that
D E DerK(K*, L). If now X is an indeterminate and g(X) E K*[X], then
by gD(X) will be understood the polynomial (with coefficients in L) that
is obtained by applying D to the coefficients of g(X). Note that if
g(X) = p(X)q(X), where p(X), q(X) are in K*[X], then

(4. 3. 1)

Again the notation g'(X) will be used for the formal derivative of g(X).
Hence, in addition to (4.3.1), we have

g'(X) = p'(X)q(X) ~ p(X)q'(X). (4.3.2)

Let ~ ELand consider the possibility of extending D, in

DerK(K*, L), so as to yield a member of DerK(K*W, L).

Lemma 9. Let K, K*, L and D be as described above. Suppose

that ~ belongs to L, is algebraic over K*, and has f(X) as its irredu-
cible polynomial over K*. Finally let A E L. Then in order that there
should exist ~ E DerK(K*W, L) which (i) extends D, and (ii) satisfies
~~ = A, it is necessary and sufficient that

~W + f'(~)A = O. (4. 3. 3)

If ~ exists, then it is unique.

Proof. If ~ exists, then, by applying it to the equation f(~) = 0,

we immediately obtain

~w + f'(~)A = O.

From noW on we shall assume that (4. 3.3) holds.

Suppose that h(X) E K*[X] and h(~) = O. Then h(X) = f(X)p(X) for
some p(X) E K*[X] and hDW = fD(~)p(~), h'W = f'(~)p(~). Accordingly

hDW + h'(~)A = O.
It follows that if g (X), g. (X) belong to K*[X] and g W = g
1 2 1 2
W, then

105
and therefore we can define a mapping ~: K*[~] -+ L by means of the
formula

But K*[~] = K*(~). Also it is a simple matter to verify that

L\ E DerK(K*W, L) and satisfies conditions (i) and (ii). Finally it is
obvious that at most one member of DerK(K*W, L) can have these
properties.

Corollary 1. Let K, K*, L, D be as above and suppose that ~ EL

and is algebraic over K*. Then the following statements are equivalent:
(a) D has exactly one extension which is a member of
DerK(K*W, L);
(b) ~ is separable over K*.

Proof. We use the same notation as in the proof of the lemma.

The number of extensions of D is the same as the number of solu-
tions of

~W + f'W>t = 0,

where >t has to belong to L. Hence if ~ is separable over K*, that is

if f'(~) "* 0, then D has exactly one extension. However, if f'(~) = 0,
then there may be no solution; but if there is a solution, then there will
be more than one.

Corollary 2. Let ~ E L and be algebraic over K. Then the L-

space DerK(K(~), L) has dimension zero if ~ is separable over K,
whereas it has dimension one if ~ is not separable over K.

Proof. In Corollary 1 take K* = K. Certainly DerK(K, L) = °

and if ~ is separable over K, then the zero derivation of K into L has
only the zero extension in DerK(K(~), L). On the other hand, if ~ is not
separable over K, then Lemma 9 shows that, for every >t E L, there is
a unique ~ in DerK(K(~), L) such that ~~ = >to

106
 In the next theorem, K* denotes a field between K and L, and
~l' ~2' ••• , ~nbelong to L.

Theorem 7. Suppose that ~ , ~ , .•. , ~ are all algebraic and

1 2 n
separable over K*. Then each D in DerK(K*, L) has a unique extension
to a member of DerK(K*(~ , ... , ~ ), L).
1 n
This follows from successive applications of Lemma 9 Cor. 1.

Theorem 8. Let ~ , ~ , ... , ~

-- 1 2 n belong to L and be algebraic
over K. Then

if and only if K(~ , ~ , •.. , ~ ) is a separable extension of K.

1 2· n

Proof. If K( ~ ., ~ , ..• , ~ ) is separable over K, then

1 2 n
DerK(K(~ , •.. , ~ ), L) = 0 by virtue of Theorem 7. Now suppose that
1 n
K( ~ , ~ , ..• , ~) is not separable over K. Then K has characteristic
1 2 n
p, where p is a prime. Let K' be the separable closure of K in
K(~
. 1
, ~ , ... , ~). Then K'
2 n·
"* K(~
!J 1
, ~
2
, •.. , ~ ) and there exists a
n
positive integer !J such that ~~ E K' for i = 1, 2, ... , n.
1
Let j be the smallest integer such that

K'(~l' ~2' ••• , ~j)=K(~l' ~2' ••• , ~n)

andput K*=K'(~l' ~2' ••• , ~j-l)' Then K*(~j)=K(~l' ~2' ••• , ~n)
and ~j is not separable over K*. Accordingly

is not zero by Lemma 9 Cor. 2. It follows a fortiori that

DerK(K(~ , ... , ~ n ), L) "*
1
o.
Suppose that, ~l' ~2 , ••• , ~n belong to L but are not necessarily
algebraic over K. Let the transcendence degree of K(~l' ~2' ••• , ~n)

over K be r.

107
 Definition. We say that K(~ , ~ , ... , ~ ) is a 'separable'
1 2 n
extension of K if there is a transcendence base TJ , TJ , ••• , TJ , for
1 2 r
K(~ , ~ , ... , ~ ) over K, such that K(~ , ~ , ... , ~ ) is a separable
1 2 n 1 2 n
algebraic extension of K(1] , TJ , ••• , TJ ).
1 2 r
Such a transcendence base is called a separating transcendence
base. Thus K(~ , ~ , ... , ~ ) is separable over K if and only if a
1 2 n
separating transcendence base exists. Of course if K has characteristic
zero, then K( ~ , ~ , ... , ~
1 2 n) is necessarily a separable extension of K.

Lemma 10. t Let

---
~l' ~ 2 , ••• , ~n belong to L and let
K( ~l' ~2' ••• , ~n) be a separable extension of K. Then the dimension of
DerK(K(~ , ... , ~ ), L), considered as a vector space over L, is equal
1 n
to the transcendence degree of K( ~1 ~ , ... ,
, 2 ~ ) over K.
n----

Proof. Let TJ l' TJ 2' ••• , 1] r be a separating transcendence base.

By Example 4 in section (4.2), the r derivations

form a base for the L-space DerK (K(TJ 1 , ••• , TJ r ), L) and, by Theorem 7,
%TJ. has a unique extension, D. say, to a derivation of K(~ , ~ , ... , ~ )
1 1 1 2 n
into L. It is clear that each member of DerK(K(~ , ... , ~ ), L) has a
1 n
unique representation in the form

AD +AD + ... +AD,

1 1 2 2 r r

where A. E L. The theorem follows.

1

Theorem 9. Suppose that ~l' ~ , ..• , ~ belong to L and that

2 n
DerK(K(~l' ..• , ~n)' L) = O. Then each ~i is separable and algebraic
over K.

Proof. By Theorem 8, it is enough to prove that the tare

1
algebraic over K. Assume the contrary and let 1] , TJ , ••• , 1] be a
1 2 r
transcendence base for K(~ , ~ , •.. , ~ ) over K. Then r ~ 1 and
1 2 n

t The converse is also true. See Theorem 12.

108
each ~i is algebraic over K( 1J l' 1J 2' ••• , 71 r)· Now K( ~1' ~2' ••• , ~n)
cannot be separable over K(71 1 , 1] 2 , ••• , 1J r ) = K* (say) for otherwise
Lemma 10 gives a contradiction. Hence, by Theorem 8,

is non-zero and therefore a fortiori

Thus in any event we arrive at a contradiction.

Theorem 10. Let ~ , ~ , .•• , ~ belong to L and let

-- 1 2 n
F .(X , X , ... , X ), where 1:s j :s n, belong to K[X , X , •.. , X ].
]12 n-- 12 n
If now F.(~ , ~ , •.. , ~ ) = 0 for all j and the determinant
---J12 n ---
I of./o~.1
] 1
is not zero, then K(~ , ~ , ••• , ~ ) is a separable algebraic extension
1 2 n
of K.

Proof. By Theorem 9, it will suffice to show that

DerK(K(~l' ••. , ~n)' L) = O. Suppose then that D is a derivation of
K( ~1' ~ 2' ••• , ~n) into Lover K. The theorem will follow if we show
that D~. = 0 for all i. However
1

n of.
L at-D~. = 0
i=l "i 1

for j = 1, 2, ••• , n and therefore D~.

1
=0 as required.
Once again let ~1' ~ , ••• , ~
belong to L. The polynomials
n 2
F(X1 , X 2 ' ••• , Xn ), with coefficients in K, that satisfy F( ~1' ~2' ••• '~n)=O
form a prime ideal, ~ say, in K[X , X , ... , X]. Let the polynomials
1 2 n
F.(X , X , •.• , X ), where 1:s j :s m, generate ~.
J 1 2 n

Theorem 11. Let the notation be as explained above. Further let

the rank of the matrix I of.J
/o~.II, where 1:s i :s nand 1 =:. j :s m,
l--- --
be
-
n - s and let the transcendence degree of K( ~ 1 , n) - over
- K ~
2
, •.. , ~

be r. Then r:s s. Moreover r = s if and only if K(~ 1 , ~ 2 , ••• , ~ n )

is separable over K.

109
 Proof. By Lemma 6 Cor., DerK(K(~l' ... , ~n)' L) and
DerK(K[ ~ , ~ , ... , ~ ], L) are isomorphic L-spaces. On the other
1 2 n
hand Theorem 6 shows that DerK(K[ ~ , ... , ~ ], L) is isomorphic to
1 n
the space of solutions (in L) of the equations

of. of. of.

o~ J '\ + o~ J A2 + ... +
1
An
2
at n
=0 (l:s j :s m).

Accordingly s is the dimension of DerK(K(~l' ... , ~n)' L) considered

as a vector space over L. Hence, by Lemma 10, if K(~l' ~2' ••• , ~n)
is a separable extension of K, then r = s.
Now let us renumber F l' F 2' ••• , F m and ~l' ~2' ••• , ~n so
that the (n - s) x (n - s) matrix "oF v /o~Il ", where 1:S)J:S n-s and
s+l :s Il :s n, has a non-zero determinant! Then Theorem 10 shows that
K( ~l' ~2' ••• , ~n) is a separable algebraic extension of
K( ~l' ~2' ••• , ~s)· Thus, in particular, r :s s. Moreover if r = s,
then ~ , ~ , ... , ~ must be a separating transcendence base for
1 2 r
K(~ , ~ , .•. , ~
1 2 n) over K and therefore the extension is separable.

Theorem 12. Let

belong to L, -
let- ~l' ~ , •.. , ~
-- 2 n
K( ~ 1 , ~ 2 , ... , ~n ) have transcendence degree rover
---
K, and let
DerK(K(~ , •.• , ~ ), L) be an L-space of dimension s. Then r:S s
1 n --
and there is equality if and only if K( ~ , ~ , ... , ~ ) is a separable
1 2 n
extension of K.

All these assertions were established during the proof of the last
theorem. At the same time we also proved

Theorem 13. Let ~l' ~2' ••• , ~n belong to L and suppose that
K(~l' ~2' ••• , ~n) is separable over K. Then a separating transcendence
base can be chosen from among ~ , ~ , ... , ~ .
1 2 n
The field K can only have non-separable extensions if its character-
istic is a prime p. Suppose that this is the case and let us regard L as
being contained in its algebraic closure L. Put

t The subsequent argument is trivial if s = n.

110
This is a subfield of L containing K.

Lemma 11. Suppose that I; , I; , ... , I; belong to L and are

1 2 n lip
algebraic and separable over K. Then K( I; , I; , ••• , I; ) and K
-- 1 2 n --
are linearly disjoint over K.

Proof. Assume the contrary. Then we can find 11 , 11 , ••. , 11 ,

12m
in K(I; , I; , •.• , I; ), linearly independent over K but not linearly
1 2 n
independent over KI/P. Let us arrange that m is as small as possible.
Then there exist non-zero elements 1', I' , ••• , I' , in Kllp , such
12m
that I'111 1+2
I' 11 +. .. + I' 11
2 mm = 0 and we can fix it that I'm = 1.
Note that I' , I' , ••• , I' are not all in K and therefore
12m
K(y , I' , ••• , I' ) is not a separable extension of K.
12m 1/
By Theorem 8, DerK(K(y , ... , I' ), K p) contains a non-zero
1 1/ m
derivation D (say). Thus Dyo E K P and at least one of
1
Dy , Dy , ••• , Dy 1 is not zero. Naturally Dy = O.
1 2 m- m
Next K(y , ••• , I' , I; , ••• , I; ) is a separable algebraic ex-
1 mIn
tension of K(Yl' ••• , I'm) and therefore, by Theorem 7, D has an
extension which belongs to

where L denotes the algebraic closure of L. (This extension will also

be denoted by D.) Now Theorem 8 shows that D must vanish on
K(I;I' 1;2' ••• , I;n) and, in particular, D11i =0 for i = 1, 2, •.. , m.
Accordingly
m-I
L (DYo)11 0 =0
i=I 1 1

and this contradicts the minimal property of the integer m.

Lemma 12. Suppose that the characteristic of K is a prime p

and let I; , I; , •.• , I;
belong to L. If now K( I; , I; , •.• , I; ) is a
1 n 2 --- 1 2 lip n - -
separable extension of K, -
then
-
K(I; 1 , I; 2 , ••• , I;n ) -
and
-
K are
linearly disjoint over K.

111
 Proof. Let y , y , ••• , y belong to K I / p and be linearly
12m
independent over K. It will suffice to prove that they are linearly inde-
pendent over K(~ , ~ , ••• , ~ ).
1 2 n
Take a separating transcendence base for K( ~ , ~ , •.• , ~ ) over
1 2 n
K, say TJ , TJ , ••• , TJ , and put K
1 2 r 1/ 1
= K(71 1 , 71 , ••• , 71 ). Then
2 r
y , y , ••• , y belong to Kl p and it is clear that they are linearly
12m
independent over K. That y , y , ••• , yare linearly independent
1 12m
over K( ~ , ~ , ••• , ~ ) = K (~ , ~ , ••• , ~ ) now follows from Lemma
12 n 112 n
11.
The lemma just proved has a converse. The full result is contained
in

Theorem 14. Let the characteristic of K be the prime p and let

~l' ~2' ••• , ~n belong to L. Then the following statements are equi-
valent:
(a) K(~l' ~2' . . . , ~n) is a separable extension of K;
(b) K(~I' ~2' ••• , ~n) and K I / p are linearly disjoint over K.

Proof. We shall assume (b) and show that (a) follows. (In view of
Lemma 12 this will be sufficient.) The demonstration uses induction on n
and we begin by observing that for n =0 the result in question is obvious.
From here on it will be assumed that n::=: 1 and that we know that (b)
implies (a) in the case of extensions generated by only n - 1 elements.
Note that we may suppose that ~ , ~ , ... , ~ are not algebraically
1 2 n
independent over K for otherwise there would be no problem.
There exists a non- null polynomial, F(X , X , ... , X ) say, with
1 2 n
coefficients in K such that F( ~ , ~ , ..• , ~ ) = O. In what follows
1 2 n
F(X , X , ... , X ) is to be chosen so that its degree is as small as
1 2 n
possible. Put

We claim that it will suffice to show that at least one of the

F.(X , X , ... , X ) is not nUll. For suppose that F.(X , X , ... , X )
112 n J12 n
is not nUll. Then F.(~ , ~ , .•. , ~ ) *- 0 and therefore ~. is separably
J 1 2 n J
algebraic over K( ~l' ••• , ~j-I' ~j+ l' •.. , ~n)' But induction shows that

112
K(~ , ... , ~. J' ~·.Ll' ... , ~ ) is separable over K and, by combining
1 J-. J ' n
these facts, we conclude that K( ~1' ~ , ••• , ~ ) is separable over K.
2 n
From this point onwards we assume that, for each i,
F. (X , X , ••• , X ) is the null polynomial and we seek a contradiction.
1 1 2 n
Our new assumption ensures that

F(X1 , X 2 , ... , Xn) = [G(X1 , X2, ... , Xn)]p

for some G(X 1 , X 2 , ••• , X ) in K1 / p [X , X , ••• , X]. Choose

n 1 2 n
Y1' Y 2 , ••• , Y s in K1 / p so that they are linearly independent over K
and
S
G(X1 , X 2 ' ••• , Xn) = L y II G)X 1 , X2' ••• , Xn),
11=1

where G (X , X , ••. , X ) E K[X , X , •.• , X]. We can choose 1L so

II 1 2 n 1 2 n
that G(X 1 , X 2 ' ••• , Xn) and GlL (X 1 , X 2 ' ••• , Xn) have the same degree
and this will be smaller than the degree of F(X , X , •.. , X). Con-
I 2 n
sequently G (~ , ~ , •.• , ~ ) "* O. On the other hand G(~ , ~ , •.• , ~ )=0,
1L 1 2 n 1 2 n
because F(~ , ~ , ..• , ~ ) = 0, and Y , Y , ••• , yare linearly inde-
1 2 n 1 2 S
pendent over K( ~ 1 , ~ 2 , .•• , ~).
n
It follows that G (~ ,
1L 1
~
2
, ..• , ~
n
)= 0
and with this we have the desired contradiction.

Corollary. Let ~ , ~ , ••. , ~ belong to L and let

-- 1 2 n
TI l' TI 2' ••• , TIm belong to K( ~1' ~ 2 , ••• , ~).
n- - K( ~1, 2
now
If - ~ , ... , ~n )
is a separable extension of K, then K( TI l' TI , ••• , TI ) is also a --- 2 m
separable extension of K.

This holds regardless of the value of the characteristic of K.

We recall that K(~
, ~ , ••• , ~ ) is called a regular extension
1 2 n
of K if K( ~1' ~ , ••• , ~ ) and the algebraic closure of K are linearly
2 n
disjoint over K.

Theorem 15. Let ~1' ~ , ••• , ~ belong to L and suppose that

-- 2 n
K( ~1' ~2' . . . , ~n) is a regular extension of K. -
Then
-
K( ~1 , ~ 2 , •.• , ~ n )
is a separable extension of K.

113
 Proof. If the characteristic of K is zero, then the assertion is
obvious, whereas if the characteristic is a prime p, then Theorem 15
follows from Theorem 14.
It may be shown t that K{~ 1 , ~ 2 , °
.•. , ~n ) is a regular extension
of K if and only if (i) K( ~ , ~ , .•. , ~ ) is separable over K, and
1 2 n
(ii) K is algebraically closed in K( ~ , ~ , ..• , ~). However we shall
1 2 n
not be using this result.

4. 4 Tangent spaces and simple points

It is time to apply the results of the preceding sections to the theory

of affine sets. Accordingly, throughout section (4. 4), V will denote an
affine set (defined over K) and L will denote an extension field of K.
When x E V we use M
to describe the corresponding rational maximal
x
ideal of K[V] and, as before, QV denotes the local ring of V at x.
,x
On this occasion the unique maximal ideal M QV ,of Q_- will be
x ,x ""V,x
designated by ~-
_0--Y, x•
From time to time some auxiliary notation will be needed. To
avoid tedious repetition we shall explain it here once for all.
The notation arises in the following way. We choose ~l' ~2'···' ~n
in K[V] so that

(4.4.1)

(There are many ways of doing this.) Then we let X , X ,

1 2
... , X n be
indeterminates and construct the surjective homomorphism

(4.4. 2)

of K-algebras in which e(x.) = C Next we put

1 1

III = Ker e (4. 4. 3)

and select a finite set of polynomials, say

F . (X , X , ... , X ) where 1:s j :s m,

] 1 2 n

t See [(ll) Theorem 5, p. 18].

114
which generates 2I. Finally when x E V we write

x. = ~.(x) (4.4.4)
1 1

for 1:5 i :5 m. Naturally our main interest will be in results whose

statements do not involve ~ , ~ , •.• , ~ nor any of the entities defined
1 2 n
in terms of them.
Let x E V. This point gives rise to a homomorphism

w : K[V] .... K (4.4. 5)

x

in which w (f) = f(x). Put

x

(4.4.6)

The members of DerK(V, x) are called local derivations on V at x.

Thus if D E DerK(V, x), then D : K[V] .... K is K-linear and

D(fg) = (Df)g(x) + f(x)(Dg)

for all f, g in K[V].

Definition. Let x E V. Then the K- space DerK (V, x) is called

the 'tangent space' to V at x.

Of course from elementary geometry one has an intuitive idea of a

tangent space. It is therefore desirable to show how the above definition
is connected with our previously acquired concept. This, however, will be
postponed for the moment. Our immediate concern will be to examine the
effect of K-morphisms on tangent spaces.
To this end suppose that cfJ: V .... W is a K-morphism of affine sets
and let cfJ* : K[W] .... K[V] be the corresponding homomorphism of K-alge-
bras. If x V and w is defined as in (4.4.5), then, by (4. 1. 6), we
E
x
have a K-linear mapping

But w 0 ·cfJ* : K[W] .... K maps g into g(cfJ(x)). Thus we obtain

x

115
 Theorem 16. Let ¢: V .... W be a K-morphism of affine sets and
let x E V. Then there is a linear mapping

d(l/>, x) : DerK(V, x) .... DerK(W, 1/1 (x))

of K-spaces in which D E DerK(V, x) is mapped into D 0 1/1*.

Obviously if 1/1 is a K-isomorphism, then d(l/>, x) is an isomorphism

of K-spaces. Also if 1/1: W .... U is a further K-morphism, then

d(1/I, I/>(x)) 0 d(l/>, x) = d(1/I 0 1/1, x). (4.4. 7)

Let x E V and consider the equations

aF 2 aF 2 aF 2
"'. a 1 + -ax- a 2 +... + ax-n an
UA I
=0 (4.4. 8)
2

aF aF aF
----..!!!a
aX l 1
+ ~a2
ax
+ ... + ",:n an =
UAn
D.

We regard these as equations to be solved, in K, for aI' a 2 , ••• , an'

Theorems 5 and 6 applied to the present situation now give

Theorem 17. There is a natural isomorphism between the K-space

formed by the solutions of the equations (4.4.8) and the tangent space,
DerK(V, x), .Q!. V ~ x.

Corollary. Let n - s be the rank of the matrix "aF /3x i ". Then
the dimension of DerK(V, x), considered as a vector space over K, is s.

We shall now digress briefly in order to explain the name tangent

space that has been given to DerK(V, x).
Suppose, for the moment, that V is a closed subset of Kn. Each
indeterminate X. induces a coordinate function on V and if we call these
1
functions ~ , ~ , •.• , ~ , then (4.4. 1) holds. The ideal III of (4.4.3) now
1 2 n
consists of all polynomials F(X , X , ••. , X ) which vanish everywhere
1 2 n

116
on V. Moreover, if x E V, then x , x , •.. , x , as defined by (4. 4. 4),
1 2 n
are just the coordinates of x considered as a point of Kn.
Let x E V and suppose that a , a , •.• , a belong to K. Further
1 2 n
let Y be a new indeterminate. If F{X 1 , X2 , ••• , Xn) belongs to Ill,
then the constant term in

F{x + a Y, x + a Y, ••• , x + a Y)
1 1 2 2 n n

is zero and the coefficient of Y is

n of
L Ox a ..
i=l i 1

If this coefficient is zero for all F EIll, then (a , a , ••• , a ) is called

1 2 n
a tangent vector to V at x. The tangent vectors to V at x form a
vector space, TV
,x say, over K. Evidently (a 1 , a 2 , •.• , an) is in
TV if and only if the equations (4. 4. 8) are satisfied. We can therefore
,x
restate Theorem 17, for this special situation, as

Theorem 18. Let V be a closed subset of Kn and let x E V.

Then the space TV of tangent vectors and the abstract tangent space
,x
DerK(V, x) are naturally isomorphic K-spaces.

This ends the digression. From now on we assume, once again,

that V is an abstract affine set.
Let V be irreducible and suppose that x E V. If now D EDerK(V, x)
then, since

our discussion of derivations and fractions shows that there is a unique

D E DerK{QV, x' K) which extends D. (Note that, because there is only
one possible homomorphism CL
-VV,X - K of K-algebras, there is no need
to make the notation more explicit. )

Theorem 19. Let V be an irreducible affine set and let x E V.

Then each D E DerK(V, x) has a unique extension D in DerK{QV , x' K)
and the mapping

117
in which D r+ D is an isomorphism of K-spaces.

This follows from Lemma 6 Cor.

The next theorem provides another K-space that is isomorphic to
DerK(~,x' K). This result does not require V to be irreducible. We
recall that --Y,x
M.._ is being used to denote the maximal ideal of Q..-
~,x
•

Suppose that D E DerK(QV,x' K). By restriction D gives rise to

a K-linear mapping ---V,x M3_ • Thus D
M.._ -+ K and this vanishes on --'V,x
induces a K-linear mapping

A : M.._ !M!_ -+ K,
--'V, x' --'V, x

i. e. --'V, x'!M!_
A E HomK(M..- --'V, x ,K). Furthermore the mapping

DerK(Qv , x ,K) -+
--'V, x'!M3_
HomK(M..- --'V, x ,K) (4.4. 9)

which sends D into A is K-linear.

Theorem 20. Let x be a point of the affine set V. Then (4.4.9)

is an isomorphism of the K-space DerK(C:L ,K) on to the K-space
2 ~,x
Hom.._(M.._ IM:_ ,K).
K ---Y, x' --'V, x

Proof. Suppose that QV • Then ~ has a unique representa-

~ E
,x
tion in the form ~ = k + u, where k E K and u E --'V,x
M.._ • Also if
D E DerK(~, x' K), then D ~ = Du. Hence if D belongs to the kernel of
(4.4.9), then Du = 0 and therefore D~ = O. Thus our mapping is an
injection.
Next assume that

M.._ !M!_
A : ---V -+ K
, x' --'V, x

M.._ ... K be obtained by combining A with the

is K-linear and let .6.: --'V,x
natural mapping

Now define D: Qv -+ K as follows.

,x
If ~ E Q_-
~,x
and ~ =k + u

118
(where k E K and u ---Y, x ) put D~ = .:lu.
E M..._ An easy verification shows
that D E DerK(Qv, x' K) and it is clear that D 1'+ X under (4.4.9).
The next series of results is aimed at obtaining information about
the actual dimension of the tangent space Der K(V, x).

Theorem 21. Let V be an irreducible affine set defined over K.

Then K(V) is a separable extension of K.

This follows by combining Theorem 29 of Chapter 3 with Theorem

15 of this chapter.

Corollary. Let V be an irreducible affine set and put Dim V = r.

Then (with the notation explained at the beginning of this section) the
matrix "oF.jot"
J 1
has rank n - r.

Proof. We have K(V) = K(~l' ~ , ... , ~ ) and we have just shown

2 n
that this is a separable extension of K. The corollary therefore follows
from Theorem 11.
Still supposing that V is irreducible, let x E V. If w is the
x
homomorphism described in (4.4. 5), then

(OF~
w _J =_J
of.
x o~i Ox i

It therefore follows, from the last corollary, that the rank of II of J.lOx.1 "
is at most n - r. This observation, combined with Theorem 17 Cor. ,
yields

Theorem 22. Let x be a point of the irreducible variety V. Then

the dimension of the tangent space to V at x is at least Dim V.

We are now ready for the

Definition. A point x of the irreducible affine set V is called a

'simple point' if the dimension of the tangent space DerK(V, x), con-
sidered as a vector space over K, is equal to Dim V. Points of V which
are not simple are called 'multiple points'.

119
 Theorem 23. Let V be an irreducible affine set and suppose that
x E V. Then x is simple on V if and only if the dimension of the K-~ace
2
M._ !'M:_
---Y, x' "--Y, x
is equal to Dim V.

Proof. By Theorems 19 and 20, x is a simple point of V if and

only if the dimension of the K-space

2
HomK(M.- ---Y, x ,K)
--Y, x'!'M:_
2
M._ x'~­
equals Dim V. However this happens when and only when ---Y, ---Y, x
is a K-space whose dimension is Dim V.

Theorem 24. Let V be an irreducible affine set. Then the simple

points of V form a non-empty open subset.

Remark. Note that this result shows that an irreducible affine set
has at least one simple point.

Proof. Put r = Dim V and consider the matrix II 3F. /3X. ",
J 1
where the notation is that introduced at the beginning of the section. The
sUbdeterminants t of order n - r will be certain polynomials, say
G.(X , X , .•• , X ) where 1:s j :S s, and, by Theorem 21 Cor., we
J I 2 n
°
have G.(~l' ~ , ..• , ~ ) '" for at least one value of j. Without loss
J 2 n
of generality we may suppose that G.(~ , ~ , •.• , ~ ) is non-zero for
J 1 2 n
every j in the range 1:S j :S J.I., whereas it is zero whenever J.I.+ 1 :S j :S s.
Note that J.I. 2:: 1.
Let x EV and define xl' X2 ' ••• , xn as in (4.4.4). We obtain
3F./3x. by applying the homomorphism w , of (4.4.5), to 3F./3~ ..
J 1 . x J 1
Since "3F./3~." has rank n - r, the rank of "3F./ax." is at most
J 1 J 1
n - r; and x is a simple point precisely when the latter rank has this
value. Thus to sum up: x is simple on V if and only if there is a
(1 :S j :::: J.I.) such that G.(x , x , ..• , x ) '" 0.
J 1 2 n
Suppose that 1:s j :S J.I. and put f. = G.(~ , ~2' . . . , ~n)' Then
J J 1
fj E K[V], fj '" 0, and f/x) = Gj (Xl' x 2 ' ••• , xn). Accordingly

t Note that if r = n, then all the points of V are simple.

120
 N. = {x Ix EV and f. (x) "* 0 }
J J

is a non-empty open subset of V and each of its points is simple on V.

Finally NuN u... U N is the set of all simple points.
1 2 JJ.

Theorem 25. Let c/J: V .... W be a K-isomorphism of affine sets

and suppose that V, W are irreducible. If now x E V, then x is simple
on V if and only if c/J(x) is simple on W.

Proof. Since Dim V = Dim W, this follows from the remarks that
come immediately after Theorem 16.
We next make a brief investigation of the effect of enlarging the ground
field on tangent spaces and simple points. We recall that L denotes an
extension field of K.
Let f E K[V]. Then f has a natural prolongation, f say, to a
function on V(L). Since K[V] = K[~l' ~ 2, .•. , ~n ], we have
(L) A A A

L[V ] = L[~l' ~2' ... , ~n]' (4.4. 10)

Denote by

(4. 4. 11)

the surjective homomorphism of L-algebras in which e(xi ) = ~i' and put

'll = Ker e. (4. 4. 12)

Let G(X 1 , X 2 ' ••• , Xn) belong to L[X 1 , X 2 ' ••• , Xn] and write
G in the form

G=AG +AG + ... +AG,

1 1 2 2 q q

where A , A , .•• , A belong to L and are linearly independent over K

1 2 q
and G. = G.(X , X , •.• , X ) belongs to K[X , X , •.• , X]. Then
J J1 2 n 12 n
A q A
e(G) = ~ A.e(G.)
i=1 1 1
A

and now it follows that G E 'll if and only if each G. is in 'll (see (4.4.3)).
1
Consequently

121
and therefore the m polynomials F.(X , X , ... , X ) which generate
l 1 ,,2 n
III in K[X1 , X 2 , ••• , Xn] also generate ~r in L[X 1 , X 2 , ••• , Xn ].

Lemma 13. Let V be an affine set defined over K, let x E V,

and let L be an extension field of K. Then the dimension of the K-space

DerK(V, x) is equal to the dimension of the L-space DerL(V(L), x).--

Proof. Since t(x)

1
= ux)
1
= x.,1 Theorem 17 Cor. shows that each
of the spaces has dimension s, where n - s is the rank of the matrix
II of.lax.ll.
l O;ce again suppose that x E V ~ V(L). If

w : K[V] -+ K
x
is the homomorphism in which w (f) = f(x), then, since K[V]L = L[V(L)]
L x
and K = L,

is the homomorphism of L-algebras in which wL(u) = u(x). Let

x
D E DerK(K[V], K, wx ). Then, by Lemma 7,
L (L) L
D E DerL(L[V ], L, wx ).

Thus when D belongs to DerK(V, x), DL is in DerL (V(L), x).

Theorem 26. Let V be an affine set defined over K, let x E V,

and let L be an extension field of K. If now D , D , ... , D is a
--L-L 1 2 L q--
base for the K-space DerK(V, x), then Dl , D 2 , ••• , Dq is a base for
the L-space Der L (V(L), x).

Proof. By Lemma 13, it will suffice to show that the derivations

D~1 are linearly independent over L. Now the rows of the matrix

122
 Dl ~l Dl ~2 ... Dl ~n
D2~1 D2~2 D2~n

are linearly independent over K because otherwise D1, D2' .•. , D

would not be linearly independent over this field. Next t D~t = D.~:
L 1 J 1 J
(by the definition of D. ) and therefore the rows of the matrix
1

must be linearly independent over K and hence also over L. The theorem
follows.

Theorem 27. Let Y be an irreducible affine set defined over K,

let x E: Y, and let L be an extension field of K. Then x is simple on
;-if and only if it is simple on y(L).

Proof. By Lemma 13, the K-space DerK(V, x) and the L-space

Dett (V(L), x) have the same dimension. The theorem follows because
y( is irreducible and Dim Y = Dim y(L).

Theorem 28. Let Y be an irreducible affine set defined over K,

and let L be an exte~n field of K. Suppose that 7J E: y(L) and is a
generic point of y. Then 1/ is simple on y(L).

Proof. We have an isomorphism K[Y].:t. K[7J] of K-algebras in

which ~. is mapped into U1/)
1 1
= 1/.1 (say). (This means that if ~. is
1
the natural prolongation of ~. to y(L), then
1
t1 (1/) = 1/ 1.• ) Accordingly
we have an isomorphism

of K-algebras which matches ~. with 1/.. Hence, by Theorem 21 Cor. ,

the matrix "oF./o1/.11 has ra~ n - r, ~here r = Dim Y = Dim y(L).
J 1 (L)
Consequently 1/ is simple on Y .

t We use the same notation as in (4.4.10).

123
4.5 Tangent spaces and products

This topic is treated in a separate section because it does not require

the auxiliary notation introduced at the beginning of section (4.4).
Let V, W be affine sets (defined over K) and suppose that x E V,
yEW. Suppose that Dl E DerK(V, x) and D2 E DerK(W, y). If we make
the identification

K[V,x W] = K[V] ®K K[W), (4. 5. 1)

then there is a K-linear mapping

D: K[V x W) ~K (4. 5. 2)

in which

(4.5. 3)

A simple verification shows that D belongs to DerK(V x W, (x, y)).

Thus we have a mapping

DerK(V, x) E9 DerK(W, y) ~ DerK(V x W, (x, y)) (4.5.4)

Theorem 29. The mapping (4.5.4) is an isomorphism of the K-

space DerK(V, x) E9 DerK(W, y) on to the K-space DerK(V x W, (x, y)).

Proof. It is easily checked that (4.5.4) is K-linear and an injection.

Now suppose that D is in DerK(V x W, (x, y)). Define

D : K[V] ~ K and D : K[W) ~ K

1 2

by D f
1
= D(f ® 1) and D g
2
= D(l ® g). Then D E DerK(V, x),
1
D E DerK(W, y) and D is the image, under (4.5.4), of (D , D ).
2 1 2

Theorem 30. Let V and W be irreducible affine sets and suppose

that x E V and yEW. Then (x, y) is a simple point of V x W if and
only if x is a simple point of V and also y is a simple pOint of W.

124
 Proof. We know, from Chapter 3 Theorems 16 and 25, that
V x W is irreducible and that its dimension is Dim V + Dim W. Next,
by Theorem 29, the dimension of the K-space DerK(V x W, (x, y)) is the
sum of the dimensions of DerK(V, x) and DerK(W, y). Theorem 30
now follows from Theorem 22 and the definition of a simple point.

4.6 Differentials

In this section we shall give another application of the theory of

derivations. This will be useful when we come to study the Lie algebra of
an affine group.
Throughout section (4. 6) it will be assumed that K is infinite. In
what follows A denotes an n-dimensional (n ~ 1) vector space over K
and we put A * = Ho~(A, K) so that A * consists of all the linear forms
on A.

Lemma 14. Let cp: A * .... K[A] be a K-linear mapping. Then there
exists a unique D € DerK(K[A]) such that DF = cp(F) for all F € A *.

Proof. Let F l' F 2' ••• , F n be a base for A * over K. Then

K[A] = K[F l' F 2' ••• , F n] and the F i are algebraically independent over
K. Now
n 0
D = L I/J(F.) of
i=l 1 i
belongs to DerK(K[A]) and has the required property. Uniqueness is
obvious.
Let a € A and take for cp: A * .... K[A] the K-linear mapping in
which I/J(F) = F(a). By Lemma 14, this mapping gives rise to a derivation,
.6a say, of K[A] over K. Thus .6a € DerK(K[A]) and

.6 F
a
= F(a) (4. 6. 1)

for all F A *. Naturally.6 extends to a derivation of K(A) over K.

€
a
This extension will be denoted by the same symbol.
Suppose that H € K(A) so that H is a rational function on A. If
b € A and H is defined at b, then .6a H is also defined at b. Put

125
 Def(H) = {b /b E A and H is defined at b} (4. 6. 2)

and define

dH : Def(H) x A -K (4.6.3)

by

(4.6.4)

The mapping dH is called the differential of H.

We record some properties of the differential. In the following
formulae H, H ,H are rational functions on A, all defined at b, and
1 2
k E K. The formulae are:

(dH)(b, a 1 + a 2) = (dH)(b, a1) + (dH)(b, a 2), (4. 6. 5)

(dH)(b, ka) = k«dH)(b, a)), (4.6.6)

(dkH)(b, a) = k«dH)(b, a)), (4.6. 7)

(d(H 1 + H 2))(b, a) = (dH I )(b, a) + (dH 2)(b, a), (4. 6. 8)

(d(H I H))(b, a) = HI (b)(dH 2)(b, a)) + H2 (b)(dH I )(b, a)), (4.6.9)

(dF)(b, a) = F(a) for all F E A *. (4.6.10)

Finally if H = p;Q, where P, Q E K[A] and Q(b) '* 0, then

(dH)(b, a) = Q(b)(dP)(b, a)) - P(b)(dQ)(b, a)) . (4.6. 11)

(Q(b))2
We continue with the same assumptions but now we suppose that L
is an extension field of K. Then, with the same notation as in section
(1.6), A L is an n-dimensional vector space over L. By Chapter 2
Theorem 32, A L = A (L) because K is infinite.
Let H E K(A). Then H is a rational function on A and therefore
" say, on A L. Note
it has a natural prolongation to a rational function, H
that if F E A *, so that F is a linear form on A, then its prolongation F
is a linear form on A L.
Suppose now that a EA ~ A L . Then we can form A F" and A F.
a a
It is clear that A.
a
F ·is the prolongation of A. F. We conclude at once
a

126
from this that A}I is the prolongation of AaH and therefore

(ctH)(b, a) = (dH)(b, a) (4.6. 12)

provided that b € A and H is defined at b.

Now assume that D € DerK(L). Choose a base a , a , .•• , a
1 2 n
for A over K. Then these same elements will constitute a base for AL
over L. Hence if x € A L we can write x in the form
x = A a + A a +... + A.a with A. € L. Put
11 22 nn 1

Dx = (DA)a + (DA)a +... + (DA )a • (4.6.13)

1 1 2 2 n n
L
Then when x, x ,x belong to A and A € L, we have
1 2

D(x + x ) = DX 1 + Dx , (4.6. 14)

1 2 2

D(Ax) = (DA)X + A(Dx). (4.6. 15)

Also

Db = 0 for all b € A. (4.6.16)

In particular it follows that, when i\, A2 , ••• , A

m
be long to Land
b l' b 2' ••• , b m to A,
m
D(\ b 1 + \b 2 + ... + A b ) = L (DA.)b .• (4.6. 17)
m m i=1 1 1

This shows that the definition of Dx as provided by (4.6.13) is independent

base
of the choice of the K- - - 1
a , a 2, ••• , a nof
-
A. Note too that

F(Dx) = D(F(x», (4. 6. 18)

for all F € A*.

A

Lemma 15. Let H € K(A) and let H be its prolongation to a ration-

al function on AL. Let D € DerK(L). Suppose that x € A L and that H
is defined at x. Then

D(H(x» = (ctH)(x, Dx).

127
 Proof. Let R consist of all the rational functions on A whose
x
prolongations are defined at x. Evidently R x is a subalgebra of the
K-algebra K(A). Let the mapping cp : R ... L be given by cp(S) = §(x).
x
This is a homomorphism of K-algebras.
Define ~: R ... Land
x
n : R x ... L by

n(s) = (dS)(x, Dx)

and

n(s) = D(S(x».

We wish to show that ~ and "IT coincide. Clearly they are additive and
they both vanish on K. Also if Sl' S E R , then
2 x

~(S S ) = ¢(S )~(S ) + cp(S 2)~(S 1 )

12 1 2

and

Again if F E A*, then FER and, by (4.6.10),

x

~(F) = F(Dx)

which is equal to D(F(x» = "IT(F) by (4. 6. 18). It follows that ~ and n

agree on K[A].
Suppose that S E R. We can choose t P, Q E K[A] so that S = P /Q
... x
and Q(x) "* O. By (4.6. 11), ~(S) is equal to

Q(x)(dP)(x, Dx» - P(x)(dQ)(x, Dx»

(Q(x»2
_ Q(x)(DP(x» - P(x)(nQ(x»
(Q(x» 2
= D(P(x);Q(x»
= D(S(x»
= "IT(S).

t C. f. the proof of Theorem 31 of Chapter 3.

128
The lemma follows.
We continue to assume that K is an infinite field but now suppose
that A is a non-trivial, unitary, associative K-algebra whose dimension
as a K-space is finite. Let a A and define A : A'" A by A (b) = abo
€
a a
This mapping induces a linear mapping A* : A*'" A* of K-spaces and now
a
we can define a K-linear mapping cp: A * ... K[A] by cp(F} = -A*(F}.
a
Lemma 14 shows that cp gives rise to a derivation, D say, of K[A]
a
over K. Thus Da € Der K(K[A]) and

D F = -(F 0 A ) (4.6.19)
a a

for all F € A *. Accordingly

(D F)(b) = -F(ab} (4. 6. 20)

a
provided that F € A * and a, b € A.
Next we establish a connection between the derivation Da and the
notion of a differential. Observe first that Da extends to a derivation
of K(A} over K (we use D also to denote the extension) and if
a
H € K(A) is defined at b, where b € A, then D H is defined at b as
a
well.

Theorem 31. Let H be a rational function on A which is defined

at b. Then

(D H)(b) = -(dH)(b, ab)

a

for all a € A.

Proof. Denote by III the set of all P, in K[A], for which

(DaP}(b) = -(dP)(b, ab).

(Here a and b are kept fixed.) It is clear that K ~ III and, by (4.6. 20)
and (4.6. 10), we see that A* ~ III as well.
Next assume that P ,P € Ill. An easy verification shows that
1 2
PI + P 2 and PIP 2 also belong to Ill. This shows that III = K[A].

129
 Finally we can write H in the form P /Q, where P, Q E K[A] and
Q(b) "* O. If we now use (4.6. ll) and the fact that Da is a derivation we
obtain the desired result.
We add a few remarks about prolongations. Let L be an extension
field of K, then AL is a unitary, associative, L-algebra. t As before
if H E K(A), then we use H to denote its prolongation to a rational
function on AL. If now F EA * and a, b E A, then

Thus D F is the prolongation of D F. It follows immediately that for

a a -
any H E K(A), D H is the prolongation of D H.
- a a
The results of section (4.6) will find applications in Chapter 6.

t See section (1. 6).

130
Part II. Affine Groups
5 . Affine groups

General remarks

Throughout Chapter 5, K will denote an arbitrary field and if X

is a subset of a topological space, then X will be used to denote its
closure.
In the case of a group G, the letter e is used to denote its identity
element. Should it be necessary to be more explicit the symbol eG will
be employed. In the case of an affine group {see section (5. 1)) we shall
always use G to denote the connected component containing the identity
o
element.

5. 1 Affine groups

Let G be an affine set and suppose that G also has the structure of
an abstract (multiplicative) group. Consider the mappings

G x G - G and G - G

in which (x, y) 1-+ xy and x 1-+ x -1 respectively. If both of these are

K-morphisms, then it would be perfectly reasonable to describe the whole
situation by saying that G is an affine group. However it will be con-
venient to give an equivalent definition in rather different terms. We
therefore make a fresh start.
Let G be an affine set (defined over K) and suppose that we are
given K-morphisms

/1.: G x G-G, (5. 1. 1)

j : G - G, (5. 1. 2)

and a point e of G.

133
 Definition. The quadruplet [G, p, j, e] is called an 'affine group'
provided that
(i) p. (p (x, y), z) = p(x, p(y, z)),
(ii) p(x, e) = x = p(e, x),
(iii) p(x, j(x)) = e = p(j(x), x),
for all x, y, z in G.

Suppose that we have this situation. Then G is a group with p as

its law of ,composition, j the operation of taking inverses, and e the
identity element. We normally, and without special comment, use the
multiplicative notation, that is we write p(x, y) = xy and j{x) = x-I.
However we shall translate certain results so as to make addition the law
of composition on the few occasions when this is more convenient.
Let [G, p, j, e] be an affine group. We can construct two K-
morphisms

GxGxG-+G

as follows: in the first (x, y, z) r+ p(p(x, y), z) and in the second

(x, y, z) ~ p(x, p(Y, z)). Condition (i) says that these coincide. Similarly
(ii) says that the K-morphisms x r+ p(x, e) and x 1-+ p(e, x) are the same
as the identity morphism G -+ G. Finally (iii) asserts that the K-morphism
x -+ p(x, j(x)) and x -+ 1L(j(x), x) both agree with the constant K-morphism
G -+ G in which every point is mapped into e.
One advantage of presenting the definition in this way is that it
enables us to show that the property of being of an affine group is pre-
served under an extension of the ground field
To see how this arises let us assume that [G, IL, j, e] is an affine
group and that L is an extension field of K. By (2. 10. 9),
(G x G)(L) = G(L) x G(L) and therefore p(L) is an L-morphism

Theorem 1. Let [G, .P, j, e] be an affine group defined over K,

and let L be an extension field of K. Then [G(L), p(L), /L), e] is an
affine group defined over L.

134
 Proof. Define L-morphisms

and
q : G(L) x G(L) x G(L) -+ G(L)

by p(~, 'f/, ~) = J.L(L)(f.L(L)(~, 71), ~) and q(~, 'f/, ~) = J.L(L)(~, J.L(L)('f/, ~)).
Since G(L) x G(L) x G(L) = (G x G x G)(L) and since p and q extend
the same K-morphism

G x G x G -+G,

it follows, by Chapter 2 Lemma 15 Cor., that p = q. This verifies the

first of the three conditions in the definition of an affine group and the
other two may be checked similarly.

Corollary 1. Let G be an affine group defined over K, and let

L be an extension field of K. Then there is just one way to endow the
affine set G(L) with the structure of an affine group so that G becomes
a subgroup.

Proof. We use the same notation. Let [G(L), ~, j, e] be an affine

group having G as a subgroup. It will suffice to show that [i = J.L(L).
Now the L-morphism

extends the K-morphism f.L: G x G -+ G by hypothesis and therefore

[i = /L) as required

Corollary 2. If G is a commutative affine group. then so is G(L).

Proof. Define an L-morphism I/J: G(L) x G(L) -+ G(L) by

CP(~, 'f/) = f.L(L)('f/, ~). This extends the K-morphism J.L : G x G - G so
cP = f.L(L). The corollary follows.
The notation [G, f.L, j, e] is more or less indispensable for the
proper statement of Theorem 1, but it is too cumbersome for general use.

135
Frequently we shall suppress any direct reference to fl, j, and e and
say simply that G is an affine group. Also for the rest of section (5. 1)
we shall use the multiplicative notation and employ xy as an alternative
to fl(x, y).
Let G be an affine group (defined over K) and let x E G. The
mapping

A : G"'G (5.1. 3)
x

in which A (y) = xy is a K-morphism called left translation by means of

x
x. In fact A is a K-isomorphism (of the affine set G on to itself)
x
whose inverse is A • The right translation by means of x is the K-
-1
X
morphism

(5.1.4)

given by P (y)
x
= yx. This too is a K-isomorphism of the affine set G on
to itself.
The next theorem asserts that an affine group, when regarded as an
affine set, is homogeneous.

Theorem 2. Let G be an affine group (defined over K) and let

x, y belong to G. Then there exists a K-isomorphism I/J, of the affine
set G on to itself, such that cp(x) = y.
-1
In fact we may take ¢ to be the left translation by yx .
We next turn our attention to the closed subgroups of an affine group.
In preparation ·for this we prove a general lemma.

Lemma 1. Let V, W be affine sets defined over K and let

I/J : V"'W and 1J; : V"'W be K-morphisms. Then

{xix E V and I/J(x) = 1J;(x)}

is a closed subset of V.

Proof. Define a K-morphism p: V ... W x W by p(x) = (I/J(x), 1J;(x)).

Then

136
 {xix f.V and ¢(x) = tJ;(x)} = p-l(A),
where A denotes the diagonal of W x W. The lemma follows because,
by section (2.9), A is closed in W x W.
Now let [G, p" j, e] be an affine group and let H be a closed sub-
group of G, that is H is a closed subset of the affine set G and a sub-
group of the abstract group G. By restriction fJ and j induce K-
morphisms

fJ':HXH-+H

and

j': H-+ H

and it is clear that [H, fJ', j', e] is an affine group. We embody these
observations in

Theorem 3. Let G be an affine group defined over_ K and let H

be a closed subgroup of G. Then H has an induced structure as an affine
group defined over the same field

From here on, every closed subgroup of an affine group will auto-
matically be regarded as an affine group. Suppose that H is a closed
subgroup of the affine group G and let L be an extension field of K.
Then on the one hand H(L) is an affine group with p,,(L) as its law of
composition (we are retaining our previous notation), and on the other it
is a closed subset of G(L). Using the inclusion mapping H(L) -+ G(L)
we can construct, in an obvious manner, the L-morphisms

H (L) x H(L) fJ,(L) (L) (L)

---_~ H --i>G
and

Since these extend the same K-morphism H x H -+ G, it follows, again by

Chapter 2 Lemma 15 Cor., that they coincide. Accordingly the inclusion
mapping H(L) - G(L) is a homomorphism of groups.

137
 Corollary. Let G be an affine group defined over K and let H be
a closed, normal subgroup of G. If now L is an extension field of K,
then H(L) is a closed normal subgroup of G(L).

Proof. Let h E H. Then

is the inverse image of H(L) with respect to the L-morphism ~ 1-+ ~h(l.
The inverse image is therefore closed in G(L) and so, since it contains
G, it is G(L) itself. Thus ~(l E H(L) for ~ E G(L) and h E H.
Now let ~ E G(L). This time

is closed in G(L) and it contains H. Consequently, by Chapter 2

Theorem 30, it must contain H(L). Accordingly ~1J (1 E H(L) for all
~ E G(L) and TJ E H(L), so the corollary is proved.
Let us investigate further the question of closed subgroups. Suppose
that G is an affine group and assume that (J E G. Put

C (J = {x, X E G and (]X = X(J J. (5. 1. 5)

Lemma 2. C (J is a closed subgroup of G.

Proof. Let A(J respectively P(J denote the left respectively right
translation by means of (J. Then

Consequently, by Lemma 1, C(J is a closed subset of G. It is obvious

that it is a subgroup of G.

Theorem 4. Let G be an affine group. Then its centre is a closed

subgroup of G.

Proof. Define C (J as in (5. 1. 5). The intersection of all the C (J's

is the centre of G. The theorem therefore follows from Lemma 2.

138
 Theorem 5. Let G be an affine group and H a subgroup of G.
If H is the closure of H. in G, then H is also a subgroup of G.

Proof.
The mapping x J-+ x-I is a homeomorphism of the affine
set G on to itself. Consequently (Hf1 is the closure of H- 1 = Hand
therefore (Hf1 = H. Thus H is closed under the taking of inverses.
Let h € H and consider left translation by means of h. Since this
also provides a homeomorphism, hH is the closure of hH = H, that is
hH=H.
Finally assume that x € H. Then Hx is the closure of Hx and
Hx !: HH = H. Accordingly Hx!: H which shows that H is also closed
under multiplication. The theorem follows.
If we are given a subset of an affine group G, then there will be a
smallest closed subgroup of' G which contains the given subset. What
follows now is preparation for an investigation of the connection between
the subset and the closed subgroup which it generates.
Let x € G. We have already remarked that the right translation
p : G - G is a K-isomorphism of affine sets. It therefore induces an
x
isomorphism

p*x : K[G] - K[G] (5. 1. 6)

of K-algebras. Note that p: is the identity mapping of K[G] and that

p* 0 p* = p* (5. 1. 7)
x y xy

for all x, y € G. Also if f € K[G], then

(p*(f))(y)
x
= f(yx). (5. 1. 8)

Theorem 6. Let G be an affine group defined over K, and let S

be a non-empty multiplicatively closed subset of G. Further let f € K[G]
and denote by Uf the K-subspace of K[G] generated by the family
{p*(f) } S. Then the dimension of Uf is finite and p*(U f ) = Uf for all
z Z€ Z ---
Z € S.

139
 Remark. There is, of course, a similar result based on left trans-
lations, but we shall not state it separately.

Proof. Define w: G x G - K by

w(y, x) = f(yx) = (P~(f))(y).

Then w E K[G x G] and therefore there exist f l , f 2 , ••• , fm and

gl' g , .•. , g in K[G] such that
2 m
m
w(y, x) = L f. (y)g. (x)
i=l 1 1

for all x, y in G. Thus

m
p*(f) = L g. (z)f.
z i=l 1 1

which proves the first assertion. Next when z, XES we have

p*(p*(f)) = p* (f) E U
z X zx f

which shows that P~(Uf) k: Ur Consequently, when z E S, P~ induces an

endomorphism of the K-space U r
Since the dimension of Uf is finite and
this particular endomorphism is an injection, we see that p;(Uf ) = Uf as
required.

Theorem 7. Let G be an affine group (defined over K) and let U

be a subspace of K[G] when K[G] is considered as a vector space over K.
Suppose that x E G is such that p*(U)
x
~ U. Then p*(U) =
-- x
u.
Remark. Naturally there is an analogous result for left translations.

Proof. Let S consist of e and the elements x n, where n 2: l.

Then S is a multiplicatively closed subset of G and p~(U) k: U for all
z E S. Let fEU and define Uf as in Theorem 6. Then

The theorem follows.

140
 The next result is very striking. t

Theorem 8. Let G be an affine group and S a non-empty multi-

plicatively closed subset of G. Then the closure S, of S in G, is the
smallest closed subgroup of G containing S.

Proof. It is enough to show that S is a subgroup of G. Put

'U = IG(S) = IG{S) and let f E'U.. The mapping w: G x G -+ K given by

w(y, x) = f(yx) = (p*(f))(y)

x

belongs to K[G x G] and it vanishes on S x S because S is multiplica-

tively closed. It must therefore vanish on the closure of S x S in G x G.
However, by Chapter 2 Theorem 28 Cor., this closure is S x S. Hence
if 7, a E S, then

(p*(f))(7)
a
= W(7, a) =0
whence p*(f) E 'U. This shows that p* ('U) k 'U. It follows, from Theorem
a a
a
=
7, that p*('U) 'U. In particular f = p*(fl) for some f' E 'U. Accor-
a
dingly f(e) = f'(a) = O. Thus to sum up: p;('U) = 'U for all a E S and
furthermore e E S.
Next suppose that x E G and p*('U)
x = 'U. Then 'U = p* 1 ('U) and
x-
so for f E 'U we can choose f1 E 'U so that f = p* -1 (f 1). It follows that
x
f(x) = f1 (e) = 0 and therefore

Combining our results so far we find that

S = {xix E G and p*('U)

x
= 'U 1
and now it is a trivial matter to check that S is a subgroup of G.

t See C. Chevalley [(2) Proposition 2, p. 82].

141
 Corollary. Let G and S be as in the theorem and put 'U=IG(S).
Then

{xix E G and p*('U) = 'U }

-- x

is the smallest closed subgroup of G containing S.

The above proof of Theorem 8 is somewhat indirect but it has the

advantage of introducing ideas that will be useful in other contexts. A
short self-contained proof of the theorem will now be given. t
Assume that the hypotheses of Theorem 8 are satisfied. By
adapting the proof of Theorem 5 we can readily check that S is multi-
plicatively closed. We may therefore suppose that S itself is closed and
seek to prove that S is a group.
Let XES. The collection of subsets of S that are (i) topologically
and multiplicatively closed, and (ii) contain a positive power of x is
non-empty. In fact it has a minimal member M (say) because G is a
Noetherian topological space. By construction xn EM for some n 2:: 1.
Consider x~. This is topologically closed, because left trans-
lations are homeomorphisms, and it is clearly multiplicatively closed
as well. Also xnM k: M, because M is multiplicatively closed, and
xnM contains in. Consequently, by the choice of M, x~ = M and
. 2n.... n 2n
therefore M = x M. ThIS shows that x = x y for some y E M k: S
and now we have xny == e. It follows that x -1 E S and with this the
theorem is proved,
We return to the general discussion in order to record some simple
observations on direct products of affine groups. To this end let Gl' G2
be affine groups defined over K. Then G1 x G2 is an affine set and it
has a group structure. Indeed it is readily verified that G1 x G2 is an
affine group. More generally, if G1 , G2 , ••• , Gm are affine groups,
then the direct product

GXGX ••. XG
12m

is also an affine group. Next let L be an extension field of K. Then

t The argument which follows was communicated to me by P. Vamos.

142
Theorem 1 Cor. 1 shows that we can identify

(G x G X ••• x G )(L)
12m

and

as affine groups in the obvious way.

5. 2 Components of an affine group

Throughout section (5.2), G will denote an affine group defined

over K, and we shall use the multiplicative notation. Since G is a space
by virtue of its affine topology, it is the disjoint union of its connected
components. We shall normally use Go to denote the particular con-
nected component that contains the identity element e ~ G. It is cus-
tomary to refer to Go as the connected component of the identity though
this is an abuse of language.

Theorem 9. The connected component Go that contains the identity

element of G is a closed normal subgroup of G.
-1 -1
Proof. The homeomorphism x t-+ x shows that G is a con-
-1 0
nected component of G containing e. Hence G = G. Let Z EGO.
-1 0 0
Then Z EGO and now the left translation of G by means of Z shows
that zG is a connected component of G containing zz -1 = e. Con-
o
sequently zG o = Go and we have proved that Go is a subgroup of G. It
must be a closed subgroup because the connected components of a topo-
logical space are always closed.
Let x E G and define p : G -+ G by p(y) = xyx- 1 • Then p is
compounded from two translations and is therefore a homeomorphism.
This makes it clear that p(G ) = G , that is to say G is a normal sub-
0 0 0
group of G.
We shall next examine the connection between Go and the irreducible
components, VI' V 2' ••• , Vr say, of G. We arrange the numbering so
{ -1 -1 -1 }
that e EV1. Note that VI' V 2 ' ••• , Vr is a permutation of

143
{VI' V 2 , ... , Vr } and so too are

for every Z E G.
Since G is the irredundant union of V , V , .•• , V , we can
1 2 r
choose x E V so that x,. V , ••. , x ,. V. This secures that, for each
1 2 r
y E G, xy belongs to just one of V , V , .•• , V .
1 2 r
Suppose that i
r. Then x-1V i = Vk for some k. Let us
1:5 :5

determine k. Since V y = V. if and only if xy EV., that is to say if

1 1 1
and only if y E Vk' it follows that V1 Vk ~ Vi" However e E V1 and
therefore we may conclude that k = i. Accordingly V V. = V. for
1 1 1
i = 1, 2, •.. , r. In this relation replace Vi by V~ 1 • This shows that
-1 -1 -1
V1 VI = VIand hence that VI V1 = VI. It has now been proved that
V 1 is a subgroup of G and, because it is connected, we have V1 ~ Go.
We recall that each of V , V , •.. , V is closed.
1 2 r
Assume that 2:5 i :5 r and choose y. E V. so that it does not
1 1
belong to any other irreducible component. Then V y.
1 1
= V.1 and thus we
see that the irreducible components of G are none other than the right
cosets of V. In particular, they are disjoint. However G is connected
1 0
and it contains e. Accordingly G does not meet V u V u... u V
o 2:3 r
and therefore we must have G
o = V.
1
This proves

Theorem 10. Let G be an affine group and Go the connected

component of the identity element. Then the irreducible components of
G are just the cosets of G in G. Thus the irreducible components of
0-
G are disjoint, and the index of G in G is finite.
o
Corollary. Any two irreducible components of the affine group G
are K-isomorphic and so have the same dimension. This is equal to the
dimension of G. In particular Dim Go = Dim G.
Theorem 11. Let H be a closed subgroup (of the affine group G)
whose index in G is finite. Then G0 ~ H.

Proof. The (left) cosets of H in G are closed. Consequently H

and G \H are disjoint closed subsets of G. Since e E Hand Go is

144
connected, we must have G 0 ~ H.
An important fact which is an immediate consequence of Theorem 10
is stated separately as

Theorem 12. Let G be an affine group. Then the following two

statements are equivalent:
(a) as an affine set G is connected;
(b) as an affine set G is irreducible.
Thus for affine groups connected and irreducible amount to the same
thing. In what follows we shall find it more convenient to speak of con-
nected rather than irreducible groups.

Theorem 13. Let G be a connected affine group. Then every point

of G is a simple point.

Remark. This result is sometimes described by saying that con-

nected affine groups are smooth.

Proof. Since G is irreducible, Theorem 24 of Chapter 4 shows

that it has at least one simple point. The fact that all of its points are
Simple follows by combining Theorem 2 with Theorem 25 of Chapter 4.

Theorem 14. Let G be a connected affine group and let U, V ~

non- empty open subsets of G. Then G = UV.

Proof. Let x E G. Then U and xV- i are non-empty open sub-

sets of G and, furthermore, G is irreducible. Hence, by Chapter 3
Theorem 3, there exists u E U which also belongs to XV-i. Accordingly
x = uv for some v E V.

Corollary. .!!.. U is a non-empty open subset of a connected affine

group G, then G = UU.

The next result has important applications in the theory of solvable

affine groups.

Theorem 15. Let G be an affine group and let the ground field K
be algebraically closed. Suppose that S , S , ..• , Sr are subsets of G
1 2

145
having the following properties:
(i) e E Si for i = 1, 2, ... , r;
(ii) for each i the closure S.,
1 -
of S.l -
in G, is irreducible;
(iii) for each i, S. contains a non-empty open subset of S ..
1 1
If now S is the smallest multiplicatively closed subset of G containing
Sl' S2 , ... , Sr , -
then
-
S is a closed connected subgroup of G.

Remark. Condition (iii) means that there is a non-empty subset

of S. which is open in the topology of S. that is contained in S..
1 1 1

Proof. Put w = S x S x ... x S. Then W is an irreducible

1 2 r
affine set and it contains

w = s x s x ... xS.
o 1 2 r
For each positive integer n put

n
W =w x W x ... x W,
n
W=wxwx ... xW,
o 0 0 0

where in each case the product contains n factors. Now although the
affine topology on a product is different from the ordinary product topology:
none the less w~ contains a non-empty open subset of Wn. But Wn is
a product

Sl X ••• x Sr x Sl x . .. x Sr x . .. X Sl X ••• x Sr

so each element of Wn is a sequence of elements of G. Let Wn - G

be the K-morphism obtained by multiplying together the terms of the
sequence without disturbing their order. Then if W[n] denotes the image
of Wn and W[n]0--
that of w n0 , we arrive at an almost surJ'ective K-
m[r~hism Wn - W[ n] of affine sets. Note that Wn (and hence also
W n) is irreducible and therefore~Chapter 3 Theorem 33, w~n] con-
tains a non-empty open subset of WlnJ .
Since e E S. for i = 1, 2, ... , r, we have w[n] k w[n+l]. Thus
1

146
and each term in the sequence is irreducible. It follows that there exists
a pos itive integer t such that

for all n ~ t. Now S is contained in the union of the W[ n) and therefore

S ~ w[tJ. Thus S ~ S ~ w[t). Also W[t) contains a non~empty open
--[- 0
subset of W t) and therefore this is also true of S. Accordingly, by
Chapter 3 Theorem 2, S = w[t]. This shows that S is connected and, by
Theorem 8, it is a closed subgroup o~ However we know that S
contains a non-empty open subset of W[t] = S. It therefore follOWS, from
Theorem 14 Cor., that S = SS = S. The proof is now complete.

Corollary 1. Let the ground field K be algebraically closed, let

W be an irreducible affine set, and let q, : W -. G be a K- morphism of
affine sets, where G is an affine group. If now e E q,(W), then the
smallest multiplicatively closed subset of G containing q,(W) is a Closed,
connected subgroup of G.

Proof. Put Sl = q,(W). Then conditions (i) and (ii) of the theorem
hold trivially, whereas condition (iii) holds by virtue of Chapter 3 Theorem
33. The corollary follows.
We recall that if H is any group, then the commutator subgroup
(H, H) of H is the subgroup that one obtains by taking all finite products
-1 -1
of elements of the form aTa T ,where a and T belong to H. This
is a normal subgroup of H. In fact it is the smallest normal subgroup of
G with an abelian factor group.

Corollary 2. Let the affine group G be connected and let K be

algebraically closed. Then the commutator subgroup (G, G) is a closed
connected subgroup of G.

Proof. The product G x G is irreducible and the mapping

q, : G x G -. G in which q,(a, T) = aTa- 1T- 1 is a K-morphism. Also
e E q,(G x G) and the smallest multiplicatively closed subset of G con-
taining q,(G x G) is (G, G). The desired result therefore follows from
Corollary 1.

147
5. 3 Examples of affine groups

All the affine groups considered in section (5. 3) will be defined

over K. If G is a group, then I-' : G x G - G will denote the composition
mapping (multiplication or addition as the case may be) and j : G - G
the mapping which takes each element into its inverse.

Example 1. Let G be a finite group. Then G is an affine set

and Theorem 18 of Chapter 2 shows that [G, 1-', j, e] is an affine group.
G is connected if and only if e is the only element of G.

Example 2. Let V be an n-dimensional vector space over K,

put t E = En~(V), and for each gEE denote by D(g) the determinant
of g. If now

GL(V) = {g /gEE and D(g) '" 0 },

then GL(V) is a group and, as was shown in section (2.5), it has a

natural structure as an affine set.
Suppose first that K is a finite field. Then GL{V) is a finite group
and therefore it is an affine group by Example l.
Now assume that K is infinite. The remarks made at the end of
section (2. 5) show that

I-' : GL(V) x GL(V) - GL{V)

and

j : GL(V) -+ GL(V)

are K-morphisms. Hence GL(V) is again an affine group.

It is usual to call GL(V) the general linear group of V. This has
a particularly important role in the theory of affine groups. t
We add a little extra information in

t See section (2. 5) Example 5.

t See Theorem 27.

148
 Theorem 16. Let V be an n-dimensional vector space over an
infinite field K. Then GL(V) is a connected affine group whose dimension
• 2
IS n

That GL(V) is connected and has dimension n 2 are consequences

of (2.5. 7). In fact the formula quoted shows that K(GL(V)) is a pure
transcendental extension of K.

Example 3. Let n;::: 1 be an integer and let GL (K) be the multi-

n
plicative group formed by all non-singular n x n matrices with entries
in K. Indeed GL (K) is, to all intents and purposes, just GL(V) when
n
V = Kn. In particular it is an affine group. Its highly explicit form has
some advantages. For instance the mapping

T : GL (K) - GL (K)
n n

which replaces each matrix by its transpose, is a K-isomorphism of

affine sets and an anti-isomorphism of groups.
If K is infinite then, of course, GL (K) is connected and of dimen-
n
sion n2 •

Example 4.
The affine group GL I (K), obtained by taking n = 1
in the last example, consists of the non-zero elements of K the law of
composition being multiplication. It is therefore known as the multiplica-
tive group of K. We shall sometimes denote it by G .
m
Suppose that K is infinite. Then G is connected, its dimension
m
is one, and (2. 5. 7) shows that

K[G ] = K[X, X-I], (5. 3. 1)

m

where X denotes an indeterminate.

Consider the direct product

G XG X .•. XG (5. 3. 2)
m m m

where there are n factors. Such a group is called a torus.

Assume once more that K is infinite. Then the torus (5. 3. 2) is
connected and its dimension is n. Indeed, with a self- explanatory notation,

149
its coordinate ring may be taken to be

where X , X , ... , X are distinct indeterminates.

1 2 n

Example 5. Let us regard K as an affine set and define

/J : K x K - K and j : K - K by /J(x, y) = x + y, j(x) = -x. Then
[K, /J, j, 0] is a commutative affine group called the additive group of K.
If K is a finite field, then its additive group is a finite group. However
if K is infinite the additive group is connected, of dimension one, and its
coordinate ring may be taken to be K[X], where X is an indeterminate.

Example 6. Let V be an n-dimensional (n 2: 1) vector space over

K. Put E = En%(V) and

SL(V) = {gig € E and D(g) = I}.

Then SL(V) is an affine group. It is known as the special linear group

of V.
Assume that K is infinite and let X.. be indeterminates. Note
IJ
that SL(V) may be regarded either as a closed subset of E or as a
closed subgroup of the general linear group GL(V); however in both cases
we obtain the same coordinate ring namely

where I is the ideal formed by the members of

which vanish at all points of SL(V).

Next the determinant Det /IX .. /I is an irreducible polynomial and
IJ
therefore Det /IX .. " - 1 is also irreducible. Let A
IJ
= A(X 22 , ... , X )
nn
be the cofactor of X in the matrix "X .. " and suppose that
11 IJ
F(X11 , ••• , Xnn) belongs to I. Then, by long division, we obtain a
relation

150
 AmF = (Detl/x1J.. 11 - 1) Q + B,

where Q, B E K[X , •.• , X ] and XII does not occur in B.

11 nn
Let a ,a belong to K and be such that
, ••. , a
12 13 nn
A(a
22
, ••. , a
nn
) *- O. We can choose a E K so that Det II 0' ..
11 1J
11 = 1.
Since F E I, the value of F(a ,a ) is zero and therefore
, ••. , a
11 nn 12
B( a 12 , ... , a nn ) = O. Thus for all choices of 0'12' 0'13' ••• , a
nn the
polynomial

vanishes and therefore, because K is an infinite field, it is the null

polynomial. It follows that B(X ,X , ••• , X ) is null. Accordingly
12 13 nn
A mF belongs to the ideal generated by Det Ilx .. 11 - 1. But, because
1J
DetIIX .. 11 - 1 is irreducible, this ideal is prime and now we see that it
1J
contains F. It follows that

I = (Det Ilx1J.. 11 - 1),

and thus we obtain

Theorem 17. Let K be an infinite field and V an n-dimensional

(n ~ 1) vector space over K. Then the special linear group SL(V) is
connected, its dimension is n 2 - 1, and its coordinate ring may be taken
to be

in the manner explained above.

We mention, in passing, that when working with GL (K), rather

n
than with GL(V), the notation for the special linear group, that is the
closed subgroup formed by all n x n matrices with determinant 1, is
SL (K).
n

Example 7. Let V be an n-dimensional (n ~ 1) vector space over

K and put E = En%(V). If x E E and v E V, then (in what follows) we
shall write xv rather than x(v).

151
 Let U, W be subspaces of V with U £: W. Put

GL(U, W) = {xix E GL(V), xw-w E U all w EW L

For any x in GL(V), the mapping x : V - V is an automorphism and

therefore Wand xW have the same dimension. Likewise U and xU
have the same dimension. But if x E GL(U, W), then xW £: Wand
xU £: U. Consequently, xW = Wand xU = U. A trivial verification now
shows that GL(U, W) is a subgroup of GL(V).
Assume that v E V, ~ E HomK(V, K) = V*, and define

t/I :E-K
v,~

by t/I v, .,t(x) = ~(xv). Evidently t/Iv, .,t belongs to HOIn._(E,

.IS..
K) and there-
fore t/I t E K[E]. Hence the restriction of t/I t to GL(V) belongs to
v,., v,.,
K[GL(V)].
Suppose that x E GL(V) and w EW. Then xw - w E U if and only
if ~(xw - w) = 0 for all ~ E V* such that ~(U) = O. But ~(xw - w) is
the value of t/I ~(w) at x. Consequently GL(U, W) consists of all
w,.,t -

elements of GL(V) that are common zeros of the functions t/I t - ~(w),
w, .,
where w ranges over W and ~ E V* satisfies ~(U) = O. This shows,
in particular, that GL(U, W) is a closed subgroup of GL(V) and hence
an affine group.
From here on we shall assume that K is an infinite field, w will
always denote an element of W, and ~ E V* will be assumed to satisfy
~(U) = O. Select wi' w 2 ' ••• , wp and ~i' ~2' ••• , ~p so that

{t/lw ,~ }1=:oJ.t:sp (5.3.3)

J.t J.t
is a base for the K-space spanned by all the functions t/I t and note that
w,"
if i denotes the identity of E, then t/Iw, .,t(i) = ~(w). It follows that every
t/IW,,,t - ~(w) is a K-linear combination of the functions

(5.3.4)

Accordingly GL(U, W) consists precisely of the elements of GL(V) that

are common zeros of (5.3.4).

152
 Now (5.3.3) can be enlarged to a base of HomK(E, K) and, because
K is infinite, K[E] is essentially a polynomial.ring with the members of
the base playing the role of indeterminates. t Accordingly the functions
in (5. 3. 4) generate a prime ideal, P say, in K[E] and P is the inter-
section of all the K-rational maximal ideals that contain it. Let X be
the locus of P. Then X is closed in E, it is irreducible, and
X n GL(V) = GL(U, W). Thus GL(U, W) is:r non-empty open subset of
X and therefore, by Chapter 3 Theorem 2, the closure of GL(U, W) in
E is X. It follows that if F E K[E], then F vanishes everywhere on
GL(U, W) if and only if F E P. Consequently the ideal of K[GL(V)] that
corresponds to GL(U, W) is just the extension of P to K[GL(V)] and,
in particular, it is prime. This proves

Theorem 18. Let K be an infinite field and U, W, where U ~ W,

subspaces of the n-dimensional K-space V. Then (with the above nota-
tion) GL(U, W) is a closed, connected subgroup of GL(V).

Example 8. Let A be a non-trivial, unitary and associative

K-algebra. (It need not be commutative.) Further let A, considered
as a vector space over K, have finite dimension. Then A has a natural
structure as an affine set. In what follows we use U(A) to denote the group
of units of A, that is U(A) consists of those elements which have two-
sided inverses.
Suppose that a EA and let Aa : A -+ A be the mapping in which
Aa(a) = aa. Then Aa is K-linear and if we put N(a) equal to the
determinant of A , we have
a
N(,Ba) = N(,B)N(a) (a, ,B E A)

and N(lA) = l K. In particular, if a has a right inverse, then N( a) *" O.

Conversely suppose that N(a) *" O. Then Aa is surjective and
therefore a,B = 1A for some ,B EA. But now N(,B) *" 0 and hence ,Br=1A
for some YEA. It follows that a = y and hence that a has a two-
sided inverse. Thus

t See Chapter 2 Theorem 19.

153
 U(A) = {a' a EA and N(a) *- 0 }.

It is clear that N, considered as a mapping of A into K, belongs

to K[A] and therefore we can turn U(A) into an affine set by using the
construction described in Example 4 of section (2.5).
We wish to consider U(A) as an affine group. If K is finite there
is no problem so from here on it will be assumed that K is infinite. In
this case

K[U(A)] = (K[A ])[1/N] (5. 3. 5)

and K[A] is a polynomial ring in n variables, where n is the dimension

of A as a vector space over K.
It is clear that the multiplication mapping

U(A) x U(A) .... U(A)

is a K-morphism. Define j : U(A) .... U(A) by j{a) = a-I. In order to

investigate j we select a base aI' a 2, •.• , a for A over K. If
n
a E A, then
n
all' = L c (a)a,
s r=1 rs r

where c : A .... K belongs to K[A] and the determinant of the matrix

rs
liers (a) II is N(a).
Let 1A· = t a + t a +... + t a , where t E K. Then, for
1 1 2 2 n n II
a E U(A),

-1
a = d (a) a + d (a)a + ... + d (a)a
1 1 2 2 n n

provided that
n
L c (a)d (a) =t (r = 1, 2, ••• , n).
s=1 rs s r
Solving these equations for d (a) shows that d : U(A) .... K belongs to
s s
(K[A])[1/N] = K[U(A)]

154
and thereby establishes that is a K-morphism. Accordingly we have
proved

Theorem 19. Let K be an infinite field and let the notation be as

above. Then U(A) is a connected affine group with (K[A ])[1 iN] as its
coordinate ring. The dimension of U(A) equals the dimension of A as
a vector space over K.

Let V be an n-dimensional (n 2:: 1) vector space over K and put

A = En%(V). Then Theorem 19 applies in this case. It is obvious that
U(A) and GL(V) coincide as abstract groups. Also if f : V -+ V belongs
to A and D(f) denotes the determinant of f, then N(f) = [D(f)t. This
shows that U(A) and GL(V) have the same coordinate ring and therefore
they coincide as affine groups.

Example 9. In this example (and the special cases derived from it)
K is assumed to be an infinite field whose characteristic is different
from 2.
Let n 2:: 1 be an integer and put G = GL (K). This is an affine
group. If n is an extension field of K, then ~(n) = GL (n) as may be
n
seen by applying Chapter 2 Lemma 14 and Theorem 1 Cor. 1. If A is
an n x n matrix we use Det(A) and AT to denote its determinant and
transpose respectively. We also use I to denote the n x n identity mat-
rix.
Suppose that B E GL (K) and put
n

H = {AlA EM (K) and ATBA

n
= BJ.
Evidently if A E H then Det(A) *- O. An easy verification shows that H
is a closed subgroup of G.
Now put

II = {p IP E M (K) and P T B + BP
n
= O}, (5. 3. 6)

sothat A isasubspaceoftheK-space Mn(K), let PI' P 2 ' ... , P q be

a base for A, and set

155
 p* = P Z
1 1
+ P 2 Z 2 +... + P q Zq , (5. 3. 7)

where Zl' Z2' .•. , Zq are indeterminates. The entries in p* are

linear forms in Z , Z , ... , Z and among these we can find q that
1 2 q T
are linearly independent. Also Det(1 - p*) *0 and Det(1 - p* ) *0 as
may be seen by putting all the Z. equal to zero. We may therefore define
1
a matrix A * by

A* = (I + P*)(I - p*fl. (5. 3. 8)

T
Then, because p* B + P*B = 0, we have

whence

A*TBA* = B. (5. 3. 9)

Again

(I - A*)(I - P*) = -2P*

and

(I + A*)(1 - P*) = 21.

Since the characteristic of K is not 2, the latter relation shows that

Det(1 + A *) * O. Consequently

p* = -(I - A*)(I + A*fl. (5.3.10)

It follows, from (5.3.8) and (5.3.10), that the entries in p* generate,

over K, the same field as those in A * and hence that K(P*) = K(A *).
(Here p* is regarded as a generalized point of M (K) and A * as a
n
generalized point of G.) Accordingly

(5.3.11)

The entries in A * are rational functions in Z l' Z2' ••• , Zq and

they can be written with

156
(say) as a common denominator. If Z l' Z 2' ..• , Zq are assigned values
in K which do not make tP vanish, then (5. 3. 9) shows that A * is turned
into a matrix belonging to H. It follows that if F E IG(H), then F(A *) = O.
Consequently, by Chapter 2 Theorem 34, A* E H(K(Z)) where K(Z) is
used as an abbreviation for K(Z , Z , ..• , Z ).
1 2 q
Let H 0 be the connected component of the identity of H and let
f E ~(Ho)' Then we can find g E K[H] which vanishes on all the irre-
ducible components of H other than Ho and is such that g(I)"* O. Then
fg = 0 and hence f(A *)g(A *) = O. Since I is a specialization of A *
when they are regarded as points of G, the same holds t for H. Con-
sequently g(A*)"* 0 and therefore f(A*) = O. Thus, by Chapter 2
Theorem 34 A* E H(K(Z))
, o'
We claim that A * is a generic point of H o' To see .this first
suppose that A E Hand Det(I + A) "*
a
o. Put

-1
P = -(I - A)(I +A)

so that
pTB = -(I + A Tf1(I _ A T)B

BP =-B(I-A)(I+Af 1

and therefore P T B + B P = O. It follows that P can be obtained from

P * by giving Z l' Z 2' •.• , Zq suitable values in K. Next

(I - P)(I + A) = 21

whence Det(I - P) "* 0 and now it follows that

A = (I + P)(I - Pf1.

Thus in view of (5. 3. 8) A can be obtained from A * by giving

Z1' Z2' •.. , Zq suitable values in K and therefore A is a specialization

t See Chapter 2, Theorem 34 Cor. 1.

157
of A *. Here in the first instance we regard A * and A as a generalized
point and an ordinary point respectively of GL (K); but then Chapter 2
n
Theorem 34 Cor. 1 shows that our conclusion remains valid if we regard
them as belonging to H •
a
Finally suppose that f E K[Ho] and f(A*) = O. The above remarks
show that f(A)Det(I + A) =0 for all A E Ho' Thus fh = 0, where
h E K[Ho] and h(A) = Det(I + A) when A E Ho' Now K[Ho] is an
integral domain and h *- 0 because h(I) *- O. Accordingly f =0 and we
have established our claim that A * is a generic point of H. It follows
a
that Dim H = Dim Ho is equal to the transcendence degree of K(A *)
over K and this is q by (5.3. ll). We combine our main conclusions in

Theorem 20. Let K be an infinite field whose characteristic is

different from 2 and let B E GL n(K). -
Then
-

H = {A IA EMn (K) -
and
-
ATBA = B}
is a closed subgroup of GLn (K). -
Put
-

A = {p IP E M (K) and P T B
n --
+ BP = 0 }.

Then Dim H is equal to the dimension of 11 as a vector space over K.

Moreover if A* is defined as in (5.3.8), then A* is a generic point of
the connected component of the identity of H.

Theorem 21. Let the assumptions and notation be as in Theorem

20. If the matrix B is symmetric, then Dim H = n(n - 1)/2. On the
other hand, if B is skew symmetric we have Dim H = n(n + 1)/2.

Proof. First assume that B is symmetric and let P E M (K).

n
Then PEA if and only if (BP? = -BP. Accordingly A consists of all
matrices B-1Q, where Q is a typical skew symmetric matrix. Thus the
vector space 11 has the same dimension as the space of all skew sym-
metric n x n matrices and this is n(n - 1)/2.
Now assume that B is skew symmetric. This time A consists of
all matrices B-1S, where S EM (K) and is symmetric. These symmet-
n
ric matrices form a vector space of dimension n(n + 1)/2 and this there-

158
fore is the dimension of A. The desired results follow by virtue of
Theorem 20.
If B is the n x n identity matrix I , then B is non-singular and
n
symmetric. In this case the group H of Theorem 20 is denoted by
on (K) and called the orthogonal group.
Finally if n is an even integer, say n = 2m, we may take B to
be the skew symmetric matrix

This time H is known as the symplectic group. The notation for the
symplectic group is SP (K).
m

5.4 K-homomorphisms of affine groups

Throughout section (5.4) G and G' denote affine groups defined

over K.

Definition. A mapping cp: G'" G' is called a 'K-homomorphism'

of affine groups if cP is a K-morphism of affine sets and cp(xy) =CP(x)cp(y)
for all x, y e: G.

Note that the identity mapping of G is a K-homomorphism and that

K-homomorphisms ¢: G ... G' and 1/1: G' ... G" combine to give a K-
homomorphism 1/1 0 cp of G into G".
A K-homomorphism cP : G'" G' is called a K-isomorphism (of
affine groups) if (a) cp is a bijection, and (b) cP -1 : G' ... G is a K-
homomorphism as well. The affine groups G and G' are said to be
K-isomorphic if there exists a K-isomorphism of G on to G'. This
relation is reflexive, symmetric and transitive. For example if V is an
n-dimensional (n 2::: 1) vector space over K, then GL(V) and GL (K)
n
are K-isomorphic affine groups.
Let cP : G'" G' be a K-homomorphism of affine groups and put
N = Ker cpo Then N is a closed subgroup of G because cP is continuous
and N is the inverse image of the identity of G'. For instance we have

159
a K-homomorphism GL(V) - GL (K) in which each f in GL(V) is
1
mapped into its determinant. In this case the kernel is SL(V).

Theorem 22. Let C/l: G - G' be a K-homomorphism of affine

groups and suppose that K is algebraically closed Then cp(G) is a
closed subgroup of G'.

Proof. It is enough to show that cp(G) is closed in G'. Let Go

be the component of the identity of G and assume that we can show that
C/l(G ) is closed in G'. Since, by Theorem 10, cp(G) is the union of a
o
finite number of cosets of cp(G o)' the desired result will then follow. We
shall therefore assume that G is connected and hence irreducible.
Put H = cp(G). Then H is a connected, closed subgroup of G' and
cp induces an almost surjective K-homomorphism I/J: G - H.
- Theorem 33
of Chapter 3 now shows that I/J(G) = cp(G) contains a non-empty subset U
which is open in H. However, by Theorem 14 Cor., H = UU ~ I/>(G).
The theorem follows.

Corollary. Let I/J : G - G' be a K-homomorphism of affine groups.

!! I/J is almost surjective and K is algebraically closed, then cp is sur-
jective.

For the general situation, where K is not necessarily algebraically

closed, we have

Theorem 23. Let 1/>: G - G' be an almost surjective K-homomor-

phism of affine groups. Further let Go respectively G~ be the connected
component of the identity of G respectively G'. Then cp(G )
-- 0
= G'.0
Proof. By Theorem 10, the index of Go in G is finite. Let
xl' x 2 ' ••• , xm be representatives of the different cosets. Then
m
cp(G) = U CP(x. )I/J(G )
i=l 1 0

and therefore
m
G' = cp(G) = U cp(x.)~.
i=l 1 0

160
Thus "fl.GJ
o
is a closed subgroup of G' whose index in G' is finite.
Consequently, by Theorem 11, G' ~"fl.GJ. On the other hand lj>(G ) is
o 0 0
connected and contains the identity of G'. Accordingly "fl.GJ
o
~ G' and
0
the theorem is proved

Corollary. Let the assumptions be as in the theorem and suppose

that K is algebraically closed. Then lj>(G ) = G' •
-- 0 0

This now follows from Theorem 22.

Before proceeding to the next theorem we note that if cp: G'" G' is
a K-homomorphism of affine groups and L is an extension field of K,
then lj>(L) : G(L) ... G,(L) is an L-homomorphism of affine groups. (This
is easily seen by using Lemma 1.) Also if lj> is almost surjective, then
so too is lj>(L). The latter assertion follows from (2.10.10).

Theorem 24. Let lj>: G'" G' be an almost surjective K-homomor-

phism of affine groups and put N = Ker cpo Then Dim N :::; Dim G - Dim G'.
If K is algebraically closed, then Dim N = Dim G - Dim G'.

Proof. We begin by considering the case where K is algebraically

closed Let Go respectively G~ be the connected component of the
identity of G respectively G'. Then Dim Go = Dim G, Dim G~=DimG',

and by Theorem 23 Cor., lj> induces a surjective K-homomorphism

G ... G' whose kernel is N n G = N (say). Since the index of G in
o 0 0 1 0
G is finite, the index of N1 in N is finite as well. It follows that
Dim N1 = Dim N and we have reduced the problem (for the algebraically
closed case) to the situation where G and G' are connected and hence
irreducible.
By Chapter 3 Theorem 37 Cor., there exists y € G' such that
cp-l( {y}) has a component whose dimension is Dim G - Dim G'. Let x
belong to this component. Then lj> -1 ( {y }) = xN. But Nand xN are
K-isomorphic and every component of N has dimension equal to Dim N
by virtue of the corollary to Theorem 10. This shows that
Dim N = Dim G - Dim G' and we have dealt with the case where K is
algebraically closed
Now let K be an arbitrary field and let L be its algebraic closure.

161
We know that cp extends to an almost surjective L-homomorphism
cp(L) : G(L) _ G,(L). Put N* = Ker cp(L). Then N ~ N* and

Dim N* = Dim G(L) - Dim G,(L) = Dim G - Dim G'.

By Chapter 2 Theorem 30, N(L) ~ N*. Consequently

Dim N = Dim N(L) :S Dim N*

and with this the proof is complete.

5.5 K-morphic actions on an affine set

Let G be an affine group and V a non-empty affine set where both

G and V are defined over K. A left action of G on V is a mapping

a:GXV-V,

where
(i) a(a, a(T, v)) = a(aT, v) for all a, T E G and v E V;
a(e, v) = v, where e = e G and v E V.
(ii)
When we have such an action we normally write av in place of a(a, v).
In this notation (i) and (ii) become a(TV) = (aT)v and ev =v respectively.
Suppose that a is a left action of G on V and that a is also a
K-morphism of affine sets. We then say that a is a K-morphic left
action of G on V or, less formally, that G acts morphically on the
left of V.
A K-morphic right action of G on V is a K-morphism V x G - V
with analogous properties. Thus in a right action (va)T = v(aT) and
ve = e. Of course results concerning right and left actions tend to come
in pairs. We shall develop the theory mainly in terms of left actions.
Suppose that we have a K-morphic left action of G on V. The
mapping

A 'V-V (5. 5. 1)
a'

in which Aa(V) = av is a K-automorphism of V with A -1 as its

a

162
inverse. It therefore induces an automorphism

A~ : K[V] - K[V] (5. 5. 2)

of K-algebras. It will be convenient to put

(5. 5. 3)

Accordingly we have the following formulae for a left action:

fa(v) = f(av),

(f +f ) a = fa + fa
1 2 1 2'
(f f)a = fal (5. 5. 4)
1 2 1 2'
(fa) T = faT,

ka =k for k E K.

Similar remarks apply to K- morphic right actions. The corresponding

formulae for a right action are

fa(v) = f(va),

(f +f ) a = fa + fa
1 2 1 2'
(f f)a = fafa (5. 5. 5)
1 2 1 2'

(fa)T = f Ta,

ka = k for k E K.

If G is an affine group, then the multiplication mapping

/J.:GXG-G

is a K-morphism and it defines both a left and a right action of G on

itself. These are known respectively as the regular left action and
regular right action. Contrary to what was said above, when we consider
a regular action of G on itself it will usually be the regular right action
that will concern us. The reason for this will appear later.

163
 Theorem 25. Let the affine group G act morphically on the left
of the affine set V and let f E K[V]. Then the K-space spanned by all
fa, where f is fixed and a varies in G, has finite dimension.

This is essentially a generalization of part of Theorem 6. Since

the proof involves no new ideas we shall not give details.
At this point it is convenient to note that if a: G x V - V is a
K-morphic left action of G on V and L is an extension field of K,
then a(L): G(L) x V(L) - V(L) is an L-morphic left action of G(L) on
V(L).
Once again let the affine group G act morphically on the left of the
affine set V, and for v E V put

StabG(v) = {ala E G and av = v}.

Obviously StabG(v) is a subgroup of the abstract group G. Now a ..... av

is a K-morphism G - V and we have a second K-morphism of G into V
in which every element of G is mapped into the given element v. An
application of Lemma 1 at once shows that StabG(v) is a closed subgroup
of G.
Suppose that VI' V 2 E V and let us write VI - V2 if v 2 = av 1 for
some a E G. This is an equivalence relation and the equivalence classes
are known as orbits. The orbit that contains the element v is Gv.
Let X be a subset of V. We say that X is stable under G if
aX ~ X whenever a E G. When this is the case we have oX = X for all
a € G, and the orbit of any x E X is contained in X.
Suppose that X is stable under G. Then its closure X, in V, is
also stable under G. It follows that X\X is a stable set as well. Hence
if N is an orbit, then N\N is stable under G.

Theorem 26. Let the connected affine group G act morphically on

the left of the affine set V, let N be an orbit, and let K be algebraically
closed. Then N is open in its closure N. If the orbit N is such that
Dim N is as small as possible (for the different orbits of G in V), then
N = N (that is N is a closed orbit).

164
 Proof. Let v EN and consider the K-morphism G - N in which
a t-+ avo The morphism is almost surjective and its image is N. Con-
sequently, by Chapter 3 Theorem 33, there exists a non-empty subset
T, of N, which is Open in N and contained in N. Next

N
U
= aEG aT.

Also N is stable under G and therefore each aT is open in N. It

follows that N is open in N.
Now assume that w EN\N and put N = Gw. Then N ~ N\N
o 0
whence N ~ N\N because N is open in
o N. Thus N 0 ~ N, N *-
0
N.
Moreover N is irreducible because G is irreducible. Accordingly, by
Chapter 3 Theorem 22, Dim N < Dim N. Hence if Dim N is minimal,
o
then N = N.

Corollary. Let G be an affine group (not necessarily connected)

which acts morphically on the left of the affine set V. Further let N
be one of the orbits and suppose that Dim N is minimal (in the sense of
the theorem). Then N = N.

Proof. First suppose that N is any orbit. It is clear that Go

acts rnorphically on V. Choose a = e, a , ••• , a
o 1 m to represent the
different cosets of G in G and let v € N. Put N = G v and
o 0 0

N. = a.N = a.G
11010
v = G (a. v).
01

Then N , N , ..• , N
o 1 m are orbits of Go'

and

Next each Nt is irreducible and N. = a.(N). Thus the N. are pairwise

110 1
K-isomorphic. It follows that Dim N = Dim N. for all i. Accordingly
1
when Dim N has the smallest possible value the same holds for the
closure of the G -orbit N1.• Consequently
o
N.1 = N.,1 by the theorem, and

165
hence N = N.

5. 6 G- modules

Let G be an arbitrary group and V a vector space over K.

Suppose that a mapping

Ct : G x V"'V

has been given and let us denote the image of (a, v), where a E G and
v E V, by avo If now
(i) a(v
1
+ v2 ) = av
1
+ av2 for v ,v E V and a E G;
1 2
(ii) a(kv) = k(av) for k E K, a E G, and v E V;
(iii) a (a v) = (a a )v for a, a E G and v E V;
1 2 1 2 1 2
(iv) ev = v when e = eG and v EV;
then we say that V is a left (G, K)-module. If we are in a situation
where K is being kept fixed, then we often say simply that V is a left
G-module. This applies particularly when G is an affine group. In this
case it is always to be understood that ~( is the ground field over which
G is defined.
The notion of a right (G, K)-module is obtained by making the obvious
modifications to the above definition. Thus a right G-module has to do
with a mapping V x G ... V in which the image of (v, a) is denoted by
va, (va) 7 = v(O"T), and so on.
We can express the above ideas in ring-theoretic terms. To be
explicit, let KG denote the group ring of G with coefficients in K.
Thus KG is the K-space which has the members of G as a base; and if

L kaa and L k' 7

aEG uG 7

are two elements of KG, then their product is given by

L k a) (L k'
( aEG a 7EG 7
7) = LL k k' (a7).
a 7
It is clear that the notion of a left respectively right (G, K)-module is
essentially the same as a left respectively right KG-module as these
latter terms are understood in the theory of rings and modules.

166
 Suppose now that V and Ware left (say) G-modules. A mapping
f :V ~ W is called a G-homomorphism if it is a linear mapping of K-
spaces and f(av) = af(v) for all (J E G and v E V. If in addition to being
a G-homomorphism f is also bijective, then f is called a G-isomorphism.
In terms of the group ring KG, a G-homomorphism respectively G-
isomorphism is the same as a homomorphism respectively isomorphism
of KG-modules.
Now suppose that G is an affine group defined over K and that
V is a (G, K)-module, where the dimension of V over K is finite.
(For definiteness we shall suppose that V is a left G-module.) If :J E G,
then the mapping

~ (J''V-V (5. 6. 1)

in which ~(J(v) = av, belongs to En~(V). Consequently, there is a

mapping

~ : G - En~(V), (5. 6. 2)

where ~((J) = ~(J' (A similar situation is obtained if V is a right G-

module. )

Definition. If the G-module V has finite dimension over K, then

V is called a 'rational' G-module if (5.6.2) is a K-morphism of affine
sets.

Once again let V be a finite-dimensional G-module and let

ul ' u2 ' ••• , un be a base for V over K. Then
n
~ (u.)
(J J
= L=1 ~IJ.. ((J)u.,1
i
where the ~.. are mappings of G into K.
IJ

Lemma 3. Let G be an affine group and V a finite-dimensional

left G-module. Further let the notation be as described above. In these
circumstances the following statements are equivalent:
(a) V is a rational G-module;
(b) all tl1e functions ~ .. belong to K[G]:
IJ

167
 (c) the mapping a: G x V - V, where a(a, v) = av, is K-
morphic.

Remark. Naturally a similar result holds concerning right G-

modules.

Proof. We may assume that K is infinite, for if K is finite,

then (a), (b) and (c) all hold trivially.
Assume (a) and define wIll) : EncL(V)
-X
- K for 1::'S Il, l) ::'S n by
W (f):::; a ,where 1/ a .. // is the matrix of f with respect to
Ill) Ill) IJ

1 2
u.
U , U , ••• ,
n
Then wIll) E K[EncL(V))
-X
and so AIll) :::; WIll) 0 A E K[G)
because A: G - En%(V) is a K-morphism by hypothesis. Thus (a)
implies (b).
Assume (b). There exist Y , Y , ••• , y in K[V] such that if
1 2 n
V € V, then

v = y 1 (v)u1 + Y (v)u
2 2
+ ... + Yn (v)un.

Hence for f E K[V) we have

f(av) :::; f(Aa(V))

= fe L Y.(V)A (u.)\
j J a J')
=f(LL
i j
Y.(V)A . .(a)u.)
J IJ 1

= q(L A1.(a)y.(v), L A2 ·(a)y.(v), ••• , L A .(a)y.(v~,

j J J j J J j nJ J ')
where q(X , X , ••• , X ) is a certain polynomial with coefficients in K.
1 2 n
Hence if we use the notation introduced in (2. 7.2), then

W 0 a:::; q(L A.. v y.,

\'j IJ J
L
j
A2 . v y., ••• , LA. v
J J j nJ
Y.) J
and this belongs to K[G x V). Thus (b) implies (c).
n
Finally assume (c). Since au. =L A.. (a)u., it follows that
J i=1 IJ 1
A.. (a) = y.(au.) :::; (y. 0 a)(a, u.),
IJ 1 J 1 J

where y. is defined as in the last paragraph. But y. 0 a E K[G x V)

1 1

168
and thus we see that \j E K[V] for all i and j. Let X E K[En~(V)].
There is a polynomial p(X11 ,X1 2, ...
·
, Xnn ), with coefficients in K,
such that if f E EncL(V) and has matrix II a .. ", then
-X 1J
X (f) = p(a11 , a 1 2' ••• , ann)' Accordingly

(X o oX)(0-)=p(oX11 (a), oX (a), ••• , oX (0-))

12 nn

and therefore

Thus oX is a K-morphism and we have shown that (c) implies (a). This
establishes the lemma.
As before let V be a left G-module of finite dimension. The
mapping oXo-: V -+ V is not just an endomorphism but an automorphism
with inverse oX -1' Thus we arrive at a homomorphism
0-
oX o .. G -+ GL(V) (5. 6.3)

of abstract groups, where oX (0-) = oX. (In the case of a right G-module
o 0-
we obtain an anti-homomorphism.) If (5.6.3) is a K-homomorphism of
affine groups, then we say that we have a rational representation of G
by means of automorphisms of V.

Corollary. Let G be an affine group and V a finite-dimensional

left G-module. Then V is a rational G-module if and only if (5.6.3)
is a rational representation of G.

Proof. It is clear that we may assume that K is infinite. Next the

natural mapping GL(V) -+ En~(V) is a K-morphism so half the corollary
is clear. Now assume that V is a rational G-module and let XEK[GL(V)].
If f with matrix "a .. 1/ is in GL(V), then
IJ
h
X(f) = p(al l , ••. , ann)![Det(f)] ,

where P(X 11 , ••• , Xnn) belongs to the polynomial ring K[X11 , ••• , Xnn]
and h> 0 is an integer. Hence

169
 h
(x 0 A )(0') = P(A (a), ••• , A (a))/[Det(A (a))] •
o 11 nn 0

Thus

h
X 0
o = p(A 11 , ••. , Ann )/[Det
A 0 A]
0

and we have to show that this belongs to K[G]. By the lemma, all the
A.. are members of K[G]. Consequently P(A 11 , ... , A ) E K[G]. Con-
~ M
sider the mapping a 1-+ Det(A (a)). Since this is just Det 0 A it belongs
o -1 h
to K[G]. It follows that 0'1-+ [Det(Ao(a ))] also belongs to K[G]. But

[Det(A o(a- 1 ))]h = [Det Ao(a)r h

because A (a-I) is the inverse of A (a). Accordingly [Det 0 A rh
0 0 0
belongs to K[G] and the corollary follows.
Observe that this result shows that the notion of a finite-dimensional
rational G-module is equivalent to that of a rational representation of G
by automorphisms of a finite-dimensional K-space.

Example 1. Let V be a finite-dimensional vector space over K.

For f E GL(V) and v E V, put fv = f(v). This turns V into a rational
left GL(V)-module.

Example 2. Let G be an affine group and let G act on itself by

means of the regular right action. For f E K[G] and a E G put af = fa.
Then, by (5.5.5), the mapping f 1-+ af is K-linear, ef =f and

0'( Tf) = (f)T a = f aT = (aT)f.

Thus K[G] has become a left (G, K)-module.

Let us keep f fixed and denote by N the K-subspace of K[G] which
is generated by the elements fa (a E G). By Theorem 25, the dimension
of N as a vector space over K is finite. Also aN = N for all a E G
and, moreover, fEN. Thus N is a left (G, K)-module and it contains f.
The next example shows that N is a rational G-module.

Example 3. Let G be an affine group and suppose that M is a

subspace of the K-space K[G]. We assume that the dimension of M

170
(over K) is finite and that aM = M for all a E G. Here K[G] is regar-
ded as a left G-module as in Example 2.

Lemma 4. With the above assumptions M is a rational G-module.

Proof. Let f 1 , f 2 , ••• , fn be a base for Mover K. Then

a n
ai, = L = L A.. (a)f.
J J i=1 IJ 1

for a E G, where A.. : G .... K. Accordingly for y E G,

IJ
n
L(ya) = L L(a)L(y).
] i=1 1] 1

If Y is kept fixed, then the mapping G .... K in which a"'" L(ya)

J
belongs to K[G). Next, by Chapter 2 Theorem 13, we can find
y l' Y2' ••• , y n in G so that the determinant Ifi (yk) I is non- zero.
Hence, by solving the equations
n
L (Yk a ) = i=1
L A.. (a)f. (Yk) (k = 1, 2, •.. , n),
] IJ 1

we find that L(a) is a certain linear combination (with coefficients in K)

IJ
of L(y a), L(y a), ••• , L(y a). This shows that A.. E K[G] and therefore
J 1 J 2 J n IJ
the desired result follows from Lemma 3.
We shall now give an important application of the ideas contained in
Examples 2 and 3.
Let G be an affine group, choose f , f , ••• , f so that
s 1 2
K[G) = K[f , f , •.. , f ), and let M be the subspace of the K-space
1 2 S a
K[G) generated by the set fi (1:5 i:5 s, a E G). Theorem 25 shows that
the dimension of M is finite and evidently aM = M for all a E G.
Accordingly, by Lemma 4, M is a rational G-module. Let us select
g , g , .•• , g
1 2 n so that they form a base of Mover K. Note that
K[G] = K[gl' g , •.• , g ] and, by Lemma 3 Cor., we have a K-homo-
2 n
morphism

p : G .... GL(M) (5.6.4)

a
of affine groups. Suppose that a € Ker p. Then gi = gi and thus right

171
translation by means of a induces the identity automorphism of the
K-algebra K[G]. Accordingly a = e and therefore Ker p = {e 1.
From here on we shall assume that K is algebraically closed. By
Theorem 22, p(G) is a closed subgroup of GL(M). Hence p induces a
bijective K-morphism

p : G'" p(G) (5. 6. 5)

of affine groups.
To p there corresponds a homomorphism

p* : K[GL(M)] ... K[G]

of K-algebras. We claim that p* is surjective. To see this define

p .. : G-K by
IJ
n
(p(a))(g.) = g~ = L p .. (a)g .•
J J i=1 IJ 1

Then
a n
g.(a) = g. (e) = L p .. (O')g.(e).
J i=l 1J 1
Next let the polynomial F .(X , ... , Xnn) be given by
J 11
n
F.(X11 , ••• , X ) = L X .. g.(e)
J nn i=1 IJ 1
and regard it as belonging to K[GL(M)]. Then

(F j 0 p)(O') = Fj(p(a))
= Fj{Pll (a), ••• , Pnn{a))
n
=L p .. {a)g.{e)
i=l IJ 1
= g.{O').
J

Consequently p*{F.) = g. and our claim follows.

J J
Next we have homomorphisms

G - p{G) .... GL{M)

172
where the first is -p and the second is an inclusion mapping. These give
rise to homomorphisms

K[GL(M)] .... K[p(G)] - K[G]

of K-algebras. We have just seen that

p* : K[GL(M)] - K[G]

is surjective and now it follows that

15* : K[p(G)] - K[G]

is surjective as well. Let g € K[G]. Then g is the image of some

l/I € K[p(G)], that is to say l/I 0 15 = g. But 15 is bijective. Consequently

g 015- 1 = l/I € K[p(G)].

Accordingly

15-1 : p(G) .... G

is a K-morphism and therefore

15: G -p(G)
is a K-isomorphism of affine groups. Thus we have proved the following
striking result.

Theorem 27. Let G be an affine group defined over an algebraically

closed ground field K. Then there exists a finite-dimensional K-space
M such that the group G is K-isomorphic to a closed subgroup of
GL(M).

5. 7 General rational G- modules

If G is an affine group and V is a G-module, then already in

section (5. 6) we have explained what is meant by saying that V is a
rational G- module in the restricted situation where the dimension of V
(over the ground field) is finite. This concept will now be extended. First,

173
however, we shall introduce some general terminology.
Suppose that G is an abstract group and V a left (G, K)-module.
Further let U be a subset of V. Then U is said to be a (G, K)-sub-
module of V provided
(a) U is a K-subspace of V;
(b) au E U for all CT E G and u E U.
Note that if KG denotes the group ring of G, then a (G, K)-submodule
of V is the same as a KG-submodule in the sense in which this term is
used in the theory of rings and modules. (Naturally similar definitions
and observations apply to right (G, K)-modules.) Often we omit any
explicit reference to K and speak of a G-module and its G-submodules.
In what follows a G-module is said to be finite-dimensional when its
dimension as a K-space is finite.

Lemma 5. Let G be an affine group and V a finite-dimensional

G-module. If now V is a rational G-module, then all its G-submodules
are rational.

This follows immediately from Lemma 3.

Let G be an affine group, VaG-module of arbitrary dimension,
and U a G-submodule of V. Then U is a G-module and if it happens to
be finite-dimensional, then it could be a rational G-module in the sense
of section (5. 6).

Lemma 6. Let U , U , .•• , U be finite-dimensional rational

- 12m
G submodules of the G- module V. Then U +U +... + Umis-a-
--12
finite-dimensional rational G-submodule of V.

Proof. We may assume that m = 2. Choose a base for U n U

1 2
and extend it so as to obtain bases for U and U. In this way we obtain
1 2
a special base for U1 + U2 ' The desired result follows by applying
Lemma 3 to this base.
Once ag'lin let G be an affine group and V a G- module.

General definition. We say that V is a 'rational' G-module if

every element of V is contained in a finite-dimensional rational G-sub-
module.

174
 Note that if V itself is finite-dimensional, then the new definition
agrees with the old This is so in view of Lemma 6.

Theorem 28. Let G be an affine group and V a rational G-

module. Then every G- module of V is a rational G- module.

This follows from Lemma 5 and the definition of a general rational

G-module.
We have already seen that, when G is an affine group, the coordi-
nate ring K[G] can be regarded as a left G-module by making use of the
regular right action of G on itself. In this situation we have af = l",
when f E K[G] and (J E G.

Theorem 29. Let G be an affine group and let K[G] be con-

sidered as a left G-module in the manner described above. Then K[G]
is a rational G- module.

This follows from Examples 2 and 3 in section (5. 6).

Definition. A G-module V is said to be 'simple' if V '" 0 and

the only G-submodules of V are the zero submodule and V itself.

Theorem 30. Let G be an affine group and V a rational G-

module. If V is a simple G-module, then the dimension of V (over the
ground field) is finite.

This is clear from the definitions.

Theorem 31. Let G be an affine group and V '" 0 a rational

G-module. Then V contains a simple G-submodule.

Proof. We can find a G-submodule U '" 0 whose dimension is finite.

If we arrange that the dimension is as small as pOSSible, then U will be a
simple G-submodule.

5. 8 Linearly reductive affine groups

Let G be an abstract group.

175
 Definition. A (G, K)-module V is said to be 'completely
reducible' if given any (G, K)-submodule U there exists a second
(G, K)-submodule U' such that V = U EB U' this being a direct sum of
vector spaces.

Evidently every (G, K)-submodule of a completely reducible (G, K)-

module is itself completely reducible.

Lemma 7. Let V be a (G, K)-module. Then the following state-

ments are equivalent:
(a) V is completely reducible;
(b) V is a sum of simple (G, K)-modules;
(c) V is a direct sum of simple (G, K)-modules.

Proof. If one restates the lemma in terms of modules over the

group ring KG, then it becomes a special case of a familiar result in the
theory of semi-simple modules over a ring. t

Definition. An affine group G is said to be 'linearly reductive'

if every rational G- module t is completely reducible.

Theorem 32. Let G be an affine group and suppose that every

finite-dimensional rational G-module is completely reducible. Then every
rational G-module is completely reducible, that is to say G is linearly
reductive.

Proof. Let V be a rational G-module. Then V is a sum of

finite-dimensional rational G-submodules and each of these is a sum of
simple G-submodules. Thus V itself is a sum of simple G-submodules
and therefore it is completely reducible.
Let G be an affine group and let U1 ' U 2 ' ••• , Urn be G-modules.
Their direct sum

U=U EBU EB ... ffiU

12m

t See, for example, [(3) Theorem 11, p. 61].

t For the remainder of this chapter all G- modules will be understood to
be left G-modules.

176
has a natural structure as a G-module. Evidently if each U. is a
1
rational G-module, then U is a rational G-module. Again if each U.
1
is a completely reducible G-module, then the same holds for U.
Suppose that V is a G- module and that d > 0 is an integer. As
is customary we write

d
V =VEfJVEfJ ... EfJV,

where there are d summands.

At this point we make a fresh start. Let G be an affine group and
V an n-dimensional, rational G-module. By Theorem 29, K[G] has a
natural structure as a rational G-module. Let F E V* = Ho~(V, K),
let v EV and define

F(v) : G-K

by

(F(v))(a) = F(av).

Since V is a rational G-module, it follows readily that F(v) E K[G].

Evidently

F: V -K[G]

is K-linear; also if T E G, then

(F(TV)(a) = F(aTV)
= (F(v))(aT)
= (F(v)) T(a)

and therefore F(TV) = TF(v). Consequently F : V - K[G] is a G-homo-

morphism.
Let F , F , ••. , F be a base for V* over K and define a
1 2 n
mapping V - (K[G])n by

v t-+ (F (v), F (v), ... , F (v)).

1 2 n

This is both a G-homomorphism and an injection. We have therefore

proved

177
 Theorem 33. Let G be an affine group and let K[G] be regarded
as a rational left G-module as in Theorem 29. If now V is an n-
dimensional rational G-module, then there is a G-submodule of (K[G])n
which is G-isomorphic to V.

Theorem 34. Let G be an affine group and let K[ G] be regarded

as a rational G-module as in Theorem 29. Then G is linearly reductive
if and only if K[G] is completely reducible.

This follows from Theorems 32 and 33.

Let us return to the proof of Theorem 33. The argument used shows
that if V *- 0, then, for some i, the G-homomorphism F. : V'" K[G] is
1
not null. Hence if V is a simple rational G-module, then we can find
F E V* such that the G-homomorphism F: V ... K[G] is not null and hence
an injection. This observation establishes

Theorem 35. Let G be an affine group and let K[G] be regarded

as a rational G-module as in Theorem 29. Let V be a simple, rational
G-module. Then V is G-isomorphic to a G-submodule of K[G].

If G is a finite group, then every finite-dimensional G-module is

rational, and therefore all G-modules are rational. The classical theory
of finite groups shows that if K has characteristic zero or the character-
istic of K does not divide the order of G, then every G-module is com-
pletely reducible. This is because KG is a semi-simple ring by virtue
of Maschke's Theorem. t
It is not easy to give other examples of linearly reductive groups
without making use of the theory of Lie algebras, but an exception arises
in the case of tori. These groups are of considerable interest and we
shall take this opportunity to discuss some of their properties.
In section (5. 3) we defined a torus to be a finite direct product

G = G x G x ... x G (s factors), (5.8.1)

m m m

where G = GL (K) is the multiplicative group of K. It is convenient

m 1
to extend this definition and describe as a torus any affine group which is

t See, for example, [(10) §125, p. 193].

178
K-isomorphic to one of these groups. We shall also regard a trivial
group (that is a group with no elements apart from its identity) as a torus.
This amounts to admitting the possibility that, in (5.8.1), the integer s
may be zero.
Let us suppose that K is infinite. Then the coordinate ring of the
group G in (5. 8. 1) is

where Xl' X 2 ' ••• , Xs are indeterminates. Assume that vI' V 2 , ... , V s
are integers (not necessarily positive), let c E K and put
vI v 2 Vs
F = cX 1 X2 ••• Xs' If now 0-= (0-1 , 0-2 , ••• , o-s), where o-i E Gm ,
0- vI v2 liS
then it is easy to check that F = 0- 0- • •• 0- F. Hence the subspace
1 2 s
V V v
of K[G] generated by XlIX 2 ••• X s is a G-submodule and, since it is
2 s
one-dimensional, it is simple. Note that we have shown that K[G) is a
sum of one-dimensional G-submodules.
Next suppose that H is a closed subgroup of G. If V is a G-
module, then it has a natural structure as an H-module; and if it is a
finite-dimensional rational G-module, then it is also a finite-dimensional
rational H-module. Hence any rational G-module can be regarded as a
rational H-module.

Theorem 36. Let G be a torus defined over an infinite field K

and let H be a clos.ed subgroup of G. Then K[H), considered as a
rational H-module as in Theorem 29, is a direct sum of one-dimensional
(and hence simple) H-submodules. Consequently H is linearly reductive.

Proof. We have a surjective homomorphism K[G) - K[H) of K-

algebras and each of K[G) and K[H] is a rational H-module. An easy
verification shows that K[G) - K[H) is a homomorphism of H-modules.
Now K[G) is a sum of one-dimensional G-submodules and these sub-
modules are also H-submodules. It follows that K [H) is a sum of one-
dimensional H-submodules. Finally this sum can be refined to give a
direct sum by omitting suitably selected summands.

179
 Corollary. Let G, K and H be as in Theorem 36 and let V ~

a rational H-module. Then V is a direct sum of one-dimensional H-

submodules.

Proof. Let S be a rational, simple H-module. In view of

Theorem 36 it will suffice to show that S has dimension one. By
Theorem 35, S is H-isomorphic to a submodule of the completely
reducible H-module K[H] and therefore it is a homomorphic image of
K[H]. Accordingly the simple H-module S is a sum of one-dimensional
H-submodules and therefore it too must be one-dimensional.
Let us take these ideas a little further.

Theorem 37. Let H be an affine group defined over an algebraically

closed field K. Then the following statements are equivalent:
(i) every rational H-module is a direct sum of one-dimensional
H-submodules i
(ii) every finite-dimensional rational H-module is a direct sum of
one-dimensional H-submodules;
(iii) H is a closed subgroup of a torus.

Proof. The equivalence of (i) and (ii) is clear. Also the corollary
to Theorem 36 shows that (iii) implies (i).
Assume (ii). We shall deduce (iii) and in view of Theorem 27 we
may suppose that H is a closed subgroup of GL(V), where V is some
finite-dimensional vector space over K. Now V is a rational GL(V)-
module and therefore it is a rational H-module. Consequently, since we
are assuming (ii), we can find a base e , e , •.• , en of V with the
1 2
property that each Ke. is an H-submodule of V.
1
With the usual notation

K[GL(V)]=K[X l' X , ... , X ][ 1 ]

1 12 nn Det II x .. "
1J
so if we put

G= {flf€GL(V) and X .. (f) = 0 whenever i*j},

1J

then G is a closed subgroup of GL(V) containing H. However the mapping

180
is a K-isomorphism of G on to

G xG x ... xG (n factors)
m m m

and therefore we see that H is a closed subgroup of a torus. The proof

is now complete.

5.9 Characters and semi-invariants

Let G be an affine group defined over K.

Definition. A 'rational character' of G is a K-homomorphism

X : G -+ GL I (K) of affine groups.

Thus a rational character of G is a function X E K[G] which gives

rise to a group-homomorphism of G into the group formed by the non-
zero elements of the ground field
In order to illustrate one way in which rational characters can
occur naturally we introduce a further definition. To this end let M
be a rationalleft G-module and suppose that m EM.

Definition. We say that m is an 'eigenvector' of G if (i) m'" 0,

and (ii) om E Km for all a E G.

Suppose that m is an eigenvector of G and define X: G - K by

am = x(a)m. Then, because M is a rational G-module, X E K[G] and,
since a-I (om) = m, we have x(a) '" O. In fact X is a rational character
of G. Thus a rational G-module gives rise to rational characters via its
eigenvectors.

Lemma 8. Let the rational G-module M have finite dimension.

Then the number of rational characters, of G, to which M gives rise is
finite.

Proof. Choose linearly independent eigenvectors m1 , m 2 , ••• , ms

with s as large as possible and let x. be the rational character associa-
1
ted with m.. Now let m be any other eigenvector, in M, and X the
1

181
corresponding rational character. In this situation we have a relation
m = a m + a m +... + am, where a. E K and we may suppose that
1122 SS 1
a 1 *- O. Let a E G and apply a to the relation. This yields
a l x(a) = a l Xl (a) whence X = Xl. This proves the lemma.
Let G be an affine group over K, H a closed subgroup of G, and
M a rational left G-module. Then M is also a rational H-module.

Lemma 9. Suppose that G is connected and that H is a closed,

normal subgroup of G. Assume that m, in M, is an eigenvector of H
and let X be the corresponding rational character of H. Then
-1
x(aAa ) = X(A) for all a E G and A E H.

Proof. Without loss of generality we may assume that the dimension

of M, when considered as a vector space over K, is finite. Next for
-1
a E G and A E H we have aAa (am) = x(A)(am). Consequently am is
an eigenvector of H and if we denote the corresponding rational
character of H by Xu, then Xu(aAa-1 ) = X( A). In particular Xe = X.
By Lemma 8, M gives rise to only finitely many different rational
characters of H. Let these be X = X , X , ... , X. Then, for fixed
1 2 r
A E H and fixed j (1:s j :s r),

{ala E G, x(a-1Aa) = X.(A)}

J
is a closed subset of G, and therefore

{ala EG, X = X.} =V.

a J J

say is also closed in G. But

G = VI U V 2 U • •• U V r'

the union is disjoint and G is .::onnected. Thus one of the sets V. is G

1
itself and the others are empty. It follows that Xa = X for all a E G
and now the lemma follows.
Let G be an affine group over K. By Theorem 29, K[G] has a
natural structure as a left G-module. Let H be a closed subgroup of G.

182
 Definition. An element of K[G] which is an eigenvector of H is
called a 'semi-invariant' of H.

Suppose that f E K[G], f "* 0 is a semi-invariant of H and let

X : H - K be the corresponding rational character. We say, in these
circumstances, that the semi-invariant f has weight X. Note that
l = X(A)f for all A E H and that X(A) is the restriction to H of a
member of K[G].

Theorem 38. Let G be an affine group defined over K and let H

be a closed subgroup of G. Then we can find semi-invariants
p 1 , p 2 , .•• , Pr of
-
H such that
(a) PI' P2' ••. , Pr all have the same weight;
(J
(b) H consists precisely of those CT E G such that p ~ Kp
J.J. J.J.
for J.J. = 1, 2, ••• , r.
In particular the closed subgroups of G are determined by their semi-
invariants.

Proof. We may suppose that H"* G. Let gI' g2' ••• , gt

generate the ideal IG(H) and let M be the K-subspace of K[G] spanned
by all g~ ((J E G, 1 ~ i ~ t). By Theorem 25, the dimension of M is
1
finite; moreover M is a rational left G-module. On the other hand, if
f E IG(H) and A E H, then also l E IG(H). Consequently if we put

then N is not only a finite-dimensional K subspace of M but also a

rational H-module.
Suppose that (J E G and N(J ~ N. Since gI' g2' ... , gt are in N,
the ideal NK[G] , generated by N, is none other than IG(H). This shows
(1
that if f E IG(H), then also f E IG(H) and therefore
(1
f(1) = f (e) = O.

It follows that

H= I
{T T E G and NT ~ N }•

183
 Let the dimension of N (as a vector space over K) be d and
consider the exterior power AdM. This has an obvious structure as a
left G-module and as such it is rational; moreover AdM contains the
one-dimensional K-space A ~ and this subspace is a rational H-module.
Choose u € A ~ so that u *- O. If A € H then certainly AU € Ku.
Now suppose that a € G and au € Ku = A dN• We claim that a € H.
For let us take a base of N and extend it to a base of M. Then consider
the effect of a on the elements which make up the base of N. Our claim
follows from the fact that if a d x q (d!S q) K-matrix has exactly one
non-singular d x d submatrix, then the d x q matrix has only d non-
zero columns. Thus we see that

H = {a a I € G and au € Ku },

Construct a base u = eo' el , e2 , ••• , e

s
of AdM over K and
for a € G write
s
ae. =L f..(a)e.,
J i=O IJ 1

where f ij : G .... K. In fact f ij € K[G]. Next for all a, T € G

s
f..(aT) = L f. (a)f .(T).
IJ IL=O III ILJ

Also T € H if and only if Te € Ke , that is T € H if and only if

o 0
fILO(T) =0 for IL = 1, 2, ••• , s.
Suppose now that a € G and A € H. Then

and so we see that f~o =fOO(A)fiO ' Accordingly the non-zero members of
{flO' f 20 , ••• , fsO} are semi-invariants of H and they all have the
same weight namely the restriction of fOO to H.
Finally assume that a € G and f~O € Kf iO for i = 1, 2, ••• , s.
Then, for 1::; i::; s, Lo(e) = 0 because e € H and therefore
a 1
fiO(a) = fiO(e) = O. It follows that a € H. Thus we can complete the
proof by taking PI' P2' •.. , Pr to be the non-zero members of the set
{flO' f 20 , ... , fsO},

184
 We shall now interrupt the main discussion in order to make some
fairly general observations concerning operations with rational modules
over an affine group. These will assist us when we come to apply
Theorem 38.
To this end let G be an affine group defined over K, and let
M, N be finite-dimensional rational left G-modules. Then M ®K N has
a natural structure as a rational left G- module in which

a(m ® n) = am ® an.
Here, of course, m € M, n € N and a € G.
Now put M* = HomK(M, K). For f € M* and a € G define af
to be the mapping m t-+ f(a- 1 m) of Minto K. Then af € M* and it is
easily checked that M* has become a left (G, K)-module. In order to
examine this structure more closely, let e 1 , e 2 , ••• , et be a base of
Mover K and e~, e;, ... , et' the dual base. Suppose that

t
(Te. = L f..(a)e.,
J i=1 1J 1

where fij : G - K. Then the f ij are in K[G] and

t
ae* = L f (a- 1 )e*
v p=1 vp P
and we see, in particular, that M* is a rational left G-module. Note
that if it happens that for a certain a € G we have ae. = ae. (where
-1 J J
a € K) for j = 1, 2, ... , t, then (Te~ = a e~ for v = 1, 2, ... , t.
The considerations of the preceding two paragraphs enable us to
regard M ®K M* as a rational left G-module. Let a E G and suppose
that when it acts on M the effect is the same as multiplying all the
elements of M by a non-zero scalar a. Then the endomorphism of
M ®K M* induced by a is the identity endomorphism. Now the converse
also holds. For suppose that T € G and

T(e. ® e*) = e. ® e*
] v ] v

for all j and v. Then Te. = a.e. and Te* = {3 e*, where a., {3 € K
]]] v vv J v

185
and a.f3
]v
=1 for all j and v. It follows that a
1
= a 2 = ... = at and
therefore, when T operates on M, the effect is the same as that pro-
duced when we multiply by this scalar.

Lemma 10. Let G be an affine group over K and let H be a

closed, normal subgroup of G. Assume that either (i) G is connected
or (ii) G is commutative. Then there exists a finite-dimensional
rational left G-module N such that the corresponding rational repre-
sentation . G - GL(N) has kernel H.

Proof. Choose semi-invariants PI' P2' ... , Pr of H as in

Theorem 38 and let X be their common weight. Denote by M the K-
subspace of K[G] spanned by all p~, where a E G and 1::: i::: r. Then
1
M is a finite-dimensional rational left G-module, and p~ is a semi-
1
invariant of H whose weight is the mapping ;\ .... x(a-l;\a). But
X(a- l ;\a) = X(;\) by Lemma 9 if G is connected and the relation holds
trivially if G is commutative. Consequently every non-zero element of
M is a semi-invariant of H of weight x.
Let a E G. By Theorem 38, a E H if and only if a acts on M like
a scalar multiplier. But this happens if and only if a induces the identity
mapping on M ®K M*. It follows that the rational representation

G - GL(M ®K M*)

has kernel H.

Theorem 39. Let G be a torus over an infinite ground field K

and H a closed subgroup of G. Then there exist rational characters
Xl' X2' ... , Xs of G such that H is the intersection of their kernels.

Proof. By Lemma 10, there exists a rational representation

G - GL(V) of G whose kernel is H. (Here V is a finite-dimensional,
left G-module.) By Theorem 36 Cor., we can find a base e , e , ... , e
12m
of V that is composed of eigenvectors of G. Let X. be the rational
1
character of G that corresponds.o e.. Then an element a, of G, belong,
1
to H if and only if Xi (a) = 1K for i = 1, 2, ••• , m. The theorem
follows.

186
 We insert here a few remarks about character groups. These will
enable us to exploit the theorem just proved.
Let G be an affine group and Xl' X2 rational characters of G.
We define their sum t as characters by

(X
I
+ X2 )(0-) = XI (o-)x 2 (0-).

The effect of this is to turn the set of rational characters into an additive
A

abelian group. This group will be denoted by G. Note that every member
A

of G is a unit of K[G].
Now suppose that K is an infinite field and consider the torus

G = G ® G ® ••• ® G (n factors)
m m m

so that, with the usual notation,

It is easily verified that, in this case, the rational characters are the
v v v
functions X IX 2 ••• X n, where the v. are integers; and so it follows
I 2 n 1
that (with respect to the addition introduced above) the rational characters
form a free abelian group on n generators. Indeed X , X , .•. , X
I 2 n
is one of the bases of this group. Let X , X , ••• , X be an arbitrary
I 2 n
base. We obtain an automorphism of G by means of the mapping

Observe that this is not just an automorphism of abstract groups. Both

the mapping and its inverse are K-morphisms of affine sets.

Theorem 40. Let G be a torus, H aclosed subgroup of G, and

suppose that the ground field K is infinite. Then H is K-isomorphic to
an affine group of the form

r xr x ... xr xG xG X ... XG,

I 2 q m m m

where r.1 denotes a finite subgroup of Gm.

t This addition corresponds to multiplication in K[G].

187
 Proof. In what follows X denotes a typical rational character
of G. Put

Ii = {x / X(A) = lK for all A E H l.

A

This is a subgroup of G and, by Theorem 39,

H = {a/a E G and x(a) = lK for all X E lil.

But G is 'a free abelian group. We can therefore find a base

Xl' X2' ... , Xn for G and positive integers_ t l , t 2, .•. , tq (q:s n)
so that tl Xl' t2 X2' ... , tq Xq is a bastfOr H. Thus an element a,
of G, belongs to H if and only if xi(a) 1 = lK for i = 1, 2, ••• , q.
Consider the K-automorphism of

G = Gm x G
m
x ••• x G
m
(n factors)

in which a- (X (a), X (a), .•. , X (a)). The image of H consists of all

1 2 n t.
elements (y l' Y2' ••• , y n) such that Yi 1 = lK for i = 1, 2, ••• , q.
Hence if
t.
r i = {y lye: K and y 1 = lK },

then H and the group

r 1 xr 2 x ... xr q xG
m
xG
m
X ... XG
m
(n factors)

are K-isomorphic.

Corollary 1. Let H be a closed subgroup of a torus and suppose that

the ground field K is algebraically closed. Then the elements of H that
are of finite order are everywhere dense in H.

Proof. The theorem allows us to assume that

H=r 1 x ... xrq xG

m
X ... XG,
m

where r. is a finite subgroup of G . Consider the elements of finite

1 m
order in G • There are infinitely many of them and so they are every-
m

188
where dense in G . The corollary now follows by applying Theorem 28
m
Cor. of Chapter 2.
The next corollary holds for an arbitrary ground field.

Corollary 2. Let H be a closed connected subgroup of a torus.

Then H itself is (K-isomorphic to) a torus.

Proof. If K is finite, then H is a finite group and therefore,

because it is connected, it has no elements other than its identity element.
We may therefore suppose that K is infinite in which case our theorem
allows us to assume that

H=r x ..• xr xG x ••. XG.

1 q m m

where r. is a finite subgroup of G • However in this case r. is

1 m 1
connected because it is a continuous image of the connected group H.
Accordingly the groups r. are all trivial and therefore H is a direct
1
product G x G x. •. x G , i. e. it is a torus.
m m m

5.10 Linearly reductive groups and invariant theory

The notion of a linearly reductive group is especially useful in

invariant theory, and in this section we shall endeavour to explain why
this is so. In what follows G always denotes an affine group defined
over the ground field K,. and we recall our earlier agreement that all
G-modules which occur are understood to be left G-modules unless there
is an explicit statement to the contrary.
Suppose that M is a rational G-module. Then, by Theorem 28,
every G-submodule of M is also rational. Put

MG = {x Ix E M and ax = x for all C1 E G }• (5. 10. 1)

This is a G-submodule of M. We shall call it the submodule of G-

invariant elements of M.

Theorem 41. Suppose that G is linearly reductive and let M be

a rational G-module. Then the set of G-submodules N, of M, such that

189
NG = 0 has a (unique) member, MG say, which contains all the others.
This satisfies M = MG EB MG and it is, moreover, the unique comple-
mentary G-submodule of MG in M.

Proof. It is clear that the G-submodules N form a non-empty

inductive system with respect to inclusion. Let Nt be a maximal
member of this system.
Let N be any G-submodule of M such that NG = O. We claim
that N k: Nt. (Note that when this is established we shall be able to
define MG') For suppose that N $I Nt. Then, because G is linearly
reductive, N contains a simple G-submodule P such that P $I Nt.
Accordingly P n Nt = 0 and therefore the sum Nt + P is direct. We
also have pG = 0 because P k: N. It follows that

However this contradicts the maximality of Nt and thereby establishes

our claim.
Let Q be a simple G-submodule of M. Then either QG =Q or
QG = 0, that is either Q k: MG or Q k: MG' Thus, in any event,
Q~ MG + MG' But M is a sum of simple G-submodules. Consequently
M k: MG + MG and therefore, since MG n MG = 0, we have
G
M = M EBMG .
Finally suppose that M = MG EB T, where T is some G-sub-
G G
module of M. Then T n M = 0 and therefore T = O. Thus T k: MG
and now we see that T must be equal to MG' This completes the proof.
Suppose for the moment that G is linearly reductive and let M be
a rational G-module. We have just seen that

(5.10.2)

Denote by

PM: M_MG (5.10.3)

G
the associated projection of M on to M . Note that PM is a homo-
morphism of G-modules and that

190
 (5.10.4)

This projection is known as the Reynold's operator of M.

Let X and Y be rational G- modules and w: X ... Y a homomor-
phism of G-modules.

Lemma 11. Let G be linearly reductive. Then w(XG) k yG and

w(XG ) k YG. Hence if P x and P y are the Reynold's operators of X
and Y respectively, then

for all x EX.

Proof. It is obvious that w(XG ) k yG. Now let Q be a simple

G-submodule of XG. Then either w(Q) is G-isomorphic to Q or
G
w(Q) = O. In any event (w(Q» = a and therefore w(Q) k YG' But
XG is the sum of its simple G-submodules and so we see that w(XG) kYG'
The rest of the lemma follows.

Corollary. Let G be linearly reductive, M a rational G-module,

and N a G-submodule of M. Then the restriction of PM to N coincides
with P N.

We now turn our attention from modules to algebras. Let R be a

non-trivial, unitary, associative and commutative K-algebra. Suppose
further that the affine group G acts on R in such a way that (i) R is a
rational G-module, and (ii) a(fg) = (af)(ag) for all f, g in R and a in
G. Then if we fix the element a, of G, the mapping f 1-+ at is an auto-
morphism of the K-algebra R. In view of this we shall describe the whole
situation by saying that G acts rationally on R by means of K-algebra
automorphisms. If we have this situation, then RG is a K-subalgebra of
R. We refer to it as the K-algebra of G-invariant elements.

Lemma 12. Let the situation be as described above and let G be

linearly reductive. Then RG and RG are RG -submodules of R, and
for fER and g ERG we have PR(gf) == gPR(f).

191
 Proof. It is clear that RG is an RG -submodule of R. Suppose
now that g ERG. Then the mapping R - R in which f ~ gf is a G-
homomorphism and therefore gR G ~ RG by Lemma ll. Accordingly
RG is an RG-submodule of R. The final assertion of Lemma 12 is now
trivial.

Corollary 1. Let the linearly reductive group G act rationally on

R by means of K-algebra automorphisms and let 'U be an ideal of RG.
Then 'UR 'n RG = 'U.

Proof. Since R:= RG ffi RG and RG is an RG-module, we have

'UR = 'URG ffi 'URG = 'U ffi 'URG

and 'URG ~ R G• Consequently 'UR n RG = 'U as required.

Corollary 2. Let the linearly reductive group G act rationally

on R by means of K-algebra automorphisms. Further let {ffi.}. J be
- J J€ -
a family of ideals of R, where each ill. is also a G-submodule of R.
J
Then

}; (ill. n RG) = (); ill.) n RG.

j EJ J j EJ J

Proof. We need only prove that the right hand side is contained in
the left hand side since the converse is trivial. Suppose therefore that
f belongs to (2::ill.) n RG. Then
J

f =L f.,
jEJ J
where f. E ill. and almost all the f. are zero. Next
J J )

f = PR(f) == L PR(f.).
jEJ )
However PR(f.) E ill. n RG because, by Lemma II Cor., the restriction
J J
of PH. to illj is the Reynolds operator of illr Accordingly f belongs
to 2::(illj n R ~ and the lemma follows.
The next lemma provides a general result from the theory of graded
rings. It will be needed in the proof of our main result concerning algebras

192
of G-invariant elements.
Suppose that R is a commutative ring with an identity element
(not necessarily a K-algebra) which is graded by the non-negative integers.
Thus

R=R EElR EElR EEl ... ,

012

where R is a subgroup of the additive group of Rand R R ~ R +

Jl Jl II Jl II
for all Jl ~ 0 and II ~ O. The elements of R are said to be hOmo-
Jl
geneous of degree IJ. and any ideal which can be generated by homogeneous
elements is called a homogeneous ideal. For example

R+ = R EEl R EB REEl ...

123

is a homogeneous ideal of R. Should it happen that the ideal R + is

finitely generated, then we can find a finite system of generators that is
composed of homogeneous elements of positive degree. Note that the
identity element of R is necessarily homogeneous of degree zero, that
R is a subring of R, and that each R is an R -module.
a IJ. a
Lemma 13. Suppose that R is a commutative graded ring with an
identity element. If R+ is a finitely generated ideal of R, then it is
possible to find a finite set {u, u , ..• , u } of elements of R such
1 2 P --
that R = RO[u1 , u 2 ' ••• , up]'

Proof. Choose f 1 , f 2 , ••• , fs' where fi is homogeneous of degree

d.
1
> 0, so that

fER n- df 1 +Rn- df 2 + ... +R n- df.

s
1 2 S

Consider the subring S, of R, that is generated by R 0' R 1 , ••• , ~

and the elements f , f , ... , f. It is easily verified, using induction,

1 2 s
that RIJ. ~ S for every IJ. ~ O. Consequently S = R. The theorem will
therefore follow if we show that each R is a finitely generated R -module.
IJ. a

193
 Assume that this is not so and choose p 2: 0 as small as possible
so that R is not a finitely generated R -module. Then p 2: 1. Also
P 0

is a homogeneous ideal of R. If we factor out this ideal, then we obtain

a new graded ring

R' = R 0 EB • •• EB Rp EB 0 EB 0 EB •••

and, because R+ is mapped on to Wf-' R+ must be a finitely generated

ideal of R'.
Now put

'U' = R 1 Rp- 1 + R 2Rp- 2 + ... + Rp- l R 1 .

This is a homogeneous ideal of R' and, by the choice of p, it is finitely

generated as an R 0 - module. Let us factor out 'U'. This leads to a new
graded ring R" whose grading is given by

R" = REB ... EB R

o p-
1 EB (R
p
I'll ') EB 0 EB 0 EB ...

because 'U' ~ Rp' Also this construction ensures that R +is a finitely-
generated R"-ideal.
Let us take a finite homogeneous system of generators for the ideal
R" and pick out those that have degree p. These will generate
+
R" = R I'll' as an R -module because R"Rb" = 0 if a> 0, b> 0 and
p p Q a
a + b = p. Hence Rp/'U' is a finitely generated R 0 - module and therefore
R is a finitely generated R -module. This contradiction completes the
p 0
proof.

Corollary. Let R be a commutative ring with an identity element

which is graded by the non-negative integers. If now Ro is a Noetherian
ring, then the following two statements are equivalent:
(a) R is a Noetherian ring;
(b) R = Ro[u1 , u 2 ' ••• , un] for suitable elements u1 ' u2 ' ••• , un
in R.

194
 Proof. Assume (a). Then R+ is a finitely generated ideal of R
and therefore (b) follows by the lemma. The converse holds by virtue
of Hilbert's Basis Theorem.
We come now to the main result of this section.

Theorem 42. Let G be a linearly reductive affine group and let

R be a unitary associative and commutative K-algebra. Further let G
act rationally on R by means of K-algebra automorphisms. If now R
is a finitely generated K-algebra, then RG is also a finitely generated
K-algebra.

Proof. By Lemma 12 Cor., CUR n RG = 'U for every ideal 'U of

RG. 8ince R is Noetherian, this observation shows that RG is
Noetherian as well.
Choose VI' v 2 ' ••• , Vs in R so that R = K[v l , v 2 ' ••• , vs ] and
then choose a finite-dimensional G-submodule M, of R, so that v. EM
1
for all i. Next select a K- base uI ' u2 ' ••• , un for M. We now have
R = K[ u I , u 2 ' ••• , un] and Ku I + Ku 2 +... + Kun is a rational G-
module.
Let XI , X 2 , ... , Xn be indeterminates. There is an isomorphism

KXI + KX 2 +... + KXn '" KuI + Ku2 +... + Kun

of K-spaces in which X. is matched with u.. We use this to turn

1 1
KX + KX + ... + KX into a rational G-module in such a way that our
I 2 n
isomorphism of K-spaces becomes an isomorphism of G-modules.
Put 8 = K[X , X , ... , X]. Then 8 is a graded ring with grading
I 2 n

where 8 0 = K and 811 consists of all forms of degree II. Each a E G

determines a K-algebra automorphism of 8 in which X. is mapped into
1
aX.. Thus 8 is a G-module. Indeed each 8 is a finite-dimensional
1 II
rational G-module. Thus, to sum up, G acts rationally on S by means
of K-algebra automorphisms.
The first paragraph of the proof now shows that SG is a Noetherian
ring. But SG is graded with grading

195
 SG = SG EB SG EB SG EB ...
o 1 2

and SG = K. It therefore follows, from Lemma 13 Cor., that SG is a

o
finitely generated K-algebra.
Finally there is a surjective homomorphism S - R of K-algebras
in which X. t-+ u.. This is also a homomorphism of G-modules. Con-
I I G G
sequently, by Lemma 11, S gets mapped into R and indeed the
induced mapping SG - RG is surjective. But SG is a finitely generated
K-algebra 'and therefore the same must hold for RG. The proof is now
complete.

5. 11 Quotients with respect to linearly reductive groups

Let V be a non-empty affine set and G an affine group, both

defined over K, and suppose that G acts morphically on the right of V.
Now suppose that we have an affine set V and a K-morphism
o
11 •• V-V 0

such that 11(va) = 11(V) for all v E V and a E G. We say that (V 0' 11) is
a quotient of V with respect to G or a quotient of V for the action of G
if the following condition is satisfied. Given any K-morphism 11' : V- V~

of V into an affine set V~ such that 11'(va) = 11'(V) for all v and a,
there exists a unique K- morphism c/J: V - V' such that c/J 0 11 = 11'.
o 0
Suppose that (V 0' 11) and (V~, 11') are both of them quotients of
V with respect to G and let c/J : V - V' and I/> : V' - V be the
1 0 0 2 0 0
K-morphisms such that c/J
1
0 11 = 11' and I/>
2
0 11' = 11. Then c/J
2
0 c/J
1
and I/> 0 I/> are the identity mappings of V and V' respectively.
1 2 0 0
Thus 1/>1 is a K-isomorphism and c/J 2 is its inverse. Consequently if
V has a quotient with respect to G, then the quotient is essentially unique.
The question of the existence of quotients is more difficult. Let
a E G. There is a K-automorphism of V in which v f-+ va and corres-
ponding to this we have an automorphism of the K-algebra K[V) in which
a a
f .... f , where f (v) = f(va). Put af = f.a Then, because G acts on
the right of V, this turns K[V) into a left (G, K)-module. Suppose that
f E K[V). Then, by Theorem 25 adapted to the case of right actions, the

196
K-space spanned by the elements {of} aEG has finite dimension. This
K-space is also a rational G-module containing f as may be seen by
adapting the proof of Lemma 4. Thus, to sum up, G acts rationally on
K[V] by means of K-algebra automorphism.
The K-subalgebra K[V]G, of K[V], consisting of the G-invariant
coordinate functions on V IS rationally reduced. This is because K[V]
is rationally reduced and every rational maximal ideal of K[V] contracts
to a rational maximal Ideal of K[V]G (see Chapter 2 Theorem 1). Thus
we obtain

Lemma 14. Let the situation be as described above. Then K[V]G

is an affine K-algebra if and only if it is finitely generated as a K-algebra.

Suppose that v E V and let M denote the rational maximal ideal

v
of K[V] corresponding to v. Then f EM if and only if l" EM.
G va v
Consequently when f E K[V] we have f E Mva if and only if f E Mv'
Accordingly

M n K[V]G :;: M n K[V]G (5.11.1)

v va

for all v E V and a E G.

Lemma 15. Suppose that K[V]G is a finitely generated K-algebra.

Then there exists a quotient of V with respect to G. Furthermore if
1T : V - V 0 is such a quotient, then, in the corresponding homomorphism
1T* : K[V 0] - K[V] of K-algebras, K[V 0] is mapped isomorphically on to

K[V]G.

Proof. By Lemma 14, K[V]G is an affine algebra. Consequently

we can find an affine set V such that K[V ] is isomorphic, as a K-
GOO
algebra, to K[V] . We can therefore construct a homomorphism
1T* : K[V 0] - K[V], of K-algebras, which maps K[V 0] isomorphically on
to K[V]G and this in turn will induce a morphism 1T: V - V 0 of affine
sets. Moreover, if v E V and a E G, then 1T(V):;: 1T(va) by virtue of
(5.11.1).
Now let A: V - U be a K-morphism of affine sets which is such
that A(va):;: A(V) for all v E V and a E G. This determines a homo-

197
morphism A* : K[U] - K[V] of K-algebras. Let h E K[U] and put
f = A*(h). Then
a
f (v) = f(va) = h(A(va» = h(A(v» = f(v)

and therefore fa = f for all a E G. This shows that A*(h) E K[V]G and
thus we see that there is a unique K~algehra homomorphism
w* : K[U] - K[V 0] such that 71* e w* = A*. But this is equivalent to
saying that there is a unique K- morphism w: V - U such that
o
we 71=A.
This establishes that 71: V - V is a quotient of V with respect
o
to G. Since quotients are essentially unique, the lemma follows.
The next corollary is more or less a restatement of the lemma.

Corollary 1. Let the affine group G act morphically on the right

of the affine set V and let 71: V - V 0 be an almost surjective K-
morphism of affine sets. If now K[V 0]' when considered as a subalgebra
of K[V], coincides with K[V]G, then (V 0' 71) is a quotient of V with
respect to G.

Corollary 2. Suppose that the affine group G is linearly reductive.

Then V .possesses a quotient with respect to G and, if 71: V - V 0 is
such a quotient, the associated K-algebra homomorphism 71*: K[V oJ-K[V]
maps K[V
-- ·0
] isomorphically on to K[V]G. If, in addition, K is algebraic-
ally closed, then 71 is surjective.

Proof. By Theorem 42, K[V]G is a finitely generated K-algebra.

Consequently we need only establish the final assertion.
Suppose then that K is algebraically closed, and let N be a rational
maximal ideal of K[V]G. We have to show that N is the contraction of a
rational maximal ideal of K[V]. But, because K is algebraically closed,
every maximal ideal of K[V] is rational and therefore we need only show
that NK[V] * K[V]. However this is clear because, by Lemma 12 Cor. 1,
NK[V] n K[V]G = N.
Until we come to the next lemma, we shall assume that our group G
is linearly reductive and that the ground field K is algebraically closed.
As before 71: V - V denotes a quotient of V with respect to G.
o

198
 First suppose that '11 is an ideal of K[V] and that W is its locus.
Then 1T* -1 ('U) is an ideal . of K[V].
0
The locus of this will be X say.
If now g E K[V 0] and g vanishes at all points of the closure 1T(W), of
1T(W), then 1T*{g) vanishes on Wand therefore 1T*{gm) E'U for some
positive integer m. The converse also holds. Thus g vanishes every-
where on 1T(W) if and only if some power of g belongs to 1T*-1{'U).
Since K is algebraically closed, this means that g vanishes everywhere
on 1T(W) if and only if it vanishes everywhere on X. Consequently
X = 1T(W) that is to say 1T(W) is the locus, on V , of the ideal
o

A subset W of V will be said to be G-invariant if Wa = W for all

a E G or, equivalently, if W is a union of orbits. Suppose that W is a
closed G-invariant subset of V and put '11 = ly(W). If now f E '11,
a E G and w E W, then

a
f (w) = f{wa) = 0
a
and therefore f . E'U. This shows that the ideal '11 is a G-submodule
of K[V].

Lemma 16. Let G be linearly reductive, K algebraically closed,

and 1T: V - V 0 a quotient of V with respect to G. If now {W j } j EJ
is a family of closed G- invariant subsets of V, then

1T{ n W.) = n 1T(W.) •

j EJ J jEJ J
Also if X is a closed G-invariant subset of V, then 1T{X) is closed in V .
o

Proof. Put 'Uj = ly(Wj) and

W = n W .•
jEJ J
Then W. is the locus of '11., ~ is the IOC11S of 1T* -1 ('U. n K[V]G), W
J J] -1] G
is the locus of ~'U.,
]
and 1T(W) is the locus of 1T* ]
n K[V] ). ({~'U.)
.

However, '11. is a G-submodule of K[V] and therefore, by Lemma 12

J
Cor. 2,

199
 (L 'U.) n K[V]G =L ('U. n K[V]G).
j EJ ] j EJ ]

We now see that 1i{W) is the locus of

L 71*-1 ('U. n K[v]G)

jEJ ]
and this shows that

1T1WJ = n 1ifWJ .
jEJ ]
Now let X be a closed G-invariant subset of V, and let
v € 71(X). Put X' = 71- 1 ( {v }). Then X' is also a closed G- invariant
o 0
subset of V and therefore

71(X n Xi) = 71(X) n {v }

o
= {v 0 }
by what has just been proved. It follows that X n X' is not empty and
therefore v 0 E 71(X). Consequently 71(X)' is closed in Vo'
We now introduce a concept which is stronger than that of a quotient.
Suppose that the affine group G acts morphically on the right of the
affine set V, and let 71: V - V 0 be a K-morphism of affine sets.

Definition. We say that (V 0' 71) is a 'strict quotient' of V with

respect to G provided that the following three conditions are satisfied:
(1) 71 is surjective and for each y E V , 71- 1 ( {y }) is an orbit
o
of G in V;
(2) as a subalgebra of K[V], K[V 0] coincides with K[V]G;
(3) 71 : V - V0 is an open mapping, that is to say open subsets
of V are mapped on to open subsets of Vo'

Note that, by Lemma 15 Cor. 1, a strict quotient of V with respect

to G is a quotient in the original sense.
Our principal result on quotients with respect to linearly reductive
groups is due to D. Mumford and may be stated as follows.

Theorem 43. Suppose that the ground field K is algebraically

closed, that G is a linearly reducUve affine group, and that G acts
morphically on the right of V. Suppose further that all the orbits of G

200
in V are closed. Then V possesses a strict quotient with respect to G.

Proof. By Lemma 15 Cor. 2, there is a quotient 7T: V - V , of

o
V with respect to G, where 7T is surjective and K[V 0] = K[V]G.
Let y E V o. Then 7T-1 ( {y }) is non- empty and it is a union of
orbits. Let W, W' be orbits in this union. Then they are closed (by
hypothesis) and they are G-invariant.
We claim that W = W'. For assume the contrary. Then W n W'
is empty and therefore

Iy(W) + Iy(WI) = K[V]

because K is algebraically closed. Accordingly, by Lemma 12 Cor. 2,

and therefore f + g =1 for a suitable f in Iy(W) n K[V]G and g in

Iy(WI) n K[V]G. Note that f vanishes everywhere on Wand takes the
value 1 at all points of W'. Furthermore f = 7T*(h) for some h in
K[V 0]. But now· if x E Wand x' E W', then

f(x) = hey) = f(x')

an~ thus we have a contradiction. This establishes our claim and shows
that, for each y E V 0' 7T-1 ( {y }) is an orbit of G in V. It remains only
for us to show that 7T: V - V 0 is an open mapping.
To this end let T be an open subset of V and put S = 7T(T). In
view of what has just been proved we have

7T -1 (S) = U TO"
O"EG
and this is open in V. It follows that 7T- 1 (V \S) is a closed G-invariant
o -1
subset of V and therefore, by Lemma 16, 7T(7T (V \S)) = V \S is closed
o 0
in Vo. But this means that S is open in V 0 and with this the proof is
complete.

201
5. 12 Quotients with respect to finite groups

The discussion of quotients set out in section (5. 11) deals primarily
with linearly reductive groups, and therefore it does not cover the case
of finite groups. These are easier to handle and will be given an ad hoc
treatment in this section.
Let R be a non-trivial, unitary, associative and commutative
K-algebra. (For the moment K is an arbitrary field.) Let G be a
finite group which acts on R by means of K-algebra automorphisms so
we have, in effect, a homomorphism of G into the group of K-algebra
automorphisms of R. Those elements of R that are left fixed by the
elements of G form the K-subalgebra RG. If fER, then the effect of
operating on f with the element (1 of G will be denoted by f(1. Note
that (f(1) 7 = f 7(1.
Now suppose that X is an indeterminate. Each (1 in G induces,
in an obvious manner, an automorphism of the polynomial ring R[X].
Also, for fER, the polynomial

II (X _ f(1)
(1EG
belongs to R[X] and is invariant under the automorphisms induced by the
members of G. It follows that its coefficients, say a , a , ... , a ,
1 2 n
all belong to RG. Since the polynomial is monic and has f as a root,
it follows that f is integral over K[a , a , ... , a ] and therefore it is
G 1 2 n
integral over the larger algebra R . In particular we see that R is an
integral extension of RG. ~-

Theorem 44. Let G be a finite group and let the situation be as

described above. If now R is a finitely generated K-algebra, then RG
is also a finitely generated K-algebra. Furthermore under these con-
ditions R is not only an integral extension of RG but it is in fact a
finitely generated RG-module. Finally every maximal ideal of RG is the
contraction of a maximal ideal of R.

Proof. Let R = K[f1 , f 2 , ••• , fs] and denote by an' a i2 , ... , a in

the coefficients of the polynomial

202
 II (X - f~).
aEG 1
G
Then a ij E R . Now put

Ro = K[a ll , a 12 , ••• , a sn]'

Evidently R is a Noetherian K-subalgebra of RG and each f. is

integral ove: R. It follows t that R
o
= R 0 [f1 , f , ..• , f ] is
2 S
~ finitely
generated R - module. However this implies that R G is a finitely genera-
o
ted K-algebra. All the assertions of the theorem are now clear except
possibly the one concerning maximal ideals. This is true simply because
R is an integral extension of RG.
We now turn our attention to affine sets. Suppose therefore that V
is a non-empty affine set defined over K and that G is a group that acts
on the right of V. We recall that if G is finite, then it is automatically
an affine group.

Lemma 17. Suppose that G is a finite group. Then in order that

the action of G on V be K-morphic it is necessary and sufficient that,
for each a E G, the mapping V'" V in which v 1-+ va be a K-morphism.

Proof. The condition is clearly necessary. Now assume that it is

satisfied. For each G define h T : G ... K by h T (a) = I) Ta,where I)
T E

denotes Kronecker's function. Then hT E K[G] because G is finite.

Next assume that f E K[V] and define f T : V ... K by f T(V) = f(vT).
Our assumptions ensure that f T E K[V] and therefore

b fT v h E K[V x G].
rEG T

Moreover the value of this particular function at (v, a) is f(va). Con-

sequently the mapping V x G'" V which is the action of G on V is a
K-morphism and the lemma follows.
We shall now assume that V is a non-empty affine set defined over
an algebraically closed field K and that G is a finite group which acts
morphically on the right of V. These assumptions will remain in force
until the proof of Lemma 18 has been completed.

t For an account of the theory of integral extensions see, for example,

[(9) pp. 86- 93].

203
 As we saw in section (5.11), these conditions ensure that G acts
rationally on K[V] by means of K-algebra automorphisms. Next
Theorem 44 shows that K[V]G is a finitely generated K-algebra and
therefore, by Lemma 15, there exists a quotient 1T: V -V 0 of V with
respect to G. The same lemma also shows that the associated K-algebra
homomorphism 1T* : K[V 0] - K[V] maps K[V 0] isomorphically on to
K[V]G. Finally 1T: V - V 0 is a surjection. This is a consequence of the
final assertion of Theorem 44 and our assumption that K is algebraically
closed.

Lemma 18. Let the situation be as described above. Then for every
closed subset U, of V, 1T(U) is a closed subset of V 0' that is 1T is a
closed mapping.

Proof. Let '11 = Iy(U), <B = '11 n K[V]G, and let X be the locus of
1T*-1 (<B) in Vo. Then K[V]G /<B is a subring of K[V]I'U and the latter is
an integral extension of the former. It follows that every maximal ideal
of K[V]G /<B is the contraction of a maximal ideal of K[V]/'U and there-
fore every maximal ideal of K[V]G containing <B is the contraction of a
maximal ideal of K[V] containing '11. On the other hand, every maximal
ideal of K[V] that contains '11 contracts, in K[V]G, to a maximal ideal
containing <B. If all this is translated into geometrical terms it says
simply that 1T(U) = X. In particular 1T(U) is closed in V .
o

Theorem 45. Let V be a non-empty affine set defined over an

algebraically closed field K and let G be a finite group which acts
K-morphically on the right of V. Then V possesses a strict quotient
with respect to G.

Proof. We know that there exists a quotient 1T: V - V , of V

o
with respect to G, where 1T is a surjection and K[V ] - K[V] maps
1T* :
o
K[V ] isomorphically on to K[V]G. Furthermore, by Lemma 18, 1T
o
takes closed subsets of V on to closed subsets of V .
-1 0
Let y EO Vo. Then 1T ({ Y }) is non-empty and it is a union of orbits.
Let Wand W' be orbits contained in 1T -1 ({y }).
We claim that W = W'. For suppose that W '* W'. Then W n W'

204
is empty and each of Wand W' is a finite set and therefore closed in
V. It follows, because K. is algebraically closed, that

Iy(W) + Iy(W') = K[V]

and therefore we can find h E Iy(W) and h' E Iy(W') such that h + h' = 1.
Of course h takes the value 0 at all points of Wand the value 1 at
all points of W'.
Let x EW, x' EW' andput

h = n h a.
a aEG
Then ha E K[V]G and therefore h a = 1T*(f) for some f E K[V].
0
Also
ha(x) = 0 and

h (x') == n h(x'a) =1
a aEG
because x'a EW' for all a E G. However h (x) = f(1T(x)) = f(y) =
a
f(1T(x')) = h (x') and this is the desired contradiction. Thus our claim
a
is established and it follows that 1T- 1 ( {y}) is an orbit.
It only remains for us to show that 1T: V .... Va is an open mapping.
However this is no problem because the argument used to complete the
proof of Theorem 43 works again here.

An example. Let V be a non- empty affine set defined over an

algebraically closed field K and let n (n 2: 1) be an integer. Put

W = V x V x .•. x V (n factors)

and let G be the permutation group of {I, 2, ... , n}, i. e. G is the

symmetric group of degree n.
For w = (v 1 , v 2 ' ••• , vn ) in Wand a in G define wa by

wa = (v a (l)' v a (2)' .•. , va(n))'

This leads to a right action of G on Wand Since, for fixed a E G,

w 1-+ wa is a K-automorphism of the affine set W, Lemma 17 shows that
G acts K-morphically on W. Accordingly, by Theorem 45, W possesses

205
a strict quotient 11: W -W withrespect to G.
o
Let us examine this quotient. The points of W 0 are matched with
the orbits of G in V. Also two points (vI' v , ... , v ) and
2 n
(v', v', ... , v') of W belong to the same orbit if and only if
1 2 n
(v', v', ... , v') is a permutation of (v , v , ... , v). Thus the
1 2 n 1 2 n
orbits of G are the unordered n-tuples of points of V. Accordingly
these unordered n-tuples may be regarded, in a natural way, as the points
of an affine set W . When this is done the associated surjection 11 :W-W
o 0
is a closed morphism, and the coordinate ring of W 0 consists of all
f E K[W] with the property that

whenever v~, v;, ... , v~ is a permutation of vI' V 2' ••• , vn.

5. 13 Quotient groups of affine groups

Let G be an affine group and H a closed, normal subgroup of G.

We shall now enquire whether the quotient group G;H has a natural
structure as an affine group and, if so, what is the coordinate ring of
G;H and what are the properties of the natural homomorphismG - G;H?
Answers to these questions will be found in Theorem 48 and, roughly
speaking, all one might reasonably hope for is realised, at least in the
case where the ground field is algebraically closed. However we shall
only arrive at these conclusions after some comparatively lengthy con-
siderations and we shall need to draw rather more heavily on the theory
of commutative Noetherian rings than we have done so far.
We begin by supplementing our previous results on affine sets.

Lemma 19. Let V be an affine set defined over an algebraically

closed field K and let V , V , ... , V be disjoint closed subsets of V
1 2 s
whose union is V itself. Then K[V] can be regarded as the direct
product

K[V ] x K[V ] x ... x K[V ]

1 2 S

of K-algebras with the projection K[V] - K[V.] corresponding to the

1

206
inclusion mapping of V.1into
--
V.

Proof. PutWi=V1U ... UVi_1UVi+1U ... UV s . Thenno

maximal ideal of K[V) contains both \T(V i) and \T(W i) and therefore
TJ.
I
+ E.I = 1 for suitable functions
L(V.) and E. € L(W .). Note
I-VI I-VI
TJ. €

that E. takes the value 1 everywhere on V. and the value 0 at all

I I
other points of V; that E , E , .•. , E are orthogonal idempotents
1 2 s
of K[V); and that E + E + ... + E = 1. Put A. = K[V)E.. Then
1 2 s I I
A. is a K-algebra and K[V) is the direct product of A , A , ... , A •
I 1 2 S
We must now identify A .. The K-algebra homomorphism
I
K[V) -+ A.
in which f 1-+ fE. is surjective and its kernel is L(V .). Hence
I I -V I
we have A. = K[V.) by identifying fE. with the restriction of f to VI"
I I I
and then the projection homomorphism of K[V) on to A. = K[V.)
I . I
corresponds to the inclusion V. ~ V.
I t
The lemma also has a converse. For suppose that affine sets
V , V , ..• , V are given and let us form the direct product
1 2 S

This is an affine K-algebra. Also its rational maximal ideals correspond

to the rational maximal ideals of the various factors and so they are in a
natural one-one correspondence with the points of the disjoint union, V
say, of VI' V 2' ••• , Vs. Thus we may regard V as an affine set with

K[V) = K[V 1 ] x K[V ] x ... x K[V ].

2 S

Let 'U. be the kernel of the projection homomorphism K[V] -+ K[V.). The
I I
rational maximal ideals of K[V) that contain 'U. correspond to the points
I
of V .. Thus V. = CV('U.) and therefore V. is closed when regarded as
I I I I
a subset of V. Also, because K[V.]
I is rationally reduced, Rad..._('U.
·V I)='U.I
and therefore L(V.) = 'U .. It follows from this that the affine structure
-V I I
of V. is not changed by regarding it as a closed subset of V, and that
I
the projection of K[V) on to K[V.) corresponds to the inclusion mapping
I
of V. into V.
I
Now let V', V', ... , V' be additional affine sets defined over K.
1 2 S

t This part of the argument does not require K to be algebraically closed.

207
Their disjoint union, V' say, may be regarded as an affine set with

K[V'] = K[V'] X K[V'] x ..• X K[V'].

1 2 S

Assume that we are givenK-morphisms 1f. : V.

1 1
~ V~,
1
where i = 1, 2, •.. ,s.
The corresponding K-algebra homomorphisms 1f": K[V~] ~ K[V.] induce
1 1 1
a K-algebra homomorphism

1f* : K[V'] ~ K[V]

and this corresponds to a K- morphism 1f: V ~ V'. Observe that, for

each i, 1f is an extension of 1f. : V. ~V~.
1 1 1
Our next result extends Theorem 38 of Chapter 3.

Theorem 46. Let V and W be irreducible affine sets defined

over an algebraically closed field K. If now cp: V ~W is both an
almost surjective K-morphism and an injection, then K(V) is a purely
inseparable, algebraic extension of K(W).

Proof. The hypotheses ensure that K(W) is a subfield of the field

K(V) and we know, from Chapter 3 Theorem 38, that Dim V = Dim W.
Since K[V] is a finitely generated K-algebra, it follows that K(V) is an
algebraic extension of K(W) of finite degree.
Let n be the separable closure of K(W) in K(V). We claim that
n = (K(W))(f) for some f E K[V]. In establishing this claim we shall
suppose that the characteristic of K is the prime p. (If K has
characteristic zero, then the proof is similar but without certain compli-
cations. )
Certainly n = (K(W))(h) for some h E K(V) and
n
n = (K(W) )(hP )

for every positive integer n. Now write h as the quotient of two elements
n
of K[V]. Then, provided n is large enough, hP is the quotient of two
elements both of which belong to K[V] n n. Accordingly n = (K(W))(f1 , f ),
2
where f ,f are in K[V]. But there are only finitely many fields between
1 2
K(W) and n, and the field K is infinite. It follows that

208
 Q = (K(W))(f1 + cf 2 )
for a suitable c E K. Since f + cf is in K[V], this establishes our
1 2
claim.
By multiplying f by a suitable non-zero element of K[W] we can
arrange for it to have an additional property. Thus if

F(X) = Xn + w1 Xn-I + w2 Xn-2 + ... + wn

is the irreducible polynomial for f over K[W], we can secure that the
w. are in K[W].
1
The polynomial F(X) is irreducible and separable over K(W) and
therefore F(X) and its formal derivative F'(X) satisfy a relation

F(X)G(X) + F'(X)H(X) = q,
where G(X), H(X) are in (K[W])[X] and q E K[W], q"* O.
Consider the K-algebras

K[W] h (K[W])[f] h K[V].

By Chapter :3 Lemma 8, there exists g E (K[W])[f], g"* 0 such that every

maximal ideal of (K[W])[f] not containing g is the contraction of a
maximal ideal of K[V]. Also our choice of f ensures that (K[W])[f]
is an integral extension of K[W] and therefore g satisfies a relation

m m-I m-2
g + y g
1 2 m
+Y g + ... + y = 0,
where y. E K[W]. Since g"* 0 we may suppose that y "* 0 and then it
1 m
follows that every maximal ideal of K[W] not containing y is the
m
contraction of a maximal ideal of (K[W])[f] not containing g, and this in
turn will be the contraction of a maximal ideal of K[V].
Choose wW so that q(w) "* 0 and y (w)"* O. This determines a
E
m
maximal ideal M of K[W]. But y 1 M so M is the contraction of a
m
maximal ideal of K[V]. However cf>: V -+W is an injection. Consequently
there is only one maximal ideal of K[V] that contracts to M. It follows
that there is exactly one maximal ideal of (K[W])[f] that contracts to M.

209
 The homomorphism K[W]'" K of K-algebras in which p 1-+ p(w)
induces a homomorphism

(K[W])[X]'" K[X]

of the polynomial rings. For Q(X) € (K[W])[X] let us use Q(X) to denote
its image. Then

F(X)G(X) + F'(X)li(X) = q(w)

and therefore, since q(w)"* 0, F(X) has n distinct roots, where n is

the degree of F(X). Let O! be anyone of these roots. Then the homo-
morphism

(K[W])[X] ... K

in which Q(X) 1-+ Q(O!) induces a homomorphism (K[W])[f]'" K in which

f 1-+ o!. Thus there are n different maximal ideals of (K[W])[f] that
contract to M and therefore n = 1. However this means that n = K(W)
and so the proof is complete.
Before we leave this preliminary section on affine sets, it is
convenient to put on record an elementary result concerning finitely
generated field extensions.

Lemma 20. Let E be an extension field of a field F and let L

be a field between F and E. If now E is a finitely generated field
extension of F, then so too is· L.

Proof. Let ~1' ~2' ••• , ~s be a transcendence base for Lover

F and let this be extended to a transcendence base ~1' ••• , ~s' 7] 1" •. ,7]t
for E over F. Note that 7] , 7] , ... , 7]t will be algebraically inde-
1 2
pendent over L.
Let -\' A2 , ••• , Aq belong to L and be linearly independent over
F(~ , ~ , .•• , ~). These elements will remain linearly independent over
1 2 S
F(~l' ••. , ~s' 1]1' ••• , 7]t)' Hence, if m is the degree (necessarily
finite) of E over F(~l' ••. , ~s' 7]1' ..• , 7]t)' we must have q:::: m.
Thus the degree of Lover F( ~ , ~ , .•• , ~ ) is finite and therefore L
1 2 S

210
is a finitely generated field extension of F.
We are now ready t<;> begin the investigation of quotient groups of
affine groups. To avoid excessive repetition we shall make the following
assumptions which are to remain in force until we come to the statement
of Lemma 25:
(1) the field K is algebraically closed;
(2) G is a connected affine group defined over Kj
(3) H is a closed, normal subgroup of Gj
(4) there is given a. connected affine~ r and a surjective
K-homomorphism 1/>: G'" r such that Ker I/> = H.
Note that if (1), (2) and (3) are satisfied, then we can always find I/> and
r so that (4) holds as well. For, by Lemma 10, there exists a finite-
dimensional rational left G-module such that if p : G'" GL(V) is the
corresponding rational representation, then the kernel of p is H. Put
r = p(G). By Theorem 22, r is a closed subgroup of GL(V) and there-
fore (4) is satisfied if I/> is taken to be the induced K-homomorphism
G'" r. Of course r will be connected because it is a continuous image
of G.
The mapping G x H'" G in which (a, T) ':'+ aT provides a K-morphic
action by H on the right of G. Consequently H acts rationally on K[G]
by means of K-algebra automorphisms. As in similar situations we use
K[G]H to denote the subalgebra formed by the functions that H leaves
invariant. Since I/> is a surjective K-homoniorphism with kernel H, we
have

Moreover K[G] is an integral domain because G is connected.

Lemma 21. The quotient field of K[G]H is a purely inseparable,

algebraic extension of K(r).

Proof. By Lemma 20, we can choose f l , f2 , ••• , fs in K[G]H

so that

(5.13.1)

211
and K[f , f , ... , f ] has the same quotient field as K[G]H. Evidently
1 2 s
K[f , f , ... , f ] is an affine K-algebra. Consequently there exists an
1 2 S
irreducible affine set V with K[V] = K[f , f , ... , f ] and then the
1 2 S
inclusions in (5.13.1) will give rise to almost surjective K-morphisms
e : G -+ V and l/J : V -+ r such that l/J 0 e = cpo By Chapter 3 Theorem 33,
e(G) contains a non-empty open subset U, of V, and now, by making
U smaller if necessary, we may suppose that U consists of the points of
V where some member of K[V] does not vanish. This ensures (see
Chapter 2 Theorem 20) that U has a natural structure as an affine set
with K[V] ~ K[U] and K(V) = K(U).
Next, again by Chapter 3 Theorem 33, l/J(U) contains a non-empty
open subset of r. Accordingly the mapping U -+ r induced by l/J is an
almost surjective K-morphism.
Suppose that f E: K[V] ~ K[G]H. Then f(crT) = f(cr) for all cr E: G
and T E: H, and from this it follows that e : G -+ V is constant on the
cosets of H in G. Suppose that u ,u E: U, say u. = e(cr.), and that
1 2 1 1
l/J(u ) = l/J(u). Then cp(u1 ) = cp(u). Thus cr and cr belong to the
1 2 2 1 2
same coset of H and therefore u = u. Thus our K-morphism U -+ r
1 2
is an injection as well as being almost surjective. Consequently, by
Theorem 46, K(U) = K(V) is a purely inseparable, algebraic extension
of K(r). The lemma follows.

Lemma 22. With the above assumptions

K(r) n K[G]H = K[r].

H
Proof. Suppose that f E: K(r) n K[G] and let y E: r. By (3.3.5)
we can regard the local ring Qr. ,of r at y, as lying between K[r]
,y
and K(r). On this understanding we claim that f belongs to Q r . For
,y
assume that f t Q r . By Theorem 13, y is a simple point of r.
,y ---
Certainly Qr is a Noetherian ring with a single maximal ideal and
,y
now we see, by Chapter 4 Theorem 23, that it is what is called a regular
local ring. Now it is known t that every regular local ring is a unique

t See [(13) Vol. 2, Appendix 7, p. 404] or [(12) Theorem 5, p. 22].

212
factorization domain. This fact is used in the argument that follows.
Write f = f' If", where f', f" are in Qr and have no common
,y
irreducible factor, and then, by multiplying numerator and denominator
by a suitable unit of Qr ,arrange that f', f" E K[r]. Since f I Qr '
,y ,y
f" has at least one irreducible factor. If w is such a factor, then
wQ r = PQr ,where P is a prime ideal of K[r], f" E P whereas
,y ,y
f' I P. This shows that f' is not in the radical of the ideal f"K[r] and
therefore there exists z E r such that f"(z) = 0 and f'(z) i- O. Choose
a E G so that cp(a) = z.
Now ff" = f' and we can regard all of f, f', f" as belonging to
K[G] in which case f(a)f"(a) = f·(a). However the value of f' at a is
the same as its value at z because of the way we embed K[r] in K[G].
Consequently f'(a) i- O. Similar considerations show that f"(a) = O.
The relation f(a)f"(a) = f'(a) now yields a contradiction and with it our
claim that f E Qr, y is established. Accordingly we can find () y E K[r]
such that () f E K[r] and () (y) i- O.
Y Y
Let us choose such a ()
y
for each y in r. The ideal they generate
in K[r] is K[r] itself because there is no single point where they all
vanish. Since () f E K[r] for every y it follows that f E K[r].
y H
It has now been proved that K[r] d K(r) n K[G] and, as the reverse
inclusion is trivial, this completes the proof.

Lemma 23. The algebra K[G]H is an integral extension of K[r]

and each prime ideal of K[r] is the contraction of exactly one prime ideal
of K[G]H. Furthermore K[G]H is an affine K-algebra.

Proof. For the moment we shall leave the final assertion on one
side.
First suppose that K has characteristic zero. In this case Lemma
21 shows thatk:(r) contains K[G]H and therefore K[r] = K[G]H. Thus
our assertions are trivial. Now assume that the characteristic of K is the
prime p and let f E K[G]H. By Lemma 21, fpn is in K(r) n K[G]H=K[r]
provided that n is large enough. Hence in this case too the first two
assertions are clear. Note that we have established more about the rela-
tion between K[G]H and K[r] than is actually stated in the lemma.

213
Indeed the extra information will be needed later.
Let us now turn our attention to the final assertion. By Lemmas
20 and 21, the quotient field of K[G]H is an algebraic extension of K(r)
of finite degree. Consequently, because K[r] is a finitely generated
K-algebra, the integral closure of K[r] in this quotient field is a finitely
generated K[r]-module t which has K[G]H as a K[r]-submodule. It
follows, because K[r] is Noetherian, that K[G]H is also a finitely
generated K[r]-module and therefore it must be a finitely generated K-
algebra. Finally K[G]H is rationally reduced because it is a K-subalgebr
of K[G]. With this the proof is complete.
Weare now ready to take a major step forward. Let us regard G
as an affine set with H acting morphically on its right.

Lemma 24. There exists a quotient 11: G - S for the right action
of H on G. Such a quotient is surjective and its fibres are precisely
the cosets of H in G. Furthermore if we regard K[S] as a subalgebra
of K[G], then K[S] = K[G]H.

Proof. By Lemma 15, there exists a quotient 1T: G - S of G

with respect to H; moreover such a quotient will be almost surjective
and it will lead to K[S] being identified with K[G]H. Next, since

K[r] ~ K[G]H = K[S],

there exists a K-morphism A: S - r which satisfies A 0 11 = cpo However

Lemma 23 shows that A is a bijection. The remaining assertions con-
cerning 11 are now consequences of the corresponding properties of cpo
Let 11: G - S be as in Lemma 24. Then there is a unique way of
putting a group structure on S so that 11 becomes a surjective homo-
morphism of abstract groups whose kernel is H. What is not immediately
clear is that this will turn S into an affine group. The next lemma shows
that this is indeed the case. It should be noted that, in the lemma, we
drop the requirement that G be connected because the more general
version will be needed later. (We also do not require K to be algebraicall

t See, for example, [(13) Vol. 1, Theorem 9, p. 267].

214
closed but this is less significant.) As usual our affine groups and sets
are assumed to be defined over K.

Lemma 25. Let G be a (not necessarily connected) affine group,

H a closed normal subgroup of G, and 11: G - S a surjective K-mor-
phism of affine sets whose fibres are precisely the cosets of H in G.
If now K[S], when considered as a subalgebra of K[G], is just K[G]H,
then there is a unique way to endow S with a group structure so that
S becomes an affine group and 11 a group homomorphism.

Remarks. Of course if S is given the structure referred to,

then 11 becomes a surjective K-homomorphism of affine groups whose
kernel is H. For G connected, S represents an improvement on r
to the extent that it has the natural coordinate ring.

Proof. We turn S into a group in such a way that 11 is a homo-

morphism of groups and verify that, for S, multiplication and inversion
are K-morphisms.
Let W E K[S]. Since W 0 11 belongs to K[G], the mapping
h : G x G - K given by h(a , a ) = w(11(a a )) belongs to K[G x G].
1 2 1 2
Consequently h can be represented in the form
m
h = L f. v g., (5. 13. 2)
i=l 1 1

where L, g. E K[G]. We claim that we can arrange that all the f. are
1 1 1 ---
in K[ G]H. To see this we first make gl' g2' .•. , gm linearly inde-
pendent over K. Then, for u ,a E G and T E H,
1 2

h(u1 T, u) = W(11(a1 7O")) = W(11(a 1a2)) = h(a1 , a)

whence
m m
L L(a T)g.(a ) = L L(a )g.(a )
i=l 1 1 1 2 i=l 1 1 1 2

and therefore
m m
L L(a T)g. = L L(a )g .•
i=l 1 1 1 i=l 1 1 1

It follows that L(a T) = L(a ) and hence that f. E K[G]H. This estab-
1 1 1 1 1

215
lishes our claim.
Let us make a fresh start. Suppose that (5. 13. 2) holds but now
with all the f. in K[G]H. Without destroying this property we can also
1
suppose that f , f , ... , f are linearly independent over K, and then
12m
an argument similar to that just used shows that the g. are in K[G]H
1
as well. Thus for each i we can find u., v. E K[S] so that u. 0 n = f.
1 1 1 1
and v. 0 n = g.. By construction the value of
1 1
m
L u. v v.
i=l 1 1

at (n(a ), n(a )) is w(n(a a ))

1 2 12
= w(n(a1 )n(a2 )) and so the mapping
S X S - K in which

(n(a ), n(a )) 1-+ w(n (a )n(a ))

1 2 1 2

belongs to K[S x S]. However, because w is an arbitrary element of

K[S], this is equivalent to saying that the multiplication mapping
S x S - S is a K-morphism.
Next the mapping p : G - K given by p(a) = w(n(a- 1 )) belongs to
K[G] and, for T E H, p(aT) = p(a) so that p E K[G]H. Hence p = eon
for some e E K[S]. However e: S - K is the inversion mapping S - S
followed by w: S -K. Consequently the inversion mapping on S is a
K-morphism and now the proof is complete.
We have now progressed a considerable way towards a satisfactory
definition of a quotient group of an affine group. The next theorem pro-
vides the main ingredient that is missing.

Theorem 47. Let G and r be connected affine groups defined over

an algebraically closed field K and let n: G - r be a surjective K-
homomorphism. Then n is an open mapping.

Proof. This will be divided into two stages. Put H = Ker n so that
H is a closed, normal subgroup of G.
First stage. Here it will be assumed that H itself is connected.
We set R = K[G] and 1= K[r]. These are integral domains, I is a K-
subalgebra of R, and there exist elements u1 ' u 2 ' ••• , urn in R such
that R = I[ u1 ' U , ••• ,
2
u ], that is to say R is a finitely generated
m

216
I-algebra.
By Theorem 13, all the points of G are Simple and therefore, as
we remarked earlier, the local rings ofthe points of G are unique
factorization domains. Moreover the intersection of these unique
factorization domains is K[G] = R. Consequently R is integrally closed
and, for similar reasons, I is integrally closed as well.
There exist t a E I, a*-O and z , z , ... *, z in R so that
1 2 q -1
z , Z , .•. , Z are algebraically independent over I and R[a ] is an
1 2 q -1 -1
integral extension of I[a ,z, .•. , Z ]. We note that R[ a ],
-1 1 q
I[ZI' Z2' .•. , Zq] and I[a ,zI' •.. , Zq] are all of them integrally
closed. l
Choose a maximal ideal M, of R, so that a I. M and let x be
the corresponding point of G. Then y = l1(X) corresponds to the maxi-
mal ideal N = M n I of I = K[r].
Let U be an open neighbourhood of x in G. We shall prove that
l1(U) is a neighbourhood of y. Since 11 is a surjective homomorphism,
the fact that 11 is open will then follow by homogeneity. Before pro-
ceeding note there exists W E R such that w(x) *- 0 and

{ ~ I~ E G and wW"* o} k u.
It follows that we may assume that U itself has the form
{ ~ I~ EG and w(~) *- O} for a suitably chosen w.
We have 11- 1 ( {y }) = Hx and this is irreducible because we are
assuming that H is connected. It follows that NR has exactly one
minimal prime ideal and, because w(x) *- 0, w does not belong to this
minimal prime ideal. But a I. M and therefore a I. N. We now see that
NR[ a -1] has precisely one minimal prime ideal and w does not belong to it.
Since Z , Z , ..• , Z are algebraically independent over I,
1 2 q
NJ[ \ ' Z , •.• , Z ] is a prime ideal of I[ Z , Z , ..• , Z ] which does
2 q -1 1 2 q
not contain a. Consequently NJ[a ,z, ... , Z ] is a prime ideal of
1 q
I[a- I , Zl' ..• , Zq]' Furthermore, because R[a- I ] is an integral ex-

t See [(8) Theorem 14.4, p. 45].

l Note that if R is an integrally closed integral domain, then so is the
polynomial ring R[X]. See [(13) Vol. 2, Theorem 29, p. 85].

217
tension of I[a-I, Z , ... , Z ], every prime ideal of R[ a -1] that con-
I q
tracts to NI[a- 1 , ZI' ... , Zq] in I[a- 1 , zl' ••• , Zq] hastobeamini-
mal prime ideal of NR[a- 1 ]. Thus we reach the following conclusion:
there is only one prime ideal of R[ a -1] that contracts to NI[a -1, Z , .•• , Z
1 q
and it does not contain w.
Since W E R it is integral over the integrally closed ring
I[a -1, Z , •.. , Z ], and therefore the irreducible polynomial for w over
1 q
the quotient field of I[a -1, Zl' .•• , Zq] has the form

h cf\ (zl' •.. , Z ) 1 I/>h (Zl' ... , Zq)

X + t q Xh- +... + t
a a

where t 2': 0 and l/>i(zl' z2' •.. , Zq) E I[zl' z2' .•• , Zq]. Also
9>h(zl' Z2' .~~' Zq)I.NI[zl' z2' •.. , Zq] because W is not in any prime
ideal of R[a ] that contracts to NI[a- 1 , Z , ••. , z]. Let v be a
1 q
coefficient of I/>h(z , Z , ••• , Z ) that is not in N. Then av E K[r]
1 2 q
and av I. N. Accordingly

is open in r and it contains y.

We assert that T <;;;: l1(U). For let N' be any maximal ideal of I
that corresponds to a point of T. Then av I. N'. Choose 0' , 0' , ••• , 0'
1 2 q
in K so that I/>h(O' 1 ,2 q
I. N'. (This is possible because
0' , ••• , 0' )

v I. N'.) The existence of the 0', shows that there is a maximal ideal of
1
I[ Zl' Z2' ..• , Zq] that contracts to N' in I, but contains neither a
nO~l I/>h (Zl' z2' ..• , Zq). Consequently there is a maximal ideal of
I[ a ,z, ..• , Z ] that contracts to N' but does not contain
1 q t
I/>h(Zl' Z2' .•. , zq)/a. Now

h 1/>1 (Zl' •.. , Zq) h-1

W + t W + ...
a

and R[a- 1 ] is an integral extension of I[a-\ Z , •.. , Z ]. Accordingly

1 q
there is a maximal ideal of R[ a-I] that contracts to N' but does not
contain W. Finally we see there is a maximal ideal M', of R, such
that M' n I = N' and w I. M'. Thus if N' corresponds to the point

218
y' E T and M' to x' E G, then x' E U and y' = 1T(X'). This proves that
T ~ 1T(U) and establishes that 1T is open for the special case where H
is connected.
Second stage. Here we remove our assumption concerning H. To
this end let H be the connected component of the identity of H. If
o -1 -1
X E G, then xH x is a closed connected subgroup of xHx = H. Con-
~1
sequently xH x ~ H and therefore Ho is a closed, connected,
o 0
normal subgroup of G. By Lemmas 24 and 25, the group G;Ho can be
given an affine structure in such a way that (i) the natural mapping
e : G -+ G;H o is a K-homomorphism, and (ii) with the obvious identification
H
K[G;Ho] = K[G] o.

Also, in view of what was established in the first stage of the present
proof, e is an open mapping.
Consider the finite group H;Ho' This is a normal subgroup of
G;Ho and its finiteness ensures that it is a closed subgroup. Let us
regard H;H as acting morphically on the right of G;H. By Theorem
o 0
45, we can find a strict quotient, X : G;H -+ S say, for this action and
o
then we have
(H;H ) H (H;H)
K[S] = K[G;Ho] 0 = (K[G] 0) 0 = K[G]H.

Also, by Lemma 25, we can impose a group structure on S so as to turn

it into an affine group and X into a K-homomorphism. Put p = X 0 e.
Then p : G -+ S is an open, surjective K-homomorphism with H = Ker p
and K[S] = K[G]H.
Since K[r] ~ K[S] (see (5.13.1)), there is a K-morphism A :s-+r,
of affine sets, such that A 0 P = 1T. To complete the proof we have only to
show that A is open.
If the characteristic of K is zero, then (as we saw in the proof of
Lemma 23) K[r] = K[G]H = K[S] so A is an isomorphism and there is
no problem. Suppose therefore that the characteristic of K is the prime
H n
p and let f E K[G] . Then (again by the proof of Lemma 23) fP E K[r]
for a suitable positive integer n. In fact K[S] is an integral extension
of K[r] and the maximal ideals of K[S] that do not contain f have as

219
their contractions precisely the maximal ideals of K[r] that do not
n
contain fP. This ensures that A : S -+ r is an open mapping.
We come now to the main result of this section.

Theorem 48. Let G be an affine group defined over an algebraic-

ally closed field K and let H be a closed, normal subgroup of G. Then
there exists an affine group S and an open, surjective K-homomorphism
H
1T : G -+ S such that Ker 1T = H and, as a subalgebra of K[G], K[S]=K[G] .

Proof. Let G be the connected component of the identity of G

o
and let G , G , ..• , G be the different cosets of G in G. These
o 1 r 0
are disjoint, closed, irreducible subsets of G.
Assume, for the moment, that H ~ G. By combining Lemmas 24
o
and 25 with Theorem 47, we see that there exists a strict quotient
1To : Go -+ So for the right action of H on Go'
Choose (J. E G so that G. = (J.G. Then left multiplication by (J.
1 1 1 0 1
induces a K-isomorphism ¢. : G
1 0
=. G..1 Now H acts morphically on the
right of both G and G., and ¢. commutes with the action of the ele-
o 1 1
ments of H. It follows that there exists a strict quotient 1T. : G. -+ S.
1 1 1
for the action of H on G..
1
Let S be the disjoint union of So' Sl' ... , Sr' We know that S
can be regarded as an affine set with

K[S] = K[G ]H x K[G1 ]H x ... x K[G ]H.

o r
Also each T E H induces a K-algebra automorphism of K[G.]
1
(i = 0, 1, ..• , r) and thereby a K-algebra automorphism of their direct
product. On this understanding

By Lemma 19,

K[G] = K[G ] x K[G ] x ••. x K[G ]

o 1 r
whence K[S] = K[G]H. Moreover the inclusion mapping of K[G]H in
K[G] gives rise to a K-morphism 1T: G -+ S which extends the various

220
morphisms 1I. : G. -+ S. (see the discussion following the proof of
1 1 1
Lemma 19). It is now clear that 1I: G -+ S is a strict quotient for the
action of H on G, and that if we use Lemma 25 to turn S into an affine
group, then 1I: G -+ S meets the requirements of the theorem.
It remains for us to remove the assumption that H is contained in
G. To this end let H denote the connected component of the identity
of°H. Then Ho is not°only a closed, normal subgroup of H, but also t
a closed normal subgroup of G. Because H ~ G , the first part of
° ° group
this proof now shows that we can regard the abstract G;H as an
affine group; °
moreover this can be done in such a way that the natural
mapping e: G -+ G;H is an open K-homomorphism allowing us to

identify K[G;H ]
° H
with K[G] o. Next the finite group H;H is a closed,
°
normal subgroup of G;H. As in the proof of Theorem 47, we can now °
find an affine group S ° an open surjective K-homomorphism
and
X : G;Ho -+ S such that (i) Ker X = H;Ho and (ii) K[S], as a subalgebra
of K[G;Ho]' is
H;H H H;H
K[G;Ho] °
= (K[G] 0) 0 = (K[G])H.

If therefore we put 1I = X0 e, then 1I: G -+ S is a K-homomorphism with

all the required properties.

t See the second part of the proof of Theorem 47.

221
6 . The associated Lie algebra

General remarks

As usual K denotes an arbitrary field. In this chapter we continue

with the study of affine groups defined over K. If G is such a group
then e (or e G) is used to denote the identity element of G, and Go the
connected component containing e. For f E K[G] and a E G, the
notation fa is al~Y~~!p-EI~oyed to describe that member of K[G] which
satisfies l'"(T) = f(ra) for all T E G. Consequently the formulae (5.5.5)
(and not (5. 5. 4» will be applicable.
The associated Lie algebra (see section (6.3)) of the affine group
G is denoted by ~ and this notation is adapted in a natural way to deal
with special situations. For example, the Lie algebra of GL(V) is
written as ~~ (V).

6. 1 General K-algebras t

So far the K-algebras that have concerned us have all been associative
It is now necessary to consider a wider class of algebras.
A general K-algebra is a pair (M, /1), where M is a vector space
over K and

(6.1.1)

is a bilinear mapping of M x Minto M. We shall call /1 (x, y), where

x, y EM, the product of x and y. However multiplication is no longer
assumed to be associative.
Let (M, /1) be a general K-algebra and suppose that N is a sub-
space of M with the property that /1(x, y) EN whenever x, YEN. In

t General algebras are more commonly called non-associative algebras.

222
these circumstances /1 induces a bilinear mapping

Ii:NxN .... N

thereby making (N, Ii) a general K-algebra. We call (N, Ii) a sub-
algebra of (M, /1).
To see how general K-algebras can arise, let M be any vector
space over K and {m.}. I a base for Mover K. Further let {~ .. },
1 IE ~
where (i, j) varies freely in I x I, be an arbitrary family of elements
of M. Then there is precisely one way to turn M into a general K-
algebra so that the product of m. and m. (in that order) is ~ ...
1 J IJ
As an application of this observation assume that (M, /1) is a
general K-algebra, let L be an extension field of K, and define the
L- space ML as in section (1. 6). Then M is contained in ML and
there is a unique way to turn ML into a general L-algebra so that
multiplication on ML extends multiplication on M.
Suppose next that (M, /1) and (M ' , IL') are general K-algebras.
A mapping f: M .... M' is called a homomorphism of general K-algebras
if it is K-linear and

f(/1(x, y)) = IL'(f(x), f(y))

for all x, y E M. In appropriate circumstances two such homomorphisms

can be combined to give a new homomorphism.
Let f: M .... M' be a homomorphism of general K-algebras. If f
is a bijection, then it is said to be an isomorphism. In such a case
f- 1 : M' .... M is also an isomorphism and M and M' are said to be
isomorphic. The identity mapping of M is, of course, an isomorphism
of general K-algebras.
Finally let (M 1 , /11 ) and (M 2 , /1 2 ) be general K-algebras and
denote by M the direct sum of the K-spaces Ml and M2 • Then the
mappin§,

given by

223
 /1((x, y), (x', y'» = (/11 (X, X'), /12(Y' y'»

is bilinear and therefore (M, /1) is a general K-algebra. We shall call

it the direct product of M1 and M2 and denote it by M1 x M2 •
Before we leave this section it should be noted that, with two
exceptions, the terminology used above is compatible with that employed
in Chapter 1 in connection with associative algebras. The exceptions
occur in connection with unitary K-algebras. Here subalgebras are
always required to share the identity element of the ambient algebra and
homomorphisms have to preserve identity elements.

6. 2 Lie algebras

Let (M, /1) be a general K-algebra in the sense of section (6.1).

Definition. The K-algebra (M, /1) is said to be a 'Lie algebra'

provided that
(i) /1 (x, x) = 0 for all x E M;
(ii) 1L(/1(x, y), z) + J.l(J.l(y, z), x) + J.l(J.l(z, x), y) =0 whenever
x, y, z are in M.
Condition (ii) is known as Jacobi's identity.
Obviously if two general K-algebras are isomorphic and one is a
Lie algebra, then so is the other. Every subalgebra of a Lie algebra is
a Lie algebra. Any vector space can be turned into a Lie algebra by
defining the product J.l(x, y) to be zero t for all choices of x and y.

Theorem 1. Let (M, J.l) be a Lie algebra. Then

J.l(x, y) + J.l(y, x) = 0

for all x, y E M.

Proof. We have J.l(x+y, x+y) = O. The desired result follows by

expanding the left hand side and using the fact that J.l(x, x) and J.l(y, y)
are both zero.

t A Lie algebra in which all products are zero is said to be abelian.

224
 Let (M, J.l.) be a general K-algebra and {e.}. I a base for M
1 IE
over K. Evidently (M, J.l.) is a Lie algebra if and only if the following
three conditions are satisfied;
(a) J.l.(e.,
1
e.)
1
=0 for
---
all i;
(b) J.l.(e., e.)
1J
+ J.l.(e.,
J1
e.) = 0 for all i and j;
-----
(c) J.l.(J.l.(e i , ej ), ek ) + J.l.(J.l.(e j , ek ), e i ) + J.l.(J.l.(e k , e i ), e j ) = 0 for all
i, and k.

Theorem 2. Let (M, J.l.) be a Lie algebra over K and let L be

an extension field o~ Then (M L , J.l.L) is a Lie algebra over L.

Remark. ML was defined in section (1. 6). By J.l.L is meant the

unique extension of J.l. which turns ML into an algebra over L.

Proof. We have only to take a base for Mover K and use the
criterion described above.

Theorem 3. The direct product of two Lie algebras is again a Lie

algebra.

The proof involves no more than a trivial verification.

We add a few words about notation and terminology. If the K-space
M is provided with a multiplication which makes it a Lie algebra, then the
usual notation for the product of x and y is [x, y]. We call [x, y]
the Lie product of x and y. Hence in addition to being bilinear the Lie
product has the following properties:

[x, x] = 0
[x, y]+ [y, x] = 0,
1 (6.2.1)
[[x, y], z] + [[y, z], x] + [[z, x], y] = O.

So far as terminology is concerned, it is customary to refer to M itself

as a Lie algebra and to suppress any direct reference to the product
although the latter is an essential part of the structure.
We shall now illustrate the concept of a Lie algebra by means of a
few examples.

225
 Example 1. Let A be an associative K-algebra in which the product
of the elements x, y of A is written as xy. Then A becomes a Lie
algebra if we put

[x, y] = xy - yx. (6. 2. 2)

If in the sequel we have occasion to treat an associative K-algebra as a

Lie algebra, it is to be understood that the Lie product is defined as in
(6.2.2).

Example 2. Let V be any vector space over K. Then

Ho~(V, V) = En~(V) is an associative K-algebra which becomes a
Lie algebra if we put

[f, g) = fg - gf.

Here fg stands for fog.

Example 3. Let n be a positive integer and M (K) the associative

n
K-algebra formed by all n x n matrices with entries in K. This be-
comes a Lie algebra if we put

[A, B) = AB - BA

for A, B in M (K). Note that the matrices with zero trace form a Lie
n
subalgebra.

Example 4. Let R be a unitary and associative K-algebra. We

know that DerK(R) is a subspace of Ho~(R, R) and Example 2 shows
that Ho~(R, R) has a natural structure as a Lie algebra. Suppose that
D1 , D2 belong to DerK(R) and that a, b € R. Then

Theorem 4. Let R be a unitary and associative K-algebra and

let HomK(R, R) be considered as a Lie algebra as in Example 2. Then
DerK(R) is a Lie subalgebra of HomK(R, R).

226
 We return to the general theory in order to introduce some extra
terminology. Let M be a Lie algebra over K. By the dimension of M
we mean simply its dimension as a K-space. Next let x E M and define
cP :M-M by
x

cP (z)
x
= [x, z]. (6.2. 3)

Then CPx E En%(M) and there results a K-linear mapping

ad : M - En~(M) (6.2.4)

in which x 1-+ cp. An easy verification using Jacobi's identity now shows
x
that

Thus (6.2.4) is a homomorphism of Lie algebras. This homomorphism

is called the adjoint representation of M.

6.3 The Lie algebra of an affine group.

Throughout section (6.3) G will denote an affine group defined over

K and Go the connected component of its identity element e. Let
a E G and f E K[G]. The right translation of G by means of a induces
an automorphism of the K-algebra K[G] and the image of f under the
a
automorphism will be denoted by f. Thus the formulae (5.5.5) are
applicable and, in particular,

a
f (7) = f(ra) (6.3.1)

for all T E G.
Let D E DerK(K[G]).

Definition. We say that D is an 'invariant derivation' t if

a
D(f ) = (Df) a for all a E G and f E K[G].

t More precisely D is a right invariant derivation. Left invariant

derivations are obtained by making the obvious changes.

227
 Denote by DerK(K[G])G the set of all invariant derivations. By
Theorem 4, DerK(K[G]) is a Lie algebra and now a trivial verification
establishes

G
Theorem 5. The space DerK(K[G]) of invariant derivations is a
Lie subalgebra of Der K(K[G]).

Definition. The Lie algebra DerK(K[G])G will be called the 'Lie

algebra of G'.

It will next be shown that there is an intimate connection between

the Lie algebra of G and the tangent space DerK(G, e) to G at e.
To this end let D € DerK(K[G]), define

D : K[G]"'K (6.3.2)
e

by

D f= (Df)(e), (6.3.3)
e

and note that D belongs to DerK(K[G], K, w ), where w : K[G] ... K

e e e
maps f into f(e). Thus De € DerK(G, e) and we obtain a K-linear
mapping

in which D 1-+ De' This in turn induces a K-linear mapping

(6.3.4)

Theorem 6. The mapping

of (6.3.4) is an isomorphism of K-spaces.

Proof. First suppose that D € DerK(K[G])G and De = D. If

a € G and f € K[G], then

(Df)(a) = (Df) a(e) = (Dfa)(e) = Def a = D.

228
It follows that Df = 0 for all f in K[G] and hence that D = O. Thus
(6. 3. 4) is an injection.
Now assume that a E DerK(G, e). For each f E K[G] we define
a mapping

Df: G-K

by (Df)(a) = a(f a). Now the mapping G x G - K in which (a, T) 1-+ f(aT)
belongs to K[G x G]. Consequently there exist f , f , ... , f and
12m
gI' g, ... , g
2 m in K[G] such that
m
f(crT) = L f.(a)g.(T)
i=l 1 1

for all a, T in G. Since fT(a) = f(aT) it follows that

T m
f = L g.(T)f.
i=l 1 1

and hence that

m
(Df)(T) = L g.(T)af..
i=l 1 1

Accordingly
m
Df = L (af.)g.
i=l 1 1

and thus we see that Df E K[G]. Consequently f t-+ Df provides a K-

linear mapping

D : K[G] - K[G].

Suppose next that f, g E K[ G]. Then

(D(fg))(a) = a(faga)
= ga(e)((Df)(a)) + fa(e)((Dg)(a))
= g(a)((Df)(a)) + f(a)((Dg)(a))
= (g(Df) + f(Dg))(a)

and this shows that D E DerK(K[G]). Obviously De = a. Finally, for

a, T E G and f E K[G], we have

229
 (DfT)(a) = ~((fTt)
= ~(faT)
= (Df)(aT)
= (Df) T (a)
which shows that D E: DerK(K[G])G. This establishes that (6.3.4) is a
surjection and completes the proof.

G
Corollary. Let D, D' E: DerK(K[G]) . Then D = D' if and only
if D = D'.
e e
It will now be shown how a homomorphism of affine groups leads
to a homomorphism of their Lie algebras. To this end let cp: G -. G'
be a K-homomorphism of affine groups. By Theorem 6, we have iso-
morphisms

and

of K-spaces, and, by Chapter 4 Theorem 16, we have a K-linear mapping

(Here e' = cp(e) is the identity element of G'.) We can therefore define,
in a unique manner, a K-linear mapping

(6.3.5)

so as to make

DerK(K[G])G-----+. DerK(G, e)

d.l 1d(O. e) (6.3. 6)

DerK(K[G,])G' • DerK(G', e')

a commutative diagram. Thus if D E: DerK(K[G])G, then

230
 (d</>D)
e
,= De 0 </>*,

where </>*: K[G'] - K[G] is the homomorphism of K-algebras induced by

!/J, and therefore

«d</>D)f')(e') = (D(f' 0 </»)(e) (6.3. 7)

for all f' E K[G'].

G
Lemma 1. Suppose that a E G, f' E K[G'] and D E DerK(K[G]) •
Then

(a) (f' 0 </»a = f,</>(a) 0 </>;

(b) (d</>D)f' 0 </> = D(f' 0 </».

Proof. Let 7 E G. Then

a
(f' 0 </» (7) = f'(</>(7a))
= f'(</>(7)I/>(a))
= (f'I/>(a) 0 1/>)(7).

This proves (a). Next

«dl/>D)f' 0 I/>)(a) = «dI/>D)f,)I/>(a)(e')

= «dI/>D)f,l/>(a))(e')
= (dl/>D) ,(£,I/>(a))
e
= D (f'I/>(a) 0 1/»
e

by (6.3. 7). Hence, using (a),

«dl/>D)f' 0 I/>)(a) = (D(f' 0 I/»a)(e)

= (D(f' 0 I/»)(a)

and (b) is established as well.

Weare now ready to prove the important

Theorem 7. Let I/> : G - G' be a K-homomorphism of affine groups.

Then

231
is a homomorphism of Lie algebras.

Proof. We use the following notation. When D E DerK(K[G])G we

put 15 = dcpD and A = D.
e
Thus when f' E K[G'] we have, from (6.3. 7),

15 ,f'
e
= A(f' 0 cp). (6. 3. 8)

Now let D(l), D(2) belong to DerK(K[G])G. Then

([15(1), n(2 )]f')(e ') = (D(1 )n(2 )f' _ B(2 )n(l )f')(e ')

= n~~)(D(2)f') - i5~)(D(l)f')

and hence

by (6. 3. 8). On the other hand (6.3. 8) also shows that

( [D(l), D(2)] )e'f' = [D(l), D(2\(f' 0 cp)

= (D(1)D(2)(f' 0 cp))(e) - (D(2)D(1)(f' 0 cp))(e)

= A(1)(D(2)(f' 0 cp)) _ A (2)(D(1)(f' 0 cp))

= A(1)(j)(2)f' 0 cp) _ A(2)(j)(1)f' 0 cp)

by (b) of Lemma 1. It therefore follows that

( [D(l) , D(2)]) e' f' = [:5(1) ':5(2)]

e' .
f'

Consequently ([D(l), D(2)])e' = [:5(1), :5(2\, whence

[D(l), D(2)] = [:5(1), :5(2)]

by Theorem 6 Cor. and this completes the proof.

Theorem 8. Let ¢ : G - G' and 1/1: G' - Gil be K-homomorphisms

of affine groups. Then the following hold:
(a) if ¢ is an identity mapping, then so is d¢;
(b) d(1/I 0 cp) = d1/l 0 d¢;

232
 (c) if cp is a K-isomorphism of affine groups, then dcp is an
isomorphism of Lie groups and d(cp-l) == (dcpfl.

Proof. The first two assertions follow from the properties of the
diagram (6.3.6) when taken in conjunction with the remarks that immedi-
ately follow the statement of Theorem 16 of Chapter 4. The final assertion
is a consequence of the first two.
We next investigate the connection between the Lie algebra of G
and that of a closed subgroup.

Theorem 9. J.et H be a closed subgroup of the affine group G

and let cp: H - G be the inclusion homomorphism. Then

is an injection and therefore the Lie algebra of H can be regarded as a

subalgebra of the Lie algebra of G.

H
Proof. Let D E DerK(K[H]) and be such that dcpD = 0, and let
f E K[G]. By (6. 3. 7)

D (f 0 cp) = ((dcpD}f)(e) = O.
e

But f 0 ¢ is a typical member of K[H]. Consequently D

e
=0 and
therefore D = O. The theorem follows.

Theorem 10. Let H be a closed subgroup of the affine group G

and let the Lie algebra of H be regarded as a subalgebra of the Lie
algebra of G. Further let D belong to the Lie algebra of G. Then D
belongs to the Lie algebra of H if and only if De(IG(H» = (0).
Proof. Let cp: H - G be the inclusion homomorphism and let
D' E DerK(K[H])H. If now f E IG(H), then

(dcpD') e (f) = D'e (f 0 cp) = D'e

0=0
·

This shows that if D belongs to the Lie algebra of H, then De (IG (H»
contains only zero.

233
 Now suppose that De(IG(H)) = (0). The homomorphism
¢* : K[G] - K[H] of K-algebras that is induced by ¢ has kernel IG(H)
and therefore there exists ~ E DerK(H, e) such that

D = ~ 0 ¢*.
e
H
Next, by Theorem 6, there exists D' E DerK(K[H]) such that D~ = ~.
Accordingly

(d¢D') = D' 0 ¢* = D
e e e

and now it follows that d¢D' = D. In other terms D belongs to the Lie
algebra of H.
This theorem will now be recast in a different form.

Theorem 11. Let H be a closed subgroup of G and let D belong

to the Lie algebra of G. Then D belongs to the Lie algebra of H if
and only if D(IG(H)) <;::: IG(H).

Proof. Suppose first of all that D(IG(H)) <;::: IG(H) and let f E IG(H).
Since e E Hand Df E IG(H), we have (Df)(e) =0 and therefore Def = O.
Thus De(IG(H)) = 0 and hence D belongs to the Lie algebra of H by
Theorem 10.
Next assume that D is in the Lie algebra of H. Let f E IG(H) and
let TEH. Then f(H) = {oj andtherefore fT(HT- 1) {O}. But =
-1 T T
HT = H and so it follows that f E IG(H). By Theorem 10, D f = 0
T e
that is to say (Df )(e) = o. We now have

T
(Df)(T) = (Df) (e) = (Df T)(e) = o.
This shows that Df E IG(H) and hence that D(IG(H)) <;::: IG(H) as required.
Let Go be the connected component of the identity of G. Then, by
Theorem 9, the Lie algebra of G is contained in the Lie algebra of G.
o
In fact we have

Theorem 12. The Lie algebra of Go coincides with the Lie algebra
algebra of G.

234
 Proof. Suppose that D belongs to the Lie algebra of G and let
f E IG(G o)' By Theorem 10, the desired result will follow if we show
that D f is zero.
e
We can choose h E K[G] so that (0 h vanishes on all the cosets of
G
o
in G other than G
0
itself, and (ii) h(e) * O. This secures that
fh = 0 and we also know that f(e) = O. Accordingly

o = De (fh) = (Def)h(e)
and therefore D f = 0 as required.
e
We next determine the dimension of the Lie algebra of G.

Theorem 13. The dimension of the Lie algebra of the affine group
G is equal to Dim G.

Proof. By Theorem 12 we may suppose that G is connected in

which case Chapter 5 Theorem 13 shows that e is a simple point of G.
Consequently DerK(G, e) is a vector space whose dimension is Dim G
and now the desired result follows from Theorem 6.
At this point we insert a few observations that have to do with the
case where G is commutative. In this situation the mapping j : G - G
in which x 1-+ X-I is a K- homomorphism. Our aim will be to identify
the associated endomorphism of the Lie algebra. First we prove

Theorem 14. Lett [G, 11, j, e] be an affine group. Then

maps each element of DerK(G, e) into its negative.

-1
Proof. Let f E K[G] and define w: G x G - K by w(a, T) =f(aT ).
Then w E K[ G x G]. Consequently there exist f 1 , f , ••• , f and
2 m
gI' g , •.• , g in K[G] with the property that
2 m
m
f(aT- I ) = L f.(a)g.(T).
i=l 1 1

t The notation is the same as that employed in section (5. 1).

235
 m
Note that L f.g. not only belongs to K[G] but is, in fact, a constant.
i=1 1 1
Let D E DerK(G, e). Then
m m
L (Df.)g.(e) + L f.(e)(Dg.) = D.
i=1 1 1 i=1 1 1

Now
m
L g.(e)f. = f
i=1 1 1

and therefore
m
Df = L (Df.)g.(e).
i=1 1 1

On the other hand

m
f 0 j = L f.(e)g.
i=1 1 1

and so we obtain
m
(d(j, e)D)f = D(f 0 j) = L f.(e)(Dg.).
i=1 1 1

Thus (d(j, e)D)f = (-D)f and the theorem follows.

If G is commutative, then the inversion mapping j : G - G is a
K-automorphism of G and therefore dj is an automorphism of the Lie
algebra of G.

Theorem 15. Let [G, /-L, j, e] be an affine group and suppose that
G is commutative. Then dj maps each element of the Lie algebra of G
into its negative.

This follows from Theorem 14 and the properties of the diagram

(6.3.6).
Still assuming that G is commutative, let D, D' belong to the
Lie algebra of G. Since

dj[D, D'] = [djD, djD'].

We conclude that -[D, D'] = [D, D']. Hence if the characteristic of K

236
is not 2, then [D, D'] = O. However this restriction on the characteristic
. not necessary. t
IS

We postpone the consideration of examples and continue with the

general theory. In what follows G and H denote affine groups defined
over K and we put

(6.3.9)

and

H
~ = Der K(K[H]) . (6.3.10)

Thus ~ respectively ~ is the Lie algebra of G respectively H. This

will be standard notation from here on.
A K-homomorphism cp : G - H induces a Lie homomorphism
dcp : ~ - ~. So far we have no general result that will tell us when dcp is
surjective. It is to this question that we now turn our attention.
Suppose that the K-homomorphism cp: G - H is almost surjective
and let G respectively H be the connected component of the identity
o 0
of G respectively H. By Chapter 5 Theorem 23, cp induces an almost
surjective K-homomorphism cp o : G0 - H.0
Now K[G 0 ] is an integral
domain and it contains K[H ] as a subalgebra. Consequently K(H ) is a
o 0
subfield of K(G).
0 0 _
Also G and G have the same Lie algebra g, and the
Lie algebra ~ of H is also that of Ho' Moreover dcp: ~ - ~ coincides
with dcp : g -+ h.
o - -
Definition. The almost surjective K-homomorphism cp: G - H is
said to be 'separable' if K(G ) is a separable extension of K(H ).
o 0

Lemma 2. If the almost surjective K-homomorphism cp: G - H

is separable, then dcp : ~ - ~ is a surjection.

Proof. The preliminary discussion shows that we may assume that

G and Hare connected. Let D E~. Then D E DerK(K[H])H and it
extends to a derivation of K(H) over K. This extension will also be
denoted by D.

t See Theorem 30.

237
 By hypothesis, there exists a field n, between K(H) and K(G),
such that (i) n is a pure transcendental extension of K(H), (ii) K(G) is
algebraic and separable over n, and (iii) the degree of K(G) over n
is finite. It is clear that D can be extended to a derivation of n over
K. By Chapter 4 Theorem 7 it can be extended further to provide a
derivation, D' say, of K(G) over K. Thus D' E DerK(K(G)) and
D'p = Dp for all p E K[H].
Let a E G. There is an automorphism of the K-algebra K[G] in
which f t-+ fa, and this extends to an automorphism of K(G) over K.
We use ~a to denote the image of ~,where ~ E K(G), under the extended
automorphism.
For the moment we keep a fixed and define D" E DerK(K(G)) by
-1
D"~ = (D,~a )a.

Then, for p E K[H] and T E H, we have

(pq,(a) 0 q,)(T) = p(q,(T)q,(a))

= p(q,(m))
= (p 0 q,)( Ta)
= (p 0 q,)a(T).

Thus pq,(a) 0 q,)a and we can simplify this to pq,(a) = pa by

q, = (p 0

regarding K[H] as being embedded in K[G]. Likewise pq,(a- 1) = pa- 1•

Accordingly

But

because D is an invariant derivation, and therefore D"p = Dp. Thus for

any choice of a E G, D" E DerK(K(G)) and it extends to D.
We now impose a condition on a. Let K[G] = K[u , u , •.• , u ].
1 2 n
Then D'u. = v./g,
11
where v , v , .•. , v , g are in K[G] and g '" O.
12 n aa a
Choose a so that g(a) '" O. Then K[G] = K[u 1 , u2 ' ••• , un] and

238
 a
a a v.
D"u. = (D'u.) = --.!. E Q
1 1 a G, e '
g
a
because g (e) = g(a) is not zero. Accordingly DTT(K[G)) ~ QG and
,e
we can define

Il. : K[G) -+ K

by M = (D"f)(e). Clearly Il. E DerK(G, e). Also t for p E K[H)

~p = (D"p)(e) =. (Dp)(E) = D p.
E

Thus

maps ~ into D. It follows that d(l/l, e) is surjective and therefore

E
dl/l is surjective as well.

Theorem 16. Let I/l: G -+ H be a surjective K-homomorphism,

let N = Ker I/l, and denote by j : N -+ G the inclusion homomorphism.
If now K is algebraically closed and I/l is separable, then (with a self-
explanatory notation) the sequence

is exact.

Proof. We know that dj is an injection and Lemma 2 shows that

dl/l is a surjection. Also, because I/l 0 j can be factored through a trivial
group and the Lie algebra of a trivial group is zero-dimensional, dl/l 0 dj
is null. Finally Theorem 24 of Chapter 5 shows that Dim G = Dim N + Dim H
and hence the sum of the dimensions of ~ and ~ equals that of ~. The
theorem follows.
We next examine the Lie algebra of the direct product of two affine
groups G and H. To this end suppose that D g and D' h. Since
K[G x H] can be identified with
- E
- E

K[G] ®K K[H], it follows that there is a

K-linear mapping

t From here on E denotes the identity element of H.

239
 AD , D' : K[G x H] - K[G x H]

in which

AD D,{f
, v g) = Df v g+f v D'g. (6. 3. ll)

{Here f e: K[G], g e: K[H], and f v g is defined as in (2. 7.1).) Indeed an

easy verification shows that AD D' belongs to DerK{K[G x H]). Note
,
that the mapping

(6. 3. 12)

given by (D, D') ~ A.D D' is not only K-linear but also an injection.
,
We next observe that if CT e: G and T e: H, then

( CT T) CT T
(f v g) , = f v g

from which it follows easily that A. D, D' belongs to the Lie algebra of
G x H. Consequently (6. 3.12) gives rise to a K-linear injection

(6. 3. 13)

However the two terms in (6. 3. 13) have the same dimension namely
Dim G + Dim H. Accordingly (6. 3. 13) is an isomorphism of K-spaces.
Now, by Theorem 3, ~ x ~ is a Lie algebra and

A[{D D') (D D')] = [AD D" AD D']

l' l' 2' 2 l' 1 2' 2

as may be verified without difficulty. Thus (6. 3. 13) is an isomorphism of

Lie algebras. These observations are recorded in

Theorem 17. The Lie algebra of the direct product G x H is

naturally isomorphic to the direct product ~ x ~ of the corresponding
Lie algebras.
- Let D e: g and D' e: h.
-- - Then (in this isomorphism) (D, D')
is matched with the invariant derivation of K[G x H] which (for f e: K[G]
and g e: K[H]) maps f v g into Df v g+f v D'g.

Thus, when convenient, we may regard ~ x ~ as being the Lie

algebra of G x H.

240
6. 4 Extension of the ground field

Let G be an affine group defined over K and L an extension field

of K. Then G has a Lie algebra g and G(L) a Lie algebra ~* say.
By Theorem 2, gL is a Lie algebra-: Our aim is to investigate the con-
- L
nection between ~ and ~*. First we show that there is a natural em-
bedding of ~ in ~*.

Suppose that D E g. Since D is a K-linear mapping of K[G] into

itself it has a unique e';ension to an L-linear mapping DL of
K[G]L = L(G(L)] into itself. We wish to show that DL is in ~*. To this
end put .d = De' Then .d E DerK(G, e) and it has a unique extension
.dL to an L-linear mapping of K[G]L = L(G(L)] into KL = L. Further
it has already been shown t that .d LEDer L (G (L), e). Next, by Theorem 6,
there exists a unique -D E g* such that -D =.d L .
We claim that i5 = DL. To prove t~iS it will suffice to show that
they agree on K[G]. Assume therefore that f E K[G] and let us use f
to denote that member of L(G(L)] which is its natural prolongation. Then
_A LA
(Df)(e) =.d f = .df = (Df)(e).

If now G, then fa is the natural prolongation of fa, so we may re-

(J E
A a Aa _
place f and f by f and f respectively. Since D and Dare
invariant derivations this leads to

(ni)( a) = (Df)( a).

Accordingly Dr is the natural prolongation of Df and our claim is

established. It follows that DL E ~*. We therefore have a mapping

~ ..... ~*, (6.4. 1)

given by D 1-+ DL which is K-linear and an injection, and which enables

us to regard g as being embedded in ~*. Before proceeding we record
some of these observations in

t See section (4. 4) and, in particular, the remark just before Theorem
26 in that section.

241
 Theorem 18. Let g be the Lie algebra of G and let L be an
extension field of K. If n~w D E g, then DL (see ab~elong~he
Lie algebra of G(L). Furthermor~ (DL) = (D )L.
e e
Now let D , D , ••• , D be a K-base for g and put a. = (D.).
1 2 P - I Ie
By Theorem 6, a , a , ... , a is a K-base for DerK(G, e) and there-
1 2 P L L L
fore, by Chapter 4 Theorem 26, a , a , ... , a is an L-base for
(L) 1 L 2 L P
DerL(G ,e). But we know that (D.) = a.. It follows, by Theorem 6,
L L L Ie 1
that D , D , ••• , D is an L-base for g*. Hence when g is regarded
1 2 P L- -
as a K- subspace of ~* we have ~* = ~ , where each side is regarded as
an L-space. However

L
for all D, D' in ~ and thus we see that ~* and ~ coincide as Lie
algebras. This establishes

Theorem 19. Let g be the Lie algebra of G and let L be an

extension of the grOU~e~d K. Then gL is the Lie algebra of G(L).
(The embedding of g in the Lie ai;bra -of G(L) is obtained by mapping
L -
D, of g, into D .)

Let us now examine the effect of enlarging the ground field on the
Lie homomorphism dcp : ~ - ~ obtained from a K-homomorphism
cp : H - G. To this end suppose that D E h, f E K[G] and, as before, let
f denote the member of L[G(L)] that pr~longS f. From cp we obtain
an L-horoomorphism cp(L) : H(L) _ G(L) and, from (6.3. 7), we have t

But ; 0 cp(L) is the prolongation of f 0 CP. Consequently

from which we conclude (using the fact that dcpD arid dCP(L)D L are
invariant derivations) that

t We use e respectively E to denote the identity element of G res-

pectively H.

242
for all (J E G. Accordingly (dl/>(L)DL)f is the prolongation of ((dl/»D)f,
~) L and dl/>D agree on K[G]. Thus dl/> ~) DL = (dl/>D) L ,
that is to say dl/> D
a result which we now restate as

Theorem 20. Let 1/>: H'" G be a K-homomorphism of affine groups

and let L be an extension field of K. Then

extends dl/>:!!"'!f'
By way of illustration let us consider the case where H is a closed
subgroup of G and 1/>: H ... G is the inclusion homomorphism. Then
I/>(L) : H(L) ... G(L) is also an inclusion homomorphism. As a temporary
measure let h* respectively g* denote the Lie algebra of H(L) res-
pectively G(L). We then have ~njections ~"'~, ~* ... ~*, ~ ... ~* and
~ ... ~*, all of which preserve Lie products, and Theorem 20 shows that
the diagram

is commutative. Thus we can embed ~, ~, ~* in ~* = ~L without dis-

turbing their interrelations in any way. The L-subspace of gL that is
spanned by !! is a Lie algebra. It is moreover the Lie algebra of H(L).

6.5 A basic example

Throughout section (6.5) we shall be concerned with a non-trivial,

unitary, and associative K-algebra A whose dimension (as a K-space) is
finite. To avoid unimportant special ~ases it will be assumed, for the
duration of section (6.5), that the ground field K is infinite.
It was shown, in section (5.3) Example 8, that the units of A form
an affine group. As before this group will be denoted by U(A) and we

243
 supplement the notation by using ~(A) to describe the associated Lie
algebra. If V is an n-dimensional (n 2:: 1) vector space over K, then
we may take A to be En%(V) in which case U(A) becomes t GL(V).
, In view of Theorem 27 of Chapter 5, it is clear that the study of the closed
subgroups of U(A) has important implications for the general theory.
Let a EA. Then, as in (4.6.19), we can construct a derivation
D of K[A] over K. This extends naturally to a derivation of
a
K[U(A)] over K. It will be convenient to denote the extension by the
same symbol. Thus Da E DerK(K[U(A)]) and if we put

(6. 5. 1)

then, by (4. 6. 20),

(D F)(b) = -F(ab) (6.5.2)

a
for all F EA * and b EA. Again, by Chapter 4 Theorem 31,

(D P)(O')
a
= -(dP)(O', aO') (6.5.3)

for all P E K[U(A)] and 0' E U(A).

Lemma 3. Let T E U(A). Then

(6. 5. 4)

for all P E K[U(A)].

Proof. If F EA * and 0' E U(A), then

T
((D F) )(0')
a
= (Da F)(O'T) = -F(aO'T).
On the other hand F TEA * and therefore

T
(D F )(0')
a
= -F T(aO') = -F(aO'T).

t See the remarks following Theorem 19 of Chapter 5.

244
It follows that (6. 5. 4) holds for all F EA * and it is clear that it holds
for all constants. Again it is easy to see that if (6. 5. 4) holds for P = PI
and P=P 2 ' then it also holds when P=P 1 +P 2 andwhen P=P 1 P 2 '
It follows that (D P) T = D (P T) for all P E K[A] and now the extension
a a
to K[U(A)] is immediate.
Lemma 3 shows that, for all a EA, Da belongs to the Lie algebra.
!:!(A) of U(A). We recall that A itself is a Lie algebra with [a, b) = ab-ba.

Theorem 21. The mapping

A - ~(A) (6. 5. 5)

in which a 1-+ Da is an isomorphism of Lie algebras.

Proof. It is clear that (6.5.5) is K-linear. Also, by (6.5.2), if

D = 0, then F(a) = 0 for all F EA * and therefore a = O. Consequently
a
(6.5.5) is an injection. However A and !:!(A) have the same dimension
as K-spaces and therefore (6.5. 5) is an isomorphism of K-spaces.
Let a, c E A. The proof will be complete if we show that

D[a, c) -- [Da' D)
c

and this will follow if we prove that

D[a, c ]F = [Da , Dc ]F

forallFEA*. Now,forbEA,

(D[
a, c ]F)(b) = -F([a, c]b)
= F(cab) - F(acb).

On the other hand

([D , D ]F)(b) = (D D F - D D F)(b).

a c ac ca

But DcF E A*. Consequently

(D D F)(b) = -(D F)(ab) = F(cab)

a c c

245
and likewise (D D F)(b) = F(acb). The theorem follows.
c a
On the basis of Theorem 21 we can identify ~(A) with A con-
sidered as a Lie algebra. Indeed in what follows we shall put

~(A) = A. (6.5.6)

We next turn our attention to a typical closed subgroup G of U(A).

As usqa,l ~ denotes the Lie algebra of G and in view of (6.5.6) we may
consider .~ as a Lie subalgebra of A. To simplify our notation a little
we shall put

I(G) = IU(A)(G). (6. 5. 7)

Thus I(G) is an ideal of K[U(A)].

Theorem 22. Let G be a closed subgroup of U(A), let a € A

and let a € G. Then the following statements are equivalent:
(1) a €~;
(2) (dP)(a, aa) =0 for all P € I(G).

Proof. First suppose that a € ~ and P € I(G). By Theorem 11,

D P € I(G) and therefore (D P)(a) = O. It follows, from (6. 5. 3), that
a a
(dP)(a, aa) =0 and we have shown that (1) implies (2). -1

Next assume that (2) holds and let P € I(G). Then pa € I(G) and
therefore
-1
a
(dP )(a, aa) = 0,
-1
that is to say (napa )(a) = O. But
-1 -1
(D pa )
a
= (Da p)a
and thus (Da P)(1A) = o. As this holds for all P € I(G) it follows, from
Theorem 10, that a € ~.
There is a companion to Theorem 22 which will be proved shortly.
First we recall that if a € U(A), then left translation by means of a
induces an automorphism ~~ of the K-algebra K[U(A)]. Note that the
inverse of ~ *a is ~ *-1 and that
a

246
 'x~(A *) = A *. (6.5. 8)

Lemma 4. Let a E A and a E U(A). Then

D 'x*
a a
= A*D
a aaa-1
.

Proof. D 1 and ;\* D A* both belong to DerK(K(U(A)])

aaa - a-I a a
and it will suffice to show that they are the same. For this it is sufficient
to prove that they agree on A*.
Assume therefore that F E A* and b E A. Then

((;\* D A*)F(b)
a-I a a
= (Da (,X*F))(a-1b)
a
-1
= - (,X *F)(aa b)
a

by (6. 5. 8) and (6.5.2). Furthermore

(,X*F)(aa-1b) = F(aaa-1b)
a
= - (D F)(b)
aaa -1

and with this the lemma follows.

Corollary. 1L P E K[U(A)], then

(d,X;;'P) (lA' a) = (dP)(a, aa). (6. 5. 9)

Proof. We have

(d'x;;'P)(lA' a) = -((Da'x~P)(lA)

= -(A*(D _lP))(lA)
a aaa

= -(D P)(a)
aaa- 1
= (dP)(a, aa)

by (6. 5. 3).
The next result is the companion to Theorem 22.

Theorem 23. Let G be a closed subgroup of U(A), let a EA and

let a E G. Then the following statements are equivalent:

247
 (1) a E ~;

(2) (dP)(a, oa) = 0 for all P E I(G).

Proof. Since a E G, it follows that .\*(I(G)) = I(G). Consequently,

a
by (6.5.9), condition (2) is equivalent to

(dP)(lA' a) = 0

for all P E I(G). However, lA E G and so the desired result follows from
Theorem 22.
Now suppose that L is an extension field of K. Then AL (see
section (1. 6)) is a non-trivial, unitary and associative L-algebra whose
dimension as a vector space over L is finite. By Chapter 2 Theorem 32,
we have AL = A (L) because K is infinite, and an easy application of
Lemma 14 of Chapter 2 shows that

U(A)(L) = U(A L).

Accordingly AL is the Lie algebra of U(A)(L).

Suppose that a EA. Then a EALso D can stand for either a
a
derivation of K[U(A)] over K or a derivation of L[U(A L)] over L. Now
if P E K[U(A)] and P denotes its prolongation, then, as we saw at the
end of section (4. 6), D
a
P is the prolongation of Da P. This means that if
D stands for the derivation of K[U(A)] over K, then what we obtain by
a
regarding a as belonging to A L is DL , where the notation is that used in
a
Theorem 18.
Thus if G is a closed subgroup of U(A), then G(L) is a closed
subgroup of U(A L); and the Lie algebra gL of G(L) is obtained, with
all its structure, by taking the L-subspac; of AL that is spanned by the
K-subspace ~ of A.
Now suppose that D E DerK(L) and for x EA L define Dx E A L
as in (4. 6. 13). We recall that if x, x' E ALand .\ E L, then

D(x + x') = Dx + Dx',

D(.\x) = (m)x + .\(Dx),

and

D(b) =0 (b E A).

248
Every element of AL can be written in the form A b + A b +... + A b
11 22 qq
with A. ELand b. EA, and then
1 1

(6.5. 10)

from which it follows that

D(xx') = (Dx)x' + x(Dx'). (6.5. 11)

Theorem 24. Suppose that D E DerK(L) and that x E G(L). Then

(DX)£1 belongs to ~L.

Proof. Put y = (Dx)x -1 and suppose that P E I(G). Then, with the
usual notation for prolongations, P(x) == 0 and therefore, by Chapter 4
Lemma 15, (dP) (x, Dx) = O. Now the prolongations P generate the ideal
I(G)L[U(A L)] and this is the ideal, 'U say, of L[U(A L)] which is
associated with G(L). Hence, by (4.6. 8) and (4. 6.9), we have
(dQ)(x, Dx) = 0, that is (dQ)(x, yx) = 0, for all Q E 'U. Consequently
L
y Eg by Theorem 22.

Theorem 25. Let G be a closed connected subgroup of U(A), and

let L be an extension field of K. Further let x E G(L) and be a generic
point of G. Then given a E ~ there exists a unique D E DerK(K(x)) such
that Dx = ax.

Proof.
Since K[G] is an integral domain, the derivation D
a
extends to a derivation of K(G) over K. (The extension will be denoted
by the same symbol.) Next, because we have an isomorphism K(G) "" K(x)
over K, Da will induce a derivation, ~ say, of K(x) over K. We put
D= -~.

Let a 1 , a , ••• , a
the dual base.
2 n
Then a , a , ••• , a is also a base for A
r: over
be a base for A over K and F , F , .•. , F
2
Land
n
1 2 n
this time the prolongations F1 , F2 , ... , Fn form the dual base. Now

K(x) = K(F 1 (X), F 2 (X), .•. , Fn(x)) (6. 5. 12)

L
and, for any YEA ,

249
 y= F1 (y)a 1 + F2 (y)a 2 +... + Fn (y)a n.

Next, by construction,

D{F.{x))
1
= -(Da
F.)(x) = F.{ax)
ll

and therefore
n
Dx = ~ (DF.{x))a.
i=l 1 1
n
= ~ F.{ax)a.
i=l 1 1

= ax.
Finally if D' also belongs to DerK{K{x)) and D'x = ax, then
D{F.{x))
1
= D'{F.{x))
1
for all i and therefore D = D' by (6. 5. 12).

6. 6 Further examples

In this section we shall give several examples of Lie algebras

associated with affine groups. In many cases our account will depend on
the results obtained in section (6. 5).

Example 1. Let G be a finite group. Then Dim G = 0 and

therefore, by Theorem 13, its Lie algebra is zero-dimensional and hence
null.

Example 2. Let V be an n-dimensional (n 2: 1) vector space

over K. We know that GL{V) is an affine group and we shall denote the
associated Lie algebra by ~£ (V). In considering this Lie algebra, Example
1 shows that we may assume that K is infinite for otherwise the Lie
algebra is trivial.
Put A = En%(V). Then A is a K-algebra to which the results of
section (6. 5) are applicable and we have already seen, in Example 8 of
section (5. 3), that

U{A) = GL{V). (6. 6. 1)

250
It therefore follows, from (6. 5. 6), that we may make the identification

(6. 6. 2)

Of course if f, g E En<1<:(V)' then their Lie product is given by

[f, g] = fog - g 0 f. (6. 6. 3)

Example 3. Suppose that n ~ 1, let K be an infinite field, and

denote the Lie group associated with the general linear group GL (K)
n
by gl (K). This situation is dealt with by taking A, in section (6. 5),
--n
to be the K-algebra M (K) of n x n matrices. If this is done, then
n
U{A) = GL (K) and therefore
n

gl (K) = M (K). (6. 6. 4)

--n n
Here if A, B E M (K), then their Lie product is
n

[A, B] =AB - BA. (6. 6.5)

Now let

X .. : M (K)-K (6. 6. 6)
IJ n

be the mapping which maps B, in M (K), into its (i, j)-th entry. Then
n
Xij is a linear form on Mn{K) ,

and the X.. are algebraically independent over K.

IJ
Suppose that A E M (K). By (6. 5. 2), the associated invariant
n
derivation DA , of K[GLn{K)], satisfies

(6. 6. 7)

for all B in M (K).

n
We recall that the special linear group SL n (K) is a closed
. subgroup
of GLn (K). Consequently its Lie algebra, which is denoted by --nsl (K),
may be regarded as a Lie subalgebra ofM (K).
n

251
 Lemma 5. Suppose that DA , where A is the n x n matrix
n
"a .. 11, belongs to sl (K). Then L a .. = O.
IJ --n - - i=1 11

Proof. By Chapter 5 Theorem 17, the associated ideal, '11 say,

of SL (K) in K[GL (K)] is generated by Det Ilx .. 11 - 1. Consequently,
n n ~
by Theorem 11,

DA(Detllx .. 11 - 1) E'U. (6. 6. 8)

IJ

Put Y ..
IJ
= XIJ.. - 15... Then
IJ

Det II Xij II - 1 = Y11 + Y22 + ... + Ynn + q (Y 11' Y12' ... , Ynn)'

where q(Y , ••. , Y ) belongs to K[Y ,Y , ••• , Y ] and each of

11 nn 11 12 nn
its terms has degree at least 2. Hence, if I denotes the n x n identity
matrix, Y.. (1)
IJ
=0 for all i, j and therefore

Next, by (6.6.8),

(DA(Det Ilx.IJ·11 - 1»(1) =0

and so we see that
n
L (DAY .. )(I) = O.
i=1 11

Finally, by (6.6. 7),

(DAY .. )(I)
IJ
= (DAXIJ.. )(I) = -aIJ..
and therefore a 11 + a 22 +. .. + a nn =0 as required.

Theorem 26. Let K be an infinite field and let the Lie algebra,
~~n(K), of SLn(K) be regarded as a Lie subalgebra of Mn(K) in the
manner explained above. Then sl (K) consists of all matrices with
--n
zero trace.

Proof. By Chapter 5 Theorem 17, Dim SL (K) = n 2 - 1 and

n
therefore n 2 - 1 is the dimension of sl (K) as a vector space over K.
--n

252
Next Lemma 5 shows that if A E sl (K), then its trace is equal to zero.
--n
The theorem now follows because the set of all matrices with zero trace
forms a vector space of dimension n 2 - 1.

Example 4. In this example, K denotes an infinite field whose

characteristic is different from 2. We suppose that n ~ 1 is an integer
and that B is a non-singular n x n matrix with entries in K. In what
follows we use the same notation as that previously employed in section
(5.3) Example 9. In particular we put G = GL (K) and
n

H = {A /A EM (K) and AT BA
n
= B}, (6. 6. 9)

A = {p / P E M (K) and P T B + BP T
n
= 0 }. (6. 6. 10)

(Here the superfix T denotes a transpose.) We already know that H

is a closed subgroup of G and that by suitably choosing B we can obtain
both the orthogonal and the symplectic groups. Our aim is to show that
the Lie algebra h, of H, when considered as a Lie subalgebra of
~ = Mn (K) is just A.
We recall that A is a vector space over K whose dimension is
equal to Dim H (see Chapter 5 Theorem 20). If L is an extension field
of K, then

M (K)(L) =M (L)
n n'

and

G(L) -- GL n (L) •

Also in our discussion of H in Chapter 5 we showed how to choose L

and A * E M (L) so that A * was a generic point for the connected
n
component, H say, of the identity of H.
o
Let C belong to h. Then C also belongs to the Lie algebra of
Ho and therefore, by Theorem 25, there exists D E DerK(K(A*» such
that

DA* = CA*. (6.6.11)

Note that if Q, Q' EM (K(A*» then, by (6.5.10), DQ is obtained by

n

253
 T T
applying D to the entries in Q. Hence D(Q ) = (DQ) ,

D(QQ') = (DQ)Q' + Q(DQ')

and DB = O. Now, by (5.3.9),

A*T BA* = B.

Consequently

Thus

whence

T
because A * and A * are invertible. Accordingly C EA and therefore
~ ~ A. However, by Chapter 5 Theorem 20, ~ and A have the same
dimension as K-spaces. Accordingly ~ =A and we have proved

Theorem 27. Let K be an infinite field whose characteristic is

different from 2, let B E GLn(K) and let H be the closed subgroup of
GL (K) defined by (6.6.9). If now the Lie algebra of GL (K) is identified
n n
with Mn (K), then the Lie algebra of its subgroup H is ]I., where ]I. is
defined in (6.6.10).

Example 5. Let K be an infinite field and V an n-dimensional

(n ?: 1) vector space over K. Further let U and W be subspaces of V
with U ~ W.
In Example 7 of section (5. 3) we constructed a connected, closed
subgroup GL(U, W) = H (say) of GL(V) = G (say). By (6.6.2),
g= En~(V). Our aim is to determine the Lie algebra of GL(U, W) as
a Lie subalgebra of En~(V).

In what follows w denotes an element of W; ~ E HO~(V, K) and

254
satisfies ~(V) = 0; and 1/Iw , ~ belongs to Honx(En~(V), K) and is
defined by t

1/1w,.,t(f) = Ww).
Now when discussing Example 7 of section (5. 3) it was shown that the
ideal IG(H) is generated by the functions 1/Iw , ~ - ~(w). Let f EEn~(V)
and use I to denote the identity mapping of V. Then, by Theorem 10,
f E h if and only if

for aU w and ~. Now, by (6.5.2) applied to the case where A = En~(V),

(D/1/Iw, ~ - ~(w))(I) = (Df1/lw, ~)(I)

= -1/1w, .,t(f)
= -Ww).

Accordingly f E ~ if and only if fw E V for all w E W. This proves

Theorem 28. Suppose that K is an infinite field, V is an n-

dimensional (n?: 1) vector space over K, and V, W are subspaces of
V with V k W. Let gf,(V) be identified with En~(V) and let
f E En~(V). Then f belongs to the Lie algebra of GL(V, W) if and
only if f(W) k V.

6. 7 Adjoint representations

Let H be an affine group (defined over K), ~ its Lie algebra, and
let x E H. The inner automorphism

(6. 7. 1)

in which y 1-+ xyx -1 induces an automorphism

d1/l : h .... h (6. 7. 2)

x - -

t We use fw as an alternative to f(w).

255
of the Lie algebra. Define

(6. 7. 3)

by (x, a) 1-+ dl/lx (a). Then ~ becomes a left (H, K)- module and, in regard
to this structure, we say that H acts on ~ by conjugation. Arising from
this module structure we have a homomorphism

Ad: H -GL(~) (6. 7.4)

of abstract groups. This is a representation of H by means of auto-

morphisms of ~ and it is known as the adjoint representation of H. The
main purpose of this section is to establish two claims concerning this
representation. The first of these can be stated at once.

Claim 1. The adjoint representation (6. 7.4) is a rational representa-

Suppose, for the moment, that this has been established. Then
from the adjoint representation of H we obtain a homomorphism

(6. 7. 5)

of the Lie algebra of H into the Lie algebra of GL(~). In other words we
obtain a representation of the Lie algebra ~ by means of endomorphisms
of ~. We can now formulate our second claim.

Claim n. The representation of ~ given by (6. 7.5) is the same

as the adjoint representation ad: ~ - En~(~) as defined in (6.2.4).
If K is a finite field, then both claims are trivial so let us assume
that K is infinite. The aim of the remarks which follow is to show that
we may suppose that K is algebraically closed.
Let L be an extension field of K and assume that
The L-homomorphism H(L) - H(L) in which 1) 1-+ x1)x -1 extends 1/1x
and therefore it is just 1/1 (L). Moreover, by Theorem 20,
(L) L L x
dl/l : h - h extends dl/l : h - h.
x - - x - -
Let e , e , ••• , e be a base for hover K. Then it is also a
L1 2 n
base for hover L. Next

256
 n (L)
dtf; (e.) = L y .. (x)e. = dtf; (e.),
x J i=l IJ 1 x J
where y .. : H ... K, anc l;)y Lemma 3 Chapter 5 and its corollary, the
IJ
representation Ad : H -. GL(h) is rational if and only if y .. E K[H] for
- D
all i and j.
Assume that the adjoint representation H(L) ... GL(h L ) is rational
and select i, j so that l:s i, j :s n. Then there exists -FE L[H(L)]
with the property that its restriction to H is y... Also we can find a
IJ
base for Lover K consisting of 1 and a family {A } _A of elements
a at:fi
of L. We can now write

F=f+ L A f.
aEA a a
(Here f and f belong to K[H] and f and f denote their natural
a (L) a
prolongations to H . Also only finitely many of the fa are non-zero. )
This makes it clear that y ..
IJ
=f E K[H] and the rationality of Ad: H-.aL(h)
-
follows.
Next, because

the adjoint representation H(L) ... GL(h L ) is none other than (Ad)(L) and,
by Theorem 20,

d((Ad)(L)) : gL ... En~ (gL)

extends d(Ad): g ... En%(g). Consequently if d((Ad)(L)) is the adjoint

representation of ~L, then d(Ad) is the adjoint representation of g.
Let us sum up our conclusions so far. It has been shown that if our
two claims hold for H(L) then they also hold for H. This in turn implies
that, when we come to establish these claims, we may add the assumption
that the ground field K is algebraically closed.
We now make a fresh start. It will be assumed that K is an infinite t
field and that A denotes a K-algebra of the kind discussed in section (6. 5).
Free use will be made of the results of that section and of the notation used in

t This assumption is to remain in force until we come to Theorem 29.

257
deriving them. In particular U(A) will denote the affine group formed
by the units of A and, as in (6.5.6), we shall identify its Lie algebra
!!(A) with A itself. We now proceed to investigate the adjoint representa·
tion of U(A).
Let F E A*, where A* = Honx(A, K). Then
s
F(abc) = L F.(a)F~(b)F~(c) (6. 7. 6)
i=l 1 1 1

for all a, b, c E A, where F., F~, F~ belong to A*. Next suppose

1 1 1
that x E U(A), a E A, and denote by D the corresponding invariant
a
derivation of K[U(A)]. By (6.3. 7), we have for the inner automorphism
ljIx : U(A) ~ U(A)

«dljl D )F)(lA) = (D (F 0 ljI ))(lA)

x a a x
= (Da G)(l A)'

-1
where G E K[U(A)] and satisfies G(j3) = F(x{3x ). Accordingly
s
G= L F.(x)F~(£l)F~
i=l 1 1 1

and therefore, because (D F~)(lA) = -F~(a),

all
s
«dljl D )F)(lA) = - L F.(x)F~(a)F~'(x-1)
x a i=l 1 1 1
-1
= -F(xax )

It follows that

«dljl D )P)(lA) = (D 1P)(lA)

x a xax-

for all P E K[U(A)] whence dljl D = D by the corollary to Theorem

x a xax-1
6. Consequently, when ~(A) is identified with A,

(6. 7. 7)

Define

258
 w : U(A) x ~(A) ... ~(A)

by
-1
w(x, a) = (dtJ; )a = xax •
x
Then
s
(F 0 w)(x, a) = F(xa£l) = L F.(x)F!(a)F~(£l)
i=l 1 1 1

from which it is clear that F 0 W E K[U(A) x ~(A)]. However A * gener-

ates K[A] as a K-algebra and so it follows that w is a K-morphism of
affine sets. Thus when U(A) acts on ~(A) by conjugation ~(A) is a
rational U(A)-module.
We now turn our attention to the closed subgroups of U(A).

Lemma 6. Let H be a closed subgroup of U(A). Then the adjoint

representation

Ad: H'" GL(~) (6. 7. 8)

is a rational representation.

Proof. Let x E H, define tJ; : H - H as in (6. 7.1), and

A -1 X A

tJ; : U(A) - U(A) by (3 .... x(3x . Then tJ; extends tJ; and therefore
x x x
dtJ; : u(A) - u(A) extends dtJ; : h - h. Thus, by (6. 7. 7),
x - - x - -
-1
dtJ; (a) = xax
x

for all a E h <;;; l!(A) = A. We know that when U(A) acts on 1!(A) by
conjugation, this makes ~(A) a rational U(A)-module. It is now clear
that the corresponding action of H on ~ makes ~ a rational H-module.
However, by Chapter 5 Lemma 3 Cor., this is equivalent to our assertion.
Lemma 6 shows that we have a homomorphism d(Ad) of ~ into
the Lie algebra En%(h). The lemma which follows will help us to in-
vestigate d(Ad).
Suppose that F EA * and a E h. Define

(6. 7. 9)

259
by

QF
,a
(f) = F(f(a)). (6. 7. 10)

Evidently QF, a is a linear form on En~(~).

Lemma 7. The QF a' where F

, EA * and a E l,!, span the space
of linear forms on En~(~).

Proof. Let aI' a 2 , ••• , a p be a base for h and choose

F l' F 2' ••• , FP in A * so that their restrictions to h form the dual
base. Then the QF
.,a. are linearly independent and the lemma follows .
1] .
To simplify the notation put I/J = Ad in (6. 7. 8) and note that for
a,b E!!

where I denotes the identity mapping of h. Now when x E H,

QF, b(l/J(x)) = F((dl/lx)b)

= F(xbx-1 )
s
= ~ F.(x)F!(b)F~(£\
i=l 1 1 1

where we have reverted to the notation previously employed in (6. 7. 6).

Hence, using a bar to indicate the restriction of a function to H,
s
QF b
,
0 I/J = i=l
~ F!(b)F.G~
1 1 1
,

where G~ E K[U(A)] and G~({3) = F~(tr\ But

1 .1 1

(D F.)(lA )
a 1
= (Da
F.)(lA ) = -F.(a)
ll

and, by Theorem 14,

(D G~)(lA)
a 1
= (Da G~)(lA)
1

= -(Da F~)(l
1 A
)

= F~(a).
1

260
Accordingly
s s
((dcjlD )QF b)(I) = L F.(IA)F!(b)F!'(a) - L F.(a)F!(b)F~(IA)
a, i=1 1 1 1 i=1 1 1 1
= F(ba) - F(ab).

Put ~ = ad(a). Then ~(b) = ab - ba and, since ~ E En<\c(~), it

determines an invariant derivation, D~ say, on K{GL(~)]. Moreover

and now we see that

In view of Lemma 7 this immediately generalizes to

for all P E K[GL(~)].

However dcjlDa and D~ are invariant derivations
on K[GL(h)] and therefore

by the corollary to Theorem 6. We are now ready to establish

Theorem 29. Let H be an affine group over the. (arbitrary) field

K. Then

Ad: H'" GL(h)

is a rational representation of Hand

coincides with ad:!!'" En~(!!).

Proof. It was shown at the beginning of the section that it will

suffice to prove the theorem when K is algebraically closed. But in
that case Theorem 27 of Chapter 5 shows that H is isomorphic to a
closed subgroup of U(A), where A is a suitable chosen K-algebra of

261
the type we have just been considering. In view of this all the assertions
follow from the previous discussion.
The next results give applications of the adjoint representations.

Theorem 30. Suppose that the affine group G is commutative.

Then [D, D'] = 0 for all D, D' E~ that is to say ~ is an abelian Lie
algebra.

Proof. Since G is commutative, Ad : G -+ GL(~) factors through

a trivial group and therefore d(Ad) is the null mapping of ~ into
En%(~). The theorem follows, because, by Theorem 29, d(Ad) = ad.
We recall that if A is a Lie algebra over K, then an ideal of A
is a subspace N such that [n, x] E N for all n E N and x E A.

Theorem 31. Let G be an affine group and let H be a closed

normal subgroup. Then ~ is an ideal of the Lie algebra ~.

Proof. We may suppose that the ground field is infinite. Let

x EG and define the K-morphism l/I : G -+ G by l/I (y) = xyx- 1 • Then
x x
dl/lx E GL(~) and dl/lx(~) ~~. Accordingly dl/lx E GL(~, ~), where the
notation is that employed in Example 7 of section (5. 3). It follows that
Ad(G) ~ GL(~, ~) and therefore

satisfies ad(g) ~ ~~ (~, ~). However Theorem 28 gives a full description

of ~~ (~, ~) and this shows that if a E ~ and {3 E~, then [a, {3] E~.
Consequently !! is an ideal of ~.

262
7 . Power series and exponentials

General remarks

Let G be a connected affine group defined over a field K, and let

~ be its Lie algebra. It is a fact that the connection between G and ~

is particularly close when the characteristic of the ground field is zero,

and in this chapter we shall develop a theory which explains why this is
so. Some of our remarks apply regardless of the value of the character-
istic of K. When we need to assume that the characteristic is zero this
will be stated at the beginning of the relevant section, and the reader
will also be reminded of this underlying assumption in the statement of
theorems.

7.1 Rings of formal power series

Let K be a field and let T l' T 2' ••• , Tq be indeterminates. The

set of all power series in T l' T 2' ••• , Tq will be denoted by
K[[T l' T 2' . . . , Tq]]. Thus a typical member of K[[T l' T 2' ... , Tq]]
is an infinite formal sum
I' I' 1/
'\' T IT 2 T q
/.J al/ 1/1 "'q'
1/
1 ->0 , ••• , 1/q-
>0 1/1 2'" q 2

where the 1/. are non-negative integers and the coefficients a

1 I/lI/2 . . . l/q
all belong to K. These power series can be added and multiplied in an
obvious way and, as a result, K[[T , T , .•. , T ]] becomes an integral
1 2 q
domain which has K as a subfield.
Let us now put

S = K[[T , T , •.. , T ]]
1 2 q

and suppose that P belongs to S. Then P can be written (in a unique

263
way) as an infinite formal sum

P=P +P +P + •.• , (7.1.1)

o 1 2

where P. = P .(T , T , .•. , T ) is a homogeneous polynomial of degree

] ] 1 2 q
j in T , T , •.. , T with coefficients in K. If P is not the null
1 2 <I. 1
power series we put IP I = t' where t is the smallest non-negative
2
integer such that P t "* 0, and if P is the null polynomial we put IP I = o.
We now have a real-valued function defined over S with the following
properties:

IP I :s: 1 for all PES; (7.1. 2)

IP I = 0 when and only when P = 0; (7.1. 3)

Ip± QI :s: max { Ipl, IQI} for P, QE S; (7.1.4)

IpQI = IpllQI for P, QE S; (7.1.5)

Ia I = 1 whenever a E K, a "* O. (7.1.6)

Note that (7.1. 4) can be readily extended so that if P, P', ••. , p* all
belong to S, then

IP ± P' ±... ± p* I :s: max { IpI, IP' I, ... , Ip* I }.

The next step is to turn S into a metric space by defining a distance
function d(P, Q) according to the formula

d(P, Q) = P - Q I I. (7. 1. 7)

It is easily verified that this has the properties

d(P + P', Q+ Q'):S: max {d(P, Q), d(P', Q')} (7. 1. 8)

and

d(PP', QQ'):'S max {d(P, Q), d(P', Q')), (7. 1. 9)

from which it readily follows that addition, subtraction and mt.1tiplication

are continuous operations on K[[T , T , ... , T ]].
1 2 q

264
 Now assume that we have an infinite sequence cp , cp , cp , ••• of
o 1 2
power series and that d(cp , cp ) -+ 0 as /1 and v tend to infinity inde-
1 v
pendently. If d(cp , cp ) < t ' then the homogeneous constituents of
/1 v 2
¢ and cp agree at least as far as the terms of degree t. For the
/1 v
moment let us keep t fixed. Then, for sufficiently large /1, the term
of degree t in CP/1 does not depend on /1 so we may denote it by P f
Put

P=P o +P 1 +P 2 + •.•

and note that the construction of P ensures that d(cp , P) -+ 0 as /1

/1
tends to infinity. This argument shows that, as a metric space,
K[[T l ' T 2' ••• , Tq]] is complete.
The theory of convergence for formal power series is in some
respects remarkably simple. This is exemplified by the following
theorem.

Theorem 1. Let cp 0' ¢l' ¢ 2' .•. be an infinite sequence of formal

power series in T l' T 2' ••• , Tq. Then the series

¢ +cp +cp +¢ + ••.

o 1 2 3

converges if and only if CPh -+ 0 as h -+ 00. If the series does converge,

then any rearrangement of it converges to the same sum.

Proof. It is clear that if the series converges, then CPh -+ O. Now

assume that ¢h -+ O. Put

l/J m=cpO+cpl + .•. +¢.

m

Then for m < n we have

Hence, for m:::: 0 and n:::: 0, we have

(7.1. 10)

265
where k = min(m, n), and the right hand side of (7.1. 10) tends to zero
as m and n tend to infinity. Since K[[T , T , ..• , T ]] is complete
1 2 q
the sequence l/J , l/J , l/J , •.. tends to a limit and therefore the original
o 1 2
series converges.
Next suppose that the series

c/J o + c/J 1 + c/J 2 + c/J 3 + ...

converges and let

c/J'o + c/J'1 + c/J'2 + c/J'3 + •..

be a rearrangement of it. If now E> 0 is given we can choose k so

that 'c/Jk+l', 'c/Jk+2', 'c/Jk+3', ••. are all smaller than E. This done
we choose 110 so that c/Jo' c/Jl' c/J2' ... , c/Jk occur among c/J~, c/J~, •.• , c/J~
whenever 11) 11 0 , Then, provided that 11 > 11 0 ,

This shows that

(rpo + rpl + •.. + rp11 ) - W0 + rp'1 + ••• + rp')

11

tends to zero as 11'" 0() and from this we conclude that the rearranged
series not only converges but that it has the same sum as the series with
which we started.
At this point it is convenient to insert a few remarks about power
series in a single indeterminate. The indeterminate will be denoted by
T so, for the moment, we are concerned with the ring K[[T]].
Suppose that the power series

2 3
c/J=a o +aT+aT
1 2
+aT
3
+ ...

belongs to K[[T]]. If c/J is a unit in the ring then clearly a o *- O. Now

suppose that a *- O. If
o
2 3
l/J=b +bT+bT +bT + ...
o 1 2 3

then c/Jl/J = 1 provided that the equations

266
 a b
o 0
= 1'
a ob 1 + a 1 b 0 = 0,
a b +a b +a b
02 11 20
= 0,

all hold. But these equations can be solved in succession to yield

b , b , b , •.. and in this wayan inverse for IjJ is obtained. Thus, to
o 1 2
sum up, IjJ is a unit in K[[T]] if and only if a o *- O.
Now assume that IjJ E K[[T]] and IjJ *- O. Then we can write

r 2
where a r *- O. Thus IjJ = T (a r + a rH T + a r + 2T + ... ) and therefore
IjJ = T r 11 where 11 is a unit in K[[T]].

Theorem 2. Let '11 be a non-zero ideal of K[[T]]. Then '11 = (T s )

for a unique s ~ O.

Proof. The remarks just preceding the statement of the theorem

show that '11 contains a power of T. Let T S be the first power of T
to belong to '11. Now suppose that IjJ E '11 and IjJ *- O. Then IjJ = T r 11
for some r ~ 0 and some unit 11 of K[[T]]. It follows that T r E '11
and therefore r ~ s. Thus IjJ E (T s ) and now we see that '11 = (T s ).
The assertion concerning uniqueness is clear.

7. 2 Modules over a power series ring

Once again let T l' T 2' ••• , T q be indeterminates and put

S = K[[T l' T 2' ••• , Tq]]'

Since S is an integral domain it has a quotient field L say, and this is

an extension field of K.
Now suppose that V is a finite-dimensional vector space over K
and define the L-space VL as in section (1. 6). Then VL can be regarded
as an S-module in which V is embedded. The S-submodule of VL that

267
is generated by V will be denoted by VS.
Put V* = Ho~(V,
K) and let F EV*. We know from section
(1. 6) that F has a unique extension to an L-linear mapping of VL into
KL = L. (The extension will also be denoted by F.) Choose a base
v , v , ••• , v
1 2 n
for V over K. Then each element of vS has a unique
representation in the form Q v + Q V +... + a v , where a 1• E S.
11 22 nn
Since

F(a 1 v1 + ••• + anv n) = a 1F(v1) + a 2F(v) + •.• + a F(v ),

n n

it follows that Fev S) k S.

S
For x E V put

Ixl = sup IF(x) I (7.2.1)

FEV*
and observe that, since IF(x) I is either zero or -\- for some t::::: 0,
2
the supremum is attained. Observe too that

Ixl ::=; 1. (7.2.2)

Lemma 1. Let v1 , v 2 , ••• , vn be a base for V over K and let

-
aI' a 2, ... , an belong to S. Then

Ia 11
v + a 22
v +... + a nn
v I = max { Ia 1 I, Ia 21, ... , Ian I },

Proof. Let F E V*. Then

IF(a 1v 1 + a 2v 2 + ... + anv n) I

= Ia 1 F(v1) + a F(v ) + • •• + a F(v ) I

2 2 n n
::=;max{l a 1 1I F (v1)1, la 2 1I F (v)l, ... , Ia n "F(vn) I }
:s max { Ia 1 I, Ia I, ... , Ia I}
2 n

because IF(Vi ) I ::=; 1. Now let F 1 , F 2, ... , Fn be the K-base of V*

that is dual to v , v , ••• , v. Then
1 2 n

lav +av + ... +avl:::::IF.(av +av + ... +av)1

1122 nn 11122 nn
= la.11

268
and now the lemma follows.

S
Lemma 2. Let x, y E Y and let a, (3 E S. Then
(a) Ixl::: 1,
(b) Ix±yl:::max{lxi,lyl),
(c) laxl = lal/xl,
(d) Iax - j3y I ::: max { Ix-y I, Ia-pll.

Proof. The assertion (a) has already been established and (b) and
(c) follow from Lemma 1. Finally

Iax - j3y I = Ia(x - y) + (a - my I

::: max { lallx - yl, la -pllyl)
::: max { Ix - yl, la - pi).

In addition to the formulae contained in Lemma 2, we note that we

can add

Ix I = 1 wheneve!.' x E Y and x '* O. (7.2. 3)

S
The S-module Y becomes a complete metric space if we define
the distance between x and y to be 1x - y I. Addition and subtraction
are now continuous operations on yS. Also if x - x in yS and a -a
n n
in S, then a x - ax.
n n

Theorem 3. With the above notation let xo' xl' x 2 ' ••• , be an
infinite sequence of elements of yS. Then the series

x +x +x +x + ...
o 1 2 3

converges if and only if xn - 0 as n - 00. Should the series converge,

then any rearrangement of it converges to the same sum.

The proof is virtually the same as that of Theorem 1 so we shall

not give any details.
As before let Y be a finite-dimensional vector space over K but
now suppose that D E DerK(L). Then for x E yL we can define Dx E yL
by using the construction first described (in a slightly different context)

269
in (4. 6. 13). This leads to a mapping D : V L - V L with the following
properties. For x, x' E V L , A (L, k E K and v E V:

D(x + x') = Dx + Dx', (7.2.4)

D(Ax) = (DA)X + A{Dx), (7.2.5)

D(kx) = k(Dx), (7. 2. 6)

D(v) = O. (7. 2. 7)

Hence if u 1 ' u 2 ' ... , Us belong to V and T}1' r)2' ... , r)s belong to L,
then

D{r) u + r) u + ... + r) U ) = (Dr) )u + (Dr) )u + ... + (Dr) )u • (7.2.8

1122 S8 1122 SS

Note that if the derivation D satisfies D{S) <;;; S, then D(V S) <;;; VS. More-
over if the mapping S - S induced by D is continuous, then the mapping
VS _ VS in which x fo+ Dx is continuous as well.
We must now give some attention to the case where V is not merely
a finite-dimensional K·space but has the further property of being a
general K-algebra. In this situation we know, from section (6. 1), that
yL is a general L-algebra. Clearly if x, y E yS then the product
S
xy EV •
Suppose next that n E DerK{L) and D(S) <;;; S. Then we have

D(xy) = (Dx)y + x(Dy) (7. 2. 9)

whenever x and y belong to VS• {This may be seen by expressing x

and y as linear combinations of elements of V (with coefficients in S)
and applying the formula (7. 2. 8). )
Finally suppose that vI' V 2' ••• , vn is a base for V over K and
let

x=av +av + ... +av,

11 22 nn
y={3v +{3v + ... +{3v,
11 22 nn

where a. and {3. belong to S. We know, from Lemma 1, that

1 1

270
 Ixl = max { Ia 1 I, Ia 2 I, .•. , Ia n I 1

and

But

xy = L L a.{3.v.v.,
ij 1]1]

and Iv. v.1 :5 1. Thus Ixy I is at most equal to the maximum of the
1 ]
numbers Ia.{3.1 and therefore
1 ]

Ixy I :5 Ix/ly I (7.2.10)

for all x, y in VS• It follows that

Ixy-x'y'l :::smax{lx-x'l, Iy-y'll (7. 2. 11)

whenever x, y, x', y' are in VS and therefore multiplication is con-

tinuous on VS•

7.3 Exponentials

Throughout section (7. 3) we shall assume that the characteristic of

K is zero and A will be used conSistently to denote a K-algebra which is
non-trivial, unitary and associative (but not necessarily commutative) and
whose dimension as a K-space is finite. As before we set

where T l' T 2' ••• , T q are indeterminates, and we use L to denote the
quotient field of S.
Algebras such as A have already received considerable attention
in these pages. In Example 8 of section (5.3) we introduced a function
N : A - K which is useful in the study of the units of A. Let us call this
the norm function on A and recall t!lat (i) N E K[A], and (ii) an element
a (of A) is a unit of A if and only if N(a)"* O. Of course the L-algebra
A L has its own norm function. However this coincides with N on A
so we may also use N to denote the norm function of the larger algebra.

271
 S
Next, because A is closed under multiplication, it follows that
N(A S) ~ S. But N is a multiplicative function and N(1A) = lK" Con-
sequently if x is a unit of AS, then N(x) is a unit of S.
Afte; these preliminaries suppose that x E AS and Ix I < 1. By
(7.2.10), we have Ixlll::s Ixlll for 11:2: 0 and therefore the series
v
2:;
II.
converges. Put
00 II
x
exp x = l !if (Ixl < 1). (7. 3. 1)
11=0
S
Then exp x EA.
Next let x and y be elements of AS such that Ix I < 1 and
Iyl < 1. By Lemma 2, we also have Ix + yl < 1 and therefore exp x,
exp y and exp(x + y) are all defined. Indeed if x and y commute, then

exp(x + y) = (exp x)(exp y) (7. 3. 2)

by virtue of considerations very similar to those encountered in the

classical theory of the exponential function. Since x and -x commute
and exp(O) = 1, it follows that

(exp x)(exp(-x)) = 1 = (exp(-x))(exp x). (7.3.3)

Thus, when Ix I < 1, exp x is a unit of AS and N(exp x) is a unit in S.

The power series ring S is a K-algebra. Let D E Der K(S) and
observe that D has a natural extension to a derivation of Lover K.
0Ne denote this extension by the same letter.) For x E A L we define
Dx as in section (7.2). Of course D(A S) ~ AS. As we noted earlier, if
the derivation D: S'" S is continuous, then the induced mapping
AS ... AS is continuous as well.

S
Theorem 4. Let D E DerK(S) and be continuous. If now x E A ,
Ix I < 1 and x and Dx commute, then

D(exp x) = (exp x)(Dx) = (Dx)(exp x).

ProoL Since x and Dx commute, we have, by virtue of (7.2.9),

272
 D( I ~~) == (~1 (~~~~!) (Dx) == (DX)(I 1 (~~~~! )
Il== 0 Il== 0 ~J== 0

for all m 2: 1. But multiplication on AS is continuous and so too is the

mapping AS -+ AS induced by D. The theorem therefore follows by
letting m tend to infinity.
Now put

(7.3.4)

for i == 1, 2, •.. , q. We may regard Di as belonging either to DerK(S)

or to DerK(L). As a member of DerK(S), the derivation Di is continuous.
Of course D.T. == 1 whereas D.T. == 0 when i if j.
1 1 1 J
Suppose that a , a , .•. , a belong to A. Then
1 2 q
1 a. T.I
1 1
< 1 and
therefore exp(a.T.) is defined. Moreover a.T. and D.(a.T.) commute
S 11 11 J 11
in A. Consequently, by Theorem 4,

D.(exp(a.T.)) == a.(exp(a.T.)) == (exp(a.T.))a. (7.3.5)

1 11 1 11 III

and

D.(exp(a.T .)) == 0 for i if. j. (7. 3. 6)

J 1 1 -

Moreover exp(a.T.) is a unit of AS.

1 1
We now put

W == exp(a T )exp(a T ) . •. exp(a T )

1 1 2 2 q q

and, for 1::0 i ::0 q,

W. == exp(a T )exp(a T ) ... exp(a.T.),

1 11 22 11

where by wql is to be understood l A . Then W == W WI and therefore, by

i i
(7. 2. 9), (7. 3. 5) and (7. 3. 6),

D.w == (D.w.)w:
1 III
+ w.(D.w:)
III

== (D.w.)w~
1 1 1

273
It follows that

-1 -1
(D.w)w
1
= w.a.w.
1 1 1
•

-1 -1 S
Of course w, w ,wi' wi and a i all belong to A •

Lemma 3. Assume that aI' a 2, •.• , a q belong to A and are

linearly independent over K. Then (with the above notation) the elements

-1 -1
(D.w)w
1
= w.a.w.
1 1 1
(i = 1, 2, ..• , q)

are linearly independent over L.

Proof. Let cp , cp , ••• , cp belong to S. We shall prove that

1 2 q

f cp.w.a.w~ll = max l'cp

Ii=l 1 1 1 1 1
I, Icp I, ... , Icp I]
2 q
(7.3.7)

and from this the lemma will follow.

It is clear that lexp(a.T.) - 1AI < 1 and since (exp(a.T.)r 1 =
J J -1 J J
exp(-a.T.) we have I (exp(a.T.)) - 1AI < 1 as well. Next successive
J J J J
applications of (7.2. 11) show that, for ~ , ~ , .•• , ~ and
S 1 2 P
17 l' 17 2' ... , 17 P in A , we have

1~1~2·"~p-171172 .. ·17pl :5max{I~1-1711, 1~2-1721, ... , l~p-17pl}

so we now see that Iwi - 1A I < 1 and Iw ~ 1 - 1A I < 1. From this it

follows that

Iw.a.w~l - a.1 < 1. (7. 3. 8)

1 1 1 1

Since our aim is to establish (7. 3. 7), we may suppose that at least
one cpo is not zero. Put
1

max { ICP11, ICP2 1, ... , Icpql} = ~.

2
Then Icp.(w.a.w~l - a.) I < --L and therefore
1 1 1 11 2 m

f cp.(w.a.w~l
Ii=l 1 1 1 11
- a.) I < _1_.
2m
(7.3.9)

274
On the other hand a , a , •.. , a are linearly independent over K and
1 2 q
therefore

f cp.a·1 =max{lcp
Ii=1 1 1 1
I, ICP21, .•• , Icpql} =
2
~.
It follows from this and (7. 3. 9) that

qL cp.w.a.w.-11 = -1
Ii=1 1 1 11 2m
and with this the proof is complete.
Weare now ready to prove

Theorem 5. Suppose that K has characteristic zero and that A

is a non-trivial, unitary and associative K-algebra. If now aI' a 2, ..• ,aq
are elements of A that are linearly dependent over K, and

w = exp(a T )exp(a T ).•. exp(a T ),

1 1 2 2 q q

then w is a unit of AS (and hence of A L) and the transcendence degree

of K(w) over Kisat least q. (Here L is the quotient field of the ring
S=K[[T 1 , T 2, •.. , Tq]] of formal power series in T 1 , T 2, •.• , Tq.)

Proof. Since the characteristic of K is zero it follows, from

Chapter 4 Theorem 12, that the transcendence degree of K(w) over K
is equal to the dimension of DerK(K(w), L) considered as an L-space.
As before put D. = a/aT. and let a. be the restriction of D. to K(w).
1 1 1 1
Then a , a , •.. , a belong to DerK(K(w), L) and it will suffice to
1 2 q
show that they are linearly independent with respect to L.
Suppose theref~re that \ a 1 + A2 a 2 + ••. + Aqaq = 0, where
A. E L. Since w EA we can write it in the form
1

w=cpb
11
+cpb + ••• +cpb,
22 qq

where CPl' CP2' ••• , CPq belong to Sand b 1 , b 2 , ••• , bq both belong to
A and are linearly independent over K.
Now let F l' F 2' ... , F q be K-linear mappings of A into K such
that F.(b.) = 1 and F.(b.) = 0 whenever i"* j. Then F. E K[A] and has
1 1 . lA J L (1)
a natural prolongation F. to a member of L[A ] = L[A L]. Also
1

275
F.(w)
1
= cpo1 and therefore ¢.
1
E K(w). Accordingly

'- (D cp.) + A (D cp.) + . .. + A (D cp.)

111221 qql
=0
whence

,\ (D w) + ,\ (D w) + . .. +,\ (D w) = 0
1 1 2 2 q q

because

D.w
J
= (D.cp)b
J11
+ (D.C/J)b + .•. + (D.1jJ )b
J22 Jqq

for all j. Thus

-1 -1 -1
A ((D w)w ) + A ((D w)w ) + .•. + A ((D w)w )= 0
1 1 2 2 q q

and therefore, by Lemma 3, the A. are all zero. The theorem follows.
1
Let U(A) be the group of units of A. This group has already
received considerable attention particularly in section (6. 5). We recall
some basic facts.
The group can be regarded as an affine group with coordinate ring
K[A][l/N], where N denotes the norm function, and (U(A))(Q) = U(A Q)
for any extension field Q of K. Also the Lie algebra )!(A) , of U(A), may
be identified with A (considered as a Lie algebra). Consequently if H
is a closed subgroup of U(A), then it is possible to regard its Lie algebra
h as a subalgebra of A.

Theorem 6. Let H be a closed subgroup of U(A) and let

aI' a , •.• , a belong to the Lie algebra h of H. Put
2 q

w = exp(a T )exp(a T ) ... exp(a T ).

1 1 2 2 q q

Then w E H(L) and therefore w is a generalized point of H.

Proof. Because H(L) is a group it is enough to show that

exp(a T ) belongs to H(L).
1 1
Put sI= K[[T ]] and let
lL
L
1
be its
quotient field. Then exp(a T ) is a unit of A 1 and it will suffice to
(L1 ) 1
show that it belongs to H I . Hence for the remainder of the proof we

276
may suppose that q = 1, i. e. that we are dealing with just one indeter-
minate. This indeterminate will be denoted by T and we shall write a
in place of a 1 •
Let Q E K[U(A)]. Then Q = P~, where P E K[A], h> 0 and N
is the norm function. Now Q and P have natural prolongations, Q
and P say, to coordinate functions on (U(A)) (L) = U(A L ). It is clear that
A

P(exp(aT)) E S and we have noted earlier that N(exp(aT)) is a unit of S.

Consequently Q(exp(aT)) E S.
Let 'U be the ideal of H in K[U(A)] and denote by 'U' the ideal
of S that is generated by the elements Q(exp(aT)) as Q varies over 'U.
a
Put D = aT and den0te by Da the invariant derivation, of K[U(A)] or
L[ (U(A))(LJ], that is associated with the element a. (Note that a EA ~A L. )
By Chapter 4 Lemma 15,

D(Q(exp(aT))) = (dQ)(exp(aT), D exp(aT))

= (dQ)(exp(aT), a exp(aT))
A

= -(DaQ)(exp(aT)).

(Here we have made use of Theorem 4 of this chapter and Theorem 31 of

Chapter 4.) But, as we saw in section (4.6), D Q is the natural pro-
a
longation of D Q and, since a E h, we have D ('U) ~ 'U by Chapter 6
a - a
Theorem 11. Accordingly for Q E 'U

(D Q)(exp(aT)) E 'U'
a
and now it follows that D('U ') ~ 'U'.
We claim that 'U' = (0). For assume the contrary. Then, by
Theorem 2, 'U' = T~[[T]] for some m ~ O. Now exp(aT) - 1A is in
TA S so if (with our earlier notation) Q = P /Nh belongs to 'U, then
P(lA) = 0 and therefore

belongs to TK[[T]]. Accordingly

Q(exp(aT)) E TK[[T]]

277
and this shows that m 2: 1. Next T m E 'U '. It follows that D(T m } E 'U'
that is to say mT m - 1 E 'U'. But K has characteristic zero so
Tm- 1 E 'U'. This gives a contradiction and thereby establishes our claim
that 'U' = (O).
It has now been established that Q(exp(aT)) =0 whenever Q E 'U.
Accordingly exp(aT} E H(L} by Chapter 2 Theorem 34.

Theorem 7. Let K have characteristic zero and let H be a

closed connected subgroup of U(A}. Let h ~ A be the Lie algebra of H
and let a , a , .•. , a be a base for hover K. Furthermore suppose
1 2 q - --
that T 1 , T 2 , ••. , T q are indeterminates and denote by
-- -
L the quotient
field of K[[TI' T 2 , . . . , Tq]]' Then

w = exp(al T 1 }exp(a 2 T 2} ••• exp(aqTq}

belongs to H(L} and is a generic point of H.

Proof. Since q is the dimension of h as a K-space, we have

Dim H = q and we also have W E H(L} by T~eorem 6. Next the trans-
cendence degree of K(w} over K is at least q by Theorem 5, whereas
Theorem 23 of Chapter 3 shows that it does not exceed this value. In
fact the result just quoted shows that w is a generic point of H as we
wished to prove.
The lemma which follows will be superseded by an important general
result to be found in the next section. The role of the lemma is to provide
a stepping-stone to this result.

Lemma 4. Let Hand H be closed connected subgroups of

1 2
U(A} and let hI and ~2 be their Lie algebras. If now ~I ~ ~2' then
H ~ H.
1 2

Proof. be a base for ~I over K and let

Let a , a , •.• , a
1 q 2 -
T l ' T 2' ••• , T q be indeterminates. Define Land w as in Theorem 7.
Then w E H~L} and is a generic pOint of HI' Furthermore, by Theorem 6
we also have W E H(L}. Now every point of H is a specialization of w
2 1
and therefore, by Chapter 2 Theorem 34 Cor. 2 applied to U(A}, every
such point belongs to H. Accordingly H ~ H .
2 1 2

278
7. 4 Applications to affine groups

We begin by generalizing Lemma 4.

Theorem 8. Let G be an affine group defined over a field K of

characteristic zero, and let HI and H2 be closed connected subgroups
of G with ~1 and ~2 as their Lie algebras. (The Lie algebras ~1
and ~2 are to be regarded as subalgebras of ~.) Then the following
statements are equivalent:
(a) H ~H .
1 2'
(b) h ~ h •
-1 -2
Hence H1 = H 2 if and only if h = h .
--- -1 -2

Proof. It is clear that (a) implies (b). From here on we assume

that h ~ h .
-1 -2
Let &1 be the algebraic closure of K. By Theorem 27 of Chapter
5, we can regard G(&1) as a closed subgroup of U(A), where A is an
&1-algebra of the kind discussed in the last section. Thus H~&1)~G(&1)~U(A)
1
for i = I, 2. Let h~ and g* denote the Lie algebras of H.(&1) and
-1 ~ 1
d m respectively. Then

and we know, from the discussion following Chapter 6 Theorem 20, that
h~ is the &1-subspace of A spanned by -1h.. Since h- 1~- h2 , it follows
-1
that h* ~ h* and therefore H(&1) ~ H(&1) by Lemma 4. But, by Chapter 2
-1 -2 1 (&1) 2
Theorem 30 Cor. 2, H. = G n H. and therefore HI ~ H as required.
1 1 2
At this point we interrupt our discussion of affine groups in order to
introduce the idea of a module over a Lie algebra.
To this end let A be a Lie algebra over K (K is an arbitrary field)
and let V be a vector space over K (the dimension of V need not be
finite). Then En~(V) has a natural structure as a Lie algebra. Any
homomorphism

(7.4.1)

of Lie algebras is called a representation of A by means of endomorphisms

279
of V. For example, the adjoint representation t is of this kind.
Suppose that we have such a representation of A and put
AV = (X(A))(V) for all A E A and v E V. Then, with a self-explanatory
notation,

A(V
1
+ v 2 ) = AV 1 + AV 2 ,
A(kv) = k(AV) = (kA)V for k E K,
(7.4.2)
(A + A )v = A v + A v,
1 2 1 2

[A , A ]v = A (A v) - A (A v).
1 2 1 2 2 1

It will be convenient to describe the properties set out in (7. 4.2) by saying
that the K-space V is a (left) module with respect to the Lie algebra A.
Evidently the notion of a A-module is equivalent to that of a representation
of A by endomorphisms of a K-space.
Let us introduce some further terminology. Suppose that A is a
Lie algebra over K and that V is a A-module. By a A-submodule of V
we mean a K-subspace W which satisfies AW EW for all A E A and
w E W. Again if V "* 0 and it has no A-submodules apart from 0 and V
itself, then we say that V is a simple A-module. Particularly important
for us is the

Definition. The A-module V is said to be 'completely reducible'

if for every A-submodule W, of V, there exists a second A-submodule
W' such that V = W EB W' this being a direct sum of K-spaces.

Now suppose that G is an affine group over K and that V is a

finite-dimensional rational G-module. t Then, by the corollary to Lemma 3
in Chapter 5, there results a rational representation

p : G - GL(V) (7. 4. 3)

of G. This in turn gives rise to a homomorphism

t See section (6.2).

t For the remainder of Chapter 7, all G-modules are to be understood
to be left G-modules.

280
 (7.4.4)

of Lie algebras. Thus the finite-dimensional rational G-module V has a

natural structure as a ~-module.

Now suppose that K is an infinite field, let W be a K-subspace of

V and consider GL(W, W), where the notation is the same as that em-
ployed in Example 7 of section (5. 3). By Chapter 5 Theorem 18,
GL(W, W) is a closed connected subgroup of GL(V). Also, as is easily
verified from the definition,

GL(W, W) = {ala E GL(V) and oW ~ W l.

Consequently W is a G-submodule of V if and only if p(G) ~ GL(W, W).

Assume that this is the case. Then dp(~) ~ !Z~ (W, W) and therefore, by
Chapter 6 Theorem 28, AW E W whenever A E~ and w E W. Accordingly
.!i W is a G-submodule of V, then it is also a !Z-submodule of V.
As the next theorem shows, provided that K has characteristic
zero and G is connected, the converse holds.

Theorem 9. Let G be a connected affine group defined over a

field K of characteristic zero and let V be a finite-dimensional rational
G-module. Then the following statements are equivalent:
(i) W is a G-submodule of V,
(ii) W is a g-submodule of V.

Proof. We shall assume (ii) and deduce (i). Our previous remarks
show that this will be sufficient.
Let p : G - GL(V) be the rational representation to which V gives
rise and let H be the closure of p(G) in GL(V). Then H is a closed
connected subgroup of GL(V) and there is induced an almost surjective
K-homomorphism G - H of affine groups. Since the characteristic of K
is zero, Lemma 2 of Chapter 6 shows that the induced Lie homomorphism
~ - ~ is a surjection and therefore dp(!Z) =~. But we are assuming (ii)
to be true and this implies that dp(~) ~ ~~ (W, W). Accordingly
~ ~ !Z~ (W, W) and so, by Theorem 8,

281
 p(G) <;;; H <;;; GL(W, W).

Finally it follows from this that W is a G-submodule of V as required.

Theorem 10. Let G be a connected affine group defined over a

field of characteristic zero and let ~ be its Lie algebra. Further let V
be a finite-dimensional rational G-module. Then V is a completely
reducible respectively simple G-module if and only if it is a completely
reducible respectively simple ~-module.

This is an immediate consequence of Theorem 9.

Theorem 11. Let G be a connected affine group defined over a

field of characteristic zero. Assume that every finite-dimensional
~-module is completely reducible. Then G is linearly reductive.

Proof. Let V be a finite-dimensional rational G-module. By

Theorem 10, V is a completely reducible G-module. That G is linearly
reductive follows from Theorem 32 of Chapter 5.
These last results are powerful tools which can be used to show
that certain special affine groups are linearly reductive. This is because
they enable the classical theory of Lie algebras to be employed.

282
References

1. Borel, A. Linear algebraic groups, W. A. Benjamin Inc. (1969).

2. Chevalley, C. Theorie des groupes de Lie, Tome 2, Groupes
algebriques, Hermann, Paris (1951).
3. Chevalley, C. Fundamental concepts of algebra, Academic Press
(1956).
4. Fogarty, J. Invariant theory, W. A. Benjamin Inc. (1969).
5. Goldman, O. Hilbert rings and the Hilbert Nullstellensatz,
Math. Zeit. 54 (1951), 136-40.
6. Hochster, M. and Eagon, J. A. Cohen Macaulay rings, invariant
theory, and the generic perfection of determinantal loci, Amer.
Journ. Math. 93 (1971), 1020-58.
7. Krull, W. Jacobsonches Radikal und Hilbertscher Nullstellensatz,
Proc. International Congress of Mathematicians Vol. 2, Cambridge
Mass. (1950), 54-64.
8. Nagata, M. Local rings, Interscience Tracts in Pure and Applied
Mathematics, No. 13, Interscience Publishers (1962).
9. Northcott, D. G. Lessons on rings, modules and multiplicities,
Cambridge Univ. Press (1968).
10. Van der Waerden, B. L. Moderne Algebra, Zweiter Teil, Springer
(1931).
11. Weil, A. Foundations of algebraic geometry, Amer. Math. Soc.
Colloquium Publications, Vol. 29 (1946).
12. Zariski, O. The concept of a simple point on an abstract algebraic
variety, Trans. Amer. Math. Soc. 62 (1974), 1-52.
13. Zariski, O. and Samuel, P. Commutative Algebra, Vols. 1, 2,
Univ. Series in Higher Mathematics, van Nostrand (1960).

283
Index

additive group of the ground dual space 29

field 150
adjoint representation of a Lie finitely generated algebra 5
algebra 227
finite sets topology 23
adjoint representation of an affine
formal power series 263
group 256
function algebra 17
affine group 134
affine K-algebra 27
generalized point 55
affine n-space 30
general K-algebra 222
affine set 24
general liner group 33
affine topology 24
general point 56
almost surjective K-morphism 84
associated ideal of a set of Hilbert ring 25
points 21
homomorphism of affine groups 159
associative K-algebra 3
homomorphism of K-algebras 4

Basis Theorem 64
invariant derivation 227
irreducible affine set 64
completely reducible module 176
irreducible components of a space 62
completely reducible module rela-
tive to a Lie algebra 280 irreducible space 60
connected component of the isomorphism of affine sets 35
identity 143
coordinate function 30 Jacobi's identity 224
coordinate ring of an affine Jacobson ring 25
set 24
left translation 136
derivation 95 Lie algebra 224
derived function algebra 18
Lie algebra of an affine group 228
differential of a rational Lie product 225
function 126
linear form on a vector space 29
dimension of an affine set 75
linearly disjoint fields 82
dual base 29

284
linearly reductive affine group 176 rational G-module (finite-dimen-
local derivation ll5 sional case) 167
rational G-module (general case) 174
local homomorphism 74
rationally reduced K-algebra 15
local ring of an affine set at a
point 71 rational maximal ideal 13
locus of a set of functions 19 rational representation 169
regular extension of a field 82
minimal prime ideals of an ideal 67
regular left action 163
module with respect to a Lie
regular right action 163
algebra 280
representation of a Lie algebra 279
morphic action of an affine group
on an affine set 162 Reynold's operator 191
morphism of affine sets 34 right translation 136
multiple point ll9
semi-invariant 183
multiplicative group of the ground
field 149 separable extens ion of a field 108
separable K-homomorphism 237
natural prolongation of a coordinate
function 48 separating transcendence base 108
simple G-module 175
natural prolongation of a rational
function 83 simple point ll9
Noetherian ring 64 specialization 56
Noetherian space 63 special linear group 150
non-associative algebra 222 strict quotient for the action of an
affine group 200
non-trivial K-algebra 4
structural homomorphism of an
algebra 4
orbit 164
symplectic group 159
orthogonal group 159

tangent space at a point ll5

principal locus 20
tangent vector 117
products of affine sets 40
tensor products of algebras 8
products of function algebras 38
tensor products of vector spaces 6
quotient for the action of an torus 149
affine group 196
unitary K-algebra 3
rational character 181
rational function on an irreducible weight of a semi-invariant 183
affine set 75

285